Power equipment parameter acquisition and intelligent monitoring method, system and device
By combining multiple neural network models and algorithms, real-time acquisition and intelligent monitoring of power equipment parameters are achieved, and the problem of insufficient accuracy of power equipment parameter monitoring in the existing technology is solved, and the reliability and safety of power equipment operation are improved. , providing more accurate failure prediction and classification capabilities.
Patent Information
- Application Number
- CN202410911037.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-09
- Publication Date
- 2024-11-01
AI Technical Summary
In the prior art, the accuracy and reliability of power equipment parameter monitoring is insufficient, making it difficult to effectively evaluate the operating status and fault diagnosis of equipment, affecting the safety and stability of the power system.
A variety of neural network models and algorithm combinations are adopted, including the Elman neural network model of denoising autoencoder-PSO, NARX neural network model, LightGBM model-BiLSTM neural network model and BiLSTM neural network-metabolic gray GM(1,1) model , combined with the BiLSTM neural network model and ANFIS neural network classifier, real-time acquisition and intelligent monitoring of power equipment parameters are carried out through triangular fuzzy numbers and intuitive fuzzy numbers to realize the classification of power equipment safety levels.
Improve the accuracy and scientificity of power equipment parameter detection, enhance the reliability and safety of power equipment operation, provide more accurate fault prediction and classification capabilities, overcome the black box characteristics of simple neural networks, and have high non-number Linear mapping capability and adaptive capability.
Smart Images

Figure CN118889673A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of power grid power equipment automation technology, and in particular to a power equipment parameter collection and intelligent monitoring method, system and device. Background Art
[0002] With the rapid development of my country's smart grid, the power system has an increasing demand for sensor technology, Internet technology, and artificial intelligence technology. In order to ensure the safety and stability of the power system, intelligent sensing technology, automation technology, and signal processing technology are used to monitor the status of power equipment, evaluate the operating status of equipment based on the power equipment operation data, and provide a decision-making basis for power equipment maintenance and fault diagnosis. Power equipment operation status data, fault data, maintenance data, etc. play an important role in the operation, management, and deployment of the power system. Therefore, the collection and storage of real-time data of power equipment has become an important part of the operation and management of the power system. However, the accuracy and reliability of monitoring power parameters in the existing technology need to be improved. Summary of the invention
[0003] Purpose of the invention: The purpose of the present invention is to provide a method, system and device for collecting and intelligently monitoring power equipment parameters, which can improve the accuracy and scientificity of collecting detected parameters, and improve the accuracy, reliability and robustness of predicting power equipment parameter detection information. Real-time data collection of power equipment is achieved by sensing power equipment information through sensor technology, and the Internet of Things technology is used as the connection between power equipment and equipment management network. The real-time data collection and intelligent monitoring system and device of power equipment are designed to realize real-time data collection and supervision of power equipment, which is of great significance to improving the reliability and safety of power equipment operation.
[0004] Technical solution: The method for collecting and intelligently monitoring electric power equipment parameters of the present invention comprises the following steps:
[0005] Collecting power parameters of power equipment and inputting them into an intelligent monitoring model of power equipment, and classifying the power equipment into different safety levels according to the output results of the intelligent monitoring model of power equipment;
[0006] The electrical parameters include temperature parameters, current parameters and vibration parameters;
[0007] The intelligent monitoring model of electric power equipment includes an electric power parameter detection module, a first model, a second model, a third model and a fourth model; the electric power parameter detection module is used to calculate the triangular fuzzy number of the electric power parameter to obtain the triangular fuzzy number of temperature, the triangular fuzzy number of current and the triangular fuzzy number of vibration;
[0008] The temperature triangular fuzzy number, the current triangular fuzzy number and the vibration triangular fuzzy number are respectively input into the first model, the second model and the third model for data prediction, the output of the first model is used as the input of the third model and the fourth model, the output of the second model is used as the input of the first model, the third model and the fourth model, and the output of the third model is used as the input of the fourth model; the fourth model is used to classify the safety level of the power equipment according to the outputs of the first model, the second model and the third model; the output of the fourth model is input into the first model, the second model, the third model and the fourth model after passing through the TDL beat delay device;
[0009] The first model, the second model and the third model are all a LightGBM model connected in series with a BiLSTM neural network model, and the fourth model is a BiLSTM neural network model connected in series with an ANFIS neural network classifier.
