JP6132739B2 - Method for detecting conditions in a power grid - Google Patents

Description

  The present invention relates generally to detecting conditions in a power grid, and in particular to detecting transient events.

Grid transients In grids, transients are short-term disturbances in voltage or current. Transient events are usually due to unexpected loads or generator mismatches. Some transients are due to external natural phenomena such as lightning. However, the majority of transient events are caused internally by, for example, load switching, breaker switching, fuse blown, short circuit or islanding. During the transient event, the auxiliary generator or distributed generator continues to supply power locally in the grid, even when there is no longer any power from the main power source.

  The amount of change in voltage or current during a transient event depends on the case. For example, a short circuit can cause a large current increase. The cumulative effects of transients can damage semiconductors that are heavily integrated into modern power grids. Unintended transients can be dangerous for workers, consumers and equipment. Therefore, transient events must be detected.

  Spectral analysis of transient events is known. Conventional approaches include analysis based on Fourier transforms and wavelet transforms. A power grid can be described by a differential equation. Another approach uses signal parameter estimation (ESPRIT) with rotational invariance techniques.

  FIG. 1 illustrates prior art transient event detection. A rectifier 110 is connected to an auxiliary power source 100, eg, a distributed power source such as a solar panel, wind turbine or backup generator, and to an inverter 120, eg, a single or three phase inverter. The inverter is also connected to the main power supply 140 via the circuit breaker 130. The main power source may be one or more conventional large generators or other power grids. The output of the prediction controller 190 is connected to the inverter. The input to the controller includes a sample 180 of the voltage to be measured (V) and a moving average 170 of frequency (F). The inverter and the main power source can be connected to the load 150.

  U.S. Pat. No. 6,057,097 detects a transient event condition by changing the phase error response in a phase locked loop and determining whether the transient is based on the changed phase error response.

  Patent Document 2 detects a transient event by superimposing a detectable signal on a transmission line voltage at a power transmission network location outside the power plant and monitoring a detectable signal different from the transmission line voltage at the power plant.

  Patent document 3 detects a transient event using a photovoltaic inverter. The inverter calculates the degree of correlation between the grid AC frequency and the electrically closest AC frequency. If the degree of correlation falls below a predetermined value or exhibits a certain pattern or behavior, this indicates a loss of main power.

Power saving (Brown out)
Patent document 4 of a related application describes detecting an unbalanced state in a power transmission network using a probability density function. The unbalanced state may be due to a significant voltage drop such as during power saving. This condition can extend for a relatively long period of time, eg, minutes to hours, compared to transients on the order of milliseconds.

U.S. Pat. No. 8,195,414 US Pat. No. 7,376,491 US Patent Application Publication No. 2011/0276192 US Patent Application No. 13 / 329,720

  Embodiments of the present invention provide a method and system for detecting transient events in a power grid. The detection uses a binary hypothesis test and a general likelihood ratio test (GLRT). GLRT involves maximum likelihood estimation (MLE) of unknown model parameters.

  Since this is a complex problem with respect to computation, embodiments of the present invention obtain unknown signal parameters via signal parameter estimation (ESPRIT) by rotational invariance techniques and replace MLE for parameter estimation. Used for.

  This method does not require prior knowledge of the fundamental frequency, initial phase or amplitude of the voltage. Therefore, the performance of the method is maintained even in non-ideal situations such as frequency deviation from the nominal frequency of the grid.

  Embodiments of the present invention construct a probability distribution function (pdf) of a sample under measurement of a voltage signal in a power grid to detect a transient event sample. Since the normal pdf and the isolated pdf have different means and covariances, they can be used.

The likelihood ratio test determines whether a signal sample corresponds to a normal state hypothesis (H ₀ ) or a transient event state hypothesis (H ₁ ).

  In one embodiment, transient event detection is formulated as a parametric test, and the problem is solved by using a generalized likelihood ratio test (GLRT). As used by this embodiment, GLRT is a statistical test to compare the suitability of two models, one of which is a model for normal conditions and the other is It is a model for transient event conditions. The test is based on the likelihood ratio of these models.

