Generator stator single-phase earth fault positioning method and system based on data driving

Abstract

The embodiment of the invention provides a generator stator single-phase earth fault positioning method and system based on data driving. The method comprises the following steps: constructing a generator stator single-phase earth fault electromagnetic transient model; transition resistance of a fault point of the generator stator single-phase earth fault electromagnetic transient model is set, and a training set and a test set of fault data are generated; an MLP neural network model is constructed, and the MLP neural network model comprises an MLP fault positioning model, an MLP transition resistance calculation model and an MLP fault branch discrimination model; training an MLP neural network model according to the training set of the fault data; and inputting new data during a fault into the MLP neural network model, and outputting a fault position, transition resistance and a fault branch. According to the method, the fault position can be accurately positioned, the transition resistance is effectively calculated, and the fault branch is reliably judged.

Description

Technical Field

[0001] The embodiments of the present invention belong to the technical field of generator faults, and specifically relate to a method and system for locating a generator stator single-phase grounding fault based on data driving. Background Art

[0002] The single-phase grounding fault of the generator stator is a frequent fault type during the operation of the generator, and it is very easy to evolve into a more serious inter-turn or inter-phase short circuit fault. As the capacity of the generator set increases, the capacitance of the stator winding to the ground continues to increase. Once a grounding fault occurs, the harm to the generator becomes more serious. In addition, in addition to internal faults of the generator, a series of external factors may also cause the stator grounding protection to operate, such as inter-turn short circuit of the voltage transformer connected to the generator end, broken excitation transformer conductor, dampness of the closed busbar, and failure of the main transformer.

[0003] At present, the fundamental zero-sequence voltage protection, third harmonic voltage protection and injection-type stator single-phase grounding protection are generally used in generator sets. Although they can isolate the grounding fault, they cannot accurately locate the fault. For large generator sets, the shutdown and maintenance procedures after the fault are complicated, which will cause huge economic losses. Summary of the invention

[0004] The embodiments of the present invention aim to solve at least one of the technical problems existing in the prior art, and provide a method and system for locating a single-phase grounding fault of a generator stator based on data driving.

[0005] An embodiment of the present invention provides a data-driven method for locating a single-phase ground fault of a generator stator, characterized in that the locating method comprises:

[0006] Construct electromagnetic transient model of generator stator single-phase ground fault based on PSCAD / EMTDC;

[0007] Setting the transition resistance of the fault point of the generator stator single-phase grounding fault electromagnetic transient model, and forming a training set and a test set;

[0008] Constructing an MLP neural network model, wherein the MLP neural network model includes: an MLP fault location model, an MLP transition resistance calculation model and an MLP fault branch discrimination model;

[0009] Train the MLP neural network model based on the training set of fault data, find the optimal hyperparameter combination of the MLP neural network model, evaluate the MLP neural network model, and verify the MLP neural network model based on the test set of fault data;

[0010] The new data at the time of fault is input into the MLP neural network model, and the fault location, transition resistance and fault branch are output.

[0011] In some embodiments of the present invention, after constructing the electromagnetic transient model of the generator stator single-phase grounding fault based on PSCAD / EMTDC and setting the parameters of the electromagnetic transient model of the generator stator single-phase grounding fault according to the parameters of the generator, it also includes:

[0012] Selecting the fault point of each phase of the stator winding in the electromagnetic transient model of the generator stator single-phase grounding fault;

[0013] The potential of each phase of the stator winding in the generator stator single-phase grounding fault electromagnetic transient model is set.

[0014] In some embodiments of the present invention, the selecting of the fault point of each phase of the stator winding in the electromagnetic transient model of the generator stator single-phase grounding fault specifically includes:

[0015] Each phase of the stator winding of the generator stator single-phase ground fault electromagnetic transient model is divided into two branches, and each branch is divided into n basic unit circuits;

[0016] According to the 2n basic unit circuits of each phase of the stator winding, 2n-1 fault points of each phase are set.

