CN117254570A - Energy recovery method, system, medium, device and server - Google Patents

Energy recovery method, system, medium, device and server Download PDF

Info

Publication number
CN117254570A
CN117254570A CN202311517860.8A CN202311517860A CN117254570A CN 117254570 A CN117254570 A CN 117254570A CN 202311517860 A CN202311517860 A CN 202311517860A CN 117254570 A CN117254570 A CN 117254570A
Authority
CN
China
Prior art keywords
energy
motor
energy recovery
value
storage device
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202311517860.8A
Other languages
Chinese (zh)
Other versions
CN117254570B (en
Inventor
刘育廷
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Suzhou Metabrain Intelligent Technology Co Ltd
Original Assignee
Suzhou Metabrain Intelligent Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Suzhou Metabrain Intelligent Technology Co Ltd filed Critical Suzhou Metabrain Intelligent Technology Co Ltd
Priority to CN202311517860.8A priority Critical patent/CN117254570B/en
Publication of CN117254570A publication Critical patent/CN117254570A/en
Application granted granted Critical
Publication of CN117254570B publication Critical patent/CN117254570B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J7/00—Circuit arrangements for charging or depolarising batteries or for supplying loads from batteries
    • H02J7/14—Circuit arrangements for charging or depolarising batteries or for supplying loads from batteries for charging batteries from dynamo-electric generators driven at varying speed, e.g. on vehicle
    • H02J7/1469—Regulation of the charging current or voltage otherwise than by variation of field
    • H02J7/1492—Regulation of the charging current or voltage otherwise than by variation of field by means of controlling devices between the generator output and the battery
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
    • G06F1/16—Constructional details or arrangements
    • G06F1/20—Cooling means
    • G06F1/206—Cooling means comprising thermal management
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02P—CONTROL OR REGULATION OF ELECTRIC MOTORS, ELECTRIC GENERATORS OR DYNAMO-ELECTRIC CONVERTERS; CONTROLLING TRANSFORMERS, REACTORS OR CHOKE COILS
    • H02P27/00—Arrangements or methods for the control of AC motors characterised by the kind of supply voltage
    • H02P27/04—Arrangements or methods for the control of AC motors characterised by the kind of supply voltage using variable-frequency supply voltage, e.g. inverter or converter supply voltage
    • H02P27/06—Arrangements or methods for the control of AC motors characterised by the kind of supply voltage using variable-frequency supply voltage, e.g. inverter or converter supply voltage using dc to ac converters or inverters
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02P—CONTROL OR REGULATION OF ELECTRIC MOTORS, ELECTRIC GENERATORS OR DYNAMO-ELECTRIC CONVERTERS; CONTROLLING TRANSFORMERS, REACTORS OR CHOKE COILS
    • H02P9/00—Arrangements for controlling electric generators for the purpose of obtaining a desired output
    • H02P9/14—Arrangements for controlling electric generators for the purpose of obtaining a desired output by variation of field
    • H02P9/26—Arrangements for controlling electric generators for the purpose of obtaining a desired output by variation of field using discharge tubes or semiconductor devices
    • H02P9/30—Arrangements for controlling electric generators for the purpose of obtaining a desired output by variation of field using discharge tubes or semiconductor devices using semiconductor devices
    • H02P9/305—Arrangements for controlling electric generators for the purpose of obtaining a desired output by variation of field using discharge tubes or semiconductor devices using semiconductor devices controlling voltage
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02P—CONTROL OR REGULATION OF ELECTRIC MOTORS, ELECTRIC GENERATORS OR DYNAMO-ELECTRIC CONVERTERS; CONTROLLING TRANSFORMERS, REACTORS OR CHOKE COILS
    • H02P2101/00—Special adaptation of control arrangements for generators
    • H02P2101/15—Special adaptation of control arrangements for generators for wind-driven turbines

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Power Engineering (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Human Computer Interaction (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Charge And Discharge Circuits For Batteries Or The Like (AREA)

Abstract

The application discloses an energy recovery method, an energy recovery system, an energy recovery medium, an energy recovery device and an energy recovery server, which relate to the field of server heat dissipation and are used for solving the problem of energy loss in the motor deceleration process. The energy recovery device in the scheme comprises a direct-current power supply, a conversion circuit, an energy storage device and a controller. The controller can adjust the working mode of the conversion circuit according to the target requirement, so that direct current is inverted into alternating current when the temperature is reduced, and the motor is powered; at a point in time when the heat dissipation demand is low, allowing the motor speed to drop, the energy recovery mechanism can be started to boost the back electromotive force at the two ends of the motor and charge the energy storage device. Therefore, by controlling the working mode of the conversion circuit, energy can be recovered when a large amount of heat dissipation requirements are not needed, and therefore the energy use efficiency is improved.

