Power distribution network operation state analysis method, electronic equipment and medium

Abstract

The invention discloses a power distribution network operation state analysis method, electronic equipment and a medium, and relates to the technical field of deep learning. According to the scheme, a prediction result is inaccurate due to irregular change of original data when the operation state of a power distribution network is predicted through the original data of the power distribution network; according to the scheme, the unstable historical operation state information of the power distribution network is decomposed in a wavelet packet decomposition mode, multi-level frequency processing is carried out on the current operation state signal of the power distribution network, the power distribution network operation state prediction model obtained after neural network training can predict the future state of the power distribution network more accurately, and in addition, the prediction accuracy of the power distribution network operation state prediction model is improved. The power distribution network operation state prediction model has the function of monitoring the real-time operation state information of the power distribution network in real time, so that the future operation state information can be adjusted based on the real-time operation state information, the accuracy of the future operation state information is improved, and the accuracy of the power distribution network operation state analysis process is further improved.

Description

Technical Field

[0001] The present invention relates to the field of deep learning technology, and in particular to a distribution network operating status analysis method, electronic equipment, and medium. Background Art

[0002] The operating status of a distribution network directly impacts its power supply reliability, safety, and stability. Traditional methods for analyzing distribution network operating status often rely on manual feature engineering. Manual feature engineering is the process of manually defining, selecting, and constructing data features based on domain expertise. These features are typically based on physical laws, grid operation knowledge, or statistical laws. They are used to describe the distribution network's operating status and support subsequent analysis (such as fault detection, load forecasting, and stability assessment). However, this approach can lead to inaccurate predictions of the network's operating status due to irregular variations in raw data (such as voltage, current, and power). Summary of the Invention

[0003] The object of the present invention is to provide a distribution network operating status analysis method, electronic device and medium. This scheme takes into account that predicting the operating status of the distribution network through the original data of the distribution network will lead to inaccurate prediction results due to irregular changes in the original data. Therefore, this scheme decomposes the historical operating status information of the unstable distribution network through wavelet packet decomposition to perform multi-level frequency division (both low-frequency signals and high-frequency signals) on the current operating status signal of the distribution network, so that the distribution network operating status prediction model obtained after neural network training can more accurately predict the future state of the distribution network. In addition, because the distribution network operating status prediction model is obtained based on the historical operating status information of the distribution network, it also has the function of real-time monitoring of the real-time operating status information of the distribution network, and can adjust the future operating status information based on the real-time operating status information to improve the accuracy of the future operating status information, thereby improving the accuracy of the distribution network operating status analysis process.

[0004] To solve the above technical problems, the present invention provides a method for analyzing the operating status of a distribution network, comprising:

[0005] Obtain historical operating status information of the distribution network;

[0006] Decomposing the operating state signal energy of the operating state signal corresponding to the historical operating state information by wavelet packet decomposition to obtain sub-band signal energies of sub-band signals corresponding to the operating state signal at different frequencies;

[0007] Training a neural network according to the energy of each sub-band signal to obtain a distribution network operation status prediction model;

[0008] The distribution network operation status prediction model is used to monitor the real-time operation status information of the distribution network and predict the future operation status information of the distribution network in real time, and the future operation status information is adjusted based on the real-time operation status information.

[0009] Optionally, decomposing the operation status signal energy of the operation status signal corresponding to the historical operation status information by wavelet packet decomposition to obtain sub-band signal energies of sub-band signals corresponding to the operation status signal at different frequencies includes:

[0010] Determining the operating state signal energy based on the historical operating state information and a signal energy determination formula;

[0011] Decomposing the operation status signal energy of the operation status signal corresponding to the historical operation status information by wavelet packet decomposition, and determining the energy of each sub-band signal according to the sample length of each sub-band signal, the discrete point amplitude of each sub-band signal, and a sub-band signal energy determination formula;

[0012] The signal energy determination formula is: ; represents the energy of the operating status signal, represents the historical operation status information, and R represents a real number set;

[0013] The formula for determining the sub-band signal energy is: ; Represents the first Tier The sub-band signal energy of the band, Indicates the Tier The sub-band signal of the band, represents the sample length of each sub-band signal, m represents the summation variable, represents the decomposition level of each of the sub-band signals, Represents the discrete point amplitude of each sub-band signal.

[0014] Optionally, the training of the neural network according to the energy of each sub-band signal includes:

[0015] Determining a eigenvector corresponding to each of the sub-band signals based on the energy of each of the sub-band signals;

[0016] The neural network is trained using the feature vectors.

[0017] Optionally, the using the distribution network operation status prediction model to monitor the real-time operation status information of the distribution network in real time and predict future operation status information of the distribution network includes:

[0018] Determining a feature sharing layer and a first linear transformation layer in the distribution network operation state prediction model;

[0019] Processing the energy of each sub-band signal through the feature sharing layer to obtain a target high-level feature corresponding to the energy of each sub-band signal, wherein the target high-level feature is a feature that meets a preset high-level feature condition;

[0020] Inputting each of the target high-level features into a first linear transformation layer, and processing each of the target high-level features using the first linear transformation layer to obtain real-time operating status information of the distribution network corresponding to each of the target high-level features;

[0021] The future operating status information of the distribution network is predicted based on each of the target high-level features.

