A method and device for adaptive synchronous phasor measurement

Abstract

The present application discloses an adaptive synchronous phasor measurement method and device, the method comprising: collecting power signals of a power grid; extracting waveform features of the power signals, and inputting the waveform features into a transient recognition model and a steady-state recognition model for recognition, respectively, so as to obtain the signal state of the power grid, the signal state comprising a transient mode, a steady-state mode and a dynamic mode; selecting a corresponding synchronous phasor measurement method according to the signal state of the power grid to perform synchronous phasor calculation on the power grid, thereby improving the technical problem that the prior art applies the same measurement and control algorithm to the three modes of the power grid, and cannot simultaneously guarantee the measurement accuracy requirements and response speed requirements when measuring the synchronous phasors of the power grid.

Description

Technical Field

[0001] The present application relates to the technical field of synchronized phasor measurement, and in particular to an adaptive synchronized phasor measurement method and device. Background Art

[0002] The access of large-scale distributed power sources and electric vehicles to the power grid has greatly increased the complexity and insecurity of the power grid. In order to meet the requirements of real-time monitoring, fault diagnosis and rapid control of the power grid, it is necessary to measure the synchronous phasors of the power grid. The synchronous phasor measurement device is used to measure the synchronous phasors of the power grid.

[0003] The synchronous phasor measurement algorithm is the core of the synchronous phasor measurement device. The performance of the synchronous phasor measurement algorithm directly affects the accuracy of the state estimation, fault diagnosis and location, and source-grid-load coordinated control of the power grid. The power grid mode is divided into three types: steady state, transient, and dynamic. When the traditional synchronous phasor measurement algorithm measures the synchronous phasor of the power grid, it applies the same measurement and control algorithm to the three modes of the power grid, which has poor flexibility and can only guarantee one of the measurement accuracy requirements or the response speed requirements when measuring the synchronous phasor. It cannot guarantee both the measurement accuracy requirements and the response speed requirements when measuring the synchronous phasor of the power grid. Summary of the invention

[0004] The present application provides an adaptive synchronous phasor measurement method and device, which is used to solve the technical problem that the prior art applies the same measurement control algorithm to three modes of the power grid, and cannot simultaneously guarantee the measurement accuracy requirements and response speed requirements when measuring the synchronous phasors of the power grid.

[0005] In view of this, the first aspect of the present application provides an adaptive synchronized phasor measurement method, comprising:

[0006] Collect power signals from the power grid;

[0007] Extracting waveform features of the power signal, and inputting the waveform features into a transient recognition model and a steady-state recognition model for recognition, respectively, to obtain a signal state of the power grid, wherein the signal state includes a transient mode, a steady-state mode, and a dynamic mode;

[0008] A corresponding synchronized phasor measurement method is selected according to the signal state of the power grid to perform synchronized phasor calculation on the power grid.

[0009] Optionally, extracting the waveform feature of the power signal includes:

[0010] Sampling the power signal based on a preset sampling frequency to obtain a waveform sequence of the power signal;

[0011] The waveform sequence of the power signal is subjected to wavelet decomposition to obtain the waveform characteristics of the power signal.

[0012] Optionally, the inputting the waveform characteristics into a transient identification model and a steady-state identification model for identification respectively to obtain a signal state of a power grid includes:

[0013] Inputting the waveform characteristics into a transient identification model to identify whether the power grid is in a transient mode;

[0014] Inputting the waveform characteristics into a steady-state identification model to identify whether the power grid is in a steady-state mode;

[0015] If the transient identification model identifies that the power grid is in transient mode, the signal state of the output power grid is in transient mode;

[0016] If the steady-state identification model identifies that the power grid is in a steady-state mode, the signal state of the output power grid is in a steady-state mode;

[0017] If the transient identification model identifies that the power grid is in a non-transient mode and the steady-state identification model identifies that the power grid is in a non-steady-state mode, the signal state of the output power grid is a dynamic mode.

[0018] Optionally, the training process of the transient recognition model and the steady-state recognition model is:

[0019] Acquire historical power signals of the power grid, and extract waveform features of the historical power signals to obtain a training set;

[0020] Labeling some waveform features in the training set by category to obtain a labeled training set, and not labeling the remaining waveform features to obtain an unlabeled training set;

[0021] Training the long short-term memory network with the labeled training set to obtain a trained long short-term memory network;

[0022] Recognize the unlabeled training set by using the trained long short-term memory network, and label the unlabeled training set according to the recognition result to obtain a new training set;

[0023] Input random noise into the generator to obtain a noisy data set;

[0024] Training a generative adversarial network using the noise data set and the new training set to obtain a trained generative adversarial network;

[0025] Combining the trained generative adversarial network and the trained long short-term memory network to obtain a first recognition model;

[0026] When the categories marked in the training set are transient mode and non-transient mode, the first recognition model obtained by corresponding training is a transient recognition model;

[0027] When the categories annotated in the training set are steady-state mode and non-steady-state mode, the first recognition model obtained by corresponding training is a steady-state recognition model.