Description

Energy recovery method, system, medium, device and server
Technical Field
The present disclosure relates to the field of server heat dissipation, and in particular, to an energy recovery method, system, medium, device, and server.
Background
In the server system, the main function of the cooling fan is to take away heat generated by the devices by flowing air onto the devices, and the rotation speed control of the fan is adjusted according to the temperature rise degree of each device. When the load on the server system is higher, each device generates more heat, and the rotational speed of the fan needs to be increased to provide more air flow so as to reduce the operating temperature of the device. The devices in the server system generate less heat when the system is in an idle or low load condition, and if the fan continues to operate at a higher rotational speed in this situation, additional noise and energy waste will result, and therefore the rotational speed of the fan will typically be reduced as the load on the server system is reduced.
However, during a decrease in fan speed, some of the energy may naturally dissipate or be converted to thermal energy, or be otherwise lost. This results in a decrease in energy use efficiency. That is, the energy consumed in reducing the fan speed is not completely converted to the effect of dissipating heat from the server system, but is otherwise lost.
Disclosure of Invention
The application aims to provide an energy recovery method, an energy recovery system, an energy recovery medium, an energy recovery device and an energy recovery server, wherein a controller can adjust the working mode of a conversion circuit according to target requirements, so that direct current is inverted into alternating current when the system is required to cool down, and power is supplied to a motor; when the energy recovery is needed, the energy storage device does not provide electric energy for the motor, so the rotating speed of the motor is reduced. And the back electromotive force at the two ends of the motor can be boosted and the energy storage device can be charged. Therefore, the energy can be recovered when the system cooling requirement is low by controlling the working mode of the conversion circuit, so that the energy utilization efficiency is improved.
In order to solve the technical problem, the application provides an energy recovery method, which is applied to a controller in an energy recovery device, wherein the energy recovery device further comprises a direct current power supply, a conversion circuit and an energy storage device, the direct current power supply is connected with a motor in a fan through the conversion circuit, and the energy storage device is respectively connected with the direct current power supply and the conversion circuit, and the energy recovery method comprises the following steps:
When the target requirement is determined to be system cooling, controlling the conversion circuit to work in an inversion mode so as to invert direct current output by the direct current power supply or the energy storage device into alternating current and supply power for the motor;
and when the target requirement is determined to be energy recovery, controlling the conversion circuit to work in a boosting mode so as to boost back electromotive force at two ends of the motor and charge the energy storage device.
In one embodiment, when the target demand is determined to be energy recovery, controlling the conversion circuit to operate in a boost mode to boost back emf across the motor and charge the energy storage device, comprising:
when the target requirement is energy recovery, determining the final rotating speed after the motor is reduced and the duration time of the motor from the current rotating speed to the final rotating speed according to a target optimization strategy, and recovering the energy of the motor in the duration time;
and controlling the conversion circuit to work in a boosting mode in the duration time so as to boost back electromotive force at two ends of the motor and charge the energy storage device.
In one embodiment, controlling the conversion circuit to operate in a boost mode for the duration to boost back emf across the motor and charge the energy storage device comprises:
And controlling the conversion circuit to work in a boosting mode within the duration time, so that the voltage obtained by boosting back electromotive force at two ends of the motor by the conversion circuit is larger than the input voltage of the energy storage device, and the difference value between the boosted voltage and the input voltage of the energy storage device is within a preset range.
In one embodiment, the target optimization strategy is determined in the following manner:
acquiring a historical data set of a server in a preset time period, wherein the historical data set comprises historical data corresponding to a plurality of moments, the historical data corresponding to each moment comprises data used for representing the state and action of the server at the current moment, and the action is used for adjusting the rotating speed of the motor;
constructing an objective function and a neural network model according to the historical data set, and training the neural network model according to the historical data set;
if the neural network model meets the preset condition, taking the neural network model meeting the preset condition as a target neural network model;
and when the target requirement is energy recovery, acquiring the current state of the server, and determining a target optimization strategy according to the current state and the target neural network model.
In one embodiment, after training the neural network model according to the historical dataset, further comprising:
in each iteration process, obtaining a predicted value output by the neural network model;
judging whether the difference value between the predicted value and the actual value of the objective function is in a first threshold range or not;
if the neural network model is within the first threshold range, judging that the neural network model meets the preset condition; otherwise, the next iteration is entered.
In one embodiment, determining whether the difference between the predicted value and the actual value of the objective function is within a first threshold range comprises:
calculating an output value of a loss function according to the predicted value and the actual value;
judging whether the output value of the loss function is in a second threshold range or not;
and if the difference value is within the second threshold range, judging that the difference value is within the first threshold range.
In one embodiment, determining whether the output value of the loss function is within a second threshold range includes:
judging whether the output values of the loss function are all in the second threshold range in the continuous several iterative processes;
and if the output values of the loss function are all in the second threshold range in the continuous several iteration processes, judging that the difference value is in the first threshold range.
In one embodiment, the loss function is expressed as:
；
loss is Loss function, x 1 Is constant and 0.ltoreq.x 1 Less than or equal to 1, s is the state of the server at the current moment, a is the action determined at the current moment, Q p (s, a) is a predicted value output by the neural network model, and Q (s, a) is an actual value.
In one embodiment, the constraints of the objective function are: the target optimization strategy is satisfied to maintain or decrease the system temperature within the server.
In one embodiment, the data in the history data used for representing the state of the server at the current moment includes system power data, temperature data and motor rotation speed data of the server at the current moment.
In one embodiment, further comprising:
pre-establishing a corresponding relation between a state and an action of the server and a reward value, wherein the reward value is in negative correlation with the temperature in the server, and the reward value is in positive correlation with the electric quantity for boosting the counter electromotive force at two ends of the motor and charging the energy storage device;
constructing an objective function from the historical dataset, comprising:
and constructing a target rewarding function according to the historical data set.
In one embodiment, the target reward function is expressed as:
；
Where s is the state of the server at the current time, a is the action determined at the current time, s 'is the state of the server at the next time, a' is the action determined at the next time, Q (s, a) is the prize value obtained by taking action a in state s, r is the prize value existing at the current time, Q (s ', a') is the prize value obtained by taking action a 'in state s', and max is the maximum value.
In one embodiment, the target reward function is expressed as:
；
wherein s is the state of the server at the current time, a is the action determined at the current time, s 'is the state of the server at the next time, a' is the action determined at the next time, Q (s, a) is the prize value obtained by taking action a in the state s, r is the prize value existing at the current time, Q (s ', a') is the prize value obtained by taking action a 'in the state s', max is the maximum value, and c is the discount factor.
In one embodiment, training the neural network model from the historical dataset includes:
in each iteration process, historical data of the current moment and data used for representing the state of the server in the next moment are input into the neural network model, and the neural network model is triggered to determine a predicted rewarding value of the next moment corresponding to each action according to the historical data of the current moment, the data used for representing the state of the server in the next moment and the corresponding relation;
Determining a maximum predicted prize value for a next time of the predicted prize values for each next time;
calculating a predicted rewarding value corresponding to the current moment output by the neural network model according to the maximum predicted rewarding value at the next moment and the target rewarding function;
and determining whether the neural network model meets the preset condition according to the predicted value corresponding to the current moment and the actual rewarding value of the current moment.
For solving above-mentioned technical problem, this application still provides an energy recuperation system, is applied to the controller among the energy recuperation device, energy recuperation device still includes DC power supply, converting circuit and energy memory, DC power supply passes through converting circuit is connected with the motor in the fan, energy memory respectively with DC power supply with converting circuit connects, energy recuperation system includes:
the power supply unit is used for controlling the conversion circuit to work in an inversion mode when the target requirement is system cooling, so as to invert the direct current output by the direct current power supply into alternating current and supply power for the motor;
and the energy recovery unit is used for controlling the conversion circuit to work in a boosting mode when the target requirement is determined to be energy recovery, so as to boost the back electromotive force at two ends of the motor and charge the energy storage device.
To solve the above technical problem, the present application further provides a computer readable storage medium, on which a computer program is stored, which when executed by a controller, implements the steps of the energy recovery method as described above.
In order to solve the technical problem, the application also provides an energy recovery device which comprises a controller, a direct current power supply, a conversion circuit and an energy storage device, wherein the direct current power supply is connected with a motor in a fan through the conversion circuit, the energy storage device is respectively connected with the direct current power supply and the conversion circuit, and the controller is respectively connected with the conversion circuit and the motor;
the controller, when executing a computer program, is adapted to implement the steps of the energy recovery method as described above.
In one embodiment, the energy storage device is a capacitor.
In one embodiment, the conversion circuit is a three-phase bridge circuit.
In one embodiment, the output positive terminal of the dc power supply is connected to the positive terminal of the energy storage device and the positive terminal of the conversion circuit, and the output negative terminal of the dc power supply is connected to the negative terminal of the energy storage device and the negative terminal of the conversion circuit, respectively, and the energy recovery device further includes:
The anti-backflow device is arranged between the output positive end of the direct current power supply and the positive end of the energy storage device, and is used for inverting the direct current output by the direct current power supply into alternating current and conducting the alternating current when supplying power to the motor, boosting the counter electromotive force at the two ends of the motor and stopping the direct current when charging the energy storage device.
In one embodiment, the anti-backflow device comprises a field effect tube and a diode, wherein the source electrode of the field effect tube is respectively connected with the output positive end of the direct current power supply and the anode of the diode, the drain electrode of the field effect tube is respectively connected with the positive end of the energy storage device and the cathode of the diode, and the grid electrode of the field effect tube is connected with a driving circuit for driving the field effect tube;