  For example, if the GLRT ratio is not equal to 1, a transient event is detected. In one variation of this embodiment, a transient event is detected if the difference between 1 and its ratio is greater than a threshold value.

  The power grid is an unsteady network, and transients occur in voltage and current as a result of frequent changes in internal and external circuits. Transient events can degrade power quality, cause damage to dedicated electrical devices, or can render vital electrical equipment inoperable.

The present invention provides a method and system for detecting transient events using binary hypothesis testing. The normal state hypothesis (H ₀ ) in the normal state is modeled as a single sinusoid. An alternative transient event state hypothesis (H ₁ ) is modeled as a sum of decaying sinusoids.

  Since the parameters of the model are unknown, a general likelihood ratio test for the probability distribution function constructed from the voltage samples of the grid is used. The likelihood ratio of pdf is compared with a threshold value to determine whether the network condition is normal or an abnormal transient.

1 is a block diagram of a prior art transient event detector. FIG. 1 is a block diagram of a method and system for detecting transient events in a power grid according to one embodiment of the invention. FIG. FIG. 2 is a block diagram of a stochastic transient detector and process according to an embodiment of the present invention. FIG. 4 is a block diagram of parameter estimation under transient event conditions according to one embodiment of the present invention. It is a figure of the signal waveform used by embodiment of this invention.

System Method Overview FIG. 2 illustrates a method and system for detecting transients within a single-phase or three-phase power grid. A rectifier 210 is connected to an auxiliary power source 200, for example a distributed power source such as a solar panel, wind turbine or backup generator, and to an inverter 220. The inverter is also connected to the main power supply 240 via the circuit breaker 230. The output of the stochastic detector 270 is connected to the inverter. The input to the detector includes a sample 260 of measured voltage (V), which will be described in more detail later. These samples can be normal samples or transient samples. The inverter and main power supply 240 can also be connected to a load 250. The output of the detector is a determination 280 of whether a transient event has been detected. The determination can be used to “connect” a power source in the grid according to the detected state. As used herein, “connecting” can mean connecting or disconnecting any of the power sources in the power grid.

  Since the time to respond to this condition is relatively short, the method measures 260 a voltage signal sample 265 in the grid and obtains the sample. Stochastic detectors detect transient event samples during transient event conditions by analyzing those samples in real time and fitting the samples to a probability density function, as described in more detail later, or normal Detect normal samples during the condition.

  Waveform 241 shows a typical transient occurrence, and the average amplitude 242 of the output voltage of main power supply 240 decreases rapidly in less than a few milliseconds. The waveform example lasts for 1 second. Therefore, the present inventor continuously samples at about 8 kHz and uses a sliding window to test the current short-term grid conditions.

  This situation is quite different from the situation described in the related application. Instead of a long-term decrease in voltage that can be recovered at any time, here, for example, the voltage amplitude suddenly decreases to zero. It is important to quickly recognize this condition so that an auxiliary power supply can be connected. This is particularly important for applications where loss of power to the electrical device can be fatal. Alternatively, an appropriate response may be to disconnect the main power source.

Voltage Signal Representation Normal During normal power grid conditions, the voltage signal can be represented as a sine curve of equation (1) below.

は、下式によって得ることができる。 Where n is an integer sampling index, a ₀ is the complex amplitude of the fundamental frequency component, ie, the product of amplitude and initial phase, v (n) represents noise, and ω ₀ is normalized Fundamental frequency. Analog board frequency

Can be obtained by the following equation.

  However, Δt represents a sampling interval. Note that the signal models for voltage and current are mathematically the same.

When a transient event occurs If the grid is damaged by disturbances due to load switching and abnormal conditions such as transient events, the voltage or current signal is subjected to a short-term transient event. In general, transient events can be modeled as a sum of decaying sinusoids.

Where γ _i > 0 is the normalized attenuation coefficient for the i-th component of a particular sinusoid. The number of sinusoids is M> 0.

Binary Hypothesis Testing Based on the above, the normal state hypothesis H ₀ is modeled as a sinusoid and the transient event state hypothesis H ₁ is modeled as a sum of decaying sinusoids.

  Therefore, the normal state hypothesis is:

  Moreover, the hypothesis of the transient event state is as follows.