[0028] Optionally, the training process of the transient recognition model and the steady-state recognition model is:

[0029] Acquire historical power signals of the power grid, extract waveform features of the historical power signals, obtain a training set, and label the training set by category;

[0030] The first convolutional neural network model and the second convolutional neural network model are trained respectively by the training set to obtain a trained first convolutional neural network model and a trained second convolutional neural network model, wherein the first convolutional neural network model and the second convolutional neural network model are different convolutional neural network models;

[0031] Performing decision fusion on the trained first convolutional neural network model and the trained second convolutional neural network model to obtain a second recognition model;

[0032] When the categories marked in the training set are transient mode and non-transient mode, the second recognition model obtained by corresponding training is a transient recognition model;

[0033] When the categories annotated in the training set are steady-state mode and non-steady-state mode, the second recognition model obtained by corresponding training is a steady-state recognition model.

[0034] Optionally, selecting a corresponding measurement method according to the signal state of the power grid to measure the power grid includes:

[0035] When the signal state of the power grid is in transient mode, a rectangular window is selected as the sampling rate window function of the power signal, and the window length is set to a quarter cycle, and a transient algorithm is used to perform synchronized phasor calculation on the power signal in the data window to obtain a synchronized phasor measurement result of the power grid;

[0036] When the signal state of the power grid is in steady-state mode, the Rife_vincent window is selected as the sampling rate window function of the power signal, and the window length is set to four cycles. The steady-state algorithm is used to perform synchronized phasor calculation on the power signal in the data window to obtain the synchronized phasor measurement result of the power grid;

[0037] When the signal state of the power grid is in dynamic mode, the Hanning window is selected as the sampling rate window function of the power signal, and the window length is set to two cycles. A dynamic algorithm is used to perform synchronous phasor calculation on the power signal in the data window to obtain the synchronous phasor measurement result of the power grid.

[0038] Optionally, the adopting a dynamic algorithm to perform synchronized phasor calculation on the power signal in the data window to obtain a synchronized phasor measurement result of the power grid includes:

[0039] Adding harmonics, interharmonics and attenuated DC components to the signal model corresponding to the power signal to obtain a first power signal model, and discretizing the first power signal model to obtain a first discrete signal model;

[0040] Performing a windowed discrete Fourier transform on the first discrete signal model based on a Hanning window to obtain a first transformed signal model;

[0041] Performing a summation operation on the harmonic and interharmonic parts in the first transformed signal model to obtain a first harmonic and interharmonic model;

[0042] Performing a summation operation on the attenuated DC component part in the first transformed signal model to obtain an attenuated DC component model;

[0043] A first set of equations is constructed according to the first transformed signal model, the first interharmonic model and the attenuated DC component model, and a first phasor measurement result of the power signal is obtained by solving the first set of equations.

[0044] Optionally, the adopting a transient algorithm to perform synchronized phasor calculation on the power signal in the data window to obtain a synchronized phasor measurement result of the power grid includes:

[0045] adding harmonic and interharmonic components to the signal model corresponding to the power signal to obtain a second power signal model, and discretizing the second power signal model to obtain a second discrete signal model;

[0046] Performing a quarter-cycle discrete Fourier transform on the second discrete signal model based on a rectangular window to obtain a second transformed signal model;

[0047] Performing a summation operation on the harmonic and interharmonic parts in the second transformed signal model to obtain a second harmonic and interharmonic model;

[0048] A second set of equations is constructed according to the second transformed signal model and the second interharmonic model, and a second phasor measurement result of the power signal is obtained by solving the second set of equations.

[0049] Optionally, the steady-state algorithm is a synchronized phasor measurement method using fast Fourier transform correction.

[0050] A second aspect of the present application provides an adaptive synchronized phasor measurement device, comprising:

[0051] A collection unit, used for collecting power signals from the power grid;

[0052] an identification unit, configured to extract waveform features of the power signal, and input the waveform features into a transient identification model and a steady-state identification model for identification, so as to obtain a signal state of a power grid, wherein the signal state includes a transient mode, a steady-state mode, and a dynamic mode;

[0053] The calculation unit is used to select a corresponding synchronized phasor measurement method according to the signal state of the power grid to perform synchronized phasor calculation on the power grid.

[0054] It can be seen from the above technical solutions that this application has the following advantages:

[0055] The present application provides an adaptive synchronized phasor measurement method, comprising: collecting power signals of a power grid; extracting waveform features of the power signals, and inputting the waveform features into a transient recognition model and a steady-state recognition model for recognition, respectively, to obtain the signal state of the power grid, the signal state including a transient mode, a steady-state mode and a dynamic mode; and selecting a corresponding synchronized phasor measurement method according to the signal state of the power grid to perform synchronized phasor calculation on the power grid.

[0056] In the present application, it is considered that the power grid is in a relatively specific state at the same time, and there are relatively specific requirements for the phasor measurement algorithm. When the power grid is in a transient process such as a short circuit, the reaction is a step in the power signal. At this time, the performance indicators of the algorithm step response are relatively high; when the power grid is in a dynamic process, the power signal contains components such as attenuated DC and interharmonics. At this time, the algorithm is required to have the ability to suppress attenuated DC components, interharmonics and other components; and when the power grid is in a stable state, the algorithm is required to have higher accuracy. The present application identifies the state mode of the power grid through the waveform characteristics of the power signal, and switches to the synchronous phasor measurement method corresponding to the state mode to perform synchronous phasor calculation on the power grid. It can take into account the measurement accuracy requirements and response speed requirements when measuring the synchronous phasors of the power grid, and has high flexibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0058] Figure 1 A schematic flow chart of an adaptive synchronous phasor measurement method provided in an embodiment of the present application;

[0059] Figure 2 A schematic diagram of switching of data windows and measurement algorithms of a power grid under various signal states provided by an embodiment of the present application;

[0060] Figure 3 A structural schematic diagram of an adaptive synchronized phasor measurement device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0061] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0062] For easier understanding, see Figure 1 , the embodiment of the present application provides an adaptive synchronized phasor measurement method, comprising:

[0063] Step 101: Collect power signals from the power grid.