the field effect transistor is turned on when the direct current output by the direct current power supply is inverted into alternating current and is used for supplying power to the motor, and is turned off when the counter electromotive force at two ends of the motor is boosted and the energy storage device is charged.
In order to solve the technical problem, the application further provides a server, which comprises the energy recovery device.
The application provides an energy recovery method, an energy recovery system, an energy recovery medium, an energy recovery device and an energy recovery server, relates to the field of server heat dissipation, and is used for solving the problem of energy loss in the process of speed reduction of a fan motor. The energy recovery device in the scheme comprises a direct-current power supply, a conversion circuit, an energy storage device and a controller. The controller can adjust the working mode of the conversion circuit according to the target requirement, so that direct current is inverted into alternating current when the temperature is reduced, and the motor is powered; the energy storage device does not provide electrical energy to the motor when the energy recovery is required, so the rotational speed of the motor is reduced. And the back electromotive force at the two ends of the motor can be boosted and the energy storage device can be charged. Therefore, the energy can be recovered when the system cooling requirement is low by controlling the working mode of the conversion circuit, so that the energy utilization efficiency is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following description will briefly explain the drawings needed in the prior art and embodiments, and it is obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a block diagram of an energy recovery device provided herein;
FIG. 2 is a flow chart of a method of energy recovery provided herein;
FIG. 3 is a flow chart of neural network model optimization provided herein;
FIG. 4 is a schematic diagram of a motor speed reduction provided herein;
FIG. 5 is a schematic diagram of an energy recovery system provided herein;
fig. 6 is a schematic diagram of a computer readable storage medium provided herein.
Detailed Description
The core of the application is to provide an energy recovery method, an energy recovery system, an energy recovery medium, an energy recovery device and an energy recovery server, wherein a controller can adjust the working mode of a conversion circuit according to target requirements, so that direct current is inverted into alternating current when the temperature is reduced, and power is supplied to a motor; the energy storage device does not provide electrical energy to the motor when the energy recovery is required, so the rotational speed of the motor is reduced. And the back electromotive force at the two ends of the motor can be boosted and the energy storage device can be charged. Therefore, the energy can be recovered when the system cooling requirement is low by controlling the working mode of the conversion circuit, so that the energy utilization efficiency is improved.
For the purposes of making the objects, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.
The application provides an energy recovery method, which is applied to a controller in an energy recovery device (shown in fig. 1), wherein the energy recovery device further comprises a direct current power supply 11, a conversion circuit 13 and an energy storage device 12, the direct current power supply 11 is connected with a motor in a fan through the conversion circuit 13, and the energy storage device 12 is respectively connected with the direct current power supply 11 and the conversion circuit 13. Vemf on the right in fig. 1 is the back emf generated when the motor is decelerating, and Vemf and the inductor are combined to form an equivalent circuit of the motor.
On the basis of fig. 1, as shown in fig. 2, the energy recovery method includes:
s1: when the target requirement is determined to be system cooling, the control conversion circuit 13 works in an inversion mode to invert direct current output by the direct current power supply 11 or the energy storage device 12 into alternating current and supply power for the motor;
This step is described in terms of determining that the target demand is cooling, which means that the load in the server system may be high, and that a large amount of heat is generated by each device. In order to reduce the operating temperature of the device, the rotational speed of the fan needs to be increased to provide greater air flow. In this case, the controller instructs the conversion circuit 13 to operate in an inversion mode, and the conversion circuit 13 in the inversion mode inverts the direct current output from the direct current power supply 11 or the energy storage device 12 into alternating current and supplies power to the motor in the fan. In the inversion mode, the conversion circuit 13 converts the direct current into alternating current by the inversion operation to adapt to the type of power source required by the motor in the fan.
Through such operation, the energy recovery device can convert the energy stored by the direct current power supply 11 or the energy storage device 12 into the alternating current power required by the fan, thereby increasing the rotation speed of the fan, providing greater air flow, improving the heat dissipation effect, and reducing the working temperature of the server system.
Further, when the dc power supply 11 supplies power, the dc power output by the dc power supply 11 may also supply power to the energy storage device 12.
S2: when it is determined that the target demand is energy recovery, the control conversion circuit 13 operates in a boost mode to boost the back emf across the motor and charge the energy storage device 12.
In this step, if the target demand is determined to be energy recovery, if the load of the server is smaller than the preset value, the target demand is determined to be energy recovery, otherwise, the target demand is determined to be cooling. When energy recovery is performed, the conversion circuit 13 converts the direct-current power supply 11 into alternating current by an inversion operation and supplies power to the fan motor, and at the same time, boosts the counter electromotive force generated at both ends of the motor. The purpose of this is to convert the power of the motor into electrical energy and store the electrical energy in the energy storage device 12. During the energy recovery process, the rotational speed of the motor is reduced and the electrical energy required by the motor is reduced, so that the control and conversion circuit 13 is operated in boost mode to provide sufficient voltage to ensure proper operation of the motor and to transfer excess electrical energy to the energy storage device 12. By the method, the energy which is originally lost in the rotation speed reduction process can be recovered and utilized, namely, the energy is recovered aiming at the back electromotive force output by the fan, so that the maximum utilization of the energy and the improvement of energy efficiency are realized.
The energy storage device 12 is used to store energy, typically by means of a battery, supercapacitor, or the like. In this stage, the boosted electric energy is charged into the energy storage device 12 through the conversion circuit 13. In this way, energy may be released from the energy storage device 12 when additional energy supply is required, improving energy efficiency.
The switching circuit 13 may be a three-phase bridge circuit, and then the switching tube in the switching circuit 13 and the inductance in the motor equivalent circuit may be specifically controlled to form a Boost circuit (Boost chopper) to Boost Vemf.
In one embodiment, when the target demand is determined to be energy recovery, the control converting circuit 13 operates in a boost mode to boost the back EMF across the motor and charge the energy storage device 12, including:
when the target demand is determined to be energy recovery, determining the final rotating speed of the motor after the motor is reduced and the duration time of the motor from the current rotating speed to the final rotating speed according to a target optimization strategy, and recovering the energy of the motor within the duration time;
the switching circuit 13 is controlled to operate in a boost mode for a duration to boost the back emf across the motor and charge the energy storage device 12.
In this embodiment, the target requirement is determined to be energy recovery, that is, the back electromotive force at the two ends of the motor is boosted by controlling the conversion circuit, and the energy is stored in the energy storage device. To achieve this objective, it is necessary to determine the final rotational speed after the motor is decelerated and the duration of the deceleration from the current rotational speed to the final rotational speed, and to recover the energy of the motor for the duration.
First, according to a target optimization strategy, the final speed and duration after motor deceleration need to be determined. This can be determined by analyzing the current load and speed of the motor. The goal is to recover the energy of the motor into the energy storage device 12 as efficiently as possible, thus requiring the motor to be reduced in speed and for some time during the deceleration process without affecting overall system performance.
During the duration, the control conversion circuit 13 operates in the boost mode. This means that the switching circuit 13 boosts the back emf across the motor to the charge requirement of the energy storage device 12. The boost mode causes the voltage of the back emf to rise to reliably charge the energy storage device 12, while the conversion circuit 13 is also responsible for transferring energy into the energy storage device 12, ensuring that the energy is stored efficiently.
In one embodiment, the control of the switching circuit 13 to operate in boost mode for a duration to boost the back emf across the motor and charge the energy storage device 12 includes:
the converting circuit 13 is controlled to operate in a boost mode in a duration time, so that the voltage obtained by boosting the back electromotive force at two ends of the motor by the converting circuit 13 is greater than the input voltage of the energy storage device 12, and the difference between the boosted voltage and the input voltage of the energy storage device 12 is within a preset range.
In one embodiment, the control conversion circuit 13 boosts the back EMF across the motor and charges the energy storage device 12. Specifically, the operation mode of the conversion circuit 13 is controlled to operate in the boost mode for a duration. In the boost mode, the conversion circuit 13 boosts the voltage of the back electromotive force across the motor. In order to ensure that the boosted voltage can effectively charge the energy storage device 12, a condition is set, i.e. the difference between the boosted voltage and the input voltage of the energy storage device 12 is within a preset range, which can be determined according to design requirements and application requirements.
By this method, the back emf of the motor can be effectively utilized to charge the energy storage device 12. When the rotation speed of the motor is reduced, the control conversion circuit 13 is automatically switched to a boost mode to boost the back electromotive force and store the back electromotive force into the energy storage device 12, so that the recovery and storage of energy can be realized, and the energy utilization efficiency and the energy saving effect of the whole system are improved.
As shown in fig. 3, in one embodiment, the target optimization strategy is determined in the following manner:
acquiring a historical data set of a server in a preset time period, wherein the historical data set comprises historical data corresponding to a plurality of moments, the historical data corresponding to each moment comprises data used for representing the state and action of the server at the current moment, and the action is used for adjusting the rotating speed of a motor;
Constructing an objective function and a neural network model according to the historical data set, and training the neural network model according to the historical data set;
if the neural network model meets the preset condition, taking the neural network model meeting the preset condition as a target neural network model;
when the rotation speed of the motor is reduced, the current state of the server is obtained, and a target optimization strategy is determined according to the current state and the target neural network model.
This embodiment describes one way of determining a target optimization strategy. In this way, firstly, a historical data set of the server in a preset time period is obtained, wherein the historical data set comprises historical data corresponding to a plurality of moments, the historical data of each moment is used for representing the state and the action of the server at the current moment, and the action is used for adjusting the rotating speed of the motor. Next, an objective function and a neural network model are constructed from the historical dataset and the neural network model is trained using the historical dataset. This means that the neural network model is trained using input and output data in the historical dataset, enabling it to be predicted and optimized from the current state and goal. If the trained neural network model meets the preset condition, namely the performance of the model meets certain requirements, the neural network model meeting the preset condition is used as a target neural network model for determining a target optimization strategy. And finally, when the rotating speed of the motor is reduced and the kinetic energy recovery starts, acquiring the current state of the server, and determining a target optimization strategy according to the current state and the target neural network model. This means that it is determined how to adjust the rotational speed of the motor to achieve a better energy recovery effect based on the current state of the server collected in real time and the target predicted by the target neural network model.
