Low-voltage active transformer area power grid fault identification model construction and identification method and system

Abstract

The invention provides a low-voltage active area power grid fault identification model construction and identification method and system, and the method comprises the steps: constructing a low-voltage active area power grid model based on a distributed new energy output prediction value, and obtaining the simulation fault operation data of a power grid under different set fault types; based on the time-frequency domain characteristics of the simulation fault operation data and the fault type corresponding to the simulation fault operation data, training a single hidden layer feed-forward neural network by using an extreme learning machine optimized by a particle swarm algorithm to obtain a trained low-voltage active area power grid fault identification model; according to the method and the system, a refined low-voltage active area power grid model is established through a distributed new energy output predicted value, so that the randomness and the uncertainty of distributed new energy output can be considered in simulated fault operation data, and a fault identification model can be more suitable for power grid fault identification of distributed new energy access; and the identification result is more accurate.

Description

Technical Field

[0001] The present invention belongs to the technical field of low-voltage distribution transformer substations in distribution networks, and specifically relates to the construction of a low-voltage active substation fault identification model, an identification method and a system. Background Art

[0002] With the vigorous development of new power systems dominated by new energy and the proposal of the "dual carbon" goal, new energy technologies represented by wind and solar energy have opened up new ways to optimize the energy structure and promote green and sustainable development. Photovoltaic power generation is known for its cleanliness, low carbon footprint and sustainability, and has become one of the most promising forms of renewable energy generation. In the low-voltage distribution area, the integration of 380 / 220V low-voltage distributed photovoltaics into the low-voltage distribution network has brought a series of new problems. The low-voltage distribution area has gradually evolved from the traditional form of one-way energy transmission, single user characteristics, and clear "source-grid-load" interface to a new form with two-way flow, complex roles of producers and consumers, and coexistence of large and small power points, which poses a severe challenge to the safe use of low-voltage electricity. The large-scale access of high-proportion distributed photovoltaics has an impact on the traditional distribution method. The traditional distribution network fault analysis and relay protection schemes are no longer suitable for the distribution network system of photovoltaic systems. Therefore, after the access of distributed new energy, such as photovoltaics, how to better identify the faults of the low-voltage distribution network so that the distribution network does not cause the protection system to malfunction or refuse to operate when a short-circuit fault occurs has become a key issue that needs to be solved at present. Summary of the invention

[0003] In order to overcome the above-mentioned deficiencies of the prior art, the present invention proposes a method for constructing a low-voltage active area power grid fault identification model, comprising:

[0004] Based on the typical topology of the low-voltage active area, the pre-acquired predicted value of the distributed new energy output in the low-voltage active area is connected to the typical topology to construct a low-voltage active area power grid model;

[0005] Performing a power grid fault simulation on the low-voltage active substation power grid model to obtain simulated fault operation data of the power grid where the low-voltage active substation is located under different set fault types;

[0006] Constructing a sample set based on the time-frequency domain characteristics of the simulated fault operation data and the fault type corresponding to the simulated fault operation data;

[0007] Based on the sample set, an extreme learning machine optimized by a particle swarm algorithm is used to train a single hidden layer feedforward neural network to obtain a trained low-voltage active substation area power grid fault identification model.

[0008] Preferably, based on the sample set, the extreme learning machine optimized by the particle swarm algorithm is used to train the single-hidden-layer feedforward neural network to obtain a trained low-voltage active power grid fault identification model, including:

[0009] Based on the sample set, the input layer and the hidden layer of the single-hidden-layer feedforward neural network are trained by the particle swarm algorithm to obtain a trained input layer and hidden layer;

[0010] The weights of the trained input layer and the biases of the hidden layer are used as the initial parameters of the extreme learning machine;

[0011] Based on the sample set, the single-hidden-layer feedforward neural network is trained by the extreme learning machine to obtain a trained low-voltage active power grid fault identification model.

[0012] Preferably, the training of the input layer and the hidden layer of the single-hidden-layer feedforward neural network by the particle swarm algorithm based on the sample set to obtain a trained input layer and hidden layer includes:

[0013] Initialize the parameters of the particle swarm algorithm, randomly generate the number of particles, positions and velocities of the particle swarm, and the position coordinate values of each particle represent a set of weights of the input layer and biases of the hidden layer of the single-hidden-layer feedforward neural network;

[0014] Based on the sample set, calculate the fitness value of the position of each particle according to the fitness function, determine the individual extreme value and the global extreme value of the particle; update the velocity and position of the particle, and obtain the new individual extreme value and global extreme value of the particle according to the fitness function; repeatedly iterate to update the velocity and position of the particle until the maximum number of iterations is reached or the change in the global extreme value is less than the set threshold, and use the position coordinate values of the particle corresponding to the global extreme value at this time as the final weights of the input layer and biases of the hidden layer of the single-hidden-layer feedforward neural network to complete the training of the input layer and the hidden layer.

[0015] Preferably, the training of the single-hidden-layer feedforward neural network by the extreme learning machine based on the sample set to obtain a trained low-voltage active power grid fault identification model includes:

[0016] The sample set is divided into a training set and a test set according to a set ratio;

[0017] Based on the training set, using the time-frequency domain features of the simulated fault operation data in the training set as the input and the fault type corresponding to the simulated fault operation data as the output, train a single-hidden-layer feedforward neural network; use the extreme learning machine to calculate the output layer weights of the neural network prediction model, and use an optimizer to optimize and update the network parameters of the single-hidden-layer feedforward neural network; use the test set to evaluate the single-hidden-layer feedforward neural network during the training process until the loss function of the single-hidden-layer feedforward neural network converges, and obtain a trained low-voltage active distribution network fault identification model.

[0018] Preferably, constructing a sample set based on the time-frequency domain features of the simulated fault operation data and the fault type corresponding to the simulated fault operation data includes:

[0019] Perform set time-domain feature extraction on the simulated fault operation data to obtain the time-domain feature extraction result of the simulated fault operation data, and the time-domain feature extraction result includes peak-to-peak value, rectified average value, variance, standard deviation, effective value, skewness, kurtosis, peak factor, and waveform factor;

[0020] Perform spectrum analysis on the simulated fault operation data to obtain the frequency components of the simulated fault operation data, and perform set frequency-domain feature extraction on the frequency components to obtain the frequency-domain feature extraction result of the simulated fault operation data, and the frequency-domain feature extraction result includes center frequency, average frequency, root mean square frequency, and frequency standard deviation;

[0021] Use the time-domain feature extraction result and the frequency-domain feature extraction result as the time-frequency domain features, use the time-frequency domain features of several simulated fault operation data as input data, and use the fault type corresponding to the simulated fault operation data as output data to construct several samples and obtain a sample set.

[0022] Preferably, the process of obtaining the predicted value of distributed new energy output includes:

[0023] Obtain the historical data of the distributed new energy output of the low-voltage active distribution network;

[0024] Use the historical data as the input and use a pre-constructed neural network prediction model to perform prediction to obtain the predicted value of the distributed new energy output.

