A method, device and storage medium for identifying short-circuit faults in an active distribution network
Through the improved LSTM network model and convolutional layer extraction features, combined with the Matlab/Simulink simulation model, the problem of low accuracy of short-circuit fault recognition in the existing technology is solved, and the accuracy of identification and distribution of short-circuit faults in the active distribution network is improved. The reliability of the power grid.
Patent Information
- Application Number
- CN202411055853.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-08-02
AI Technical Summary
The prior art fails to fully consider the time series characteristics when identifying short-circuit failures of the active distribution network, resulting in low recognition accuracy.
The improved LSTM network model is adopted, combined with the convolutional layer to perform feature extraction and time series processing on short-circuit fault data, and a network model for short-circuit fault identification of active distribution network based on improved LSTM is built, and simulation models are built through Matlab/Simulink, and the simulation model is collected and processed. Three-phase current data, build a short-circuit fault data feature library, and use the Adam optimizer for model training.
It significantly improves the accuracy of short circuit fault identification, reduces the difficulty of data acquisition, saves human resources, and improves the reliability and safety of distribution network operation.
Smart Images

Figure CN118585911B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of short-circuit fault identification in an active distribution network, and in particular relates to a method, a device and a storage medium for identifying short-circuit faults in an active distribution network, and more specifically to a method, a device and a storage medium for identifying short-circuit faults in an active distribution network based on an improved LSTM. Background Art
[0002] With the rapid development of my country's power industry, more and more power users have begun to use self-provided power supplies such as photovoltaic power generation systems, energy storage systems, and self-provided power plants to improve the reliability and economy of electricity use. As a large number of self-provided power supplies on the user side are continuously connected to the distribution network, the analysis of the operating status of the distribution network has become more and more complicated, and the evolution principle of the short-circuit fault risk of the distribution network has become more and more complicated, which makes short-circuit faults difficult to detect. Therefore, short-circuit fault identification and detection of active distribution networks containing self-provided power supplies has become an urgent problem to be solved.
[0003] The rapid application of artificial intelligence technology in the electrical field has made data-driven and artificial intelligence-based models a popular method for short-circuit fault diagnosis in distribution networks with self-contained power supplies. Such methods require the use of massive short-circuit fault data for model training. Secondly, most existing artificial intelligence methods use conventional networks such as convolutional neural networks and BP networks for short-circuit fault diagnosis, which fails to fully consider the time series characteristics of actual distribution network short-circuit fault information, so the recognition accuracy is not high. Summary of the invention
[0004] In view of the shortcomings of the above-mentioned prior art, the present invention provides a method, device and storage medium for identifying short-circuit faults in an active distribution network. The purpose is to consider the short-circuit fault timing information and improve the local feature extraction capability of LSTM through the convolution layer, improve the accuracy of short-circuit fault identification, and realize the invention purpose of simple and effective active distribution network short-circuit fault identification.
[0005] The technical solution adopted by the present invention to achieve the above-mentioned purpose is:
[0006] A method for identifying short-circuit faults in an active distribution network, comprising:
[0007] Collect the three-phase current data of the line during the simulation process, build an active distribution network model with a self-contained power supply, conduct joint adjustment and preliminary simulation processing on the distribution network model, and obtain the original data samples of different types of short-circuit faults; convert the collected three-phase current data into zero-sequence current data to obtain different types of short-circuit fault samples, digitally annotate the different types of short-circuit fault data samples in sequence, build an active distribution network short-circuit fault data feature library, pre-process the original short-circuit fault data, and randomly divide the processed data feature library into a short-circuit fault training set and a short-circuit fault test set;
[0008] By constructing the input layer of the improved LSTM active distribution network short-circuit fault identification network model in the form of a one-dimensional matrix, the hidden layer of the network model and the output layer of the LSTM network model, a network model for short-circuit fault identification of active distribution network based on improved LSTM is built;
[0009] The active distribution network short-circuit fault training dataset is used to train the active distribution network short-circuit fault identification network model based on improved LSTM;
[0010] The active distribution network short-circuit fault test set was used to test the active distribution network short-circuit fault identification network model based on improved LSTM, and the identification results of the active distribution network short-circuit fault were obtained.
[0011] Furthermore, the active distribution network short-circuit fault data feature library includes: a short-circuit fault training set and a short-circuit fault test set.
[0012] Furthermore, the active distribution network short-circuit fault identification network model based on improved LSTM includes an input layer, a hidden layer and an output layer; the size of the input layer is determined by the feature quantity in the data feature library, the hidden layer includes two convolutional layers and five LSTM memory units, and the output layer includes a fully connected layer and a Softamx classifier; the LSTM memory unit includes: an input gate, a forget gate and an output gate.
[0013] Furthermore, the simulation process uses Matlab / Simulink to build an active distribution network simulation model containing multiple types of self-contained power supplies, and simulates different types of short-circuit faults, including:
[0014] The programmable three-phase power supply module is equivalent to replacing the upper power grid to supply power to the distribution network;
[0015] The transformer module converts the power supply voltage of the upper power grid into a 10kV voltage suitable for this distribution network;
[0016] Self-provided power supply, which includes photovoltaic, energy storage, electric vehicle charging piles and self-provided power plants. The addition of self-provided power supply makes the power flow distribution of the original distribution network more complicated;
[0017] The circuit module uses a π-type equivalent circuit to simulate the actual circuit;
[0018] The measurement module uses a three-phase voltage and current measurement module to measure the three-phase fault current waveform;
[0019] The fault module is used to simulate single-phase short circuit fault, phase-to-phase short circuit fault, two-phase short circuit to ground fault and three-phase short circuit fault;
[0020] The simulation of different types of short-circuit faults is to set the parameters of the established active distribution network simulation model, set different short-circuit fault point locations, short-circuit fault phase angles, grounding impedances and different short-circuit fault types; through the fault module in the established active distribution network model, simulate multiple situations under no fault, three single-phase short-circuit faults, three phase-to-phase short-circuit faults, three two-phase short-circuit grounding faults and one three-phase short-circuit fault, and randomly set the distribution network model parameters, set different fault point locations, short-circuit fault phase angles, and grounding impedances for each short-circuit fault, and use the measurement module to collect three-phase current waveform information during the simulation process.