The target optimization strategy in fig. 3 is derived based on a trained target neural network model. The working environment of the server after executing the target action comprises corresponding s and a data (s is the state of the server at the current moment, and a is the action determined at the current moment), and the s and a data are stored in a replay memory (which is a storage module for storing data).
In this way, an optimization strategy can be determined from historical data and neural network model predicted objectives, thereby enabling a more efficient energy recovery method.
In one embodiment, after training the neural network model according to the historical dataset, further comprising:
in each iteration process, obtaining a predicted value output by the neural network model;
judging whether the difference value between the predicted value and the actual value of the objective function is in a first threshold range or not;
if the neural network model is within the first threshold range, judging that the neural network model meets the preset condition; otherwise, the next iteration is entered.
The present embodiment describes a process of determining whether the neural network model satisfies a preset condition. The process mainly comprises the following steps: in each iteration process, obtaining a predicted value output by the neural network model, which means that after each operation of the neural network model, a predicted result is obtained, and the result can be used for evaluating the accuracy of the model; judging whether the difference value between the predicted value and the actual value of the objective function is within a first threshold value range, comparing the actual value of the objective function with the predicted value output by the neural network model in the step, and evaluating the accuracy of the model by calculating the difference value between the two; if the difference value between the predicted value and the actual value of the objective function is within the preset threshold range, the neural network model can be considered to meet the preset condition and meet the accuracy requirement, otherwise, the next iteration is carried out; if the difference value between the predicted value and the actual value of the objective function is not in the preset threshold range, the accuracy of the model is not in accordance with the requirement, in this case, further training or adjustment is needed to be carried out on the neural network model, and then prediction and judgment are carried out again; this process may be iterated until the neural network model meets preset conditions.
The purpose of the embodiment is to ensure that the model can effectively realize the determination of the target optimization strategy by judging whether the accuracy of the neural network model meets the preset requirement. In this way, the efficiency and performance of the energy recovery device can be improved and efficient charging of the energy storage device 12 achieved.
In one embodiment, determining whether the difference between the predicted value and the actual value of the objective function is within a first threshold range includes:
calculating an output value of the loss function according to the predicted value and the actual value;
judging whether the output value of the loss function is in a second threshold range or not;
if the difference value is within the second threshold range, the difference value is judged to be within the first threshold range.
The present embodiment describes a method for determining whether or not the difference between the predicted value and the actual value of the objective function is within the first threshold range. Specifically, the method comprises the steps of: step 1: calculating an output value of a loss function according to the predicted value and the actual value, wherein the loss function is used for measuring the difference between the predicted value and the actual value of the objective function, the predicted value is a result output by the neural network model, and the actual value is the actual value of the objective function; step 2: judging whether the output value of the loss function is in a second threshold range, wherein the second threshold is a preset threshold for judging whether the output value of the loss function is close to a desired range, and if the output value of the loss function is in the second threshold range, considering that the difference between the predicted value and the actual value is smaller; step 3: if the output value of the loss function is within the second threshold range, the difference value is determined to be within a first threshold range, which is a threshold value for determining the difference between the predicted value and the actual value of the objective function, and if the output value of the loss function is within the second threshold range, the difference between the predicted value and the actual value of the objective function may be considered to be within a preset first threshold range.
By using such a method of judging whether the difference is within the first threshold range, the predictive performance of the neural network model can be effectively evaluated, thereby determining whether the preset condition is satisfied.
In one embodiment, determining whether the output value of the loss function is within a second threshold range includes:
judging whether the output values of the loss function are all in a second threshold range in the continuous several iteration processes;
if the output values of the loss function are all in the second threshold range in the continuous several iteration processes, the difference value is judged to be in the first threshold range.
The present embodiment describes a method of determining whether the output value of the loss function is within the second threshold range. In particular, it is mentioned that in several consecutive iterations it is necessary to determine whether the output values of the loss function are all within the second threshold range. If the output values of the loss function are within the second threshold range during these iterations, it may be determined that the difference (i.e., the difference between the predicted value and the actual value of the objective function) is within the first threshold range.
Such a judgment manner may be used to determine whether the neural network model satisfies a preset condition, thereby selecting whether to use the neural network model as the target neural network model. In the technical scheme, a target neural network model is determined through training of a historical data set and the neural network model.
For this determination method, the accuracy requirement of the model can be controlled by setting an appropriate second threshold range. If the output values of the loss function are all within the second threshold range, the prediction capability of the model is better, and the difference value can be considered to be within the first threshold range, namely, the difference between the predicted value and the actual value of the objective function is relatively smaller. Such decisions may be used to evaluate the effectiveness of the target optimization strategy, thereby guiding decisions and controls in actual operation.
In one embodiment, the expression for the loss function is:
；
loss is Loss function, x 1 Is constant and 0.ltoreq.x 1 Less than or equal to 1, s is the state of the server at the current moment, a is the action determined at the current moment, Q p (s, a) is a predicted value output by the neural network model, and Q (s, a) is an actual value.
In training the neural network model, a gradient descent method is employed to minimize the value of the loss function, thereby adjusting the parameters of the neural network model so that it can more accurately predict the target value. Therefore, in each iteration process, the derivative of the loss function with respect to the neural network model parameter needs to be calculated and used as the gradient descent direction to continuously adjust the neural network model parameter. Through the multi-round iterative process, the model gradually converges to reach an optimal state, so that the state of the server can be accurately predicted and corresponding actions can be determined later.
In the expression of the loss function, it is required to calculate a square error between the predicted value and the actual value corresponding to the objective function. The square error can measure the distance between the predicted value of the neural network model and the actual value corresponding to the objective function, thereby guiding the neural network model to continuously adjust the parameters thereof to approachNear true values. Adding a constant x to an expression 1 The reason for (a) is that in order to control the influence degree of the error on the neural network model, x 1 The value of (2) is usually between 0.1 and 0.9, x in this embodiment 1 The value of 0.5 can be adjusted according to specific problems.
In summary, the embodiment describes a training method based on a loss function, which can optimize parameters of a neural network model, so as to improve the prediction accuracy of the neural network model on the state and the action of a server, and further determine a more accurate energy recovery optimization strategy.
As shown in fig. 4, the intermediate slope is the angular acceleration when the motor is decelerated from rpm1 to rpm 2. How much energy the fan motor can recover during the deceleration process. As can be seen from the following electrical torque energy equation (1), J is the inertial mass of the rotor and the weight load above. (3) Respectively from angular velocity Deriving (2) angular acceleration of angular velocity +.>Is a value of (2).
（1）；
Which is the torque produced when the motor speed increases.
（2）；
（3）；
（4）；
For the angular velocity corresponding to time t2, +.>The angular velocity corresponding to time t 1.
Whereas the energy recovery of the fan motor if in a reduced speed state can be expressed as follows:
（5），/>for recovered power.
（6）；
（7）；
Because of (6)Is generated by back emf (i.e. recovered power),>is the load torque, which also consumes power and is therefore indicated by a negative sign. While (7) the resistor consumes power +.>Is made up of the resistance of the motor coil winding and the three-phase current through the winding>Composition is prepared. Because of->Is power consuming and therefore is indicated by a negative sign, ">Is the resistance on the winding. In this way, the energy recovery formula of (8) can be deduced.
（8），/>Is recovered energy.
Wherein the method comprises the steps ofIs the power generated by the mechanical rotation of the rotor, +.>Is the power consumed by the mechanical rotation on the rotor. />Is the power dissipated by the resistor on the winding. From the energy recovery formula of (8), the amount of recovered energy is equal to +.>(angular acceleration at time t1 falling to t 2) and average angular velocity +.>Related to the following. If the motor speed of the fan drops faster although +.>Will be relatively large and the positive power produced by the kinetic energy recovery will be relatively large. However, the average angular velocity +. >Also become larger, let ∈ ->The power consumed mechanically will also be relatively large. The target neural network model Qp (s, a) requiring kinetic energy recovery determines the target time of descent (t 2-t 1) and target rotation of descent based on the state of the s server, such as system power, system temperature, and fan speed …Speed->Let->Is optimized. So as to achieve the best kinetic energy recovery effect. The target time (t 2-t 1) of the decrease of the fan speed and the decreased target speed can be adjusted by recovering the kinetic energy>The obtainedThe value is substituted into the formula (8) to obtain the recovered energy. And the larger the recovered energy, the higher the prize value.
In one embodiment, the constraints of the objective function are: the target optimization strategy is satisfied to keep the temperature within the server unchanged or drop.
The present embodiment is a description about the constraint condition of the objective function. In such an embodiment, the constraint of the objective function is to keep the temperature within the server constant or drop. That is, the temperature of the server needs to be considered as an important factor in determining the target optimization strategy.
To take into account constraints of the objective function, the current state of the server may be obtained: this may include temperature data inside the server, operating conditions, and other relevant parameters. Such information may be obtained by a sensor or the like. Then determining a target optimization strategy according to the current state and the target neural network model: according to the current state of the server and the pre-trained target neural network model, the optimal optimization strategy can be determined. Such a policy should be able to meet the constraint of keeping the temperature within the server unchanged or dropping. Finally, according to the target optimization strategy, the rotating speed of the motor can be adjusted to realize the required control. By adjusting the rotational speed, the running condition of the fan can be influenced, thereby influencing the temperature in the server.
