Single-phase ground fault determination method for small current grounding system based on bi-lstm

Abstract

The application belongs to the field of power system fault diagnosis, and discloses a single-phase grounding fault determination method for a small current grounding system based on Bi-LSTM, which comprises the following steps: analyzing a model of a power system and actual dispatching data, building a computer numerical simulation model through the analyzed data, labeling fault data and non-fault data in the simulation data and the actual dispatching data, normalizing simulation and real-time dispatching characteristic data, making a training set and a test set of a Bi-LSTM model, and designing a fault line selection model based on Bi-LSTM and training and testing of the model. In the case of single-phase grounding, the application can quickly determine a fault line according to discrete dispatching data; compared with existing zero sequence current identification methods, the application does not require zero sequence current data, and has a higher identification rate in the case of neutral point grounding through an arc suppression coil or a fault point with high transition resistance, thereby realizing simple and efficient small current grounding fault line selection.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of power system fault diagnosis, and particularly relates to a single-phase grounding fault determination method for a small-current grounding system based on Bi-LSTM. BACKGROUND

[0002] With the increase of power consumption in urban development, in order to ensure that the line can still operate when a fault occurs, the small-current grounding mode is currently used in the distribution network of the power system. The advantage of the small-current grounding system is that when a grounding fault occurs in the power transmission line, the faulty line can still operate for two hours, which ensures the power supply reliability and increases the difficulty of fault line selection of the power system due to the unobvious current change characteristics.

[0003] The traditional processing method for the small-current grounding fault is the manual line pulling method, that is, when the busbar insulation monitoring device sends a single-phase grounding fault signal, the on-duty personnel adopts the sequential line pulling method to find the faulty line, and the faulty line is cut off after the load is transferred. This method causes some non-faulty line users to also be short-time power off, reduces the power supply reliability, prolongs the time of the system operating with single-phase grounding, and increases the possibility of expanding the fault and misoperation. SUMMARY

[0004] The purpose of the application is to provide a single-phase grounding fault determination method for a small-current grounding system based on Bi-LSTM, which can effectively extract larger range of features in the time series discrete scheduling data to improve the fault diagnosis rate and realize simple and efficient small-current grounding fault line selection.

[0005] To achieve the above purpose, the application provides a single-phase grounding fault determination method for a small-current grounding system based on Bi-LSTM, which comprises the following steps:

[0006] S1: analyzing the CIM model of the power system to obtain the actual power system topology structure diagram;

[0007] S2: building a power system numerical simulation model by using the Matlab software according to the power system topology structure diagram;

[0008] S3: collecting data, including real scheduling data of the power grid and computer simulation data, to form original data and perform normalization processing; when the fault data in the real scheduling data of the power grid is insufficient, simulation data is obtained by using the Matlab software simulation, and the simulation model is constructed based on the actual power system topology structure diagram, and the simulation data and the real scheduling data of the power grid have consistency;

[0009] S4: make a model training data set, and classify and label; the model training data set includes a training set and a test set, due to the adoption of computer simulation technology, the method can efficiently expand fault and non-fault data, so that a large amount of data can be obtained to train the small current grounding fault line selection model based on Bi-LSTM;

[0010] S5: establish a deep learning Bi-LSTM model, including a small current grounding fault line selection model based on Bi-LSTM;

[0011] The small current grounding fault line selection model based on Bi-LSTM contains 6 layers, including an input layer, four hidden layers, and an output layer, the input layer is determined by the characteristic quantity, the hidden layer is an LSTM module, the output layer is a Dense module, and the output represents the result of determining whether the line is faulty;

[0012] S6: train the Bi-LSTM neural network and test, including training the small current grounding fault line selection model based on Bi-LSTM by using the characteristic quantity of the sample data, then obtaining the single-phase grounding fault determination result by the trained small current grounding fault line selection model based on Bi-LSTM, and completing fault line selection.

[0013] As a further technical improvement, the computer simulation data includes voltage effective value data, current effective value data, and reactive power data, and the voltage effective value data, the current effective value data, and the reactive power data are used as characteristic quantities.

[0014] As a further technical improvement, in step S1, the power system CIM model is parsed by Java or Python to obtain an actual power system topology graph.

[0015] As a further technical improvement, in step S2, the Matlab simulation model is built according to the connection sequence of the transformer winding, bus, disconnecting switch, circuit breaker, line and load in the power system topology graph.