[0021] Furthermore, the three-phase current data of the line in the simulation process is collected, an active distribution network model with a self-contained power supply is built, and the distribution network model is preliminarily processed by joint adjustment and simulation to obtain original data samples of different types of short-circuit faults; the collected three-phase current data is converted into zero-sequence current data to obtain different types of short-circuit fault samples, and the different types of short-circuit fault data samples are digitally labeled in sequence to build an active distribution network short-circuit fault data feature library, and the original short-circuit fault data is preprocessed, and the processed data feature library is randomly divided into a short-circuit fault training set and a short-circuit fault test set; including:
[0022] Step 11. Collect the three-phase current data of the line during the simulation process;
[0023] Step (1) Use Matlab / Simulink tools to build an active distribution network model with self-contained power supply;
[0024] Step (2) Write a Matlab script program to control the parameters of each module of the distribution network model through the script code to achieve joint debugging between the two;
[0025] Step (3) Simulation parameter setting: setting each type of fault in turn, randomly setting the fault location, fault phase angle, and grounding resistance, keeping the line length unchanged, to set the fault location;
[0026] Step (4) start simulation;
[0027] Step (5) stops the simulation after running for 0.2 seconds;
[0028] Step (6) automatically saves the fault three-phase current simulation data as a CSV file;
[0029] Step (7) repeats steps (5)-(7) to generate the original single-class fault samples required for each class of faults;
[0030] Step (8) Repeat steps (4)-(8) to obtain original samples of all types of faults;
[0031] Step 12: Convert the collected three-phase current data into zero-sequence current data, that is, process the original sample into a one-dimensional data form, and the formula is as follows:
[0032] ;
[0033] in, I 0 ( k ) is the zero-sequence short-circuit current, They are the original three-phase current data information of the line at time t, where t=1,2,...,T, and T is the maximum time sampling point;
[0034] The above zero-sequence current is processed by Fourier algorithm to extract the zero-sequence current under the fundamental wave and the zero-sequence current under the 5th harmonic, as shown in the following formula:
[0035]
[0036] In the formula, x(t) is the Fourier function, ∞ is positive infinity, n is the harmonic number, n=0,1,2,...∞; a n and b n are the amplitudes of the cosine and sine terms of each harmonic, t0 is the starting time of the extracted data, t is the current time, and ω1 is the angular frequency;
[0037] According to the Fourier series principle, we can get:
[0038]
[0039]
[0040] Where N is the number of sampling points in one cycle, For a cycle k Current sampling value, is the current sampling error in the present invention, which indicates the deviation between the sampling value and the actual value during the sampling process. ; n is the extracted zero-sequence current component, when extracting the fundamental component, n=1, when extracting the fifth harmonic component, n=5, For the k The current sampling value corresponds to the cosine term of each harmonic. For the k Each current sampling value corresponds to the sinusoidal term of each harmonic;
[0041] The short-circuit fault characteristics are expanded into a one-dimensional vector form, as shown in the following formula:
[0042] ƒ = [ I 1 1 , I 21 ,... I N 1 , I 1 5 , I 2 5 ,... I N 5 ];
[0043] in, f is a short circuit fault sample, I N 1 Represents the zero-sequence current fundamental wave data of the Nth sampling point, I N 5 Represents the zero-sequence current 5th harmonic data of the Nth sampling point;
[0044] Step 13. The obtained short-circuit fault samples containing the zero-sequence current fundamental and the fifth harmonic components are digitally labeled in the order of 1-11 for 11 data samples under non-fault, three single-phase short-circuit faults, three phase-to-phase short-circuit faults, three two-phase short-circuit grounding faults and one three-phase short-circuit fault, and all the labeled samples are placed in the same CSV table file to form a fault data set.
[0045] Among them, the three single-phase short-circuit faults include the three phases A, B and C being grounded respectively; the three interphase short-circuit faults include AB, AC and BC interphase short circuits; the three two-phase short-circuit grounding faults include AB_G, AC_G and BC_G two-phase short-circuit grounding faults; the one three-phase short-circuit fault includes ABC three-phase short circuit.
[0046] The data is preprocessed, the formula is as follows:
[0047] ;
[0048] ;
[0049] ;
[0050] In the above formula, is the data feature library, x is the original data, N is the sample value, is the i-th sample value, is the sample mean, is the standard deviation, is the sample deviation rate;
[0051] Step 14. After the preprocessed data set is constructed, the processed data feature library is shuffled and randomly divided into a short-circuit fault training set and a short-circuit fault test set in a ratio of 4:1.
[0052] Furthermore, an active distribution network short-circuit fault identification network model based on improved LSTM is built, including:
[0053] Step 1. Construct the input layer of the improved LSTM active distribution network short-circuit fault identification network model in the form of a one-dimensional matrix;
[0054] Step 2. Construct a hidden layer of the network model to extract local short-circuit fault feature information and time series features of the short-circuit fault data sample;
[0055] Step 3. Construct the LSTM network model output layer to output the corresponding short-circuit fault type identification result; including:
[0056] By constructing a fully connected layer, the local features of the short-circuit fault data samples are combined to form a one-dimensional vector, and the output value of multiple classifications is obtained;
[0057] Among them, the fully connected layer combines the local short-circuit fault information containing time series feature information, connects all local short-circuit fault features together, and forms a one-dimensional vector containing all local short-circuit fault features;
[0058] The multi-classification output values are converted into probability values in the range of [0, 1] through the Softmax classifier to form a one-dimensional vector representing the probability of each class label. The label with the largest probability represents the result of the short-circuit fault type identification.