By taking into account constraints of the objective function, the controller of such an energy recovery device may operate more intelligently to achieve energy recovery while maintaining a steady operating temperature of the server. This has an important meaning for improving energy utilization efficiency and protecting the operation state of the server.
In one embodiment, the data in the history data used to characterize the state of the server at the current time includes system power data, temperature data, and motor speed data at the current time of the server.
In this embodiment, the data used to characterize the state of the server in the history data includes system power data, temperature data, and motor rotation speed data of the server. The system power data refers to the power consumption of the server at a specific moment. This can be obtained by monitoring the current and voltage of the server. Changes in system power data may reflect the workload and energy consumption of the server. The temperature data is the temperature change of the server in different components and environments. Sensors may be used to measure the temperature of the internal and external environments of the server. The temperature data may reflect the heat distribution and heat dissipation performance of the server, thereby knowing its energy consumption and heat dissipation requirements. Motor speed data refers to the rotational speed of the motor driving the fan, which can be measured by a sensor. The motor rotation speed data can reflect the working state and the wind output of the fan, and further deduce the heat dissipation effect and the energy consumption level of the server.
These data are collected and recorded and used to construct objective functions and neural network models. These models and functions may predict server states and control strategies based on information provided by historical data. Thus, by analyzing and optimizing these status data, the objective of energy recovery and optimization control according to the actual situation of the server can be achieved.
In one embodiment, further comprising:
the corresponding relation of the state-action-rewarding value of the server is pre-established, the rewarding value is in negative correlation with the temperature in the server, and the rewarding value is in positive correlation with the electric quantity which is used for boosting the counter electromotive force at two ends of the motor and charging the energy storage device 12;
constructing an objective function from the historical dataset, comprising:
a target reward function is constructed from the historical dataset.
In this embodiment, the corresponding relationship between the state and the action and the prize value of the server needs to be established in advance. This means that in implementing the energy recovery method, the prize value needs to be assessed according to the different states of the server and the different actions taken. Specifically, the prize value is inversely related to the temperature in the server and positively related to the amount of power that is used to boost the back emf across the motor and charge the energy storage device 12. Further, the objective reward function constructed from the historical data set includes an optimization objective aimed at maximizing the reward value in the process of boosting the back emf across the motor and charging the energy storage device 12 by controlling the conversion circuit 13.
For example, the correspondence of rewards in a server state is shown in table 1, and table 1 is a correspondence table of server states and rewards values:
TABLE 1
Specifically, in order to prevent the temperature of the system (specifically, the temperature of the important devices in the system) from rising during energy recovery, qp (s, a) can be set to be higher than the temperature of the system, the fan speed, and the system power of the server system according to s (state)To determine the target of the descent. Server system power, if it is in a downward trend, indicates that the system does not require much power consumption. The need for heat dissipation will decrease. Also indicating that there is more likelihood of cooling in the future at the current fan speed. Therefore, the rotation speed of the fan can be reduced more and can be increased>The value is higher. More energy is recovered. If the server system power has a tendency to be flat, the fan speed is not reduced too muchMany. In order to avoid a sudden rise in the temperature of the system. If the server system power is in an upward trend, it indicates that the current heat dissipation requirement is increasing. The fan is not suitable for reducing the speed and starting the function of recovering kinetic energy. So as to increase the system power, system temperature and fan speed of the system for the system state s (state) server >Conditions for r (reorder) rewarding targets. The system temperature is given a reward point as in table 1, the system temperature is increased to obtain a negative reward point (deduction), and the system temperature is decreased to obtain a positive reward point (deduction), so that the target that the temperature of the system part is not increased is higher than the target that the system obtains more kinetic energy to recover energy.
Therefore, the embodiment further perfects the implementation of the energy recovery method, and by pre-establishing the corresponding relation between the state and the action and the rewarding value and constructing the objective function, the working mode of the conversion circuit 13 can be controlled more accurately, and the boosting of the back electromotive force at the two ends of the motor and the charging of the energy storage device 12 can be maximized, so that the energy recovery efficiency can be effectively improved.
In one embodiment, the expression for the target rewards function is:
；
where s is the state of the server at the current time, a is the action determined at the current time, s 'is the state of the server at the next time, a' is the action determined at the next time, Q (s, a) is the prize value obtained by taking action a in state s, r is the prize value existing at the current time, Q (s ', a') is the prize value obtained by taking action a 'in state s', and max is the maximum value.
The meaning of this expression is that the prize value r at the current time plus the maximum prize value max (Q (s ', a')) at the next time is taken as the total prize value Q (s, a) obtained after taking the current action a. Thus, by selecting an action that maximizes the total prize value, the energy recovery strategy can be optimized to maximize recovered energy.
The expression of this target reward function may be calculated and optimized in the historical dataset. By training the historical data set, parameters in the reward function can be determined, so that the reward function can accurately evaluate reward values obtained by taking different actions under different states. In this way, by maximizing the total prize value, an optimal energy recovery strategy model can be determined, thereby improving energy recovery efficiency.
In one embodiment, the expression for the target rewards function is:
；
where s is the state of the server at the current time, a is the action determined at the current time, s 'is the state of the server at the next time, a' is the action determined at the next time, Q (s, a) is the prize value obtained by taking action a in state s, r is the prize value existing at the current time, Q (s ', a') is the prize value obtained by taking action a 'in state s', max is the maximum value, and c is the discount factor.
This embodiment incorporates a discount factor c as compared to the above embodiment. This discount factor may be used to balance current and future prize values so that future rewards are more comprehensively considered in calculating the prize values. In particular, the discount factor c may control the decay rate of the prize value, a larger discount factor may make consideration of future prize values more important, and a smaller discount factor may pay more attention to the current prize value.
By introducing the discount factor c, the present embodiment may better enable optimization of the energy recovery strategy. The temperature of the server can be effectively controlled so as not to rise too fast while the recovered energy is maximized. The introduction of the discount factor may better balance the trade-off of these two objectives so that the energy recovery strategy model can calculate the optimal rotational speed at each moment more accurately.
In one embodiment, training the neural network model from the historical dataset includes:
in each iteration process, the historical data of the current moment and the data used for representing the state of the server in the next moment are input into the neural network model, and the neural network model is triggered to determine a predicted rewarding value of the next moment corresponding to each action according to the historical data of the current moment, the data used for representing the state of the server in the next moment and the corresponding relation;
Determining a maximum predicted prize value of the predicted prize values at the next time;
calculating a predicted rewarding value corresponding to the current moment output by the neural network model according to the maximum predicted rewarding value at the next moment and the target rewarding function;
and determining whether the neural network model meets the preset condition according to the predicted value corresponding to the current moment and the actual rewarding value of the current moment.
This embodiment describes a method of training a neural network model. First, in each iteration of training the neural network model, it is necessary to prepare history data at the present time and data for characterizing the state of the server at the next time. Then, the history data at the present time and the state data at the next time are input into the neural network model. With these inputs, the neural network model is triggered to calculate predicted prize values for each action at the next time. The action here refers to an operation of adjusting the rotation speed of the motor. Next, the one with the largest predicted prize value is selected from all the predicted prize values at the next time. That is, the optimal action at the next time is determined so as to obtain the maximum predicted prize value at that time. And then, calculating the predicted rewarding value corresponding to the current moment output by the neural network model according to the definition of the maximum predicted rewarding value and the target rewarding function at the next moment. The target reward function is used to evaluate the contribution of the current action to system performance. And finally, determining whether the neural network model meets the preset condition according to the comparison of the predicted rewarding value and the actual rewarding value at the current moment. This comparison process can be used to evaluate the training effect of the neural network model, determine if it can accurately predict the prize value, and if the expected performance requirements can be met.
In summary, the present embodiment provides a training method based on historical data and a target reward function, and predicts an optimal action of motor rotation speed adjustment by training a neural network model, so as to implement effective operation of an energy recovery device. The training method can improve the energy recovery efficiency of the system and can adapt to different working conditions and requirements.
In order to solve the above technical problems, the present application further provides an energy recovery system, on the basis of fig. 1, as shown in fig. 5, a controller applied in an energy recovery device, the energy recovery device further includes a dc power supply 11, a conversion circuit 13, and an energy storage device 12, the dc power supply 11 is connected with a motor in a fan through the conversion circuit 13, the energy storage device 12 is respectively connected with the dc power supply 11 and the conversion circuit 13, and the energy recovery system includes:
a power supply unit 51 for controlling the conversion circuit 13 to operate in an inversion mode to invert the direct current output from the direct current power supply 11 or the energy storage device 12 into alternating current and supply power to the motor when it is determined that the target demand is a temperature decrease;
the energy recovery unit 52 is configured to control the conversion circuit 13 to operate in a boost mode to boost the back emf across the motor and charge the energy storage device 12 when the target demand is determined to be energy recovery.
In one embodiment, the energy recovery unit 52 includes:
the strategy determining unit is used for determining the final rotating speed of the motor after the motor is reduced and the duration time from the current rotating speed to the final rotating speed according to a target optimization strategy when the target requirement is energy recovery, and recovering the energy of the motor within the duration time;
the circuit control unit is used for controlling the conversion circuit to work in a boosting mode in a duration time so as to boost the back electromotive force at two ends of the motor and charge the energy storage device 12.