[0025] Preferably, the fault types include solid single-phase grounding fault, single-phase grounding fault with small resistance, single-phase grounding fault with high resistance, leakage and electric shock fault, and phase-to-phase short circuit fault.

[0026] Based on the same inventive concept, the present invention also provides a low-voltage active distribution network fault identification model construction system, including:

[0027] A power grid model construction module, which is used to access the predicted values of distributed new energy output in the low-voltage active area obtained in advance into the typical topology based on the typical topology of the low-voltage active area, and construct a low-voltage active area power grid model;

[0028] A simulation data acquisition module, which is used to perform power grid fault simulation on the low-voltage active area power grid model, and obtain the simulation fault operation data of the power grid where the low-voltage active area is located under different set fault types;

[0029] A sample construction module, which is used to construct a sample set based on the time-frequency domain characteristics of the simulation fault operation data and the fault types corresponding to the simulation fault operation data;

[0030] An identification model training module, which is used to train a single-hidden layer feedforward neural network based on the sample set by using an extreme learning machine optimized by a particle swarm algorithm, and obtain a trained low-voltage active area power grid fault identification model.

[0031] Preferably, the identification model training module includes:

[0032] A first sub-module, which is used to train the input layer and the hidden layer of the single-hidden layer feedforward neural network by using a particle swarm algorithm based on the sample set, and obtain a trained input layer and hidden layer;

[0033] A second sub-module, which is used to use the weight of the trained input layer and the bias of the hidden layer as the initial parameters of the extreme learning machine;

[0034] A third sub-module, which is used to train the single-hidden layer feedforward neural network by using the extreme learning machine based on the sample set, and obtain a trained low-voltage active area power grid fault identification model.

[0035] Preferably, the first sub-module is specifically used for:

[0036] Initialize the parameters of the particle swarm algorithm, randomly generate the number of particles, positions and velocities of the particle swarm, and the position coordinate values of each particle represent a set of weights of the input layer and biases of the hidden layer of the single-hidden layer feedforward neural network;

[0037] Based on the sample set, calculate the fitness value of the position of each particle according to the fitness function, determine the individual extreme value and the global extreme value of the particle; update the velocity and position of the particle, and obtain the new individual extreme value and global extreme value of the particle according to the fitness function; repeatedly iterate to update the velocity and position of the particle until the maximum number of iterations is reached or the change of the global extreme value is less than the set threshold, and use the position coordinate value of the particle corresponding to the global extreme value at this time as the final weights of the input layer and biases of the hidden layer of the single-hidden layer feedforward neural network, and complete the training of the input layer and the hidden layer.

[0038] Preferably, the third sub-module is specifically configured to:

[0039] Divide the sample set into a training set and a test set according to a set ratio;

[0040] Based on the training set, use the time-frequency domain features of the simulated fault operation data in the training set as the input, and the fault type corresponding to the simulated fault operation data as the output to train a single-hidden-layer feedforward neural network; use the extreme learning machine to calculate the output layer weights of the neural network prediction model, and use an optimizer to optimize and update the network parameters of the single-hidden-layer feedforward neural network; use the test set to evaluate the single-hidden-layer feedforward neural network during the training process until the loss function of the single-hidden-layer feedforward neural network converges, and obtain a trained low-voltage active distribution network fault identification model.

[0041] Preferably, the sample construction module is specifically configured to:

[0042] Extract set time-domain features from the simulated fault operation data to obtain the time-domain feature extraction result of the simulated fault operation data, where the time-domain feature extraction result includes peak-to-peak value, rectified average value, variance, standard deviation, effective value, skewness, kurtosis, peak factor, and waveform factor;

[0043] Perform spectrum analysis on the simulated fault operation data to obtain the frequency components of the simulated fault operation data, and extract set frequency-domain features from the frequency components to obtain the frequency-domain feature extraction result of the simulated fault operation data, where the frequency-domain feature extraction result includes center frequency, average frequency, root mean square frequency, and frequency standard deviation;

[0044] Use the time-domain feature extraction result and the frequency-domain feature extraction result as the time-frequency domain features, use the time-frequency domain features of several simulated fault operation data as input data, and use the fault type corresponding to the simulated fault operation data as output data to construct several samples and obtain a sample set.

[0045] Preferably, the power grid model construction module is specifically configured to:

[0046] Obtain the historical data of the distributed new energy output of the low-voltage active distribution area;

[0047] Use the historical data as input and use a pre-constructed neural network prediction model to perform prediction to obtain the predicted value of the distributed new energy output.

[0048] Preferably, the fault types include metallic single-phase grounding fault, small-resistance single-phase grounding fault, high-resistance single-phase grounding fault, leakage and electric shock fault, and phase-to-phase short circuit fault.

[0049] Based on the same inventive concept, the present invention also provides a method for identifying grid faults in a low-voltage active power distribution area, including:

[0050] Obtain the real-time operation data of the grid in the low-voltage active power distribution area;

[0051] Using the real-time operation data of the grid as input, perform fault identification by using a pre-constructed fault identification model for the grid in the low-voltage active power distribution area to obtain the real-time fault type of the low-voltage active power distribution area;

[0052] The fault identification model for the grid in the low-voltage active power distribution area is constructed based on the method for constructing a fault identification model for the grid in the low-voltage active power distribution area as described above.

[0053] Based on the same inventive concept, the present invention also provides a system for identifying grid faults in a low-voltage active power distribution area, including:

[0054] An operation data acquisition module for obtaining the real-time operation data of the grid in the low-voltage active power distribution area;

[0055] A fault identification module for using the real-time operation data of the grid as input and performing fault identification by using a pre-constructed fault identification model for the grid in the low-voltage active power distribution area to obtain the real-time fault type of the low-voltage active power distribution area;

[0056] The fault identification model for the grid in the low-voltage active power distribution area is constructed based on the method for constructing a fault identification model for the grid in the low-voltage active power distribution area as described above.

[0057] Based on the same inventive concept, the present invention also provides a computer device, including: one or more processors;

[0058] A memory for storing one or more programs;

[0059] When the one or more programs are executed by the one or more processors, a method for constructing a fault identification model for the grid in the low-voltage active power distribution area as described above, or a method for identifying grid faults in the low-voltage active power distribution area as described above is implemented.

[0060] Based on the same inventive concept, the present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed, a method for constructing a fault identification model for the grid in the low-voltage active power distribution area as described above, or a method for identifying grid faults in the low-voltage active power distribution area as described above is implemented.

[0061] Compared with the closest prior art, the beneficial effects of the present invention are as follows:

[0062] The present invention provides a method and system for constructing a low-voltage active distribution network fault identification model, including: based on the typical topology of the low-voltage active distribution network, connecting the predicted values of distributed new energy output obtained in advance in the low-voltage active distribution network to the typical topology to construct a low-voltage active distribution network model; performing power grid fault simulation on the low-voltage active distribution network model to obtain simulation fault operation data of the power grid where the low-voltage active distribution network is located under different set fault types; constructing a sample set based on the time-frequency domain characteristics of the simulation fault operation data and the fault types corresponding to the simulation fault operation data; based on the sample set, using an extreme learning machine optimized by a particle swarm algorithm to train a single-hidden layer feedforward neural network to obtain a trained low-voltage active distribution network fault identification model; by connecting the predicted values of distributed new energy output to the typical topology, the method and system realize the establishment of a refined low-voltage active distribution network model, enabling the simulation fault operation data to consider the randomness and uncertainty of distributed new energy output. Using this simulation fault operation data for training the fault identification model can make the fault identification model more suitable for power grid fault identification with distributed new energy access, and the identification result is more accurate; in addition, by optimizing the extreme learning machine through the particle swarm optimization algorithm and then training the single-hidden layer feedforward neural network, the training speed and identification performance of the model can be improved.

[0063] The present invention also provides a method and system for identifying faults in a low-voltage active distribution network, including: obtaining the real-time operation data of the power grid in the low-voltage active distribution network; using the real-time operation data of the power grid as input, and performing fault identification using a pre-constructed low-voltage active distribution network fault identification model to obtain the real-time fault type of the low-voltage active distribution network; the low-voltage active distribution network fault identification model is constructed based on the method for constructing a low-voltage active distribution network fault identification model as described above; by using the optimized and constructed low-voltage active distribution network fault identification model, the method and system can identify different fault types while considering the randomness and uncertainty of distributed new energy output, greatly improving the accuracy of fault identification of the power grid after the access of distributed new energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a schematic flow chart of a method for constructing a low-voltage active distribution network fault identification model provided by the present invention;

[0065] Figure 2 It is a schematic diagram of a distributed photovoltaic output access system solution provided by the present invention;

[0066] Figure 3 It is a schematic diagram of the typical topology structure of a low-voltage active distribution network provided by the present invention;

[0067] Figure 4 It is a schematic diagram of a radial power grid model of a low-voltage active distribution network provided by the present invention;

[0068] Figure 5 Schematic diagram of the transmission path of the low-voltage active substation area fault detection data provided by the present invention;

[0069] Figure 6 Schematic diagram of the network architecture of the extreme learning machine provided by the present invention;

[0070] Figure 7 Schematic diagram of the construction process of the low-voltage active substation area power grid fault identification model provided by the present invention;

[0071] Figure 8 Schematic diagram of the fitness curve of the particle swarm optimization algorithm provided by the present invention;

[0072] Figure 9 Schematic diagram of the confusion matrix of the training set and the test set of the low-voltage active substation area power grid fault identification model provided by the present invention;

[0073] Figure 10 Schematic diagram of the structure of a low-voltage active substation area power grid fault identification model construction system provided by the present invention;

[0074] Figure 11 Schematic diagram of the process of a low-voltage active substation area power grid fault identification method provided by the present invention;

[0075] Figure 12 Schematic diagram of the structure of a low-voltage active substation area power grid fault identification system provided by the present invention;

[0076] Figure 13 Schematic diagram of the structure of an electronic device provided by the present invention. Specific implementation manners

[0077] The following further elaborates on the specific implementation manners of the present invention with reference to the accompanying drawings.

[0078] Example 1:

[0079] A method for constructing a low-voltage active substation area power grid fault identification model provided by the present invention, as Figure 1 shown, includes:

[0080] S1. Based on the typical topology of the low-voltage active substation area, connect the predicted distributed new energy output values in the pre-acquired low-voltage active substation area to the typical topology to construct a low-voltage active substation area power grid model;

[0081] S2. Perform power grid fault simulation on the low-voltage active substation area power grid model to obtain the simulated fault operation data of the power grid where the low-voltage active substation area is located under different set fault types;

[0082] S3. Construct a sample set based on the time-frequency domain characteristics of the simulated fault operation data and the corresponding fault types of the simulated fault operation data;

[0083] S4. Based on the sample set, use the extreme learning machine optimized by the particle swarm algorithm to train a single-hidden-layer feedforward neural network to obtain a trained low-voltage active distribution network fault identification model.

[0084] Considering that the existing low-voltage active distribution network fault identification technologies do not consider the impact of distributed new energy such as distributed photovoltaic on the low-voltage distribution network fault after access, nor do they consider the new risk states brought by the access of distributed new energy; this method realizes the establishment of a refined low-voltage active distribution network model by connecting the predicted output value of distributed new energy to the typical topology, enabling the simulated fault operation data to consider the randomness and uncertainty of the distributed new energy output. Using this simulated fault operation data for the training of the fault identification model can make the fault identification model more suitable for the fault identification of the power grid with distributed new energy access, and the identification result is more accurate; in addition, by optimizing the extreme learning machine through the particle swarm optimization algorithm and then training a single-hidden-layer feedforward neural network, the training speed and identification performance of the model can be improved.

[0085] Considering that the construction fineness of the typical topology of the low-voltage active distribution network will directly affect the construction accuracy of the low-voltage active distribution network model, therefore, taking the distributed new energy as distributed photovoltaic as an example, in the above S1, the refined establishment of the typical topology of the low-voltage active distribution network is carried out, and the establishment steps include:

[0086] According to the principle of accessing the user's internal power grid at the 380V voltage level, combined with the photovoltaic project plan provided by the user and the main wiring situation of the 10kV distribution room in the photovoltaic park, and referring to the typical design scheme XGF380-Z-Z1 of multi-point access in the "Typical Design of Distributed Photovoltaic Power Generation Access System of State Grid", adopt the Figure 2 access system scheme shown. This scheme is mainly applicable to photovoltaic power stations with spontaneous self-use / remaining power on the grid (accessing the user's power grid). The reference installed capacity of a single grid connection point is not more than 300kW, and three-phase access is adopted; for installed capacity of 8kW and below, single-phase access can be adopted. The access of distributed photovoltaic in this article considers the randomness and uncertainty of distributed photovoltaic output, and connects the predicted output value of distributed photovoltaic to the distribution area; specifically, as Figure 2As shown in the figure, the 380 / 220V distribution box and the 380 / 220 overhead line are connected to the 380 / 220V user distribution box and the 380 / 220 branch overhead line through the common connection point and the property right demarcation point CB4. The common connection point adopts the common connection points CB1, CB2 or CB3. The 380 / 220V user distribution box and the 380 / 220 branch overhead line supply power to the internal loads of the users. The predicted value of the distributed photovoltaic output is connected to the 380 / 220V branch distribution box and the 380 / 220 branch overhead line through the grid connection point CB5.

[0087] According to Figure 3 the typical topological structure of the low-voltage active distribution area, through on-site visits and investigations and the research on the wiring method of the low-voltage active distribution area, a low-voltage active distribution area power grid model as shown in Figure 4 the figure is established, specifically a radial power grid model of the low-voltage active distribution area. Combining Figure 3 and Figure 4 to explain the structure of the radial power grid model of the low-voltage active distribution area. Specifically, this model is connected to a 10kV / 0.4kV distribution transformer through a branch line by an infinite power grid G on the feeder side, that is Figure 3 the step-down transformer in the typical topological structure. The grounding method of the step-down transformer is direct grounding of the neutral point, and the capacity of the step-down transformer is 400kV·A. This step-down transformer has a total of three outgoing lines. The voltage level of the low-voltage lines of each outgoing line is 380V. Each outgoing line is radially connected to different user loads through a branch box. The loads consider user single-phase loads and three-phase loads, simulating different household users. Distributed photovoltaics are connected to users according to the typical design scheme, considering the three-phase five-wire low-voltage active distribution area topology. The line model is LGJ-25mm2, with a total of three outgoing lines. The active power consumed by each user ranges from 8kW to 12kW. The overall power factor of the distribution area is set to 0.92. The distance between every two households ranges from 0.02 to 0.04km. The impedance per unit length of the line is 1.131 + j0.393Ω / km.

[0088] Specifically, based on the established model, the model is trained and tested, and the batch simulation of faults is realized by the interaction between Matrix Laboratory Matlab and the simulation link Simulink. Write a Matlab custom script to call the Simulink automation library, abstract the model components and assign values to their parameters, traverse the parameter combinations through a loop structure, and finally generate and export the simulation data.

[0089] In addition, with the development of smart meters, the real-time data acquisition situation of the current low-voltage distribution network has been improved. By introducing artificial edge machine algorithms for calculation, processing the distribution network data nearby and diagnosing faults, the fault handling efficiency can be effectively improved. The transmission path of the fault detection data in the current low-voltage distribution network is as shown in Figure 5As shown in the figure, the data stream of the low-voltage line is collected by the smart meter, and the smart meter transmits the normal operation data or the data during the fault to the edge Internet of Things terminal equipped with the edge machine algorithm. The edge Internet of Things terminal transmits the data to the regional cloud master station via the Internet, processes the distribution network data nearby and conducts fault diagnosis, which can effectively improve the fault handling efficiency.

[0090] In this embodiment, the process of obtaining the predicted value of the distributed new energy output includes:

[0091] Obtain the historical data of the distributed new energy output of the low-voltage active area;

[0092] Taking the historical data as the input, use the pre-constructed neural network prediction model for prediction to obtain the predicted value of the distributed new energy output.

[0093] In a possible implementation manner, the construction process of the neural network prediction model includes:

[0094] Obtain the actual output values of the distributed new energy at each historical moment, divide the actual output values at each historical moment into a training set and a test set, set the initial learning rate and the number of hidden layers of the long short-term memory neural network, take the training set as the input, and take the actual output value at the next moment of the moment where the training set is located as the output, train the long short-term memory neural network, and use the stochastic optimization algorithm to iteratively update the parameters of the long short-term memory neural network until the loss function of the long short-term memory neural network converges;

[0095] Based on the test set, use the trained long short-term memory neural network for prediction to obtain the predicted output value at the next moment of the moment where the training set is located. Calculate the prediction result accuracy of the trained long short-term memory neural network based on the predicted output value at the next moment of the moment where the training set is located and the actual output value at the next moment of the moment where the training set is located. If the prediction result accuracy is less than the preset value, re-divide the training set and the test set for training until the prediction result accuracy of the trained long short-term memory neural network is not less than the preset value, and obtain the trained neural network prediction model.

[0096] In another possible implementation manner, the neural network prediction model can also be constructed based on neural networks such as recurrent neural networks or gated recurrent units.

[0097] In this embodiment, in the above S2, select one of the outgoing lines of the low-voltage topology of the active distribution network, set five types of faults as shown in Table 2, and respectively set the vector labels required for ELM training for each type of fault; among them, the fault types include metallic single-phase grounding fault, small-resistance single-phase grounding fault, high-resistance single-phase grounding fault, leakage and electric shock fault, and phase-to-phase short circuit fault.

[0098] By changing the fault sending location, the location of distributed power access, the distributed photovoltaic output considering randomness, and different fault transition resistances, different simulated fault operation data are obtained. Each set of simulated fault operation data corresponds to a fault type, and 120 samples are finally simulated for each fault type, with a total of 600 samples as the fault set.

[0099] Table 2 Fault Type Labels

[0100]

[0101] Considering that the dimension of the simulated fault operation data is relatively high, before introducing the machine learning algorithm, when constructing the sample set in the above S3, feature extraction is first performed on the simulated fault operation data, which can reveal the dynamic characteristics of the simulated fault operation data and improve the model training efficiency and prediction performance.

[0102] In this embodiment, when constructing the sample set in the above S3, it may include:

[0103] Perform set time-domain feature extraction on the simulated fault operation data to obtain the time-domain feature extraction result of the simulated fault operation data. The time-domain feature extraction result includes peak-to-peak value, rectified average value, variance, standard deviation, effective value, skewness, kurtosis, peak factor, and waveform factor;

[0104] Perform spectrum analysis on the simulated fault operation data to obtain the frequency components of the simulated fault operation data, and perform set frequency-domain feature extraction on the frequency components to obtain the frequency-domain feature extraction result of the simulated fault operation data. The frequency-domain feature extraction result includes center frequency, average frequency, root mean square frequency, and frequency standard deviation;

[0105] Take the time-domain feature extraction result and the frequency-domain feature extraction result as time-frequency domain features, use the time-frequency domain features of several sets of simulated fault operation data as input data, and use the fault type corresponding to the simulated fault operation data as output data to construct several samples and obtain the sample set.

[0106] In this paper, 9 time-domain statistical features are extracted, including 5 dimensional features and 4 dimensionless features. Subsequently, spectrum analysis is performed on the current signal of the fault phase to obtain the frequency components of the signal, and then 4 frequency-domain features of the signal are extracted, with a total of 13-dimensional time-frequency domain features for identification. The calculation formulas for feature extraction are shown in Table 1.

[0107] Table 1 Calculation Formulas for Thirteen-Dimensional Time-Frequency Domain Feature Extraction

[0108]

[0109]

[0110] Among them, x iThe current of the i-th data point representing the simulation fault operation data waveform, N is the total number of data points of the simulation fault operation data waveform, is the average current of the simulation fault operation data waveform, T4 is the standard deviation of the current of the simulation fault operation data waveform, f(k) is the frequency of the k-th frequency component of the current signal, K and n are both the total number of frequency components; s(k) is the amplitude of the k-th frequency component, F c is the center frequency.

[0111] In summary, this paper details the topological structure and hierarchy of the low-voltage active distribution network. Based on field research, a refined simulation model of the low-voltage active distribution network is established. The abnormal risk states and common faults of the low-voltage distribution network are simulated, and a fault simulation set is established. The simulation model is accurate and has a faster operation speed; common fault types of the low-voltage distribution network are batch-simulated, thirteen-dimensional time-frequency domain features are extracted from them, a sample set of fault features is established, and based on this, a data-driven fault identification model based on the extreme learning machine is established.

[0112] The Extreme Learning Machine (ELM) is a machine learning algorithm used to train a Single Hidden Layer Feedforward Neural Network (SLFN). Its basic principle is as follows:

[0113] Given the input samples {(E I ,t I ), I = 1, 2, …, J}, where J represents the number of samples in the sample set, the sample input E I ={e I1 ,e I2 ,…,e Im} T , the vector label t I ={t I1 ,t I2 ,…,t I5} T , e Im represents the m-th input vector of the i-th sample, that is, the m-th fault feature extracted, and t I1 -t I5 represents each element in the vector label of the i-th sample, that is, the corresponding fault type; for example, m = 13. The network structure of the extreme learning machine is as Figure 5 shown. D is the number of features. In this example, D = 13; M is the fault type. In this example, M = 5; h(x) is the hidden layer, L is the number of neurons in the hidden layer, and β is the hidden layer bias;

[0114] Considering that there is a certain randomness in initializing the parameters of the hidden layer nodes in ELM, which may affect the performance of the subsequent network, in the above S4, the particle swarm algorithm is used to optimize ELM, which can improve the network performance, enhance the generalization ability, reduce randomness, achieve fast convergence, and is easy to implement and has strong adaptability;

[0115] In this embodiment, when constructing the low-voltage active distribution network fault identification model in the above S4, it may include:

[0116] Based on the sample set, use the particle swarm algorithm to train the input layer and the hidden layer of the single-hidden-layer feedforward neural network to obtain the trained input layer and hidden layer;

[0117] Take the weights of the trained input layer and the biases of the hidden layer as the initial parameters of the extreme learning machine;

[0118] Based on the sample set, use the extreme learning machine to train the single-hidden-layer feedforward neural network to obtain the trained low-voltage active distribution network fault identification model.

[0119] In this embodiment, based on the sample set, using the particle swarm algorithm to train the input layer and the hidden layer of the single-hidden-layer feedforward neural network to obtain the trained input layer and hidden layer, including:

[0120] Initialize the parameters of the particle swarm algorithm, randomly generate the number of particles, positions and velocities of the particle swarm, and the position coordinate values of each particle represent a set of weights of the input layer and biases of the hidden layer of the single-hidden-layer feedforward neural network;

[0121] Based on the sample set, calculate the fitness value of the position of each particle according to the fitness function, determine the individual extreme value and the global extreme value of the particle; update the velocity and position of the particle, and obtain the new individual extreme value and global extreme value of the particle according to the fitness function; repeatedly iterate to update the velocity and position of the particle until the maximum number of iterations is reached or the change of the global extreme value is less than the set threshold, and take the position coordinate value of the particle corresponding to the global extreme value at this time as the final weights of the input layer and biases of the hidden layer of the single-hidden-layer feedforward neural network to complete the training of the input layer and the hidden layer.

[0122] In this embodiment, based on the sample set, using the extreme learning machine to train the single-hidden-layer feedforward neural network to obtain the trained low-voltage active distribution network fault identification model, including:

[0123] Divide the sample set into a training set and a test set according to a set ratio;

[0124] Based on the training set, the single-hidden-layer feedforward neural network is trained with the time-frequency domain features of the simulated fault operation data in the training set as the input and the corresponding fault types of the simulated fault operation data as the output. The extreme learning machine is used to calculate the output layer weights of the neural network prediction model, and the optimizer is used to optimize and update the network parameters of the single-hidden-layer feedforward neural network. The test set is used to evaluate the single-hidden-layer feedforward neural network during the training process until the loss function of the single-hidden-layer feedforward neural network converges, and a trained low-voltage active distribution network fault identification model is obtained.

[0125] Specifically, as Figure 7 shown, this paper uses the particle swarm optimization algorithm to optimize the initial weights and biases of the ELM, and the steps to construct the low-voltage active distribution network fault identification model are as follows:

[0126] (1) Define the ELM model parameters. Specifically, this paper takes the fault features after time-frequency domain feature extraction as the input, and sets the number of neurons in the input layer to 13. Combining with the empirical formula, the number of neurons in the hidden layer is set to 11. Taking the five fault types identified in this paper as the output, the number of neurons in the output layer is set to 5. The activation function of the hidden layer neurons is selected as the logistic function Sigmoid, and the output vector is processed by the softmax function to convert the value of each element in the output vector to between 0 and 1;

[0127] (2) Divide the training set and the test set, then input the fault feature vector data composed of waveform features, and define the vector labels of each grounding fault cause, that is, the fault type. The particle swarm optimization algorithm (Particle Swarm Optimization, PSO) is used to optimize the initial weights and biases to complete the training and optimization of the model; for example, randomly select 83% of each sample in the sample set as the training set, and the remaining 17% as the test set, and use the fault classification model for training and classification.

[0128] (3) The basic process of the particle swarm optimization algorithm is as follows:

[0129] ① Particle swarm initialization: Randomly generate a group of particles, and randomly set the initial velocity v 0 and the initial position x 0 of each particle.

[0130] ② Calculate the fitness value of each particle; at the same time, calculate the individual extreme value P best and the global extreme value G best of the particle, and record the corresponding particle position.

[0131] ③ Iterative update: Update the velocity and position of each particle according to equations (1) and (2). After each update, recalculate the fitness value, compare the fitness value of the updated particle with the fitness value at the historical optimal position to determine the optimal fitness value, and then assign a value to the ELM. The schematic diagram of the fitness curve of the particle swarm algorithm is as shown in Figure 8 Figure []. The abscissa represents the number of updates, and the ordinate represents the fitness value. As the number of updates increases, the fitness value gradually converges.

[0132] V ad h+1 = wV ad h + c 1 rand 1 (P bestd h - X ad h ) + c 2 rand 2 (G bestd h - X ad h ) (1)

[0133] X ad h+1 = X ad h + V ad h (2)

[0134] Where: V ad h , X ad h , P bestd h , G bestd h represent the components of the velocity, position vector, individual extreme value, and global extreme value of the ath particle in the d-dimensional space at the hth iteration, respectively. V ad h+1 , X ad h+1 represent the components of the velocity and position vector of the ath particle in the d-dimensional space at the (h + 1)th iteration, respectively. The learning factors c 1 , c 2 are both set to 1.5, the number of iterations is set to 1000, and w is the inertia factor; rand 1 and rand 2 are random constants within the range of [0 - 1].

[0135] (4) Use the test set data to evaluate the model performance.

[0136] Figure 9 is the confusion matrix of the fault classification results of the ELM model on the training set and the test set, and its classification accuracies are 93.6% and 89% respectively, as shown in Table 3 and Table 4, which proves that the data-driven active low-voltage active substation area fault extreme learning machine identification model proposed in this paper has good classification effects.

[0137] Table 3 Classification accuracy of the model under the training set

[0138]

[0139] Table 4 Classification accuracy of the model under the test set

[0140]

[0141] Considering that after a large number of distributed power sources are connected to the low-voltage distribution network, the operation risk of the distribution network is greatly challenged, especially the fault identification of grounding faults, short-circuit faults and leakage and electric shock faults is more affected. This application is based on the data collection of smart meters and the training of deep learning algorithms such as extreme learning machines to identify the fault types of low-voltage active substation areas with distributed photovoltaics connected, laying a foundation for the subsequent protection of power grid faults. Under the background of large-scale access of power sources and loads, different fault types are identified to improve the accuracy of identification. Specifically, this application establishes a refined low-voltage active substation area simulation topology model based on Matlab / Simulink, conducts simulation analysis considering key elements such as low-voltage distribution networks, user loads, and grid-connected inverters, obtains the current low-voltage topology data and types through visits and field investigations, establishes a refined low-voltage active substation area power grid simulation model, and considering the uncertainty and randomness of distributed photovoltaics, connects the predicted values of distributed new energy output to the typical topology to further improve the authenticity of fault simulation. Secondly, based on the fault data obtained from batch simulations and a large number of simulations, a machine learning algorithm, the extreme learning machine optimized by the particle swarm algorithm, is used to conduct batch training and testing on the fault types for fault identification to improve the accuracy of fault type identification. Finally, based on the established fault identification model, different extracted features are analyzed, the causality and correlation between multi-dimensional transient state features and faults and abnormal conditions are evaluated, and a method for identifying different fault types of active substation areas based on extreme learning machines is proposed. Experiments show that the identification accuracy of the data-driven fault identification model is 89%, and this method is effective and has strong generalization; future research will consider more ways of photovoltaic access and more common faults in low-voltage distribution networks to enrich the identification methods and databases.

[0142] Example 2:

[0143] Based on the same inventive concept, the present invention also provides a system for constructing a low-voltage active distribution network fault identification model, as Figure 10 shown, including:

[0144] A power grid model construction module, configured to access the predicted values of distributed new energy output in a low-voltage active distribution area into a typical topology based on the typical topology of the low-voltage active distribution area, and construct a low-voltage active distribution network model;

[0145] A simulation data acquisition module, configured to perform power grid fault simulation on the low-voltage active distribution network model, and obtain simulation fault operation data of the power grid where the low-voltage active distribution area is located under different set fault types;

[0146] A sample construction module, configured to construct a sample set based on the time-frequency domain characteristics of the simulation fault operation data and the fault types corresponding to the simulation fault operation data;

[0147] An identification model training module, configured to train a single-hidden-layer feedforward neural network based on the sample set by using an extreme learning machine optimized by a particle swarm algorithm, and obtain a trained low-voltage active distribution network fault identification model.

[0148] In this embodiment, the identification model training module includes:

[0149] A first sub-module, configured to train the input layer and the hidden layer of the single-hidden-layer feedforward neural network by using a particle swarm algorithm based on the sample set, and obtain a trained input layer and hidden layer;

[0150] A second sub-module, configured to use the weights of the trained input layer and the biases of the hidden layer as the initial parameters of the extreme learning machine;

[0151] A third sub-module, configured to train the single-hidden-layer feedforward neural network by using an extreme learning machine based on the sample set, and obtain a trained low-voltage active distribution network fault identification model.

[0152] In this embodiment, the first sub-module is specifically configured to:

[0153] Initialize the parameters of the particle swarm algorithm, randomly generate the number of particles, positions and velocities of the particle swarm, and the position coordinate values of each particle represent a set of weights of the input layer and biases of the hidden layer of the single-hidden-layer feedforward neural network;

[0154] Based on the sample set, calculate the fitness value of the position of each particle according to the fitness function, and determine the individual extreme value and the global extreme value of the particle; update the velocity and position of the particle, and obtain the new individual extreme value and the global extreme value of the particle according to the fitness function; repeatedly iterate to update the velocity and position of the particle until the maximum number of iterations is reached or the change in the global extreme value is less than the set threshold, and use the position coordinate value of the particle corresponding to the global extreme value at this time as the weights of the final input layer and the bias of the hidden layer of the single-hidden-layer feedforward neural network, thus completing the training of the input layer and the hidden layer.

[0155] In this embodiment, the third sub-module is specifically used for:

[0156] Divide the sample set into a training set and a test set according to a set ratio;

[0157] Based on the training set, use the time-frequency domain features of the simulated fault operation data in the training set as the input and the fault type corresponding to the simulated fault operation data as the output to train the single-hidden-layer feedforward neural network; calculate the output layer weights of the neural network prediction model using an extreme learning machine, and optimize and update the network parameters of the single-hidden-layer feedforward neural network using an optimizer; use the test set to evaluate the single-hidden-layer feedforward neural network during the training process until the loss function of the single-hidden-layer feedforward neural network converges, thus obtaining a trained low-voltage active distribution network fault identification model.

[0158] In this embodiment, the sample construction module is specifically used for:

[0159] Perform set time-domain feature extraction on the simulated fault operation data to obtain the time-domain feature extraction result of the simulated fault operation data. The time-domain feature extraction result includes peak-to-peak value, rectified average value, variance, standard deviation, effective value, skewness, kurtosis, peak factor, and waveform factor;

[0160] Perform spectrum analysis on the simulated fault operation data to obtain the frequency components of the simulated fault operation data, and perform set frequency-domain feature extraction on the frequency components to obtain the frequency-domain feature extraction result of the simulated fault operation data. The frequency-domain feature extraction result includes center frequency, average frequency, root mean square frequency, and frequency standard deviation;

[0161] Use the time-domain feature extraction result and the frequency-domain feature extraction result as time-frequency domain features, use the time-frequency domain features of several simulated fault operation data as input data, and use the fault type corresponding to the simulated fault operation data as output data to construct several samples and obtain a sample set.

[0162] In this embodiment, the power grid model construction module is specifically used for:

[0163] Obtain the historical data of the distributed new energy output of the low-voltage active distribution network area;

[0164] Using historical data as input, a pre - constructed neural network prediction model is employed for prediction to obtain the predicted values of distributed new - energy output.

[0165] In this embodiment, the fault types include metallic single - phase grounding fault, single - phase grounding fault with small resistance, single - phase grounding fault with high resistance, leakage and electric shock fault, and phase - to - phase short - circuit fault.

[0166] Embodiment 3:

[0167] Based on the same inventive concept, the present invention also provides a method for identifying faults in a low - voltage active distribution network area, as Figure 11 shown, including:

[0168] A1. Obtain the real - time operation data of the power grid in the low - voltage active distribution network area;

[0169] A2. Using the real - time operation data of the power grid as input, a pre - constructed fault - identification model for the low - voltage active distribution network area is employed for fault identification to obtain the real - time fault types in the low - voltage active distribution network area;

[0170] The fault - identification model for the low - voltage active distribution network area is constructed based on the fault - identification model construction method for the low - voltage active distribution network area in the above - mentioned embodiment.

[0171] By using the optimized - constructed fault - identification model for the low - voltage active distribution network area, different fault types are identified while considering the randomness and uncertainty of distributed new - energy output, greatly improving the fault - identification accuracy of the power grid after the access of distributed new - energy.

[0172] Embodiment 4:

[0173] Based on the same inventive concept, the present invention also provides a system for identifying faults in a low - voltage active distribution network area, as Figure 12 shown, including:

[0174] An operation - data acquisition module, configured to obtain the real - time operation data of the power grid in the low - voltage active distribution network area;

[0175] A fault - identification module, configured to use the real - time operation data of the power grid as input and employ a pre - constructed fault - identification model for the low - voltage active distribution network area to perform fault identification, obtaining the real - time fault types in the low - voltage active distribution network area;

[0176] The fault - identification model for the low - voltage active distribution network area is constructed based on the fault - identification model construction method for the low - voltage active distribution network area in the above - mentioned embodiment.

[0177] Embodiment 5

[0178] As Figure 13As shown, the present invention also provides an electronic device, which may be a computer device, a single-chip microcomputer device, a smart mobile device, etc. The electronic device in this embodiment may include a processor, a memory, a transceiver component, etc. The memory, the processor, and the transceiver component are connected by a bus; the memory can be used to store an execution program, and an exemplary execution program may include instructions; the processor is used to execute the instructions stored in the memory. The memory can also be used to store data, and this data can be called and / or modified when the instructions are executed.

[0179] The processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, and is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions in the storage medium to implement the corresponding method flow or corresponding function, so as to implement a method for constructing a low-voltage active substation grid fault identification model as in the above embodiment, or a method for identifying low-voltage active substation grid faults as in the above embodiment.

[0180] Embodiment 6

[0181] Based on the same inventive concept, the present invention also provides a readable storage medium, specifically an electronic device-readable storage medium (Memory). The electronic device-readable storage medium is a memory device in the electronic device, and is used to store programs and data. It can be understood that the storage medium here can include both the built-in storage medium in the electronic device, and of course can also include the extended storage medium supported by the electronic device. The storage medium provides a storage space, and this storage space stores the operating system of the terminal. And, one or more instructions suitable for being loaded and executed by the processor are also stored in this storage space. These instructions can be one or more execution programs (including program codes). It should be noted that the storage medium here can be a high-speed RAM memory, or a non-volatile memory, such as at least one disk memory. By the processor loading and executing one or more instructions stored in the storage medium, a method for constructing a low-voltage active substation grid fault identification model as in the above embodiment, or a method for identifying low-voltage active substation grid faults as in the above embodiment can be implemented.

[0182] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can 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.) that contain computer-usable program code.

[0183] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0184] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing devices to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0185] These computer program instructions can also be loaded onto a computer or other programmable data processing devices, such that a series of operation steps are executed on the computer or other programmable devices to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable devices provide steps for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0186] 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 the scope of its protection. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that after reading the present invention, various changes, modifications, or equivalent replacements can still be made to the specific implementation manners of the application. However, these changes, modifications, or equivalent replacements are all within the scope of the protection of the claims of the present invention.

Claims

1. A method for constructing a low-voltage active area power grid fault identification model, characterized in that: include: Based on the typical topology of the low-voltage active area, the pre-acquired predicted value of the distributed new energy output in the low-voltage active area is connected to the typical topology to construct a low-voltage active area power grid model; Performing a power grid fault simulation on the low-voltage active substation power grid model to obtain simulated fault operation data of the power grid where the low-voltage active substation is located under different set fault types; Constructing a sample set based on the time-frequency domain characteristics of the simulated fault operation data and the fault type corresponding to the simulated fault operation data; Based on the sample set, an extreme learning machine optimized by a particle swarm algorithm is used to train a single hidden layer feedforward neural network to obtain a trained low-voltage active substation area power grid fault identification model.

2. The method according to claim 1, characterized in that Based on the sample set, an extreme learning machine optimized by a particle swarm algorithm is used to train a single hidden layer feedforward neural network to obtain a trained low-voltage active area power grid fault identification model, including: Based on the sample set, a particle swarm algorithm is used to train the input layer and the hidden layer of the single hidden layer feedforward neural network to obtain a trained input layer and a hidden layer; Using the trained weights of the input layer and the bias of the hidden layer as initial parameters of the extreme learning machine; Based on the sample set, the single hidden layer feedforward neural network is trained using the extreme learning machine to obtain a trained low voltage active substation area power grid fault identification model.

3. The method according to claim 2, characterized in that The method of training the input layer and the hidden layer of the single hidden layer feedforward neural network based on the sample set using a particle swarm algorithm to obtain the trained input layer and the hidden layer includes: Initializing the parameters of the particle swarm algorithm, randomly generating the number, position and speed of particles of the particle swarm, wherein the position coordinate value of each particle represents a set of input layer weights and a hidden layer bias of the single hidden layer feedforward neural network; Based on the sample set, the fitness value of the position of each particle is calculated according to the fitness function, and the individual extreme value and the global extreme value of the particle are determined; the speed and position of the particle are updated, and the new individual extreme value and the global extreme value of the particle are obtained according to the fitness function; the speed and position of the particle are updated repeatedly until the maximum number of iterations is reached or the change of the global extreme value is less than the set threshold value, and the position coordinate value of the particle corresponding to the global extreme value at this time is used as the weight of the final input layer of the single hidden layer feedforward neural network and the bias of the hidden layer, so as to complete the training of the input layer and the hidden layer.

4. The method according to claim 2 or 3, characterized in that Based on the sample set, the single hidden layer feedforward neural network is trained by the extreme learning machine to obtain a trained low voltage active area power grid fault identification model, including: Dividing the sample set into a training set and a test set according to a set ratio; Based on the training set, the single hidden layer feedforward neural network is trained by taking the time-frequency domain characteristics of the simulated fault operation data in the training set as input and taking the fault type corresponding to the simulated fault operation data as output; the extreme learning machine is used to calculate the output layer weights of the neural network prediction model, and the optimizer is used to optimize and update the network parameters of the single hidden layer feedforward neural network; the test set is used to evaluate the single hidden layer feedforward neural network in the training process until the loss function of the single hidden layer feedforward neural network converges, thereby obtaining a trained low-voltage active substation power grid fault identification model.

5. The method according to any one of claims 1 to 3, characterized in that: The constructing of a sample set based on the time-frequency domain features of the simulated fault operation data and the fault type corresponding to the simulated fault operation data comprises: Performing a set time domain feature extraction on the simulated fault operation data to obtain a time domain feature extraction result of the simulated fault operation data, wherein the time domain feature extraction result includes a peak-to-peak value, a rectified mean value, a variance, a standard deviation, an effective value, a skewness, a kurtosis, a peak factor, and a waveform factor; Performing spectrum analysis on the simulated fault operation data to obtain frequency components of the simulated fault operation data, performing set frequency domain feature extraction on the frequency components to obtain frequency domain feature extraction results of the simulated fault operation data, wherein the frequency domain feature extraction results include centroid frequency, average frequency, root mean square frequency and frequency standard deviation; The time domain feature extraction result and the frequency domain feature extraction result are used as the time-frequency domain features, the time-frequency domain features of several simulated fault operation data are used as input data, and the fault type corresponding to the simulated fault operation data is used as output data to construct several samples and obtain a sample set.

6. The method according to any one of claims 1 to 3, characterized in that: The process of obtaining the predicted value of the distributed new energy output includes: Obtaining historical data of distributed renewable energy output of the low-voltage active station area; The historical data is used as input and a pre-built neural network prediction model is used to perform prediction to obtain the distributed renewable energy output prediction value.

7. The method according to any one of claims 1 to 3, characterized in that: The fault types include metallic single-phase grounding fault, low-resistance single-phase grounding fault, high-resistance single-phase grounding fault, leakage electric shock fault and phase-to-phase short circuit fault.

8. A low voltage active area power grid fault identification model construction system, characterized in that: include: A power grid model building module, which is used to connect the pre-acquired distributed renewable energy output prediction value in the low-voltage active area to the typical topology based on the typical topology of the low-voltage active area to build a low-voltage active area power grid model; A simulation data acquisition module, used to perform a power grid fault simulation on the low-voltage active substation power grid model, and obtain simulated fault operation data of the power grid where the low-voltage active substation is located under different set fault types; A sample construction module, used to construct a sample set based on the time-frequency domain characteristics of the simulated fault operation data and the fault type corresponding to the simulated fault operation data; The identification model training module is used to train a single hidden layer feedforward neural network based on the sample set using an extreme learning machine optimized by a particle swarm algorithm to obtain a trained low-voltage active substation area power grid fault identification model.

9. The system according to claim 8, characterized in that The recognition model training module includes: The first submodule is used to train the input layer and the hidden layer of the single hidden layer feedforward neural network based on the sample set by using a particle swarm algorithm to obtain the trained input layer and the hidden layer; The second submodule is used to use the trained weights of the input layer and the bias of the hidden layer as initial parameters of the extreme learning machine; The third submodule is used to train the single hidden layer feedforward neural network based on the sample set using the extreme learning machine to obtain a trained low-voltage active substation area power grid fault identification model.

10. The system according to claim 9, characterized in that The first submodule is specifically used for: Initializing the parameters of the particle swarm algorithm, randomly generating the number, position and speed of particles of the particle swarm, wherein the position coordinate value of each particle represents a set of input layer weights and a hidden layer bias of the single hidden layer feedforward neural network; Based on the sample set, the fitness value of the position of each particle is calculated according to the fitness function, and the individual extreme value and the global extreme value of the particle are determined; The speed and position of the particle are updated, and a new individual extreme value and a global extreme value of the particle are obtained according to the fitness function; the speed and position of the particle are updated repeatedly until the maximum number of iterations is reached or the change of the global extreme value is less than the set threshold value, and the position coordinate value of the particle corresponding to the global extreme value at this time is used as the weight of the final input layer of the single hidden layer feedforward neural network and the bias of the hidden layer, so as to complete the training of the input layer and the hidden layer.

11. The system according to claim 9 or 10, characterized in that The third submodule is specifically used for: Dividing the sample set into a training set and a test set according to a set ratio; Based on the training set, the single hidden layer feedforward neural network is trained by taking the time-frequency domain characteristics of the simulated fault operation data in the training set as input and taking the fault type corresponding to the simulated fault operation data as output; the extreme learning machine is used to calculate the output layer weights of the neural network prediction model, and the optimizer is used to optimize and update the network parameters of the single hidden layer feedforward neural network; the test set is used to evaluate the single hidden layer feedforward neural network in the training process until the loss function of the single hidden layer feedforward neural network converges, thereby obtaining a trained low-voltage active substation power grid fault identification model.

12. The system according to any one of claims 8 to 10, characterized in that: The sample construction module is specifically used for: Performing a set time domain feature extraction on the simulated fault operation data to obtain a time domain feature extraction result of the simulated fault operation data, wherein the time domain feature extraction result includes a peak-to-peak value, a rectified mean value, a variance, a standard deviation, an effective value, a skewness, a kurtosis, a peak factor, and a waveform factor; Performing spectrum analysis on the simulated fault operation data to obtain frequency components of the simulated fault operation data, performing set frequency domain feature extraction on the frequency components to obtain frequency domain feature extraction results of the simulated fault operation data, wherein the frequency domain feature extraction results include centroid frequency, average frequency, root mean square frequency and frequency standard deviation; The time domain feature extraction result and the frequency domain feature extraction result are used as the time-frequency domain features, the time-frequency domain features of several simulated fault operation data are used as input data, and the fault type corresponding to the simulated fault operation data is used as output data to construct several samples and obtain a sample set.

13. The system according to any one of claims 8 to 10, characterized in that: The power grid model building module is specifically used for: Obtaining historical data of distributed renewable energy output of the low-voltage active station area; The historical data is used as input and a pre-built neural network prediction model is used to perform prediction to obtain the distributed renewable energy output prediction value.

14. The system according to any one of claims 8 to 10, characterized in that: The fault types include metallic single-phase grounding fault, low-resistance single-phase grounding fault, high-resistance single-phase grounding fault, leakage electric shock fault and phase-to-phase short circuit fault.

15. A method for identifying faults in a low voltage active area power grid, characterized in that: include: Obtain real-time operation data of the power grid in low-voltage active substations; Taking the real-time operation data of the power grid as input, a pre-built low-voltage active area power grid fault identification model is used to perform fault identification to obtain the real-time fault type of the low-voltage active area; The low-voltage active substation area power grid fault identification model is constructed based on the low-voltage active substation area power grid fault identification model construction method as described in any one of claims 1-7.

16. A low voltage active area power grid fault identification system, characterized in that: include: Operation data acquisition module, used to obtain real-time operation data of the power grid in the low-voltage active substation area; A fault identification module is used to use the real-time operation data of the power grid as input, and adopt a pre-built low-voltage active area power grid fault identification model to perform fault identification, so as to obtain the real-time fault type of the low-voltage active area; The low-voltage active substation area power grid fault identification model is constructed based on the low-voltage active substation area power grid fault identification model construction method as described in any one of claims 1-7.

17. An electronic device, characterized in that: include: at least one processor and memory; The memory and the processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, a low-voltage active substation area power grid fault identification model construction method as described in any one of claims 1 to 8, or a low-voltage active substation area power grid fault identification method as described in claim 15 is implemented.

18. A readable storage medium, characterized in that: An execution program is stored thereon, and when the execution program is executed, a low-voltage active substation area power grid fault identification model construction method as described in any one of claims 1 to 8, or a low-voltage active substation area power grid fault identification method as described in claim 15 is implemented.

Images