[0010] Further, the power parameter detection module includes a fifth model to a fourteenth model;
[0011] The power parameters are input into the fifth model and the sixth model, the input of the fifth model is used as the input of the sixth model, the seventh model and the eighth model, and the output of the sixth model is used as the input of the seventh model, the eighth model and the ninth model;
[0012] The difference between the outputs of the seventh model and the eighth model is used as the input of the ninth model, the tenth model and the eleventh model, and the output of the ninth model is used as the input of the tenth model, the eleventh model and the twelfth model;
[0013] The difference between the outputs of the tenth model and the eleventh model is used as the input of the twelfth model and the thirteenth model, and the output of the twelfth model is used as the input of the thirteenth model;
[0014] The outputs of the eighth model, the eleventh model and the thirteenth model are used as the inputs of the fourteenth model, and the output of the fourteenth model is used as the inputs of the eighth model, the eleventh model and the thirteenth model;
[0015] The output values of the eighth model, the eleventh model, and the thirteenth model are a, b, and c respectively, and the triangular fuzzy numbers constituting the power parameters are:<a,b,c> ;
[0016] The fifth model is a denoising autoencoder and an Elman neural network model connected in series;
[0017] The sixth model, the ninth model and the twelfth model are all NARX neural network models;
[0018] The seventh model, the tenth model, and the thirteenth model are all LightGBM models connected in series with the BiLSTM neural network model;
[0019] The eighth model, the eleventh model and the fourteenth model are all a series connection of the BiLSTM neural network model and the metabolic grey GM (1, 1) model.
[0020] Furthermore, the PSO algorithm is used to optimize the ANFIS neural network classifier in the fourth model, the Elman neural network model in the fifth model, and the sixth model, the ninth model and the twelfth model.
[0021] Furthermore, the fourth model outputs five parameters x, y, z, j and k to form the triangular intuitionistic fuzzy number of the power equipment state [<x,y,z> ; j; k], where x, y and z are triangular fuzzy numbers of the power equipment state<x,y,z> , j is the credibility of the triangular fuzzy number of the power equipment state, k is the unreliability of the triangular fuzzy number of the power equipment state, and the sum of j and k is not greater than 1.
[0022] Furthermore, the triangular intuitionistic fuzzy number of the power equipment state [<x,y,z> ;j;k] and the corresponding relationship between the safety level of power equipment is as follows:
[0023] The triangular intuitionistic fuzzy number [<0.7, 0.8, 0.9>; 0.9; 0.1], the safety level of power equipment is very safe;
[0024] The triangular intuitionistic fuzzy number [<0.6, 0.7, 0.8>; 0.8; 0.1], the safety level of power equipment is relatively safe;
[0025] The triangular intuitionistic fuzzy number [<0.5, 0.6, 0.7>; 0.7; 0.2], the safety level of power equipment is safe;
[0026] The triangular intuitionistic fuzzy number [<0.4, 0.5, 0.6>; 0.6; 0.2], the safety level of power equipment is unsafe;
[0027] The triangular intuitionistic fuzzy number [<0.3, 0.4, 0.5>; 0.5; 0.3], the safety level of power equipment is very unsafe.
[0028] Furthermore, the temperature parameters and vibration parameters are input into the power parameter detection module after the clustering results are obtained by the subtraction clustering classifier.
[0029] The power equipment parameter collection and intelligent monitoring system of the present invention comprises an intelligent monitoring model for power equipment, the power parameters of the power equipment are input into the intelligent monitoring model for power equipment, and the power equipment is classified into different safety levels according to the output result of the intelligent monitoring model for power equipment;
[0030] The electrical parameters include temperature parameters, current parameters and vibration parameters;
[0031] The intelligent monitoring model of electric power equipment includes an electric power parameter detection module, a first model, a second model, a third model and a fourth model; the electric power parameter detection module is used to calculate the triangular fuzzy number of the electric power parameter to obtain the triangular fuzzy number of temperature, the triangular fuzzy number of current and the triangular fuzzy number of vibration;
[0032] The temperature triangular fuzzy number, the current triangular fuzzy number and the vibration triangular fuzzy number are respectively input into the first model, the second model and the third model for data prediction, the output of the first model is used as the input of the third model and the fourth model, the output of the second model is used as the input of the first model, the third model and the fourth model, and the output of the third model is used as the input of the fourth model; the fourth model is used to classify the safety level of the power equipment according to the outputs of the first model, the second model and the third model; the output of the fourth model is input into the first model, the second model, the third model and the fourth model after passing through the TDL beat delay device;
[0033] The first model, the second model and the third model are all a LightGBM model connected in series with a BiLSTM neural network model, and the fourth model is a BiLSTM neural network model connected in series with an ANFIS neural network classifier.