In one embodiment, the circuit control unit is specifically configured to control the converting circuit to operate in a boost mode during a duration time, so that a voltage obtained by boosting back electromotive force at two ends of the motor by the converting circuit 13 is greater than an input voltage of the energy storage device 12, and a difference between the boosted voltage and the input voltage of the energy storage device 12 is within a preset range.
In one embodiment, further comprising:
the system comprises an acquisition unit, a motor and a control unit, wherein the acquisition unit is used for acquiring a historical data set of a server in a preset time period, the historical data set comprises historical data corresponding to a plurality of moments, the historical data corresponding to each moment comprises data used for representing the state and action of the server at the current moment, and the action is used for adjusting the rotating speed of the motor;
The construction unit is used for constructing an objective function and a neural network model according to the historical data set and training the neural network model according to the historical data set;
the model determining unit is used for taking the neural network model meeting the preset conditions as a target neural network model when the neural network model meets the preset conditions;
and the target optimization strategy determining unit acquires the current state of the server when the rotating speed of the motor is reduced, and determines the target optimization strategy according to the current state and the target neural network model.
In one embodiment, further comprising:
the predicted value acquisition unit is used for acquiring a predicted value output by the neural network model in each iteration process; judging whether the difference value between the predicted value and the actual value of the objective function is in a first threshold range or not; if the neural network model is within the first threshold range, judging that the neural network model meets the preset condition; otherwise, the next iteration is entered.
In one embodiment, the predicted value obtaining unit is specifically configured to calculate an output value of the loss function according to the predicted value and the actual value; judging whether the output value of the loss function is in a second threshold range or not; if the difference value is within the second threshold range, the difference value is judged to be within the first threshold range.
In one embodiment, the predicted value obtaining unit is specifically configured to calculate an output value of the loss function according to the predicted value and the actual value; judging whether the output values of the loss function are all in a second threshold range in the continuous several iteration processes; if the output values of the loss function are all in the second threshold range in the continuous several iteration processes, the difference value is judged to be in the first threshold range.
In one embodiment, the expression for the loss function is:
；
loss is Loss function, x 1 Is constant and 0.ltoreq.x 1 Less than or equal to 1, s is the state of the server at the current moment, a is the action determined at the current moment, Q p (s, a) is a predicted value output by the neural network model, and Q (s, a) is an actual value.
In one embodiment, the constraints of the objective function are: the target optimization strategy is satisfied to keep the temperature within the server unchanged or drop.
In one embodiment, the data in the history data used to characterize the state of the server at the current time includes system power data, temperature data, and motor speed data at the current time of the server.
In one embodiment, further comprising:
the corresponding relation establishing unit is used for pre-establishing a corresponding relation between the state of the server and the action of the server and the rewarding value which is inversely related to the temperature in the server and is positively related to the electric quantity for boosting the counter electromotive force at two ends of the motor and charging the energy storage device 12;
The construction unit is specifically used for constructing the target rewarding function and the neural network model according to the historical data set, and training the neural network model according to the historical data set.
In one embodiment, the expression for the target rewards function is:
；
where s is the state of the server at the current time, a is the action determined at the current time, s 'is the state of the server at the next time, a' is the action determined at the next time, Q (s, a) is the prize value obtained by taking action a in state s, r is the prize value existing at the current time, Q (s ', a') is the prize value obtained by taking action a 'in state s', and max is the maximum value.
In one embodiment, the expression for the target rewards function is:
；
where s is the state of the server at the current time, a is the action determined at the current time, s 'is the state of the server at the next time, a' is the action determined at the next time, Q (s, a) is the prize value obtained by taking action a in state s, r is the prize value existing at the current time, Q (s ', a') is the prize value obtained by taking action a 'in state s', max is the maximum value, and c is the discount factor.
In one embodiment, a building unit comprises:
A function and model construction unit for constructing a target rewarding function and a neural network model based on the historical data set,
the data input unit is used for inputting the historical data of the current moment and the data used for representing the state of the server in the next moment into the neural network model in each iteration process, and triggering the neural network model to determine a predicted rewarding value of the next moment corresponding to each action according to the historical data of the current moment, the data used for representing the state of the server in the next moment and the corresponding relation;
a predicted prize unit for determining a maximum predicted prize value at a next time among the predicted prize values at each next time;
the calculating predicted rewarding unit is used for calculating a predicted rewarding value corresponding to the current moment output by the neural network model according to the maximum predicted rewarding value of the next moment and the target rewarding function;
and the condition determining unit is used for determining whether the neural network model meets the preset condition according to the predicted value corresponding to the current moment and the actual rewarding value of the current moment.
For other descriptions of the energy recovery system, refer to the above embodiments, and the description thereof is omitted herein.
In order to solve the above technical problem, the present application further provides a computer readable storage medium 61, as shown in fig. 6, where a computer program 62 is stored on the computer readable storage medium 61, and the computer program 62 implements the steps of the energy recovery method as described above when executed by a controller.
For other descriptions of the computer readable storage medium 61, refer to the above embodiments, and the description is omitted herein.
In order to solve the technical problem, the application also provides an energy recovery device which comprises a controller, a direct current power supply 11, a conversion circuit 13 and an energy storage device 12, wherein the direct current power supply 11 is connected with a motor in a fan through the conversion circuit 13, the energy storage device 12 is respectively connected with the direct current power supply 11 and the conversion circuit 13, and the controller is respectively connected with the conversion circuit 13 and the motor;
the controller, when executing the computer program, is adapted to implement the steps of the energy recovery method as described above.
In one embodiment, the energy storage device 12 is a capacitor.
A capacitor is a device that stores electrical energy and is composed of two metal plates and a medium (e.g., alumina) that forms an insulating layer between the two metal plates. When the capacitor is connected to a power source, an electric field is formed between the metal plates, causing charge to accumulate on the metal plates.
By using a capacitor as the energy storage device 12, such an energy recovery device has the characteristics of high energy storage and release. The capacitor has a fast charge and discharge rate and good voltage stability, making it a suitable energy storage device 12. In addition, the capacitor has the advantages of long service life, no maintenance, small volume and the like.
In one embodiment, the conversion circuit 13 is a three-phase bridge circuit.
In one embodiment, the positive output terminal of the dc power supply 11 is connected to the positive terminal of the energy storage device 12 and the positive terminal of the conversion circuit 13, and the negative output terminal of the dc power supply 11 is connected to the negative terminal of the energy storage device 12 and the negative terminal of the conversion circuit 13, respectively, and the energy recovery device further includes:
the anti-backflow device is arranged between the output positive end of the direct current power supply 11 and the positive end of the energy storage device 12, and is used for conducting when the direct current output by the direct current power supply 11 is inverted into alternating current and supplying power to the motor, and stopping when the counter electromotive force at the two ends of the motor is boosted and the energy storage device 12 is charged.
The present embodiment further perfects the design of the energy recovery device. In this embodiment, the energy storage device 12 is connected in parallel with the bridge arms of the three-phase bridge circuit, the dc power supply 11 is connected in parallel with the energy storage device 12, and a backflow preventing device is added.
When the rotation speed of the fan is reduced, the controller can adjust the switching tube in the three-phase bridge circuit to be in a boosting mode, and back electromotive force can be generated when the fan operates. The anti-backflow device has the function of stopping when the energy storage device 12 is charged, preventing back electromotive force from backflow to the direct current power supply 11, and protecting the normal work of the direct current power supply 11. In other words, when the fan is decelerated, the backflow preventing means prevents the counter electromotive force from flowing reversely to the direct current power source 11, effectively improving the energy utilization efficiency and the reliability of the system.
In one embodiment, the anti-backflow device comprises a field effect tube and a diode, wherein the source electrode of the field effect tube is respectively connected with the positive output end of the direct current power supply 11 and the anode of the diode, the drain electrode of the field effect tube is respectively connected with the positive end of the energy storage device 12 and the cathode of the diode, and the grid electrode of the field effect tube is connected with a driving circuit for driving the field effect tube;
the field effect transistor is turned on when the dc power output from the dc power supply 11 is inverted to ac power to supply power to the motor, and turned off when the counter electromotive force at both ends of the motor is boosted to charge the energy storage device 12.
The present embodiment defines that the anti-backflow device comprises a field effect transistor and a diode. The source electrode of the field effect tube is respectively connected with the anode of the direct current power supply 11 and the anode of the diode, the drain electrode of the field effect tube is respectively connected with the anode of the energy storage device 12 and the cathode of the diode, and the grid electrode of the field effect tube is connected with the driving circuit for driving the field effect tube.
When the motor is powered, the field effect transistor is turned on, allowing the power of the dc power supply 11 to flow to the motor. This ensures that the motor will work properly. Meanwhile, the diode is cut off by the on state of the field effect transistor, so that electric energy is prevented from flowing back from the energy storage device 12 to the direct current power supply 11, and the backflow phenomenon is prevented.
When the energy storage device 12 is charged, the field effect transistor is cut off, so that the electric energy of the direct current power supply 11 is prevented from flowing to the energy storage device 12. By switching off the field effect transistor, the anti-backflow device ensures that the direct current power supply 11 does not deliver electrical energy into the energy storage device 12, but rather boosts the counter-electromotive force into the energy storage device 12 for charging by the voltage difference between the motor and the energy storage device 12. Therefore, the anti-backflow device plays a role in preventing the reverse flow of electric energy, and ensures that the energy storage device 12 is charged only when appropriate, thereby ensuring the normal operation and performance of the energy recovery device.
For other descriptions of the energy recovery device, please refer to the above embodiments, and the description is omitted herein.
In order to solve the technical problem, the application also provides a server, which comprises the energy recovery device. For other descriptions of the server, please refer to the above embodiments, and the description is omitted herein.
It should also be noted that in this specification, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (22)