[0016] As a further technical improvement, in step S3, the sampling time of the dispatching data in the power system is 3 seconds once, which is a discrete effective value, and the sampling time in the Matlab software is set to 3 seconds once.

[0017] As a further technical improvement, in step S4, a part of the simulation data and the real dispatching data of the power grid are selected as the training set, and the other part of the real dispatching data of the power grid is selected as the test set.

[0018] As a further technical improvement, the small current grounding system single-phase grounding fault determination method based on Bi-LSTM is applied to the fault line selection of the urban distribution network.

[0019] Compared with the prior art, the present invention has the following beneficial technical effects:

[0020] 1. This invention provides a method for determining single-phase grounding faults in low-current grounding systems based on Bi-LSTM. It designs a low-current grounding fault line selection model based on Bi-LSTM and trains and tests the model. Based on the extracted feature quantities, it can effectively extract a wider range of features from time-series discrete scheduling data, thereby improving the fault diagnosis rate.

[0021] 2. This invention supplements the lack of actual fault data with Matlab simulation data, and uses a deep learning-based Bi-LSTM-based low-current grounding fault selection model to achieve low-current grounding fault selection. It does not require zero-sequence current data and has a higher recognition rate when the neutral point is grounded through an arc suppression coil or the fault point has a high transition resistance, thus realizing simple, efficient and automated intelligent fault selection.

[0022] 3. This invention employs computer simulation technology, which can efficiently expand fault and non-fault data, thus obtaining massive amounts of data to train the Bi-LSTM-based low-current grounding fault line selection model. Attached Figure Description

[0023] To more clearly illustrate the embodiments of the present invention, the accompanying drawings required for the embodiments will be briefly described below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is an overall flowchart of the embodiment;

[0025] Figure 2 This is a schematic diagram of the structure of the low-current grounding fault selection model based on Bi-LSTM in the embodiment. Detailed Implementation

[0026] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0027] The following specific examples illustrate the implementation of this disclosure. Those skilled in the art can easily understand other advantages and effects of this disclosure from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. This disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this disclosure. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0028] Example

[0029] As attached Figures 1-2 As shown, this embodiment provides a method for determining single-phase grounding faults in a low-current grounding system based on Bi-LSTM, including the following steps: S1: Analyze the CIM model of the power system to obtain the actual power system topology diagram;

[0030] The analytical power system CIM model involves obtaining the connection sequence and method of each feeder device based on the urban power distribution network CIM model, thereby deriving the line model. This step primarily uses Python or Java to parse the power system CIM model, drawing the power system topology diagram through the connection IDs of several components to make the Matlab simulation more realistic. Key components include: transformer windings, buses, circuit breakers, disconnectors, lines, and loads. The form of the transformer windings and the location of the circuit breakers and disconnectors affect the data characteristics of the lines. Buses, lines, and loads are mainly used to measure relevant data of the power system feeders.

[0031] S2: Based on the power system topology diagram, a numerical simulation model of the power system is built using Matlab software. This model simulates normal power system operation, three-phase grounding faults on the same line, load switching, line disconnection, and data transmission failures from the substation to the dispatching station. Simultaneously, a dynamic load module is used in the load module, employing a sine function to simulate load increases and decreases, corresponding to the daily changes in user electricity consumption. The load function is as follows:

[0032] sin(′Amplitude′,1e6,′Frequency′,0.01,′Phase′,1)

[0033] Where Amplitude represents the amplitude of the sine function, and its value is 1 × 10⁻⁶. 6Frequency represents the frequency of this sine function, with a value of 0.01, and Phase represents the phase, with a value of 1.

[0034] Simultaneously, the same topology should be used to build a non-fault model, ensuring a 1:1 ratio of non-fault data to fault data. Because faults are extremely rare in actual power system operation, parameter adjustments are necessary in the simulation to simulate as many different operating conditions as possible. The magnitude of the capacitor current is the primary cause of voltage and current changes in faulty lines, as shown in the formula:

[0035]

[0036] Where k is the grounding point, U A U is the phase voltage of phase A. B For phase B voltage, U C Let UAK be the phase voltage of phase C, and UAK be the line voltage at the short-circuit point of phase A. Since the voltage at the grounding point is 0, the voltages of phases B and C change from phase voltages to line voltages, thus increasing the voltage. The voltage is several times higher. A crucial factor affecting voltage magnitude is the line's capacitance to ground; the larger the capacitance, the greater the current flowing through the circuit after a short circuit. Therefore, the capacitance ranges for overhead lines are selected as follows: positive sequence: 0.0096-0.015uf / km, zero sequence: 0.003-0.009uf / km; and for cables, positive sequence: 0.15-0.75uf / km, zero sequence: 0.09-0.45uf / km.