[0034] The power equipment parameter collection and intelligent monitoring device described in the present invention includes a power equipment parameter collection terminal, an environment control terminal, a power equipment field monitoring terminal, a power equipment gateway node, a power equipment cloud platform server, a power equipment management terminal and a power equipment monitoring mobile phone APP;
[0035] The power equipment parameter collection terminal is used to collect power equipment parameter information, and the environment control terminal is used to control the operating environment control device of the power equipment. The power equipment parameter collection terminal, the environment control terminal, the power equipment on-site monitoring terminal, the power equipment cloud platform server, the power equipment management terminal and the power equipment monitoring mobile phone APP are used to realize two-way communication through the power equipment gateway node;
[0036] The electric power equipment on-site monitoring terminal is used to implement the electric power equipment parameter collection and intelligent monitoring method as described in any one of claims 1 to 6;
[0037] The power equipment management terminal and the power equipment monitoring mobile phone APP access the power equipment cloud platform server through the 5G network to collect and monitor the power equipment parameters.
[0038] The electronic device described in the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded into the processor, the method for collecting and intelligently monitoring electric power equipment parameters is implemented.
[0039] The computer-readable storage medium of the present invention stores a computer program, and when the computer program is executed by a processor, the method for collecting and intelligently monitoring electric power equipment parameters is implemented.
[0040] Beneficial effects: Compared with the prior art, the advantages of the present invention are:
[0041] 1. According to the nonlinear, time-varying and uncertain characteristics of the detection parameters, the power parameter detection module includes the Elman neural network model of the noise reduction encoder-PSO, the NARX neural network model 1-3 of the PSO, the LightGBM model-BiLSTM neural network model 1-3 and the BiLSTM neural network model-metabolic gray GM (1, 1) model 1-3 and other combined models to convert the detected parameters into the form of triangular fuzzy numbers, thereby improving the accuracy and scientificity of the collection of the detected parameters.
[0042] 2. The BiLSTM neural network model of the present invention can be composed of two LSTM neural network models with opposite directions. The power equipment parameter detection information input BiLSTM neural network model will always correspond to two LSTM neural network models with opposite directions, and the output is also affected by the two LSTM neural network models with opposite directions. The input of the bidirectional BiLSTM neural network model mainly depends on the input of the power equipment parameter detection signal, the output of the hidden layer state corresponding to the detection signal at the previous moment, and the output of the hidden layer state in different directions of the detection signal, which improves the accuracy, reliability and robustness of the prediction of the power equipment parameter detection information.
[0043] 3. The input of the NARX neural network model adopted in the present invention includes input and output historical feedback for a period of time. This part of the feedback input can be considered to include the historical information of the power equipment parameter status for a period of time to participate in the prediction of the power equipment parameters. For a suitable feedback time length, the prediction has a good effect. The NARX neural network model of the present invention provides an effective method for predicting power equipment parameters.
[0044] 4. The present invention uses the NARX neural network to establish a power equipment parameter prediction model. Since the dynamic recursive network of the model is established by introducing a delay module and output feedback, it introduces the delayed feedback of the input and output vectors of the power equipment parameters into the network training to form a new input vector, and has good nonlinear mapping ability. The input of the NARX neural network includes not only the original input power equipment data, but also the trained output data. The generalization ability of the NARX neural network is improved, so that it has better prediction accuracy and adaptability in the time series prediction of nonlinear power equipment parameters than the traditional static neural network.
[0045] 5. The LightGBM model adopts the Gradient-based one-side sampling (GOSS) algorithm. Its idea is to retain sample points with larger gradients when downsampling samples, and randomly sample sample points with smaller gradients in proportion, thereby reducing the number of samples and reducing computational costs; it adopts the Exclusive feature bundling (EFB) strategy. High-dimensional data often have multiple eigenvalues. Feature bundling can reduce the number of features and redundant information between eigenvalues, thereby reducing computational complexity; it adopts a leaf-wise decision tree growth strategy with depth limitation, which has higher prediction accuracy than traditional strategies, and prevents overfitting while ensuring high efficiency.
[0046] 6. BiLSTM neural network model of triangular intuitionistic fuzzy number-ANFIS neural network classifier of PSO. Since the parameters of power equipment have complex nonlinear characteristics and the parameters vary greatly under different working conditions, it is difficult to establish an accurate mathematical model. The input and output characteristics of the control parameters of power equipment can be accurately identified by using the ANFIS neural network classifier, which has good nonlinear approximation ability. The ANFIS neural network classifier has both the reasoning function of the fuzzy reasoning system and the training and learning function of the neural network. Combining the advantages of the two, it overcomes the black box characteristics of a simple neural network and has a certain degree of transparency. According to the engineering practice of power equipment safety, the power equipment status is classified into very safe, relatively safe, safe, unsafe and very unsafe levels. The BiLSTM neural network model of triangular intuitionistic fuzzy number-ANFIS neural network classifier of PSO improves the accuracy and reliability of the classification of power equipment safety levels. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1 This is a structural diagram of the power parameter detection module of the present invention;
[0048] Figure 2 This is a structural diagram of the intelligent monitoring model of electric power equipment of the present invention;
[0049] Figure 3 This is the architecture diagram of the power equipment parameter collection and intelligent monitoring system of the present invention;
[0050] Figure 4 This is a structural diagram of the electric power equipment parameter collection terminal of the present invention;
[0051] Figure 5 It is a structural diagram of the environment control terminal of the present invention;
[0052] Figure 6 This is a node structure diagram of the power equipment gateway of the present invention;
[0053] Figure 7 This is a functional diagram of the software for the on-site monitoring terminal of the power equipment of the present invention. DETAILED DESCRIPTION
[0054] The technical solution of the present invention is further described below in conjunction with the accompanying drawings. The power equipment parameter collection and intelligent monitoring method of the present invention includes the following contents.