[0064] The power signal of the power grid collected in the present application can be a current signal or a voltage signal. The current signal can be collected by a current sensor, or the voltage signal can be collected by a voltage sensor. The embodiment of the present application preferably collects the current signal as the power signal. The current transformer is divided into a metering current transformer, a measuring current transformer and a protection current transformer. When the signal state of the power grid is identified according to the power signal, the current signal collected by the measuring current transformer can be used as the input signal for signal state identification. That is, in the embodiment of the present application, the current signal collected by the measuring current transformer in the power grid is obtained as the power signal of the power grid.

[0065] Step 102: extract the waveform features of the power signal, and input the waveform features into a transient recognition model and a steady-state recognition model for recognition, so as to obtain the signal state of the power grid, where the signal state includes a transient mode, a steady-state mode and a dynamic mode.

[0066] Consider that the power grid is in a relatively specific state at the same time, and has relatively specific requirements for the phasor measurement algorithm. For example, when the power grid is in a transient process such as a short circuit, the reaction on the power signal is a step, and at this time, the performance index of the algorithm step response is required to be high; when the power grid is in a dynamic process, the power signal contains attenuated DC, interharmonics and other components, and at this time, the algorithm is required to have the ability to suppress attenuated DC components, interharmonics and other components; and when the power grid is in a stable state, the algorithm is required to have higher accuracy. Based on this idea, a steady-state transient multivariate data acquisition algorithm based on high-precision synchronous sampling is proposed. The mode of the power grid is identified by the waveform characteristics of the power signal, and the algorithm corresponding to the mode is switched to measure.

[0067] Specifically, the power signal is sampled based on a preset sampling frequency to obtain a waveform sequence of the power signal; the waveform sequence of the power signal is subjected to wavelet decomposition to obtain waveform characteristics of the power signal. When performing wavelet decomposition, the initial scale function and wavelet function are first determined, and corresponding high-pass filters and low-pass filters are constructed. The waveform sequence is filtered multiple times using the high-pass filter and low-pass filter to obtain multiple wavelet coefficients and multiple scale coefficients, and the multiple wavelet coefficients and multiple scale coefficients are used as waveform characteristics of the power signal; the waveform characteristics are input into a transient recognition model to identify whether the power grid is in a transient mode; the waveform characteristics are input into a steady-state recognition model to identify whether the power grid is in a steady-state mode; if the transient recognition model identifies that the power grid is in a transient mode, the signal state of the output power grid is a transient mode; if the steady-state recognition model identifies that the power grid is in a steady-state mode, the signal state of the output power grid is a steady-state mode; if the transient recognition model identifies that the power grid is in a non-transient mode and the steady-state recognition model identifies that the power grid is in a non-steady-state mode, the signal state of the output power grid is a dynamic mode.

[0068] In one embodiment, the training process of the transient recognition model and the steady-state recognition model may be:

[0069] Obtain historical power signals of the power grid, extract waveform features of the historical power signals, and obtain a training set;

[0070] Some waveform features in the training set are labeled by category to obtain a labeled training set, and the remaining waveform features are not labeled to obtain an unlabeled training set;

[0071] Train the long short-term memory network through a labeled training set to obtain a trained long short-term memory network;

[0072] The unlabeled training set is identified through the trained long short-term memory network, and the unlabeled training set is labeled according to the identification results to obtain a new training set;

[0073] Input random noise into the generator to obtain a noisy data set;

[0074] Train the adversarial generative network through the noise data set and the new training set to obtain a trained adversarial generative network;

[0075] Combine the trained generative adversarial network and the trained long short-term memory network to obtain the first recognition model;

[0076] When the categories marked in the training set are transient mode and non-transient mode, the first recognition model obtained by corresponding training is a transient recognition model;

[0077] When the categories annotated in the training set are steady-state mode and non-steady-state mode, the first recognition model obtained by corresponding training is a steady-state recognition model.

[0078] The transient recognition model in the embodiment of the present application can be a GAN-LSTM model, or it can be other trained convolutional neural network models. The training process of the GAN-LSTM model can be: first, obtain the historical power signal of the power grid, use the above method to extract the waveform characteristics of the historical power signal, divide the waveform characteristics into two parts, and classify the waveform characteristics of one part, the labeled categories include transient mode and non-transient mode, and obtain a labeled training set, and the waveform characteristics of the other part are not labeled to obtain an unlabeled training set. Use the labeled training set to train the long short-term memory network (LSTM network) to obtain a trained LSTM model; use the trained LSTM model to identify the unlabeled training set, identify each waveform feature in the unlabeled training set as a transient mode or a non-transient mode, and label the mode of each waveform feature in the unlabeled training set based on the recognition result to obtain a new training set; then use the new training set to train the generator and discriminator in the adversarial generative network (GAN network), and finally obtain the GAN-LSTM model.