[0037] S3: Data acquisition is performed, including real power grid dispatch data and computer simulation data, forming raw data, which is then normalized. When the fault data in the real power grid dispatch data is insufficient, simulation data is obtained through Matlab software simulation. The simulation model is constructed based on the real power system topology diagram, and the simulation data and the real power grid dispatch data are consistent. The sampling time of the dispatch data in the power system is once every 3 seconds, and its current and voltage values ​​are valid values, which are discrete valid values. The sampling time in the Matlab software is set to once every 3 seconds.

[0038] After supplementing the real data with simulation data, all data are normalized using the following formula:

[0039]

[0040] This method achieves proportional scaling of the original data, where X norm The data is normalized, and X is the original data. max X min These are the maximum and minimum values ​​of the original dataset, respectively.

[0041] The voltage, current, and reactive power of each sample constitute a 9-dimensional array, with each of the voltage, current, and reactive power having a three-dimensional array. Each dimension of data represents the data of each phase.

[0042] S4: Create a model training dataset and classify and label it; the model training dataset includes a training set and a test set. Due to the use of computer simulation technology, this method can efficiently expand fault and non-fault data, thus obtaining massive amounts of data to train the Bi-LSTM-based low-current grounding fault selection model.

[0043] S5: Establish a deep learning Bi-LSTM model, including a Bi-LSTM-based low-current grounding fault line selection model; use the Bi-LSTM (Long Short-Term Memory) and Dense (fully connected layer) modules in the PyTorch framework to build the Bi-LSTM-based low-current grounding fault line selection model.

[0044] The Bi-LSTM-based low-current grounding fault selection model comprises six layers: an input layer, four hidden layers, and an output layer. The input layer is determined by a feature quantity and has nine neurons. The hidden layers are LSTM modules with a size of 120. The output layer is a Dense module with two neurons, and the output represents the result of determining whether the line is faulty. Preferably, the activation function of the hidden layers is tanh (hyperbolic tangent function), and the activation function of the output layer is sigmoid (S-shaped function).

[0045] Therefore, the loss function is the binary classification loss function "binary_crossentropy", and the corresponding formula is:

[0046]

[0047] In the formula, error represents the loss value, and y represents the actual value. This represents the predicted value.

[0048] To prevent overfitting, a Dropout layer and an EarlyStopping function were added. The keep_prob parameter in dropout was set to 0.5, and the patience parameter in EarlyStopping (the number of periods without improvement before training stops) was set to 30. The number of iterations was set to 200, and the learning rate was 0.001.

[0049] The above model structure design can prevent overfitting to a certain extent and reduce running time while ensuring accuracy. The parameter settings are determined according to this embodiment (parameters obtained by tuning with accuracy as the evaluation metric) to ensure the accuracy of the LSTM fault line selection model. Four hidden layers are set to better fit the nonlinear relationship between input and output.

[0050] See Figure 2 This is an LSTM fault line selection model built based on Keras encapsulation modules. The input layer's x1 to x9 represent the selected feature quantities, namely the values ​​of three-phase current, voltage, and reactive power. The numbers in each encapsulation module indicate the number of neurons in that module. The arrows between two modules indicate the input-output relationship. Y is the final output, representing the result of determining whether the line is faulty. The input layer features x1, x2, x3, ..., x7, x8, x9 are connected to the Dropout layer, the Dropout layer's output is connected to the Dense module of the output layer, and the output of the Dense module of the output layer is the final output Y.

[0051] S6: Train and test the Bi-LSTM neural network, including training a Bi-LSTM-based low-current grounding fault selection model using the feature values ​​of the sample data, and then obtaining the single-phase grounding fault determination result through the trained Bi-LSTM-based low-current grounding fault selection model to complete the fault selection.

[0052] The computer simulation data includes effective voltage data, effective current data, and reactive power data, which are used as characteristic quantities.

[0053] In step S2, the Matlab simulation models are built according to the connection sequence of transformer windings, busbars, disconnecting switches, circuit breakers, lines and loads in the power system topology diagram.

[0054] In step S4, a portion of the simulation data and the actual power grid dispatch data are selected as the training set, and another portion of the actual power grid dispatch data is selected as the test set.

[0055] The single-phase grounding fault determination method for low-current grounding systems based on Bi-LSTM described in this embodiment is applied to fault line selection in urban power distribution networks.

[0056] This embodiment uses voltage, current, and reactive power. Experimental results show that after outlier handling and parameter tuning of the model, the accuracy can reach 90%.

[0057] This embodiment is based on power system dispatch data and Matlab simulation data, which can be divided into fault and non-fault data. The data is substituted into a pre-built Bi-LSTM network based on the nine-dimensional features of three-phase voltage, current, and reactive power to obtain the trained model. This model can be used to identify single-phase grounding faults occurring during power system operation with high accuracy.

[0058] Those skilled in the art will readily understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. The above embodiments are merely illustrative of the technical solutions of the present invention, and not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for determining single-phase grounding faults in a low-current grounding system based on Bi-LSTM, characterized in that, Includes the following steps: S1: Analyze the CIM model of the power system to obtain the actual power system topology diagram; S2: Based on the power system topology diagram, a numerical simulation model of the power system is built using Matlab software; This simulation addresses scenarios such as normal power system operation, three-phase grounding faults on the same line, load switching, line disconnection, and data transmission failure from the substation to the dispatching station. The load module utilizes a dynamic load module with a sine function to simulate load increases and decreases, corresponding to the daily variation in user electricity consumption. The load function is as follows: sin(′Amplitude ′ ,1e6, ′ Frequency ′ ,0.01, ′ Phase ′ ,1) Where Amplitude represents the amplitude of the sine function, and its value is 1 × 10⁻⁶. 6 Frequency represents the frequency of this sine function, with a value of 0.01, and Phase represents the phase, with a value of 1. S3: Data acquisition is performed, including real power grid dispatch data and computer simulation data, which constitute the raw data and are then normalized. When the fault data in the real power grid dispatch data is insufficient, simulation data is obtained through Matlab software simulation. The simulation model is constructed based on the real power system topology diagram, and the simulation data and the real power grid dispatch data are consistent. S4: Create a model training dataset and classify and label it; the model training dataset includes a training set and a test set; S5: Establish deep learning Bi-LSTM models, including a Bi-LSTM-based low-current grounding fault line selection model; The low-current grounding fault selection model based on Bi-LSTM includes an input layer, four hidden layers, and an output layer. The input layer is determined by a feature quantity, the hidden layer is an LSTM module, the output layer is a Dense module, and the output represents the result of determining whether the line is faulty. S6: Train and test the Bi-LSTM neural network, including training a Bi-LSTM-based low-current grounding fault selection model using the feature values ​​of the sample data, and then obtaining the single-phase grounding fault determination result through the trained Bi-LSTM-based low-current grounding fault selection model to complete the fault selection.

2. The method for determining single-phase grounding faults in a low-current grounding system based on Bi-LSTM according to claim 1, characterized in that, The computer simulation data includes effective voltage data, effective current data, and reactive power data, which are used as characteristic quantities.

3. The method for determining single-phase grounding faults in a low-current grounding system based on Bi-LSTM according to claim 1, characterized in that, In step S1, the power system CIM model is parsed using Java or Python to obtain the actual power system topology diagram.

4. The method for determining single-phase grounding faults in a low-current grounding system based on Bi-LSTM according to claim 1, characterized in that, In step S2, the Matlab simulation models are built according to the connection sequence of transformer windings, busbars, disconnecting switches, circuit breakers, lines and loads in the power system topology diagram.

5. The method for determining single-phase grounding faults in a low-current grounding system based on Bi-LSTM according to claim 1, characterized in that, In step S3, the sampling time of the dispatch data in the power system is once every 3 seconds, and it is a discrete effective value. The sampling time is set to once every 3 seconds in the Matlab software.

6. The method for determining single-phase grounding faults in a low-current grounding system based on Bi-LSTM according to claim 1, characterized in that, In step S4, a portion of the simulation data and the actual power grid dispatch data are selected as the training set, and another portion of the actual power grid dispatch data is selected as the test set.

7. The method for determining single-phase grounding faults in a low-current grounding system based on Bi-LSTM according to any one of claims 1-6, characterized in that, Fault location is applied to urban power distribution networks.

Images