Magnetic recloser adaptive control method and system for circuit design

Abstract

The invention relates to the technical field of automatic control, and discloses a magnetic recloser adaptive control method and system for circuit design. The method comprises the steps of collecting multi-dimensional electrical parameters of a distribution line through an intelligent sensor array, constructing a mixed fault feature vector by utilizing a time-frequency domain fusion algorithm, separating instantaneous and permanent fault components by adopting three-stage fault recognition, obtaining fault severity level evaluation based on dynamic weight distribution, and performing fault severity level evaluation based on dynamic weight distribution. And generating a self-adaptive control instruction through grading reclosing strategy matching to execute the opening and closing operation of the magnetic recloser. According to the invention, the technical problem that the magnetic recloser cannot carry out accurate fault recognition and adaptive hierarchical control in a multi-transient fault and permanent fault mixed scene is solved.

Description

Technical Field

[0001] This application relates to the field of automatic control technology, and in particular to an adaptive control method and system for magnetic reclosers used in circuit design. Background Technology

[0002] In existing technologies, adaptive control methods for magnetic reclosers mainly employ a single threshold judgment mechanism based on basic electrical parameters such as current and voltage for fault detection and classification. Reclosing operations are executed through preset fixed time delays and reclosing counts. Traditional control methods typically use simple overcurrent protection and voltage monitoring to identify fault states and perform reclosing control based on empirically set parameters. This method is effective in handling single-type faults and can meet basic distribution network automation requirements.

[0003] However, when multiple faults of different natures occur simultaneously in power distribution lines, traditional methods lack an effective mechanism for accurate fault identification. They cannot accurately distinguish between the mixed characteristics of transient and permanent faults, leading to low fault identification accuracy and a simplistic control strategy. Existing methods use static discrimination thresholds and fixed reclosing parameters, which cannot cope with the nonlinear changes and mutual interference of fault characteristic parameters in mixed fault scenarios. This often results in misjudgments and improper reclosing operations, which not only fail to effectively restore power supply but may also cause secondary impacts on equipment.

[0004] Further analysis revealed that existing technologies lack the ability to extract multi-dimensional features and intelligently separate fault components for mixed fault scenarios. They are unable to establish independent identification mechanisms for transient and permanent fault components, resulting in an inability to perform accurate fault component analysis and formulate corresponding hierarchical control strategies in complex fault environments. Furthermore, existing technologies lack dynamic weight allocation and adaptive parameter adjustment mechanisms, failing to intelligently select reclosing strategies and optimize parameters based on the severity and mutual influence of different fault components. This severely impacts the control performance and power supply reliability of magnetic reclosers in mixed fault scenarios. Summary of the Invention

[0005] This application provides an adaptive control method and system for magnetic reclosers in circuit design, which solves the technical problem that magnetic reclosers cannot perform accurate fault identification and adaptive hierarchical control in mixed scenarios of multiple transient and permanent faults.

[0006] In a first aspect, this application provides an adaptive control method for magnetic reclosers in circuit design. The method includes: acquiring and processing multi-dimensional electrical parameters of a power distribution line using an intelligent sensor array to obtain an original fault dataset; constructing a fault feature vector based on the original fault dataset using a time-frequency domain fusion algorithm to obtain a hybrid fault feature vector; separating the hybrid fault components of the hybrid fault feature vector using a three-level fault identification process to obtain instantaneous fault component data and permanent fault component data; dynamically assigning weights to each fault component based on the instantaneous fault component data and permanent fault component data to obtain a fault quantification assessment result containing instantaneous fault severity levels and permanent fault severity levels; and performing a graded reclosing strategy matching process on the fault quantification assessment result to obtain an adaptive control command containing reclosing delay parameters, reclosing frequency parameters, and protection setting parameters, and executing magnetic recloser opening and closing control operations.

[0007] Optionally, the step of acquiring and processing multi-dimensional electrical parameters of the power distribution line through an intelligent sensor array to obtain the original fault dataset includes: Based on current transformers and voltage transformers, the three-phase current value, three-phase voltage value, zero-sequence current value and negative-sequence current value of the power distribution line are collected and processed in real time to obtain basic electrical parameter data. The basic electrical parameter data is input into a digital signal processor for waveform analysis to obtain dynamic electrical characteristic data including current surge, voltage drop depth, phase angle change rate and frequency offset. The dynamic electrical characteristic data is processed by Fast Fourier Transform to obtain frequency domain electrical parameter data containing the fundamental component, harmonic components, and harmonic distortion rate. The basic electrical parameter data, dynamic electrical characteristic data, and frequency domain electrical parameter data are fused together to obtain the original fault dataset containing time domain electrical quantities, frequency domain electrical quantities, and time-frequency domain electrical quantities.

[0008] Optionally, the step of constructing a hybrid fault feature vector based on the original fault dataset using a time-frequency domain fusion algorithm includes: The time-domain electrical quantities in the original fault dataset are input into the wavelet transform module for multi-scale decomposition processing to obtain time-domain feature component data containing wavelet coefficients at 8 scale levels. The frequency domain electrical quantities in the original fault dataset are processed by Hilbert-Huang transform to obtain frequency domain feature component data containing instantaneous frequency change trajectory and instantaneous amplitude change trajectory; Based on the time-domain feature component data, statistical feature extraction processing is performed on the current surge, voltage drop depth and phase angle change rate to obtain time-domain statistical feature parameters including mean, variance, kurtosis and skewness. Based on the frequency domain characteristic component data, the energy distribution of the fundamental component, each harmonic component, and harmonic distortion rate is calculated to obtain frequency domain energy characteristic parameters including frequency domain energy spectral density and frequency band energy proportion. Based on the time-domain statistical feature parameters and frequency-domain energy feature parameters, a 96-dimensional vector construction process is performed to obtain a hybrid fault feature vector containing 32-dimensional time-domain features, 32-dimensional frequency-domain features, and 32-dimensional time-frequency fusion features.

[0009] Optionally, the step of separating the mixed fault feature vector through a three-level fault identification process to obtain instantaneous fault component data and permanent fault component data includes: The hybrid fault feature vector is input into a radial basis function support vector machine for coarse classification of fault-normal states, resulting in a first-level classification result containing fault state identifiers and normal state identifiers. Based on the fault state identifier in the first-level classification result, the hybrid fault feature vector is classified by a multilayer perceptron neural network to obtain a second-level classification result that includes pure transient fault category, pure permanent fault category and hybrid fault category. Based on the mixed fault categories in the second-level classification results, the mixed fault feature vector is decomposed using a blind source separation algorithm to obtain fault signal separation data containing instantaneous fault component signals and permanent fault component signals; Independent component analysis is performed on the fault signal separation data to obtain instantaneous fault component data containing instantaneous fault component amplitude and instantaneous fault component frequency characteristics, and permanent fault component data containing permanent fault component amplitude and permanent fault component impedance characteristics.

[0010] Optionally, the step of performing multilayer perceptron neural network classification on the hybrid fault feature vector based on the fault state identifier in the first-level classification result to obtain a second-level classification result containing pure transient fault categories, pure permanent fault categories, and hybrid fault categories includes: The hybrid fault feature vector is input into the first hidden layer of 128 neurons and processed by linear transformation and activation function to obtain the output data of the first neural network containing 128-dimensional intermediate features. Based on the output data of the first layer neural network, weight calculation and nonlinear mapping are performed on the second hidden layer of 64 neurons to obtain the output data of the second layer neural network containing 64-dimensional compressed features. Based on the output data of the second layer neural network, feature fusion and dimensionality reduction are performed on the third hidden layer of 32 neurons to obtain the output data of the third layer neural network containing 32 core features; The output data of the third layer neural network is input into the output layer of the three neurons and subjected to softmax normalization to obtain a second-level classification result containing pure instantaneous fault probability values, pure permanent fault probability values, and mixed fault probability values.

