Method for predicting development situation of mechanical vibration state of GIS (Geographic Information System) equipment by applying autoregressive convolutional recurrent neural network

Abstract

The invention relates to a method for predicting the mechanical vibration state development situation of GIS equipment by applying an autoregressive convolutional recurrent neural network, and belongs to the technical field of mechanical vibration situation prediction, and the method comprises the following steps: S1, collecting vibration signals of the GIS equipment under typical mechanical defects by using a piezoelectric acceleration sensor, and dividing the vibration signals into a training set and a test set after preprocessing; s2, constructing a GIS equipment mechanical vibration state development situation prediction model based on an autoregressive convolutional recurrent neural network by using a multi-scale convolution module, a gating cycle unit module and a linear transformation module; s3, training a prediction model by using the training set, and optimizing network parameters of the prediction model by using vibration signals of the GIS equipment in different working states; s4, evaluating effectiveness by taking a decision coefficient of a predicted future result and actual future data as an evaluation index, and finding an optimal model; and S5, testing the prediction model by using the test set, and using the prediction model after the test is completed.

Description

Technical Field

[0001] The present invention belongs to the technical field of mechanical vibration trend prediction, and relates to a method for predicting the development trend of the mechanical vibration state of GIS equipment by applying an autoregressive convolutional recurrent neural network. Background Art

[0002] Gas insulated switchgear (GIS) equipment is widely used in the power transmission field due to its long maintenance cycle, high operation reliability, and low maintenance workload. Mechanical vibration defects are one of the key factors for sudden failures of GIS. Since GIS equipment will be subjected to varying degrees of mechanical vibration during operation, these vibrations may cause wear, loosening, or other forms of damage to equipment components over time, thereby triggering equipment failures. Therefore, accurately predicting the development trend of the mechanical vibration state of GIS equipment can detect potential fault hazards in advance, providing strong support for equipment maintenance and safety management. This not only helps improve the operation reliability of GIS equipment but also effectively enhances its active safety protection level, reduces the probability of equipment failures, extends the service life of the equipment, and ensures the stable operation of the power system. Therefore, it is of great significance to carry out the prediction of the development trend of the mechanical vibration state of GIS equipment.

[0003] To analyze the development trend, it is necessary to perform time series modeling on time series signals to capture the time, trend, and seasonal characteristics of the time series. Sun Muxin of Neusoft Group Co., Ltd. input time series data into multiple decision models respectively, obtained multiple prediction results, and fused them to obtain the final time series prediction result. Wang Jinsong et al. of Beijing University of Technology realized the prediction of the resource load of the server cluster based on the RNN neural network.

[0004] Research scholars have carried out certain research on state prediction methods in the industrial field. Hao Jian et al. of Chongqing University constructed a prediction model for the development trend of the mechanical vibration state of the target GIS equipment by integrating the attention mechanism and the bidirectional gated recurrent unit, which can accurately predict the development trend of the severity of different types of mechanical defects. Chen Lin et al. of Guangxi University established a battery aging model based on the grey prediction model and combined it with the particle filter algorithm to realize the accurate online life prediction of the battery. Huang Nan et al. of State Grid Beijing Electric Power Company realized the prediction of transformer load by calculating the trend coefficient and the regression model. None of the above methods involve combining two types of neural network models into a hybrid model to realize the prediction of the development trend of the mechanical vibration state of GIS equipment. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a method for predicting the development trend of the mechanical vibration state of GIS equipment by applying an Autoregressive Convolutional Recurrent Neural Network (ACRNN), aiming to improve the accuracy of mechanical defect early warning of GIS equipment and being applicable to the field of operation state and fault diagnosis of gas-insulated equipment. The technical advantage of the method of the present invention is that it does not rely on complex modeling of the development mechanism of mechanical defects, but directly uses vibration signal data to predict the future development trend of the vibration state of GIS equipment, supporting early warning of equipment failures.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for predicting the development trend of the mechanical vibration state of GIS equipment by applying an autoregressive convolutional recurrent neural network, comprising the following steps:

[0008] S1: Use a piezoelectric acceleration sensor to collect the vibration signals of GIS equipment under typical mechanical defects. After preprocessing, divide the signal samples of each defect type into a training set and a test set according to a certain proportion;

[0009] S2: Use a multi-scale convolution module, a gated recurrent unit module and a linear transformation module to construct a prediction model for the development trend of the mechanical vibration state of GIS equipment based on an autoregressive convolutional recurrent neural network. This model analyzes the signal development trend based on the existing GIS vibration signals and predicts the future vibration signals as the representation of future mechanical state information;

[0010] S3: Use the training set to train the prediction model, and optimize the network parameters of the prediction model by using the vibration signals of GIS equipment under different working conditions;

[0011] S4: Take the coefficient of determination between the predicted future results and the actual future data as an evaluation index to evaluate the effectiveness of the proposed prediction model for the development of GIS equipment state and find the optimal model;

[0012] S5: Use the test set to test the prediction model, and use the prediction model after the test is completed.

[0013] Further, the prediction model for the development trend of the mechanical vibration state of GIS equipment based on an autoregressive convolutional recurrent neural network described in step S2 specifically includes:

[0014] Adopt a convolution module to extract the features of time series modeling. By sliding a convolution kernel on the input data, gradually calculate the weighted sum of the local area, and the calculation formula is as follows:

[0015] X out =ΣXin *W + b (1)

[0016] Wherein, X in is the input data, X out is the output data, W is the convolutional kernel, and b is the bias coefficient;

[0017] The gated recurrent unit is used to encode and decode the time series. The encoding part compresses the input sequence into a representative vector to capture important information; the decoding part generates the output sequence, that is, the future prediction value, according to the important information. The calculation formula is as follows:

[0018]

[0019] r t is the reset gate, σ is the activation function, is the candidate hidden state, z t is the update gate, h t is the updated memory data, x t is the time series data at the corresponding moment, Wz and W r are the parameters corresponding to the update gate and the reset gate, and the subscripts t - 1 and t respectively indicate that the data belongs to the previous time step and the current time step;

[0020] Finally, a linear transformation is performed through the linear transformation unit, and added to obtain the future time series of the mechanical vibration state.

[0021] Furthermore, in step S3, the vibration signals of the GIS device under different working states in the training set part are used to set the number of iteration rounds and the learning rate, and through forward propagation and backward propagation, the network parameters of the prediction model are optimized.

[0022] Furthermore, in step S4, the coefficient of determination is selected as the effectiveness evaluation index of the prediction model for the development trend of the mechanical vibration state of the GIS device, which is used to quantify the goodness of fit of the linear regression model and measure the closeness between the predicted dependent variable value of the model and the actual observed value. The larger the coefficient of determination value, the higher the prediction accuracy. The expression of the coefficient of determination R2 is as follows:

[0023]

[0024] Wherein, y i is the true value of the time series data, is the average value of the time series data, is the predicted value given by the prediction model;

[0025] Set the coefficient of determination threshold. When the coefficient of determination is greater than the coefficient of determination threshold, the model is considered to be the optimal.

[0026] The beneficial effects of the present invention are as follows: By combining the convolutional layer with the gated recurrent unit, this method effectively extracts and encodes the characteristics of the mechanical vibration signals of GIS equipment, predicts the changing trends of the equipment status, can not only give early warnings of the mechanical defects of the equipment in advance, but also improve the operation reliability and safety of the equipment, extend its service life, and ensure the stable operation of the power system. The technical advantage of the method of the present invention is that it does not rely on complex modeling of the development mechanism of mechanical defects, but directly uses vibration signal data to predict the future development trends of the vibration status of GIS equipment, supports early warning of equipment failures, and can effectively achieve current prediction under different loads and different mechanical states.