[0059] The classifier converts the output value into a [0, 1] probability value as shown below:
[0060] ;
[0061] in, z i For the i The output value of the node, Y is the number of output nodes, that is, the number of classification categories, and e is an exponential function The base of is a constant;
[0062] Step 4. Set activation functions between each network layer of the network model. Use Relu activation function between convolutional layers and between convolutional layers and LSTM layers. Use Sigmoid activation function between other layers to complete the construction of the improved LSTM active distribution network short-circuit fault identification network model. The mathematical expression of the activation function is expressed as:
[0063] ;
[0064] ;
[0065] in, x is the value of the neural unit in the previous network layer, f is a short circuit fault sample, e is an exponential function The base of is a constant.
[0066] Furthermore, the active distribution network short-circuit fault training data set is used to train the active distribution network short-circuit fault identification network model based on the improved LSTM; the active distribution network short-circuit fault test set is used to test the active distribution network short-circuit fault identification network model based on the improved LSTM, and the identification result of the active distribution network short-circuit fault is obtained; including:
[0067] After completing the construction of the active distribution network short-circuit fault identification network model based on the improved LSTM, the network model is trained and tested. 1000 data samples are selected for each type of short-circuit fault in the short-circuit fault data set, totaling 1100 data samples; the data samples are shuffled, and the data set is divided into an active distribution network short-circuit fault training set and an active distribution network short-circuit fault test set in a ratio of 4:1, and input into the improved LSTM network;
[0068] The training parameters of the active distribution network short-circuit fault identification network model based on improved LSTM are set as follows:
[0069] The Adam solver was selected, the maximum number of training times was set to 500, the batch size was set to 2, the network initialization learning rate was set to 0.001, and the learning rate was set to drop to 0.0001 after 300 training times;
[0070] After completing the network model parameter setting, train the network model to obtain the recognition result;
[0071] The Adam optimizer is set to adaptively adjust the learning rate to accelerate the convergence speed of the network model. The active distribution network short-circuit fault training set and the active distribution network short-circuit fault test set are input into the improved LSTM network. The constructed improved LSTM network is trained and tested to complete the active distribution network short-circuit fault identification.
[0072] An active power distribution network short circuit fault identification device, comprising:
[0073] The acquisition and construction module is used to acquire the three-phase current data of the line during the simulation process, build an active distribution network model with a self-contained power supply, perform joint adjustment and preliminary simulation processing on the distribution network model, and obtain original data samples of different types of short-circuit faults; convert the acquired three-phase current data into zero-sequence current data, obtain samples of different types of short-circuit faults, digitally annotate the data samples of different types of short-circuit faults in sequence, construct an active distribution network short-circuit fault data feature library, and pre-process the original short-circuit fault data, and randomly divide the processed data feature library into a short-circuit fault training set and a short-circuit fault test set;
[0074] Building module, used to build an active distribution network short-circuit fault identification network model based on improved LSTM by constructing the input layer of the improved LSTM active distribution network short-circuit fault identification network model in the form of a one-dimensional matrix, the hidden layer of the network model and the output layer of the LSTM network model
[0075] The training module is used to train the active distribution network short-circuit fault identification network model based on improved LSTM using the active distribution network short-circuit fault training data set;
[0076] The testing module is used to test the active distribution network short-circuit fault identification network model based on improved LSTM using the active distribution network short-circuit fault test data set to obtain the identification result of the active distribution network short-circuit fault;
[0077] An active distribution network short-circuit fault identification device is used to implement the steps of any of the active distribution network short-circuit fault identification methods.
[0078] A computer device comprises a storage medium, a processor and a computer program stored in the storage medium and executable on the processor, wherein when the processor executes the computer program, the steps of any one of the methods for identifying short-circuit faults in an active power distribution network are implemented.
[0079] A computer storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any of the methods for identifying short-circuit faults in an active power distribution network.
[0080] The present invention has the following beneficial effects and advantages:
[0081] The present invention proposes to collect data sets by establishing a relevant distribution network model, which greatly reduces the difficulty of actual data collection.
[0082] The present invention extracts short-circuit fault features through a convolution layer, which greatly saves human resources and improves efficiency compared to the traditional method of manually extracting short-circuit fault features.
[0083] The present invention also fully considers the timing characteristics of short-circuit fault samples and uses an improved LSTM network model to process its timing characteristics, so that the accuracy of short-circuit fault identification is significantly improved.
[0084] The present invention also helps to improve the accuracy of short-circuit fault identification in the active distribution network, while improving the reliability and safety of the distribution network operation.
[0085] At present, existing algorithms mainly use wavelet analysis and machine learning algorithms to identify short-circuit faults, which rely too much on manual extraction of short-circuit fault features, have low extraction efficiency and fail to consider the temporal characteristics of short-circuit fault information. It is difficult to extract all short-circuit fault features, which easily leads to the loss of short-circuit fault feature information and low accuracy of short-circuit fault identification. This method uses convolutional layers to actively and fully mine short-circuit fault features, and uses LSTM memory units to process the temporal characteristics of short-circuit fault features, which greatly improves the recognition efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0086] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:
[0087] Figure 1 is a flow chart of the steps of the method of the present invention;
[0088] Figure 2 It is a structural diagram of a distribution network simulation model of the present invention;
[0089] Figure 3 is a flow chart of the sample simulation collection steps of the present invention;
[0090] Figure 4 It is the improved LSTM structure diagram of the present invention;
[0091] Figure 5 This is a comparison chart of the prediction results of the test set;
[0092] Figure 6 This is a comparison chart of the prediction results of the training set. DETAILED DESCRIPTION
[0093] In order to more clearly understand the above-mentioned objectives, features and advantages of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.
