Partial discharge fault diagnosis method, device, equipment, medium and product

Abstract

The invention relates to a partial discharge fault diagnosis method and device, equipment, a medium and a product. The method comprises the following steps: performing feature extraction on sample partial discharge data of sample gas insulated switchgear under different fault types to obtain sample partial discharge features; constructing an initialized population; population individuals in the initialized population correspond to parameter combinations of to-be-trained parameters; updating individual positions of population individuals in the initialized population according to the sample partial discharge features to optimize the initialized population and obtain a target population; according to a target parameter combination in the target population, initializing model parameters of the extreme learning machine model; and performing model training on the initialized extreme learning machine model according to the sample partial discharge characteristics to obtain a fault diagnosis model which is used for diagnosing the fault type of the partial discharge signal. By adopting the method, the fault type of partial discharge can be accurately diagnosed.

Description

Technical Field

[0001] This application relates to the field of fault diagnosis technology, and in particular to a method, apparatus, equipment, medium and product for diagnosing partial discharge faults. Background Technology

[0002] In power systems, gas-insulated switchgear (GIS) is prone to insulation defects during installation and operation. With the development of partial discharge fault diagnosis technology, the safety risks of partial discharge caused by such defects have attracted more attention; and partial discharge itself can seriously threaten the operational reliability of equipment. Therefore, fault diagnosis of partial discharge in GIS is extremely important to ensure the safe and stable operation of the power grid.

[0003] Current partial discharge fault diagnosis methods rely on diagnostic models tuned manually based on experience, making them prone to getting trapped in local optima. Furthermore, their generalization ability and diagnostic accuracy are difficult to guarantee when faced with complex and variable discharge characteristics. Therefore, existing partial discharge fault diagnosis methods suffer from insufficient accuracy, a problem that urgently needs to be addressed. Summary of the Invention

[0004] Therefore, it is necessary to provide a fault diagnosis method, device, equipment, medium, and product for partial discharge that can accurately diagnose fault types, addressing the aforementioned technical problems.

[0005] In a first aspect, this application provides a method for diagnosing partial discharge faults, comprising:

[0006] Feature extraction was performed on the partial discharge data of the sample gas-insulated switchgear under different fault types to obtain the sample partial discharge characteristics.

[0007] Construct an initial population; the individuals in the initial population correspond to the parameter combinations of each parameter to be trained;

[0008] The individual positions of the individuals in the initial population are updated based on the partial discharge characteristics of the sample to optimize the initial population and obtain the target population.

[0009] The model parameters of the extreme learning machine model are initialized based on the target parameter combinations in the target population.

[0010] The initialized Extreme Learning Machine model is trained based on the partial discharge characteristics of the sample to obtain a fault diagnosis model, which is used to diagnose the fault type of the partial discharge signal.

[0011] In one embodiment, updating the individual positions of individuals in the initial population based on the sample partial discharge characteristics to optimize the initial population and obtain the target population includes: for each iteration, evaluating the population fitness of each parameter combination in the population to be optimized in the current iteration based on the sample partial discharge characteristics, and obtaining the population fitness of each parameter combination in the current iteration; wherein, the population to be optimized in the current iteration is the population optimization result of the previous iteration; the optimized population in the first iteration is the initial population; determining the individual role of each parameter combination in the population to be optimized in the current iteration based on the population fitness of each parameter combination in the current iteration; updating the position of the population to be optimized in the current iteration based on the population fitness of each parameter combination in the current iteration and the corresponding individual role, and obtaining the population optimization result of the current iteration; if the current iteration is not less than the iteration number threshold, using the population optimization result of the current iteration as the target population.

[0012] In one embodiment, the step of updating the position of the population to be optimized in the current iteration based on the population fitness and corresponding individual role of each parameter combination in the current iteration to obtain the population optimization result of the current iteration includes: determining the adaptive weight and step size adjustment factor corresponding to the current iteration; for the individual role belonging to the discoverer in the population to be optimized, updating the position of the individual role corresponding to the discoverer in the population to be optimized based on the individual position data of the population individual corresponding to the discoverer, the preset flight strategy parameters, the adaptive weight, the step size adjustment factor, and the individual position data of the individual at the boundary position corresponding to the discoverer; for the individual role belonging to the follower in the population to be optimized, updating the position of the individual role corresponding to the discoverer based on the adaptive weight and step size adjustment factor; and updating the position of the individual role corresponding to the follower in the population to be optimized based on the adaptive weight and step size adjustment factor. The position of the follower in the population to be optimized is updated based on the individual position data of the individual in the population corresponding to the follower, the individual position data of the individual at the boundary position corresponding to the follower, the individual position data of any individual in the population to be optimized, the individual index of the individual in the population corresponding to the follower, and the population size. For the individual role belonging to the vigilant in the population to be optimized, the position of the vigilant in the population to be optimized is updated based on the individual position data of the individual in the population corresponding to the vigilant, the individual position data of the individual at the boundary position corresponding to the vigilant, the population fitness of the individual in the population corresponding to the vigilant in the current iteration round, and the maximum population fitness of the population to be optimized in the current iteration round.

[0013] In one embodiment, the calculation formula for updating the position of the individual role corresponding to the discoverer in the population to be optimized, based on the individual position data of the individual population corresponding to the discoverer, preset flight strategy parameters, the adaptive weight, the step size adjustment factor, and the individual position data of the individual at the boundary position corresponding to the discoverer, is as follows: , where X i，j (t) represents the individual location data of the discoverer in the corresponding population, w(t) is the adaptive weight, and X g，j (t) represents the individual location data of the individual corresponding to the boundary position of the discoverer. The step size adjustment factor is denoted by rand, which is a random number between 0 and 1. L is the preset flight strategy parameter, and X is the step size adjustment factor. i，j (t+1) represents the updated individual position data of the individual in the population corresponding to the discoverer, where t is the current iteration round and t+1 is the next iteration round of the current iteration round.

[0014] In one embodiment, for individual roles belonging to the follower in the population to be optimized, the position update of the individual roles corresponding to the follower in the population to be optimized is performed based on the individual position data of the individual in the population corresponding to the follower, the individual position data of the individual at the boundary position corresponding to the follower, the individual position data of any individual in the population to be optimized, the individual index of the individual in the population corresponding to the follower, and the population size. This includes: for individual roles belonging to the follower in the population to be optimized whose corresponding population fitness is within a preset division ratio in the fitness ranking of the follower group, the calculation formula based on the individual position data of the individual in the population corresponding to the follower, the individual position data of the individual at the boundary position corresponding to the follower, the individual index of the individual in the population corresponding to the follower, and the population size is as follows: , where X i，j (t+1) represents the updated individual position data of the follower's corresponding population individual, X i，j (t) represents the individual location data of the follower in the population, X best，j (t) represents the individual position data of the individual with the maximum fitness in the population to be optimized, i represents the individual index of the individual corresponding to the follower, n represents the population size, t represents the current iteration round, and t+1 represents the next iteration round; for individual roles in the population to be optimized that belong to the follower and whose corresponding population fitness is not within the preset division ratio in the fitness ranking of the follower group, the calculation formula based on the individual position data of the individual corresponding to the follower, the individual position data of the individual corresponding to the boundary position of the follower, and the individual position data of any individual in the population to be optimized is as follows: , where X i，j(t+1) represents the updated individual position data of the follower's corresponding population individual, X i，j (t) represents the individual location data of the follower in the population, X worst，j (t) represents the individual location data of the individual with the minimum fitness in the population to be optimized, X. r，j (t) represents the individual location data of any individual in the population to be optimized. `rand` is a random number between [0, 1], and `rand` is a random number between 0 and 1. Follows a normal distribution t is the current iteration round, and t+1 is the next iteration round after the current iteration round.

[0015] In one embodiment, the step of updating the position of the individual role corresponding to the vigilant in the population to be optimized, based on the individual position data of the vigilant's corresponding population individual, the individual position data of the vigilant's corresponding boundary position individual, the population fitness of the vigilant's corresponding population individual in the current iteration round, and the maximum population fitness of the population to be optimized in the current iteration round, includes: , where X i，j (t+1) represents the updated individual location data of the vigilant in the corresponding population, X i，j (t) represents the individual location data of the vigilant population, X best，j (t) represents the individual location data of the individual with the maximum fitness in the population to be optimized. A random number between [-1, 1] X represents the population fitness of the individual corresponding to the vigilant in the current iteration round. i Let i be the individual location data of the i-th individual in the population. This represents the maximum fitness value of the population to be optimized in the current iteration round. Let t be a very small positive number, where t is the current iteration round, and t+1 is the next iteration round after the current iteration round.

[0016] Secondly, this application also provides a partial discharge fault diagnosis device, comprising:

[0017] The processing module is used to extract features from the partial discharge data of the sample gas-insulated switchgear under different fault types to obtain the sample partial discharge features.

[0018] A construction module is used to construct an initial population; the individuals in the initial population correspond to the parameter combinations of each parameter to be trained.

[0019] The update module is used to update the individual positions of the population individuals in the initial population according to the partial discharge characteristics of the sample, so as to optimize the initial population and obtain the target population.

[0020] The optimization module is used to initialize the model parameters of the extreme learning machine model based on the target parameter combination in the target population.

[0021] The training module is used to train the initialized extreme learning machine model based on the partial discharge characteristics of the sample to obtain a fault diagnosis model, which is used to diagnose the fault type of the partial discharge signal.

[0022] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0023] Feature extraction was performed on the partial discharge data of the sample gas-insulated switchgear under different fault types to obtain the sample partial discharge characteristics.

[0024] Construct an initial population; the individuals in the initial population correspond to the parameter combinations of each parameter to be trained;

[0025] The individual positions of the individuals in the initial population are updated based on the partial discharge characteristics of the sample to optimize the initial population and obtain the target population.

[0026] The model parameters of the extreme learning machine model are initialized based on the target parameter combinations in the target population.

[0027] The initialized Extreme Learning Machine model is trained based on the partial discharge characteristics of the sample to obtain a fault diagnosis model, which is used to diagnose the fault type of the partial discharge signal.

[0028] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0029] Feature extraction was performed on the partial discharge data of the sample gas-insulated switchgear under different fault types to obtain the sample partial discharge characteristics.

[0030] Construct an initial population; the individuals in the initial population correspond to the parameter combinations of each parameter to be trained;

[0031] The individual positions of the individuals in the initial population are updated based on the partial discharge characteristics of the sample to optimize the initial population and obtain the target population.

[0032] The model parameters of the extreme learning machine model are initialized based on the target parameter combinations in the target population.

[0033] The initialized Extreme Learning Machine model is trained based on the partial discharge characteristics of the sample to obtain a fault diagnosis model, which is used to diagnose the fault type of the partial discharge signal.

[0034] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0035] Feature extraction was performed on the partial discharge data of the sample gas-insulated switchgear under different fault types to obtain the sample partial discharge characteristics.

[0036] Construct an initial population; the individuals in the initial population correspond to the parameter combinations of each parameter to be trained;

[0037] The individual positions of the individuals in the initial population are updated based on the partial discharge characteristics of the sample to optimize the initial population and obtain the target population.

[0038] The model parameters of the extreme learning machine model are initialized based on the target parameter combinations in the target population.

[0039] The initialized Extreme Learning Machine model is trained based on the partial discharge characteristics of the sample to obtain a fault diagnosis model, which is used to diagnose the fault type of the partial discharge signal.

[0040] The aforementioned methods, devices, equipment, media, and products for partial discharge fault diagnosis extract sample partial discharge features from gas-insulated switchgear under different fault types. By constructing an initial population and updating individual positions based on these features, the system adaptively searches for the globally optimal parameter combination, significantly improving the accuracy and efficiency of model parameter initialization. Then, the optimized target parameter combination is used to initialize the extreme learning machine model, and further combined with sample partial discharge features to complete model training. This significantly improves the model's accuracy and generalization ability in identifying GIS partial discharge fault features under complex operating conditions and varying discharge characteristics. Based on the final fault diagnosis model, accurate identification of GIS partial discharge signal fault types can be achieved, significantly improving the reliability of fault diagnosis for gas-insulated switchgear under complex operating conditions. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart illustrating a partial discharge fault diagnosis method in one embodiment;

[0043] Figure 2 This is a schematic diagram of the PRPD (partial discharge fault diagnosis) pattern of a partial discharge fault diagnosis method in one embodiment;

[0044] Figure 3 This is a schematic diagram of the test platform for building a partial discharge fault diagnosis method in one embodiment;

[0045] Figure 4 This is a flowchart illustrating a partial discharge fault diagnosis method in another embodiment;

[0046] Figure 5 This is a flowchart illustrating a partial discharge fault diagnosis method in another embodiment;

[0047] Figure 6 This is a flowchart illustrating a partial discharge fault diagnosis method in another embodiment;

[0048] Figure 7 This is a structural block diagram of a partial discharge fault diagnosis device in one embodiment;

[0049] Figure 8 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0051] In one embodiment, such as Figure 1 As shown, a fault diagnosis method for partial discharge is provided. This embodiment illustrates the method by applying it to a computer device. It is understood that this method can also be applied to a server, and to a system including both a computer device and a server, and is implemented through the interaction between the computer device and the server. In this embodiment, the method includes the following steps:

[0052] S101, feature extraction is performed on the partial discharge data of the sample gas-insulated switchgear under different fault types to obtain the sample partial discharge features.

[0053] The sample gas-insulated switchgear is a pre-selected GIS device known to have typical insulation defects, used for sample partial discharge data acquisition. In this embodiment, typical insulation defects are specific physical flaws that cause distortion of the electric field distribution and constitute potential insulation hazards. These defects are the root cause of partial discharge, and their nature directly determines the physical mechanism and external manifestation mode of partial discharge, i.e., the fault type. Fault types include, but are not limited to, conductor spikes, floating discharge, etc.; the sample partial discharge data consists of partial discharge signal samples containing different fault type labels; the sample partial discharge features are quantitative parameters extracted from the sample partial discharge data that characterize the essential differences between different fault types.

[0054] Optionally, feature extraction can be performed on the partial discharge data of sample GIS equipment under different fault types, using the following method: For the partial discharge data of sample GIS equipment under different fault types, the phase intervals of the partial discharge data are subdivided according to a pre-set power frequency cycle. The pulse amplitude and occurrence frequency within each phase segment are statistically analyzed. A high-resolution PRPD (Phase-Resolved Partial Discharge) map of the partial discharge data is generated based on the pulse amplitude and occurrence frequency within each phase segment. The horizontal axis of the PRPD map represents phase, which refers to the number of phase intervals of the partial discharge data subdivided according to a pre-set power frequency cycle. The vertical axis represents partial discharge (pC), i.e., the partial discharge quantity, measured in picocoulombs, used to measure the intensity of the partial discharge. In this embodiment, the partial discharge quantity is determined based on the pulse amplitude and occurrence frequency within each phase segment. Further time-domain features are extracted from the PRPD map. These time-domain features include, but are not limited to, peak value, mean value, root mean square value, kurtosis, skewness, peak factor, impulse factor, and waveform factor. The temporal features of all extracted sample partial discharge data are subjected to dimensionality reduction processing to obtain sample partial discharge features. In this embodiment, the specific form of the sample partial discharge features is a low-dimensional feature matrix, but other forms are also possible, and this embodiment does not limit this.

[0055] The traditional method of extracting partial discharge features from samples by constructing PRPD maps relies on fixed phase segmentation and single statistical rules, which results in low phase resolution and shallow temporal feature mining, leading to fuzzy discharge pattern recognition.

[0056] Therefore, in this embodiment, a signal-phase correlation analysis environment is built. Based on the collected sample partial discharge data, a power phase acquisition module (with an accuracy set to ≤0.1°) is simultaneously connected. Using a pre-set power frequency period as the time dimension, the phase interval is divided into multiple sub-intervals with a granularity of 1°. For example, taking a power frequency period of 360° and a corresponding time of 0.02s as an example, 360 sub-intervals can be obtained, and the phase sub-intervals... i = 0, 1, 359. For each phase subinterval The number of discharge pulses and their corresponding amplitudes are statistically analyzed to construct a pulse number-amplitude-phase mapping relationship and generate a phase-resolved partial discharge (PRPD) map of the corresponding insulation defect.

[0057] Among them, the generated PRPD is as follows Figure 2 As shown, the horizontal axis represents the phase angle from 0° to 360°, and the vertical axis represents the pulse amplitude. The color level / dot density is correlated with the number of pulses, visually presenting the discharge phase distribution pattern. Based on the PRPD spectrum, eight time-domain features are extracted according to the time-domain feature definition: peak value, mean value, root mean square value, kurtosis, skewness, peak factor, pulse factor, and waveform factor. For the extracted feature sequences, envelope analysis is used to construct upper and lower envelopes: assuming the feature sequence length is L and the sliding window length is w=5, the upper envelope (maximum value within the window) and lower envelope (minimum value within the window) are calculated. Normalization is used to suppress noise interference and enhance feature recognition, forming a preprocessed 8-dimensional time-domain feature set. For the 8-dimensional time-domain features after envelope analysis, the t-SNE algorithm is used for dimensionality reduction. High-dimensional similarity is calculated. Low-dimensional similarity Minimize the above similarity This yields a low-dimensional feature matrix. Where x i and x j For two sample points in a high-dimensional space, σ i It is the sample point x i The standard deviation of y, where k is the index of other sample points in the calculated data. Similarly, y i and y j These are two sample points in a low-dimensional space.

[0058] In addition to the above-mentioned method of extracting sample partial discharge features by constructing PRPD maps, other methods can also be used to extract features from sample partial discharge data of sample GIS equipment under different fault types. For example, sample partial discharge features can be obtained by performing time domain analysis, frequency domain analysis, or time-frequency domain analysis on sample partial discharge data. This embodiment does not limit this.

[0059] Regarding how to collect sample partial discharge data of GIS equipment under different fault types, traditional GIS partial discharge signal acquisition relies on a single sensor and a fixed acquisition process, which has problems such as incomplete signal coverage (e.g., weak discharges are easily missed) and weak anti-interference ability (affected by electromagnetic noise, features are easily distorted), resulting in insufficient reliability of diagnostic data.

[0060] In this embodiment, a pre-built test platform can be used to collect partial discharge data of the sample gas-insulated switchgear under different fault types. The test platform is as follows: Figure 3 As shown. The platform construction and data acquisition process are as follows: A GIS test prototype equipped with typical insulation defects (conductor spikes: copper, 0.4 mm in diameter, 4 mm in length; floating discharge: aluminum foil, 0.15 mm thick, 1.2 cm² in area; metal particles: 0.3-0.5 mm in diameter; insulator surface particles: 150 μm in diameter) was selected. A high-frequency current sensor (HFCT, bandwidth 5-25 MHz, sensitivity 10 mV / mA), a pulse generator (nanosecond-level pulse, 0-20 kV adjustable), an oscilloscope (8 GS / s sampling rate, 2 GHz bandwidth), and a data acquisition device (8-channel parallel) were configured to construct a partial discharge signal acquisition system. The system was tested by debugging the sample GIS equipment. The initial voltage is set to 0kV, and the chamber pressure is maintained at 0.4MPa. The voltage is gradually increased from the initial voltage in 5kV steps, stabilizing for 2 minutes at each step, until it precisely reaches 1.2 times the discharge initiation voltage and operates stably for 10 minutes, creating a test condition where partial discharge signals can be stably acquired. The data acquisition device is then activated, and for each typical insulation defect, raw partial discharge signals (i.e., sample partial discharge data) are continuously acquired according to a preset acquisition cycle (e.g., one set every 10 seconds, each set covering a 1-2 second pulse cycle). Multiple sets (e.g., no less than 200 sets) of samples are acquired for each defect, with a 2-minute pause every 50 sets to prevent signal drift, forming a partial discharge signal sample library containing different defect type labels. In this embodiment, all parameters involved in the construction of the test platform and the data acquisition process can be adjusted according to actual needs, and this embodiment does not impose any limitations on this.

[0061] S102, Construct an initial population; the individuals in the initial population correspond to the parameter combinations of each parameter to be trained.

[0062] The initial population is a set of candidate solutions generated when the intelligent optimization algorithm (such as the improved sparrow search algorithm) is started; each candidate solution is a set of parameters of the Extreme Learning Machine (ELM); each individual in the initial population is a single candidate solution in the initial population and uniquely corresponds to a set of parameters to be trained.

[0063] It should be noted that the parameters to be trained refer to the key ELM parameters that need to be optimized through intelligent optimization algorithms, including but not limited to input weights and hidden layer biases; the parameter combination is a specific set of values ​​composed of the above-mentioned parameters to be trained.

[0064] Optionally, to improve initial diversity and avoid the algorithm getting stuck in local regions in the initial stage, the improved Sparrow Search Algorithm (ISSA) can be used to construct the initial population in this embodiment. The specific process is as follows: Initialize the individuals in the population based on a chaotic sequence. Determine the total dimension and the value range of each dimension of the parameters to be trained according to the structure of the Extreme Learning Machine (ELM); then generate a chaotic sequence matching the target number through Logistic chaotic mapping, where the target number is determined by the product of the population size and the individual dimension; then convert the chaotic sequence into the position of the individuals in the population through linear mapping to form the initial population. In this embodiment, the population size ranges from 20 to 30, and can also be adjusted according to actual needs; this embodiment does not impose any restrictions on this. In addition to the above method of constructing the initial population, the initial population can also be constructed according to random initialization; this embodiment does not limit the method of constructing the initial population.

[0065] The formula for the Logistic chaotic mapping involved in the construction process is as follows: Z k For an operator, Z k+1 This is the value of the operator in the next iteration. When a specific value is taken (e.g., 4), the Logistic mapping is in a chaotic state. Then through Mapping chaotic sequences to the location space of individuals in the population. and These are the minimum and maximum values ​​of the j-th dimension variable, respectively. An initial population is constructed using chaotic mapping, resulting in a more even distribution of the initial population and coverage of a wider search region, laying the foundation for subsequent optimization.

[0066] It should be noted that the Extreme Learning Machine (ELM) model is a fast learning algorithm for single-hidden-layer feedforward neural networks. It simplifies the training process by randomly generating the weights and biases from the input layer to the hidden layers and keeping them unchanged during training. Suppose there are N arbitrary samples... ,in x in This is the nth input feature of the i-th sample; , t im Let m be the m-th output feature of the i-th sample. A single-hidden-layer neural network with L hidden nodes can be represented as:

[0067]

[0068] In the formula: For output weights; b is the activation function; i This is the bias of the i-th hidden layer unit; Input weights; For W i and X j The inner product; o j Let be the output of the j-th hidden layer. The learning objective of a single-hidden-layer neural network is to minimize the output error, which can be expressed as:

[0069]

[0070] Among them, t j This is the j-th output result;

[0071] That is, it exists W i and b i , so that:

[0072]

[0073] Represented in matrix form as follows:

[0074]

[0075] In the formula: H is the output of the hidden layer node. Let T be the output weight and T be the desired output.

[0076] Make It becomes:

[0077]

[0078]

[0079] In order to train a single-hidden-layer neural network, we hope to obtain... , and , so that: In the formula, W L b is the Lth weight; L Let m be the Lth threshold; N be the number of hidden layers; L be the total number of input features, and m be the total number of output features; this formula is equivalent to minimizing the loss function:

[0080]

[0081] In the ELM model, if the input weights W i and the bias b of the hidden layer iSince the hidden layer is randomly determined, its output matrix H is also uniquely determined. Training a single-hidden-layer neural network can be transformed into solving a linear system. The output weights It can be determined as:

[0082]

[0083] In the formula, Let H be the Moore-Penrose generalized inverse of matrix H. Furthermore, it can be proven that the obtained norm... It is the smallest and only solution.

[0084] This implementation uses the Sigmoid, Sine, and Hardlim functions, and selects the optimal activation function through an improved Sparrow Algorithm (ISSA) to achieve better classification performance. The population corresponding to the ELM model is constructed using ISSA, and the basic parameters of the ELM model are configured.

[0085] S103, update the individual positions of the individuals in the initial population according to the partial discharge characteristics of the sample to optimize the initial population and obtain the target population.

[0086] The target population is the set of candidate solutions with better fitness obtained after one or more rounds of iterative optimization of the initial population.

[0087] Optionally, the initial population can be optimized based on the partial discharge characteristics of the samples to obtain the target population. An improved sparrow search algorithm can be used, with the following process: The partial discharge characteristics of the samples are pre-divided into a training set and a validation set. Then, the training set is input into the Extreme Learning Machine (ELM) corresponding to each individual in the initial population to calculate the fitness value of each individual. Next, the individual positions are updated according to the ISSA optimization rules: first, based on the fitness values ​​of individuals in each population, three roles are identified: discoverer, follower, and vigilant. Then, the position update strategy for each role is executed. After each iteration, the training set is re-input into the updated ELM corresponding to each individual in the population to calculate the new fitness value. This iterative process is repeated until the preset maximum number of iterations is reached or the fitness value converges, ultimately obtaining the target population. The classification accuracy of the Extreme Learning Machine on the training set is used as the basis for calculating the fitness value, and the formula for the fitness function is: , where N correct The number of correctly classified samples, N totol This represents the total number of samples in the training set. In this embodiment, the preset maximum number of iterations can range from 50 to 80, and can also be adjusted according to actual needs. This embodiment does not impose any restrictions on this.

[0088] S104, Initialize the model parameters of the Extreme Learning Machine model according to the target parameter combination in the target population.

[0089] The target parameter combination refers to the parameter combination corresponding to the individual with the best fitness in the target population. In this embodiment, the best fitness is the maximum fitness value in the population.

[0090] Optionally, by screening the fitness of individuals in various population groups within the target population, the parameter combination corresponding to the population individual with the optimal fitness is determined and used as the target parameter combination. Based on the parameter combination corresponding to the population individual with the optimal fitness, the parameters of the Extreme Learning Machine model (such as input weights and hidden layer biases) are initialized to obtain the initialized Extreme Learning Machine model, thereby solving the problems of poor feature fitting and weak robustness caused by parameter randomization in traditional models.

[0091] S105, the initialized extreme learning machine model is trained according to the partial discharge characteristics of the sample to obtain a fault diagnosis model, which is used to diagnose the fault type of the partial discharge signal.

[0092] Among them, the fault diagnosis model is the initialized Extreme Learning Machine model after training.

[0093] Optionally, the initialized Extreme Learning Machine (ELM) model is trained using the training set obtained after segmenting the partial discharge features of the samples, resulting in a trained ELM model. The trained ELM model is then evaluated using the validation set obtained after segmenting the partial discharge features of the samples. If the trained ELM model meets the performance indicators, it is used as a fault diagnosis model. The partial discharge signal to be diagnosed in the GIS equipment can then be input into the fault diagnosis model for fault diagnosis, and the fault type output by the fault diagnosis model can be obtained.

[0094] The specific process for training and evaluating the performance of the initialized Extreme Learning Machine (ELM) model based on the training and validation sets, as described above, is as follows: The sample partial discharge features are divided into a training set and a validation set according to a preset ratio (e.g., 8:2); the initialized ELM model with ISSA-optimized parameters is called, using the training set data as input, to drive the model to learn the mapping relationship between the input layer (feature dimension) and the output layer (fault type label), thus completing the model training process. A diagnostic system verification operation is then performed on the trained ELM model: the feature data corresponding to the validation set is extracted and input into the trained ELM model; the fault type prediction results output by the trained ELM model are obtained and compared with the actual defect labels in the validation set; based on the comparison results, the system is evaluated according to the formula... , The diagnostic accuracy and recall of the trained Extreme Learning Machine model are calculated to quantitatively evaluate the model's diagnostic capability. The calculated performance indicators are verified by setting corresponding performance thresholds (e.g., diagnostic accuracy ≥ 95%). If the performance indicators are greater than the corresponding performance thresholds, the diagnostic system is deemed usable.

[0095] When performing fault diagnosis on partial discharge signals of GIS equipment using a fault diagnosis model, the specific process is as follows: For the GIS equipment to be diagnosed, the partial discharge signals of the GIS equipment to be diagnosed are collected, and the corresponding feature data is obtained by extracting features from the partial discharge signals; the fault diagnosis model outputs the partial discharge fault type corresponding to the feature data based on the trained mapping relationship, thus completing the actual diagnosis process.

[0096] In the aforementioned partial discharge fault diagnosis method, computer equipment extracts sample partial discharge features from partial discharge data of gas-insulated switchgear under different fault types. By constructing an initial population and updating individual positions based on the sample partial discharge features, it can adaptively search for the globally optimal parameter combination, thereby significantly improving the accuracy and efficiency of model parameter initialization. Then, the optimized target parameter combination is used to initialize the extreme learning machine model, and further combined with the sample partial discharge features to complete model training, significantly improving the model's recognition accuracy and generalization ability for GIS partial discharge fault features under complex operating conditions and varying discharge characteristics. Based on the final fault diagnosis model, accurate identification of GIS partial discharge signal fault types can be achieved, significantly improving the reliability of fault diagnosis for gas-insulated switchgear under complex operating conditions.

[0097] In one embodiment, the method for optimizing the initialization of the population is further elaborated, such as... Figure 4 As shown, the individual positions of the population in the initial population are updated according to the partial discharge characteristics of the sample to optimize the initial population and obtain the target population, specifically including S201 to S204:

[0098] S201, for each iteration, based on the partial discharge characteristics of the sample, the population fitness of each parameter combination in the population to be optimized in the current iteration is evaluated to obtain the population fitness of each parameter combination in the current iteration; wherein, the population to be optimized in the current iteration is the population optimization result of the previous iteration; the optimized population in the first iteration is the initial population.

[0099] Population fitness refers to the classification accuracy of each parameter combination in the current iteration round.

[0100] Optionally, in each iteration, the parameter combinations of the current population to be optimized are substituted into the Extreme Learning Machine (ELM) model under the current iteration. The training set corresponding to the partial discharge features of the samples drives the ELM model under the current iteration to calculate the prediction results. The population fitness of each parameter combination is obtained by comparing the number of correctly classified samples with the total number of samples in the training set. In the first iteration, the above evaluation is performed directly using the initialized population as the population to be optimized. The evaluation results are used to guide the position update of individuals in the next round.

[0101] S202, based on the population fitness of each parameter combination in the current iteration round, determine the individual role of each parameter combination in the population to be optimized in the current iteration round.

[0102] In this context, the individual role refers to the functional positioning assigned to individuals in the population to be optimized based on the population fitness. Taking the improved sparrow search algorithm as an example, the individual roles include discoverer, follower, and vigilant. Global optimization is achieved through the division of roles among discoverer, follower, and vigilant.

[0103] Optionally, the parameter combinations of the population to be optimized in the current iteration are sorted from best to worst according to the population fitness. The parameter combination with the best fitness is assigned as the discoverer responsible for global exploration according to a preset ratio (e.g., 20%), and the rest (e.g., the remaining 80%) are assigned as followers to search with the discoverer. A small number of individuals (e.g., 10%) are randomly selected as vigilant individuals to monitor environmental changes. The individual role division is completed in the above way to guide subsequent position updates.

[0104] S203, based on the population fitness and corresponding individual roles of each parameter combination in the current iteration round, update the position of the population to be optimized in the current iteration round to obtain the population optimization result of the current iteration round.

[0105] The population optimization result is the population obtained after optimizing the position of each individual in the population to be optimized in the current iteration round.

[0106] Optionally, for individuals with different roles in the population to be optimized in the current iteration, their positions are updated according to the corresponding strategies: the discoverer explores the global optimal area by flying through Levy based on the population fitness, the follower moves closer to high-quality individuals or randomly perturbs according to the fitness ranking, and the watcher makes fine adjustments around the optimal position to avoid local optima. After all individuals are updated, the population optimization result of the current iteration is formed.

[0107] S204, if the current iteration round is not less than the iteration number threshold, the population optimization result of the current iteration round is taken as the target population.

[0108] The iteration threshold is a critical value used to determine whether to terminate the iteration.

[0109] Optionally, if the current iteration number is not less than the critical number of iterations used to determine whether to terminate the iteration, it indicates that the iteration termination condition has been met, and the optimization result of the population to be optimized in the current iteration is determined as the target population.

[0110] In addition to the above methods, an iteration termination condition can be triggered if the fitness function value does not improve during a series of preset iterations (e.g., 5 times), thereby determining the optimization result of the population to be optimized in the current iteration as the target population.

[0111] In this embodiment, the population fitness of each parameter combination in the current iteration is calculated based on the partial discharge characteristics of the samples. Then, based on the ranking of the population fitness of each parameter combination, the individual role of each parameter combination in the target population in the current iteration is determined. The position of the target population in the current iteration is updated according to the assigned individual roles to obtain the corresponding population optimization result. Finally, when the iteration termination condition is met in the current iteration, the population optimization result of the current iteration is used as the target population. This scheme improves the model's ability to recognize complex features of partial discharge in GIS.

[0112] In one embodiment, the method for optimizing the population to be optimized based on the population fitness and corresponding individual roles of various parameter combinations is further elaborated, such as... Figure 5 As shown, the position of the population to be optimized in the current iteration is updated according to the population fitness and corresponding individual role of each parameter combination in the current iteration, to obtain the population optimization result of the current iteration, specifically including S301 to S304:

[0113] S301, determine the adaptive weight and step size adjustment factor corresponding to the current iteration round.

[0114] Among them, the adaptive weight is a weight parameter that is dynamically adjusted with each iteration round, used to balance the global exploration and local development capabilities. The step size adjustment factor is a parameter that controls the magnitude of the individual position update in the population, which changes dynamically with the number of iterations, used to avoid convergence oscillations caused by an excessively large step size or getting stuck in a local optimum due to an excessively small step size.

[0115] Optionally, in this embodiment, to better balance global exploration and local development, ISSA introduces adaptive weights, the formula of which is:

[0116]

[0117] Where w(t) is the weight parameter corresponding to the current iteration round, and t is the current iteration round. max(Maximum number of iterations).

[0118] In this embodiment, the step size adjustment factor is only used in the discoverer, and its calculation formula is as follows:

[0119]

[0120] in, As the initial factor, This represents the current iteration number. (Maximum number of iterations). Initial search phase Approaching 2, the step size is larger to expand the exploration range, and gradually decreases to 0 in the later stage to focus on fine-grained local search, balancing global exploration and local development capabilities.

[0121] S302, for the individual role belonging to the discoverer in the population to be optimized, the position of the individual role corresponding to the discoverer in the population to be optimized is updated according to the individual position data of the individual in the population corresponding to the discoverer, the preset flight strategy parameters, the adaptive weight, the step size adjustment factor, and the individual position data of the individual at the boundary position corresponding to the discoverer.

[0122] In this embodiment, the individual location data corresponding to the boundary position of the discoverer is the location data corresponding to the individual with the worst fitness among all historical populations. The individual with the worst fitness is the individual with the smallest fitness value in the population.

[0123] Optionally, for the discoverer in the population to be optimized, a preset flight strategy is invoked, and its current individual position data is combined with the adaptive weight to adjust the ratio of global exploration and local development. The update magnitude is controlled by the step size adjustment factor. At the same time, the position data of the population individual with the worst fitness in all historical populations and the individual position data of the discoverer's corresponding population individual are updated to complete the position update of the discoverer's individual role.

[0124] It should be noted that the formula for calculating the location update for the discoverer's corresponding individual role is as follows:

[0125]

[0126] The formula for incorporating adaptive weights into the location update of the discoverer is as follows:

[0127]

[0128] Among them, X i，j (t) represents the individual location data of the discoverer in the corresponding population (i.e., the value of the j-th variable of the i-th individual in the t-th iteration), w(t) is the adaptive weight, and X g，j(t) represents the individual location data of the individual corresponding to the boundary position of the discoverer. The step size adjustment factor is denoted by rand, which is a random number between 0 and 1 (i.e., a random number operator). L is a preset flight strategy parameter, and X is the random number generator. i，j (t+1) represents the updated individual position data of the individual in the population corresponding to the discoverer, t represents the current iteration round, t+1 represents the next iteration round of the current iteration round, i represents the individual number, and j represents the variable number on that individual.

[0129] S303, for the individual roles belonging to followers in the population to be optimized, update the position of the individual roles corresponding to the followers in the population to be optimized based on the individual position data of the individual in the population corresponding to the followers, the individual position data of the individual in the boundary position corresponding to the followers, the individual position data of any individual in the population to be optimized, the individual number of the individual in the population corresponding to the followers, and the population size.

[0130] Among them, the individual position data of the follower corresponding to the boundary position includes the worst individual position and the best individual position in the current population.

[0131] Optionally, for followers in the population to be optimized, the update amount is calculated based on their individual index and the population size, taking the position data of randomly selected high-quality individuals in the population as a reference, and combining it with their own current individual position data. At the same time, the new position is constrained to be within an effective range by the followers according to the worst individual position and the best individual position in the current population, thus completing the position update of the followers.

[0132] It should be noted that when followers are in the latter half of the population, the position update formula is:

[0133]

[0134] Among them, X i，j (t+1) represents the updated individual position data of the follower's corresponding population individual, X i，j (t) represents the individual location data of the follower in the population, X worst，j (t) represents the individual location data of the individual with the minimum fitness in the population to be optimized, X. r，j (t) represents the individual location data of any individual in the population to be optimized. `rand` is a random number between [0, 1], and `rand` is a random number between 0 and 1. Follows a normal distribution .

[0135] When followers are in the first half of the population, the position update formula is:

[0136]

[0137] Among them, X i，j (t+1) represents the updated individual position data of the follower's corresponding population individual, X i，j (t) represents the individual location data of the follower in the population, X best，j (t) represents the individual position data of the individual with the maximum fitness in the population to be optimized, i is the individual index of the individual in the population corresponding to the follower, and n is the population size.

[0138] S304, for the individual role belonging to the vigilant in the population to be optimized, update the position of the individual role corresponding to the vigilant in the population to be optimized based on the individual position data of the individual in the population corresponding to the vigilant, the individual position data of the individual at the boundary position corresponding to the vigilant, the population fitness of the individual in the population corresponding to the vigilant in the current iteration round, and the maximum population fitness of the population to be optimized in the current iteration round.

[0139] Among them, the individual position data of the individual corresponding to the boundary position of the vigilant is the optimal individual position of the current population, and the maximum population fitness value in the current iteration of the population to be optimized is the optimal fitness value of the current population.

[0140] Optionally, for the vigilant in the population to be optimized, its current population fitness is compared with the maximum population fitness. If its fitness is poor (e.g., large error), its individual position data is updated by random perturbation, and the new position is constrained within the effective range by the vigilant's corresponding boundary position data. If its fitness is good, the position is maintained or finely adjusted to complete the vigilant's position update.

[0141] It should be noted that the calculation formula for the location update of the vigilant for the corresponding individual role is as follows:

[0142]

[0143] Among them, X i，j (t+1) represents the updated individual location data of the vigilant in the corresponding population, X i，j (t) represents the individual location data of the vigilant population, X best，j (t) represents the individual location data of the individual with the maximum fitness in the population to be optimized. A random number between [-1, 1] X represents the population fitness of the individual corresponding to the vigilant in the current iteration round. i Let i be the individual location data of the i-th individual in the population. This represents the maximum fitness value of the population to be optimized in the current iteration round. It is a very small positive number.

[0144] In this embodiment, by dynamically determining the adaptive weight and step size adjustment factor of the current iteration, the algorithm's global exploration and local development capabilities are balanced. At the same time, for the discoverer, efficient global search is achieved by combining flight strategy, adaptive parameters and boundary constraints. Followers follow in a targeted manner based on individual index, population size and information on high-quality individuals to improve optimization efficiency. The vigilant updates by comparing its own fitness with the optimal fitness of the population and combining boundary constraints to avoid local optima. Finally, the accurate search and efficient convergence of the population in the solution space are achieved, which solves the problem of insufficient diagnostic stability caused by random parameter initialization in ELM.

[0145] To more comprehensively demonstrate this solution, this embodiment presents an optional approach to energy dispatching, such as... Figure 6 As shown, specifically including S401 to S411:

[0146] S401, extract features from the partial discharge data of the sample gas-insulated switchgear under different fault types to obtain the sample partial discharge features;

[0147] S402, Construct an initial population; the individuals in the initial population correspond to the parameter combinations of each parameter to be trained;

[0148] S403, for each iteration, based on the partial discharge characteristics of the sample, the population fitness of each parameter combination in the population to be optimized in the current iteration is evaluated to obtain the population fitness of each parameter combination in the current iteration; wherein, the population to be optimized in the current iteration is the population optimization result of the previous iteration; the optimized population in the first iteration is the initial population.

[0149] S404, Based on the population fitness of each parameter combination in the current iteration round, determine the individual role of each parameter combination in the population to be optimized in the current iteration round;

[0150] S405, determine the adaptive weight and step size adjustment factor corresponding to the current iteration round;

[0151] S406, for the individual role belonging to the discoverer in the population to be optimized, the position of the individual role corresponding to the discoverer in the population to be optimized is updated according to the individual position data of the individual in the population corresponding to the discoverer, the preset flight strategy parameters, the adaptive weight, the step size adjustment factor and the individual position data of the individual at the boundary position corresponding to the discoverer.

[0152] S407, for the individual roles belonging to the followers in the population to be optimized, the position of the individual roles corresponding to the followers in the population to be optimized is updated according to the individual position data of the individual in the population corresponding to the followers, the individual position data of the individual in the boundary position corresponding to the followers, the individual position data of any individual in the population to be optimized, the individual number of the individual in the population corresponding to the followers, and the population size.

[0153] S408, for the individual role belonging to the vigilant in the population to be optimized, update the position of the individual role corresponding to the vigilant in the population to be optimized based on the individual position data of the vigilant corresponding to the individual in the population, the individual position data of the vigilant corresponding to the boundary position individual, the population fitness of the vigilant corresponding to the individual in the current iteration round, and the maximum population fitness of the population to be optimized in the current iteration round.

[0154] S409, if the current iteration round is not less than the iteration number threshold, the population optimization result of the current iteration round is taken as the target population;

[0155] S410, Initialize the model parameters of the Extreme Learning Machine model according to the target parameter combination in the target population;

[0156] S411, The initialized extreme learning machine model is trained according to the partial discharge characteristics of the sample to obtain a fault diagnosis model, which is used to diagnose the fault type of the partial discharge signal.

[0157] The specific processes of S401 to S411 described above can be found in the description of the above method embodiments. Their implementation principles and technical effects are similar, and will not be repeated here.

[0158] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0159] Based on the same inventive concept, this application also provides a partial discharge fault diagnosis device for implementing the partial discharge fault diagnosis method described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations in one or more partial discharge fault diagnosis device embodiments provided below can be found in the limitations of the partial discharge fault diagnosis method described above, and will not be repeated here.

[0160] In one exemplary embodiment, such as Figure 7 As shown, a partial discharge fault diagnosis device is provided, comprising: a processing module 51, a construction module 52, an update module 53, an optimization module 54, and a training module 55, wherein:

[0161] The processing module 51 is used to extract features from the partial discharge data of the sample gas-insulated switchgear under different fault types to obtain the sample partial discharge features.

[0162] The construction module 52 is used to construct an initial population; the individuals in the initial population correspond to the parameter combinations of each parameter to be trained.

[0163] The update module 53 is used to update the individual positions of the population individuals in the initial population according to the sample partial discharge characteristics, so as to optimize the initial population and obtain the target population.

[0164] The optimization module 54 is used to initialize the model parameters of the extreme learning machine model according to the target parameter combination in the target population.

[0165] The training module 55 is used to train the initialized extreme learning machine model based on the partial discharge characteristics of the sample to obtain a fault diagnosis model, which is used to diagnose the fault type of the partial discharge signal.

[0166] In one embodiment, the above Figure 7 The update module 53 includes:

[0167] For each iteration, based on the partial discharge characteristics of the sample, the population fitness of each parameter combination in the population to be optimized in the current iteration is evaluated to obtain the population fitness of each parameter combination in the current iteration; wherein, the population to be optimized in the current iteration is the population optimization result of the previous iteration; the optimized population in the first iteration is the initial population;

[0168] Based on the population fitness of each parameter combination in the current iteration round, determine the individual role of each parameter combination in the population to be optimized in the current iteration round;

[0169] Based on the population fitness and corresponding individual roles of each parameter combination in the current iteration round, the position of the population to be optimized in the current iteration round is updated to obtain the population optimization result of the current iteration round.

[0170] If the current iteration round is not less than the iteration count threshold, the population optimization result of the current iteration round is taken as the target population.

[0171] In one embodiment, the above Figure 7 The update module 53 includes:

[0172] Determine the adaptive weights and step size adjustment factors corresponding to the current iteration round;

[0173] For the individual role belonging to the discoverer in the population to be optimized, the position of the individual role corresponding to the discoverer in the population to be optimized is updated according to the individual position data of the individual in the population corresponding to the discoverer, the preset flight strategy parameters, the adaptive weight, the step size adjustment factor, and the individual position data of the individual at the boundary position corresponding to the discoverer.

[0174] For individual roles belonging to followers in the population to be optimized, the position of the individual role corresponding to the follower in the population to be optimized is updated based on the individual position data of the individual in the population corresponding to the follower, the individual position data of the individual at the boundary position corresponding to the follower, the individual position data of any individual in the population to be optimized, the individual number of the individual in the population corresponding to the follower, and the population size.

[0175] For individuals in the population to be optimized that belong to the vigilant role, the position of the vigilant individual in the population to be optimized is updated based on the individual position data of the vigilant corresponding to the individual in the population, the individual position data of the vigilant corresponding to the boundary position individual in the population, the population fitness of the vigilant corresponding to the individual in the current iteration round, and the maximum population fitness of the population to be optimized in the current iteration round.

[0176] In one embodiment, the above Figure 7 The update module 53 includes:

[0177]

[0178] Among them, X i，j (t) represents the individual location data of the discoverer in the corresponding population, w(t) is the adaptive weight, and X g，j (t) represents the individual location data of the individual corresponding to the boundary position of the discoverer. The step size adjustment factor is denoted by rand, which is a random number between 0 and 1. L is the preset flight strategy parameter, and X is the step size adjustment factor. i，j(t+1) represents the updated individual position data of the individual in the population corresponding to the discoverer, where t is the current iteration round and t+1 is the next iteration round of the current iteration round.

[0179] In one embodiment, the above Figure 7 The update module 53 includes:

[0180] For individual roles in the population to be optimized that are followers and whose fitness ranks within a preset proportion in the follower population, the calculation formula based on the individual position data of the follower's corresponding population individual, the individual position data of the follower's corresponding boundary position individual, the individual index of the follower's corresponding population individual, and the population size is as follows:

[0181]

[0182] Among them, X i，j (t+1) represents the updated individual position data of the follower's corresponding population individual, X i，j (t) represents the individual location data of the follower in the population, X best，j (t) represents the individual position data of the individual in the population with the maximum fitness in the population to be optimized, i is the individual index of the individual in the population corresponding to the follower, n is the population size, t is the current iteration round, and t+1 is the next iteration round of the current iteration round.

[0183] For individual roles in the population to be optimized that are followers and whose fitness is not within a preset division ratio in the fitness ranking of the follower group, the calculation formula is as follows, based on the individual position data of the individual in the population corresponding to the follower, the individual position data of the individual at the boundary position corresponding to the follower, and the individual position data of any individual in the population to be optimized:

[0184]

[0185] Among them, X i，j (t+1) represents the updated individual position data of the follower's corresponding population individual, X i，j (t) represents the individual location data of the follower in the population, X worst，j (t) represents the individual location data of the individual with the minimum fitness in the population to be optimized, X. r，j (t) represents the individual location data of any individual in the population to be optimized. `rand` is a random number between [0, 1], and `rand` is a random number between 0 and 1. Follows a normal distribution t is the current iteration round, and t+1 is the next iteration round after the current iteration round.

[0186] In one embodiment, the above Figure 7 The update module 53 includes:

[0187]

[0188] Among them, X i，j (t+1) represents the updated individual location data of the vigilant in the corresponding population, X i，j (t) represents the individual location data of the vigilant population, X best，j (t) represents the individual location data of the individual with the maximum fitness in the population to be optimized. A random number between [-1, 1] X represents the population fitness of the individual corresponding to the vigilant in the current iteration round. i Let i be the individual location data of the i-th individual in the population. This represents the maximum fitness value of the population to be optimized in the current iteration round. Let t be a very small positive number, where t is the current iteration round, and t+1 is the next iteration round after the current iteration round.

[0189] Each module in the aforementioned partial discharge fault diagnosis device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0190] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 8As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When executed by the processor, the computer program implements a method for diagnosing partial discharge faults. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0191] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0192] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0193] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0194] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0195] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0196] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for diagnosing partial discharge faults, characterized in that, The method includes: Feature extraction was performed on the partial discharge data of the sample gas-insulated switchgear under different fault types to obtain the sample partial discharge characteristics. Construct an initial population; the individuals in the initial population correspond to the parameter combinations of each parameter to be trained; The individual positions of the individuals in the initial population are updated based on the partial discharge characteristics of the sample to optimize the initial population and obtain the target population. The model parameters of the extreme learning machine model are initialized based on the target parameter combinations in the target population. The initialized Extreme Learning Machine model is trained based on the partial discharge characteristics of the sample to obtain a fault diagnosis model, which is used to diagnose the fault type of the partial discharge signal.

2. The method according to claim 1, characterized in that, The step of updating the individual positions of individuals in the initial population based on the partial discharge characteristics of the sample to optimize the initial population and obtain the target population includes: For each iteration, based on the partial discharge characteristics of the sample, the population fitness of each parameter combination in the population to be optimized in the current iteration is evaluated to obtain the population fitness of each parameter combination in the current iteration; wherein, the population to be optimized in the current iteration is the population optimization result of the previous iteration; the optimized population in the first iteration is the initial population; Based on the population fitness of each parameter combination in the current iteration round, determine the individual role of each parameter combination in the population to be optimized in the current iteration round; Based on the population fitness and corresponding individual roles of each parameter combination in the current iteration round, the position of the population to be optimized in the current iteration round is updated to obtain the population optimization result of the current iteration round. If the current iteration round is not less than the iteration count threshold, the population optimization result of the current iteration round is taken as the target population.

3. The method according to claim 2, characterized in that, The step of updating the position of the population to be optimized in the current iteration based on the population fitness and corresponding individual role of each parameter combination in the current iteration, to obtain the population optimization result of the current iteration, includes: Determine the adaptive weights and step size adjustment factors corresponding to the current iteration round; For the individual role belonging to the discoverer in the population to be optimized, the position of the individual role corresponding to the discoverer in the population to be optimized is updated according to the individual position data of the individual in the population corresponding to the discoverer, the preset flight strategy parameters, the adaptive weight, the step size adjustment factor, and the individual position data of the individual at the boundary position corresponding to the discoverer. For individual roles belonging to followers in the population to be optimized, the position of the individual role corresponding to the follower in the population to be optimized is updated based on the individual position data of the individual in the population corresponding to the follower, the individual position data of the individual at the boundary position corresponding to the follower, the individual position data of any individual in the population to be optimized, the individual number of the individual in the population corresponding to the follower, and the population size. For individuals in the population to be optimized that belong to the vigilant role, the position of the vigilant individual in the population to be optimized is updated based on the individual position data of the vigilant corresponding to the individual in the population, the individual position data of the vigilant corresponding to the boundary position individual in the population, the population fitness of the vigilant corresponding to the individual in the current iteration round, and the maximum population fitness of the population to be optimized in the current iteration round.

4. The method according to claim 3, characterized in that, The formula for updating the position of the individual role corresponding to the discoverer in the population to be optimized, based on the individual position data of the individual population corresponding to the discoverer, the preset flight strategy parameters, the adaptive weight, the step size adjustment factor, and the individual position data of the individual at the boundary position corresponding to the discoverer, is as follows: Among them, X i，j (t) represents the individual location data of the discoverer in the corresponding population, w(t) is the adaptive weight, and X g，j (t) represents the individual location data of the individual corresponding to the boundary position of the discoverer. The step size adjustment factor is denoted by rand, which is a random number between 0 and 1. L is the preset flight strategy parameter, and X is the step size adjustment factor. i，j (t+1) represents the updated individual position data of the individual in the population corresponding to the discoverer, where t is the current iteration round and t+1 is the next iteration round of the current iteration round.

5. The method according to claim 3, characterized in that, The step of updating the position of the follower-related individual role in the population to be optimized, based on the individual position data of the follower-related population individual, the individual position data of the follower-related boundary position individual, the individual position data of any population individual in the population to be optimized, the individual index of the follower-related population individual, and the population size, includes: For individual roles in the population to be optimized that are followers and whose fitness ranks within a preset proportion in the follower population, the calculation formula based on the individual position data of the follower's corresponding population individual, the individual position data of the follower's corresponding boundary position individual, the individual index of the follower's corresponding population individual, and the population size is as follows: Among them, X i，j (t+1) represents the updated individual position data of the follower's corresponding population individual, X i，j (t) represents the individual location data of the follower in the population, X best，j (t) represents the individual position data of the individual in the population with the maximum fitness in the population to be optimized, i is the individual index of the individual in the population corresponding to the follower, n is the population size, t is the current iteration round, and t+1 is the next iteration round of the current iteration round. For individual roles in the population to be optimized that are followers and whose fitness is not within a preset division ratio in the fitness ranking of the follower group, the calculation formula is as follows, based on the individual position data of the individual in the population corresponding to the follower, the individual position data of the individual at the boundary position corresponding to the follower, and the individual position data of any individual in the population to be optimized: Among them, X i，j (t+1) represents the updated individual position data of the follower's corresponding population individual, X i，j (t) represents the individual location data of the follower in the population, X worst，j (t) represents the individual location data of the individual with the minimum fitness in the population to be optimized, X. r，j (t) represents the individual location data of any individual in the population to be optimized. `rand` is a random number between [0, 1], and `rand` is a random number between 0 and 1. Follows a normal distribution t is the current iteration round, and t+1 is the next iteration round after the current iteration round.

6. The method according to claim 3, characterized in that, The step of updating the position of the individual role belonging to the vigilant in the population to be optimized, based on the individual position data of the vigilant's corresponding population individual, the individual position data of the vigilant's corresponding boundary position individual, the population fitness of the vigilant's corresponding population individual in the current iteration round, and the maximum population fitness of the population to be optimized in the current iteration round, includes: Among them, X i，j (t+1) represents the updated individual location data of the vigilant in the corresponding population, X i，j (t) represents the individual location data of the vigilant population, X best，j (t) represents the individual location data of the individual with the maximum fitness in the population to be optimized. A random number between [-1, 1] X represents the population fitness of the individual corresponding to the vigilant in the current iteration round. i The species is the location data of the i-th individual in the population. This represents the maximum fitness value of the population to be optimized in the current iteration round. Let t be a very small positive number, where t is the current iteration round, and t+1 is the next iteration round after the current iteration round.

7. A fault diagnosis device for partial discharge, characterized in that, The device includes: The processing module is used to extract features from the partial discharge data of the sample gas-insulated switchgear under different fault types to obtain the sample partial discharge features. A construction module is used to construct an initial population; the individuals in the initial population correspond to the parameter combinations of each parameter to be trained. The update module is used to update the individual positions of the population individuals in the initial population according to the partial discharge characteristics of the sample, so as to optimize the initial population and obtain the target population. The optimization module is used to initialize the model parameters of the extreme learning machine model based on the target parameter combination in the target population. The training module is used to train the initialized extreme learning machine model based on the partial discharge characteristics of the sample to obtain a fault diagnosis model, which is used to diagnose the fault type of the partial discharge signal.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images