[0027] Other advantages, objectives and features of the present invention will be described to some extent in the subsequent description, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. Brief Description of the Drawings

[0028] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be described in detail preferably with reference to the accompanying drawings, where:

[0029] Figure 1 It is a flowchart of a method for predicting the development trend of the mechanical vibration status of GIS equipment by applying an autoregressive convolutional recurrent neural network;

[0030] Figure 2 It is a schematic structural diagram of a prediction model for the development trend of the mechanical vibration status of GIS equipment based on an autoregressive convolutional recurrent neural network;

[0031] Figure 3 It is a coefficient of determination matrix under different states;

[0032] Figure 4 It is the prediction effect under different states. Detailed Embodiments

[0033] The following illustrates the embodiments of the present invention through specific specific examples. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present invention schematically, and the following embodiments and the features in the embodiments can be combined with each other without conflict.

[0034] It should be noted that the illustrations provided in the following embodiments only schematically illustrate the basic concept of the present invention. Therefore, only the components related to the present invention are shown in the drawings, rather than being drawn according to the number, shape, and size of the components in actual implementation. The type, quantity, and ratio of each component in actual implementation can be arbitrarily changed, and the component layout type may also be more complex.

[0035] In the following description, a large number of details are explored to provide a more thorough explanation of the embodiments of the present invention. However, it is obvious to those skilled in the art that the embodiments of the present invention can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present invention difficult to understand.

[0036] As Figure 1 shown, the present invention provides a method for predicting the development trend of the mechanical vibration state of GIS equipment by applying an autoregressive convolutional recurrent neural network, including the following steps:

[0037] The first step: Construction of the vibration signal dataset of the GIS equipment in different states. Use a piezoelectric acceleration sensor to collect the vibration signals of the GIS equipment under typical mechanical defect conditions. For each type of mechanical defect, first preprocess the collected vibration signals, and then divide the signal samples of each type of defect into a training set and a test set according to a ratio of 7:3. The training set is used for the learning and parameter optimization of the model, and the test set is used to evaluate the performance and generalization ability of the model, so as to ensure the accuracy and reliability of fault diagnosis.

[0038] Specifically, a section of GIS vibration signal is divided into two sections of signals, as the existing signal and the future signal. The existing GIS vibration signal is used as the input parameter, and the future GIS vibration signal is used as the output parameter to establish the vibration signal dataset of the GIS equipment in different states.

[0039] The second step: Establishment of the prediction model for the development trend of the mechanical vibration state of the GIS equipment. Use common neural network structures such as a multi-scale convolutional module, a gated recurrent unit module, and a linear transformation module to construct a prediction model for the development trend of the mechanical vibration state of the GIS equipment based on the principle of an autoregressive convolutional recurrent neural network. This model can analyze the signal development trend based on the existing GIS vibration signals and predict the future vibration signals as the representation of future mechanical state information.

[0040] To address the problem of difficult feature extraction in time series modeling, the present invention uses a convolution module for feature extraction. Convolution calculation (Convolution) is mainly used to extract features from input data. It slides a convolution kernel over the input data and gradually calculates the weighted sum of the local area, which has the advantages of being easy to capture the local spatial features of the data, reducing the amount of calculation, and alleviating the problem of model overfitting. The convolution calculation is shown in Equation (1).

[0041] Regarding the encoding and decoding problems in time series prediction, the present invention uses gated recurrent units. The encoding part compresses the input sequence into a representative vector to capture important information; the decoding part generates the output sequence based on this representation, usually the future predicted values. The encoder-decoder structure helps the model better process time series data, capture dependencies, and thus improve the prediction accuracy and generation ability. The gated recurrent unit (GRU) is a variant of the recurrent neural network (RNN). It solves problems such as the vanishing gradient that occur in RNN during long-term memory and backpropagation through a gating mechanism. Compared with LSTM, the internal structure of GRU is more concise. The expression of the gated recurrent unit is shown in Equation (2).

[0042] X out =ΣX in *W + b (1)

[0043]

[0044] In the formula, X in is the input data, X out is the output data, W is the convolution kernel, b is the bias coefficient; r t is the reset gate, σ is the activation function, is the candidate hidden state, z t is the update gate, h t is the updated memory data, x t is the time series data at the corresponding moment, Wz and W r are the parameters corresponding to the update gate and the reset gate, and the subscripts t - 1 and t respectively indicate that the data belongs to the previous time step and the current time step.

[0045] The specific structure of the GIS equipment status development prediction model based on the autoregressive convolutional recurrent neural network designed by the present invention is as Figure 2 shown.

[0046] Step 3: Optimization of the parameters of the GIS equipment mechanical vibration status development trend prediction model. The vibration signals of the GIS equipment under different working conditions are used to optimize the network parameters of the prediction model.

[0047] Divide the actual vibration data of GIS equipment into two parts: a training set and a test set. Use the vibration signals of GIS equipment under different working conditions in the training set part, set the number of iteration rounds to 300 rounds, and the learning rate to 0.001. Through forward propagation and backward propagation, optimize the network parameters of the prediction model.

[0048] Step 4: Evaluate the effectiveness of the prediction model for the development trend of the mechanical vibration state of GIS equipment. Use the coefficient of determination between the predicted future results and the actual future data as the evaluation index to evaluate the effectiveness of the proposed prediction model for the development trend of GIS equipment state.

[0049] Select the coefficient of determination as the evaluation index for the effectiveness of the prediction model for the development trend of the mechanical vibration state of GIS equipment. It is one of the common indicators for evaluating the prediction performance of a model. This method is mainly used to quantify the goodness of fit of a linear regression model and measure the closeness between the predicted dependent variable values and the actual observed values. It is widely used in the fields of statistical analysis and machine learning. When the coefficient of determination value is large, it means that the deviation between the prediction results of the model and the actual data is small, and the prediction accuracy is higher. The expression of the coefficient of determination R2 is as shown in Equation (3):

[0050]

[0051] In the formula, y i is the true value of the time series data, is the average value of the time series data, is the predicted value given by the prediction model.

[0052] In this embodiment, the total coefficient of determination is set to reach above 0.9. When the coefficient of determination of the calculated model reaches 0.9, it can be considered that the model is already optimal.

[0053] Step 5: Select the vibration signals of GIS equipment under different working conditions in the test set part, conduct prediction on the development trend of the mechanical vibration state of GIS equipment, input some signal segments as existing signals into the prediction model for the development trend of the mechanical vibration state of GIS equipment, output the predicted future data, and compare it with the actual future data to intuitively show the actual application effect of the proposed method.

[0054] This embodiment conducts experiments based on a 550kV full-scale GIS mechanical defect fault simulation experimental platform, which can withstand a voltage range of 0 - 500kV and a current range of 0 - 5000A. During the experiment, the defect of poor contact of the contact was simulated under different severity levels, with contact resistances of 150 ohms and 303 ohms respectively, currents of 1200, 1800, and 2400A were passed through, and the normal state was set as a control. According to Figure 2For the network structure shown, a model was constructed and trained. Using a data acquisition card with a sampling rate of 125000 Hz and a piezoelectric sensor, data was collected and downsampled to 1250 Hz. With the first 1 second as the input information and the subsequent 0.4 seconds as the output information, through the means of windowing and framing, under different currents and different mechanical states, a total of 14556 training sets and 2556 test sets were constructed. The effectiveness evaluation parameters for the prediction effect and the prediction test effect are as Figure 3 and Figure 4 shown. It can be seen from the results that the prediction effect is good.

[0055] In the above embodiments, the mention of "this embodiment" in the specification means that the specific features, structures or characteristics described in connection with the embodiments are included in at least some embodiments, but not necessarily all embodiments. Multiple occurrences of "this embodiment" do not necessarily all refer to the same embodiment.

[0056] In the above embodiments, although the present invention has been described in connection with specific embodiments of the present invention, many substitutions, modifications and variations of these embodiments will be apparent to those of ordinary skill in the art based on the previous description. For example, other storage structures (e.g., dynamic RAM (DRAM)) can be used in the embodiments discussed. Embodiments of the present invention are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the appended claims.

[0057] This embodiment also provides a computer-readable storage medium having a computer program stored thereon, and when the program is executed by a processor, it implements any one of the methods in this embodiment.

[0058] This embodiment also provides an electronic terminal, including: a processor and a memory;

[0059] The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory so that the terminal executes any one of the methods in this embodiment.

[0060] For the computer-readable storage medium in this embodiment, those of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiments can be completed by hardware related to a computer program. The foregoing computer program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps including the above method embodiments; and the foregoing storage medium includes: various media such as ROM, RAM, magnetic disk or optical disk that can store program codes.

[0061] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver, and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication therebetween. The memory is used to store computer programs, the communication interface is used for communication, and the processor and the transceiver are used to run the computer programs so that the electronic terminal executes each step of the above method.

[0062] In this embodiment, the memory may include a Random Access Memory (RAM) and may also include a non-volatile memory, such as at least one disk memory.

[0063] The above-mentioned processor may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0064] The present invention can be used in numerous general-purpose or special-purpose computing system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and so on.

[0065] The present invention can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present invention can also be practiced in a distributed computing environment where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the present technical solution, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting the development trend of mechanical vibration state of GIS equipment using autoregressive convolutional recurrent neural network, characterized by: The following steps are involved: S1: Use piezoelectric accelerometers to collect vibration signals of GIS equipment under typical mechanical defects. After preprocessing, divide the signal samples of each defect type into training sets and test sets in proportion; S2: Using multi-scale convolution modules, gated recurrent unit modules and linear transformation modules, a GIS equipment mechanical vibration state development trend prediction model based on autoregressive convolutional recurrent neural network is constructed. This model is based on the existing GIS vibration signal, analyzes the signal development trend, and predicts the future vibration signal as a representation of the future mechanical state information; S3: Use the training set to train the prediction model, and use the vibration signals of GIS equipment under different working conditions to optimize the prediction model network parameters; S4: Using the determination coefficient of the predicted future results and the actual future data as the evaluation index, the effectiveness of the proposed GIS equipment status development prediction model is evaluated and the optimal model is found; S5: Use the test set to test the prediction model, and use the prediction model after the test is completed.

2. The method for predicting the development trend of mechanical vibration state of GIS equipment using autoregressive convolutional recurrent neural network according to claim 1 is characterized in that: The GIS equipment mechanical vibration state development trend prediction model based on the autoregressive convolutional recurrent neural network described in step S2 specifically includes: The convolution module is used to extract the features of time series modeling. By sliding a convolution kernel on the input data, the weighted sum of the local area is gradually calculated. The calculation formula is as follows: X out =∑X in *W+b (1) Where, X in is the input data, X out is the output data, W is the convolution kernel, and b is the bias coefficient; The gated recurrent unit is used to encode and decode the time series. The encoding part compresses the input sequence into a representative vector to capture important information; the decoding part generates an output sequence based on the important information, that is, the future prediction value. The calculation formula is as follows: r t is the reset gate, σ is the activation function, is the candidate hidden state, z t is the update gate, h t is the updated memory data, x t is the time series data at the corresponding moment, Wz and W r To update the parameters corresponding to the gate and reset the gate, the t-1 and t subscripts indicate that the data belongs to the previous time step and the current time step respectively; Finally, a linear transformation is performed through a linear transformation unit, and the future time series of the mechanical vibration state is obtained by addition.

3. The method for predicting the development trend of mechanical vibration state of GIS equipment using autoregressive convolutional recurrent neural network according to claim 1 is characterized in that: In step S3, the vibration signals of the GIS equipment under different working conditions in the training set are used to set the number of iterations and the learning rate, and the network parameters of the prediction model are optimized through forward propagation and back propagation.

4. The method for predicting the development trend of mechanical vibration state of GIS equipment using autoregressive convolutional recurrent neural network according to claim 1 is characterized in that: In step S4, the determination coefficient is selected as the effectiveness evaluation index of the GIS equipment mechanical vibration state development trend prediction model, which is used to quantify the goodness of fit of the linear regression model and measure the closeness between the dependent variable value predicted by the model and the actual observed value. The larger the determination coefficient value, the higher the prediction accuracy. The determination coefficient R2 expression is as follows: In the formula, y i is the actual value of the time series data, is the average value of the time series data, The predicted value given by the prediction model; The determination coefficient threshold is set. When the determination coefficient threshold is greater than the determination coefficient threshold, the model is considered to be optimal.

Images