Power distribution network protection setting decision system and method based on big data

Abstract

The invention discloses a big data-based power distribution network protection setting decision system and method, and relates to the technical field of power distribution network protection, and the system comprises a multi-source data collection module which collects the operation data of a power distribution network; the big data feature processing module is used for carrying out feature extraction and constructing a multi-dimensional feature vector library containing time sequence features and frequency domain features; the protection setting model module is used for constructing a protection setting decision model based on a deep learning framework and carrying out supervised learning training by utilizing historical fault data and a setting scheme; the multi-scene simulation verification module is used for constructing a power distribution network digital simulation model; and the scheme evaluation optimization module is used for establishing an evaluation index system and outputting an optimal protection setting scheme. According to the method, deep learning and historical data are combined, an optimal protection setting scheme is automatically generated, faults are recognized through wavelet transform and Fourier transform, protection parameters are adjusted in real time, mistaken and leaked protection is reduced, timeliness is improved through edge calculation, the strategy is optimized, and the safety of the power distribution network is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network protection, and in particular to a distribution network protection setting decision-making system and method based on big data. Background Art

[0002] As distribution networks continue to expand, traditional protection setting methods struggle to adapt to the complex operating environments and sudden fault conditions of modern distribution networks. Traditional protection setting methods, relying primarily on manual settings and experience, suffer from inaccuracies, poor real-time performance, and difficulty coping with diverse operating conditions. This is particularly true when faced with multiple fault types and load variations, making it difficult for traditional methods to achieve intelligent and dynamic adjustments. To address this, a big data-based distribution network protection setting decision system has emerged. Through the real-time collection and processing of multi-source data, the introduction of deep learning algorithms, and the application of digital simulation verification technology, more accurate protection setting decisions can be achieved. This approach not only enables real-time adjustment of distribution network protection setting parameters, but also enables flexible optimization for different fault types and operating conditions, significantly improving the intelligence and automation of distribution network protection.

[0003] Currently, the field of distribution network protection and setting has begun to explore data-driven intelligent decision-making systems, among which applications based on big data and artificial intelligence have become a trend. For example, methods such as using wavelet transform and Fourier transform to extract fault characteristics, using deep learning training models to make protection setting decisions, and verifying the effectiveness of protection schemes through multi-scenario simulation can effectively improve the protection performance of distribution networks. However, existing technologies still face challenges such as high data processing complexity and high real-time requirements. Summary of the Invention

[0004] In order to solve the above technical problems, a distribution network protection setting decision-making system and method based on big data are provided. This technical solution solves the above problems.

[0005] In order to achieve the above objects, the technical solution adopted by the present invention is: The distribution network protection setting decision system based on big data includes: Multi-source data acquisition module: used to collect distribution network operation data and perform real-time data collection and preprocessing through edge computing nodes; Big data feature processing module: This module is electrically connected to the multi-source data acquisition module, performs feature extraction on the received unstructured data, extracts fault features using a wavelet transform and Fourier transform fusion algorithm, and constructs a multi-dimensional feature vector library containing time series features and frequency domain features. Protection setting model module: This module is electrically connected to the big data feature processing module and builds a protection setting decision model based on a deep learning framework. This model includes an overcurrent protection threshold generation unit, a zero-sequence protection time limit calculation unit, and an adaptive coordination strategy generation unit. It uses historical fault data and setting plans for supervised learning training. Multi-scenario simulation verification module: Connects with the protection setting model module to build a digital simulation model of the distribution network, simulates protection action behaviors under different topologies, load conditions, and fault types, and generates simulated fault current curves and protection coordination timing diagrams; Scheme evaluation and optimization module: Receives simulation verification results, establishes an evaluation index system based on the four dimensions of speed, selectivity, sensitivity and reliability, uses the hierarchical analysis method to determine the index weights, iteratively optimizes the setting scheme through a multi-objective genetic algorithm, and outputs the optimal protection setting scheme.

[0006] Preferably, the big data feature processing module specifically includes: Data cleaning unit: uses the isolation forest algorithm to identify and remove abnormal data points, repairs missing data through sliding window interpolation, and establishes a data quality assessment matrix; Feature extraction unit: Analyzes the cleaned data according to time series and frequency characteristics. First, perform Fourier transform on the data to convert the time domain signal into the frequency domain signal, extract the frequency component of the signal, and then use wavelet transform to perform multi-resolution analysis on the data to obtain signal characteristics at different time scales. Finally, the features obtained from time domain, frequency domain and multi-resolution analysis are integrated to construct a multidimensional feature vector. Data fusion unit: It integrates equipment parameters, historical setting records and real-time operation data in a temporal and spatial manner to generate a comprehensive feature data set including the health status of the equipment.

[0007] Preferably, the protection setting model module specifically includes: Overcurrent protection threshold generation unit: Based on real-time collected distribution network operation data and combined with fault characteristic data, it generates adaptive overcurrent protection thresholds. It uses a deep learning model to analyze historical fault data, determine the overcurrent threshold range, and automatically adjust the protection threshold. Zero-sequence protection time limit calculation unit: obtains the corresponding protection time limit based on zero-sequence current data, and calculates the protection time limit by analyzing historical fault data, combining network topology and load conditions; Adaptive coordination strategy generation unit: Based on the protection setting requirements of different devices, it optimizes the coordination of protection strategies through adaptive algorithms. Through deep learning and supervised learning of historical fault data, it automatically adjusts the action timing and coordination strategies of different protection devices. Deep learning training: Use historical fault data and existing setting plans for supervised learning to train the deep learning model. Combined with the distribution network's operating data and fault data, the model parameters are continuously adjusted to enable it to generate the optimal protection setting plan under different fault conditions.

[0008] Preferably, the overcurrent protection threshold generation unit specifically includes: The eigenvectors related to overcurrent protection are extracted from the multidimensional eigenvector library. The probability density function of the load current during normal operation is established using the kernel density estimation method. The overcurrent threshold prediction model is constructed using the LSTM network to predict the overcurrent threshold. Combined with the real-time fault feature vector, the threshold is dynamically adjusted through the attention mechanism.

[0009] Preferably, the zero-sequence protection time limit calculation unit specifically includes: Extract the zero-sequence current related features from the fault feature vector, represent the distribution network topology as an adjacency matrix, establish a zero-sequence protection time limit calculation model based on fuzzy logic, and use the time limit correction formula to make corrections; The correction formula is: ; Where, is the corrected zero-sequence protection time limit, is the basic zero sequence protection time limit, is the fault distance correction factor, is the actual distance from the fault point to the protection installation location, is the maximum protection range of the zero-sequence protection device, is the load capacity correction factor, is the actual load capacity of the current distribution network, is the rated load capacity of the distribution network.

[0010] Preferably, the multi-scenario simulation verification module specifically includes: The digital simulation model construction unit obtains the actual topological structure data of the distribution network, including node connection relationships, line parameters, and transformer parameters. Based on the principles of graph theory, the topological structure is converted into a mathematical model, and a node admittance matrix is ​​constructed. Based on this, the basic digital model framework of the distribution network is built and the equipment parameters are entered. Model initialization unit: According to the normal operating conditions of the distribution network, the initial operating state parameters of the model are set to complete the initialization of the digital simulation model, simulate the topological structure changes during the operation of the distribution network, and generate simulation scenarios under different topological structures by modifying the node connection relationship and corresponding parameters; Load condition setting steps: Based on historical load data and load forecast results, set different load change scenarios, set seasonal load changes, daily load curve changes, and sudden load fluctuation conditions, adjust the load power parameters of each node in the model, and simulate the distribution network operation status under different load levels; Distributed power access scenario setting steps: Considering the access situation of distributed power, set distributed power scenarios of different types, capacities and access locations, and connect them to the distribution network digital model; Fault scenario generation unit: Based on the fault type, the fault location and transition resistance parameters are determined, fault conditions of different severity and characteristics are simulated, the time point and duration of the fault are set, and the fault conditions at different operation stages of the distribution network are simulated. The action response characteristics of the protection device under different time conditions are studied, and the fault location and transition resistance parameters are randomized to generate random fault parameter combinations. Protection action simulation unit: Based on the protection device parameters output by the protection setting model module, the protection device is modeled in the simulation model. After setting the operation scenario and fault scenario, the simulation calculation program is started. Based on the digital simulation model and the protection device model, the current and voltage changes of each node after the fault occurs, as well as the action behavior of the protection device, are calculated. The data during the simulation process is recorded in real time, and the simulation results are displayed in a visual form.

[0011] Preferably, the solution evaluation and optimization module specifically includes: Weight determination unit: Scoring is performed based on a dedicated questionnaire for the hierarchical analysis method. Experts score the relative importance of the evaluation indicators to obtain the weight of each evaluation dimension and the final comprehensive evaluation results; The calculation formula for the comprehensive evaluation results is: ; Where, To comprehensively evaluate the results, is the evaluation value of the speed indicator, is the weight of the evaluation value of the speed indicator, is the evaluation value of the selectivity index, is the weight of the evaluation value of the selectivity index, is the evaluation value of the sensitivity index, is the weight of the evaluation value of the sensitivity index, is the evaluation value of the reliability index, is the weight of the evaluation value of the reliability index; Multi-objective genetic algorithm optimization unit: This unit uses binary coding to encode the parameters in the protection tuning scheme and converts them into a chromosome form that can be processed by the genetic algorithm. This forms a complete chromosome individual. Each individual corresponds to a protection tuning scheme. Based on the evaluation index system and the determined weights, a fitness function is constructed. Optimal solution output unit: The candidate solutions obtained by multi-objective genetic algorithm optimization are further screened in combination with actual engineering constraints and the optimal solution is output.

[0012] The big data-based distribution network protection setting decision-making method includes: S1: Use edge computing nodes to collect distribution network operation data, equipment status data, historical fault data, and environmental data in real time and perform preliminary preprocessing; S2: Clean the preprocessed data and extract multimodal features to build a multidimensional feature vector library; S3: Based on the multi-dimensional feature vector library, the overcurrent protection threshold, zero-sequence protection time limit, and differential protection parameters are calculated through the protection setting decision model; S4: Build a digital simulation model of the distribution network, simulate the protection action behavior under different working conditions, and generate simulation verification data; S5: Based on the simulation verification data, the setting scheme is evaluated from four dimensions and iteratively optimized using a multi-objective particle swarm optimization algorithm to obtain the optimal protection setting scheme; S6: Display the optimal solution through the human-computer interaction interface, and support user adjustment and solution distribution.

[0013] Preferably, the S2 specifically includes: S201: Identify and remove abnormal data points using the isolation forest algorithm, repair missing data using the sliding window interpolation method, and establish a data quality assessment matrix; S202: Identify and remove abnormal data points using the isolation forest algorithm, repair missing data using the sliding window interpolation method, and establish a data quality assessment matrix; S203: Perform multi-resolution analysis on the data using wavelet transform to obtain signal features at different time scales, fuse the features obtained from time domain, frequency domain and multi-resolution analysis, and construct a multi-dimensional feature vector; S204: Equipment parameters, historical setting records, and real-time operation data are temporally and spatially correlated and integrated to generate a comprehensive feature data set including the health status of the equipment.

[0014] Preferably, the S3 specifically includes: S301: Extract feature vectors related to overcurrent protection from a multidimensional feature vector library, establish a probability density function of the load current during normal operation using a kernel density estimation method, and construct an overcurrent threshold prediction model using an LSTM network to predict the overcurrent threshold. S302: Dynamically adjust the overcurrent protection threshold using an attention mechanism based on the real-time fault feature vector; S303: Extracting zero-sequence current related features from the fault feature vector, representing the distribution network topology as an adjacency matrix, establishing a zero-sequence protection time limit calculation model based on fuzzy logic, and performing correction using a time limit correction formula; S304: Based on the protection setting requirements of different devices, the coordination of protection strategies is optimized through adaptive algorithms. Through deep learning and supervised learning of historical fault data, the action timing and coordination strategies of different protection devices are automatically adjusted.

[0015] Compared with the prior art, the present invention has the following beneficial effects: This invention proposes to combine deep learning with historical data to automatically generate the optimal protection setting plan, improve the protection accuracy and real-time performance, accurately identify the fault type through algorithms such as wavelet transform and Fourier transform, and adjust the protection parameters in real time to reduce false protection and missed protection. It uses edge computing to ensure the timeliness of data processing, improves adaptability based on a dynamic adjustment mechanism, and optimizes the protection strategy through multi-scenario simulation verification. It combines a multi-objective genetic algorithm to improve the effectiveness of the plan. The equipment health management function monitors the equipment status in real time, warns of faults in advance, optimizes operation and maintenance, and enhances the safety and reliability of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a system framework diagram of the present invention; Figure 2 It is a flow chart of the steps of the present invention. DETAILED DESCRIPTION

[0017] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0018] Reference Figure 1 As shown in the figure, the distribution network protection setting decision system based on big data includes: Multi-source data acquisition module: used to collect distribution network operation data and perform real-time data collection and preprocessing through edge computing nodes; Big data feature processing module: This module is electrically connected to the multi-source data acquisition module, performs feature extraction on the received unstructured data, extracts fault features using a wavelet transform and Fourier transform fusion algorithm, and constructs a multi-dimensional feature vector library containing time series features and frequency domain features. Protection setting model module: This module is electrically connected to the big data feature processing module and builds a protection setting decision model based on a deep learning framework. This model includes an overcurrent protection threshold generation unit, a zero-sequence protection time limit calculation unit, and an adaptive coordination strategy generation unit. It uses historical fault data and setting plans for supervised learning training. Multi-scenario simulation verification module: Connects with the protection setting model module to build a digital simulation model of the distribution network, simulates protection action behaviors under different topologies, load conditions, and fault types, and generates simulated fault current curves and protection coordination timing diagrams; Scheme evaluation and optimization module: Receives simulation verification results, establishes an evaluation index system based on the four dimensions of speed, selectivity, sensitivity and reliability, uses the hierarchical analysis method to determine the index weights, iteratively optimizes the setting scheme through a multi-objective genetic algorithm, and outputs the optimal protection setting scheme.

[0019] The big data feature processing module specifically includes: Data cleaning unit: uses the isolation forest algorithm to identify and remove abnormal data points, repairs missing data through sliding window interpolation, and establishes a data quality assessment matrix; Feature extraction unit: Analyzes the cleaned data according to time series and frequency characteristics. First, perform Fourier transform on the data to convert the time domain signal into the frequency domain signal, extract the frequency component of the signal, and then use wavelet transform to perform multi-resolution analysis on the data to obtain signal characteristics at different time scales. Finally, the features obtained from time domain, frequency domain and multi-resolution analysis are integrated to construct a multidimensional feature vector. Among them, the Fourier transform formula is:

[0020] Where, is the result of discrete Fourier transform, To correspond to different frequency components, is the original discrete-time signal, which is a signal with a length of sequence, where is the time domain index, is a discrete-time signal The length of , that is, the number of sampling points, is a natural constant, is the imaginary unit, is the frequency domain index, is the phase part of the exponential function; Data fusion unit: It integrates equipment parameters, historical setting records and real-time operation data in a temporal and spatial manner to generate a comprehensive feature data set including the health status of the equipment; The system uses the isolation forest algorithm for data cleaning, and combines Fourier transform and wavelet transform to implement time domain, frequency domain and multi-resolution analysis to generate a multi-dimensional feature vector library. At the same time, it performs spatiotemporal correlation fusion to form a feature set that includes the health status of the equipment, providing multi-level support for subsequent decision-making.

[0021] The protection setting model module specifically includes: Overcurrent protection threshold generation unit: Based on real-time collected distribution network operation data and combined with fault characteristic data, it generates adaptive overcurrent protection thresholds. It uses a deep learning model to analyze historical fault data, determine the overcurrent threshold range, and automatically adjust the protection threshold. Zero-sequence protection time limit calculation unit: obtains the corresponding protection time limit based on zero-sequence current data, and calculates the protection time limit by analyzing historical fault data, combining network topology and load conditions; Adaptive coordination strategy generation unit: Based on the protection setting requirements of different devices, it optimizes the coordination of protection strategies through adaptive algorithms. Through deep learning and supervised learning of historical fault data, it automatically adjusts the action timing and coordination strategies of different protection devices. Deep learning training: Utilizes historical fault data and existing protection setting plans for supervised learning to train deep learning models. Combined with distribution network operating data and fault data, the model parameters are continuously adjusted to generate the optimal protection setting plan under different fault conditions. This module uses a deep learning framework to perform supervised learning based on historical fault data, builds overcurrent protection threshold generation, zero-sequence protection time limit calculation, and adaptive coordination strategy generation units to achieve automated protection setting decisions.

[0022] The overcurrent protection threshold generation unit specifically includes: The eigenvectors related to overcurrent protection are extracted from the multidimensional eigenvector library. The probability density function of the load current during normal operation is established using the kernel density estimation method. The overcurrent threshold prediction model is constructed using the LSTM network to predict the overcurrent threshold. Combined with the real-time fault feature vector, the threshold is dynamically adjusted through the attention mechanism; The dynamic adjustment formula is:

[0023] Where, The final overcurrent protection threshold is determined by is the basic overcurrent protection threshold predicted by the deep learning model, is the number of features contained in the fault feature vector, For the The attention weight of each fault feature, is the first eigenvalues, For the The trainable weight parameters of the fault features, is to sum the exponential terms of all fault features; The LSTM network is used in combination with historical data to predict the overcurrent threshold, and the threshold is dynamically adjusted through the attention mechanism, so that the protection system can respond accurately according to real-time fault data.

[0024] The zero-sequence protection time limit calculation unit specifically includes: Extract the zero-sequence current related features from the fault feature vector, represent the distribution network topology as an adjacency matrix, establish a zero-sequence protection time limit calculation model based on fuzzy logic, and use the time limit correction formula to make corrections; The correction formula is:

[0025] Where, is the corrected zero-sequence protection time limit, is the basic zero sequence protection time limit, is the fault distance correction factor, is the actual distance from the fault point to the protection installation location, is the maximum protection range of the zero-sequence protection device, is the load capacity correction factor, is the actual load capacity of the current distribution network, is the rated load capacity of the distribution network.

[0026] Fuzzy logic is used to calculate the zero-sequence protection time limit, and the time limit correction formula is used to correct the time limit, taking into account the load capacity and the actual distance from the fault point, so that the time limit calculation is more in line with the actual working conditions.

[0027] The multi-scenario simulation verification module specifically includes: The digital simulation model construction unit obtains the actual topological structure data of the distribution network, including node connection relationships, line parameters, and transformer parameters. Based on the principles of graph theory, the topological structure is converted into a mathematical model, and a node admittance matrix is ​​constructed. Based on this, the basic digital model framework of the distribution network is built and the equipment parameters are entered. Model initialization unit: According to the normal operating conditions of the distribution network, the initial operating state parameters of the model are set to complete the initialization of the digital simulation model, simulate the topological structure changes during the operation of the distribution network, and generate simulation scenarios under different topological structures by modifying the node connection relationship and corresponding parameters; Load condition setting steps: Based on historical load data and load forecast results, set different load change scenarios, set seasonal load changes, daily load curve changes, and sudden load fluctuation conditions, adjust the load power parameters of each node in the model, and simulate the distribution network operation status under different load levels; Distributed power access scenario setting steps: Considering the access situation of distributed power, set distributed power scenarios of different types, capacities and access locations, and connect them to the distribution network digital model; Fault scenario generation unit: Based on the fault type, the fault location and transition resistance parameters are determined, fault conditions of different severity and characteristics are simulated, the time point and duration of the fault are set, and the fault conditions at different operation stages of the distribution network are simulated. The action response characteristics of the protection device under different time conditions are studied, and the fault location and transition resistance parameters are randomized to generate random fault parameter combinations. Protection action simulation unit: Based on the protection device parameters output by the protection setting model module, the protection device is modeled in the simulation model. After setting the operation scenario and fault scenario, the simulation calculation program is started. Based on the digital simulation model and the protection device model, the current and voltage changes at each node after the fault occurs, as well as the action behavior of the protection device, are calculated. The data during the simulation process is recorded in real time, and the simulation results are displayed in a visual form. This module uses digital simulation models to simulate various scenarios such as different topologies, load conditions, and distributed power supply access, generating simulation data for different fault types and protection device responses, providing all-round support for setting scheme optimization.

[0028] The solution evaluation and optimization module specifically includes: Weight determination unit: Scoring is performed based on a dedicated questionnaire for the hierarchical analysis method. Experts score the relative importance of the evaluation indicators to obtain the weight of each evaluation dimension and the final comprehensive evaluation results; The calculation formula for the comprehensive evaluation results is: ; Where, To comprehensively evaluate the results, is the evaluation value of the speed indicator, is the weight of the evaluation value of the speed indicator, is the evaluation value of the selectivity index, is the weight of the evaluation value of the selectivity index, is the evaluation value of the sensitivity index, is the weight of the evaluation value of the sensitivity index, is the evaluation value of the reliability index, is the weight of the evaluation value of the reliability index; Multi-objective genetic algorithm optimization unit: This unit uses binary coding to encode the parameters in the protection tuning scheme and converts them into a chromosome form that can be processed by the genetic algorithm. This forms a complete chromosome individual. Each individual corresponds to a protection tuning scheme. Based on the evaluation index system and the determined weights, a fitness function is constructed. Optimal solution output unit: The candidate solutions obtained by multi-objective genetic algorithm optimization are further screened in combination with actual engineering constraints and the optimal solution is output; The analytic hierarchy process is used to assign weights to evaluation indicators, and the tuning scheme is optimized through a multi-objective genetic algorithm. This not only ensures the speed, selectivity, sensitivity and reliability of the tuning scheme, but also allows for flexible adjustments based on engineering requirements.

[0029] Reference Figure 2 As shown in the figure, the distribution network protection setting decision-making method based on big data includes: S1: Use edge computing nodes to collect distribution network operation data, equipment status data, historical fault data, and environmental data in real time and perform preliminary preprocessing; S2: Clean the preprocessed data and extract multimodal features to build a multidimensional feature vector library; S3: Based on the multi-dimensional feature vector library, the overcurrent protection threshold, zero-sequence protection time limit, and differential protection parameters are calculated through the protection setting decision model; S4: Build a digital simulation model of the distribution network, simulate the protection action behavior under different working conditions, and generate simulation verification data; S5: Based on the simulation verification data, the setting scheme is evaluated from four dimensions and iteratively optimized using a multi-objective particle swarm optimization algorithm to obtain the optimal protection setting scheme; S6: Display the optimal solution through the human-computer interaction interface, and support user adjustment and solution distribution.

[0030] 9. The distribution network protection setting decision method based on big data according to claim 8, wherein S2 specifically comprises: S201: Identify and remove abnormal data points using the isolation forest algorithm, repair missing data using the sliding window interpolation method, and establish a data quality assessment matrix; S202: Identify and remove abnormal data points using the isolation forest algorithm, repair missing data using the sliding window interpolation method, and establish a data quality assessment matrix; S203: Perform multi-resolution analysis on the data using wavelet transform to obtain signal features at different time scales, fuse the features obtained from time domain, frequency domain and multi-resolution analysis, and construct a multi-dimensional feature vector; S204: Equipment parameters, historical setting records, and real-time operation data are temporally and spatially correlated and integrated to generate a comprehensive feature data set including the health status of the equipment.

[0031] 10. The distribution network protection setting decision method based on big data according to claim 8, wherein S3 specifically comprises: S301: Extract feature vectors related to overcurrent protection from a multidimensional feature vector library, establish a probability density function of the load current during normal operation using a kernel density estimation method, and construct an overcurrent threshold prediction model using an LSTM network to predict the overcurrent threshold. S302: Dynamically adjust the overcurrent protection threshold using an attention mechanism based on the real-time fault feature vector; S303: Extracting zero-sequence current related features from the fault feature vector, representing the distribution network topology as an adjacency matrix, establishing a zero-sequence protection time limit calculation model based on fuzzy logic, and performing correction using a time limit correction formula; S304: Based on the protection setting requirements of different devices, the coordination of protection strategies is optimized through adaptive algorithms. Through deep learning and supervised learning of historical fault data, the action timing and coordination strategies of different protection devices are automatically adjusted.

[0032] In summary, the advantages of the present invention are: By combining deep learning with historical data, the system can automatically generate the optimal protection setting plan, avoiding the errors and lags of traditional manual settings and ensuring that distribution network protection measures can be dynamically adjusted according to real-time operating conditions. By fusing multiple algorithms such as wavelet transform and Fourier transform to extract fault characteristics, it can efficiently and accurately identify fault types, adjust protection thresholds and time limits in real time, reduce misprotection and missed protection, and improve the stability of the distribution network. The system uses edge computing nodes to collect and preprocess data in real time, ensuring the timeliness of data processing. At the same time, the dynamic adjustment mechanism based on deep learning can flexibly adjust protection strategies according to different fault types and operating conditions, improving the adaptability of the distribution network. The multi-scenario simulation verification module can simulate the protection action behavior under different topologies, load conditions and fault types, providing actual data support for the protection setting plan and ensuring the feasibility and effectiveness of the plan; The multi-objective genetic algorithm is used to optimize the protection setting scheme, taking into account multiple indicators such as speed, selectivity, sensitivity and reliability. It can output the optimal scheme and further improve the quality and effectiveness of distribution network protection setting. The system combines spatiotemporal data with equipment health status to monitor equipment status in real time, detect potential faults in advance, and optimize maintenance plans, thereby extending equipment life and reducing operation and maintenance costs. Through the comprehensive application of big data, deep learning and simulation technologies, this system can effectively improve the safety and reliability of the distribution network under complex working conditions and reduce accidents caused by improper protection of the power system.

[0033] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. The distribution network protection setting decision system based on big data is characterized by: include: Multi-source data acquisition module: used to collect distribution network operation data and perform real-time data collection and preprocessing through edge computing nodes; Big data feature processing module: This module is electrically connected to the multi-source data acquisition module, performs feature extraction on the received unstructured data, extracts fault features using a wavelet transform and Fourier transform fusion algorithm, and constructs a multi-dimensional feature vector library containing time series features and frequency domain features. Protection setting model module: This module is electrically connected to the big data feature processing module and builds a protection setting decision model based on a deep learning framework. This model includes an overcurrent protection threshold generation unit, a zero-sequence protection time limit calculation unit, and an adaptive coordination strategy generation unit. It uses historical fault data and setting plans for supervised learning training. Multi-scenario simulation verification module: Connects with the protection setting model module to build a digital simulation model of the distribution network, simulates protection action behaviors under different topologies, load conditions, and fault types, and generates simulated fault current curves and protection coordination timing diagrams; Scheme evaluation and optimization module: Receives simulation verification results, establishes an evaluation index system based on the four dimensions of speed, selectivity, sensitivity and reliability, uses the hierarchical analysis method to determine the index weights, iteratively optimizes the setting scheme through a multi-objective genetic algorithm, and outputs the optimal protection setting scheme.

2. The distribution network protection setting decision system based on big data according to claim 1 is characterized in that: The big data feature processing module specifically includes: Data cleaning unit: uses the isolation forest algorithm to identify and remove abnormal data points, repairs missing data through sliding window interpolation, and establishes a data quality assessment matrix; Feature extraction unit: Analyzes the cleaned data according to time series and frequency characteristics. First, perform Fourier transform on the data to convert the time domain signal into the frequency domain signal, extract the frequency component of the signal, and then use wavelet transform to perform multi-resolution analysis on the data to obtain signal characteristics at different time scales. Finally, the features obtained from time domain, frequency domain and multi-resolution analysis are integrated to construct a multidimensional feature vector. Data fusion unit: It integrates equipment parameters, historical setting records and real-time operation data in a temporal and spatial manner to generate a comprehensive feature data set including the health status of the equipment.

3. The distribution network protection setting decision system based on big data according to claim 2 is characterized in that: The protection setting model module specifically includes: Overcurrent protection threshold generation unit: Based on real-time collected distribution network operation data and combined with fault characteristic data, it generates adaptive overcurrent protection thresholds. It uses a deep learning model to analyze historical fault data, determine the overcurrent threshold range, and automatically adjust the protection threshold. Zero-sequence protection time limit calculation unit: obtains the corresponding protection time limit based on zero-sequence current data, and calculates the protection time limit by analyzing historical fault data, combining network topology and load conditions; Adaptive coordination strategy generation unit: Based on the protection setting requirements of different devices, it optimizes the coordination of protection strategies through adaptive algorithms. Through deep learning and supervised learning of historical fault data, it automatically adjusts the action timing and coordination strategies of different protection devices. Deep learning training: Use historical fault data and existing setting plans for supervised learning to train the deep learning model. Combined with the distribution network's operating data and fault data, the model parameters are continuously adjusted to enable it to generate the optimal protection setting plan under different fault conditions.

4. The distribution network protection setting decision system based on big data according to claim 3 is characterized in that: The overcurrent protection threshold generation unit specifically includes: The eigenvectors related to overcurrent protection are extracted from the multidimensional eigenvector library. The probability density function of the load current during normal operation is established using the kernel density estimation method. The overcurrent threshold prediction model is constructed using the LSTM network to predict the overcurrent threshold. Combined with the real-time fault feature vector, the threshold is dynamically adjusted through the attention mechanism.

5. The distribution network protection setting decision system based on big data according to claim 3 is characterized in that: The zero-sequence protection time limit calculation unit specifically includes: Extract the zero-sequence current related features from the fault feature vector, represent the distribution network topology as an adjacency matrix, establish a zero-sequence protection time limit calculation model based on fuzzy logic, and use the time limit correction formula to make corrections; The correction formula is: ； Where, is the corrected zero-sequence protection time limit, is the basic zero sequence protection time limit, is the fault distance correction factor, is the actual distance from the fault point to the protection installation location, is the maximum protection range of the zero-sequence protection device, is the load capacity correction factor, is the actual load capacity of the current distribution network, is the rated load capacity of the distribution network.

6. The distribution network protection setting decision system based on big data according to claim 1 is characterized in that: The multi-scenario simulation verification module specifically includes: The digital simulation model construction unit obtains the actual topological structure data of the distribution network, including node connection relationships, line parameters, and transformer parameters. Based on the principles of graph theory, the topological structure is converted into a mathematical model, and a node admittance matrix is ​​constructed. Based on this, the basic digital model framework of the distribution network is built and the equipment parameters are entered. Model initialization unit: According to the normal operating conditions of the distribution network, the initial operating state parameters of the model are set to complete the initialization of the digital simulation model, simulate the topological structure changes during the operation of the distribution network, and generate simulation scenarios under different topological structures by modifying the node connection relationship and corresponding parameters; Load condition setting steps: Based on historical load data and load forecast results, set different load change scenarios, set seasonal load changes, daily load curve changes, and sudden load fluctuation conditions, adjust the load power parameters of each node in the model, and simulate the distribution network operation status under different load levels; Distributed power access scenario setting steps: Considering the access situation of distributed power, set distributed power scenarios of different types, capacities and access locations, and connect them to the distribution network digital model; Fault scenario generation unit: Based on the fault type, the fault location and transition resistance parameters are determined, fault conditions of different severity and characteristics are simulated, the time point and duration of the fault are set, and the fault conditions at different operation stages of the distribution network are simulated. The action response characteristics of the protection device under different time conditions are studied, and the fault location and transition resistance parameters are randomized to generate random fault parameter combinations. Protection action simulation unit: Based on the protection device parameters output by the protection setting model module, the protection device is modeled in the simulation model. After setting the operation scenario and fault scenario, the simulation calculation program is started. Based on the digital simulation model and the protection device model, the current and voltage changes of each node after the fault occurs, as well as the action behavior of the protection device, are calculated. The data during the simulation process is recorded in real time, and the simulation results are displayed in a visual form.

7. The distribution network protection setting decision system based on big data according to claim 1 is characterized in that: The scheme evaluation and optimization module specifically includes: Weight determination unit: Scoring is performed based on a dedicated questionnaire for the hierarchical analysis method. Experts score the relative importance of the evaluation indicators to obtain the weight of each evaluation dimension and obtain the final comprehensive evaluation results; The calculation formula for the comprehensive evaluation results is: ； Where, To comprehensively evaluate the results, is the evaluation value of the speed indicator, is the weight of the evaluation value of the speed indicator, is the evaluation value of the selectivity index, is the weight of the evaluation value of the selectivity index, is the evaluation value of the sensitivity index, is the weight of the evaluation value of the sensitivity index, is the evaluation value of the reliability index, is the weight of the evaluation value of the reliability index; Multi-objective genetic algorithm optimization unit: This unit uses binary coding to encode the parameters in the protection tuning scheme and converts them into a chromosome form that can be processed by the genetic algorithm. This forms a complete chromosome individual. Each individual corresponds to a protection tuning scheme. Based on the evaluation index system and the determined weights, a fitness function is constructed. Optimal solution output unit: The candidate solutions obtained by multi-objective genetic algorithm optimization are further screened in combination with actual engineering constraints and the optimal solution is output.

8. A distribution network protection setting decision-making method based on big data, which is applicable to the digital twin simulation system for asphalt pavement repair according to any one of claims 1 to 7, characterized in that: include: S1: Use edge computing nodes to collect distribution network operation data, equipment status data, historical fault data, and environmental data in real time and perform preliminary preprocessing; S2: Clean the preprocessed data and extract multimodal features to build a multidimensional feature vector library; S3: Based on the multi-dimensional feature vector library, the overcurrent protection threshold, zero-sequence protection time limit, and differential protection parameters are calculated through the protection setting decision model; S4: Build a digital simulation model of the distribution network, simulate the protection action behavior under different working conditions, and generate simulation verification data; S5: Based on the simulation verification data, the setting scheme is evaluated from four dimensions and iteratively optimized using a multi-objective particle swarm optimization algorithm to obtain the optimal protection setting scheme; S6: Display the optimal solution through the human-computer interaction interface, and support user adjustment and solution distribution.

9. The distribution network protection setting decision method based on big data according to claim 8 is characterized in that: The S2 specifically includes: S201: Identify and remove abnormal data points using the isolation forest algorithm, repair missing data using the sliding window interpolation method, and establish a data quality assessment matrix; S202: Identify and remove abnormal data points using the isolation forest algorithm, repair missing data using the sliding window interpolation method, and establish a data quality assessment matrix; S203: Perform multi-resolution analysis on the data using wavelet transform to obtain signal features at different time scales, fuse the features obtained from time domain, frequency domain and multi-resolution analysis, and construct a multi-dimensional feature vector; S204: Equipment parameters, historical setting records, and real-time operation data are temporally and spatially correlated and integrated to generate a comprehensive feature data set including the health status of the equipment.

10. The distribution network protection setting decision method based on big data according to claim 8, characterized in that: The S3 specifically includes: S301: Extract feature vectors related to overcurrent protection from a multidimensional feature vector library, establish a probability density function of the load current during normal operation using a kernel density estimation method, and construct an overcurrent threshold prediction model using an LSTM network to predict the overcurrent threshold. S302: Dynamically adjust the overcurrent protection threshold using an attention mechanism based on the real-time fault feature vector; S303: Extracting zero-sequence current related features from the fault feature vector, representing the distribution network topology as an adjacency matrix, establishing a zero-sequence protection time limit calculation model based on fuzzy logic, and performing correction using a time limit correction formula; S304: Based on the protection setting requirements of different devices, the coordination of protection strategies is optimized through adaptive algorithms. Through deep learning and supervised learning of historical fault data, the action timing and coordination strategies of different protection devices are automatically adjusted.

Images