[0022] Optionally, processing the energy of each sub-band signal by the feature sharing layer to obtain a target high-level feature corresponding to the energy of each sub-band signal includes:

[0023] Determine the target high-level feature corresponding to each of the sub-band signal energies according to the sub-band signal energy, the weight of the feature sharing layer, the bias of the feature sharing layer, the ReLu activation function, and the target high-level feature determination formula;

[0024] Accordingly, the processing of each of the target high-level features using the first linear transformation layer to obtain the real-time operating status information of the distribution network corresponding to each of the target high-level features includes:

[0025] According to the weight of the first linear transformation layer, each of the target high-level features, the bias of the first linear transformation layer, Function and real-time operation status information determination formula to determine the real-time operation status information of the distribution network corresponding to each target advanced feature;

[0026] The target high-level feature determination formula is: ; Indicates the target high-level features corresponding to the energy of each sub-band signal, represents the bias of the feature sharing layer, represents the signal energy of each sub-band, represents the weight of the feature sharing layer, Represents the ReLu activation function;

[0027] The formula for determining the real-time operating status information is: ; represents the real-time operating status information of the distribution network corresponding to each target advanced feature, represents the weight of the first linear transformation layer, Represents the high-level features of each target, represents the bias of the first linear transformation layer, Indicates the function.

[0028] Optionally, predicting the future operating state information of the distribution network according to each of the target advanced features includes:

[0029] Determine the association layer, GRU time series feature extraction layer and second linear transformation layer in the distribution network operation status prediction model;

[0030] Inputting each of the target high-level features into the association layer, and outputting the feedback output amount and feedback delay amount corresponding to each of the target high-level features through the association layer;

[0031] Inputting each of the sub-band signal energies, each of the feedback output amounts, and each of the feedback delay amounts into the GRU temporal feature extraction layer, and outputting prediction features corresponding to each of the sub-band signal energies through the GRU temporal feature extraction layer;

[0032] Each of the prediction features is input into the second linear transformation layer, and the future operation state information of the power distribution network is output through the second linear transformation layer.

[0033] Optionally, inputting the energy of each sub-band signal, each feedback output amount, and each feedback delay amount into the GRU temporal feature extraction layer, and outputting the prediction feature corresponding to each sub-band signal energy through the GRU temporal feature extraction layer includes:

[0034] Determine the output value of the first layer of the GRU temporal feature extraction layer according to the ReLu activation function, the weight of the first layer of the GRU temporal feature extraction layer, each of the target high-level features, each of the feedback output amounts, each of the feedback delay amounts, and the output value determination formula of the first layer of the GRU temporal feature extraction layer;

[0035] Determine the output value of each layer in the GRU temporal feature extraction layer except the first layer according to the ReLu activation function, the weight of the Lth layer of the GRU temporal feature extraction layer, the output value of the L-1th layer of the GRU temporal feature extraction layer, the bias of the L-1th layer of the GRU temporal feature extraction layer, and the output value determination formula of each layer in the GRU temporal feature extraction layer except the first layer, wherein the Lth layer of the GRU temporal feature extraction layer is any layer other than the first layer in the GRU temporal feature extraction layer;

[0036] The output value of the first layer of the GRU temporal feature extraction layer is determined by the following formula: ; Represents the output value of the first layer of the GRU temporal feature extraction layer at time t, represents the ReLu activation function, represents the weight of the first layer of the GRU temporal feature extraction layer, represents the target high-level features input into the first layer of the GRU temporal feature extraction layer at time t, represents the feedback input amount output by the association layer at time t, Indicates each feedback delay amount output by the association layer;

[0037] The formula for determining the output value of each layer except the first layer in the GRU temporal feature extraction layer is: ; represents the output value of the Lth layer except the first layer in the GRU temporal feature extraction layer at time t, represents the ReLu activation function, represents the weight of the Lth layer except the first layer in the GRU temporal feature extraction layer, represents the output value of the L-1th layer except the first layer in the GRU temporal feature extraction layer at time t, Represents the bias of the L-1th layer except the first layer in the GRU temporal feature extraction layer.

[0038] Optionally, after adjusting the future operating status information based on the real-time operating status information, the method further includes:

[0039] Classifying the real-time operating status information and the future operating status information according to the degree of abnormality, and determining a confusion matrix of the operating status of the distribution network according to the classification results;

[0040] The accuracy of the distribution network operation status prediction model is determined according to the distribution status of each classification result in the confusion matrix.

[0041] To solve the above technical problems, the present invention further provides an electronic device, comprising:

[0042] memory for storing computer programs;

[0043] A processor is used to implement the steps of the distribution network operation status analysis method as described above when executing the computer program.

[0044] To solve the above technical problems, the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the distribution network operation status analysis method as described above are implemented.

[0045] The purpose of the present invention is to provide a distribution network operating status analysis method, electronic device and medium. This scheme takes into account that predicting the operating status of the distribution network through the original data of the distribution network will lead to inaccurate prediction results due to irregular changes in the original data. Therefore, this scheme decomposes the historical operating status information of the unstable distribution network through wavelet packet decomposition to perform multi-level frequency analysis on the current operating status signal of the distribution network, so that the distribution network operating status prediction model obtained after neural network training can more accurately predict the future state of the distribution network. In addition, because the distribution network operating status prediction model has the function of real-time monitoring of the real-time operating status information of the distribution network, it can also adjust the future operating status information based on the real-time operating status information to improve the accuracy of the future operating status information, thereby improving the accuracy of the distribution network operating status analysis process. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0047] Figure 1 A process flow chart of a distribution network operating status analysis method provided by the present invention;

[0048] Figure 2 A schematic diagram of decomposing signal energy using a three-layer wavelet decomposition method provided by the present invention;

[0049] Figure 3 A schematic diagram of a confusion matrix of the operating status of a distribution network provided by the present invention;

[0050] Figure 4 A schematic diagram of the deployment of a distribution network hierarchical model based on a central platform architecture provided by the present invention;

[0051] Figure 5 This is a structural diagram of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0052] The core of the present invention is to provide a distribution network operating status analysis method, electronic device and medium. This scheme takes into account that predicting the operating status of the distribution network through the original data of the distribution network will lead to inaccurate prediction results due to irregular changes in the original data. Therefore, this scheme decomposes the historical operating status information of the unstable distribution network through wavelet packet decomposition to perform multi-level frequency division (both low-frequency signals and high-frequency signals) on the current operating status signal of the distribution network, so that the distribution network operating status prediction model obtained after neural network training can more accurately predict the future state of the distribution network. In addition, because the distribution network operating status prediction model is obtained based on the historical operating status information of the distribution network, it also has the function of real-time monitoring of the real-time operating status information of the distribution network, and can adjust the future operating status information based on the real-time operating status information to improve the accuracy of the future operating status information, thereby improving the accuracy of the distribution network operating status analysis process.