[0011] Optionally, the step of dynamically weighting each fault component based on the instantaneous fault component data and the permanent fault component data to obtain a fault quantification assessment result including the instantaneous fault severity level and the permanent fault severity level includes: Based on the instantaneous fault component data, the fault duration, current fluctuation amplitude, voltage drop depth, recovery time, and impact range are weighted and processed to obtain instantaneous fault weight allocation parameters including fault duration weight of 0.3, current fluctuation amplitude weight of 0.25, voltage drop depth weight of 0.2, recovery time weight of 0.15, and impact range weight of 0.1. Based on the permanent fault component data, the degree of insulation resistance reduction, the steady-state value of fault current, the equipment temperature rise, and the fault development trend are weighted and processed to obtain permanent fault weight allocation parameters including a weight of 0.35 for the degree of insulation resistance reduction, a weight of 0.25 for the steady-state value of fault current, a weight of 0.2 for the equipment temperature rise, and a weight of 0.2 for the fault development trend. The instantaneous fault weight allocation parameters and permanent fault weight allocation parameters are input into the weighted fusion calculation module for severity level calculation to obtain fault severity level classification data including level 1-3 minor faults, level 4-6 moderate faults and level 7-10 severe faults. The interaction effect correction coefficient is calculated on the fault severity level classification data to obtain a fault quantitative assessment result that includes the instantaneous fault severity level and the permanent fault severity level.

[0012] Optionally, the step of processing the fault quantification assessment result through a graded reclosing strategy matching process to obtain an adaptive control command including reclosing delay parameters, reclosing count parameters, and protection setting parameters, and executing the magnetic recloser opening and closing control operation, includes: Based on the instantaneous fault severity level and permanent fault severity level in the fault quantification assessment results, strategy type allocation processing is performed for minor faults of level 1-3, moderate faults of level 4-6, and severe faults of level 7-10 to obtain hierarchical reclosing strategy type data including fast reclosing strategy, delayed reclosing strategy, and intelligent discrimination reclosing strategy. Based on the hierarchical reclosing strategy type data, the parameters of the fast reclosing strategy are calculated and processed to obtain fast reclosing parameter configuration data including reclosing delay of 0.3 seconds, maximum number of reclosing times of 3, and reclosing interval of 0.3 seconds-0.6 seconds-1.2 seconds. The hierarchical reclosing strategy type data is input into the delayed reclosing strategy calculation module for instantaneous component proportion judgment processing to obtain delayed reclosing parameter configuration data, which includes a first reclosing delay of 1.5 seconds, a second delay of 3 seconds when the instantaneous component proportion is greater than 70%, and a second delay of 10 seconds when the permanent component proportion is greater than 70%. The fast reclosing parameter configuration data and the delayed reclosing parameter configuration data are processed by control instruction encapsulation to obtain an adaptive control instruction containing reclosing delay parameters, reclosing count parameters and protection setting parameters, and the opening and closing control operation instruction is sent to the magnetic recloser.

[0013] Secondly, this application provides an adaptive control system for a magnetic recloser used in circuit design, the adaptive control system for a magnetic recloser used in circuit design comprising: The data acquisition module is used to collect and process multi-dimensional electrical parameters of the power distribution line through an intelligent sensor array to obtain the original fault dataset. The construction module is used to construct fault feature vectors based on the original fault dataset using a time-frequency domain fusion algorithm, thereby obtaining a hybrid fault feature vector. The separation module is used to separate the mixed fault component data and the permanent fault component data by performing mixed fault component separation processing on the mixed fault feature vector through three-level fault identification. The allocation module is used to perform dynamic weight allocation processing on each fault component based on the instantaneous fault component data and the permanent fault component data, so as to obtain a fault quantitative assessment result including the instantaneous fault severity level and the permanent fault severity level. The control module is used to process the fault quantification assessment results through a hierarchical reclosing strategy to obtain an adaptive control command containing reclosing delay parameters, reclosing count parameters, and protection setting parameters, and to execute the magnetic recloser opening and closing control operation.

[0014] Thirdly, a magnetic re-selector adaptive control device for circuit design is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the magnetic re-selector adaptive control device for circuit design to execute the aforementioned magnetic re-selector adaptive control method for circuit design.

[0015] Fourthly, a computer-readable storage medium is provided, wherein instructions are stored therein, which, when executed on a computer, cause the computer to perform the above-described adaptive control method for magnetic reconciliation in circuit design.

[0016] The technical solution provided in this application obtains a raw fault dataset by acquiring and processing multi-dimensional electrical parameters of the power distribution line through an intelligent sensor array. Compared with existing technologies that only collect basic electrical quantities, this approach can comprehensively acquire multi-dimensional feature information such as current surges, voltage dips, phase angle change rates, zero-sequence components, and negative-sequence components, laying a data foundation for subsequent accurate fault analysis. Based on the raw fault dataset, a hybrid fault feature vector is obtained by constructing and processing fault feature vectors through a time-frequency domain fusion algorithm. This overcomes the limitations of single-domain analysis in existing technologies. By constructing a 96-dimensional feature vector using signal processing techniques such as wavelet transform and Hilbert-Huang transform, it can simultaneously capture the time-domain dynamic characteristics, frequency-domain distribution characteristics, and time-frequency coupling characteristics of the fault, significantly improving the completeness and accuracy of fault feature representation. This method separates mixed fault feature vectors into instantaneous and permanent fault component data through a three-level fault identification process, solving the technical challenge of effectively separating mixed fault components in existing technologies. By employing a progressive processing approach using radial basis function support vector machines, multilayer perceptron neural networks, and blind source separation algorithms, accurate identification and component separation of complex mixed faults are achieved, improving fault identification accuracy by more than 15% compared to existing technologies. Dynamic weight allocation is applied to each fault component based on the instantaneous and permanent fault component data to obtain a quantitative fault assessment result. This overcomes the shortcomings of existing technologies, such as fixed weights and singular assessments. By establishing a differentiated weight system and interaction correction mechanism for different fault components, the severity of mixed faults can be accurately quantified, providing a scientific basis for subsequent control strategy formulation.

[0017] In the field of adaptive control methods for magnetic reclosers used in circuit design, the results of fault quantification assessment are processed through hierarchical reclosing strategy matching to obtain adaptive control commands and execute the magnetic recloser opening and closing control operations. A three-level control system is established, consisting of a fast reclosing strategy, a delayed reclosing strategy, and an intelligent discrimination reclosing strategy. The system automatically selects the most suitable reclosing strategy and parameter configuration based on the fault characteristics of different severity levels, avoiding the drawbacks of the "one-size-fits-all" approach in existing technologies. The application of time-frequency domain fusion algorithms in fault feature extraction enables the system to accurately identify the high-frequency transient characteristics of instantaneous faults and the low-frequency steady-state characteristics of permanent faults. The introduction of blind source separation algorithms in the three-level fault identification system achieves effective decomposition of mixed signals, and dynamic weight allocation algorithms automatically adjust control parameters according to the proportion of fault components. The synergistic effect of these algorithmic features enables the magnetic recloser to make intelligent decisions in complex fault environments, significantly improving the power supply reliability and equipment protection level of the distribution network. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of an embodiment of the adaptive control method for magnetic reclosers used in circuit design in this application. Figure 2 This is a schematic diagram of an embodiment of the adaptive control system for magnetic re-closer used in circuit design in this application. Figure 3 This is a schematic block diagram of the structure of the adaptive control device for magnetic reconciliation used in circuit design in an embodiment of the present invention. Detailed Implementation

[0020] This application provides an adaptive control method and system for magnetic reclosers in circuit design. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0021] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the adaptive control method for magnetic reclosers used in circuit design in this application includes: Step S101: Collect and process multi-dimensional electrical parameters of the power distribution line through an intelligent sensor array to obtain the original fault dataset; Step S102: Based on the original fault dataset, the fault feature vector is constructed using a time-frequency domain fusion algorithm to obtain a hybrid fault feature vector; Step S103: The mixed fault feature vector is processed by three-level fault identification to separate the mixed fault components, and instantaneous fault component data and permanent fault component data are obtained. Step S104: Perform dynamic weight allocation processing on each fault component based on the instantaneous fault component data and the permanent fault component data to obtain a fault quantification assessment result that includes the instantaneous fault severity level and the permanent fault severity level. Step S105: The fault quantification assessment results are processed by the hierarchical reclosing strategy to obtain an adaptive control command containing reclosing delay parameters, reclosing count parameters and protection setting parameters, and the magnetic recloser opening and closing control operation is executed.

[0022] It is understood that the executing entity of this application can be a magnetic recloser adaptive control system used for circuit design, or it can be a terminal or a server; the specific implementation is not limited here. This application's embodiments use a server as an example for illustration.

[0023] Specifically, the intelligent sensor array first monitors the power distribution line in real time based on current transformers and voltage transformers, collecting basic electrical parameter data such as three-phase current values, three-phase voltage values, zero-sequence current values, and negative-sequence current values. Then, these data are input into a digital signal processor for waveform analysis, extracting dynamic electrical characteristic data such as current surge, voltage drop depth, phase angle change rate, and frequency offset. Through fast Fourier transform processing, frequency domain electrical parameter data such as fundamental component, harmonic components, and harmonic distortion rate are obtained. Finally, the time domain, frequency domain, and time-frequency domain electrical quantities are fused to form the original fault dataset.

[0024] The time-frequency domain fusion algorithm extracts features from the original fault dataset. It inputs the time-domain electrical quantities into the wavelet transform module for multi-scale decomposition, generating wavelet coefficients at eight scale levels as time-domain feature components. Simultaneously, it performs Hilbert-Huang transform on the frequency-domain electrical quantities to obtain frequency-domain feature components such as instantaneous frequency change trajectory and instantaneous amplitude change trajectory. Based on the time-domain feature components, it extracts statistical features of current surge, voltage drop depth, and phase angle change rate, calculating time-domain statistical feature parameters such as mean, variance, kurtosis, and skewness. Based on the frequency-domain feature components, it calculates the energy distribution of the fundamental component, each harmonic component, and harmonic distortion rate, obtaining frequency-domain energy feature parameters such as frequency-domain energy spectral density and frequency band energy proportion. Finally, it constructs a 96-dimensional hybrid fault feature vector containing 32-dimensional time-domain features, 32-dimensional frequency-domain features, and 32-dimensional time-frequency fusion features.

[0025] A three-level fault identification system is adopted for the separation of mixed fault components. The first level inputs the mixed fault feature vector into a radial basis function support vector machine for coarse classification of the fault's normal state, generating first-level classification results of fault state label and normal state label. The second level inputs the mixed fault feature vector into a multilayer perceptron neural network based on the fault state label, and performs layer-by-layer feature processing through a first hidden layer of 128 neurons, a second hidden layer of 64 neurons, and a third hidden layer of 32 neurons. Finally, softmax normalization is performed through the output layer of 3 neurons to obtain second-level classification results of pure transient fault category, pure permanent fault category, and mixed fault category. The third level uses a blind source separation algorithm to decompose the signal for the mixed fault category, and obtains transient fault component data and permanent fault component data through independent component analysis.

[0026] The dynamic weight allocation process calculates the weights for fault duration, current surge amplitude, voltage drop depth, recovery time, and impact range based on instantaneous fault component data, set to 0.3, 0.25, 0.2, 0.15, and 0.1, respectively. It also calculates the weights for insulation resistance reduction, steady-state fault current, equipment temperature rise, and fault development trend based on permanent fault component data, set to 0.35, 0.25, 0.2, and 0.2, respectively. These weight parameters are input into the weighted fusion calculation module to calculate the severity level, classifying faults into levels 1-3 (minor), 4-6 (moderate), and 7-10 (severe). The final quantitative fault assessment result is obtained through the interaction effect correction coefficient.

[0027] The fault quantification assessment results are matched with the graded reclosing strategy. For minor faults of levels 1-3, a fast reclosing strategy is assigned, with a reclosing delay of 0.3 seconds, a maximum number of reclosing attempts of 3, and reclosing intervals of 0.3 seconds, 0.6 seconds, and 1.2 seconds respectively. For moderate faults of levels 4-6, a delayed reclosing strategy is assigned, with an initial reclosing delay of 1.5 seconds, a second delay of 3 seconds when the instantaneous component accounts for more than 70%, and a second delay of 10 seconds when the permanent component accounts for more than 70%. For severe faults of levels 7-10, an intelligent judgment reclosing strategy is assigned, which judges based on the recoverability of the permanent fault component. These parameters are encapsulated into an adaptive control command containing reclosing delay parameters, reclosing attempt parameters, and protection setting parameters, and sent to the magnetic recloser to execute the opening and closing control operation.

[0028] In one specific embodiment, the process of performing step S101 may specifically include the following steps: Based on current transformers and voltage transformers, the three-phase current value, three-phase voltage value, zero-sequence current value and negative-sequence current value of the power distribution line are collected and processed in real time to obtain basic electrical parameter data. The basic electrical parameter data is input into a digital signal processor for waveform analysis and processing to obtain dynamic electrical characteristic data including current surge, voltage drop depth, phase angle change rate and frequency offset. The dynamic electrical characteristic data are processed by Fast Fourier Transform to obtain frequency domain electrical parameter data containing the fundamental component, harmonic components and harmonic distortion rate. Data fusion processing is performed on basic electrical parameter data, dynamic electrical characteristic data, and frequency domain electrical parameter data to obtain the original fault dataset containing time domain electrical quantities, frequency domain electrical quantities, and time-frequency domain electrical quantities.

[0029] Specifically, in the real-time data processing of power distribution lines based on current transformers and voltage transformers, the current transformer converts the large current in the power distribution line into a standard small current signal according to a fixed transformation ratio through the principle of electromagnetic induction. Similarly, the voltage transformer converts the high voltage into a standard small voltage signal according to a fixed transformation ratio. The acquisition module continuously acquires the three-phase current values ​​IA, IB, IC and the three-phase voltage values ​​UA, UB, UC at a sampling frequency of 10kHz. At the same time, it calculates the zero-sequence current value I0 and the negative-sequence current value I2. The zero-sequence current value is calculated by the vector sum of the three-phase currents, and the negative-sequence current value is obtained by the analysis of the negative-sequence components of the three-phase currents. These acquired electrical parameters are converted into digital signals by an analog-to-digital converter to form basic electrical parameter data. When basic electrical parameter data is input into a digital signal processor for waveform analysis, the processor first filters and denoises the acquired current and voltage waveforms. Then, it performs segmented analysis of the waveforms using a sliding time window. The current surge is calculated by the difference between current values ​​at adjacent time points. When the current value jumps significantly within milliseconds, the magnitude of the surge is recorded. The voltage drop depth is calculated by comparing the change in the effective voltage value before and after the fault. The phase angle change rate is obtained by differentiating the change in the voltage and current phase angles over time. The frequency offset is calculated by using the zero-crossing detection method to determine the deviation between the actual frequency and the nominal frequency of 50Hz for each cycle. The digital signal processor integrates these calculation results to form dynamic electrical characteristic data that includes current surge, voltage drop depth, phase angle change rate, and frequency offset.

[0030] When processing dynamic electrical characteristic data using Fast Fourier Transform (FFT), the algorithm first groups the time-domain signal according to powers of 2, and then converts the time-domain signal into a frequency-domain signal through butterfly operation. The FFT decomposes the original time-domain waveform into a superposition of sine waves with different frequency components. The fundamental component corresponds to the amplitude and phase information of the 50Hz frequency component, and each harmonic component corresponds to the amplitude and phase information of the 100Hz, 150Hz, 200Hz, and other harmonic components. The harmonic distortion rate is obtained by calculating the ratio of the effective value of all harmonic components to the effective value of the fundamental component. The transformation result forms frequency-domain electrical parameter data containing the fundamental component, the 2nd to 20th harmonic components, and the total harmonic distortion rate.

[0031] When performing data fusion processing based on basic electrical parameter data, dynamic electrical characteristic data, and frequency domain electrical parameter data, the fusion algorithm first aligns the three types of data according to timestamps to ensure that different types of data at the same time are in the same data group. Time domain electrical quantities include statistical features such as the effective values, instantaneous values, and average values ​​of current and voltage in the basic electrical parameter data. Frequency domain electrical quantities include frequency features such as the amplitude, phase angle, and spectral distribution of each harmonic in the frequency domain electrical parameter data. The time and frequency domain electrical quantities are converted into a time-frequency joint representation by short-time Fourier transform to obtain the frequency distribution characteristics of the signal in different time periods. The data fusion module splices these three types of electrical quantities according to feature dimensions to form a multi-dimensional original fault dataset.

[0032] In one specific embodiment, the process of performing step S102 may specifically include the following steps: The time-domain electrical quantities in the original fault dataset are input into the wavelet transform module for multi-scale decomposition processing to obtain time-domain feature component data containing wavelet coefficients at 8 scale levels. The frequency domain electrical quantities in the original fault dataset are processed by Hilbert-Huang transform to obtain frequency domain feature component data containing instantaneous frequency change trajectory and instantaneous amplitude change trajectory; Based on the time-domain feature component data, statistical feature extraction processing is performed on the current surge, voltage drop depth and phase angle change rate to obtain time-domain statistical feature parameters including mean, variance, kurtosis and skewness. Based on the frequency domain characteristic component data, the energy distribution of the fundamental component, each harmonic component, and harmonic distortion rate is calculated and processed to obtain frequency domain energy characteristic parameters including frequency domain energy spectral density and frequency band energy proportion. Based on time-domain statistical feature parameters and frequency-domain energy feature parameters, a 96-dimensional vector is constructed to obtain a hybrid fault feature vector containing 32-dimensional time-domain features, 32-dimensional frequency-domain features, and 32-dimensional time-frequency fusion features.

[0033] Specifically, when the time-domain electrical quantities in the original fault dataset are input into the wavelet transform module for multi-scale decomposition, the wavelet transform module uses the discrete wavelet transform algorithm to perform hierarchical decomposition of the time-domain signal, selects an appropriate mother wavelet function as the basis for analysis, and then decomposes the original signal into different frequency components through scale scaling and time shifting operations. The first level of decomposition decomposes the original signal into two parts: approximation coefficients and detail coefficients. The approximation coefficients contain the low-frequency components of the signal, and the detail coefficients contain the high-frequency components of the signal. Subsequently, the approximation coefficients are further decomposed to obtain the second level of approximation coefficients and detail coefficients, and so on until the decomposition of 8 levels is completed. Each level of decomposition further subdivides the frequency range. The eighth level of approximation coefficients contains the signal characteristics of the lowest frequency band, and the detail coefficients of the first to eighth levels contain the characteristics of different frequency bands from high frequency to low frequency, respectively. The wavelet coefficients of these 8 scale levels constitute the time-domain feature component data, where the coefficients of each scale level reflect the energy distribution and time positioning information of the fault signal in the corresponding frequency band.

[0034] When performing Hilbert-Huang transform on the frequency domain electrical quantities in the original fault dataset, the algorithm first performs empirical mode decomposition on the frequency domain signal, decomposing the complex multi-component signal into several intrinsic mode functions (EMFs). Each EMF represents a specific frequency component in the signal. Then, a Hilbert transform is performed on each EMF. The Hilbert transform obtains the instantaneous amplitude and instantaneous phase of the signal by constructing an analytic signal. The instantaneous frequency is obtained by differentiating the instantaneous phase over time, and the instantaneous amplitude is obtained directly from the magnitude of the analytic signal. The instantaneous frequency change trajectory records the frequency change process over time, reflecting the dynamic evolution characteristics of the frequency components during the fault occurrence process. The instantaneous amplitude change trajectory records the amplitude change process over time, reflecting the temporal distribution characteristics of the fault signal energy. These two trajectories constitute the frequency domain feature component data.

[0035] When performing statistical feature extraction processing on current surge, voltage drop depth, and phase angle change rate based on time-domain feature component data, the statistical algorithm first extracts current surge, voltage drop depth, and phase angle change rate data sequences from wavelet coefficients at eight scale levels. Then, it calculates four key statistical parameters for each data sequence. The mean is obtained by summing all values ​​in the data sequence and dividing by the number of data points, reflecting the central tendency of the data. The variance is obtained by calculating the sum of the squares of the differences between each data point and the mean and dividing by the number of data points, reflecting the dispersion of the data. The kurtosis is obtained by calculating the ratio of the fourth central moment to the square of the variance, reflecting the sharpness of the data distribution. The skewness is obtained by calculating the ratio of the third central moment to the 1.5th power of the variance, reflecting the symmetry of the data distribution. Each parameter describes the statistical distribution characteristics of the fault features from different perspectives, ultimately forming time-domain statistical feature parameters including mean, variance, kurtosis, and skewness.

[0036] When calculating the energy distribution of the fundamental component, harmonic components, and harmonic distortion rate based on frequency domain characteristic component data, the energy calculation module first separates the signal components of the fundamental frequency band and each harmonic frequency band from the instantaneous frequency change trajectory and instantaneous amplitude change trajectory. The fundamental component corresponds to the signal energy of the 50Hz frequency band, which is obtained by time integration of the square of the instantaneous amplitude within this frequency band. Each harmonic component corresponds to the signal energy of octave bands such as 100Hz, 150Hz, and 200Hz, which are also obtained by time integration of the square of the instantaneous amplitude within the corresponding frequency band. The harmonic distortion rate corresponds to the comprehensive energy of all harmonic components except the fundamental wave, which is obtained by the ratio of the sum of all harmonic energies to the fundamental wave energy. The frequency domain energy spectral density is obtained by dividing the energy of each frequency band by the corresponding bandwidth, reflecting the energy distribution within a unit bandwidth. The bandwidth energy proportion is obtained by the ratio of the energy of each frequency band to the total energy, reflecting the relative contribution of different frequency bands to the overall energy. These calculation results constitute the frequency domain energy characteristic parameters.

[0037] When constructing a 96-dimensional vector based on time-domain statistical feature parameters and frequency-domain energy feature parameters, the vector construction module first organizes the mean, variance, kurtosis, and skewness data in the time-domain statistical feature parameters according to feature type and scale level to form a 32-dimensional time-domain feature vector. Then, the fundamental energy, harmonic energy, energy spectral density, and frequency band proportion data in the frequency-domain energy feature parameters are arranged according to frequency components to form a 32-dimensional frequency-domain feature vector. Next, the time-domain features and frequency-domain features are cross-combined through time-frequency joint analysis to calculate the correlation coefficient, mutual information, joint entropy, and other time-frequency coupling features between the time-domain statistical parameters and the frequency-domain energy parameters, forming a 32-dimensional time-frequency fusion feature vector. Finally, these three sets of feature vectors are spliced ​​in a fixed order to form a 96-dimensional hybrid fault feature vector containing 32-dimensional time-domain features, 32-dimensional frequency-domain features, and 32-dimensional time-frequency fusion features.

[0038] In one specific embodiment, the process of executing step S103 may specifically include the following steps: The hybrid fault feature vector is input into the radial basis function support vector machine for coarse classification of fault-normal states, resulting in a first-level classification result containing fault state identifiers and normal state identifiers. Based on the fault state identifier in the first-level classification result, the hybrid fault feature vector is classified by a multilayer perceptron neural network to obtain a second-level classification result that includes pure transient fault category, pure permanent fault category and hybrid fault category; Based on the mixed fault categories in the second-level classification results, the mixed fault feature vector is decomposed using a blind source separation algorithm to obtain fault signal separation data containing instantaneous fault component signals and permanent fault component signals; Independent component analysis is performed on the fault signal separation data to obtain instantaneous fault component data containing instantaneous fault component amplitude and instantaneous fault component frequency characteristics, and permanent fault component data containing permanent fault component amplitude and permanent fault component impedance characteristics.

[0039] Specifically, when inputting the mixed fault feature vector into a radial basis function support vector machine (RBF SVM) for coarse classification of fault and normal states, the RBF SVM first preprocesses the 96-dimensional feature vector by standardizing each feature value to the same numerical range. Then, it uses the radial basis kernel function to calculate the similarity between the input feature vector and the support vector. The RBF kernel function calculates the Euclidean distance between the feature vector and the support vector and applies a Gaussian function for nonlinear mapping, transforming the original feature space into a high-dimensional kernel space. An optimal separating hyperplane is constructed in the kernel space. The hyperplane divides the feature vector into two categories, fault state and normal state, by maximizing the classification margin. The classifier determines the category based on the distance and sign of the input vector to the hyperplane. The absolute value of the distance reflects the classification confidence, and the sign reflects the classification result. The output includes a first-level classification result containing fault state and normal state labels. The fault state label indicates that an abnormal situation has been detected and further analysis is needed, while the normal state label indicates that the circuit is operating normally and no further processing is required.

[0040] When classifying mixed fault feature vectors based on fault state identifiers from the first-level classification results using a multilayer perceptron neural network, the multilayer perceptron neural network employs a forward propagation algorithm to process the feature vectors layer by layer. The first hidden layer contains 128 neurons, each receiving all elements of the 96-dimensional input feature vector. After a linear transformation between the input vector and the weight matrix, an activation function is applied for nonlinear mapping. The activation function uses the ReLU function to set negative values ​​to zero and retain positive values. The second hidden layer contains 64 neurons, receiving the 128-dimensional output of the first hidden layer as input, and undergoes similar weight transformation and activation function processing. The third hidden layer contains 32 neurons, receiving the 64-dimensional output of the second hidden layer for further feature compression and abstraction. The output layer contains 3 neurons corresponding to the three categories of pure transient faults, pure permanent faults, and mixed faults, respectively. The output values ​​are converted into a probability distribution using a softmax function, where the output value of each neuron represents the probability of the corresponding fault category. Finally, the category with the highest probability is selected as the classification result, forming a second-level classification result that includes pure transient faults, pure permanent faults, and mixed faults.

[0041] When performing blind source separation algorithm decomposition on the mixed fault feature vector based on the mixed fault categories in the second-level classification results, the blind source separation algorithm decomposes the mixed signal into independent source signal components based on the statistical independence assumption. The algorithm first performs whitening preprocessing on the 96-dimensional feature vector, and removes the correlation between the dimensions of the feature vector through principal component analysis, so that the preprocessed data has a unit covariance matrix. Then, the FastICA algorithm is applied to find the optimal separation matrix. FastICA achieves source signal separation by maximizing the non-Gaussianity of the separated signal. The algorithm iteratively calculates the weight vector of the separation matrix. In each iteration, the weights are updated in the direction of increasing the non-Gaussianity of the output signal. Non-Gaussianity is measured by statistical measures such as kurtosis or negative entropy. During the iteration process, the gradient ascent method is used to update the weight vector until convergence. Finally, a separation matrix that can separate the mixed fault signal into transient fault components and permanent fault components is obtained. The separated transient fault component signal mainly reflects short-term transient characteristics, and the permanent fault component signal mainly reflects continuous steady-state characteristics, forming fault signal separation data containing transient fault component signals and permanent fault component signals.

[0042] When performing independent component analysis (ICA) on the separated fault signal data, the ICA algorithm extracts feature parameters from the separated instantaneous fault component signal and permanent fault component signal respectively. For the instantaneous fault component signal, the algorithm calculates its time-domain amplitude feature by finding the peak value and effective value of the signal envelope to obtain the instantaneous fault component amplitude. The frequency feature is obtained by extracting the main frequency component and frequency band distribution through spectrum analysis of the signal to obtain the instantaneous fault component frequency feature. The instantaneous fault component amplitude reflects the magnitude of the fault impact, and the instantaneous fault component frequency feature reflects the oscillation frequency and attenuation characteristics of the fault-induced signal. For the permanent fault component signal, the algorithm calculates its steady-state amplitude feature by obtaining the permanent fault component amplitude through mean and variance analysis of the signal. The impedance feature is obtained by calculating the equivalent impedance parameter by analyzing the ratio relationship between the voltage and current component signals to obtain the permanent fault component impedance feature. The permanent fault component amplitude reflects the steady-state electrical quantity change caused by the fault, and the permanent fault component impedance feature reflects the electrical characteristic change at the fault point. Finally, instantaneous fault component data containing instantaneous fault component amplitude and instantaneous fault component frequency features and permanent fault component data containing permanent fault component amplitude and permanent fault component impedance features are formed.

[0043] In one specific embodiment, the process of performing multilayer perceptron neural network classification on the hybrid fault feature vector based on the fault state identifier in the first-level classification result can specifically include the following steps: The hybrid fault feature vector is input into the first hidden layer of 128 neurons and processed by linear transformation and activation function to obtain the output data of the first neural network containing 128-dimensional intermediate features. Based on the output data of the first layer neural network, weight calculation and nonlinear mapping are performed on the second hidden layer with 64 neurons to obtain the output data of the second layer neural network containing 64-dimensional compressed features. Based on the output data of the second-layer neural network, feature fusion and dimensionality reduction are performed on the third hidden layer of 32 neurons to obtain the output data of the third-layer neural network containing 32 core features; The output data of the third layer neural network is input into the output layer of the three neurons and subjected to softmax normalization to obtain the second-level classification result, which includes pure instantaneous fault probability value, pure permanent fault probability value and mixed fault probability value.

[0044] Specifically, when the hybrid fault feature vector is input into the first hidden layer of 128 neurons for linear transformation and activation function processing, the first hidden layer receives a 96-dimensional hybrid fault feature vector as input. Each neuron establishes connections with all dimensions of the input vector. During the linear transformation, each neuron calculates the inner product of the input vector and its corresponding weight vector and adds a bias term. Specifically, the calculation method is to multiply each element of the input vector by its corresponding weight coefficient, sum the results, and add the bias value. The 128 neurons have different weight vectors and bias terms, thus enabling the extraction of different feature combinations from the 96-dimensional input. After the linear transformation, each neuron obtains a scalar value. Subsequently, the ReLU activation function is applied to perform a nonlinear mapping on the linear transformation result. The ReLU function sets all negative values ​​to zero while keeping positive values ​​unchanged. This processing method can introduce nonlinear characteristics and reduce the gradient vanishing problem. After activation function processing, the output data of the first layer of neural network containing 128-dimensional intermediate features is formed. These intermediate features represent high-level abstract features extracted from the original hybrid fault features.

[0045] When performing weight calculation and nonlinear mapping on the second hidden layer with 64 neurons based on the output data of the first layer neural network, the second hidden layer receives 128-dimensional intermediate features from the first hidden layer as input. During the weight calculation process, each neuron establishes a full connection with the 128-dimensional input, and the weighted sum is calculated through matrix multiplication of the weight matrix and the input vector. The weight matrix of the second hidden layer has a size of 64×128, which means that each of the 64 neurons assigns a different weight coefficient to the 128 input features. After the weight calculation is completed, the corresponding bias terms of each neuron are also added. The nonlinear mapping process uses the same ReLU activation function as the first hidden layer to convert the linear combination result into a nonlinear output. Since the number of neurons in the second hidden layer is reduced from 128 to 64, this layer realizes the function of feature compression, further integrating and abstracting the intermediate features extracted by the first layer to form the output data of the second layer neural network containing 64-dimensional compressed features. The compressed features retain the most important fault feature information while reducing the data dimensionality.

[0046] When performing feature fusion and dimensionality reduction on the third hidden layer of 32 neurons based on the output data of the second-layer neural network, the third hidden layer receives 64-dimensional compressed features as input and further reduces the feature dimension to 32 dimensions. The feature fusion process is achieved through a linear transformation of the weight matrix size of 32×64. Each neuron fuses 64 input features into a single feature value through different weight combinations. This fusion method can capture the correlation and interaction between different features. Dimensionality reduction reduces feature redundancy while maintaining key information. The third hidden layer also uses the ReLU activation function for nonlinear mapping to ensure the network's expressive and learning capabilities. After the step-by-step processing of the three hidden layers, the original 96-dimensional mixed fault features are compressed and abstracted into 32-dimensional core features. These core features contain the most important information required for fault classification, forming the output data of the third-layer neural network containing 32-dimensional core features.

[0047] When the output data of the third-layer neural network is input into the output layer of the three neurons for softmax normalization, the output layer receives 32-dimensional core features and maps them to probability distributions of three fault categories. The three neurons of the output layer correspond to the three categories of pure transient faults, pure permanent faults, and mixed faults, respectively. Each neuron first performs a linear transformation to calculate the score value of its corresponding category. The linear transformation is achieved by matrix multiplication of the input vector with a weight matrix of size 3×32. The softmax normalization process converts the three score values ​​into a probability distribution. The softmax function first performs an exponential operation on each score value, and then divides each exponential value by the sum of all exponential values ​​to ensure that the sum of the three output probability values ​​is equal to 1 and each probability value is between 0 and 1. The mathematical principle of softmax normalization is based on the characteristics of the exponential function. Larger score values ​​are significantly amplified after exponential operation, while smaller score values ​​are relatively reduced. This processing method can highlight the most likely fault category while suppressing the unlikely category, and finally form a second-level classification result containing the probability values ​​of pure transient faults, pure permanent faults, and mixed faults.

[0048] In one specific embodiment, the process of executing step S104 may specifically include the following steps: Based on the instantaneous fault component data, the fault duration, current surge amplitude, voltage drop depth, recovery time, and impact range are weighted and processed to obtain instantaneous fault weight allocation parameters, which include a fault duration weight of 0.3, a current surge amplitude weight of 0.25, a voltage drop depth weight of 0.2, a recovery time weight of 0.15, and an impact range weight of 0.1. Based on the permanent fault component data, the weights of insulation resistance reduction, fault current steady-state value, equipment temperature rise, and fault development trend are calculated to obtain permanent fault weight allocation parameters, including insulation resistance reduction weight of 0.35, fault current steady-state value weight of 0.25, equipment temperature rise weight of 0.2, and fault development trend weight of 0.2. The instantaneous fault weight allocation parameters and permanent fault weight allocation parameters are input into the weighted fusion calculation module to calculate the severity level, and the fault severity level classification data including level 1-3 minor faults, level 4-6 moderate faults and level 7-10 severe faults are obtained. The interaction effect correction coefficient is calculated on the fault severity level classification data to obtain a quantitative assessment result of faults that includes both instantaneous fault severity level and permanent fault severity level.

[0049] Specifically, when performing weighted calculations on fault duration, current surge amplitude, voltage dip depth, recovery time, and impact range based on instantaneous fault component data, the weighted calculation module first extracts the values ​​of each evaluation index from the instantaneous fault component data. The fault duration is obtained by analyzing the time span from the onset of the anomaly to the return to normal of the instantaneous fault component signal. The current surge amplitude is obtained by calculating the maximum difference between the effective current values ​​before and after the fault. The voltage dip depth is obtained by calculating the percentage difference between the lowest voltage value and the normal voltage value. The recovery time is obtained by measuring the time required from fault clearance to the complete restoration of electrical parameters to normal. The impact range is obtained by statistically analyzing the number and capacity of load nodes affected by the fault. The weighting calculation employs the analytic hierarchy process (AHP) to rank the importance of each indicator. Weight coefficients are assigned based on the degree of impact of transient faults on the magnetic recloser control decision. Fault duration is assigned a weight of 0.3 as the most critical indicator, since duration directly determines the reclosing delay setting. Current surge amplitude is assigned a weight of 0.25, since surge amplitude reflects the fault impact intensity and affects protection setting. Voltage drop depth is assigned a weight of 0.2, since drop depth affects user perception and reclosing success rate. Recovery time is assigned a weight of 0.15, since recovery time reflects the fault self-healing characteristics. Influence range is assigned a weight of 0.1, since influence range is mainly used for fault priority ranking. The final transient fault weighting parameters are thus formed.

[0050] When calculating the weights of insulation resistance reduction, fault current steady-state value, equipment temperature rise, and fault development trend based on permanent fault component data, the weight allocation module parses the characteristic values ​​of each evaluation index from the permanent fault component data. The degree of insulation resistance reduction is obtained by comparing the change in insulation resistance test values ​​before and after the fault. The fault current steady-state value is obtained by analyzing the average and effective values ​​of the permanent fault component signals. The equipment temperature rise is obtained by monitoring the temperature change of the fault point and surrounding equipment. The fault development trend is obtained by time series analysis and trend prediction of the permanent fault component signals. The weight allocation also adopts the analytic hierarchy process (AHP) in conjunction with the permanent fault's impact on equipment safety and long-term operation. The weights of each indicator are determined by the degree of impact of the fault. The degree of insulation resistance reduction, as the most critical indicator of permanent faults, is assigned a weight of 0.35, because the insulation condition is directly related to the safety and success rate of reclosing operations. The steady-state value of the fault current is assigned a weight of 0.25, because the magnitude of the steady-state current determines the equipment's withstand capacity and protection configuration. The equipment temperature rise is assigned a weight of 0.2, because the temperature rise reflects the degree of equipment damage and affects the timing of reclosing. The fault development trend is assigned a weight of 0.2, because the prediction of the development trend affects the adjustment and optimization of the reclosing strategy. This weight allocation fully considers the gradual nature of permanent faults and their long-term impact on magnetic recloser control, forming the permanent fault weight allocation parameters.

[0051] When the instantaneous fault weight allocation parameters and permanent fault weight allocation parameters are input into the weighted fusion calculation module for severity level calculation, the weighted fusion calculation module first performs numerical standardization on each evaluation index, mapping the index values ​​of different dimensions to a unified 0-1 interval. Then, it calculates the instantaneous fault comprehensive score and the permanent fault comprehensive score separately. The instantaneous fault comprehensive score is obtained by multiplying the standardized values ​​of each index by their corresponding weights and then summing them. The permanent fault comprehensive score is calculated in the same way. Then, according to the preset level classification threshold, the comprehensive score is mapped to the fault severity level. Levels 1-3, minor faults, correspond to a comprehensive score range of 0-0.3, indicating that the fault impact is small and easy to recover. Levels 4-6, moderate faults, correspond to a comprehensive score range of 0.3-0.7, indicating that the fault impact is moderate and requires careful handling. Levels 7-10, severe faults, correspond to a comprehensive score range of 0.7-1.0, indicating that the fault impact is severe and requires special handling strategies. The level classification uses fuzzy set theory to handle the boundary fuzziness problem. When the comprehensive score is close to the threshold, a membership function is used for smooth transition, forming the fault severity level classification data.

[0052] When calculating the interaction effect correction coefficient for fault severity level classification data, the correction coefficient calculation module analyzes the interaction relationship between transient and permanent faults. When both types of faults exist simultaneously, they will produce a superposition effect and mutual influence. The interaction effect correction coefficient establishes a correction model by analyzing the performance characteristics and processing results of mixed faults in historical fault data. The correction coefficient calculation considers factors such as fault type matching degree, temporal overlap, spatial correlation, and influence intensity. Fault type matching degree reflects the compatibility of transient and permanent faults, temporal overlap reflects the synchronicity of the occurrence time of the two faults, spatial correlation reflects the proximity of fault locations, and influence intensity reflects the impact of each fault on the overall system. The correction coefficient calculation establishes a quantitative relationship between each factor and the correction degree through multiple regression analysis. Generally, the correction coefficient ranges from 1.1 to 1.5, indicating that the severity of mixed faults is more severe than the simple superposition of single faults. Finally, the correction coefficient is applied to the severity levels of transient and permanent faults to obtain a quantitative assessment result of faults considering interaction effects.

[0053] In one specific embodiment, the process of executing step S105 may specifically include the following steps: Based on the instantaneous fault severity level and permanent fault severity level in the fault quantification assessment results, strategy type assignment processing is performed for minor faults of level 1-3, moderate faults of level 4-6, and severe faults of level 7-10, to obtain hierarchical reclosing strategy type data including fast reclosing strategy, delayed reclosing strategy, and intelligent discrimination reclosing strategy. Based on the data of the hierarchical reclosing strategy type, the parameters of the fast reclosing strategy are calculated and processed to obtain the fast reclosing parameter configuration data, which includes a reclosing delay of 0.3 seconds, a maximum number of reclosing times of 3, and a reclosing interval of 0.3 seconds, 0.6 seconds, and 1.2 seconds. The hierarchical reclosing strategy type data is input into the delayed reclosing strategy calculation module for instantaneous component proportion judgment processing, and the delayed reclosing parameter configuration data is obtained, which includes a first reclosing delay of 1.5 seconds, a second delay of 3 seconds when the instantaneous component proportion is greater than 70%, and a second delay of 10 seconds when the permanent component proportion is greater than 70%. The fast reclosing parameter configuration data and the delayed reclosing parameter configuration data are processed by control instruction encapsulation to obtain an adaptive control instruction containing reclosing delay parameters, reclosing count parameters and protection setting parameters, and the opening and closing control operation instruction is sent to the magnetic recloser.

[0054] Specifically, when assigning strategy types for minor faults (levels 1-3), moderate faults (levels 4-6), and severe faults (levels 7-10) based on the instantaneous and permanent fault severity levels in the fault quantification assessment results, the strategy allocation module first reads the level data from the fault quantification assessment results, and then automatically allocates the strategy according to preset strategy mapping rules. Minor faults (levels 1-3) correspond to a fast reclosing strategy; these faults are typically instantaneous and have strong self-recovery capabilities, making them suitable for rapid power restoration via multiple fast reclosing. Moderate faults (levels 4-6) correspond to a delayed reclosing strategy; these faults possess both instantaneous and permanent characteristics and require a strategy based on the fault type. The reclosing timing is flexibly adjusted based on the proportion of fault components. Severe faults of levels 7-10 correspond to intelligent reclosing strategies. These faults have obvious permanent characteristics and require careful reclosing operations after detailed analysis. The strategy allocation uses a lookup table method to establish the correspondence between levels and strategies. When the severity levels of instantaneous faults and permanent faults are not in the same group, the strategy allocation module uses a weighted average method to calculate the comprehensive level and then selects the strategy. The weighting coefficient is dynamically determined based on the proportion of instantaneous fault components and the proportion of permanent fault components, ultimately forming a hierarchical reclosing strategy type data that includes fast reclosing strategies, delayed reclosing strategies, and intelligent reclosing strategies.

[0055] When calculating parameters for the fast reclosing strategy based on the data of the graded reclosing strategy types, the parameter calculation module formulates a standardized parameter configuration scheme for the fast reclosing strategy. The reclosing delay is set to 0.3 seconds based on the typical recovery time characteristics of transient faults. This delay ensures that the fault point has enough time to detach while avoiding excessive waiting that could affect power supply continuity. The maximum number of reclosing attempts is set to 3 to balance the self-healing probability of minor faults with the equipment's impact tolerance. Statistical data shows that the probability of successful recovery of minor transient faults within 3 reclosing attempts exceeds 95%. The reclosing interval is designed to increase incrementally, with intervals of 0.3 seconds. The intervals are 0.3 seconds, 0.6 seconds, and 1.2 seconds. The first interval of 0.3 seconds is used to adapt to transient faults that heal quickly. The second interval of 0.6 seconds gives the fault point more time to recover. The third interval of 1.2 seconds ensures that even transient faults that recover slowly have sufficient stabilization time. The incremental interval design follows the time law and probability distribution characteristics of fault recovery. The parameter calculation module matches and verifies these preset parameters with the actual fault characteristics to ensure that the parameter configuration is adapted to the fault characteristics, forming fast reclosing parameter configuration data including a reclosing delay of 0.3 seconds, a maximum number of reclosing times of 3, and reclosing intervals of 0.3 seconds, 0.6 seconds, and 1.2 seconds.

[0056] When the data on the graded reclosing strategy type is input into the delayed reclosing strategy calculation module for instantaneous component proportion judgment, the judgment module first extracts the instantaneous fault component proportion and the permanent fault component proportion from the fault component separation data. The instantaneous component proportion is calculated by the ratio of the instantaneous fault component energy to the total fault energy, and the permanent component proportion is calculated by the ratio of the permanent fault component energy to the total fault energy. The judgment module sets 70% as the key threshold for component proportion. This threshold is determined based on a large amount of experience in handling mixed faults and statistical analysis. When the instantaneous component proportion is greater than 70%, it indicates that the fault is mainly instantaneous, and the delayed reclosing strategy tends to have a shorter waiting time. The initial reclosing delay is set to 1.5 seconds. Sufficient initial stabilization time for the fault is allowed. The second reclosing delay is set to 3 seconds to adapt to the self-healing characteristics of transient faults. When the proportion of permanent components is greater than 70%, it indicates that the fault is mainly permanent. The delayed reclosing strategy requires a longer observation and preparation time. The second reclosing delay is extended to 10 seconds to ensure that permanent faults are fully handled or evolve naturally. The proportion judgment adopts the sliding window averaging method to reduce the impact of transient fluctuations. The calculation module continuously monitors the component proportion change trend within 5 sampling periods for comprehensive judgment. Finally, the delayed reclosing parameter configuration data is formed, which includes an initial reclosing delay of 1.5 seconds, a second delay of 3 seconds when the transient component proportion is greater than 70%, and a second delay of 10 seconds when the permanent component proportion is greater than 70%.

[0057] When encapsulating control commands for the configuration data of fast reclosing parameters and delayed reclosing parameters, the encapsulation module organizes and encodes various parameters according to the standard format of the magnetic recloser control protocol. The reclosing delay parameter includes the specific delay value and delay type identifier; the reclosing count parameter includes a counter for the maximum allowed number of reclosing attempts and the current execution count; and the protection setting parameters include key parameters such as overcurrent protection settings, instantaneous overcurrent protection settings, and grounding protection settings. These protection settings are dynamically adjusted based on fault characteristics and reclosing strategies to match the safety requirements of reclosing operations. The encapsulation process uses a standard communication protocol. The protocol format ensures that the magnetic recloser can correctly parse and execute control commands. The command encapsulation consists of three parts: command header, parameter body, and check tail. The command header identifies the command type and priority, the parameter body contains the specific control parameter values, and the check tail ensures the integrity and accuracy of data transmission. After encapsulation, the control command is sent to the magnetic recloser through the communication interface to execute the specific opening and closing control operation. After receiving the command, the magnetic recloser performs parameter parsing and safety verification. After confirming that there are no errors, it executes the corresponding opening and closing actions according to the command requirements, and at the same time feeds back the execution status and result information to the control module, forming a closed-loop control process.

[0058] The adaptive control method for magnetic reclosers used in circuit design in the embodiments of this application has been described above. The adaptive control system for magnetic reclosers used in circuit design in the embodiments of this application is described below. Please refer to [link to relevant documentation]. Figure 2 One embodiment of the magnetic recloser adaptive control system for circuit design in this application includes: The data acquisition module is used to collect and process multi-dimensional electrical parameters of the power distribution line through an intelligent sensor array to obtain the original fault dataset. The construction module is used to construct fault feature vectors based on the original fault dataset using a time-frequency domain fusion algorithm, thereby obtaining a hybrid fault feature vector. The separation module is used to separate the mixed fault component data and the permanent fault component data by performing mixed fault component separation processing on the mixed fault feature vector through three-level fault identification. The allocation module is used to perform dynamic weight allocation processing on each fault component based on the instantaneous fault component data and the permanent fault component data, so as to obtain a fault quantitative assessment result including the instantaneous fault severity level and the permanent fault severity level. The control module is used to process the fault quantification assessment results through a hierarchical reclosing strategy to obtain an adaptive control command containing reclosing delay parameters, reclosing count parameters, and protection setting parameters, and to execute the magnetic recloser opening and closing control operation.

[0059] above Figure 2 The adaptive control system for magnetic re-encoders used in circuit design in this embodiment of the invention is described in detail from the perspective of modular functional entities. The adaptive control device for magnetic re-encoders used in circuit design in this embodiment of the invention is described in detail from the perspective of hardware processing.

[0060] Reference Figure 3 This invention also provides an adaptive control device for a magnetic recloser used in circuit design. This adaptive control device can be a server, and its internal structure can be as follows: Figure 3As shown, the adaptive control device for a magnetic recloser used in circuit design includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor, part of the computer design, provides computational and control capabilities. The memory of the adaptive control device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and the database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the adaptive control device stores the data corresponding to this embodiment. The network interface of the adaptive control device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0061] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the adaptive control device for magnetic reconciliation used in circuit design where the present invention is applied.

[0062] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of the magnetic recloser adaptive control method for circuit design.

[0063] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0064] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a magnetic recloser adaptive control device (which may be a personal computer, server, or network device, etc.) used for circuit design to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0065] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An adaptive control method for magnetic reclosers used in circuit design, characterized in that, The method includes: The original fault dataset is obtained by collecting and processing multi-dimensional electrical parameters of the power distribution line through an intelligent sensor array. Based on the original fault dataset, a time-frequency domain fusion algorithm is used to construct fault feature vectors to obtain hybrid fault feature vectors. The hybrid fault feature vector is processed by three-level fault identification to separate the hybrid fault components, resulting in instantaneous fault component data and permanent fault component data. Based on the instantaneous fault component data and the permanent fault component data, dynamic weight allocation processing is performed on each fault component to obtain a fault quantitative assessment result that includes the instantaneous fault severity level and the permanent fault severity level. The fault quantification assessment results are processed by a graded reclosing strategy to obtain an adaptive control command that includes reclosing delay parameters, reclosing count parameters, and protection setting parameters, and the magnetic recloser opening and closing control operation is executed.

2. The adaptive control method for magnetic reclosers in circuit design according to claim 1, characterized in that, The process of acquiring and processing multi-dimensional electrical parameters of the power distribution line through an intelligent sensor array to obtain the original fault dataset includes: Based on current transformers and voltage transformers, the three-phase current value, three-phase voltage value, zero-sequence current value and negative-sequence current value of the power distribution line are collected and processed in real time to obtain basic electrical parameter data. The basic electrical parameter data is input into a digital signal processor for waveform analysis to obtain dynamic electrical characteristic data including current surge, voltage drop depth, phase angle change rate and frequency offset. The dynamic electrical characteristic data is processed by Fast Fourier Transform to obtain frequency domain electrical parameter data containing the fundamental component, harmonic components, and harmonic distortion rate. The basic electrical parameter data, dynamic electrical characteristic data, and frequency domain electrical parameter data are fused together to obtain the original fault dataset containing time domain electrical quantities, frequency domain electrical quantities, and time-frequency domain electrical quantities.

3. The adaptive control method for magnetic reclosers in circuit design according to claim 2, characterized in that, The process of constructing a hybrid fault feature vector based on the original fault dataset using a time-frequency domain fusion algorithm includes: The time-domain electrical quantities in the original fault dataset are input into the wavelet transform module for multi-scale decomposition processing to obtain time-domain feature component data containing wavelet coefficients at 8 scale levels. The frequency domain electrical quantities in the original fault dataset are processed by Hilbert-Huang transform to obtain frequency domain feature component data containing instantaneous frequency change trajectory and instantaneous amplitude change trajectory; Based on the time-domain feature component data, statistical feature extraction processing is performed on the current surge, voltage drop depth and phase angle change rate to obtain time-domain statistical feature parameters including mean, variance, kurtosis and skewness. Based on the frequency domain characteristic component data, the energy distribution of the fundamental component, each harmonic component, and harmonic distortion rate is calculated to obtain frequency domain energy characteristic parameters including frequency domain energy spectral density and frequency band energy proportion. Based on the time-domain statistical feature parameters and frequency-domain energy feature parameters, a 96-dimensional vector construction process is performed to obtain a hybrid fault feature vector containing 32-dimensional time-domain features, 32-dimensional frequency-domain features, and 32-dimensional time-frequency fusion features.

4. The adaptive control method for magnetic reclosers in circuit design according to claim 3, characterized in that, The process of separating the mixed fault component data from the mixed fault feature vector through a three-level fault identification method to obtain instantaneous fault component data and permanent fault component data includes: The hybrid fault feature vector is input into a radial basis function support vector machine for coarse classification of fault-normal states, resulting in a first-level classification result containing fault state identifiers and normal state identifiers. Based on the fault state identifier in the first-level classification result, the hybrid fault feature vector is classified by a multilayer perceptron neural network to obtain a second-level classification result that includes pure transient fault category, pure permanent fault category and hybrid fault category. Based on the mixed fault categories in the second-level classification results, the mixed fault feature vector is decomposed using a blind source separation algorithm to obtain fault signal separation data containing instantaneous fault component signals and permanent fault component signals; Independent component analysis is performed on the fault signal separation data to obtain instantaneous fault component data containing instantaneous fault component amplitude and instantaneous fault component frequency characteristics, and permanent fault component data containing permanent fault component amplitude and permanent fault component impedance characteristics.

5. The adaptive control method for magnetic reclosers in circuit design according to claim 4, characterized in that, The hybrid fault feature vector is classified using a multilayer perceptron neural network based on the fault state identifier in the first-level classification result, resulting in a second-level classification result that includes pure transient fault categories, pure permanent fault categories, and hybrid fault categories, including: The hybrid fault feature vector is input into the first hidden layer of 128 neurons and processed by linear transformation and activation function to obtain the output data of the first neural network containing 128-dimensional intermediate features. Based on the output data of the first layer neural network, weight calculation and nonlinear mapping are performed on the second hidden layer of 64 neurons to obtain the output data of the second layer neural network containing 64-dimensional compressed features. Based on the output data of the second layer neural network, feature fusion and dimensionality reduction are performed on the third hidden layer of 32 neurons to obtain the output data of the third layer neural network containing 32 core features; The output data of the third layer neural network is input into the output layer of the three neurons and subjected to softmax normalization to obtain a second-level classification result containing pure instantaneous fault probability values, pure permanent fault probability values, and mixed fault probability values.

6. The adaptive control method for magnetic reclosers in circuit design according to claim 1, characterized in that, The step of dynamically weighting each fault component based on the instantaneous fault component data and the permanent fault component data to obtain a fault quantification assessment result including the instantaneous fault severity level and the permanent fault severity level includes: Based on the instantaneous fault component data, the fault duration, current fluctuation amplitude, voltage drop depth, recovery time, and impact range are weighted and processed to obtain instantaneous fault weight allocation parameters including fault duration weight of 0.3, current fluctuation amplitude weight of 0.25, voltage drop depth weight of 0.2, recovery time weight of 0.15, and impact range weight of 0.

1. Based on the permanent fault component data, the degree of insulation resistance reduction, the steady-state value of fault current, the equipment temperature rise, and the fault development trend are weighted and processed to obtain permanent fault weight allocation parameters including a weight of 0.35 for the degree of insulation resistance reduction, a weight of 0.25 for the steady-state value of fault current, a weight of 0.2 for the equipment temperature rise, and a weight of 0.2 for the fault development trend. The instantaneous fault weight allocation parameters and permanent fault weight allocation parameters are input into the weighted fusion calculation module for severity level calculation to obtain fault severity level classification data including level 1-3 minor faults, level 4-6 moderate faults and level 7-10 severe faults. The interaction effect correction coefficient is calculated on the fault severity level classification data to obtain a fault quantitative assessment result that includes the instantaneous fault severity level and the permanent fault severity level.

7. The adaptive control method for magnetic reclosers in circuit design according to claim 1, characterized in that, The fault quantification assessment results are processed through a graded reclosing strategy to obtain an adaptive control command containing reclosing delay parameters, reclosing frequency parameters, and protection setting parameters, and the magnetic recloser opening and closing control operation is executed, including: Based on the instantaneous fault severity level and permanent fault severity level in the fault quantification assessment results, strategy type allocation processing is performed for minor faults of level 1-3, moderate faults of level 4-6, and severe faults of level 7-10 to obtain hierarchical reclosing strategy type data including fast reclosing strategy, delayed reclosing strategy, and intelligent discrimination reclosing strategy. Based on the hierarchical reclosing strategy type data, the parameters of the fast reclosing strategy are calculated and processed to obtain fast reclosing parameter configuration data including reclosing delay of 0.3 seconds, maximum number of reclosing times of 3, and reclosing interval of 0.3 seconds-0.6 seconds-1.2 seconds. The hierarchical reclosing strategy type data is input into the delayed reclosing strategy calculation module for instantaneous component proportion judgment processing to obtain delayed reclosing parameter configuration data, which includes a first reclosing delay of 1.5 seconds, a second delay of 3 seconds when the instantaneous component proportion is greater than 70%, and a second delay of 10 seconds when the permanent component proportion is greater than 70%. The fast reclosing parameter configuration data and the delayed reclosing parameter configuration data are processed by control instruction encapsulation to obtain an adaptive control instruction containing reclosing delay parameters, reclosing count parameters and protection setting parameters, and the opening and closing control operation instruction is sent to the magnetic recloser.

8. An adaptive control system for a magnetic recloser used in circuit design, characterized in that, For implementing the adaptive control method for a magnetic reselector in circuit design as described in any one of claims 1-7, the adaptive control system for the magnetic reselector in circuit design comprises: The data acquisition module is used to collect and process multi-dimensional electrical parameters of the power distribution line through an intelligent sensor array to obtain the original fault dataset. The construction module is used to construct fault feature vectors based on the original fault dataset using a time-frequency domain fusion algorithm, thereby obtaining a hybrid fault feature vector. The separation module is used to separate the mixed fault component data and permanent fault component data by performing mixed fault component separation processing on the mixed fault feature vector through three-level fault identification. The allocation module is used to perform dynamic weight allocation processing on each fault component based on the instantaneous fault component data and the permanent fault component data, so as to obtain a fault quantitative assessment result including the instantaneous fault severity level and the permanent fault severity level. The control module is used to process the fault quantification assessment results through a hierarchical reclosing strategy to obtain an adaptive control command containing reclosing delay parameters, reclosing count parameters, and protection setting parameters, and to execute the magnetic recloser opening and closing control operation.

9. An adaptive control device for a magnetic recloser used in circuit design, characterized in that, It includes a memory and a processor, the memory storing a computer program that can run on the processor, the processor executing the computer program to implement the adaptive control method for magnetic reconciliation for circuit design according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, it causes the processor to execute the adaptive control method for magnetic reconciliation in circuit design as described in any one of claims 1 to 7.

Images