[0017] In some embodiments of the present invention, setting the transition resistance of the fault point of the generator stator single-phase grounding fault electromagnetic transient model and forming a training set and a test set of fault data specifically includes:

[0018] Select all fault points, select transition resistances at equal intervals between 100 and 1000 Ω, generate fault data for all fault points, and form a training set;

[0019] Some fault points are selected, transition resistances are selected at equal intervals between 100 and 1000 Ω, fault data of some fault points are generated, and a test set is formed.

[0020] In some embodiments of the present invention, an MLP neural network model is constructed, wherein the MLP neural network model includes: an MLP fault location model, an MLP transition resistance calculation model and an MLP fault branch discrimination model, specifically including:

[0021] Construct the input layer, hidden layer and output layer of the MLP neural network model;

[0022] The fundamental zero-sequence voltage amplitude and the phase difference between the zero-sequence voltage and the single-phase potential of the fault data are used as the input layer of each model, the fault position is used as the output layer of the MLP fault location model, the resistance value of the transition resistance is used as the output layer of the MLP transition resistance calculation model, and the fault branch is used as the output layer of the MLP fault branch discrimination model.

[0023] The second aspect of the present invention proposes a data-driven generator stator single-phase grounding fault locating system, the data-driven generator stator single-phase grounding fault locating system is used to execute the data-driven generator stator single-phase grounding fault locating method according to any of the above embodiments, the data-driven generator stator single-phase grounding fault locating system includes:

[0024] Model building module: The model building module is used to build an electromagnetic transient model of a generator stator single-phase grounding fault based on PSCAD / EMTDC, and is used to build an MLP neural network model, wherein the MLP neural network model includes: an MLP fault location model, an MLP transition resistance calculation model, and an MLP fault branch discrimination model;

[0025] A data setting module, the data setting module is used to set the transition resistance of the fault point of the electromagnetic transient model of the generator stator single-phase grounding fault, and form a training set and a test set of fault data;

[0026] A model training module, wherein the model training module is used to train the MLP neural network model according to the training set of fault data, find the optimal hyperparameter combination of the MLP neural network model, evaluate the MLP neural network model, and verify the MLP neural network model according to the test set of fault data;

[0027] The fault location module is used to input new data at the time of the fault into the MLP neural network model and output the fault location, transition resistance and fault branch.

[0028] The data-driven generator stator single-phase grounding fault location method and system of the embodiment of the present invention first constructs a generator stator single-phase grounding fault electromagnetic transient model based on PSCAD / EMTDC, then sets the transition resistance of the fault point of the fault electromagnetic transient model, generates a training set and a test set of fault data, and then constructs an MLP neural network model composed of an MLP fault location model, an MLP transition resistance calculation model, and an MLP fault branch discrimination model, trains the MLP neural network model, and finds the optimal hyperparameter combination of the MLP neural network model, applies the optimal hyperparameter combination to the MLP neural network model and evaluates the model, and finally inputs new data at the time of the fault into the MLP neural network model, and outputs the fault location, fault branch, and transition resistance. The data-driven generator stator single-phase grounding fault location method can accurately locate the fault location, effectively calculate the transition resistance, and reliably discriminate the fault branch, thereby improving the efficiency of troubleshooting and processing the generator stator single-phase grounding fault. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1is a logic flow chart of a method for locating a single-phase ground fault of a generator stator based on data drive according to an embodiment of the present invention;

[0030] Figure 2 This is an electromagnetic transient model diagram of a single-phase grounding fault of a generator stator according to an embodiment of the present invention;

[0031] Figure 3 It is the MLP neural network model diagram;

[0032] Figure 4 It is a scatter plot of the actual fault location and the predicted fault location;

[0033] Figure 5 It is a relationship diagram between the actual value of transition resistance and the predicted value of transition resistance;

[0034] Figure 6 It is the confusion matrix diagram of fault branch discrimination. DETAILED DESCRIPTION

[0035] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0036] like Figure 1 As shown, an embodiment of the present invention provides a data-driven method for locating a single-phase ground fault of a generator stator, the locating method comprising:

[0037] S100: Constructing electromagnetic transient model of generator stator single-phase ground fault based on PSCAD / EMTDC;

[0038] S200: setting the transition resistance of the fault point of the electromagnetic transient model of the generator stator single-phase grounding fault, and forming a training set and a test set of fault data;

[0039] S300: constructing an MLP neural network model, wherein the MLP neural network model includes: an MLP fault location model, an MLP transition resistance calculation model and an MLP fault branch discrimination model;

[0040] S400: training the MLP neural network model according to the training set of fault data, finding the optimal hyperparameter combination of the MLP neural network model, evaluating the MLP neural network model, and verifying the MLP neural network model according to the test set of fault data;

[0041] S500: Input the new data at the time of the fault into the MLP neural network model, and output the fault location, transition resistance and fault branch.

[0042] The data-driven generator stator single-phase grounding fault location method of the embodiment of the present invention first constructs a generator stator single-phase grounding fault electromagnetic transient model based on PSCAD / EMTDC (such as Figure 2 As shown), then set the transition resistance of the fault point of the fault electromagnetic transient model, generate a training set and a test set of fault data, and then construct an MLP neural network model consisting of an MLP fault location model, an MLP transition resistance calculation model, and an MLP fault branch discrimination model, train the MLP neural network model, and find the optimal hyperparameter combination of the MLP neural network model, apply the optimal hyperparameter combination to the MLP neural network model and evaluate the model, and finally input the new data at the time of the fault into the MLP neural network model, and output the fault location, fault branch, and transition resistance. This data-driven generator stator single-phase grounding fault location method can accurately locate the fault location, effectively calculate the transition resistance, and reliably discriminate the fault branch, thereby improving the efficiency of troubleshooting and processing of generator stator single-phase grounding faults and reducing the economic losses of the generator set.

[0043] In some embodiments of the present invention, S100: constructing an electromagnetic transient model of a generator stator single-phase grounding fault based on PSCAD / EMTDC, specifically including:

[0044] S110: Obtaining parameters of the generator;

[0045] S120: construct an electromagnetic transient model of a generator stator single-phase grounding fault based on PSCAD / EMTDC, and set parameters of the electromagnetic transient model of the generator stator single-phase grounding fault according to the parameters of the generator.

[0046] Specifically, the various parameters of the generator are first obtained, and then the electromagnetic transient model of the generator stator single-phase grounding fault is constructed based on PSCAD / EMTDC, and the parameters of the electromagnetic transient model of the generator stator single-phase grounding fault are set according to the various parameters of the generator. Among them, the various parameters of the generator include: the rated voltage U n , Neutral point grounding resistance R n , stator winding resistance per phase R s , stator winding inductance per phase L s 、Electronic winding distributed capacitance C per phase G The other components connected to the generator end have a capacitance C per phase to ground J Among them, the relative distributed capacitances of stator windings A, B, and C are C a , C b , C c .

[0047] The parameters of the electromagnetic transient model of the generator stator single-phase grounding fault are shown in Table 1.

[0048] Table 1 Generator parameters in the transient model of single-phase ground fault of generator stator

[0049]

[0050]

[0051] In some embodiments of the present invention, after S120: constructing the electromagnetic transient model of the generator stator single-phase grounding fault based on PSCAD / EMTDC and setting the parameters of the electromagnetic transient model of the generator stator single-phase grounding fault according to the parameters of the generator, it also includes:

[0052] Selecting the fault point of each phase of the stator winding in the electromagnetic transient model of the generator stator single-phase grounding fault;

[0053] The potential of each phase of the stator winding in the generator stator single-phase grounding fault electromagnetic transient model is set.

[0054] Specifically, each phase of the electromagnetic transient model of the generator stator single-phase grounding fault is divided into multiple unit circuits, and corresponding fault points are set. Then, the potential of each phase of the stator winding in the electromagnetic transient model of the generator stator single-phase grounding fault is assigned.