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0054] Please refer to Figure 1 , Figure 1 This is a process flow chart of a distribution network operating status analysis method provided by the present invention. The distribution network operating status analysis method includes:

[0055] S11: Obtain historical operating status information of the distribution network;

[0056] S12: decomposing the operating state signal energy of the operating state signal corresponding to the historical operating state information by wavelet packet decomposition to obtain sub-band signal energies of sub-band signals corresponding to the operating state signal at different frequencies;

[0057] S13: training a neural network according to the signal energy of each sub-band to obtain a distribution network operation status prediction model;

[0058] S14: using the distribution network operation status prediction model to monitor the real-time operation status information of the distribution network and predict the future operation status information of the distribution network in real time, and adjusting the future operation status information based on the real-time operation status information.

[0059] In the present invention, it is considered that the prediction of the operating status of the distribution network through the original data of the distribution network may lead to inaccurate prediction results due to the irregular changes of the original data. In addition, the original operating signals of the distribution network (such as voltage and current waveforms) usually contain a lot of noise and redundant information, and the frequency components are complex. If the original operating signals of the distribution network are directly input into the neural network for model training, the computational burden of the neural network will be greatly increased due to excessive redundant information and noise information. In addition, the noise information may also cover up the key features of the distribution network, thereby affecting the robustness of the neural network. Therefore, this scheme decomposes the operating status signal energy corresponding to the historical operating status information of the distribution network by using wavelet packet decomposition to decompose the current operating status signal of the distribution network. Multi-level frequency segmentation (decomposing both low-frequency and high-frequency signals) is performed to decompose the operating status signal into sub-band signals at different frequencies and obtain the signal energy of each sub-band. This approach facilitates the neural network to focus on key frequency bands. Specifically, training is performed based on the energy of sub-band signals in certain frequency bands. The energy changes in these sub-band signals are strongly correlated with specific distribution network signals (such as faults, harmonics, and sudden load changes). This enables the distribution network operating status prediction model obtained after neural network training to more stably and accurately predict the future state of the distribution network. Furthermore, this approach can compress the distribution network operating status signal into the energy characteristics of a few sub-bands (for example, extracting 5-10 frequency bands), significantly reducing the amount of data input to the neural network. Furthermore, because the distribution network operating status prediction model is derived from historical operating status information of the distribution network, it also has the function of real-time monitoring of the distribution network's operating status information. This solution also adjusts future operating status information based on real-time operating status information, improving the accuracy of future operating status information and, in turn, the accuracy of the distribution network operating status analysis process.

[0060] It should be noted that this solution uses wavelet packet decomposition to decompose the distribution network's operating status signal into low-frequency and high-frequency parts. At the same time, a dual-branch deep learning network (neural network training) is designed to realize the monitoring and prediction of the distribution network's operating status. The adjustment factor of the real-time operating status can be added during the prediction, which can make the prediction of the distribution network's operating status more accurate.

[0061] It should also be noted that abnormal conditions in distribution networks (such as short circuits, arcing, and equipment aging) often manifest as energy variations in specific frequency bands. For example, low-frequency bands (e.g., 50 / 60 Hz) can reflect steady-state operating conditions (such as voltage fluctuations and load changes); high-frequency bands (e.g., hundreds of Hz to kHz) may correspond to transient events such as switching operations and arcing; and harmonic frequency bands (e.g., third and fifth harmonics) can indicate nonlinear loads or equipment failures. Therefore, this solution analyzes sub-band signal energy through wavelet packet decomposition. This enables the distribution network operating status prediction model trained using a neural network to more intuitively capture these abnormal conditions without blindly processing raw operating status data. This significantly improves the accuracy of the operating status prediction process and reduces the computational effort required.

[0062] It should also be noted that the energy characteristics of the sub-band signal are insensitive to small fluctuations in the signal amplitude and are more concerned with the overall energy distribution. Therefore, they are suitable for the time-varying characteristics of the distribution network operation status signal. The characteristic vector of the sub-band signal energy is essentially a knowledge distillation of the distribution network operation status signal. Through frequency domain analysis, physical features that are strongly correlated with the distribution network status are extracted. This method greatly improves the performance and efficiency of the distribution network operation status prediction model.

[0063] It should also be noted that if the sub-band energy extraction is skipped and the distribution network operation status signal is directly input into the neural network, it may result in: 1. The distribution network operation status prediction model trained by the neural network needs to learn the frequency domain characteristics by itself, which increases the training difficulty and computational complexity; 2. It is sensitive to noise signals and the prediction accuracy is reduced; 3. The computational cost increases dramatically. For example, when processing the same operation status signal, the actual amount of computing required varies greatly.