[0079] The process of training LSTM specifically includes: dividing the first part of the sample set into a training set and a validation set; inputting the training set into LSTM to train LSTM; inputting the validation set into LSTM to calculate the relative error of the trained LSTM; if the relative error meets the preset conditions, obtaining the final LSTM; if the relative error does not meet the preset conditions, re-dividing the first part of the sample set into a training set and a validation set, and continuing training.

[0080] The specific process of GAN training includes: inputting random noise into the generator to obtain a noise data set, and the noise data set is of the same type as the new training set; training the discriminator step: inputting the data set output by the generator and the new training set into the discriminator, and training the discriminator so that the discrimination rate of the discriminator for the new training set reaches a first threshold; training the generator step: inputting the new training set into the generator, and training the generator so that the similarity between the training data set generated by the generator and the new training set is greater than a second threshold; looping through the steps of training the discriminator and training the generator until the discrimination rate of the discriminator for the new training set reaches a third threshold (greater than the first threshold), and using the generator trained at this time as the final generator.

[0081] The trained generative adversarial network and the trained long short-term memory network are combined to obtain a transient recognition model, namely, a GAN-LSTM model. The generative adversarial network in the trained GAN-LSTM model is used to perform data enhancement on the input data, namely, to expand the input data and increase the amount of input data. After the GAN-LSTM model is trained, the waveform features are input into the transient recognition model, and new waveform features are generated according to the waveform features by the trained generative adversarial network to obtain waveform features after data enhancement. The waveform features after data enhancement are recognized by the trained long short-term memory network to obtain recognition results.

[0082] Among them, the transient recognition model and the steady-state recognition model can be the same model, for example, both can be GAN-LSTM models, and the training process of the two models is similar, except that the categories when the waveform features are marked are different. When training to obtain the steady-state recognition model, the category of the marked waveform features is a steady-state mode or a non-steady-state mode. The specific training process of the network is similar to that of the transient recognition model, which will not be repeated here. In addition, the transient recognition model and the steady-state recognition model can also be RNN-LSTM network models, and the training method for the RNN-LSTM network can adopt an existing training method, which will not be repeated here.

[0083] In another embodiment, the training process of the transient recognition model and the steady-state recognition model may also be:

[0084] Obtain historical power signals of the power grid, extract waveform features of the historical power signals, obtain a training set, and label the training set by category;

[0085] The first convolutional neural network model and the second convolutional neural network model are respectively trained by the training set to obtain a trained first convolutional neural network model and a trained second convolutional neural network model, wherein the first convolutional neural network model and the second convolutional neural network model are different convolutional neural network models;

[0086] Performing decision fusion on the trained first convolutional neural network model and the trained second convolutional neural network model to obtain a second recognition model;

[0087] When the categories marked in the training set are transient mode and non-transient mode, the second recognition model obtained by corresponding training is a transient recognition model;

[0088] When the categories annotated in the training set are steady-state mode and non-steady-state mode, the second recognition model obtained by corresponding training is a steady-state recognition model.

[0089] The first convolutional neural network model and the second convolutional neural network model can be a GAN-LSTM network and a RNN-LSTM network respectively. After the GAN-LSTM network and the RNN-LSTM network are trained, the GAN-LSTM network and the RNN-LSTM network can be subjected to decision fusion as a final transient recognition model or a steady-state recognition model.

[0090] Assume that the output channels of the GAN-LSTM network are X1,X2,...,X c , the output channels of the RNN-LSTM network are Y1,Y2,...,Y c After fusing the GAN-LSTM network and the RNN-LSTM network, the output channel is K is the fusion coefficient, and K i +K i+c = 1. That is, weighted fusion processing is performed on the output results of the two models, which can improve the accuracy of state recognition to a certain extent.

[0091] It is understandable that the signal state of the power grid can also be identified by real-time verification. Specifically, the power signal can be subjected to a first-level real-time verification. If the power signal passes the first-level real-time verification, the signal state of the power grid is determined to be a transient mode. If the power signal does not pass the first-level real-time verification, the power information is subjected to a second-level real-time verification. If the power signal passes the second-level real-time verification, the signal state of the power grid is determined to be a steady-state mode. If the power signal does not pass the second-level real-time verification, the signal state of the power grid is determined to be a dynamic mode.

[0092] Step 103: Select a corresponding synchronized phasor measurement method according to the signal state of the power grid to perform synchronized phasor calculation on the power grid.

[0093] Current transformers are divided into metering current transformers, measuring current transformers and protection current transformers. Limited by the principles of electromagnetic induction and magnetic saturation, a single set of current transformers cannot achieve high precision in normal current and short-circuit current at the same time, which results in the phasor algorithm being unable to achieve high precision and fast response at the same time. The embodiment of the present application proposes to solve the problem of high precision and unsaturated transient process when the transformer is normal by adopting the method of current transformer switching. That is, when the power grid is in steady-state mode and dynamic mode, the current signal collected by the measuring current transformer is used to perform power grid phasor measurement, which has high accuracy. When the power grid is in transient mode, the current signal collected by the protection current transformer is used to perform power grid phasor measurement, which has strong anti-saturation performance. In other words, when it is identified that the power grid is currently in steady-state mode and dynamic mode, the aforementioned collected power signal is used for phasor measurement. When it is identified that the power grid is currently in transient mode, the current signal collected by the protection current sensor is obtained as the power signal for phasor measurement.