[0055] Furthermore, the selecting of the fault point of each phase of the stator winding in the electromagnetic transient model of the generator stator single-phase grounding fault specifically includes:

[0056] Each phase of the stator winding of the generator stator single-phase ground fault electromagnetic transient model is divided into two branches, and each branch is divided into n basic unit circuits;

[0057] According to the 2n basic unit circuits of each phase of the stator winding, 2n-1 fault points of each phase are set.

[0058] Specifically, Figure 2 As shown, the A phase of the stator winding of the electromagnetic transient model of the generator stator single-phase grounding fault is divided into two branches, A1 and A2, and the two branches A1 and A2 are divided into 6 basic unit circuits respectively; the B phase is divided into two branches, B1 and B2, and the two branches B1 and B2 are divided into 6 basic circuits respectively; the C phase is divided into two branches, C1 and C2, and the two branches C1 and C2 are divided into 6 basic circuits respectively.

[0059] Taking the A phase voltage equal to 1∠0° as an example, the potential of each unit circuit in the electromagnetic transient model of the generator stator single-phase grounding fault is set, as shown in Table 2.

[0060] Table 2 Potential of each unit circuit of A1 and A2 branches of phase A

[0061]

[0062] Taking the A phase grounding fault as an example, each branch of A1 and A2 is divided into 6 unit circuits, and the fault position is set at 1 / 6 to 6 / 6 of each branch of the stator winding with an equal interval of 1 / 6. Therefore, there are 11 fault points in total for selection in phase A.

[0063] In some embodiments of the present invention, setting the transition resistance of the fault point of the generator stator single-phase grounding fault electromagnetic transient model and forming a training set and a test set of fault data specifically includes:

[0064] Select all fault points, select transition resistances at equal intervals between 100 and 1000 Ω, generate fault data for all fault points, and form a training set;

[0065] Some fault points are selected, transition resistances are selected at equal intervals between 100 and 1000 Ω, fault data of some fault points are generated, and a test set is formed.

[0066] Taking the A-phase grounding fault as an example, for the fault location model, 11 fault points in the two branches A1 and A2 were selected, and the transition resistance was changed at equal intervals in the range of 100-1000Ω to generate 506 groups of fault data as the training set. A total of 3 fault points at 1 / 3, 2 / 3, and 5 / 6 of the stator winding of the A1 branch were selected, and another 135 groups of fault data were generated by changing the transition resistance at equal intervals in the range of 100-1000Ω as the test set.

[0067] For the transition resistance calculation model, 11 fault points in the two branches A1 and A2 were selected, and the transition resistance was changed at equal intervals in the range of 100-1000Ω to generate 906 groups of fault data as the training set. A total of 3 fault points at 1 / 3, 2 / 3, and 5 / 6 of the stator winding of the A1 branch were selected, and the transition resistance was changed at equal intervals in the range of 100-1000Ω to generate 135 groups of fault data as the test set.

[0068] For the fault branch discrimination model, 11 fault points of the two branches A1 and A2 were selected, and the transition resistance was changed at equal intervals in the range of 100 to 1000Ω to generate 906 groups of fault data as the training set. A total of 6 fault points at 1 / 3, 2 / 3, and 5 / 6 of the stator winding of the A1 and A2 branches were selected, and the transition resistance was changed at equal intervals in the range of 100 to 1000Ω to generate 270 groups of fault data as the test set.

[0069] The fault data of the electromagnetic transient model of the generator stator single-phase grounding fault generated by the fault location model, transition electronic calculation model and fault branch discrimination model are shown in Table 3.

[0070] Table 3 Number of fault data of electromagnetic transient model for single-phase ground fault of generator stator

[0071]

[0072] For phase B, all fault points are selected, transition resistances are selected at equal intervals between 100 and 1000 Ω, fault data of all fault points are generated, and a training set of phase B is formed; some fault points are selected, transition resistances are selected at equal intervals between 100 and 1000 Ω, fault data of some fault points are generated, and a test set of phase B is formed.

[0073] For phase C, all fault points are selected, transition resistances are selected at equal intervals between 100 and 1000 Ω, fault data of all fault points are generated, and a training set of phase C is formed; some fault points are selected, transition resistances are selected at equal intervals between 100 and 1000 Ω, fault data of some fault points are generated, and a test set of phase C is formed.