[0064] It should also be noted that the historical operating status information of the distribution network refers to various data collected and monitored in the distribution network. These data are crucial to ensuring the safe, reliable and efficient operation of the distribution network. For example, the operating status information of the distribution network obtained may include: 1. Voltage information: voltage value, voltage fluctuation, and voltage over-limit of each node; 2. Current information: current value, current fluctuation, and current over-limit of each line; 3. Power information: Active power (Active Power): actual power consumption in the system, Reactive power (Reactive Power): power generated by inductance and capacitance components in the system, Apparent power (Apparent Power): power generated by inductance and capacitance components in the system Power: The total power of the system, including active power and reactive power; 4. Frequency information: The frequency value of the power grid, frequency fluctuations, and frequency over-limit conditions; 5. Load information: The power load of each user, load curve, peak load, and valley load; 6. Switch status information: The status (open or closed) and operation records of equipment such as circuit breakers and disconnectors; 7. Fault information: Fault type (short circuit, grounding, line break, etc.), fault location, fault time, and fault duration; 8. Protection action information: The action record of the relay protection device, protection setting, and protection action time; 9. Environmental information: Temperature, humidity, and weather conditions (such as thunderstorms and strong winds); 10. Equipment status information: Transformer status (temperature, oil level, gas pressure, etc.), switchgear status (number of operations, lifespan, etc.), and cable and line status (aging, insulation condition, etc.); 11. Communication status information: The communication status of the SCADA system (Supervisory Control And Data Acquisition) and the RTU (Remote Terminal Unit) 12. Power quality information: voltage harmonics, current harmonics, voltage flicker, and voltage imbalance; 13. Distributed energy information: output power of distributed generation (such as solar and wind power), charge and discharge status of distributed energy storage systems, and other information.

[0065] This embodiment provides a method for analyzing the operating status of a distribution network. This scheme takes into account that predicting the operating status of a distribution network through the original data of the distribution network may lead to inaccurate prediction results due to irregular changes in the original data. Therefore, this scheme decomposes the historical operating status information of an unstable distribution network through wavelet packet decomposition to perform multi-level frequency analysis on the current operating status signal of the distribution network, so that the distribution network operating status prediction model obtained after neural network training can more accurately predict the future state of the distribution network. In addition, because the distribution network operating status prediction model has the function of real-time monitoring of the real-time operating status information of the distribution network, it can also adjust the future operating status information based on the real-time operating status information to improve the accuracy of the future operating status information, thereby improving the accuracy of the distribution network operating status analysis process.

[0066] Based on the above embodiment:

[0067] As an optional embodiment, decomposing the operating status signal energy of the operating status signal corresponding to the historical operating status information by wavelet packet decomposition to obtain sub-band signal energies of sub-band signals corresponding to the operating status signal at different frequencies includes:

[0068] Determine the operating status signal energy based on historical operating status information and a signal energy determination formula;

[0069] Decomposing the operation status signal energy of the operation status signal corresponding to the historical operation status information by wavelet packet decomposition, and determining the energy of each sub-band signal according to the sample length of each sub-band signal, the discrete point amplitude of each sub-band signal and the sub-band signal energy determination formula;

[0070] The signal energy determination formula is: ; Indicates the energy of the operating status signal, Represents historical operating status information, and R represents a set of real numbers;

[0071] The formula for determining the sub-band signal energy is: ; Represents the first Tier The sub-band signal energy of the band, Indicates the Tier The sub-band signal of the band, represents the sample length of each sub-band signal, m represents the summation variable, Indicates the decomposition level of each sub-band signal, Indicates the discrete point amplitude of each sub-band signal.

[0072] In the present invention, because the sub-band signal energy of the sub-band signal is actually obtained by wavelet packet decomposition, and the wavelet packet decomposition process actually needs to first determine the energy of the operating status signal, and then decompose the energy of the operating status signal to obtain the sub-band signal energy of the sub-band signal, wherein the wavelet packet decomposition process needs to substitute the sample length of each sub-band signal and the discrete point amplitude of each sub-band signal into the sub-band signal energy determination formula to determine the energy of each sub-band signal, thereby ensuring the integrity of the process of determining the energy of each sub-band signal.

[0073] It should be noted that the decomposition diagram of the signal energy decomposition using the three-layer wavelet packet decomposition method is as follows Figure 2 As shown in the figure, the various steps of the wavelet packet band decomposition of a signal are shown from left to right. The details are as follows: Input signal (x): The original input signal energy. First-level decomposition: H1: High-pass filter output, L1: Low-pass filter output. Second-level decomposition: The output of H1 is further decomposed into: HH2: High-pass filter output, HL2: Low-pass filter output. The output of L1 is further decomposed into: LH2: High-pass filter output, LL2: Low-pass filter output. Third-level decomposition: The output of HH2 is further decomposed into: HHH3: High-pass filter output, HHL3: Low-pass filter output. The output of HL2 is further decomposed into: HLH3: High-pass filter output, HLL3: Low-pass filter output. The output of LH2 is further decomposed into: LHH3: High-pass filter output, LHL3: Low-pass filter output. The output of LL2 is further decomposed into: LLH3: High-pass filter output, LLL3: Low-pass filter output.

[0074] It should also be noted that the neural network in this application can be implemented by various neural network models or modules such as Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Autoencoders, Transformer Networks, Multilayer Perceptrons (MLP), Graph Neural Networks (GNNs), etc. There is no specific restriction on the method of determining the eigenvector corresponding to the signal of each sub-band according to the energy of the signal of each sub-band. For example: After the distribution network operation data is pre-processed by data cleaning and then decomposed into three layers of wavelet packets, the signal of each sub-band is obtained as follows: , and the signal energy of each sub-band can be calculated as , so the energy structure feature vector of each frequency band can be obtained as .

[0075] As an optional embodiment, training the neural network according to the energy of each sub-band signal includes:

[0076] Determine the eigenvector corresponding to each sub-band signal based on the energy of each sub-band signal;

[0077] The neural network is trained through each feature vector.

