Distributed power distribution network leakage current grading protection method and device based on deep learning

Abstract

The invention discloses a distributed power distribution network leakage current grading protection method and device based on deep learning, and the method comprises the steps: collecting the electrical parameters of a power grid system, processing the electrical parameters according to a preset deep learning model, obtaining the analysis result of leakage current, and carrying out the grading protection of the leakage current according to the analysis result. And matching the corresponding hierarchical protection strategy, generating a corresponding execution instruction, and sending the execution instruction to an execution mechanism to enable the execution mechanism to execute a corresponding protection action. Through the deep learning algorithm, rapid and accurate identification and grading protection of the leakage current are realized, the intelligent level and the protection effect of the protection system are improved, and a powerful guarantee is provided for safe operation of the distributed power distribution network.

Description

Technical Field

[0001] The present invention belongs to the technical field of hierarchical overcurrent protection of power systems, and relates to a distributed distribution network leakage current hierarchical protection method and device based on deep learning. Background Art

[0002] In recent years, with the widespread access to distributed energy (such as solar energy, wind energy, etc.), the structure of the distribution network has become more complex, and the requirements for protection systems have also increased. Leakage current is a common safety hazard in the distribution network. Its existence may not only cause electrical fires, but also pose a serious threat to personal safety.

[0003] Some existing leakage protection schemes, such as using machine learning algorithms (such as support vector machines, decision trees, etc.) to classify and identify the collected electrical parameters to determine whether there is leakage, but these algorithms have certain limitations when dealing with complex and nonlinear problems, and require artificially designed feature extraction processes, making it difficult to fully utilize the information in the original data; or further based on shallow neural networks to automatically learn features from data to overcome the shortcomings of machine learning algorithms, however, due to the shallow number of network layers, its learning ability is limited and it is difficult to process large-scale, high-dimensional data. Therefore, it is of great practical significance to study the hierarchical overcurrent protection method of power systems, realize the rapid and accurate identification and hierarchical protection of leakage current, and improve the intelligence level and protection effect of the protection system. Summary of the invention

[0004] In response to the shortcomings of the prior art, the present application provides a distributed distribution network leakage current hierarchical protection method and device based on deep learning, which can realize rapid and accurate identification and hierarchical protection of leakage current.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a distributed distribution network leakage current hierarchical protection method based on deep learning, comprising:

[0006] Collecting electrical parameters of the power grid system; the electrical parameters include current, voltage and temperature data at each location of the power grid;

[0007] According to a preset deep learning model, the electrical parameters are processed to obtain an analysis result of the leakage current; the deep learning model is trained using a Dropout regularization method based on the electrical parameters of the known analysis results; the analysis result includes the type and severity of the leakage current;

[0008] According to the analysis results, a corresponding hierarchical protection strategy is matched and a corresponding execution instruction is generated;

[0009] The execution instruction is sent to the execution mechanism so that the execution mechanism executes the corresponding protection action.

[0010] Compared with the prior art, the embodiments of the present application have the following beneficial effects: by collecting the electrical parameters of the power grid system and using a deep learning model for analysis, efficient identification and classification of leakage current can be achieved; according to the analysis results of the model, the corresponding hierarchical protection strategy is matched and the corresponding execution instructions are generated so that the actuator executes the corresponding protection action, which can effectively achieve rapid response and effective isolation of leakage current.

[0011] In some embodiments of the first aspect of the present application, the deep learning model includes a convolutional neural network, a recurrent neural network, a long short-term memory network and a fully connected layer; the convolutional neural network is used to perform feature extraction and dimensionality reduction processing on the electrical parameters to obtain parameter features; the recurrent neural network or the long short-term memory network is used to perform timing analysis and pattern recognition on the parameter features to obtain recognition results; the fully connected layer is used to input the recognition results and output them as the analysis results; the timing analysis is divided into long timing analysis and short timing analysis; the long short-term memory network is used for long timing analysis; the recurrent neural network is used for short timing analysis.

[0012] Compared with the prior art, the above-mentioned embodiment has the following beneficial effects: through the combined optimization of multiple models, the Dropout regularization method is used to train the model, the overfitting in the neural network is reduced, the generalization ability is improved, different network models are selected for calculation for different time series data, and the efficiency of processing large-scale high-dimensional data is accelerated, so as to realize efficient identification and classification of leakage current.

[0013] In some embodiments of the first aspect of the present application, matching a corresponding hierarchical protection strategy and generating a corresponding execution instruction based on the analysis result includes: dividing the leakage fault into corresponding fault levels according to the analysis result and the topological structure of the distribution network; matching a corresponding hierarchical protection strategy according to the fault level, and generating a corresponding execution instruction.

[0014] Compared with the prior art, the above embodiment has the following beneficial effects: according to the analysis results and the actual topological structure, the leakage fault can be accurately judged, the hierarchical protection strategy can be matched, and the corresponding execution instructions can be generated to achieve rapid response, optimize the protection measures, improve the maintenance efficiency, and ensure the safe and stable operation of the power grid system.

[0015] In some embodiments of the first aspect of the present application, sending the execution instruction to the execution mechanism so that the execution mechanism performs a corresponding protection action includes: converting the execution instruction into a corresponding control signal; and sending the control signal to the execution mechanism so that the execution mechanism performs the corresponding protection action.

[0016] Compared with the prior art, the above embodiment has the following beneficial effects: precise control is performed according to the execution instructions, the execution efficiency of the protection action is improved, and the leakage current is effectively isolated, thereby improving the maintenance response efficiency.

[0017] In some embodiments of the first aspect of the present application, the collecting of electrical parameters of the power grid system includes: collecting real-time electrical data of the power grid system, and preprocessing the real-time electrical data to obtain the electrical parameters, wherein the preprocessing includes the following or a combination of multiple ones: outlier elimination processing, missing value filling processing, prediction supplement processing, normalization processing or calibration processing.

[0018] Compared with the prior art, the above embodiment has the following beneficial effects: preprocessing the collected electrical data can improve the quality of the data and provide a solid foundation for subsequent model processing and analysis, including: removing outliers from electrical parameters can reduce processing errors caused by abnormal data, which is helpful for the subsequent construction of a more stable model; filling missing values ​​can maintain data integrity, reduce deviations, and improve the accuracy of subsequent model processing; predicting future values ​​to supplement electrical data can enrich the dimensions and information content of electrical data and enhance data features; normalizing electrical data can speed up model training, improve model accuracy, improve the robustness of the algorithm, help the model learn faster, and also improve the generalization ability of the model; forwarding after verification can improve the reliability and security of data transmission.

[0019] In a second aspect, the present invention also provides a distributed distribution network leakage current hierarchical protection device based on deep learning, including: an acquisition module, a processing module, a decision module and an execution module.

[0020] Wherein, the acquisition module is used to acquire electrical parameters of the power grid system; the electrical parameters include current, voltage and temperature data at each location of the power grid;

[0021] The processing module is used to process the electrical parameters according to a preset deep learning model to obtain an analysis result of the leakage current; the deep learning model is trained using a Dropout regularization method based on the electrical parameters of the known analysis results; the analysis result includes the type and severity of the leakage current;

[0022] The decision module is used to match the corresponding hierarchical protection strategy and generate corresponding execution instructions according to the analysis result;

[0023] The execution module is used to send the execution instruction to the execution mechanism so that the execution mechanism executes the corresponding protection action.

[0024] In some embodiments of the second aspect of the present application, the processing module includes: a feature extraction unit, an analysis unit, and an output unit;

[0025] The feature extraction unit is used to use a convolutional neural network to perform feature extraction and dimension reduction processing on the electrical parameters to obtain parameter features;

[0026] The analysis unit is used to use a recurrent neural network or a long short-term memory network to perform time series analysis and pattern recognition on the parameter features to obtain a recognition result; wherein the time series analysis is divided into long time series analysis and short time series analysis; the long short-term memory network is used for long time series analysis; and the recurrent neural network is used for short time series analysis;

[0027] The output unit is used to input the recognition result using a fully connected layer and output it as the analysis result.

[0028] In some embodiments of the second aspect of the present application, the decision module includes: a level division unit and an instruction generation unit;

[0029] The level classification unit is used to classify the corresponding fault level for the leakage fault according to the analysis result and the topological structure of the distribution network;

[0030] The instruction generation unit is used to match the corresponding hierarchical protection strategy according to the fault level and generate a corresponding execution instruction.

[0031] In some embodiments of the second aspect of the present application, the execution module includes: an instruction conversion unit and an instruction execution unit;

[0032] The instruction conversion unit is used to convert the execution instruction into a corresponding control signal;

[0033] The instruction execution unit is used to send the control signal to the execution mechanism so that the execution mechanism executes the corresponding protection action.

[0034] In some embodiments of the second aspect of the present application, the acquisition module includes: an acquisition unit and a preprocessing unit;

[0035] The acquisition unit is used to collect real-time electrical data of the power grid system;

[0036] The preprocessing unit is used to preprocess the electrical data to obtain the electrical parameters;

[0037] The preprocessing includes the following or a combination of the following: outlier elimination processing, missing value filling processing, prediction supplement processing, normalization processing or calibration processing.

[0038] Compared with the prior art, the above embodiments of the present application have the following beneficial effects: by collecting the electrical parameters of the power grid system and using a deep learning model for analysis, efficient identification and classification of leakage current can be achieved. According to the analysis results of the model, the corresponding hierarchical protection strategy is matched and the corresponding execution instructions are generated so that the actuator executes the corresponding protection action, which can effectively achieve rapid response and effective isolation of leakage current. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 : A flow chart of a distributed distribution network leakage current hierarchical protection method based on deep learning provided in some embodiments of the present invention.

[0040] Figure 2 : A schematic diagram of the structure of a distributed distribution network leakage current grading protection device based on deep learning provided in some embodiments of the present invention.

[0041] Figure 3 : A schematic diagram of collected data provided in some embodiments of the present invention.

[0042] Figure 4 : A data preprocessing flow chart provided in some embodiments of the present invention.

[0043] Figure 5 : A time series sliding window diagram for predicting future values ​​provided in some embodiments of the present invention.

[0044] Figure 6 : is a CRC check algorithm table used for checking in some embodiments of the present invention.

[0045] Figure 7 : A demonstration diagram of the Dropout regularization method provided in some embodiments of the present invention.

[0046] Figure 8 : Schematic diagram of the deep learning model library provided in some embodiments of the present invention.

[0047] Fig. 9 :A deep learning flowchart provided in some embodiments of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] Embodiment 1:

[0050] Please refer to Figure 1 , a distributed distribution network leakage current hierarchical protection method based on deep learning provided by an embodiment of the present invention includes steps S1 to S4:

[0051] Step S1: Collect electrical parameters of the power grid system.

[0052] In some embodiments of the present application, Figure 3 The data collection schematic diagram shown in the figure, the electrical parameters include current, voltage and temperature data at each location of the power grid, including steps S11 to S12, the specific steps are as follows:

[0053] S11: Collect real-time electrical data of the power grid system;

[0054] S12: preprocessing the real-time electrical data to obtain the electrical parameters;

[0055] Further, step S12 can be implemented by any or a combination of the following preprocessing methods, as follows:

[0056] Outlier elimination processing, wherein the method used may be the Laida criterion, and the standard deviation σ and the average value μ are calculated according to the real-time electrical parameters, and the data with values ​​outside the range of (μ-3σ, μ+3σ) are regarded as outliers and eliminated;

[0057] Missing value filling process, the formula used is: where v t is the missing value in the real-time electrical data at time t, v t-i is the earliest data that appears at time point i before time point t in the real-time electrical data, v t+i It is the earliest data appearing at time point i after time point t in the real-time electrical data;

[0058] Predicting supplementary treatment, e.g. Figure 5 The time series sliding window diagram shown in the figure can slice and predict the electrical parameter data set in a sliding window manner, and use the values ​​of the first n data samples as input to predict the next (n+1) data sample. The specific steps are as follows:

[0059] The real-time electrical data is divided into two data sets, a training set train_X and a validation set test_X, and the future value to be predicted is called data Y;

[0060] Take train_X and test_X as the input of the prediction model, and get the output data Y after the model prediction;

[0061] Incorporating the output data Y into the real-time electrical data;

[0062] Normalization processing, where the formula is as follows: Where X is the original data value in the real-time electrical data, max is the maximum value in the original data value, min is the minimum value in the original data value, and X 、 is the normalized data value;

[0063] Check processing, such as Figure 6 The CRC check algorithm table shown in the figure can be used to perform CRC check;

[0064] Preferably, all the above pretreatment methods can be used as follows Figure 4 The preprocessing flow chart shown is processed sequentially to obtain the electrical parameters.

[0065] Compared with the prior art, the above embodiment has the following beneficial effects: preprocessing the collected electrical data can improve the quality of the data and provide a solid foundation for subsequent model processing and analysis, including: removing outliers from the electrical data can reduce processing errors caused by abnormal data, which is helpful for the subsequent construction of a more stable model; filling missing values ​​can maintain data integrity, reduce deviations, and improve the accuracy of subsequent model processing; predicting future values ​​to supplement the electrical data can enrich the dimensions and information content of the electrical data and enhance data features; normalizing the electrical data can speed up model training, improve model accuracy, improve the robustness of the algorithm, help the model learn faster, and also improve the generalization ability of the model; forwarding after verification can improve the reliability and security of data transmission.

[0066] Step S2: Process the electrical parameters according to a preset deep learning model to obtain an analysis result of the leakage current.

[0067] Preferably, in some embodiments of the present application, the deep learning model is trained using a Dropout regularization method based on electrical parameters of known analysis results, wherein the Dropout regularization method is as follows: Figure 7 As shown in the figure, during network training, Dropout will randomly select a certain proportion of neurons. These neurons will be "discarded" and will not participate in the forward and backward propagation of the network in this network iteration, so that the weights connecting these neurons will not be updated in this iteration. The neurons selected to be discarded in each iteration are random, and the retained neurons will be multiplied by a weight. This weight is the ratio of the number of all neurons to the number of retained neurons, which can generally be set to 0.5;

[0068] Further, the analysis results include the type and severity of the leakage current;

[0069] Furthermore, if Figure 8The deep learning model library diagram and Fig. 9 The deep learning flow chart shown, wherein the deep learning model includes a convolutional neural network, a recurrent neural network, a long short-term memory network and a fully connected layer;

[0070] Furthermore, the convolutional neural network is used to perform feature extraction and dimensionality reduction processing on the electrical parameters to obtain parameter features, the recurrent neural network or long short-term memory network is used to perform time series analysis and pattern recognition on the parameter features to obtain recognition results, and the fully connected layer is used to input the recognition results and output them as the analysis results;

[0071] Furthermore, the time series analysis is divided into long time series analysis and short time series analysis; the long short-term memory network is used for long time series analysis; and the recurrent neural network is used for short time series analysis.

[0072] Compared with the prior art, the above-mentioned embodiment has the following beneficial effects: through the combined optimization of multiple models, the Dropout regularization method is used to train the model, the overfitting in the neural network is reduced, the generalization ability is improved, different network models are selected for calculation for different time series data, and the efficiency of processing large-scale high-dimensional data is accelerated, so as to realize efficient identification and classification of leakage current.

[0073] Step S3: According to the analysis result, the corresponding hierarchical protection strategy is matched and the corresponding execution instruction is generated.

[0074] In some embodiments of the present application, the specific steps include S31 to S32, each step is as follows:

[0075] S31: classifying corresponding fault levels for leakage faults according to the analysis results and the topological structure of the distribution network;

[0076] S32: According to the fault level, matching a corresponding hierarchical protection strategy and generating a corresponding execution instruction;

[0077] The specific hierarchical protection strategies are as follows:

[0078] Level 1 protection: When a slight leakage is detected, the system first attempts to eliminate the leakage by adjusting the operating parameters of the distribution network (such as voltage, current, etc.). If the adjustment is effective, monitoring will continue; if the adjustment is ineffective or the leakage worsens, the next level of protection will be entered;

[0079] Secondary protection: When the leakage phenomenon persists or worsens, the system will automatically cut off the power supply of the relevant branch to prevent the leakage current from further expanding. At the same time, the system will send an alarm signal to the operation and maintenance personnel so that they can arrive at the site in time for processing;

[0080] Level 3 protection: In extreme cases, if the leakage fault seriously threatens the safe operation of the distribution network or personal safety, the system will take emergency measures to cut off the power supply of the entire area to ensure safety;

[0081] In addition to the above, the current state of the power grid can be further analyzed based on the analysis results and historical data to generate the development trend of leakage current and risk assessment results.

[0082] Compared with the prior art, the above embodiment has the following beneficial effects: according to the analysis results and the actual topological structure, the leakage fault can be accurately judged, the hierarchical protection strategy can be matched, and the corresponding execution instructions can be generated to achieve rapid response, optimize the protection measures, improve the maintenance efficiency, and ensure the safe and stable operation of the power grid system.

[0083] Step S4: Send the execution instruction to the execution mechanism so that the execution mechanism executes the corresponding protection action.

[0084] In some embodiments of the present application, the specific steps include S41 to S42, and the specific steps are as follows:

[0085] S41: converting the execution instruction into a corresponding control signal; wherein the control signal includes but is not limited to a circuit breaker control signal, an isolating switch control signal, etc., for realizing control of relevant equipment in the distribution network;

[0086] S42: Send the control signal to the actuator, so that the actuator performs a corresponding protection action.

[0087] Compared with the prior art, the above embodiment has the following beneficial effects: precise control is performed according to the execution instructions, the execution efficiency of the protection action is improved, and the leakage current is effectively isolated, thereby improving the maintenance response efficiency.

[0088] In summary, compared with the prior art, the above embodiments of the present application have the following beneficial effects: by collecting the electrical parameters of the power grid system and using a deep learning model for analysis, efficient identification and classification of leakage current can be achieved. According to the analysis results of the model, the corresponding hierarchical protection strategy is matched and the corresponding execution instructions are generated so that the actuator executes the corresponding protection action, which can effectively achieve rapid response and effective isolation of leakage current.

[0089] Embodiment 2:

[0090] Please refer to Figure 2 Based on the same inventive concept, the distributed distribution network leakage current hierarchical protection device based on deep learning disclosed in an embodiment of the present invention includes: an acquisition module M1, a processing module M2, a decision module M3 and an execution module M4.

[0091] The acquisition module M1 is used to acquire electrical parameters of the power grid system; the electrical parameters include current, voltage and temperature data at each location of the power grid;

[0092] The acquisition module M1 includes: an acquisition unit and a preprocessing unit; the acquisition unit is used to acquire real-time electrical data of the power grid system; the preprocessing unit is used to preprocess the electrical data to obtain the electrical parameters; wherein the preprocessing includes the following or a combination of the following: outlier elimination processing, missing value filling processing, prediction supplement processing, normalization processing or calibration processing;

[0093] Furthermore, the specific preprocessing method includes the following:

[0094] Outlier elimination processing, wherein the method used may be the Laida criterion, and the standard deviation σ and the average value μ are calculated according to the real-time electrical parameters, and the data with values ​​outside the range of (μ-3σ, μ+3σ) are regarded as outliers and eliminated;

[0095] Missing value filling process, the formula used is: where v t is the missing value in the real-time electrical data at time t, v t-i is the earliest data that appears at time point i before time point t in the real-time electrical data, v t+i The data that appears earliest at time point i after time point t in the real-time electrical data;

[0096] Predicting supplementary treatment, e.g. Figure 5 The time series sliding window diagram shown in the figure can slice and predict the electrical parameter data set in a sliding window manner, and use the values ​​of the first n data samples as input to predict the next (n+1) data sample. The specific steps are as follows:

[0097] The real-time electrical data is divided into two data sets, a training set train_X and a validation set test_X, and the future value to be predicted is called data Y;

[0098] Take train_X and test_X as the input of the prediction model, and get the output data Y after the model prediction;

[0099] Incorporating the output data Y into the real-time electrical data;

[0100] Normalization processing, where the formula is as follows: Where X is the original data value in the real-time electrical data, max is the maximum value in the original data value, min is the minimum value in the original data value, and X 、 is the normalized data value;

[0101] Check processing, such as Figure 6 The CRC check algorithm table shown in the figure can be used to perform CRC check;

[0102] Preferably, all the above pretreatment methods can be carried out as follows Figure 4 The preprocessing flow chart shown is processed sequentially to obtain the electrical parameters.

[0103] Compared with the prior art, the above embodiment has the following beneficial effects: preprocessing the collected electrical data can improve the quality of the data and provide a solid foundation for subsequent model processing and analysis, including: removing outliers from the electrical data can reduce processing errors caused by abnormal data, which is helpful for the subsequent construction of a more stable model; filling missing values ​​can maintain data integrity, reduce deviations, and improve the accuracy of subsequent model processing; predicting future values ​​to supplement the electrical data can enrich the dimensions and information content of the electrical data and enhance data features; normalizing the electrical data can speed up model training, improve model accuracy, improve the robustness of the algorithm, help the model learn faster, and also improve the generalization ability of the model; forwarding after verification can improve the reliability and security of data transmission.

[0104] The processing module M2 is used to process the electrical parameters according to a preset deep learning model to obtain an analysis result of the leakage current; the deep learning model is trained using a Dropout regularization method based on the electrical parameters of the known analysis results; the analysis result includes the type and severity of the leakage current;

[0105] Preferably, the Dropout regularization method is as follows Figure 7 As shown in the figure: When training the network, Dropout will randomly select a certain proportion of neurons. These neurons will be "discarded" and will not participate in the forward and backward propagation of the network in this network iteration, so that the weights connecting these neurons will not be updated in this iteration. The neurons selected to be discarded in each iteration are random, and the retained neurons will be multiplied by a weight. This weight is the ratio of the number of all neurons to the number of retained neurons, which can generally be set to 0.5;

[0106] The processing module M2 includes: a feature extraction unit, an analysis unit and an output unit; the feature extraction unit is used to use a convolutional neural network to perform feature extraction and dimension reduction processing on the electrical parameters to obtain parameter features; the analysis unit is used to use a recurrent neural network or a long short-term memory network to perform time series analysis and pattern recognition on the parameter features to obtain a recognition result; wherein the time series analysis is divided into long time series analysis and short time series analysis; the long short-term memory network is used for long time series analysis; the recurrent neural network is used for short time series analysis; the output unit is used to use a fully connected layer to input the recognition result and output it as the analysis result;

[0107] Further, the analysis results include the type and severity of the leakage current;

[0108] Furthermore, if Figure 8 The deep learning model library diagram and Fig. 9 The deep learning flow chart shown, wherein the deep learning model includes a convolutional neural network, a recurrent neural network, a long short-term memory network and a fully connected layer;

[0109] Furthermore, the convolutional neural network is used to perform feature extraction and dimensionality reduction processing on the electrical parameters to obtain parameter features, the recurrent neural network or long short-term memory network is used to perform time series analysis and pattern recognition on the parameter features to obtain recognition results, and the fully connected layer is used to input the recognition results and output them as the analysis results;

[0110] Furthermore, the time series analysis is divided into long time series analysis and short time series analysis; the long short-term memory network is used for long time series analysis; and the recurrent neural network is used for short time series analysis.

[0111] Compared with the prior art, the above-mentioned embodiment has the following beneficial effects: through the combined optimization of multiple models, the Dropout regularization method is used to train the model, the overfitting in the neural network is reduced, the generalization ability is improved, different network models are selected for calculation for different time series data, and the efficiency of processing large-scale high-dimensional data is accelerated, so as to realize efficient identification and classification of leakage current.

[0112] The decision module M3 is used to match the corresponding hierarchical protection strategy and generate corresponding execution instructions according to the analysis result;

[0113] The decision module M3 includes: a level division unit and an instruction generation unit; the level division unit is used to divide the corresponding fault level for the leakage fault according to the analysis result and the topological structure of the distribution network; the instruction generation unit is used to match the corresponding hierarchical protection strategy according to the fault level and generate the corresponding execution instruction;

[0114] The specific hierarchical protection strategies are as follows:

[0115] Level 1 protection: When a slight leakage is detected, the system first attempts to eliminate the leakage by adjusting the operating parameters of the distribution network (such as voltage, current, etc.). If the adjustment is effective, monitoring will continue; if the adjustment is ineffective or the leakage worsens, the next level of protection will be entered;

[0116] Secondary protection: When the leakage phenomenon persists or worsens, the system will automatically cut off the power supply of the relevant branch to prevent the leakage current from further expanding. At the same time, the system will send an alarm signal to the operation and maintenance personnel so that they can arrive at the site in time for processing;

[0117] Level 3 protection: In extreme cases, if the leakage fault seriously threatens the safe operation of the distribution network or personal safety, the system will take emergency measures to cut off the power supply of the entire area to ensure safety;

[0118] In addition to the above, the current state of the power grid can be further analyzed based on the analysis results and historical data to generate the development trend of leakage current and risk assessment results.

[0119] Compared with the prior art, the above embodiment has the following beneficial effects: according to the analysis results and the actual topological structure, the leakage fault can be accurately judged, the hierarchical protection strategy can be matched, and the corresponding execution instructions can be generated to achieve rapid response, optimize the protection measures, improve the maintenance efficiency, and ensure the safe and stable operation of the power grid system.

[0120] The execution module M4 is used to send the execution instruction to the execution mechanism, so that the execution mechanism executes the corresponding protection action.

[0121] The execution module M4 includes: an instruction conversion unit and an instruction execution unit; the instruction conversion unit is used to convert the execution instruction into a corresponding control signal; the control signal includes but is not limited to a circuit breaker control signal, an isolating switch control signal, etc., which is used to realize the control of related equipment in the distribution network; the instruction execution unit is used to send the control signal to the actuator, so that the actuator performs the corresponding protection action.

[0122] Compared with the prior art, the above embodiment has the following beneficial effects: precise control is performed according to the execution instructions, the execution efficiency of the protection action is improved, and the leakage current is effectively isolated, thereby improving the maintenance response efficiency.

[0123] In summary, compared with the prior art, the embodiments of the present application have the following beneficial effects: by collecting the electrical parameters of the power grid system and using a deep learning model for analysis, efficient identification and classification of leakage current can be achieved. According to the analysis results of the model, the corresponding hierarchical protection strategy is matched and the corresponding execution instructions are generated so that the actuator executes the corresponding protection action, which can effectively achieve rapid response and effective isolation of leakage current.

[0124] The division of the modules described above is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules may be combined or integrated into another system.

[0125] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A distributed distribution network leakage current hierarchical protection method based on deep learning, characterized in that: include: Collecting electrical parameters of the power grid system; the electrical parameters include current, voltage and temperature data at each location of the power grid; According to a preset deep learning model, the electrical parameters are processed to obtain an analysis result of the leakage current; the deep learning model is trained using a Dropout regularization method based on the electrical parameters of the known analysis results; the analysis result includes the type and severity of the leakage current; According to the analysis results, a corresponding hierarchical protection strategy is matched and a corresponding execution instruction is generated; The execution instruction is sent to the execution mechanism so that the execution mechanism executes the corresponding protection action.

2. The distributed distribution network leakage current hierarchical protection method based on deep learning according to claim 1 is characterized in that: The deep learning model includes a convolutional neural network, a recurrent neural network, a long short-term memory network and a fully connected layer; The convolutional neural network is used to perform feature extraction and dimension reduction processing on the electrical parameters to obtain parameter features; The recurrent neural network and the long short-term memory network are used to perform time series analysis and pattern recognition on the parameter features to obtain recognition results; The fully connected layer is used to input the recognition result and output the analysis result; The timing analysis is divided into long timing analysis and short timing analysis; The long short-term memory network is used for long time series analysis; The recurrent neural network is used for short time series analysis.

3. The distributed distribution network leakage current hierarchical protection method based on deep learning as claimed in claim 2 is characterized in that: According to the analysis result, matching the corresponding hierarchical protection strategy and generating the corresponding execution instruction includes: According to the analysis result and the topological structure of the distribution network, the leakage fault is classified into corresponding fault levels; According to the fault level, a corresponding hierarchical protection strategy is matched and a corresponding execution instruction is generated.

4. The distributed distribution network leakage current hierarchical protection method based on deep learning as claimed in claim 3 is characterized in that: The step of sending the execution instruction to the execution mechanism so that the execution mechanism executes the corresponding protection action includes: Converting the execution instruction into a corresponding control signal; The control signal is sent to the actuator so that the actuator performs a corresponding protection action.

5. The distributed distribution network leakage current hierarchical protection method based on deep learning according to any one of claims 1 to 4, characterized in that: The electrical parameters of the acquisition power grid system include: Collecting real-time electrical data of the power grid system and preprocessing the real-time electrical data to obtain the electrical parameters; The preprocessing includes the following or a combination of the following: outlier elimination processing, missing value filling processing, prediction supplement processing, normalization processing or calibration processing.

6. A distributed distribution network leakage current classification protection device based on deep learning, characterized in that: include: Acquisition module, processing module, decision module and execution module. Wherein, the acquisition module is used to acquire electrical parameters of the power grid system; the electrical parameters include current, voltage and temperature data at each location of the power grid; The processing module is used to process the electrical parameters according to a preset deep learning model to obtain an analysis result of the leakage current; the deep learning model is trained using a Dropout regularization method based on the electrical parameters of the known analysis results; the analysis result includes the type and severity of the leakage current; The decision module is used to match the corresponding hierarchical protection strategy and generate corresponding execution instructions according to the analysis result; The execution module is used to send the execution instruction to the execution mechanism so that the execution mechanism executes the corresponding protection action.

7. The distributed distribution network leakage current hierarchical protection device based on deep learning according to claim 6, characterized in that: The processing module includes: a feature extraction unit, an analysis unit and an output unit; The feature extraction unit is used to use a convolutional neural network to perform feature extraction and dimension reduction processing on the electrical parameters to obtain parameter features; The analysis unit is used to use a recurrent neural network or a long short-term memory network to perform time series analysis and pattern recognition on the parameter features to obtain a recognition result; wherein the time series analysis is divided into long time series analysis and short time series analysis; the long short-term memory network is used for long time series analysis; and the recurrent neural network is used for short time series analysis; The output unit is used to input the recognition result using a fully connected layer and output it as the analysis result.

8. The distributed distribution network leakage current hierarchical protection device based on deep learning according to claim 7, characterized in that: The decision module includes: a level division unit and an instruction generation unit; The level classification unit is used to classify the corresponding fault level for the leakage fault according to the analysis result and the topological structure of the distribution network; The instruction generation unit is used to match the corresponding hierarchical protection strategy according to the fault level and generate a corresponding execution instruction.

9. The distributed distribution network leakage current hierarchical protection device based on deep learning according to claim 8, characterized in that: The execution module includes: an instruction conversion unit and an instruction control unit; The instruction conversion unit is used to convert the execution instruction into a corresponding control signal; The instruction control unit is used to send the control signal to the actuator so that the actuator performs a corresponding protection action.

10. The distributed distribution network leakage current hierarchical protection device based on deep learning according to any one of claims 6 to 9, characterized in that: The acquisition module includes: an acquisition unit and a pre-processing unit; The acquisition unit is used to collect real-time electrical data of the power grid system; The preprocessing unit is used to preprocess the electrical data to obtain the electrical parameters; The preprocessing includes the following or a combination of the following: outlier elimination processing, missing value filling processing, prediction supplement processing, normalization processing or calibration processing.

Images