[0074] In some embodiments of the present invention, S300: constructing an MLP neural network model, wherein the MLP neural network model includes: an MLP fault location model, an MLP transition resistance calculation model and an MLP fault branch discrimination model, specifically including:

[0075] Construct the input layer, hidden layer and output layer of the MLP neural network model;

[0076] The fundamental zero-sequence voltage amplitude and the phase difference between the zero-sequence voltage and the A-phase potential of the fault data are used as the input layer of each model, the fault position is used as the output layer of the MLP fault location model, the resistance value of the transition resistance is used as the output layer of the MLP transition resistance calculation model, and the fault branch is used as the output layer of the MLP fault branch discrimination model.

[0077] Specifically, we first construct an MLP neural network model (such as Figure 3 As shown in Figure 1, the MLP neural network model includes an input layer, two hidden layers and an output layer, in which each layer of neurons is fully connected. Then, the fundamental zero-sequence voltage amplitude of the fault data and the phase difference between the zero-sequence voltage and the A-phase potential are used as the input layer of each model, the fault position is used as the output layer of the MLP fault location model, the transition resistance value is used as the output layer of the MLP transition resistance calculation model, and the fault branch is used as the output layer of the MLP fault branch discrimination model. Among them, the fault position refers to the ratio of the distance from the fault point to the neutral point to the entire stator winding.

[0078] Specifically, the zero-sequence voltage of the fault data of phase A and the potential phase difference of phase A are used as the input layer of the model, and the fundamental zero-sequence voltage amplitude of the fault data of phase A is used as the input layer of the model. The zero-sequence voltage of the fault data of phase B and the potential phase difference of phase A are used as the input layer of the model, and the fundamental zero-sequence voltage amplitude of the fault data of phase B is used as the input layer of the model. The zero-sequence voltage of the fault data of phase C and the potential phase difference of phase A are used as the input layer of the model, and the fundamental zero-sequence voltage amplitude of the fault data of phase C is used as the input layer of the model.

[0079] In some embodiments of the present invention, S400: training the MLP neural network model according to the training set of fault data, finding the optimal hyperparameter combination of the MLP neural network model, and evaluating the MLP neural network model, further includes:

[0080] The fault data input into the MLP neural network model is normalized by Max-Min.

[0081] Specifically, using data normalization to solve the problem of inconsistent dimensions of different features and large differences in numerical ranges can improve algorithm stability and model robustness, and highlight the changing trend of data. At the same time, normalization processing accelerates the convergence speed of neural network training, reduces the risk of overfitting of small-scale distributed data, and enhances the generalization ability of the model.

[0082] The Max-Min normalization method, such as formula (4), is used to process the data.

[0083]

[0084] Where: X * is the normalized data; X i is the data set of the first group; X min is the minimum eigenvector; X max is the maximum eigenvector.

[0085] In some embodiments of the present invention, S400: training an MLP neural network model according to a training set of fault data, finding an optimal hyperparameter combination of the MLP neural network model, and evaluating the MLP neural network model, specifically includes:

[0086] The hyperparameter ranges of the MLP fault location model, the MLP transition resistance calculation model, and the MLP fault branch discrimination model are set respectively. The hyperparameter ranges are shown in Tables 4 to 6.

[0087] Table 4. Hyperparameter ranges of MLP fault localization model

[0088]

[0089] Table 5. Hyperparameter ranges of MLP transition resistance calculation model

[0090]

[0091] Table 6 Hyperparameter ranges of MLP fault branch discrimination model

[0092]

[0093] Then, the cross-validation method and grid search method are used to optimize the hyperparameters and determine the optimal hyperparameter combination.

[0094] Specifically, according to the training set of fault data, the cross-validation method and grid search method are used to optimize the hyperparameters and determine the optimal hyperparameter combination.

[0095] The combination of grid search and cross validation is often used as a common hyperparameter tuning method. By constructing a hyperparameter grid structure and traversing all possible combinations, the optimal hyperparameter configuration is found to achieve the purpose of optimizing model performance.