[0094] When the power grid is in steady-state mode, the sampling rate of the device AD ​​has little effect on the phasor measurement algorithm, so a lower sampling rate can be used to reduce the amount of algorithm calculations and save software and hardware resources; at the same time, because the power grid is in a steady state, the voltage / current signal fluctuations are not large, and when the master station uses a lower sampling rate for the device, advanced applications can still accept it, which can save a lot of communication resources and master station software resources. When the power grid is in dynamic mode and transient mode, when a higher data density is required, increasing the AD sampling rate can improve the accuracy and speed of the algorithm; at the same time, increasing the upload rate of the device can improve the application effect of advanced applications.

[0095] The selection of the window function can be based on the length of the data window, combined with the phasor measurement algorithm and the amplitude-frequency characteristics of the window function, to select a specific window function to achieve the best effect. Typical window functions include rectangular window, Hanning window, Hamming window and Rife_vincent window. The function of the rectangular window is:

[0096]

[0097] The function of the Hanning window is:

[0098]

[0099] The function of the Hamming window is:

[0100]

[0101] The function of the Rife_vincent window is:

[0102]

[0103] Where h(n) is the filter array, i.e., the window function; N is the number of sampling points per cycle, and n is the index of the filter value.

[0104] When the main lobe width in the window function frequency response curve is narrower, the frequency resolution is higher; and the smaller the side lobe peak is, the less leakage is, the faster the attenuation is, and the stronger the suppression of leakage is. Therefore, the embodiment of the present application requires the window function to have the main lobe of its spectrum as narrow as possible and the side lobe peak as small as possible.

[0105] Considering the main lobe width, side lobe peak and side lobe attenuation rate in the window function design, the embodiment of the present application selects a specific window function based on the data window length. The selection result is that when the data window length is 1 / 4 cycle or shorter, a rectangular window is selected; when the data window length is 2 cycles, a Hanning window is selected; when the data window length is 4 cycles or more, a Rife_vincent window is selected. The specific schematic diagram is as follows Figure 2 shown.

[0106] When it is identified that the power grid is in transient mode, a transient algorithm that takes rapidity into consideration is used to perform synchronized phasor calculations on the data in the transient data window; when it is identified that the power grid is in steady-state mode, a steady-state algorithm that takes accuracy into consideration is used to perform synchronized phasor calculations on the data in the steady-state data window; when it is identified that the power grid is in dynamic mode, a dynamic algorithm that takes the influence of attenuated DC components and interharmonics into consideration is used to perform synchronized phasor calculations on the data in the dynamic data window.

[0107] When the signal state of the power grid is in steady-state mode, the Rife_vincent window is selected as the sampling rate window function of the power signal, and the window length is set to four cycles. The steady-state algorithm is used to perform synchronous phasor calculation on the power signal in the data window to obtain the synchronous phasor measurement result of the power grid; when the power signal is judged to be in steady-state mode, the data window is taken as 4 cycles, which mainly solves the influence of frequency deviation and harmonics. The DFT-corrected synchronous phasor measurement method can be used. This method belongs to the prior art and will not be described in detail here.

[0108] When the signal state of the power grid is in dynamic mode, the Hanning window is selected as the sampling rate window function of the power signal, and the window length is set to two cycles. The dynamic algorithm is used to perform synchronized phasor calculation on the power signal in the data window to obtain the synchronized phasor measurement results of the power grid.

[0109] The dynamic algorithm can solve the interference of harmonics, interharmonics and attenuated DC components on the accurate calculation of phasors by adding them to the signal model corresponding to the power signal. The specific implementation process includes: solving the frequency of interharmonics through fast Fourier transform; using a longer data window, scanning the amplitude of interharmonics through fast Fourier transform, so as to solve the frequency of interharmonics.

[0110] Solve the decay time constant of the decaying DC component. There are two methods for reference. Method 1 can obtain the decay time constant of the decaying DC component through the existing method of three adjacent data windows. Method 2, through the calculation and identification of power system parameters and existing empirical values, derive the time constant of the decaying DC component of the system.

[0111] When the interharmonics and the decaying DC component interfere at the same time, the measurement method in the embodiment of the present application can still accurately solve the fundamental phasor of the power signal. In order to suppress the interference of harmonics and interharmonics on the accurate calculation of the phasor, the harmonics, interharmonics and decaying DC component components are considered in the signal model, and the first power signal model obtained can be expressed as:

[0112]

[0113] Where x(t) is the power signal, are harmonic and interharmonic components, R(t) is the attenuated DC component, and the calculation formula is: λ i is the amplitude of the i-th harmonic or interharmonic; α i f0 is the frequency of the ith harmonic or interharmonic; φ i is the phase angle of the i-th harmonic or interharmonic; when α i When it is an integer, it represents the order of the i-th harmonic. i When it is a non-integer, it represents the coefficient of the interharmonic frequency; C' is the initial amplitude of the attenuated DC component; T c is the decay time constant of the attenuated DC component; f0 is the rated frequency of the power system. In order to better reflect the characteristics of amplitude and phase angle changing with time in the dynamic state of the power system, the amplitude phase angle polynomial form P(t) is used to express it: P(t) = a(t)e jθ(t) , a(t) and θ(t) represent the amplitude and phase angle of the voltage signal or current signal of the power grid respectively, and P*(t) is the conjugate of P(t).