[0094] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.
[0095] Refer to the following Figure 1-Figure 6 The technical solutions of some embodiments of the present invention are described.
[0096] Example 1
[0097] The present invention provides an embodiment, which is a method for identifying short-circuit faults in an active distribution network. Figure 1 As shown, Figure 1 More specifically, a method for identifying short-circuit faults in an active distribution network based on an improved LSTM is disclosed, the method comprising:
[0098] Matlab / Simulink is used to build an active distribution network simulation model containing multiple types of self-contained power sources, and different types of short-circuit faults are simulated to provide a simulation environment and conditions for the next step of data collection. The self-contained power sources include photovoltaics, energy storage, electric vehicle charging piles and self-contained power plants.
[0099] Step S1: During the short-circuit fault simulation process, the three-phase current data of the line during the simulation process is collected and preliminarily processed to obtain the original data samples of all types of faults; the data samples of different short-circuit fault types are labeled to build a short-circuit fault data feature library for the active distribution network, and the original data is preprocessed.
[0100] The active distribution network short-circuit fault data feature library includes: a short-circuit fault training set and a short-circuit fault test set, which can provide data support for the subsequent short-circuit fault identification network model to perform training and testing for short-circuit fault identification.
[0101] Step S2: Build an active distribution network short-circuit fault identification network model based on improved LSTM, the network model includes an input layer, a hidden layer and an output layer. The LSTM active distribution network short-circuit fault identification network model can fully consider the temporal characteristics of short-circuit fault information, but it cannot effectively extract local features. Therefore, the LSTM active distribution network short-circuit fault identification network model is improved, and two convolutional layers are connected after the input layer of the LSTM network model to extract local features of short-circuit fault information. The input layer size of the constructed improved LSTM active distribution network short-circuit fault identification network model is determined by the feature quantity in the data feature library. This example is set to 1000×1×1. The hidden layer consists of two convolutional layers and five LSTM memory units. The output layer consists of a fully connected layer and a Softamx classifier to complete the model construction. A short-circuit fault identification network model is provided for the final active distribution network short-circuit fault identification.
[0102] Step S3: Combine the active distribution network short-circuit fault training set and the active distribution network short-circuit fault test set to train the active distribution network short-circuit fault identification network model based on the improved LSTM.
[0103] Step S4: Use the active distribution network short-circuit fault test set to test the active distribution network short-circuit fault identification network model based on the improved LSTM to obtain the identification result of the active distribution network short-circuit fault.
[0104] In the simulation process of the present invention, Matlab / Simulink is used to build an active distribution network simulation model containing multiple types of self-contained power supplies, and different types of short-circuit faults are simulated, specifically including:
[0105] In this embodiment, a 10kV active distribution network model with a self-contained power supply and a neutral point grounded through an arc suppression coil is established by Matlab / Simulink. Figure 2 As shown in the figure, DG is used to represent the self-contained power supply, and the model includes: a programmable three-phase power supply module, a transformer module, a voltage source module, a line module, a measurement module and a fault module. The fault module is used to simulate single-phase short circuit fault, inter-phase short circuit fault, two-phase short circuit grounding fault and three-phase short circuit fault.
[0106] The programmable three-phase power supply module is equivalent to replacing the upper power grid to supply power to the distribution network.
[0107] The transformer module converts the power supply voltage of the upper power grid into a 10kV voltage suitable for this distribution network.
[0108] Self-provided power supply, which includes photovoltaic, energy storage, electric vehicle charging piles and self-provided power plants. The addition of self-provided power supply makes the power flow distribution of the original distribution network more complicated.
[0109] The circuit module uses a π-type equivalent circuit to simulate the actual circuit.
[0110] The measurement module uses a three-phase voltage and current measurement module to measure the three-phase fault current waveform.
[0111] The fault module is used to simulate single-phase short circuit fault, phase-to-phase short circuit fault, two-phase short circuit to ground fault and three-phase short circuit fault.
[0112] Different types of short-circuit faults are simulated, and the parameters of the established active distribution network simulation model are set, including different short-circuit fault point locations, short-circuit fault phase angles, grounding impedances, and different short-circuit fault types.
[0113] In the simulation of different types of short-circuit faults, the fault module in the active distribution network model was used to simulate 11 situations including no fault, three single-phase short-circuit faults, three phase-to-phase short-circuit faults, three two-phase short-circuit grounding faults and one three-phase short-circuit fault, and the distribution network model parameters were randomly set. For each fault, different short-circuit fault point positions, short-circuit fault phase angles (0-360° random), and grounding impedances were set. The single simulation time was set to 0.2 seconds, and the sampling frequency was 5kHz, that is, 1000 short-circuit fault feature points were collected in one simulation, and each type of short-circuit fault was simulated 100 times, that is, the overall short-circuit fault data feature library had 11*100 data samples, and the measurement module was used to collect the three-phase current waveform information during the simulation.
[0114] Among them, the three single-phase short-circuit faults include the three phases A, B and C being grounded respectively; the three interphase short-circuit faults include AB, AC and BC interphase short circuits; the three two-phase short-circuit grounding faults include AB_G, AC_G and BC_G two-phase short-circuit grounding faults; the one three-phase short-circuit fault includes ABC three-phase short circuit.