1. The utility model provides an energy recuperation method which characterized in that is applied to the controller in the energy recuperation device, the energy recuperation device still includes DC power supply, converting circuit and energy storage device, DC power supply passes through converting circuit is connected with the motor in the fan, energy storage device respectively with DC power supply with converting circuit connects, the energy recuperation method includes:
when the target requirement is determined to be that the system is cooled, controlling the conversion circuit to work in an inversion mode so as to invert the direct current output by the direct current power supply or the energy storage device into alternating current and supply power for the motor;
And when the target requirement is determined to be energy recovery, controlling the conversion circuit to work in a boosting mode so as to boost back electromotive force at two ends of the motor and charge the energy storage device.
2. The energy recovery method of claim 1, wherein upon determining that the target demand is energy recovery, controlling the conversion circuit to operate in a boost mode to boost back emf across the motor and charge the energy storage device comprises:
when the target requirement is energy recovery, determining the final rotating speed after the motor is reduced and the duration time of the motor from the current rotating speed to the final rotating speed according to a target optimization strategy, and recovering the energy of the motor in the duration time;
and controlling the conversion circuit to work in a boosting mode in the duration time so as to boost back electromotive force at two ends of the motor and charge the energy storage device.
3. The energy recovery method of claim 2, wherein controlling the conversion circuit to operate in boost mode for the duration to boost back emf across the motor and charge the energy storage device comprises:
And controlling the conversion circuit to work in a boosting mode within the duration time, so that the voltage obtained by boosting back electromotive force at two ends of the motor by the conversion circuit is larger than the input voltage of the energy storage device, and the difference value between the boosted voltage and the input voltage of the energy storage device is within a preset range.
4. The energy recovery method of claim 2, wherein the target optimization strategy is determined in a manner that:
acquiring a historical data set of a server in a preset time period, wherein the historical data set comprises historical data corresponding to a plurality of moments, the historical data corresponding to each moment comprises data used for representing the state and action of the server at the current moment, and the action is used for adjusting the rotating speed of the motor;
constructing an objective function and a neural network model according to the historical data set, and training the neural network model according to the historical data set;
if the neural network model meets the preset condition, taking the neural network model meeting the preset condition as a target neural network model;
and when the target requirement is energy recovery, acquiring the current state of the server, and determining a target optimization strategy according to the current state and the target neural network model.
5. The energy recovery method of claim 4, further comprising, after training the neural network model from the historical dataset:
in each iteration process, obtaining a predicted value output by the neural network model;
judging whether the difference value between the predicted value and the actual value of the objective function is in a first threshold range or not;
if the neural network model is within the first threshold range, judging that the neural network model meets the preset condition; otherwise, the next iteration is entered.
6. The energy recovery method according to claim 5, wherein determining whether a difference between the predicted value and an actual value of the objective function is within a first threshold range comprises:
calculating an output value of a loss function according to the predicted value and the actual value;
judging whether the output value of the loss function is in a second threshold range or not;
and if the difference value is within the second threshold range, judging that the difference value is within the first threshold range.
7. The energy recovery method of claim 6, wherein determining whether the output value of the loss function is within a second threshold range comprises:
judging whether the output values of the loss function are all in the second threshold range in the continuous several iterative processes;
And if the output values of the loss function are all in the second threshold range in the continuous several iteration processes, judging that the difference value is in the first threshold range.
8. The energy recovery method of claim 6, wherein the loss function is expressed as:
；
loss is Loss function, x 1 Is constant and 0.ltoreq.x 1 Less than or equal to 1, s is the state of the server at the current moment, a is the action determined at the current moment, Q p (s, a) predicting output for the neural network modelThe value Q (s, a) is the actual value.
9. The energy recovery method of claim 4, wherein the constraints of the objective function are: the target optimization strategy is satisfied to keep the temperature within the server unchanged or drop.
10. The energy recovery method of claim 4, wherein the data in the history data that characterizes the state of the server at the current time includes system power data, temperature data, and motor speed data at the current time of the server.
11. The energy recovery method according to any one of claims 4 to 10, further comprising:
pre-establishing a corresponding relation between a state and an action of the server and a reward value, wherein the reward value is in negative correlation with the temperature in the server, and the reward value is in positive correlation with the electric quantity for boosting the counter electromotive force at two ends of the motor and charging the energy storage device;
Constructing an objective function from the historical dataset, comprising:
and constructing a target rewarding function according to the historical data set.
12. The energy recovery method of claim 11, wherein the target reward function has an expression:
；
where s is the state of the server at the current time, a is the action determined at the current time, s 'is the state of the server at the next time, a' is the action determined at the next time, Q (s, a) is the prize value obtained by taking action a in state s, r is the prize value existing at the current time, Q (s ', a') is the prize value obtained by taking action a 'in state s', and max is the maximum value.
13. The energy recovery method of claim 11, wherein the target reward function has an expression:
；
wherein s is the state of the server at the current time, a is the action determined at the current time, s 'is the state of the server at the next time, a' is the action determined at the next time, Q (s, a) is the prize value obtained by taking action a in the state s, r is the prize value existing at the current time, Q (s ', a') is the prize value obtained by taking action a 'in the state s', max is the maximum value, and c is the discount factor.
14. The energy recovery method of claim 13, wherein training the neural network model from the historical dataset comprises:
in each iteration process, historical data of the current moment and data used for representing the state of the server in the next moment are input into the neural network model, and the neural network model is triggered to determine a predicted rewarding value of the next moment corresponding to each action according to the historical data of the current moment, the data used for representing the state of the server in the next moment and the corresponding relation;
determining a maximum predicted prize value for a next time of the predicted prize values for each next time;
calculating a predicted rewarding value corresponding to the current moment output by the neural network model according to the maximum predicted rewarding value at the next moment and the target rewarding function;
and determining whether the neural network model meets the preset condition according to the predicted value corresponding to the current moment and the actual rewarding value of the current moment.
15. The utility model provides an energy recuperation system, its characterized in that is applied to the controller in the energy recuperation device, the energy recuperation device still includes DC power supply, converting circuit and energy storage device, DC power supply passes through converting circuit is connected with the motor in the fan, energy storage device respectively with DC power supply with converting circuit connects, the energy recuperation system includes:
The power supply unit is used for controlling the conversion circuit to work in an inversion mode when the target requirement is determined to be that the system is cooled, so as to invert the direct current output by the direct current power supply or the energy storage device into alternating current and supply power for the motor;
and the energy recovery unit is used for controlling the conversion circuit to work in a boosting mode when the target requirement is determined to be energy recovery, so as to boost the back electromotive force at two ends of the motor and charge the energy storage device.
16. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a computer program which, when executed by a controller, implements the steps of the energy recovery method according to any of claims 1-14.
17. The energy recovery device is characterized by comprising a controller, a direct-current power supply, a conversion circuit and an energy storage device, wherein the direct-current power supply is connected with a motor in a fan through the conversion circuit, the energy storage device is respectively connected with the direct-current power supply and the conversion circuit, and the controller is respectively connected with the conversion circuit and the motor;
the controller, when executing a computer program, is adapted to implement the steps of the energy recovery method of any one of claims 1-14.
18. The energy recovery device of claim 17, wherein the energy storage device is a capacitor.
19. The energy recovery device of claim 17, wherein the conversion circuit is a three-phase bridge circuit.
20. The energy recovery device of any one of claims 17-19, wherein an output positive terminal of the dc power source is connected to a positive terminal of the energy storage device and a positive terminal of the conversion circuit, respectively, and an output negative terminal of the dc power source is connected to a negative terminal of the energy storage device and a negative terminal of the conversion circuit, respectively, the energy recovery device further comprising:
the anti-backflow device is arranged between the output positive end of the direct current power supply and the positive end of the energy storage device, and is used for inverting the direct current output by the direct current power supply into alternating current and conducting the alternating current when supplying power to the motor, boosting the counter electromotive force at the two ends of the motor and stopping the direct current when charging the energy storage device.
21. The energy recovery device of claim 20, wherein the anti-backflow device comprises a field effect transistor and a diode, wherein a source electrode of the field effect transistor is respectively connected with an output positive terminal of the direct current power supply and an anode of the diode, a drain electrode of the field effect transistor is respectively connected with a positive terminal of the energy storage device and a cathode of the diode, and a grid electrode of the field effect transistor is connected with a driving circuit for driving the field effect transistor;
The field effect transistor is turned on when the direct current output by the direct current power supply is inverted into alternating current and is used for supplying power to the motor, and is turned off when the counter electromotive force at two ends of the motor is boosted and the energy storage device is charged.
22. A server comprising an energy recovery device according to any one of claims 17-21.
CN202311517860.8A 2023-11-15 2023-11-15 Energy recovery method, system, medium, device and server Active CN117254570B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311517860.8A CN117254570B (en) 2023-11-15 2023-11-15 Energy recovery method, system, medium, device and server