For N samples, y = [y (0), y (1),..., Y (N−1)] ^T and v = [v (0), v (1),. v (N-1)] ^T is defined. Therefore, the binary hypothesis test can be concisely expressed in matrix form as follows.

Here, a = [a ₁ , a ₂ ,..., A _M ] ^T , and C (N, M) is a maximum rank Vandermonde matrix having a geometric sequence term in each row. The matrix is a function of the frequency w _i , the damping factor γ _i , and the number of sinusoids M in the event of a transient event.

The variables a, w _i , γ _i and M are unknown.

Stochastic Transient Event Detector FIG. 3 illustrates a stochastic transient event detector and process according to an embodiment of the present invention. The detector can be implemented in a separate electronic circuit, or a general purpose processor or microprocessor. The method steps may be implemented as a circuit or process module. The method steps may be performed by a processor connected to a memory and input / output interface and a power grid.

The input to the detector is a normal / transient event sample 260 measured under normal and transient event conditions. The combined probability density function (pdf) 340 of a set of voltage samples under the normal state hypothesis H ₀ includes an unknown vector θ ₀ that includes a ₀ and w ₀ . The set of observed joint probability density functions (pdf) 350 under the transient event state hypothesis H ₁ includes an unknown vector θ ₁ that includes a _i and w _i .

が求められ（３２０）、仮説Ｈ_０またはＨ_１に対する判定を得るためにしきい値と比較される（３３０）。 The unknown vector θ ₀ is estimated using a maximum likelihood estimator (300). The parameters C and a for H ₁ are estimated using ESPRIT (310) as shown in FIG. Likelihood ratio of outputs of steps 300 and 310

Is determined (320), it is compared with a threshold to obtain a decision for hypothesis _{H 0} or _{H 1} (330).

  As shown in FIG. 4, the model order M is estimated using the maximum description length (MDL) (410), thereby allowing the selection of the hypothesis that most closely corresponds to the regularity in the data. Parameters C and a are estimated using ESPRIT (420).

In the case of H ₀ and H ₁ , after all unknown parameters a ₀ , w ₀ , C, a and M have been estimated, a transient event is detected using the generalized likelihood ratio 320 (250), which is Is compared with a predetermined threshold T (330). The threshold can be based on a desired false alarm probability (PFA). The higher the PFA, the smaller the threshold.

Attenuated signal consisting of multiple sinusoids FIG. 5 shows the original signal and three components. The first component plays a role for the fundamental frequency, for example 60 Hz in this example. For time, samples in one cycle of the fundamental frequency are used. The signal to noise ratio (SNR) between the first component and the noise is about 30 dB.

Claims (9)

A method for detecting a condition in a power grid,
Modeling the normal state hypothesis as a sinusoid and the transient event state hypothesis as a sum of damped sinusoids;
Constructing a probability density function (pdf) from voltage samples in the grid;
Comparing a likelihood ratio based on the pdf and the hypothesis with a threshold to determine a state; and
Including
A method for detecting a condition in a power grid, wherein said step is performed in a processor. The method of claim 1, further comprising connecting a device to the power grid according to the state. The method of claim 1, wherein the voltage is single phase or three phase. The method of claim 1, further comprising analyzing the sample in a stochastic detector and applying the sample to the pdf. The sample includes a normal sample and a transient event sample, the normal state hypothesis is H ₀ : y = a ₀ e, the transient event state hypothesis is H ₁ : y = Ca, and y is 2. The voltage, a is the amplitude, e is the frequency of each normal sample, and C is the frequency, damping factor and sinusoidal maximum order Vandermonde matrix for the transient event sample. the method of. _A 0 using a signal parameter estimation with the rotation invariance techniques (ESPRIT), a, method of claim 5, further comprising the step of estimating the e and C. The comparing step includes:
Determining a likelihood ratio of the normal pdf and the transient event pdf;
The method of claim 1, further comprising: comparing the likelihood ratio to a threshold value. The method of claim 5, further comprising: estimating a model order using a maximum description length to obtain the normal state hypothesis. The method of claim 7, wherein the step of comparing with the threshold is based on a desired false alarm probability.

Images