[0078] In the present invention, considering that abnormal conditions in the distribution network (such as short circuit, electric arc, and equipment aging) often manifest as energy changes in a specific frequency band, this scheme specifically analyzes the sub-band signal energy through wavelet packet decomposition. Moreover, considering that the energy characteristics of the sub-band signal are insensitive to small fluctuations in the signal amplitude and pay more attention to the overall energy distribution, it is suitable for the time-varying characteristics of the distribution network operation status signal. Therefore, this scheme will first determine the eigenvector corresponding to each sub-band signal based on the energy of each sub-band signal, and then train the neural network through each eigenvector, so that the distribution network operation status prediction model trained by the neural network can more intuitively capture the abnormal state of the distribution network, thereby greatly improving the accuracy of the operation status prediction process.

[0079] As an optional embodiment, the distribution network operation state prediction model is used to monitor the real-time operation state information of the distribution network and predict the future operation state information of the distribution network in real time, including:

[0080] Determining a feature sharing layer and a first linear transformation layer in a distribution network operation state prediction model;

[0081] The signal energy of each sub-band is processed through the feature sharing layer to obtain the target high-level features corresponding to the signal energy of each sub-band, where the target high-level features are features that meet the preset high-level feature conditions;

[0082] Inputting each target high-level feature into the first linear transformation layer, and using the first linear transformation layer to process each target high-level feature to obtain real-time operating status information of the distribution network corresponding to each target high-level feature;

[0083] The future operating status information of the distribution network is predicted based on the high-level features of each target.

[0084] In the present invention, the distribution network operation status prediction model includes a feature sharing layer and a first linear transformation layer, wherein the feature sharing layer is used to process the energy of each sub-band signal to obtain the target high-level features corresponding to the energy of each sub-band signal, and the first linear transformation layer is used to process each target high-level feature to obtain the real-time operation status information of the distribution network corresponding to each target high-level feature, that is, real-time monitoring of the real-time operation status information of the distribution network, and can also predict the future operation status information of the distribution network based on each target high-level feature, that is, predict the future operation status information of the distribution network, thereby ensuring the integrity of the solution.

[0085] As an optional embodiment, the signal energy of each sub-band is processed by a feature sharing layer to obtain target high-level features corresponding to the signal energy of each sub-band, including:

[0086] Determine the target high-level features corresponding to the signal energy of each sub-band according to the signal energy of each sub-band, the weight of the feature sharing layer, the bias of the feature sharing layer, the ReLu activation function and the target high-level feature determination formula;

[0087] Accordingly, the first linear transformation layer is used to process each target high-level feature to obtain the real-time operating status information of the distribution network corresponding to each target high-level feature, including:

[0088] According to the weight of the first linear transformation layer, the high-level features of each target, the bias of the first linear transformation layer, Function and real-time operation status information determination formula determine the real-time operation status information of the distribution network corresponding to each target advanced feature;

[0089] Among them, the target high-level feature determination formula is: ; Indicates the target high-level features corresponding to the signal energy of each sub-band, represents the bias of the feature sharing layer, represents the signal energy of each sub-band, represents the weight of the feature sharing layer, Represents the ReLu activation function;

[0090] The formula for determining real-time operating status information is: ; Indicates the real-time operating status information of the distribution network corresponding to each target advanced feature, represents the weight of the first linear transformation layer, Represents the high-level features of each target, represents the bias of the first linear transformation layer, express function.

[0091] In the present invention, the specific process of processing the energy of each sub-band signal through the feature sharing layer to obtain the target high-level features corresponding to the energy of each sub-band signal is as follows: the energy of each sub-band signal, the weight of the feature sharing layer, the bias of the feature sharing layer, and the ReLu activation function are substituted into the target high-level feature determination formula to accurately determine the target high-level features corresponding to the energy of each sub-band signal. The specific process of processing each target high-level feature using the first linear transformation layer to obtain the real-time operation status information of the distribution network corresponding to each target high-level feature is as follows: the weight of the first linear transformation layer, each target high-level feature, the bias of the first linear transformation layer, By substituting the function into the real-time operation status information determination formula, the real-time operation status information of the distribution network corresponding to each target advanced feature can be accurately determined, ensuring the integrity of the solution.

[0092] As an optional embodiment, predicting the future operating state information of the distribution network based on various target advanced features includes:

[0093] Determine the association layer, GRU time series feature extraction layer and second linear transformation layer in the distribution network operation status prediction model;

[0094] Input each target high-level feature into the association layer, and output the feedback output and feedback delay corresponding to each target high-level feature through the association layer;

[0095] Input the signal energy of each sub-band, each feedback output and each feedback delay into the GRU temporal feature extraction layer, and output the prediction features corresponding to the signal energy of each sub-band through the GRU temporal feature extraction layer;

[0096] Each prediction feature is input into the second linear transformation layer, and the future operation status information of the distribution network is output through the second linear transformation layer.

[0097] In the present invention, the distribution network operation status prediction model also includes an association layer, a GRU time series feature extraction layer and a second linear transformation layer, wherein the association layer is used to output the feedback output and feedback delay corresponding to each target high-level feature according to the input target high-level features, the GRU time series feature extraction layer is used to output the prediction features corresponding to each sub-band signal energy according to the input sub-band signal energy, each feedback output and each feedback delay, and the second linear transformation layer is used to output the future operation status information of the distribution network according to each input prediction feature, thereby ensuring the integrity of the solution.

[0098] It should be noted that the input dimension of the feature vector can be the same as or related to the acquired data dimension, and there is no specific limitation here. For example, the feature vector can be , the real-time running status can be , then the mapping function can be .