[0114] The first power signal model is discretized to obtain a first discrete signal model, and then the first discrete signal model is subjected to a windowed DFT transform, where the window function is h(n) and the DFT transform coefficient is The first transformed signal model is obtained, namely:

[0115]

[0116] Where α0 = f c / f0 is the ratio of the measured fundamental frequency to the rated frequency, which is used for the center frequency adaptive adjustment coefficient; R0 is the real part of the fundamental phasor; I0 is the imaginary part of the fundamental phasor. After obtaining the fundamental frequency, the center frequency of the filter is adjusted in real time. The filter with adaptive center frequency adjustment can provide better performance when the frequency is deviated, such as solving the problem of harmonic interference in the case of large frequency deviation.

[0117] The interharmonic parts of the harmonics are summed to obtain the first harmonic interharmonic model:

[0118]

[0119] Where, X Qk is the kth DFT transform value of the interharmonic part; X QkR and X QkI are the real and imaginary parts of the kth DFT transform value of the interharmonic part respectively.

[0120] The decaying DC component is summed to obtain the first decaying DC component model:

[0121]

[0122] Where, X Rk is the kth DFT transform value of the attenuated DC component; X RkR and X RkI are the real and imaginary parts of the kth DFT transform value of the attenuated DC component, respectively.

[0123] The first set of equations is constructed by using the first transformed signal model, the first interharmonic model and the first attenuated DC component model. Specifically, the complex domain equation is expanded into the real part and the imaginary part form to obtain the first set of equations:

[0124]

[0125] Where, X k =[X kR X kI ] T is the real and imaginary part of the kth discrete Fourier transform result of the original measurement data; M k =[M k0 ,N k1 ,...,N kK ,O k ] are the coefficients of the first system of equations, where

[0126]

[0127]

[0128]

[0129] In the formula, are the parameters of the signal model, and are the real and imaginary parts of the fundamental wave phasor respectively; R hi and I hi are the real and imaginary parts of the ith harmonic or interharmonic phasor respectively; C' is the initial amplitude of the decaying DC component.

[0130] When k=0,1,...,K, the system of simultaneous equations is simplified to:

[0131]

[0132] In the formula, are the coefficients of the equation system, is the DFT coefficient of the original power signal.

[0133] Next, solve the phasor of the power signal. Consider the following in the equation coefficient M: 1) The power system frequency real-time frequency f c It can be obtained by modifying the DFT algorithm, so the center frequency adaptive adjustment coefficient α0 = f c / f0 is known, so the coefficient M k0 2) The interharmonic frequency has been obtained by fast Fourier transform, and the interharmonic α i =f i / f0 is known, so the coefficient N ki Known; 3) The decay time constant T of the decaying DC component c It can be obtained according to the existing method, so the coefficient O k Known. Coefficients of the equations is a known quantity. Considering that in order to avoid the ill-conditioning of the equation system, the parameter g in the DFT coefficient k The selection should make the matrix M -1 The number of conditions should be as small as possible.

[0134] From the above calculation, we can get the parameter P of the signal model, and the real part R0 and imaginary part I0 of the fundamental phasor of the power signal model, so the amplitude of the power signal is The phase angle is tanθ0=I0 / R0, so the phasor of the power signal is calculated to achieve phasor measurement.

[0135] When the signal state of the power grid is in transient mode, a rectangular window is selected as the sampling rate window function of the power signal, and the window length is set to a quarter cycle. The transient algorithm is used to perform synchronous phasor calculation on the power signal in the data window to obtain the synchronous phasor measurement result of the power grid. When it is recognized that the power grid is currently in transient mode, in order to improve the response speed, a transient algorithm with an ultra-short data window is used. The transient algorithm in this application adopts a quarter-cycle frequency domain algorithm that takes into account the influence of harmonics and interharmonics. In order to suppress the interference of harmonics and interharmonics on the accurate calculation of phasors, harmonic and interharmonic components are considered in the signal model, and the second power signal model can be expressed as:

[0136]

[0137] Discretizing the second power signal model to obtain a second discrete signal model;

[0138] Based on the rectangular window, a quarter-cycle discrete Fourier transform is performed on the second discrete signal model. The discrete Fourier transform coefficient is The signal model after the second transformation is obtained, namely:

[0139]

[0140] The harmonic and interharmonic parts in the second transformed signal model are summed to obtain the second harmonic interharmonic model. The specific process is similar to the acquisition process of the first harmonic interharmonic model, which will not be repeated here. The second set of equations is constructed according to the second transformed signal model and the second harmonic interharmonic model, and the complex domain equation is expanded into the real part and imaginary part form to obtain the second set of equations:

[0141]

[0142] By solving the second set of equations, the second phasor measurement result of the power signal is obtained. The solution process can refer to the dynamic algorithm and add the harmonic and interharmonic models to the signal model to solve the interference caused by harmonics and interharmonics to the accurate calculation of the phasor.

