Method, device and equipment for predicting evolution trend of mechanical vibration defects of GIS equipment and medium

Abstract

The application discloses a GIS device mechanical vibration defect evolution trend prediction method and device, equipment and medium, relates to the computer technical field, including: for the vibration signal of GIS device of different working conditions and different load current levels, determine the training set and test set; based on the initial mechanical vibration defect evolution trend prediction model, the signal noise reduction processing of joint variation mode decomposition is carried out on the training set, the target gate recurrent unit in the initial mechanical vibration defect evolution trend prediction model is used, the hyperparameter is optimized, the training is carried out by using the hyperparameter optimization result, the target gate recurrent unit and the training set, and the target mechanical vibration defect evolution trend prediction model is determined in combination with the test set; based on the target mechanical vibration defect evolution trend prediction model, the vibration signal to be processed of the GIS device to be processed, the target defect evolution trend prediction result is determined. The application can accurately predict the GIS device mechanical vibration defect evolution trend.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method, apparatus, equipment, and medium for predicting the evolution trend of mechanical vibration defects in GIS equipment. Background Technology

[0002] As the scale of power systems continues to expand, the problem of abnormal vibration due to mechanical defects in GIS equipment (Gas Insulated Switchgear) is becoming increasingly prominent. Long-term abnormal vibration is extremely harmful to the equipment and poses a serious threat to the safe operation of the power grid.

[0003] However, existing solutions mostly focus on predicting power equipment failures in advance, and the research objects mainly include transformers, circuit breakers and other electrical equipment. Due to the latent and complex mechanical defects of GIS equipment, the prediction methods used in existing solutions are difficult to effectively deal with, that is, the prediction effect is poor. Summary of the Invention

[0004] In view of this, the purpose of this invention is to provide a method, apparatus, equipment, and medium for predicting the evolution trend of mechanical vibration defects in GIS equipment. This method enables accurate prediction of the evolution trend of mechanical vibration defects in GIS equipment and improves adaptability to different defect types and load conditions, thereby providing a basis for proactive maintenance of mechanical defects in GIS equipment and ensuring the safe and stable operation of the power system. The specific solution is as follows: Firstly, this application provides a method for predicting the evolution trend of mechanical vibration defects in GIS equipment, including: For different combinations of operating conditions and load current levels, vibration signals of GIS equipment are collected to determine the signal acquisition results. Based on the initial mechanical vibration defect evolution trend prediction model, the signal acquisition results are subjected to signal denoising processing by joint variational mode decomposition, and the training set and test set are determined using the corresponding denoised signal; wherein, the joint variational mode decomposition is a variational mode decomposition optimized by joint mutual information and particle swarm algorithm. Based on the target gated recurrent unit in the initial mechanical vibration defect evolution trend prediction model, hyperparameter optimization is performed on the training set to determine the hyperparameter optimization result. Based on the hyperparameter optimization results and the target gated recurrent unit, the mechanical vibration defect evolution trend prediction is trained for each sample in the training set to determine the prediction training results. Based on the initial mechanical vibration defect evolution trend prediction model, the test set, and the prediction training results, the target mechanical vibration defect evolution trend prediction model that has completed training is determined. Based on the target mechanical vibration defect evolution trend prediction model and the vibration signal of the GIS device to be processed, the prediction result of the target defect evolution trend corresponding to the GIS device to be processed is determined.

[0005] Optionally, the signal denoising process based on the initial mechanical vibration defect evolution trend prediction model, which involves joint variational mode decomposition of the signal acquisition results, and using the corresponding denoised signal to determine the training set and test set, includes: Construct an initial model to predict the evolution trend of mechanical vibration defects; Based on the initial mechanical vibration defect evolution trend prediction model, mutual information criterion, and signal acquisition results, the decomposition layer number is optimized to determine the target decomposition layer number. Based on the initial mechanical vibration defect evolution trend prediction model, the target decomposition layer number, the signal acquisition results, and the particle swarm optimization algorithm, the mutual information difference between the decomposition layers is optimized to determine the target mutual information difference. Based on the initial mechanical vibration defect evolution trend prediction model, the target decomposition layer number, and the target mutual information difference, variational mode decomposition is performed on the vibration signal in the signal acquisition results to determine multiple intrinsic mode functions. Based on the initial mechanical vibration defect evolution trend prediction model, and using the frequency domain features corresponding to each intrinsic mode function, as well as the correlation between the frequency domain features and the corresponding vibration signal, a selection of multiple target intrinsic mode functions is made. The denoised signal is obtained based on the target intrinsic mode function.

[0006] Optionally, determining the training set and test set using the corresponding denoised signal includes: Based on the denoised signal and the corresponding acquisition time of the denoised signal, the data is divided to determine the data division result. Based on the data partitioning results and the vibration trend prediction labels corresponding to each combination, annotations are performed to determine the training set and the test set.

[0007] Optionally, the hyperparameter optimization of the training set based on the target gated recurrent unit in the initial mechanical vibration defect evolution trend prediction model to determine the hyperparameter optimization result includes: The target intrinsic mode functions in the training set are normalized to determine the processed functions; Based on the target gated loop unit in the initial mechanical vibration defect evolution trend prediction model, the parameters of the sparrow search algorithm are initialized to determine the initial parameters; Based on the target gated recurrent unit, the initial parameters, the preset fitness function, and each of the processed functions, hyperparameter optimization is performed to determine the hyperparameter information corresponding to each of the processed functions; the hyperparameter information includes the number of target hidden units, the target training period, and the target initial learning rate. Based on the hyperparameter information, the hyperparameter optimization result is determined.

[0008] Optionally, the step of training the prediction of the mechanical vibration defect evolution trend for each sample in the training set based on the hyperparameter optimization results and the target gated recurrent unit includes: The hyperparameter information is used as the initial hyperparameter of the target gated loop unit. The signals corresponding to each of the processed functions are used to train the prediction of the mechanical vibration defect evolution trend, so as to determine the initial prediction result corresponding to each of the processed functions. The initial prediction results are reorganized to determine the prediction training results.

[0009] Optionally, determining the target mechanical vibration defect evolution trend prediction model based on the initial mechanical vibration defect evolution trend prediction model, the test set, and the prediction training results includes: The test set is input into the target gated loop unit to determine the test results; Based on the test results, the prediction training results, and the preset evaluation indicators, the current evaluation result is determined; Based on the current evaluation results, determine whether the preset model training termination condition is met, and if so, determine the target mechanical vibration defect evolution trend prediction model that has completed training.

[0010] Optionally, determining the predicted evolution trend of the target mechanical vibration defect based on the prediction model and the vibration signal of the GIS device to be processed includes: Based on the prediction model of the evolution trend of the target mechanical vibration defect, the vibration signal of the GIS equipment to be processed is subjected to joint variational mode decomposition signal denoising processing to determine the signal denoising result. Based on the target gated loop unit and the signal denoising result, hyperparameter optimization is performed to determine the target hyperparameters; Based on the target gated loop unit, the target hyperparameters, and the signal denoising results, the predicted result of the target defect evolution trend corresponding to the GIS device to be processed is determined.

[0011] Secondly, this application provides a device for predicting the evolution trend of mechanical vibration defects in GIS equipment, comprising: The signal acquisition module is used to acquire vibration signals from GIS equipment under different operating conditions and load current levels to determine the signal acquisition results. The signal denoising module is used to perform joint variational mode decomposition on the signal acquisition results based on the initial mechanical vibration defect evolution trend prediction model, and to determine the training set and test set using the corresponding denoised signal; wherein, the joint variational mode decomposition is a variational mode decomposition optimized by joint mutual information and particle swarm algorithm. The hyperparameter determination module is used to perform hyperparameter optimization on the training set based on the target gated recurrent unit in the initial mechanical vibration defect evolution trend prediction model, so as to determine the hyperparameter optimization result. The prediction training module is used to train the prediction of the mechanical vibration defect evolution trend of each sample in the training set based on the hyperparameter optimization results and the target gated recurrent unit, so as to determine the prediction training results. The training completion module is used to determine the target mechanical vibration defect evolution trend prediction model that has completed training, based on the initial mechanical vibration defect evolution trend prediction model, the test set, and the prediction training results. The prediction result determination module is used to determine the prediction result of the target defect evolution trend of the GIS device to be processed based on the target mechanical vibration defect evolution trend prediction model and the vibration signal of the GIS device to be processed.

[0012] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor is used to execute the computer program to implement the steps of the aforementioned method for predicting the evolution trend of mechanical vibration defects in GIS equipment.

[0013] Fourthly, this application provides a computer-readable storage medium for storing a computer program, which, when executed by a processor, implements the steps of the aforementioned method for predicting the evolution trend of mechanical vibration defects in GIS equipment.

[0014] As can be seen, in this application, vibration signals of GIS equipment are collected for combinations of different operating conditions and different load current levels to determine the signal acquisition results; based on the initial mechanical vibration defect evolution trend prediction model, the signal acquisition results are subjected to joint variational mode decomposition for signal denoising, and the training set and test set are determined using the corresponding denoised signals; wherein, the joint variational mode decomposition is a variational mode decomposition optimized by joint mutual information and particle swarm optimization algorithm; based on the target gated recurrent unit in the initial mechanical vibration defect evolution trend prediction model, hyperparameter optimization is performed on the training set to determine the hyperparameter optimization result; based on the hyperparameter optimization result and the target gated recurrent unit, mechanical vibration defect evolution trend prediction is trained on each sample in the training set to determine the prediction training result; based on the initial mechanical vibration defect evolution trend prediction model, the test set, and the prediction training result, the target mechanical vibration defect evolution trend prediction model that has completed training is determined; based on the target mechanical vibration defect evolution trend prediction model and the vibration signal of the GIS equipment to be processed, the target defect evolution trend prediction result corresponding to the GIS equipment to be processed is determined. In other words, this application firstly uses vibration signals from GIS equipment corresponding to different combinations of operating conditions and load current levels. Then, based on an initial mechanical vibration defect evolution trend prediction model, the signal acquisition results undergo joint variational mode decomposition for noise reduction to determine the training and test sets. Next, using the training set and the target gated recurrent unit in the initial mechanical vibration defect evolution trend prediction model, the hyperparameter optimization result is determined. Then, based on the target gated recurrent unit, the hyperparameter optimization result, and each sample in the training set, training is performed to predict the mechanical vibration defect evolution trend. Using the prediction training result and the test set, the trained target mechanical vibration defect evolution trend prediction model is determined. Finally, based on the target mechanical vibration defect evolution trend prediction model and the vibration signal of the GIS equipment to be processed, the target defect evolution trend prediction result is determined. This enables accurate prediction of the mechanical vibration defect evolution trend of GIS equipment and improves adaptability to different defect types and load conditions, thereby providing a basis for proactive maintenance of mechanical defects in GIS equipment and ensuring the safe and stable operation of the power system. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0016] Figure 1A flowchart of a method for predicting the evolution trend of mechanical vibration defects in GIS equipment provided in this application; Figure 2 A flowchart of a specific method for predicting the evolution trend of mechanical vibration defects provided in this application; Figure 3 A flowchart of a specific method for predicting the evolution trend of mechanical vibration defects in GIS equipment provided in this application; Figure 4 This application provides a schematic diagram of the structure of a gated loop unit; Figure 5 A schematic diagram of a network architecture for a Transformer-gated recurrent unit is provided in this application; Figure 6(a) is a schematic diagram of the optimization iteration curve of joint variational mode decomposition under different load current levels under normal operating conditions provided in this application; Figure 6(b) is a schematic diagram of the optimization iteration curve of joint variational mode decomposition for different load current levels under the condition of conductor contact loosening defect provided in this application; Figure 7(a) is a schematic diagram of the decomposition results of a joint variational mode decomposition under normal operating conditions provided in this application; Figure 7(b) is a schematic diagram of the decomposition results of the joint variational mode decomposition under the condition of conductor contact loosening defect provided in this application; Figure 8(a) is a schematic diagram of the vibration waveform prediction result of a GIS equipment under normal working conditions provided in this application; Figure 8(b) is a schematic diagram of the vibration waveform prediction results of GIS equipment under the condition of conductor contact loosening defect provided in this application; Figure 9 A schematic diagram of model evaluation indexes under different combinations of operating conditions and different load current levels provided in this application; Figure 10 A schematic diagram of a device for predicting the evolution trend of mechanical vibration defects in GIS equipment provided in this application; Figure 11 This application provides a structural diagram of an electronic device. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] As the scale of power systems continues to expand, the problem of abnormal vibration due to mechanical defects in GIS equipment is becoming increasingly prominent. Long-term abnormal vibration poses a significant threat to the equipment and the safe operation of the power grid. However, existing solutions mainly focus on predicting power equipment failures in advance, with research objects primarily including transformers, circuit breakers, and other electrical equipment. Due to the latent and complex nature of mechanical defects in GIS equipment, the prediction methods used in existing solutions are insufficient to effectively address these issues, resulting in poor prediction performance.

[0019] Therefore, this application provides a scheme for predicting the evolution trend of mechanical vibration defects in GIS equipment, which can accurately predict the evolution trend of mechanical vibration defects in GIS equipment and improve the adaptability to different defect types and different load conditions, thereby providing a basis for the advance maintenance of mechanical defects in GIS equipment and ensuring the safe and stable operation of the power system.

[0020] See Figure 1 As shown in the figure, an embodiment of the present invention discloses a method for predicting the evolution trend of mechanical vibration defects in GIS equipment, including: Step S11: For different combinations of operating conditions and load current levels, collect vibration signals from the GIS equipment to determine the signal acquisition results.

[0021] In this embodiment, vibration signals from the GIS equipment under different operating conditions (e.g., normal, loose conductor contacts) and different load currents (e.g., 600A, 1500A, 2400A) are first collected. It is understood that the sampling rate and the duration of sampling a single sample can also be configured based on actual needs.

[0022] Step S12: Based on the initial mechanical vibration defect evolution trend prediction model, perform signal denoising processing on the signal acquisition results using joint variational mode decomposition, and use the corresponding denoised signal to determine the training set and test set; wherein, the joint variational mode decomposition is a variational mode decomposition optimized by joint mutual information and particle swarm algorithm.

[0023] In this embodiment, a PSO-MVMD-SSA-Transformer-GRU prediction model is proposed, namely, an initial mechanical vibration defect evolution trend prediction model, combined with... Figure 2As shown, the model includes: a variational mode decomposition (VMD) denoising component that combines mutual information and particle swarm optimization (PSO); an SSA-Transformer-GRU prediction component (also known as a target-gated recurrent unit) composed of a Transformer optimized by the Sparrow Search Algorithm (SSA) and a Gated Recurrent Unit (GRU). Then, by combining the designed denoising component and the prediction component, an initial PSO-MVMD-SSA-Transformer-GRU prediction model can be constructed, enabling high-precision prediction of mechanical vibration defects in GIS equipment.

[0024] Specifically, in combination Figure 3 As shown, in this embodiment, firstly, PSO-MVMD denoising is performed on the acquired vibration signal to obtain the intrinsic mode functions (IMFs). Function (representing the basic components of data vibration modes), namely: constructing an initial mechanical vibration defect evolution trend prediction model; based on the initial mechanical vibration defect evolution trend prediction model, mutual information criterion, and signal acquisition results, optimizing the number of decomposition layers to determine the target number of decomposition layers; based on the initial mechanical vibration defect evolution trend prediction model, the target number of decomposition layers, the signal acquisition results, and particle swarm optimization algorithm, optimizing the mutual information difference between decomposition layers to determine the target mutual information difference; based on the initial mechanical vibration defect evolution trend prediction model, the target number of decomposition layers, and the target mutual information difference, performing variational mode decomposition on the vibration signal in the signal acquisition results to determine multiple intrinsic mode functions; based on the initial mechanical vibration defect evolution trend prediction model, and utilizing the frequency domain features corresponding to each intrinsic mode function, and the correlation between the frequency domain features and the corresponding vibration signal, filtering to determine multiple selected target intrinsic mode functions; based on the target intrinsic mode functions, obtaining the denoised signal.

[0025] It is important to understand that, regarding the noise reduction component, assuming the original vibration signal is x(t), after VMD decomposition, K eigenmode functions {u1(t), u2(t), ..., u...} are obtained. K (t)}, satisfying Where t is the time variable.

[0026] Determining the optimal K value is crucial when using VMD to decompose a signal. If K is too large, the signal may generate excessive noise or mode aliasing due to over-decomposition; if K is too small, the signal decomposition may be insufficient. Therefore, this embodiment introduces the Mutual Information (MI) method to find the optimal K value. MI is an information metric in information theory; it can be viewed as the amount of information contained in one random variable about another, or the reduction in uncertainty of one random variable due to knowledge of another. It can be used to represent the degree of interdependence between two random variables. For random variables... and The mutual information is defined as shown in equation (1).

[0027] (1).

[0028] In the formula, yes and The joint probability density function; and They are and The marginal probability density function.

[0029] Mutual information is used to measure the similarity between the reconstructed signal and the original signal, and is defined as the first... Layer and First The mutual information difference between the reconstructed signal and the original signal is calculated using equation (2): (2).

[0030] when When the value falls below a preset threshold (0.05 in this embodiment, but can be adjusted based on actual needs), the iteration stops. This is the optimal number of decomposition layers, which can avoid over-decomposition or insufficient decomposition of the signal.

[0031] VMD secondary penalty factor The modal bandwidth and signal fidelity are affected, and PSO (Particle Sorting) is used to optimize them. The particle number is set to 25. The optimization range is [50, 3000], and the maximum number of iterations is 50. Taking into account both signal reconstruction error and modal independence, the objective function is defined and the calculation formula is Equation (3).

[0032] (3).

[0033] in For norm, The mutual information penalty coefficient is set to 0.05 in this embodiment, but can be adjusted based on actual needs. For modality and Mutual information; Next, the particle velocity and position are updated, the fitness value of each particle is calculated, local optima and global optima are retained, and the optimal solution is output after the iteration is completed. Based on optimal and Perform VMD decomposition to remove noise-dominated IMFs components and retain effective signal components to complete the noise reduction of the original vibration signal.

[0034] Understandably, regarding the signal denoising process, a denoising component is used to denoise the original vibration signal of the acquired GIS equipment. First, the original vibration signal x(t) is input, and the optimal decomposition level K is determined according to the mutual information criterion: starting from K=1, the decomposition level is increased successively, and the mutual information difference σ between the reconstructed signal of the Kth level and the original signal of the (K+1)th level is calculated. When σ is less than a preset threshold, the iteration stops. K at this point is the optimal decomposition level, thus ensuring that the signal decomposition is both sufficient and avoids over-decomposition.

[0035] After determining the value of K, a particle swarm optimization algorithm is initiated to optimize the secondary penalty factor α. The particle swarm size and maximum number of iterations are set, with the fitness function comprehensively considering signal reconstruction error and modal independence as the objective. The optimal value of α is searched within a certain range to achieve the best purity and independence of each intrinsic mode function after decomposition. Subsequently, VMD decomposition is performed on the original vibration signal based on the optimal K and α, yielding K intrinsic mode functions IMF1, IMF2, ..., IMFK. After decomposition, based on the frequency domain characteristics of each IMF and its correlation with the original signal, high-frequency components dominated by noise are removed, retaining the effective components containing the main mechanical vibration characteristics. Finally, all retained effective IMF components are reconstructed to obtain the denoised vibration signal, effectively filtering out environmental noise and interference components from the original signal, significantly improving the signal-to-noise ratio, and providing a high-quality data foundation for subsequent prediction of the evolution trend of mechanical vibration defects in GIS equipment.

[0036] Furthermore, after signal denoising is completed, the obtained IMF is used to divide the data to obtain training and test sets. That is, the data is divided based on the denoised signal and the acquisition time corresponding to the denoised signal to determine the data division result; the data division result and the vibration trend prediction label corresponding to each combination are labeled to determine the training and test sets.

[0037] It is understood that in this embodiment, the training set and the test set can be divided proportionally according to the sampling time of the samples. For example, if the entire sampling lasts 5 seconds, the vibration signals collected in the first 4 seconds can be divided into the training set and the vibration signals collected in the last 1 second can be divided into the test set in a 4:1 ratio.

[0038] Step S13: Based on the target gated recurrent unit in the initial mechanical vibration defect evolution trend prediction model, perform hyperparameter optimization on the training set to determine the hyperparameter optimization result.

[0039] In this embodiment, combined with Figure 3 As shown, after signal denoising is completed and the training and test sets are determined, the prediction component in the PSO-MVMD-SSA-Transformer-GRU prediction model, namely the target-gated recurrent unit, will be used to perform hyperparameter optimization on each denoised IMF. Specifically: the target intrinsic mode functions in the training set are normalized to determine the processed functions; the parameters of the sparrow search algorithm are initialized based on the target-gated recurrent unit in the initial mechanical vibration defect evolution trend prediction model to determine the initial parameters; hyperparameter optimization is performed based on the target-gated recurrent unit, the initial parameters, the preset fitness function, and each processed function to determine the hyperparameter information corresponding to each processed function; the hyperparameter information includes the number of target hidden units, the target training period, and the target initial learning rate; and the hyperparameter optimization result is determined based on the hyperparameter information.

[0040] It's important to understand that, regarding the prediction component, the Transformer within this component employs a multi-head self-attention mechanism to capture long-range dependencies in vibration signals, supplementing temporal information through position encoding, and the output dimension is... The number of attention heads is 8; the GRU inherits the output features of the Transformer and dynamically adjusts the hidden state through the reset gate and update gate to capture the short-term temporal features of the signal. The activation function of the hidden layer is tanh.

[0041] GRU is a simplified and optimized model of LSTM (Long Short-Term Memory) network, with fewer parameters and higher computational efficiency. The structure of GRU is as follows: Figure 4 As shown. The GRU mainly consists of two gates: the reset gate and the update gate. The reset gate... It controls the degree to which the state information of the previous moment is forgotten, as shown in equation (4).

[0042] (4).

[0043] In the formula: It is the sigmoid function; It is a weight matrix; This indicates that the previous state will be hidden. Input at the current time splicing; It is a bias term. The output value of the reset gate is between 0 and 1, and the closer the value is to 0, the more information from the previous moment is forgotten.

[0044] Update Gate The degree to which state information from the previous moment flows into the current state is controlled is shown in equation (5).

[0045] (5).

[0046] In the formula, These are the weight matrix and the bias term, respectively. The update gate output value is also between 0 and 1, with the value closer to 1 indicating that more state information from the previous time step is retained.

[0047] Based on the reset and update gates, the current state is hidden. The calculation method is shown in equations (6) and (7).

[0048] (6); (7).

[0049] In the formula, It is a candidate hidden state; These are the weight matrix and the bias terms; This indicates element-wise multiplication.

[0050] The mesh framework diagram of the Transformer-GRU model in this component is as follows: Figure 5 As shown. The coefficient of determination R0 on the model test set is... 2 The goal is to maximize the number of hidden units N∈[50, 300], the maximum training period E∈[50, 300], and the initial learning rate η∈[0.001, 0.01]. SSA sets the sparrow population size to 5, the discoverer percentage to 0.7, the warning value to 0.6, and the maximum number of iterations to 5. It calculates the fitness value of each individual sparrow (hyperparameter combination), updates the positions of discoverers and followers, excludes warning individuals, and outputs the optimal hyperparameter combination after the iteration is complete.

[0051] Regarding the hyperparameter optimization process, this embodiment is based on SSA. For example, the number of hidden units, the maximum training period, and the initial learning rate are used as optimization objectives. An SSA algorithm is set with a sparrow population size of 5, a discoverer ratio of 0.7, a warning value of 0.6, and a maximum number of iterations of 5. The fitness function is to maximize the coefficient of determination on the model's test set. The fitness value of each individual sparrow (hyperparameter combination) is calculated, the positions of discoverers and followers are updated, and warning individuals are excluded. After iteration, the optimal hyperparameter combination is output. Furthermore, the Transformer uses an 8-head attention mechanism with an output dimension of 256, and the GRU uses the tanh activation function. Then, based on the obtained optimal hyperparameter combination, an SSA-optimized Transformer-GRU prediction model is constructed for each denoised IMF.

[0052] Step S14: Based on the hyperparameter optimization results and the target gated recurrent unit, train the prediction of mechanical vibration defect evolution trend for each sample in the training set to determine the prediction training results.

[0053] In this embodiment, combined with Figure 3 As shown, after completing the hyperparameter optimization, the normalized signals of each IMF are used as the model input. The training set is input into the hyperparameter-optimized Transformer-GRU prediction model to train for accurate prediction of the evolution trend of mechanical vibration defects in GIS equipment. L2 regularization is used to prevent overfitting, and the gradient threshold is set to 1 for training. That is, the hyperparameter information is used as the initial hyperparameters of the target gated recurrent unit. The signals corresponding to each processed function are used to train for prediction of the evolution trend of mechanical vibration defects to determine the initial prediction results corresponding to each processed function. The initial prediction results are then recombined to determine the prediction training results.

[0054] Step S15: Based on the initial mechanical vibration defect evolution trend prediction model, the test set, and the prediction training results, determine the target mechanical vibration defect evolution trend prediction model that has completed training.

[0055] In this embodiment, after training until the loss converges and obtaining the current prediction training result, the test set is input into the SSA-optimized Transformer-GRU prediction model. The prediction test results output by all SSA-optimized Transformer-GRU prediction models are reorganized and combined with four evaluation indicators: MAE (Mean Absolute Error), MSE (Mean Squared Error), RMSE (Root Mean Squared Error), and R² (Coefficient of Determination). This achieves accurate prediction of the evolution trend of mechanical vibration defects in GIS equipment. Specifically, the test set is input into the target gated loop unit to determine the test result; based on the test result, the prediction training result, and the preset evaluation indicators, the current evaluation result is determined; based on the current evaluation result, it is determined whether the preset model training termination condition is met, and if so, the target mechanical vibration defect evolution trend prediction model is determined to have completed training.

[0056] Step S16: Based on the target mechanical vibration defect evolution trend prediction model and the vibration signal of the GIS device to be processed, determine the target defect evolution trend prediction result corresponding to the GIS device to be processed.

[0057] In this embodiment, after training the mechanical vibration defect evolution trend prediction model, the obtained target mechanical vibration defect evolution trend prediction model can be used to execute the mechanical vibration defect evolution trend prediction service for GIS equipment. That is, based on the target mechanical vibration defect evolution trend prediction model, the vibration signal of the GIS equipment to be processed is subjected to joint variational mode decomposition signal denoising processing to determine the signal denoising result; based on the target gated loop unit and the signal denoising result, hyperparameter optimization is performed to determine the target hyperparameters; based on the target gated loop unit, the target hyperparameters, and the signal denoising result, the target defect evolution trend prediction result corresponding to the GIS equipment to be processed is determined.

[0058] Therefore, in this embodiment, firstly, based on the vibration signals of GIS equipment corresponding to different combinations of operating conditions and load current levels, then, based on the initial mechanical vibration defect evolution trend prediction model, the signal acquisition results are subjected to joint variational mode decomposition for noise reduction to determine the training set and test set; then, using the training set and the target gated recurrent unit in the initial mechanical vibration defect evolution trend prediction model, the hyperparameter optimization result is determined; then, based on the target gated recurrent unit, the hyperparameter optimization result, and each sample in the training set, the mechanical vibration defect evolution trend prediction is trained, and using the prediction training result and the test set, the trained target mechanical vibration defect evolution trend prediction model is determined; finally, based on the target mechanical vibration defect evolution trend prediction model and the vibration signal of the GIS equipment to be processed, the target defect evolution trend prediction result is determined. In this way, accurate prediction of the mechanical vibration defect evolution trend of GIS equipment can be achieved, and the adaptability to different defect types and different load conditions can be improved, thereby providing a basis for the pre-maintenance of mechanical defects in GIS equipment and ensuring the safe and stable operation of the power system.

[0059] The following refers to Figure 6(a) to... Figure 9 The schematic / timing diagrams disclosed herein provide a detailed description of the technical solutions of the embodiments of this application.

[0060] In one specific embodiment, vibration signals from a 126kV GIS device under different operating conditions (normal, loose conductor contacts) and different load currents (600A, 1500A, 2400A) are collected at a sampling rate of 125kHz, with a single sample duration of 5s. The denoised IMFs signals are normalized and divided into a training set (first 4s) and a test set (last 1s) at a 4:1 ratio. The training set is input into a hyperparameter-optimized Transformer-GRU model, using an L2 regularization parameter (0.01) to prevent overfitting, and a gradient threshold of 1. Training continues until the loss function converges. The test set is input into the model, and the vibration trend prediction results are output, using mean absolute error (MAE), mean squared error (MSE), root mean square error (RMSE), and coefficient of determination (CQD). ) Evaluate model performance.

[0061] (8); (9); (10); (11).

[0062] in, This is the actual value. For predicted values, This is the average of the actual values. This represents the number of samples.

[0063] Further, regarding the verification of the PSO-MVMD signal noise reduction effect.

[0064] Taking the conductor contact loosening defect signal under a 1500A load as an example, the initial quadratic penalty factor α of VMD is set to a range of 50, 1500, and 3000, and the mutual information difference threshold is set. The value range is 0.01, 0.5, and 0.1. The optimal values ​​for different parameter combinations are... The values ​​are shown in Table 1: Table 1 Comparison of different initial parameter selections ; With initial parameters as =50, Under the condition that K = 0.05, the final output optimal K value is the minimum value of 5 in all cases, and the initial parameters are the minimum values ​​that satisfy the optimal K value condition. Therefore, the initial parameters selected for the MVMD model are... =50, =0.05.

[0065] The optimization iteration curves of the PSO-MVMD denoising algorithm are shown in Figures 6(a) and 6(b). The results show that as the number of iterations increases, the model error gradually decreases. When the number of iterations reaches 30, all vibration signals reach their minimum values ​​and subsequently stabilize. Comparing the fitness values ​​before and after optimization, the initial average fitness value under normal conditions is 4.2673, while the average fitness value after optimization is 4.2191, a decrease of 1.13%. Under the condition of loose conductor contacts, the initial average fitness value is 1.3500, while the average fitness value after optimization is 1.1700, a decrease of 13.33%. This indicates that the PSO algorithm effectively reduces the secondary penalty factor. Optimization can reduce the error between the vibration signal after VMD decomposition and the original signal, and by adding a mutual information penalty term to fitness, the independence of modes can be improved and mode aliasing can be prevented.

[0066] Taking a load current of 2400A as an example, the final decomposition results of the PSO-MVMD denoising algorithm are shown in Figures 7(a) and 7(b). The results show that the original signal is divided into multiple IMFs according to different center frequencies, and the spectra of each mode exhibit significant separation with minimal overlap between frequency bands. The data generated after PSO-MVMD decomposition has clearer fluctuations, reducing the complexity of the original data and helping to more deeply mine the information contained in the original data, thereby improving the accuracy of subsequent predictions.

[0067] Further, the prediction performance of the PSO-MVMD-SSA-Transformer-GRU model was validated.

[0068] The predicted vibration waveforms of GIS equipment under different defect types are shown in Figures 8(a) and 8(b). The results show that the predicted waveform's variation period is consistent with the actual waveform. The predicted waveform has a high degree of overlap with the actual waveform, and the vibration amplitude of the predicted waveform is close to but slightly smaller than the actual waveform. Taking a load current of 2400A and a normal (Type 1) condition as an example, the maximum absolute error is 0.0013, and the minimum is 1.469 × 10⁻⁶. -7 When the load current is 2400A, under normal (Type 1) conditions, the maximum absolute error is 4.327 × 10⁻⁶. -4 The minimum is 2.285 × 10 -8 The predicted waveforms under different load currents show that as the load current increases, the amplitude of the vibration waveform increases, the waveform becomes more regular, the predicted signal is closer to the real signal, and the prediction effect is better.

[0069] Figure 9 The mean absolute error (MAE), mean square error (MSE), root mean square error (RMSE), and coefficient of determination (R²) are used to evaluate the model under different defect types and load currents. The results show that load current and defect type have a significant impact on the prediction results. Overall, the prediction effect for conductor contact loosening defects is the best, and the R² of the model increases with increasing load current, reaching a maximum value of 94.32% at 2400A, indicating that the prediction results fit the actual waveform well.

[0070] To verify that the addition of PSO-MVMD and SSA can improve the prediction accuracy and robustness of the model, two prediction models, Transformer-GRU and PSO-MVMD-Transformer-GRU, were constructed, trained and used on the same sample set. The prediction results were then compared and analyzed with those obtained by the PSO-MVMD-SSA-Transformer-GRU prediction model proposed in this embodiment. The four metrics, MAE, MSE, RMSE, and R², were still used. Taking a load current of 2400A and a conductor contact loosening defect as an example, the evaluation metrics of the three prediction models are shown in Table 2.

[0071] Table 2 Comparison of Evaluation Indicators for Three Prediction Models ; Table 2 shows that the PSO-MVMD-SSA-Transformer-GRU model has significant advantages over the Transformer-GRU and PSO-MVMD-Transformer-GRU models. Under the condition of a load current of 2400A and a loose conductor contact defect, the MAE index decreases by 0.0102 and 0.0099, respectively; the MSE index decreases by 0.0012 and 0.0011, respectively; the RMAE index decreases by 0.0129 and 0.0122, respectively; and the R² index increases by 4.51% and 4.22%, respectively. Therefore, the PSO-MVMD-SSA-Transformer-GRU model has the highest prediction accuracy.

[0072] In summary, this embodiment proposes a prediction scheme for the evolution trend of mechanical vibration defects in GIS equipment based on improved signal processing and machine learning. It combines MI and PSO to optimize VMD, solving the problems of modal aliasing, the reliance on empirically set decomposition layers and penalty factors in traditional VMD, thus achieving efficient noise reduction of vibration signals. SSA is used to optimize the Transformer-GRU hyperparameters, avoiding the subjectivity of manual parameter tuning and improving the model's adaptability to different defect types and load conditions. The model achieves a maximum R² of 94.32%, a significant improvement over the unoptimized model, enabling accurate prediction of the evolution trend of mechanical vibration defects in GIS equipment. This provides a basis for proactive equipment maintenance, reduces downtime losses, and ensures the safe and stable operation of the power system.

[0073] See Figure 10 As shown in the figure, this application also discloses a device for predicting the evolution trend of mechanical vibration defects in GIS equipment, comprising: Signal acquisition module 11 is used to acquire vibration signals of GIS equipment for different combinations of working conditions and load current levels, so as to determine the signal acquisition results; The signal denoising module 12 is used to perform joint variational mode decomposition on the signal acquisition results based on the initial mechanical vibration defect evolution trend prediction model, and to determine the training set and test set using the corresponding denoised signal; wherein, the joint variational mode decomposition is a variational mode decomposition optimized by joint mutual information and particle swarm algorithm. The hyperparameter determination module 13 is used to perform hyperparameter optimization on the training set based on the target gated recurrent unit in the initial mechanical vibration defect evolution trend prediction model, so as to determine the hyperparameter optimization result. The prediction training module 14 is used to train the prediction of the mechanical vibration defect evolution trend of each sample in the training set based on the hyperparameter optimization results and the target gated recurrent unit, so as to determine the prediction training results. Training completion module 15 is used to determine the target mechanical vibration defect evolution trend prediction model that has completed training based on the initial mechanical vibration defect evolution trend prediction model, the test set, and the prediction training results. The prediction result determination module 16 is used to determine the prediction result of the target defect evolution trend of the GIS device to be processed based on the target mechanical vibration defect evolution trend prediction model and the vibration signal of the GIS device to be processed.

[0074] In some specific embodiments, the signal denoising module 12 can be used to: construct an initial mechanical vibration defect evolution trend prediction model; optimize the number of decomposition layers based on the initial mechanical vibration defect evolution trend prediction model, the mutual information criterion, and the signal acquisition results to determine the target number of decomposition layers; optimize the mutual information difference between decomposition layers based on the initial mechanical vibration defect evolution trend prediction model, the target number of decomposition layers, the signal acquisition results, and the particle swarm optimization algorithm to determine the target mutual information difference; perform variational mode decomposition on the vibration signal in the signal acquisition results based on the initial mechanical vibration defect evolution trend prediction model, the target number of decomposition layers, and the target mutual information difference to determine multiple intrinsic mode functions; perform screening based on the initial mechanical vibration defect evolution trend prediction model and using the frequency domain features corresponding to each intrinsic mode function, and the correlation between the frequency domain features and the corresponding vibration signal, to determine multiple selected target intrinsic mode functions; and obtain the denoised signal based on the target intrinsic mode functions.

[0075] In some specific embodiments, the signal denoising module 12 can be used to: divide the data based on the denoised signal and the acquisition time corresponding to the denoised signal to determine the data division result; and label the data based on the data division result and the vibration trend prediction label corresponding to each combination to determine the training set and the test set.

[0076] In some specific embodiments, the hyperparameter determination module 13 can be used to: normalize each of the target intrinsic mode functions in the training set to determine the processed functions; initialize the parameters of the sparrow search algorithm based on the target gated recurrent unit in the initial mechanical vibration defect evolution trend prediction model to determine the initial parameters; perform hyperparameter optimization based on the target gated recurrent unit, the initial parameters, the preset fitness function, and each of the processed functions to determine the hyperparameter information corresponding to each of the processed functions; the hyperparameter information includes the number of target hidden units, the target training period, and the target initial learning rate; and determine the hyperparameter optimization result based on the hyperparameter information.

[0077] In some specific embodiments, the prediction training module 14 can be used to: use the hyperparameter information as the initial hyperparameter of the target gated loop unit, train the prediction of the mechanical vibration defect evolution trend of the signals corresponding to each of the processed functions, so as to determine the initial prediction result corresponding to each of the processed functions; and reorganize the initial prediction result to determine the prediction training result.

[0078] In some specific embodiments, the training completion module 15 may be used to: input the test set into the target gated loop unit to determine the test result; determine the current evaluation result based on the test result, the predicted training result and the preset evaluation index; determine whether the current preset model training termination condition is met based on the current evaluation result, and when so, determine the target mechanical vibration defect evolution trend prediction model that has completed training.

[0079] In some specific embodiments, the prediction result determination module 16 can be used to: perform joint variational mode decomposition signal denoising processing on the vibration signal of the GIS equipment to be processed based on the target mechanical vibration defect evolution trend prediction model, so as to determine the signal denoising result; perform hyperparameter optimization based on the target gated cyclic unit and the signal denoising result, so as to determine the target hyperparameter; and determine the target defect evolution trend prediction result corresponding to the GIS equipment to be processed based on the target gated cyclic unit, the target hyperparameter, and the signal denoising result.

[0080] Furthermore, embodiments of this application also disclose an electronic device, Figure 11 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application.

[0081] Figure 11 This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of this application. Specifically, the electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the method for predicting the evolution trend of mechanical vibration defects in GIS equipment disclosed in any of the foregoing embodiments. Alternatively, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0082] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0083] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0084] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the method for predicting the evolution trend of mechanical vibration defects in GIS equipment executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include computer programs capable of performing other specific tasks.

[0085] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned method for predicting the evolution trend of mechanical vibration defects in GIS equipment. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.

[0086] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0087] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0088] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0089] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0090] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for predicting the evolution trend of mechanical vibration defects in GIS equipment, characterized in that, include: For different combinations of operating conditions and load current levels, vibration signals of GIS equipment are collected to determine the signal acquisition results. Based on the initial mechanical vibration defect evolution trend prediction model, the signal acquisition results are subjected to signal denoising processing by joint variational mode decomposition, and the training set and test set are determined using the corresponding denoised signal; wherein, the joint variational mode decomposition is a variational mode decomposition optimized by joint mutual information and particle swarm algorithm. Based on the target gated recurrent unit in the initial mechanical vibration defect evolution trend prediction model, hyperparameter optimization is performed on the training set to determine the hyperparameter optimization result. Based on the hyperparameter optimization results and the target gated recurrent unit, the mechanical vibration defect evolution trend prediction is trained for each sample in the training set to determine the prediction training results. Based on the initial mechanical vibration defect evolution trend prediction model, the test set, and the prediction training results, the target mechanical vibration defect evolution trend prediction model that has completed training is determined. Based on the target mechanical vibration defect evolution trend prediction model and the vibration signal of the GIS device to be processed, the prediction result of the target defect evolution trend corresponding to the GIS device to be processed is determined.

2. The method for predicting the evolution trend of mechanical vibration defects in GIS equipment according to claim 1, characterized in that, The prediction model based on the evolution trend of initial mechanical vibration defects performs signal denoising processing on the signal acquisition results using joint variational mode decomposition, and uses the corresponding denoised signals to determine the training set and test set, including: Construct an initial model to predict the evolution trend of mechanical vibration defects; Based on the initial mechanical vibration defect evolution trend prediction model, mutual information criterion, and signal acquisition results, the decomposition layer number is optimized to determine the target decomposition layer number. Based on the initial mechanical vibration defect evolution trend prediction model, the target decomposition layer number, the signal acquisition results, and the particle swarm optimization algorithm, the mutual information difference between the decomposition layers is optimized to determine the target mutual information difference. Based on the initial mechanical vibration defect evolution trend prediction model, the target decomposition layer number, and the target mutual information difference, variational mode decomposition is performed on the vibration signal in the signal acquisition results to determine multiple intrinsic mode functions. Based on the initial mechanical vibration defect evolution trend prediction model, and using the frequency domain features corresponding to each intrinsic mode function, as well as the correlation between the frequency domain features and the corresponding vibration signal, a selection of multiple target intrinsic mode functions is made. The denoised signal is obtained based on the target intrinsic mode function.

3. The method for predicting the evolution trend of mechanical vibration defects in GIS equipment according to claim 2, characterized in that, The step of determining the training set and test set using the corresponding denoised signal includes: Based on the denoised signal and the corresponding acquisition time of the denoised signal, the data is divided to determine the data division result. Based on the data partitioning results and the vibration trend prediction labels corresponding to each combination, annotations are performed to determine the training set and the test set.

4. The method for predicting the evolution trend of mechanical vibration defects in GIS equipment according to claim 3, characterized in that, The target-gated recurrent unit in the initial mechanical vibration defect evolution trend prediction model performs hyperparameter optimization on the training set to determine the hyperparameter optimization result, including: The target intrinsic mode functions in the training set are normalized to determine the processed functions; Based on the target gated loop unit in the initial mechanical vibration defect evolution trend prediction model, the parameters of the sparrow search algorithm are initialized to determine the initial parameters; Based on the target gated recurrent unit, the initial parameters, the preset fitness function, and each of the processed functions, hyperparameter optimization is performed to determine the hyperparameter information corresponding to each of the processed functions; the hyperparameter information includes the number of target hidden units, the target training period, and the target initial learning rate. Based on the hyperparameter information, the hyperparameter optimization result is determined.

5. The method for predicting the evolution trend of mechanical vibration defects in GIS equipment according to claim 4, characterized in that, The training of mechanical vibration defect evolution trend prediction for each sample in the training set based on the hyperparameter optimization results and the target gated recurrent unit includes: The hyperparameter information is used as the initial hyperparameter of the target gated loop unit. The signals corresponding to each of the processed functions are used to train the prediction of the mechanical vibration defect evolution trend, so as to determine the initial prediction result corresponding to each of the processed functions. The initial prediction results are reorganized to determine the prediction training results.

6. The method for predicting the evolution trend of mechanical vibration defects in GIS equipment according to claim 1, characterized in that, The step of determining the target mechanical vibration defect evolution trend prediction model based on the initial mechanical vibration defect evolution trend prediction model, the test set, and the prediction training results includes: The test set is input into the target gated loop unit to determine the test results; Based on the test results, the prediction training results, and the preset evaluation indicators, the current evaluation result is determined; Based on the current evaluation results, determine whether the preset model training termination condition is met, and if so, determine the target mechanical vibration defect evolution trend prediction model that has completed training.

7. The method for predicting the evolution trend of mechanical vibration defects in GIS equipment according to claim 1, characterized in that, The step of determining the predicted evolution trend of the target mechanical vibration defect based on the prediction model of the target mechanical vibration defect and the vibration signal of the GIS device to be processed includes: Based on the prediction model of the evolution trend of the target mechanical vibration defect, the vibration signal of the GIS equipment to be processed is subjected to joint variational mode decomposition signal denoising processing to determine the signal denoising result. Based on the target gated loop unit and the signal denoising result, hyperparameter optimization is performed to determine the target hyperparameters; Based on the target gated loop unit, the target hyperparameters, and the signal denoising results, the predicted result of the target defect evolution trend corresponding to the GIS device to be processed is determined.

8. A device for predicting the evolution trend of mechanical vibration defects in GIS equipment, characterized in that, include: The signal acquisition module is used to acquire vibration signals from GIS equipment under different operating conditions and load current levels to determine the signal acquisition results. The signal denoising module is used to perform joint variational mode decomposition on the signal acquisition results based on the initial mechanical vibration defect evolution trend prediction model, and to determine the training set and test set using the corresponding denoised signal; wherein, the joint variational mode decomposition is a variational mode decomposition optimized by joint mutual information and particle swarm algorithm. The hyperparameter determination module is used to perform hyperparameter optimization on the training set based on the target gated recurrent unit in the initial mechanical vibration defect evolution trend prediction model, so as to determine the hyperparameter optimization result. The prediction training module is used to train the prediction of the mechanical vibration defect evolution trend of each sample in the training set based on the hyperparameter optimization results and the target gated recurrent unit, so as to determine the prediction training results. The training completion module is used to determine the target mechanical vibration defect evolution trend prediction model that has completed training, based on the initial mechanical vibration defect evolution trend prediction model, the test set, and the prediction training results. The prediction result determination module is used to determine the prediction result of the target defect evolution trend of the GIS device to be processed based on the target mechanical vibration defect evolution trend prediction model and the vibration signal of the GIS device to be processed.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the method for predicting the evolution trend of mechanical vibration defects in GIS equipment as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store a computer program, which, when executed by a processor, implements the method for predicting the evolution trend of mechanical vibration defects in GIS equipment as described in any one of claims 1 to 7.

Images