[0099] As an optional embodiment, the signal energy of each sub-band, each feedback output amount, and each feedback delay amount are input into the GRU temporal feature extraction layer, and the prediction features corresponding to the signal energy of each sub-band are output by the GRU temporal feature extraction layer, including:

[0100] Determine the output value of the first layer of the GRU temporal feature extraction layer according to the ReLu activation function, the weight of the first layer of the GRU temporal feature extraction layer, each target high-level feature, each feedback output amount, each feedback delay amount, and the output value determination formula of the first layer of the GRU temporal feature extraction layer;

[0101] Determine the output value of each layer except the first layer in the GRU temporal feature extraction layer according to the ReLu activation function, the weight of the Lth layer of the GRU temporal feature extraction layer, the output value of the L-1th layer of the GRU temporal feature extraction layer, the bias of the L-1th layer of the GRU temporal feature extraction layer, and the output value determination formula of each layer except the first layer in the GRU temporal feature extraction layer, wherein the Lth layer of the GRU temporal feature extraction layer is any layer other than the first layer in the GRU temporal feature extraction layer;

[0102] Among them, the output value determination formula of the first layer of the GRU temporal feature extraction layer is: ; Represents the output value of the first layer of the GRU temporal feature extraction layer at time t, represents the ReLu activation function, Represents the weight of the first layer of the GRU temporal feature extraction layer, Represents the target high-level features of the first layer input into the GRU temporal feature extraction layer at time t, represents the feedback input output by the associated layer at time t, Indicates the delay of each feedback output of the correlation layer;

[0103] The formula for determining the output value of each layer except the first layer in the GRU time series feature extraction layer is: ; Represents the output value of the Lth layer except the first layer in the GRU temporal feature extraction layer at time t, represents the ReLu activation function, Represents the weight of the Lth layer except the first layer in the GRU temporal feature extraction layer, Represents the output value of the L-1th layer except the first layer in the GRU temporal feature extraction layer at time t, Represents the bias of the L-1th layer except the first layer in the GRU temporal feature extraction layer.

[0104] In the present invention, considering that the number of layers of the GRU (Gated Recurrent Unit) temporal feature extraction layer is L, different determination processes are required for the weights and biases of different layers. For example, when determining the output value of the first layer of the GRU temporal feature extraction layer, the ReLu activation function, the weight of the first layer of the GRU temporal feature extraction layer, each target high-level feature, each feedback output amount, and each feedback delay amount need to be substituted into the output value determination formula of the first layer of the GRU temporal feature extraction layer; similarly, when determining the output values ​​of each layer except the first layer in the GRU temporal feature extraction layer, the ReLu activation function, the weight of the Lth layer of the GRU temporal feature extraction layer, the output value of the L-1th layer of the GRU temporal feature extraction layer, and the bias of the L-1th layer of the GRU temporal feature extraction layer need to be substituted into the output value determination formula of each layer except the first layer in the GRU temporal feature extraction layer, thereby ensuring the integrity of the solution.

[0105] As an optional embodiment, after adjusting the future operating status information based on the real-time operating status information, the method further includes:

[0106] Classify the real-time operation status information and future operation status information according to the degree of abnormality, and determine the confusion matrix of the operation status of the distribution network based on the classification results;

[0107] The accuracy of the distribution network operation status prediction model is determined according to the distribution status of each classification result in the confusion matrix.

[0108] In this invention, considering the importance of the accuracy of the distribution network operating status prediction model for actual distribution network operating status prediction, this solution uses a confusion matrix to determine the accuracy of the distribution network operating status prediction model. The specific process is as follows: first, real-time operating status information and future operating status information are classified according to the degree of abnormality, and a confusion matrix of the distribution network operating status is determined based on the classification results. The classification results can be divided into normal, slightly abnormal (does not affect system operation but reduces operational reliability), moderately abnormal (affects system operation but requires low maintenance time), and severely abnormal (affects system operation and is urgent). The distribution of each classification result in the confusion matrix can then be used to accurately determine the accuracy of the distribution network operating status prediction model, making it easier for users to observe the accuracy of the distribution network operating status prediction model in real time.

[0109] It should be noted that if Figure 3 As shown, Figure 3 This is a diagram of the confusion matrix of the running state. Figure 3 In the example, the classification results of 820 samples can be presented in the form of a confusion matrix, where 0, 1, 2, and 3 represent normal, slightly abnormal, moderately abnormal, and severely abnormal, respectively. Figure 3As can be seen, the distribution network operating status type is essentially consistent with the actual type. Due to the large number of normal samples, the classification accuracy is the highest, reaching 99.62%. At the same time, due to the similarities in the characteristics of mild and moderate anomalies, there are more misclassifications. Severe anomalies have distinct characteristics, resulting in fewer misclassifications and higher classification accuracy.

[0110] It should also be noted that the above-mentioned distribution network operation status analysis method based on the middle-platform architecture and deep learning can be deployed in the data computing and analysis layer of the distribution network hierarchical model based on the middle-platform architecture. The distribution network hierarchical model based on the middle-platform architecture adopts a layered design, which consists of a data access layer, a computing and analysis layer, a data storage layer, and a real-time service layer. The data access layer mainly uses the kafka (distributed stream processing platform) cluster as the access carrier to meet the real-time access needs of various types of information. Data computing and analysis use clustered Flink (real-time computing) and Spark (batch computing) components. The data storage adopts the HBase+redis method, among which the HBase library (measurement library) meets the storage needs of massive measurement data, and Redis (cache library) is deployed as a multi-master and multi-slave cluster to achieve rapid cache utilization. The real-time service is deployed on the k8s cluster in docker. In the above deployment method, the message queue, analysis and computing components, and the k8s cluster deployed by the real-time service can reuse the existing software and hardware resources of the power grid resource business middle platform. The business library and collection library required for data storage require new servers. The deployment framework diagram can be shown as follows Figure 4 As shown, Figure 4 This is a deployment diagram of the distribution network layered model based on the middle-office architecture. Data access is mainly divided into model access service and measurement access service. The model access service enables access to the grid resource ledger data and the archival data of the procurement / marketing system from the grid resource business middle office and the grid resource business middle office, respectively. This includes the ledgers, archives, measurement points, topology and other types of data corresponding to the main grid, distribution network medium voltage, distribution network low voltage, and electricity consumption. Build a grid relationship model for the measurement center and synchronously update the model data to the cache library to meet the grid relationship model mapping and splicing requirements when accessing measurement data; in response to the real-time requirements of measurement data access, distribution network data access uniformly adopts the form of message queues to receive telemetry and telesignaling data pushed by external systems. The main network data access is currently parsed in the form of E files, and after the parsing is completed, it is pushed to the message queue of the enterprise-level real-time measurement center.

[0111] The computing process uses the stream processing framework Flink. After subscribing to measurement data from the message queue, verification, model mapping, and business analysis are performed. During the processing process, the results are stored in the collection library and business library in real time at different processing stages as needed. At the same time, the results are pushed to the message bus of the enterprise-level real-time measurement center to provide real-time subscription capabilities. The architecture also uses Spark to provide batch processing capabilities to meet the needs of scheduled or periodic business analysis.

[0112] Data storage: Based on data categories and business needs, data storage is planned as model library, collection library, cache library, and business library. The most appropriate storage mechanism is adopted for each of the data storage and usage characteristics.

[0113] Shared services use standard RESTful to provide external query services for models, measurements, and business analysis results. Combined with the subscription capabilities of the enterprise-level real-time measurement center message bus, they meet the various data acquisition needs of upper-level applications.

[0114] Please refer to Figure 5 , Figure 5 This is a schematic diagram of the structure of an electronic device provided by the present invention. The electronic device includes:

[0115] Memory 20, for storing computer programs;

[0116] The processor 21 is configured to implement the steps of the above-mentioned distribution network operation status analysis method when executing a computer program.

[0117] The electronic device provided in this embodiment may include but is not limited to a smart phone, a tablet computer, a laptop computer, or a desktop computer.

[0118] Among them, the processor 21 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 21 can be implemented in at least one hardware form of a digital signal processor (DSP), a field programmable gate array (FPGA), and a programmable logic array (PLA). The processor 21 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a central processing unit (CPU); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 21 may be integrated with a graphics processing unit (GPU), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 21 may also include an artificial intelligence (AI) processor, which is used to process computing operations related to machine learning.

[0119] The memory 20 may include one or more computer-readable storage media, which may be non-transitory. The memory 20 may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In this embodiment, the memory 20 is at least used to store the following computer program 201, wherein, after the computer program is loaded and executed by the processor 21, it can implement the relevant steps of the distribution network operation status analysis method disclosed in any of the aforementioned embodiments. In addition, the resources stored in the memory 20 may also include an operating system 202 and data 203, etc., and the storage method may be temporary storage or permanent storage. Among them, the operating system 202 may include Windows, Unix, Linux, etc. The data 203 may include but is not limited to the distribution network operation status analysis method, etc.

[0120] In some embodiments, the electronic device may further include a display screen 22 , an input / output interface 23 , a communication interface 24 , a power supply 25 , and a communication bus 26 .

[0121] Those skilled in the art will understand that Figure 5 The structure shown in the figure does not constitute a limitation of the electronic device, and may include more or fewer components than shown in the figure.

[0122] The purpose of this embodiment is to provide an electronic device, in which the memory 20 is used to store computer programs, and the processor 21 is used to implement the steps of the above-mentioned distribution network operation status analysis method when executing the computer program, so as to make the optimization process more efficient and accurate.

[0123] The present invention also provides an embodiment corresponding to a computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the distribution network operation status analysis method as described above are implemented.

[0124] It is understood that if the methods in the above embodiments are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and executes all or part of the steps of the methods in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0125] The computer-readable storage medium provided in this embodiment corresponds to the above method and therefore has the same beneficial effects as the above method. Therefore, for the embodiments of the computer-readable storage medium part, please refer to the description of the embodiments of the method part, which will not be repeated here.

[0126] It should be noted that, in this specification, relational terms such as first and second, etc. are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "comprising" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.

[0127] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for analyzing the operating status of a distribution network, characterized in that: include: Obtain historical operating status information of the distribution network; Decomposing the operating state signal energy of the operating state signal corresponding to the historical operating state information by wavelet packet decomposition to obtain sub-band signal energies of sub-band signals corresponding to the operating state signal at different frequencies; Training a neural network according to the energy of each sub-band signal to obtain a distribution network operation status prediction model; The distribution network operation status prediction model is used to monitor the real-time operation status information of the distribution network and predict the future operation status information of the distribution network in real time, and the future operation status information is adjusted based on the real-time operation status information.

2. The method for analyzing the operating status of a distribution network according to claim 1, wherein: Decomposing the operating state signal energy of the operating state signal corresponding to the historical operating state information by wavelet packet decomposition to obtain sub-band signal energies of sub-band signals corresponding to the operating state signal at different frequencies includes: Determining the operating state signal energy based on the historical operating state information and a signal energy determination formula; Decomposing the operation status signal energy of the operation status signal corresponding to the historical operation status information by wavelet packet decomposition, and determining the energy of each sub-band signal according to the sample length of each sub-band signal, the discrete point amplitude of each sub-band signal, and a sub-band signal energy determination formula; The signal energy determination formula is: ; represents the energy of the operating status signal, represents the historical operation status information, and R represents a real number set; The formula for determining the sub-band signal energy is: ; Represents the first Tier The sub-band signal energy of the band, Indicates the Tier The sub-band signal of the band, represents the sample length of each sub-band signal, m represents the summation variable, represents the decomposition level of each of the sub-band signals, Represents the discrete point amplitude of each sub-band signal.

3. The method for analyzing the operating status of a distribution network according to claim 1, wherein: The training of the neural network according to the energy of each sub-band signal includes: Determining a eigenvector corresponding to each of the sub-band signals based on the energy of each of the sub-band signals; The neural network is trained using the feature vectors.

4. The method for analyzing the operating status of a distribution network according to claim 1, wherein: The method of using the distribution network operation state prediction model to monitor the real-time operation state information of the distribution network and predict the future operation state information of the distribution network in real time includes: Determining a feature sharing layer and a first linear transformation layer in the distribution network operation state prediction model; Processing the energy of each sub-band signal through the feature sharing layer to obtain a target high-level feature corresponding to the energy of each sub-band signal, wherein the target high-level feature is a feature that meets a preset high-level feature condition; Inputting each of the target high-level features into a first linear transformation layer, and processing each of the target high-level features using the first linear transformation layer to obtain real-time operating status information of the distribution network corresponding to each of the target high-level features; The future operating status information of the distribution network is predicted based on each of the target high-level features.

5. The method for analyzing the operating status of a distribution network according to claim 4, wherein: The processing of each sub-band signal energy by the feature sharing layer to obtain a target high-level feature corresponding to each sub-band signal energy includes: Determine the target high-level feature corresponding to each of the sub-band signal energies according to the sub-band signal energy, the weight of the feature sharing layer, the bias of the feature sharing layer, the ReLu activation function, and the target high-level feature determination formula; Accordingly, the processing of each of the target high-level features using the first linear transformation layer to obtain the real-time operating status information of the distribution network corresponding to each of the target high-level features includes: According to the weight of the first linear transformation layer, each of the target high-level features, the bias of the first linear transformation layer, Function and real-time operation status information determination formula to determine the real-time operation status information of the distribution network corresponding to each target advanced feature; The target high-level feature determination formula is: ; Indicates the target high-level features corresponding to the energy of each sub-band signal, represents the bias of the feature sharing layer, represents the signal energy of each sub-band, represents the weight of the feature sharing layer, Represents the ReLu activation function; The formula for determining the real-time operating status information is: ; represents the real-time operating status information of the distribution network corresponding to each target advanced feature, represents the weight of the first linear transformation layer, Represents the high-level features of each target, represents the bias of the first linear transformation layer, Indicates the function.

6. The method for analyzing the operating status of a distribution network according to claim 4, wherein: The predicting of the future operating state information of the distribution network according to each of the target high-level features includes: Determine the association layer, GRU time series feature extraction layer and second linear transformation layer in the distribution network operation status prediction model; Inputting each of the target high-level features into the association layer, and outputting the feedback output amount and feedback delay amount corresponding to each of the target high-level features through the association layer; Inputting each of the sub-band signal energies, each of the feedback output amounts, and each of the feedback delay amounts into the GRU temporal feature extraction layer, and outputting prediction features corresponding to each of the sub-band signal energies through the GRU temporal feature extraction layer; Each of the prediction features is input into the second linear transformation layer, and the future operation state information of the power distribution network is output through the second linear transformation layer.

7. The method for analyzing the operating status of a distribution network according to claim 6, wherein: Inputting the energy of each sub-band signal, each feedback output amount, and each feedback delay amount into the GRU temporal feature extraction layer, and outputting the prediction feature corresponding to each sub-band signal energy through the GRU temporal feature extraction layer, includes: Determine the output value of the first layer of the GRU temporal feature extraction layer according to the ReLu activation function, the weight of the first layer of the GRU temporal feature extraction layer, each of the target high-level features, each of the feedback output amounts, each of the feedback delay amounts, and the output value determination formula of the first layer of the GRU temporal feature extraction layer; Determine the output value of each layer in the GRU temporal feature extraction layer except the first layer according to the ReLu activation function, the weight of the Lth layer of the GRU temporal feature extraction layer, the output value of the L-1th layer of the GRU temporal feature extraction layer, the bias of the L-1th layer of the GRU temporal feature extraction layer, and the output value determination formula of each layer in the GRU temporal feature extraction layer except the first layer, wherein the Lth layer of the GRU temporal feature extraction layer is any layer other than the first layer in the GRU temporal feature extraction layer; The output value of the first layer of the GRU temporal feature extraction layer is determined by the following formula: ; Represents the output value of the first layer of the GRU temporal feature extraction layer at time t, represents the ReLu activation function, represents the weight of the first layer of the GRU temporal feature extraction layer, represents the target high-level features input into the first layer of the GRU temporal feature extraction layer at time t, represents the feedback input amount output by the association layer at time t, Indicates each feedback delay amount output by the association layer; The formula for determining the output value of each layer except the first layer in the GRU temporal feature extraction layer is: ; represents the output value of the Lth layer except the first layer in the GRU temporal feature extraction layer at time t, represents the ReLu activation function, represents the weight of the Lth layer except the first layer in the GRU temporal feature extraction layer, represents the output value of the L-1th layer except the first layer in the GRU temporal feature extraction layer at time t, Represents the bias of the L-1th layer except the first layer in the GRU temporal feature extraction layer.

8. The method for analyzing the operating status of a distribution network according to any one of claims 1 to 7, wherein: After adjusting the future operating status information based on the real-time operating status information, the method further includes: Classifying the real-time operating status information and the future operating status information according to the degree of abnormality, and determining a confusion matrix of the operating status of the distribution network according to the classification results; The accuracy of the distribution network operation status prediction model is determined according to the distribution status of each classification result in the confusion matrix.

9. An electronic device, characterized in that: include: memory for storing computer programs; A processor, configured to implement the steps of the method for analyzing the operating status of a distribution network as claimed in any one of claims 1 to 8 when executing the computer program.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for analyzing the operating status of a distribution network according to any one of claims 1 to 8.

Images