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311517860.8A CN117254570B (en) 2023-11-15 2023-11-15 Energy recovery method, system, medium, device and server

Publications (2)

Publication Number Publication Date
CN117254570A true CN117254570A (en) 2023-12-19
CN117254570B CN117254570B (en) 2024-02-13

Family

ID=89137204

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311517860.8A Active CN117254570B (en) 2023-11-15 2023-11-15 Energy recovery method, system, medium, device and server

Country Status (1)

Country Link
CN (1) CN117254570B (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101331673A (en) * 2005-12-14 2008-12-24 丰田自动车株式会社 Motor driving device and car provided with the same
US20150207359A1 (en) * 2012-07-05 2015-07-23 Volvo Construction Equipment Ab Battery charging system for hybrid construction machinery by using rotational force of fan and charging method therefor
CN116316755A (en) * 2023-03-07 2023-06-23 西南交通大学 Energy management method for electrified railway energy storage system based on reinforcement learning

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101331673A (en) * 2005-12-14 2008-12-24 丰田自动车株式会社 Motor driving device and car provided with the same
US20150207359A1 (en) * 2012-07-05 2015-07-23 Volvo Construction Equipment Ab Battery charging system for hybrid construction machinery by using rotational force of fan and charging method therefor
CN116316755A (en) * 2023-03-07 2023-06-23 西南交通大学 Energy management method for electrified railway energy storage system based on reinforcement learning

Also Published As

Publication number Publication date
CN117254570B (en) 2024-02-13

Similar Documents

Publication Publication Date Title
CN112519635B (en) Control method and related device for battery thermal management
JP5312504B2 (en) Energy management system
JP3905692B2 (en) Wind power generation control method
JP5553632B2 (en) Cooling control method for power conversion device of hybrid vehicle
JP5268853B2 (en) Hybrid cruise control system
CN113239563B (en) Thermal management method of IGBT module and storage medium
JP6299728B2 (en) Secondary battery management device
JP2001240323A (en) Control device of elevator
RU2011102346A (en) POWER MANAGEMENT FROM SEVERAL SOURCES IN THE ELEVATOR POWER SYSTEM
JP2010124634A (en) Charge/discharge controller
JP6215753B2 (en) How to operate an energy supply system
JP2015064120A (en) Air conditioning control system and air conditioning control method
JP2010088288A (en) Method and apparatus for power-bus stabilizer
KR20170104763A (en) System and method for managing temperature of battery
JP2014140281A (en) Control device, method, and program, and natural energy power generation device comprising the same
CN117254570B (en) Energy recovery method, system, medium, device and server
JP3897289B2 (en) Cogeneration equipment
JP2014049402A (en) Cooling method of secondary battery
WO2018235325A1 (en) Battery management device, battery management method, and power storage system
CN216389501U (en) Energy storage control device
JP5946210B2 (en) Power supply
JP7001519B2 (en) Battery management device, battery management method, power storage system
JP4810803B2 (en) Power supply
JP7146938B2 (en) Energy management systems, independent systems, and methods of operating independent systems
CN111384839A (en) Method for smoothing low-frequency junction temperature fluctuation of IGBT module in wind power converter

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant