Intelligent diagnosis system of power distribution network fault self-judgment type switching device

Abstract

The invention discloses an intelligent diagnosis system of a power distribution network fault self-judgment switch device, and relates to the technical field of power system automation and intelligent power grids, and the intelligent diagnosis system comprises an intelligent diagnosis system interconnection hub which is in communication connection with a power parameter monitoring module, an intelligent fault detection module and a visual decision support module; the electric power parameter monitoring module monitors the state of a power distribution network in real time through a mutual inductor, voltage, current, power factors and harmonic data are collected through an intelligent electric meter, the intelligent fault detection module achieves fault recognition, trend prediction and anomaly detection through an LSTM neural network, and the decision support module provides a graphical interface and automatically generates a fault response strategy. According to the method, autonomous judgment and quick response of the power distribution network fault are achieved, the accuracy and efficiency of fault detection are remarkably improved, the false alarm rate and the missing report rate are reduced, meanwhile, the decision support module of the graphical interface simplifies the fault processing flow, and the stability and safety of power grid operation are improved.

Description

Technical Field

[0001] The present invention relates to the technical fields of power system automation and smart grid, and particularly relates to an intelligent diagnosis system for a self-judging switch device for distribution network faults. Background Art

[0002] A distribution network fault switch device is a key power equipment, which aims to automatically identify fault units in the power grid, isolate the fault units through reclosing, and at the same time restore the power supply of the non-fault part. Such a device usually includes a tripping loop device, a closing loop device, and a locking loop device composed of current and voltage relays, time relays, intermediate relays, and trip devices. Generally speaking, the distribution network fault switch device is an important equipment to ensure the safe and stable operation of the power grid.

[0003] In order to solve the problems of insufficient accuracy and slow response speed in fault detection and isolation of the distribution network fault switch device, the prior art uses a fault detection method based on preset thresholds and simple logical judgments for processing. Although this method can achieve preliminary detection and isolation of faults to a certain extent, there will still be misjudgments and missed judgments. Especially in a complex and changeable distribution network environment, when the fault characteristics are not obvious or multiple faults occur simultaneously, the traditional method is often difficult to accurately judge the fault type and location, resulting in low fault handling efficiency, long power restoration time, and affecting user experience and power grid stability. In view of this, an intelligent diagnosis system for a self-judging switch device for distribution network faults is proposed. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent diagnosis system for a self-judging switch device for distribution network faults to solve the problems raised in the above background art.

[0005] To solve the above technical problems, the technical solution adopted by the present invention is: an intelligent diagnosis system for a self-judging switch device for distribution network faults, including an intelligent diagnosis system interconnection hub, and the intelligent diagnosis system interconnection hub is communicatively connected to a power parameter monitoring module, an intelligent fault detection module, and a visualization decision support module;

[0006] The power parameter monitoring module, through current transformers and voltage transformers, real-time monitors the power line state information in the distribution network, collects voltage and current data, and through an intelligent electricity meter, obtains electricity consumption information and collects the power factor and harmonic content in the power grid;

[0007] The intelligent fault detection module constructs an LSTM neural network model to process time series data and realizes functions of fault mode recognition, trend prediction, and anomaly detection;

[0008] The visualization decision support module provides a graphical user interface and generates an automated fault response strategy based on the analysis results of the LSTM model.

[0009] A further improvement of the technical solution of the present invention lies in that: in the power parameter monitoring module, through current transformers and voltage transformers, the state information of the power lines in the distribution network is monitored in real time. The process of collecting voltage and current data includes:

[0010] Current transformers and voltage transformers are deployed at the outgoing side of the substation, the low-voltage side of the distribution transformer, and the T-joints and ring main units in the distribution network. Based on the principle of electromagnetic induction, the current transformer reduces the large current on the primary side proportionally to a small current on the secondary side. The voltage transformer adopts the principle of resistor voltage division and reduces the high voltage proportionally to a low voltage level;

[0011] Preprocess the collected current data and voltage data. The preprocessing operations include removing high-frequency noise in the collected current data and voltage data through low-pass filters and moving average filtering techniques, adding timestamps to each data point, and performing outlier processing. Set thresholds to identify and mark data points that exceed the normal range, and use interpolation methods to repair outliers;

[0012] Extract information reflecting the changes in the grid state from the preprocessed current data and voltage data. Calculate the root mean square values of current and voltage to evaluate the true strength of the signal, calculate the peak factor to measure the degree of signal fluctuation, decompose the fundamental wave and harmonic components through Fourier transform to assist in understanding the non-linear load situation in the grid, calculate the average value and standard deviation to provide additional perspectives on the overall trend and volatility, and perform transient event detection to identify rapid changes and zero-crossing points of current and voltage to capture potential fault signs;

[0013] Structurally pack the preprocessed and feature-extracted current data and voltage data, attach timestamps and source identifiers, and transmit them to the intelligent fault detection module using a symmetric encryption algorithm.

[0014] A further improvement of the technical solution of the present invention lies in that: in the power parameter monitoring module, through smart meters, electricity consumption information is obtained. The process of collecting the power factor and harmonic content in the grid includes:

[0015] Smart meters are deployed at the entrance of the user side, inside large industrial users, and at distributed energy access points. The smart meters collect current and voltage data through built-in current transformers and voltage transformers, and calculate the power factor and harmonic content;

[0016] Preprocess the data collected by the smart meter. The preprocessing operations include using a low-pass filter and a moving average filtering technique to remove high-frequency noise, adding timestamps to each data point, and performing outlier processing. Set a threshold to identify and mark data points that exceed the normal range, and use interpolation to repair outliers.

[0017] Extract features from the preprocessed data. Calculate the average power factor within a fixed-length time period, analyze the overall change trend, statistically calculate the average amplitude and standard deviation of each harmonic, identify rapid changes in the power factor and the content of each harmonic, and capture potential faults and abnormal situations.

[0018] Structurally package the power factor and harmonic content data that have undergone preprocessing and feature extraction, attach timestamps and source identifiers, and transmit them to the intelligent fault detection module using a symmetric encryption algorithm.

[0019] A further improvement in the technical solution of the present invention is that in the intelligent fault detection module, the process of processing time series data and training the input layer of the LSTM model includes:

[0020] Construct an LSTM model, which includes an input layer, a hidden layer, and an output layer.

[0021] The input layer receives the time series data from the power parameter monitoring module that has undergone preprocessing and feature extraction. The time series data includes current data, voltage data, power factor, and harmonic content. The time series data is segmented into a time window containing the past T time steps and the features of current, voltage, power factor, and harmonic content, forming a multi-dimensional array X = [x 1 , x 2 , …, x t , where x t represents the data vector at the t-th time step. Normalize each feature to make the numerical ranges between different features consistent. The calculation process is as follows:

[0022]

[0023] Among them, x' t represents the normalized feature value, μ represents the feature mean, and σ represents the standard deviation.

[0024] Map the data vector of each time window to the input space of the LSTM model, forming a three-dimensional tensor X input ∈ R T ×F , where T represents the time window length, F represents the number of features. For each new time window, initialize the hidden state h 0 and the cell state C 0 as zero vectors.

[0025] A further improvement of the technical solution of the present invention lies in that: in the intelligent fault detection module, the process of training the hidden layer of the LSTM model includes:

[0026] The hidden layer uses the gating mechanism in its internal long short-term memory cell state to capture its long-term dependencies that change over time. The gating mechanism includes a forget gate, an input gate, cell state update, and an output gate;

[0027] The forget gate evaluates and decides whether to retain the past state information based on the historical hidden state h t-1 and the input at the current time step x t . If the current and voltage remain stable within a fixed-length time period in the past, the forget gate chooses to discard the old current and voltage information. If the current and voltage are unstable within a fixed-length time period in the past, the forget gate chooses to retain the old current and voltage information. Its calculation process is as follows:

[0028] f t = σ(W f ·[h t-1 ; x t +b f );

[0029] where x t represents the input vector at the current time step, including current, voltage, power factor, and harmonic content, h t-1 represents the output state at the previous time step, W f is the weight matrix corresponding to the forget gate, b f is the bias term, σ is the sigmoid activation function, which maps the result to between 0 and 1, indicating the degree of retention and discard. If f t is close to 1, the information is retained. If f t is close to 0, the information is discarded;

[0030] The input gate evaluates whether the input x t at the current time step contains a sudden increase in current and a change in power factor and allows this change to enter the cell state C t , and its calculation process is as follows:

[0031] i t = σ(W i ·[h t-1 ,x t +b i );

[0032]

[0033] where i tRepresents the input factor, and uses the Sigmoid activation function to determine whether each element is added to the cell state. Uses the Tanh activation function to generate a candidate cell state ranging from -1 to 1, W i and b i Determines the selection mechanism of the input gate, W C and b C Controls the generation of the candidate cell state;

[0034] Combines the results of the forget gate and the input gate, multiplies the cell state of the previous moment bitwise by the result output by the forget gate, and then adds the candidate state generated by the input gate. If a short - circuit fault causes a sudden increase in current, the abnormal situation is reflected by updating the cell state that stores long - term information. The calculation process is as follows:

[0035]

[0036] The output gate extracts and outputs the information related to the mutation of the distribution network current and the grid power factor from the cell state as the hidden state h t , and its calculation process is as follows:

[0037] o t =σ(W o ·[h t-1 ,x t +b o );

[0038] h t =o t ⊙tanh(C t );

[0039] Among them, o t is the output factor, uses the Sigmoid activation function to determine whether each element is output, and tanh(C t ) makes the output value fall between -1 and 1, reflecting the change of the current cell state;

[0040] For each time step t in each time window, the calculation processes of the forget gate, the input gate, and the output gate are repeated to gradually update the hidden state h t and the cell state C t , capturing the dynamic changes in the time series.

[0041] A further improvement of the technical solution of the present invention lies in: in the intelligent fault detection module, the process of training the output layer of the LSTM model includes:

[0042] The output layer receives the hidden state h output from the hidden layer t, it is mapped to the target output dimension through the fully connected layer inside the output layer, an activation function is selected according to the task requirements, a loss function is used to measure the difference between the predicted value and the true value of the LSTM model, and the parameters of the LSTM model are adjusted through backpropagation.

[0043] A further improvement of the technical solution of the present invention lies in: in the intelligent fault detection module, the process of evaluating the performance of the LSTM model using an independent validation set includes:

[0044] A validation set is divided from the time series data that has never participated in the training of the LSTM model, the current, voltage, power factor, and harmonic content data in the validation set are preprocessed and feature extracted, and are mapped to the input space of the LSTM model. The trained LSTM model is used to predict the validation set data. The mean square error, mean absolute error, and accuracy are selected as the performance indicators to measure the difference between the predicted value and the true value. The reliability and stability of the evaluation are improved through the cross-validation method, and the hyperparameters are adjusted to optimize the model performance.

[0045] A further improvement of the technical solution of the present invention lies in: in the intelligent fault detection module, the process of deploying the LSTM model, based on the newly collected data, judging whether there is a fault and its type, foreseeing potential overload situations and equipment aging problems and performing anomaly detection includes:

[0046] For fault mode recognition, the output layer generates a probability distribution including two categories: normal and fault. The Softmax activation function is used to make the sum of probabilities equal to 1. The calculation process is as follows:

[0047]

[0048] Among them, z 0 represents the unnormalized score corresponding to the normal state, z 1 represents the unnormalized score corresponding to the fault state. A fault probability threshold is set. If P (故障) is lower than the set threshold, it is considered that the power grid is in a normal state. If P (故障) reaches the set threshold, it is considered that the power grid is in a fault state, and the fault type is analyzed. The fault type includes short circuit and ground fault;

[0049] For foreseeing potential overload situations and equipment aging problems, the output layer predicts the future current value, and the linear activation function is used to maintain the original numerical range. The calculation process is as follows:

[0050]

[0051] Among them, Represents the predicted current value for the next time step, sets a safety current threshold. If it is predicted that the current will exceed the safety current threshold of the distribution network at a certain future moment, an alarm will be issued in advance to remind the operation and maintenance personnel to take preventive measures;

[0052] Calculate the root mean square value, peak factor, average value, and standard deviation of the current data and voltage data, set the corresponding threshold ranges. When the actual measured value exceeds the threshold range, it is determined as abnormal; when the actual measured value does not exceed the threshold range, it is determined as normal. Identify the rapid changes and zero-crossing points of the current and voltage, capture potential fault signs. If the power factor suddenly changes and the harmonic content abnormally increases, the intelligent diagnosis system responds and generates an automated response strategy.

[0053] A further improvement of the technical solution of the present invention lies in: In the visual decision support module, the process of providing a graphical user interface includes:

[0054] Construct a graphical interface to display the information extracted from the current, voltage, power factor, and harmonic content data, assist the operation and maintenance personnel in fault judgment, foresee potential problems, and anomaly detection. This graphical interface integrates the real-time collected and preprocessed data, predicts the power grid state, future current values, and abnormal conditions through the LSTM model, calculates the root mean square value, peak factor, average value, and standard deviation, presents the data changes and anomaly marks in the form of trend charts, heat maps, classification result charts, and prediction curves, and provides interactive functions to allow dynamic updates, detailed views, and custom configuration of display parameters. When a fault is detected and potential problems are foreseen, an alarm is automatically triggered and corresponding maintenance suggestions are recommended.

[0055] A further improvement of the technical solution of the present invention lies in: In the visual decision support module, the process of generating an automated fault response strategy based on the analysis results of the LSTM model includes:

[0056] The LSTM model predicts the real-time collected data, outputs the normal and fault probability distributions of the distribution network state and the prediction of future current values. When the fault probability exceeds the set threshold, it is determined as a fault and the fault type is analyzed. If it is predicted that the current will exceed the safety current threshold, an alarm will be issued in advance to remind the operation and maintenance personnel to take preventive measures, trigger the automated response strategy. The automated response strategy includes triggering an alarm and displaying warning information through the graphical interface, pushing notifications to the operation and maintenance personnel, disconnecting the relevant circuits to prevent damage, performing maintenance inspections and replacing equipment in advance. For the detected fault, send control instructions to the switching device to quickly isolate the fault area and continuously supply power to the non-fault area.

[0057] Due to the adoption of the above technical solution, the technical progress achieved by the present invention compared with the prior art is:

[0058] 1. The present invention provides an intelligent diagnosis system for a self - judging switch device of a distribution network fault, which significantly improves the real - time performance and accuracy of fault detection. Through the efficient processing of the LSTM neural network model, it can quickly identify fault patterns, predict fault trends, and effectively shorten the fault troubleshooting time.

[0059] 2. The present invention provides an intelligent diagnosis system for a self - judging switch device of a distribution network fault. Through the comprehensive data collection of the power parameter monitoring module and the combination of smart meter information, it realizes an in - depth insight into the state of the distribution network, provides rich and accurate data support for fault detection, and reduces the false alarm rate.

[0060] 3. The present invention provides an intelligent diagnosis system for a self - judging switch device of a distribution network fault. The design of the visual decision - making support module greatly simplifies the fault handling process. Operators can intuitively understand the fault situation through the graphical user interface, automatically generate and execute fault response strategies, improving the reliability and safety of power grid operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0062] Figure 1 It is a block diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0064] Embodiment, as Figure 1 shown, the present invention provides an intelligent diagnosis system for a self - judging switch device of a distribution network fault, including an intelligent diagnosis system interconnection hub, and the intelligent diagnosis system interconnection hub is communicatively connected to a power parameter monitoring module, an intelligent fault detection module, and a visual decision - making support module;

[0065] The power parameter monitoring module monitors the status information of power lines in the distribution network in real time through current transformers and voltage transformers, collects voltage and current data, obtains power consumption information through smart meters, and collects power factor and harmonic content in the power grid. Current transformers and voltage transformers are deployed at the outgoing side of the substation, the low-voltage side of the distribution transformer, and the T-joints and ring main units in the distribution network. Based on the principle of electromagnetic induction, the current transformer scales down the large current on the primary side to a small current on the secondary side in proportion. The voltage transformer uses the principle of resistor voltage division to reduce the high voltage to a low voltage level. The collected current data and voltage data are preprocessed. The preprocessing operations include removing high-frequency noise in the collected current data and voltage data through a low-pass filter and a moving average filtering technique, adding a timestamp to each data point, and performing outlier processing. A threshold is set to identify and mark data points that exceed the normal range, and the interpolation method is used to repair outliers. Information reflecting the change of the grid state is mined from the preprocessed current data and voltage data. The root mean square values of current and voltage are calculated to evaluate the true intensity of the signal, and the peak factor is calculated to measure the fluctuation degree of the signal. The fundamental wave and harmonic components are decomposed through Fourier transform to assist in understanding the non-linear load situation in the power grid. The average value and standard deviation are calculated to provide an additional perspective on the overall trend and volatility. Transient event detection is performed to identify rapid changes and zero-crossing points of current and voltage and capture potential fault signs. The preprocessed and feature-extracted current data and voltage data are structured and packaged with timestamps and source identifiers and transmitted to the intelligent fault detection module using a symmetric encryption algorithm. Smart meters are deployed at the user-side entrance, inside large industrial users, and distributed energy access points. The smart meters collect current and voltage data through built-in current transformers and voltage transformers, calculate the power factor and harmonic content, and preprocess the data collected by the smart meters. The preprocessing operations include removing high-frequency noise using a low-pass filter and a moving average filtering technique, adding a timestamp to each data point, and performing outlier processing. A threshold is set to identify and mark data points that exceed the normal range, and the interpolation method is used to repair outliers. Feature extraction is performed on the preprocessed data. The average power factor within a fixed-length time period is calculated to analyze the overall change trend. The average amplitude and standard deviation of each harmonic are statistically analyzed to identify rapid changes in the power factor and the content of each harmonic and capture potential faults and abnormal conditions. The preprocessed and feature-extracted power factor and harmonic content data are structured and packaged with timestamps and source identifiers and transmitted to the intelligent fault detection module using a symmetric encryption algorithm;

[0066] The intelligent fault detection module constructs an LSTM neural network model to process time series data, realizing functions of fault mode recognition, trend prediction, and anomaly detection. An LSTM model is constructed. The LSTM model includes an input layer, a hidden layer, and an output layer. The input layer receives the time series data of the power parameter monitoring module that has undergone preprocessing and feature extraction. The time series data includes current data, voltage data, power factor, and harmonic content. The time series data is segmented into time windows containing the past T time steps and features of current, voltage, power factor, and harmonic content, forming a multi-dimensional array X = [x 1 , x 2 , …, x t , where x t represents the data vector at the t-th time step. Each feature is normalized to make the numerical ranges between different features consistent. The calculation process is as follows:

[0067]

[0068] Among them, x' t represents the normalized feature value, μ represents the feature mean, σ represents the standard deviation. The data vector of each time window is mapped to the input space of the LSTM model, forming a three-dimensional tensor X input ∈ R T×F , where T represents the time window length, F represents the number of features. For each new time window, the hidden state h 0 and the cell state C 0 of the LSTM cell are initialized as zero vectors. The hidden layer uses the gating mechanism in its internal long short-term memory cell state to capture the long-term dependencies that change over time. The gating mechanism includes a forget gate, an input gate, cell state update, and an output gate. The forget gate evaluates and decides whether to retain the past state information based on the historical hidden state h t-1 and the input at the current time step x t . If the current and voltage remain stable within a fixed-length time period in the past, the forget gate chooses to discard the old current and voltage information. If the current and voltage are unstable within a fixed-length time period in the past, the forget gate chooses to retain the old current and voltage information. The calculation process is as follows:

[0069] f t = σ(W f · [h t-1 ; x t + b f );

[0070] Among them, x t represents the input vector at the current time step, including current, voltage, power factor, and harmonic content, h t-1 represents the output state at the previous time step, Wf is the weight matrix corresponding to the forget gate, and b f is the bias term, and σ is the sigmoid activation function that maps the result to between 0 and 1, representing the retention and discard degrees. If f t is close to 1, the information is retained. If f t is close to 0, the information is discarded. The input gate evaluates whether the input x t at the current time step contains a sudden increase in current and a change in power factor, and allows this change to enter the cell state C t , and its calculation process is as follows:

[0071] i t =σ(W i ·[h t-1 ,x t +b i );

[0072]

[0073] Among them, i t represents the input factor, and uses the Sigmoid activation function to determine whether each element is added to the cell state. Uses the Tanh activation function to generate a candidate cell state in the range of -1 to 1. W i and b i determine the selection mechanism of the input gate. W C and b C control the generation of the candidate cell state. Combining the results of the forget gate and the input gate, multiply the cell state at the previous moment by the result output by the forget gate bit by bit, and then add the candidate state generated by the input gate. If a short - circuit fault causes a sudden increase in current, the abnormal situation is reflected by updating the cell state that stores long - term information. Its calculation process is as follows:

[0074]

[0075] The output gate extracts and outputs the information related to the sudden change in the current of the distribution network and the power factor of the power grid from the cell state as the hidden state h t , and its calculation process is as follows:

[0076] o t =σ(W o ·[h t-1 ,x t +b o );

[0077] h t =o t ⊙tanh(C t );

[0078] Among them, ot As the output factor, the Sigmoid activation function is used to determine whether each element is output. tanh(C t ) makes the output value fall between -1 and 1, reflecting the change in the current cell state. For each time step t in each time window, the calculation processes of the forget gate, input gate, and output gate are repeated to gradually update the hidden state h t and the cell state C t , capturing the dynamic changes in the time series. The output layer receives the hidden state h t output from the hidden layer, maps it to the target output dimension through the fully connected layer inside the output layer, selects the activation function according to the task requirements, uses the loss function to measure the difference between the predicted value and the true value of the LSTM model, and adjusts the LSTM model parameters through backpropagation. A validation set is divided from the time series data that has not participated in the LSTM model training. The current, voltage, power factor, and harmonic content data in the validation set are preprocessed and feature extracted, and then mapped to the input space of the LSTM model. The trained LSTM model is used to predict the validation set data. The mean square error, mean absolute error, and accuracy are selected as the performance indicators to measure the difference between the predicted value and the true value. The cross-validation method is used to improve the reliability and stability of the evaluation, and the hyperparameters are adjusted to optimize the model performance. For fault mode recognition, the output layer generates a probability distribution including two categories: normal and fault. The Softmax activation function is used to make the sum of probabilities equal to 1. The calculation process is as follows:

[0079]

[0080] where z 0 represents the unnormalized score corresponding to the normal state, and z 1 represents the unnormalized score corresponding to the fault state. A fault probability threshold is set. If P (故障) is lower than the set threshold, it is considered that the power grid is in a normal state. If P (故障) reaches the set threshold, it is considered that the power grid is in a fault state. Analyze the fault type, and the fault type includes short circuit and ground fault. For foreseeing potential overload situations and equipment aging problems, the output layer predicts the future current value, and the linear activation function is used to maintain the original numerical range. The calculation process is as follows:

[0081]

[0082] where Represents the predicted current value for the next time step. Set a safety current threshold. If it is predicted that the current will exceed the safety current threshold of the distribution network at a certain future time, an alarm will be issued in advance to remind the operation and maintenance personnel to take preventive measures. Calculate the root mean square value, peak factor, average value, and standard deviation of the current data and voltage data. Set the corresponding threshold ranges. When the actual measured value exceeds the threshold range, it is determined as abnormal; when the actual measured value does not exceed the threshold range, it is determined as normal. Identify the rapid changes and zero-crossing points of the current and voltage to capture potential fault signs. If the power factor suddenly changes and the harmonic content abnormally increases, the intelligent diagnosis system responds and generates an automated response strategy;

[0083] The visualization decision support module provides a graphical user interface, generates an automated fault response strategy based on the analysis results of the LSTM model, constructs a graphical interface to display the information extracted from the current, voltage, power factor, and harmonic content data, assisting the operation and maintenance personnel in fault judgment, foreseeing potential problems, and anomaly detection. This graphical interface integrates the real-time collected and preprocessed data, predicts the power grid state, future current values, and abnormal situations through the LSTM model, calculates the root mean square value, peak factor, average value, and standard deviation, and presents the data changes and anomaly marks in the form of trend charts, heat maps, classification result charts, and prediction curves, and provides interactive functions to allow dynamic updates, detailed viewing, and custom configuration of display parameters. When a fault is detected and potential problems are foreseen, an alarm is automatically triggered and corresponding maintenance suggestions are recommended. The LSTM model predicts the real-time collected data, outputs the normal and fault probability distributions of the distribution network state and the prediction of future current values. When the fault probability exceeds the set threshold, it is determined as a fault and the fault type is analyzed. If it is predicted that the current will exceed the safety current threshold, an alarm will be issued in advance to remind the operation and maintenance personnel to take preventive measures and trigger the automated response strategy. The automated response strategy includes triggering an alarm and displaying a warning message through the graphical interface, pushing a notification to the operation and maintenance personnel, disconnecting the relevant circuit to prevent damage, performing a maintenance check and replacing equipment in advance, sending a control instruction to the switching device for the detected fault to quickly isolate the fault area and continuously supply power to the non-fault area.

[0084] First, through the power parameter monitoring module, the current transformer and voltage transformer are used to monitor the power line status in the distribution network in real time, and voltage and current data are collected. At the same time, the power consumption information, including the power factor and harmonic content in the power grid, is obtained through the smart meter. Secondly, the intelligent fault detection module receives the data transmitted by the power parameter monitoring module and processes the collected data using the LSTM neural network model. This model can identify fault patterns, predict fault trends, and detect abnormal conditions, thereby realizing the autonomous judgment of distribution network faults. Finally, the visualization decision support module automatically generates fault response strategies based on the analysis results of the LSTM model and displays them to the operator through the graphical user interface. The operator can quickly take corresponding measures according to these strategies to handle the faults in the distribution network. The entire system realizes the communication and data transmission between modules through the intelligent diagnosis system interconnection hub, ensuring the accuracy of fault detection and the timeliness of response.

[0085] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. An intelligent diagnosis system for a power distribution network fault self-diagnosis switch device, characterized in that: It includes an intelligent diagnosis system interconnection hub, wherein the intelligent diagnosis system interconnection hub is communicatively connected with a power parameter monitoring module, an intelligent fault detection module and a visual decision support module; The power parameter monitoring module monitors the power line status information in the distribution network in real time through current transformers and voltage transformers, collects voltage and current data, obtains power consumption information through smart meters, and collects power factor and harmonic content in the power grid; The intelligent fault detection module builds an LSTM neural network model, processes time series data, and realizes fault pattern recognition, trend prediction and anomaly detection functions; The visual decision support module provides a graphical user interface and generates an automated fault response strategy based on the LSTM model analysis results.

2. The intelligent diagnosis system of a power distribution network fault self-diagnosis switch device according to claim 1, characterized in that: In the power parameter monitoring module, the power line status information in the distribution network is monitored in real time through the current transformer and the voltage transformer. The process of collecting voltage and current data includes: Current transformers and voltage transformers are deployed on the outgoing line side of the substation, the low-voltage side of the distribution transformer, and the T-junctions and ring network cabinets in the distribution network. The current transformer is based on the principle of electromagnetic induction to proportionally reduce the large current on the primary side to the small current on the secondary side. The voltage transformer uses the principle of resistance voltage division to proportionally reduce the high voltage to a low voltage level. Preprocessing the collected current data and voltage data, wherein the preprocessing operation includes removing high-frequency noise in the collected current data and voltage data by using a low-pass filter and a sliding average filter technology, adding a timestamp to each data point, and performing outlier processing, setting a threshold to identify and mark data points that are beyond a normal range, and using an interpolation method to repair outliers; Extract information reflecting the change of power grid status from the pre-processed current and voltage data, calculate the root mean square value of current and voltage to evaluate the true strength of the signal, calculate the crest factor to measure the degree of signal fluctuation, decompose the fundamental and harmonic components through Fourier transform to assist in understanding the nonlinear load conditions in the power grid, calculate the average and standard deviation to provide additional perspectives on the overall trend and volatility, perform transient event detection, identify rapid changes and zero crossing points of current and voltage, and capture potential fault signs; The preprocessed and feature-extracted current and voltage data are packaged in a structured manner with timestamps and source identifiers, and transmitted to the intelligent fault detection module using a symmetric encryption algorithm.

3. The intelligent diagnosis system of a power distribution network fault self-diagnosis switch device according to claim 2, characterized in that: In the power parameter monitoring module, the process of obtaining power consumption information and collecting power factor and harmonic content in the power grid through a smart meter includes: Smart meters are deployed at the user entrance, inside large industrial users, and at distributed energy access points. Smart meters collect current and voltage data through built-in current transformers and voltage transformers to calculate power factor and harmonic content. Preprocessing the data collected by the smart meter, the preprocessing operation includes removing high-frequency noise using a low-pass filter and a sliding average filter technique, adding a timestamp to each data point, performing outlier processing, setting a threshold to identify and mark data points that are out of a normal range, and using an interpolation method to repair outliers; Extract features from preprocessed data, calculate the average power factor within a fixed-length time period, analyze the overall change trend, count the average amplitude and standard deviation of each harmonic, identify rapid changes in power factor and harmonic content, and capture potential faults and abnormal conditions; The pre-processed and feature-extracted power factor and harmonic content data are structured and packaged with timestamps and source identifiers, and transmitted to the intelligent fault detection module using a symmetric encryption algorithm.

4. The intelligent diagnosis system of a power distribution network fault self-diagnosis switch device according to claim 3, characterized in that: In the intelligent fault detection module, the process of processing time series data and training the input layer of the LSTM model includes: Constructing an LSTM model, wherein the LSTM model includes an input layer, a hidden layer, and an output layer; The input layer receives the time series data of the power parameter monitoring module after preprocessing and feature extraction, and the time series data includes current data, voltage data, power factor and harmonic content. The time series data is divided into a time window containing the past T time steps and the current, voltage, power factor and harmonic content features to form a multidimensional array X = [x1, x2, ..., x t ], where x t Represents the data vector at the t-th time step. Each feature is normalized to make the numerical ranges of different features consistent. The calculation process is as follows: Among them, x' t represents the normalized eigenvalue, μ represents the eigenmean, and σ represents the standard deviation; Map the data vector of each time window to the input space of the LSTM model to form a three-dimensional tensor X input ∈R T×F , where T represents the time window length and F represents the number of features. For each new time window, the hidden state h0 and cell state C0 of the LSTM unit are initialized to zero vectors.

5. The intelligent diagnosis system of the power distribution network fault self-diagnosis switch device according to claim 4 is characterized in that: In the intelligent fault detection module, the process of training the hidden layer of the LSTM model includes: The hidden layer uses the gating mechanism in its internal long short-term memory cell state to capture its long-term dependencies over time, and the gating mechanism includes a forget gate, an input gate, a cell state update, and an output gate; The forget gate is based on the historical hidden state h t-1 and the current time step x t The input of the gate is used to evaluate and decide whether to retain the past state information. If the current and voltage remain stable in the past fixed-length period of time, the forget gate chooses to discard the old current and voltage information. If the current and voltage are unstable in the past fixed-length period of time, the forget gate chooses to retain the old current and voltage information. The calculation process is as follows: f t =σ(W f ·[h t-1 ;x t ]+b f ); Among them, x t Represents the input vector of the current time step, including current, voltage, power factor and harmonic content, h t-1 represents the output state of the previous time step, W f is the weight matrix corresponding to the forget gate, b f is the bias term, σ is the sigmoid activation function, which maps the result to between 0 and 1, indicating the degree of retention and discarding. t If f is close to 1, the information is retained. t If it is close to 0, the information is discarded; The input gate evaluates the input x at the current time step t Whether to include current surges and power factor changes and allow the changes to enter cell state C t , the calculation process is as follows: i t =σ(W i ·[h t-1 ,x t ]+b i ); Among them, i t Represents the input factor, using the Sigmoid activation function to determine whether each element is added to the cell state. The Tanh activation function is used to generate candidate cell states ranging from -1 to 1, W i and b i Determines the selection mechanism of the input gate, W C and b C Controls the generation of candidate cell states; Combining the results of the forget gate and the input gate, the cell state at the previous moment is bitwise multiplied by the output of the forget gate, and then added to the candidate state generated by the input gate. If a short circuit fault occurs and causes a sudden increase in current, the abnormal situation is reflected by updating the cell state that stores long-term information. The calculation process is as follows: The output gate extracts and outputs information related to the distribution network current and power factor mutation from the cell state as the hidden state h t , the calculation process is as follows: the t =σ(W o ·[h t-1 ,x t ]+b o ); h t =o t ⊙tanh(C t ); Among them, t is the output factor, and the Sigmoid activation function is used to determine whether each element is output. Tanh(C t ) makes the output value fall between -1 and 1, reflecting the change of the current cell state; For each time step t in each time window, the calculation process of the forget gate, input gate and output gate is repeated to gradually update the hidden state h t and cell state C t , capturing dynamic changes in time series.

6. The intelligent diagnosis system of the power distribution network fault self-diagnosis switch device according to claim 5, characterized in that: In the intelligent fault detection module, the process of training the LSTM model output layer includes: The output layer receives the hidden state b from the hidden layer output t , map it to the target output dimension through the fully connected layer inside the output layer, select the activation function according to the task requirements, use the loss function to measure the difference between the LSTM model prediction value and the true value, and adjust the LSTM model parameters through back propagation.

7. The intelligent diagnosis system of the power distribution network fault self-diagnosis switch device according to claim 6, characterized in that: In the intelligent fault detection module, the process of evaluating the performance of the LSTM model using an independent validation set includes: A validation set was separated from the time series data that had not participated in the LSTM model training. The current, voltage, power factor, and harmonic content data in the validation set were preprocessed and feature extracted, and mapped to the input space of the LSTM model. The trained LSTM model was used to predict the validation set data. The mean square error, mean absolute error, and accuracy were selected as performance indicators to measure the difference between the predicted value and the true value. The reliability and stability of the evaluation were improved through the cross-validation method, and the hyperparameters were adjusted to optimize the model performance.

8. The intelligent diagnosis system for a power distribution network fault self-diagnosis switch device according to claim 7, characterized in that: In the intelligent fault detection module, the LSTM model is deployed to determine whether there is a fault and its type based on the newly collected data, foresee potential overload conditions and equipment aging problems, and perform anomaly detection. The process includes: For fault mode recognition, the output layer generates a probability distribution containing two categories: normal and fault. The Softmax activation function is used to make the sum of the probabilities equal to 1. The calculation process is as follows: Among them, z0 represents the unnormalized score corresponding to the normal state, z1 represents the unnormalized score corresponding to the fault state, and the fault probability threshold is set. If P (故障) If the value is lower than the set threshold, the power grid is considered to be in a normal state. (故障) If the set threshold is reached, the power grid is considered to be in a fault state, and the fault type is analyzed, which includes short circuit and ground fault; To foresee potential overload conditions and equipment aging problems, the output layer predicts future current values ​​and uses a linear activation function to maintain the original value range. The calculation process is as follows: in, Indicates the predicted current value of the next time step and sets the safety current threshold. If it is predicted that the current will exceed the safety current threshold of the distribution network at a certain moment in the future, an alarm will be issued in advance to remind the operation and maintenance personnel to take preventive measures; Calculate the root mean square value, peak factor, average value and standard deviation of current data and voltage data, set the corresponding threshold range, and judge it as abnormal when the actual measured value exceeds the threshold range. When the actual measured value does not exceed the threshold range, it is judged as normal. Identify the rapid changes and zero crossing points of current and voltage, and capture potential fault signs. If the power factor changes suddenly and the harmonic content increases abnormally, the intelligent diagnosis system responds and generates an automated response strategy.

9. The intelligent diagnosis system of the power distribution network fault self-diagnosis switch device according to claim 8, characterized in that: In the visual decision support module, the process of providing a graphical user interface includes: A graphical interface is constructed to display information extracted from current, voltage, power factor and harmonic content data, to assist operation and maintenance personnel in fault judgment, foreseeing potential problems and detecting anomalies. The graphical interface integrates real-time collected and pre-processed data, predicts the grid status, future current value and abnormal conditions through the LSTM model, calculates the root mean square value, peak factor, average value and standard deviation, and presents data changes and abnormal markers in trend charts, heat maps, classification result charts and prediction curves. It also provides interactive functions that allow dynamic update, detailed viewing and custom configuration of display parameters. When a fault is detected and potential problems are foreseen, an alarm is automatically triggered and corresponding maintenance suggestions are recommended.

10. The intelligent diagnosis system of the power distribution network fault self-diagnosis switch device according to claim 9, characterized in that: In the visualization decision support module, the process of generating an automated fault response strategy based on the LSTM model analysis results includes: The LSTM model predicts the data collected in real time, outputs the normal and fault probability distribution of the distribution network status and the future current value prediction. When the fault probability exceeds the set threshold, it is determined to be a fault and the fault type is analyzed. If it is predicted that the current will exceed the safe current threshold, an alarm is issued in advance to remind the operation and maintenance personnel to take preventive measures and trigger an automated response strategy. The automated response strategy includes triggering an alarm and displaying a warning message through a graphical interface, pushing a notification to the operation and maintenance personnel, disconnecting related circuits to prevent damage, performing maintenance inspections and replacing equipment in advance, and sending control instructions to the switch device for detected faults to quickly isolate the faulty area and provide continuous power to the non-faulty area.

Images