[0115] As described in step S1, the three-phase current data of the line during the simulation is collected and preliminarily processed to obtain data samples; different short-circuit fault type data samples are labeled to build an active distribution network short-circuit fault data feature library, and the original data is preprocessed. Specifically including:
[0116] Step 11. The three-phase current data of the line during the simulation is collected. The process is as follows: Figure 3 As shown, specifically including:
[0117] Step (1) Use Matlab / Simulink tools to build an active distribution network model with self-contained power supply;
[0118] Step (2) Write a Matlab script program to control the parameters of each module of the distribution network model through the script code to achieve joint debugging between the two;
[0119] Step (3) Simulation parameter setting: Set each type of fault in turn, randomly set the fault location, fault phase angle, and ground resistance, keep the line length unchanged, and set the fault location to ensure the randomness of the fault;
[0120] Step (4) start simulation;
[0121] Step (5) stops the simulation after running for 0.2 seconds;
[0122] Step (6) automatically saves the fault three-phase current simulation data as a CSV file;
[0123] Step (7) repeats steps (5)-(7) to generate the original single-class fault samples required for each class of faults;
[0124] Among them, the number of cycles = the number of single-type fault samples = 100;
[0125] Step (8) Repeat steps (4)-(8) to obtain the original samples of all types of faults.
[0126] Among them, there are 11 categories in total, and the number of cycles = fault category = 11;
[0127] Step 12: Convert the collected three-phase current data into zero-sequence current data, that is, process the original sample into a one-dimensional data form, and the formula is as follows:
[0128] ;
[0129] in, I 0 ( k ) is the zero-sequence short-circuit current, They are the original three-phase current data information of the line at time t, where t=1,2,...,T, and T is the maximum time sampling point.
[0130] The above zero-sequence current is processed by Fourier algorithm to extract the zero-sequence current under the fundamental wave and the zero-sequence current under the 5th harmonic, as shown in the following formula:
[0131]
[0132] In the formula, x(t) is the Fourier function, ∞ is positive infinity, n is the harmonic number, n=0,1,2,...∞; a n and b n are the amplitudes of the cosine and sine terms of each harmonic, t0 is the starting time of the extracted data, t is the current time, and ω1 is the angular frequency.
[0133] According to the Fourier series principle, we can get:
[0134]
[0135]
[0136] Where N is the number of sampling points in one cycle, For a cycle k Current sampling value, is the current sampling error in the present invention, which indicates the deviation between the sampling value and the actual value during the sampling process. ; n is the extracted zero-sequence current component, when extracting the fundamental component, n=1, when extracting the fifth harmonic component, n=5, As one, For the k The current sampling value corresponds to the cosine term of each harmonic. As one, For the k Each current sampling value corresponds to the sinusoidal term of each harmonic.
[0137] The short-circuit fault feature is expanded into a one-dimensional vector form, as shown in the following formula:
[0138] ƒ = [ I 1 1 , I 2 1 ,... I N 1 , I 1 5 , I 2 5 ,... I N 5 ]
[0139] in, f is a short circuit fault sample, I N 1 Represents the zero-sequence current fundamental wave data of the Nth sampling point, I N 5 Represents the zero-sequence current fifth harmonic data of the Nth sampling point, where N is 1000 in the present invention.
[0140] Step 13. The obtained short-circuit fault samples containing the zero-sequence current fundamental and the fifth harmonic components are digitally labeled in the order of 1-11 for 11 data samples under non-fault, three single-phase short-circuit faults, three phase-to-phase short-circuit faults, three two-phase short-circuit grounding faults and one three-phase short-circuit fault, and all the labeled samples are placed in the same CSV table file to form a fault data set.
[0141] Among them, the three single-phase short-circuit faults include the three phases A, B and C being grounded respectively; the three interphase short-circuit faults include AB, AC and BC interphase short circuits; the three two-phase short-circuit grounding faults include AB_G, AC_G and BC_G two-phase short-circuit grounding faults; the one three-phase short-circuit fault includes ABC three-phase short circuit.
[0142] The data is preprocessed, the formula is as follows:
[0143] ;
[0144] ;
[0145] ;
[0146] In the above formula, is the data feature library, x is the original data, N is the sample value, is the i-th sample value, is the sample mean, is the standard deviation, is the sample deviation rate, in the present invention Indicates the deviation between the standard sample and the actual sample. =1.02
[0147] Step 14. Shuffle the processed data feature library and randomly divide it into a short-circuit fault training set and a short-circuit fault test set in a ratio of 4:1.
[0148] As described in step S2, building an active distribution network short-circuit fault identification network model based on improved LSTM includes the following steps:
[0149] Step 21. Construct the input layer of the improved LSTM active distribution network short-circuit fault identification network model in the form of a one-dimensional matrix. The size of the input layer is determined by the feature quantity in the short-circuit fault data feature library and must match the one-dimensional data sample size. The input layer size is set to 1000×1×1.
[0150] Step 22. Construct a hidden layer of the network model, which is used to extract local short-circuit fault feature information and time series features of short-circuit fault data samples. Specifically, it includes two convolutional layers and five layers of active distribution network short-circuit fault identification LSTM memory units. The LSTM memory unit includes: an input gate, a forget gate, and an output gate.
[0151] Step 23. Construct the output layer of the LSTM network model to output the corresponding short-circuit fault type identification results. The size of the output layer is determined by the number of output results. The number of data set labels is set to 11×1×1, which specifically includes a fully connected layer and a Softmax classifier.
[0152] (1) By constructing a fully connected layer, the local features of the short-circuit fault data samples are combined to form a one-dimensional vector, and the output value of the multi-classification is obtained.
[0153] Among them, the fully connected layer combines the local short-circuit fault information containing time series feature information, connects all local short-circuit fault features together, and forms a one-dimensional vector containing all local short-circuit fault features.