[0096] After determining the optimal parameter combination, the MLP neural network model is evaluated and verified using a test set of fault data.

[0097] According to Table 4, the optimal parameter combination of the fault location model is selected, the number of neurons in hidden layer 1 is 16, the number of neurons in hidden layer 2 is 8, the learning rate is 0.001, and the number of iterations is 4000. In the test set, when the distance from the actual fault position to the neutral point accounts for 1 / 3, 2 / 3, and 5 / 6 of the entire stator winding, the predicted fault position is exactly the same as the actual fault position, such as Figure 4 shown.

[0098] According to Table 5, the optimal parameter combination of the transition resistance calculation model is selected, the number of neurons in the hidden layer 1 is 8, the number of neurons in the hidden layer 2 is 4, the learning rate is 0.001, and the number of iterations is 4000. In the test set, the actual value and predicted value of the transition resistance are as follows Figure 5 As shown, the error between the predicted value and the actual value in each set of data is less than 10%, and the accuracy is relatively high.

[0099] According to Table 6, the optimal parameter combination of the fault branch recognition model is selected, the number of neurons in hidden layer 1 is 8, the number of neurons in hidden layer 2 is 4, the learning rate is 0.001, and the number of iterations is 1500. In the test set, the confusion matrix of the fault branch is as follows Figure 6As shown in the figure, 264 out of 270 test data sets can be accurately identified, with an accuracy of 97.8%. Due to the structural limitations of the electromagnetic transient model used in this paper, the two branches of the stator winding converge at the generator outlet. This structure makes it impossible to distinguish the fault branch at this location, so the accuracy of the MLP fault branch discrimination model cannot reach 100%.

[0100] The second aspect of the present invention proposes a data-driven generator stator single-phase grounding fault locating system, the data-driven generator stator single-phase grounding fault locating system is used to execute the data-driven generator stator single-phase grounding fault locating method according to any of the above embodiments, the data-driven generator stator single-phase grounding fault locating system includes:

[0101] Model building module: The model building module is used to build an electromagnetic transient model of a generator stator single-phase grounding fault based on PSCAD / EMTDC, and is used to build an MLP neural network model, wherein the MLP neural network model includes: an MLP fault location model, an MLP transition resistance calculation model, and an MLP fault branch discrimination model;

[0102] A data acquisition setting module, the data acquisition setting module is used to acquire the transition resistance of the fault point of the generator stator single-phase grounding fault electromagnetic transient model, and form a training set and a test set of fault data;

[0103] A model training module, wherein the model training module is used to train the MLP neural network model according to the training set of fault data, find the optimal hyperparameter combination of the MLP neural network model, evaluate the MLP neural network model, and verify the MLP neural network model according to the test set of fault data;

[0104] The fault location module is used to input new data at the time of the fault into the MLP neural network model and output the fault location, transition resistance and fault branch.

[0105] The data-driven generator stator single-phase grounding fault location system of the embodiment of the present invention firstly builds the generator stator single-phase grounding fault electromagnetic transient model based on PSCAD / EMTDC, then the data setting module sets the transition resistance of the fault point of the fault electromagnetic transient model, and generates a training set and a test set of fault data, the model building module then builds an MLP neural network model composed of an MLP fault location model, an MLP transition resistance calculation model and an MLP fault branch discrimination model, the model training module trains the MLP neural network model, and finds the optimal hyperparameter combination of the MLP neural network model, applies the optimal hyperparameter combination to the MLP neural network model and evaluates the model, and finally, the fault location module inputs the new data at the time of the fault into the MLP neural network model, and outputs the fault location, fault branch and transition resistance. The data-driven generator stator single-phase grounding fault location system can accurately locate the fault location, effectively calculate the transition resistance, reliably discriminate the fault branch, and improve the efficiency of troubleshooting and processing the generator stator single-phase grounding fault.