[0143] In the embodiments of the present application, it is considered that the power grid is in a relatively specific state at the same time and has relatively specific requirements for the phasor measurement algorithm. When the power grid is in a transient process such as a short circuit, the reaction is a step in the power signal. At this time, the performance index of the algorithm step response is relatively high; when the power grid is in a dynamic process, the power signal contains components such as attenuated DC and interharmonics. At this time, the algorithm is required to have the ability to suppress attenuated DC components, interharmonics and other components; and when the power grid is in a stable state, the algorithm is required to have higher accuracy. The present application identifies the mode of the power grid through the characteristics of the power signal, and switches to the synchronous phasor measurement method corresponding to the mode to perform synchronous phasor calculation on the power grid. It can take into account the measurement accuracy requirements and response speed requirements when measuring the synchronous phasors of the power grid, and has high flexibility.

[0144] Please refer to Figure 3 The embodiment of the present application further provides an adaptive synchronous phasor measurement device, comprising:

[0145] A collection unit, used for collecting power signals from the power grid;

[0146] An identification unit is used to extract waveform features of the power signal, and input the waveform features into a transient identification model and a steady-state identification model for identification, so as to obtain the signal state of the power grid, where the signal state includes a transient mode, a steady-state mode and a dynamic mode;

[0147] The calculation unit is used to select a corresponding synchronized phasor measurement method according to the signal state of the power grid to perform synchronized phasor calculation on the power grid.

[0148] In the embodiments of the present application, it is considered that the power grid is in a relatively specific state at the same time and has relatively specific requirements for the phasor measurement algorithm. When the power grid is in a transient process such as a short circuit, the reaction is a step in the power signal. At this time, the performance index of the algorithm step response is relatively high; when the power grid is in a dynamic process, the power signal contains components such as attenuated DC and interharmonics. At this time, the algorithm is required to have the ability to suppress attenuated DC components, interharmonics and other components; and when the power grid is in a stable state, the algorithm is required to have higher accuracy. The present application identifies the mode of the power grid through the characteristics of the power signal, and switches to the synchronous phasor measurement method corresponding to the mode to perform synchronous phasor calculation on the power grid. It can take into account the measurement accuracy requirements and response speed requirements when measuring the synchronous phasors of the power grid, and has high flexibility.

[0149] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0150] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0151] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An adaptive synchronized phasor measurement method, characterized in that: include: Collect power signals from the power grid; Extracting waveform features of the power signal, and inputting the waveform features into a transient recognition model and a steady-state recognition model for recognition, respectively, to obtain a signal state of the power grid, wherein the signal state includes a transient mode, a steady-state mode, and a dynamic mode; The training process of the transient recognition model and the steady-state recognition model is: Acquire historical power signals of the power grid, and extract waveform features of the historical power signals to obtain a training set; Labeling some waveform features in the training set by category to obtain a labeled training set, and not labeling the remaining waveform features to obtain an unlabeled training set; Training the long short-term memory network with the labeled training set to obtain a trained long short-term memory network; Recognize the unlabeled training set by using the trained long short-term memory network, and label the unlabeled training set according to the recognition result to obtain a new training set; Input random noise into the generator to obtain a noisy data set; Training a generative adversarial network using the noise data set and the new training set to obtain a trained generative adversarial network; Combining the trained generative adversarial network and the trained long short-term memory network to obtain a first recognition model; When the categories marked in the training set are transient mode and non-transient mode, the first recognition model obtained by corresponding training is a transient recognition model; When the category marked in the training set is a steady-state mode and a non-steady-state mode, the first recognition model obtained by corresponding training is a steady-state recognition model; Or, obtaining historical power signals of the power grid, and extracting waveform features of the historical power signals to obtain a training set, and classifying the training set; The first convolutional neural network model and the second convolutional neural network model are trained respectively by the training set to obtain a trained first convolutional neural network model and a trained second convolutional neural network model, wherein the first convolutional neural network model and the second convolutional neural network model are different convolutional neural network models; Performing decision fusion on the trained first convolutional neural network model and the trained second convolutional neural network model to obtain a second recognition model; wherein the decision fusion process is weighted fusion of the output results of the trained first convolutional neural network model and the trained second convolutional neural network model; When the categories marked in the training set are transient mode and non-transient mode, the second recognition model obtained by corresponding training is a transient recognition model; When the category marked in the training set is a steady-state mode and a non-steady-state mode, the second recognition model obtained by corresponding training is a steady-state recognition model; A corresponding synchronized phasor measurement method is selected according to the signal state of the power grid to perform synchronized phasor calculation on the power grid.

2. The adaptive synchronized phasor measurement method according to claim 1, characterized in that: The extracting the waveform feature of the power signal comprises: Sampling the power signal based on a preset sampling frequency to obtain a waveform sequence of the power signal; The waveform sequence of the power signal is subjected to wavelet decomposition to obtain the waveform characteristics of the power signal.

3. The adaptive synchronized phasor measurement method according to claim 1, characterized in that: The step of inputting the waveform characteristics into a transient identification model and a steady-state identification model for identification to obtain a signal state of a power grid includes: Inputting the waveform characteristics into a transient identification model to identify whether the power grid is in a transient mode; Inputting the waveform characteristics into a steady-state identification model to identify whether the power grid is in a steady-state mode; If the transient identification model identifies that the power grid is in transient mode, the signal state of the output power grid is in transient mode; If the steady-state identification model identifies that the power grid is in a steady-state mode, the signal state of the output power grid is in a steady-state mode; If the transient identification model identifies that the power grid is in a non-transient mode and the steady-state identification model identifies that the power grid is in a non-steady-state mode, the signal state of the output power grid is a dynamic mode.

4. The adaptive synchronized phasor measurement method according to claim 1, characterized in that: The selecting a corresponding measurement method according to the signal state of the power grid to measure the power grid includes: When the signal state of the power grid is in transient mode, a rectangular window is selected as the sampling rate window function of the power signal, and the window length is set to a quarter cycle, and a transient algorithm is used to perform synchronized phasor calculation on the power signal in the data window to obtain a synchronized phasor measurement result of the power grid; When the signal state of the power grid is in steady-state mode, the Rife_vincent window is selected as the sampling rate window function of the power signal, and the window length is set to four cycles. The steady-state algorithm is used to perform synchronized phasor calculation on the power signal in the data window to obtain the synchronized phasor measurement result of the power grid; When the signal state of the power grid is in dynamic mode, the Hanning window is selected as the sampling rate window function of the power signal, and the window length is set to two cycles. A dynamic algorithm is used to perform synchronous phasor calculation on the power signal in the data window to obtain the synchronous phasor measurement result of the power grid.

5. The adaptive synchronized phasor measurement method according to claim 4, characterized in that: The method of using a dynamic algorithm to perform synchronized phasor calculation on the power signal in the data window to obtain synchronized phasor measurement results of the power grid includes: Adding harmonics, interharmonics and attenuated DC components to the signal model corresponding to the power signal to obtain a first power signal model, and discretizing the first power signal model to obtain a first discrete signal model; Performing a windowed discrete Fourier transform on the first discrete signal model based on a Hanning window to obtain a first transformed signal model; Performing a summation operation on the harmonic and interharmonic parts in the first transformed signal model to obtain a first harmonic and interharmonic model; Performing a summation operation on the attenuated DC component part in the first transformed signal model to obtain an attenuated DC component model; A first set of equations is constructed according to the first transformed signal model, the first interharmonic model and the attenuated DC component model, and a first phasor measurement result of the power signal is obtained by solving the first set of equations.

6. The adaptive synchronized phasor measurement method according to claim 4, characterized in that: The method of using a transient algorithm to perform synchronized phasor calculation on the power signal in the data window to obtain synchronized phasor measurement results of the power grid includes: adding harmonic and interharmonic components to the signal model corresponding to the power signal to obtain a second power signal model, and discretizing the second power signal model to obtain a second discrete signal model; Performing a quarter-cycle discrete Fourier transform on the second discrete signal model based on a rectangular window to obtain a second transformed signal model; Performing a summation operation on the harmonic and interharmonic parts in the second transformed signal model to obtain a second harmonic and interharmonic model; A second set of equations is constructed according to the second transformed signal model and the second interharmonic model, and a second phasor measurement result of the power signal is obtained by solving the second set of equations.

7. The adaptive synchronized phasor measurement method according to claim 4, characterized in that: The steady-state algorithm is a synchronized phasor measurement method using fast Fourier transform correction.

8. An adaptive synchronized phasor measurement device, characterized in that: include: A collection unit, used for collecting power signals from the power grid; an identification unit, configured to extract waveform features of the power signal, and input the waveform features into a transient identification model and a steady-state identification model for identification, so as to obtain a signal state of a power grid, wherein the signal state includes a transient mode, a steady-state mode, and a dynamic mode; The training process of the transient recognition model and the steady-state recognition model is: Acquire historical power signals of the power grid, and extract waveform features of the historical power signals to obtain a training set; Labeling some waveform features in the training set by category to obtain a labeled training set, and not labeling the remaining waveform features to obtain an unlabeled training set; Training the long short-term memory network with the labeled training set to obtain a trained long short-term memory network; Recognize the unlabeled training set by using the trained long short-term memory network, and label the unlabeled training set according to the recognition result to obtain a new training set; Input random noise into the generator to obtain a noisy data set; Training a generative adversarial network using the noise data set and the new training set to obtain a trained generative adversarial network; Combining the trained generative adversarial network and the trained long short-term memory network to obtain a first recognition model; When the categories marked in the training set are transient mode and non-transient mode, the first recognition model obtained by corresponding training is a transient recognition model; When the category marked in the training set is a steady-state mode and a non-steady-state mode, the first recognition model obtained by corresponding training is a steady-state recognition model; Or, obtaining historical power signals of the power grid, and extracting waveform features of the historical power signals to obtain a training set, and classifying the training set; The first convolutional neural network model and the second convolutional neural network model are trained respectively by the training set to obtain a trained first convolutional neural network model and a trained second convolutional neural network model, wherein the first convolutional neural network model and the second convolutional neural network model are different convolutional neural network models; Performing decision fusion on the trained first convolutional neural network model and the trained second convolutional neural network model to obtain a second recognition model; wherein the decision fusion process is weighted fusion of the output results of the trained first convolutional neural network model and the trained second convolutional neural network model; When the categories marked in the training set are transient mode and non-transient mode, the second recognition model obtained by corresponding training is a transient recognition model; When the category marked in the training set is a steady-state mode and a non-steady-state mode, the second recognition model obtained by corresponding training is a steady-state recognition model; The calculation unit is used to select a corresponding synchronized phasor measurement method according to the signal state of the power grid to perform synchronized phasor calculation on the power grid.

Images