[0154] (2) The multi-classification output values are converted into probability values in the range of [0, 1] through a Softmax classifier to form a one-dimensional vector representing the probability of each class label, and the short-circuit fault type identification result represented by the label with the highest probability.
[0155] Among them, the one-dimensional feature vector output value is converted into a probability value in the range of [0, 1] through a Softmax classifier to form a one-dimensional short-circuit fault type probability vector representing the probability of each type of short-circuit fault label, and the short-circuit fault type represented by the label with the largest probability is the identification result.
[0156] The classifier converts the output value into a [0, 1] probability value as shown below:
[0157] ;
[0158] in, z i For the i The output value of the node, Y is the number of output nodes, that is, the number of classification categories, and e is an exponential function The base of is a constant.
[0159] Step 24. Set activation functions between each network layer of the network model. Use Relu activation function between convolutional layers and between convolutional layers and LSTM layers. Use Sigmoid activation function between other layers. Finally, the improved LSTM active distribution network short-circuit fault identification network model is built. The mathematical expression of the activation function is expressed as:
[0160] ;
[0161] ;
[0162] in, x is the value of the neural unit in the previous network layer, f is a short circuit fault sample, e is an exponential function The base of is a constant.
[0163] As described in step S3 and step S4, the active distribution network short-circuit fault training data set is used to train the active distribution network short-circuit fault identification network model based on the improved LSTM; the active distribution network short-circuit fault test set is used to test the active distribution network short-circuit fault identification network model based on the improved LSTM, and the identification result of the active distribution network short-circuit fault is obtained, which specifically includes:
[0164] After the construction of the active distribution network short-circuit fault identification network model based on improved LSTM is completed, the model should be trained and tested. 1000 data samples are selected for each type of fault in the short-circuit fault data set, totaling 1100 data samples. The data samples are shuffled, and the data set is divided into an active distribution network short-circuit fault training set and an active distribution network short-circuit fault test set in a ratio of 4:1, and then input into the improved LSTM network model.
[0165] The training parameters of the active distribution network short-circuit fault identification network model based on improved LSTM are set as follows:
[0166] The Adam solver was selected, the maximum number of training times was set to 500, the batch size was set to 2, the network model initialization learning rate was set to 0.001, and the learning rate was set to drop to 0.0001 after 300 training times.
[0167] After completing the network model parameter setting, start training the model and get the recognition results, such as Figure 5 and Figure 6 As shown, the accuracy of the fault identification results of the present invention includes Figure 5 Comparison chart of prediction results for the test set and Figure 6 This is a comparison chart of the prediction results of the training set.
[0168] The improved LSTM structure diagram of the present invention is as follows Figure 4 shown.
[0169] The Adam optimizer is set to adaptively adjust the learning rate to accelerate the convergence speed of the network model. The active distribution network short-circuit fault training set and the active distribution network short-circuit fault test set are input into the improved LSTM network model. The constructed improved LSTM network model is trained and tested to complete the active distribution network short-circuit fault identification.
[0170] Example 2
[0171] The present invention further provides an embodiment, which is an active distribution network short-circuit fault identification device, comprising:
[0172] The acquisition and construction module is used to acquire the three-phase current data of the line during the simulation process, build an active distribution network model with a self-contained power supply, perform joint adjustment and preliminary simulation processing on the distribution network model, and obtain original data samples of different types of short-circuit faults; convert the acquired three-phase current data into zero-sequence current data, obtain samples of different types of short-circuit faults, digitally annotate the data samples of different types of short-circuit faults in sequence, construct an active distribution network short-circuit fault data feature library, and pre-process the original short-circuit fault data, and randomly divide the processed data feature library into a short-circuit fault training set and a short-circuit fault test set;
[0173] A building module is used to build an active distribution network short-circuit fault identification network model based on improved LSTM by constructing an improved LSTM active distribution network short-circuit fault identification network model input layer, a hidden layer of the network model and an LSTM network model output layer in the form of a one-dimensional matrix;
[0174] The training module is used to train the active distribution network short-circuit fault identification network model based on improved LSTM using the active distribution network short-circuit fault training data set;
[0175] The testing module is used to train and test the active distribution network short-circuit fault identification network model based on the improved LSTM using the active distribution network short-circuit fault test set to obtain the identification result of the active distribution network short-circuit fault.
[0176] The active distribution network short-circuit fault identification device described in the present invention is used to implement the steps of the active distribution network short-circuit fault identification method described in Example 1.
[0177] Example 3
[0178] Based on the same inventive concept, an embodiment of the present invention further provides a computer device, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor. When the processor executes the computer program, the steps of the active distribution network short circuit fault identification method described in Embodiment 1 are implemented.
[0179] Example 4
[0180] Based on the same inventive concept, an embodiment of the present invention further provides a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of a method for identifying short-circuit faults in an active distribution network described in Example 1 are implemented.
[0181] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0182] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0183] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0184] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0185] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.