[0106] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present invention, but the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A data-driven method for locating a single-phase ground fault of a generator stator, characterized in that: The data-driven generator stator single-phase grounding fault location method includes: Construct electromagnetic transient model of generator stator single-phase ground fault based on PSCAD / EMTDC; Setting the transition resistance of the fault point of the generator stator single-phase grounding fault electromagnetic transient model, and forming a training set and a test set of fault data; Constructing an MLP neural network model, wherein the MLP neural network model includes: an MLP fault location model, an MLP transition resistance calculation model and an MLP fault branch discrimination model; Train the MLP neural network model based on the training set of fault data, find the optimal hyperparameter combination of the MLP neural network model, evaluate the MLP neural network model, and verify the MLP neural network model based on the test set of fault data; The new fault data is input into the MLP neural network model, and the fault location, transition resistance and fault branch are output.

2. The data-driven generator stator single-phase grounding fault location method according to claim 1 is characterized in that: After constructing the electromagnetic transient model of the generator stator single-phase grounding fault based on PSCAD / EMTDC and setting the parameters of the electromagnetic transient model of the generator stator single-phase grounding fault according to the parameters of the generator, the method further includes: Selecting the fault point of each phase of the stator winding in the electromagnetic transient model of the generator stator single-phase grounding fault; The potential of each phase of the stator winding in the generator stator single-phase grounding fault electromagnetic transient model is set.

3. The data-driven generator stator single-phase grounding fault location method according to claim 2 is characterized in that: The selecting of the fault point of each phase of the stator winding in the electromagnetic transient model of the generator stator single-phase grounding fault specifically includes: Each phase of the stator winding of the generator stator single-phase ground fault electromagnetic transient model is divided into two branches, and each branch is divided into n basic unit circuits; According to the 2n basic unit circuits of each phase of the stator winding, 2n-1 fault points of each phase are set.

4. The data-driven generator stator single-phase grounding fault location method according to claim 3 is characterized in that: The setting of the transition resistance of the fault point of the electromagnetic transient model of the generator stator single-phase grounding fault and forming a training set and a test set of fault data specifically includes: Select all fault points, select transition resistances at equal intervals between 100 and 1000 Ω, generate fault data for all fault points, and form a training set; Some fault points are selected, transition resistances are selected at equal intervals between 100 and 1000 Ω, fault data of some fault points are generated, and a test set is formed.

5. The data-driven generator stator single-phase grounding fault location method according to claim 1, characterized in that: Construct an MLP neural network model, wherein the MLP neural network model includes: an MLP fault location model, an MLP transition resistance calculation model and an MLP fault branch discrimination model, specifically including: Construct the input layer, hidden layer and output layer of the MLP neural network model; The fundamental zero-sequence voltage amplitude and the phase difference between the zero-sequence voltage and the A-phase potential of the fault data are used as the input layer of each model, the fault position is used as the output layer of the MLP fault location model, the resistance value of the transition resistance is used as the output layer of the MLP transition resistance calculation model, and the fault branch is used as the output layer of the MLP fault branch discrimination model.

6. A data-driven generator stator single-phase grounding fault location system, the data-driven generator stator single-phase grounding fault location system is used to perform the data-driven generator stator single-phase grounding fault location method according to any one of claims 1 to 5, the data-driven generator stator single-phase grounding fault location system comprising: Model building module: The model building module is used to build an electromagnetic transient model of a generator stator single-phase grounding fault based on PSCAD / EMTDC, and is used to build an MLP neural network model, wherein the MLP neural network model includes: an MLP fault location model, an MLP transition resistance calculation model, and an MLP fault branch discrimination model; A data setting module, the data setting module is used to set the transition resistance of the fault point of the electromagnetic transient model of the generator stator single-phase grounding fault, and form a training set and a test set of fault data; A model training module, wherein the model training module is used to train the MLP neural network model according to the training set of fault data, find the optimal hyperparameter combination of the MLP neural network model, evaluate the MLP neural network model, and verify the MLP neural network model according to the test set of fault data; The fault location module is used to input new data at the time of the fault into the MLP neural network model and output the fault location, transition resistance and fault branch.

Images