Claims
1. A method for identifying short-circuit faults in an active distribution network, characterized in that: include: The three-phase current data of the line during the simulation process is collected, an active distribution network model with a self-contained power supply is built, and the distribution network model is preliminarily processed by joint adjustment and simulation to obtain original data samples of different types of short-circuit faults; the collected three-phase current data is converted into zero-sequence current data to obtain different types of short-circuit fault samples, and the different types of short-circuit fault data samples are digitally labeled in sequence to build an active distribution network short-circuit fault data feature library, and the original short-circuit fault data is preprocessed, and the processed data feature library is randomly divided into a short-circuit fault training set and a short-circuit fault test set; By constructing the input layer of the improved LSTM active distribution network short-circuit fault identification network model in the form of a one-dimensional matrix, the hidden layer of the network model and the output layer of the LSTM network model, an active distribution network short-circuit fault identification network model based on the improved LSTM is built; the active distribution network short-circuit fault training data set is used to train the active distribution network short-circuit fault identification network model based on the improved LSTM; the active distribution network short-circuit fault test set is used to test the active distribution network short-circuit fault identification network model based on the improved LSTM, and the identification result of the active distribution network short-circuit fault is obtained; The active distribution network short-circuit fault data feature library includes: a short-circuit fault training set and a short-circuit fault test set; the active distribution network short-circuit fault identification network model based on improved LSTM includes an input layer, a hidden layer and an output layer; the size of the input layer is determined by the feature quantity in the data feature library, the hidden layer includes two convolutional layers and five LSTM memory units, and the output layer includes a fully connected layer and a Softamx classifier; the LSTM memory unit includes: an input gate, a forget gate and an output gate; the simulation process is to use Matlab / Simulink to build an active distribution network simulation model containing multiple types of self-contained power supplies, and to simulate different types of short-circuit faults. The simulation includes: the programmable three-phase power supply module is equivalent to the upper power grid to supply power to this distribution network; the transformer module converts the power supply voltage of the upper power grid into a 10kV voltage suitable for this distribution network; the self-provided power supply includes photovoltaic, energy storage, electric vehicle charging piles and self-provided power plants. The addition of self-provided power supply makes the power flow distribution of the original distribution network more complicated; the line module uses a π-type equivalent circuit to simulate the actual line; the measurement module uses a three-phase voltage and current measurement module to measure the three-phase fault current waveform; the fault module is used to simulate single-phase short circuit fault, phase-to-phase short circuit fault, two-phase short circuit grounding fault and three-phase short circuit fault ; The simulation of different types of short-circuit faults is to set the parameters of the established active distribution network simulation model, set different short-circuit fault point locations, short-circuit fault phase angles, grounding impedances and different short-circuit fault types; through the fault module in the established active distribution network model, simulate multiple situations under no fault, three single-phase short-circuit faults, three phase-to-phase short-circuit faults, three two-phase short-circuit grounding faults and one three-phase short-circuit fault, and randomly set the distribution network model parameters, set different fault point locations, short-circuit fault phase angles, and grounding impedances for each short-circuit fault, and use the measurement module to measure the three-phase power during the simulation process. The method collects three-phase current data of the line during the simulation process, builds an active distribution network model with a self-contained power supply, performs joint adjustment and preliminary simulation processing on the distribution network model, and obtains original data samples of different types of short-circuit faults; converts the collected three-phase current data into zero-sequence current data, obtains samples of different types of short-circuit faults, digitally annotates the samples of different types of short-circuit fault data in sequence, constructs a data feature library of short-circuit faults of the active distribution network, pre-processes the original short-circuit fault data, and randomly divides the processed data feature library into a short-circuit fault training set and a short-circuit fault test set; comprising: step 11.Collect the three-phase current data of the line during the simulation process; Step (1) Use Matlab / Simulink tools to build an active distribution network model with a self-contained power supply; Step (2) Write a Matlab script program, and control the parameters of each module of the distribution network model through the script code to achieve joint adjustment between the two; Step (3) Simulation parameter setting, set each type of fault in turn, randomly set the fault location, fault phase angle, and grounding resistance, and keep the line length unchanged to set the fault location; Step (4) Start the simulation; Step (5) Stop the simulation after running for 0.2 seconds; Step (6) Automatically save the fault three-phase current simulation data as a CSV file; Step (7) Repeat steps (5)-(7) to generate the original single-type fault samples required for each type of fault; Step (8) Repeat steps (4)-(8) to obtain the original samples of all types of faults; Step 12. Convert the collected three-phase current data into zero-sequence current data, that is, process the original samples into one-dimensional data form, the formula is as follows:. Among them, I 0 (k) is the zero-sequence short-circuit current, I A (t), I B (t), I C (t) are the original three-phase current data information of the line at time t, where t = 1, 2, ..., T, and T is the maximum time sampling point; The above zero-sequence current is processed by Fourier algorithm to extract the zero-sequence current under the fundamental wave and the zero-sequence current under the 5th harmonic, as shown in the following formula: Where x(t) is the Fourier function, ∞ is positive infinity, n is the harmonic order, n = 0, 1, 2, ... ∞; a n and b n are the amplitudes of the cosine and sine terms of each harmonic, t0 is the starting time of the extracted data, t is the current time, and ω1 is the angular frequency; According to the Fourier series principle, we can get: Where N is the number of sampling points in one cycle, x k is the kth current sampling value in a cycle, λ is the current sampling error, which indicates the deviation between the sampling value and the actual value during the sampling process, and is set to 0.97<λ<1.03; n is the extracted zero-sequence current component, n=1 when extracting the fundamental component, and n=5 when extracting the fifth harmonic component, is the cosine term of each harmonic corresponding to the kth current sampling value, is the sinusoidal term of each harmonic corresponding to the kth current sampling value; the short-circuit fault feature is expanded into a one-dimensional vector form, as shown in the following formula: f = [I1 1 ,I2 1 ,...I N 1 ,I1 5 ,I2 5 ,...I N 5 ], where f is the short-circuit fault sample, I N 1 Represents the zero-sequence current fundamental wave data of the Nth sampling point, I N 5 Represents the zero-sequence current 5th harmonic data of the Nth sampling point; Step 13. The obtained short-circuit fault samples containing the zero-sequence current fundamental wave and the 5th harmonic component are digitally labeled in sequence for multiple data samples under non-fault, three single-phase short-circuit faults, three phase-to-phase short-circuit faults, three two-phase short-circuit grounding faults and one three-phase short-circuit fault, and all the labeled samples are placed in the same CSV table file to form a short-circuit fault data set; wherein the three single-phase short-circuit faults include A, B, and C three-phase grounding respectively; the three phase-to-phase short-circuit faults include AB, AC, and BC three-phase short circuits; the three two-phase short-circuit grounding faults include AB_G, AC_G, and BC_G two-phase short-circuit grounding faults; the one three-phase short-circuit fault includes ABC three-phase short circuit; the data is preprocessed, and the formula is as follows: In the above formula, x′ is the data feature library, x is the original data, N is the sample value, and x i is the i-th sample value, μ is the sample mean, σ is the standard deviation, and β is the sample deviation rate; Step 14. Shuffle the processed data feature library and randomly divide it into a short-circuit fault training set and a short-circuit fault test set in a ratio of 4:
1.
2. The method for identifying short-circuit faults in an active power distribution network according to claim 1, characterized in that: The improved LSTM active distribution network short-circuit fault identification network model input layer, the hidden layer of the network model and the LSTM network model output layer are constructed in the form of a one-dimensional matrix to build an active distribution network short-circuit fault identification network model based on the improved LSTM, including: Step 1. Construct the input layer of the improved LSTM active distribution network short-circuit fault identification network model in the form of a one-dimensional matrix; Step 2. Construct a hidden layer of the network model to extract local short-circuit fault feature information and time series features of the short-circuit fault data sample; Step 3. Construct the LSTM network model output layer to output the corresponding short-circuit fault type identification result; including: By constructing a fully connected layer, the local features of the short-circuit fault data samples are combined to form a one-dimensional vector, and the output value of multiple classifications is obtained; Among them, the fully connected layer combines the local short-circuit fault information containing time series feature information, connects all local short-circuit fault features together, and forms a one-dimensional vector containing all local short-circuit fault features; The multi-classification output values are converted into probability values in the range of [0, 1] through the Softmax classifier to form a one-dimensional vector representing the probability of each class label. The label with the largest probability represents the result of the short-circuit fault type identification. The classifier converts the output value into a [0, 1] probability value as shown below: Among them, z i is the output value of the i-th node, Y is the number of output nodes, that is, the number of classification categories, and e is an exponential function The base of is a constant; Step 4. Set activation functions between each network layer of the network model. Use Relu activation function between convolutional layers and between convolutional layers and LSTM layers, and use Sigmoid activation function between other layers to complete the construction of the improved LSTM active distribution network short-circuit fault identification network model. The mathematical expression of the activation function is: Relu:f(x)=max(0,x) Sigmoid: Among them, x is the value of the neural unit in the previous network layer, f is the short-circuit fault sample, and e is the exponential function e -x The base of is a constant.
3. The method for identifying short-circuit faults in an active power distribution network according to claim 1, characterized in that: The active distribution network short-circuit fault training data set is used to train the active distribution network short-circuit fault identification network model based on the improved LSTM; The active distribution network short-circuit fault test set is used to test the active distribution network short-circuit fault identification network model based on improved LSTM, and the identification results of the active distribution network short-circuit fault are obtained; including: After completing the construction of the active distribution network short-circuit fault identification network model based on the improved LSTM, the network model is trained and tested. 1000 data samples are selected for each type of short-circuit fault in the short-circuit fault data set, totaling 1100 data samples; the data samples are shuffled, and the data set is divided into an active distribution network short-circuit fault training set and an active distribution network short-circuit fault test set in a ratio of 4:1, and input into the improved LSTM network; The training parameters of the active distribution network short-circuit fault identification network model based on improved LSTM are set as follows: The Adam solver was selected, the maximum number of training times was set to 500, the batch size was set to 2, the network initialization learning rate was set to 0.001, and the learning rate was set to drop to 0.0001 after 300 training times; After completing the network model parameter setting, train the network model to obtain the recognition result; The Adam optimizer is set to adaptively adjust the learning rate to accelerate the convergence speed of the network model. The active distribution network short-circuit fault training set and the active distribution network short-circuit fault test set are input into the improved LSTM network. The constructed improved LSTM network is trained and tested to complete the active distribution network short-circuit fault identification.
4. An active distribution network short-circuit fault identification device, characterized in that: include: The acquisition and construction module is used to acquire the three-phase current data of the line during the simulation process, build an active distribution network model with a self-contained power supply, perform joint adjustment and preliminary simulation processing on the distribution network model, and obtain original data samples of different types of short-circuit faults; convert the acquired three-phase current data into zero-sequence current data, obtain samples of different types of short-circuit faults, digitally annotate the data samples of different types of short-circuit faults in sequence, construct an active distribution network short-circuit fault data feature library, and pre-process the original short-circuit fault data, and randomly divide the processed data feature library into a short-circuit fault training set and a short-circuit fault test set; A building module is used to build an active distribution network short-circuit fault identification network model based on improved LSTM by constructing an improved LSTM active distribution network short-circuit fault identification network model input layer, a hidden layer of the network model and an LSTM network model output layer in the form of a one-dimensional matrix; The training module is used to train the active distribution network short-circuit fault identification network model based on improved LSTM using the active distribution network short-circuit fault training data set; The testing module is used to test the active distribution network short-circuit fault identification network model based on improved LSTM using the active distribution network short-circuit fault test data set to obtain the identification result of the active distribution network short-circuit fault; An active distribution network short-circuit fault identification device is used to implement the steps of an active distribution network short-circuit fault identification method as described in any one of claims 1 to 3.
5. A computer device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for identifying short-circuit faults in an active power distribution network as described in any one of claims 1 to 3 are implemented.
6. A computer storage medium, characterized in that: The computer storage medium stores a computer program, and when the computer program is executed by a processor, the steps of a method for identifying short-circuit faults in an active distribution network as described in any one of claims 1 to 3 are implemented.