[0055] (I) Construct power parameter detection module, such as Figure 1 As shown, the power parameter detection module includes the Elman neural network model of the noise reduction encoder-PSO, the NARX neural network model 1-3 of the PSO, the LightGBM model-BiLSTM neural network model 1-3 and the BiLSTM neural network model-metabolic gray GM (1, 1) model 1-3.
[0056] The outputs of multiple parameter sensors are respectively used as the inputs of the Elman neural network model of the denoising encoder-PSO and the corresponding inputs of the NARX neural network model 1 of the PSO; the outputs of the Elman neural network model of the denoising encoder-PSO are respectively used as the corresponding inputs of the NARX neural network model 1 of the PSO, the corresponding inputs of the LightGBM model-BiLSTM neural network model 1 and the corresponding inputs of the BiLSTM neural network model-metabolism grey GM (1, 1) model 1; the outputs of the NARX neural network model 1 of the PSO are respectively used as the corresponding inputs of the outputs of the LightGBM model-BiLSTM neural network model 1, the BiLSTM neural network model-metabolism grey GM (1, 1) model 1 and the NARX neural network model 2 of the PSO;
[0057] The time series difference and the change rate of the time series difference output by the LightGBM model-BiLSTM neural network model 1 and the BiLSTM neural network model-metabolism grey GM (1, 1) model 1 are used as the corresponding inputs of the LightGBM model-BiLSTM neural network model 2, the BiLSTM neural network model-metabolism grey GM (1, 1) model 2 and the NARX neural network model 2 of PSO respectively;
[0058] The outputs of the NARX neural network model 2 of PSO are used as the corresponding inputs of the LightGBM model-BiLSTM neural network model 2, the BiLSTM neural network model-metabolic grey GM (1, 1) model 2 and the NARX neural network model 3 of PSO;
[0059] The time series difference and the change rate of the time series difference output by the LightGBM model-BiLSTM neural network model 2 and the BiLSTM neural network model-metabolism grey GM (1, 1) model 2 are used as the corresponding inputs of the LightGBM model-BiLSTM neural network model 3 and the NARX neural network model 3 of PSO respectively;
[0060] The output of NARX neural network model 3 of PSO is used as the corresponding input of LightGBM model-BiLSTM neural network model 3;
[0061] The outputs of the LightGBM model-BiLSTM neural network model 3, the BiLSTM neural network model-metabolism gray GM (1, 1) model 2, and the BiLSTM neural network model-metabolism gray GM (1, 1) model 1 are respectively used as the corresponding inputs of the BiLSTM neural network model-metabolism gray GM (1, 1) model 3;
[0062] The outputs of the BiLSTM neural network model-metabolism gray GM (1, 1) model 3 are respectively used as the corresponding inputs of the LightGBM model-BiLSTM neural network model 3, the BiLSTM neural network model-metabolism gray GM (1, 1) model 2 and the BiLSTM neural network model-metabolism gray GM (1, 1) model 1;
[0063] The output values of iLSTM neural network model-metabolism gray GM (1, 1) model 1, BiLSTM neural network model-metabolism gray GM (1, 1) model 2 and LightGBM model-BiLSTM neural network model 3 are a, b and c respectively, which constitute the triangular fuzzy number of the detected parameter:<a,b,c> .
[0064] (II) Constructing intelligent monitoring model for power equipment, such as Figure 2 As shown, the intelligent monitoring model of power equipment includes power parameter detection module, subtractive clustering classifier, LightGBM model-BiLSTM neural network model 4-6, TDL beat delay device and triangular intuitionistic fuzzy number BiLSTM neural network model-PSO ANFIS neural network classifier;
[0065] The outputs of multiple temperature sensors are used as inputs of the subtractive clustering classifier 1, and the outputs of multiple types of temperature sensors output by the subtractive clustering classifier 1 are respectively used as inputs of corresponding power parameter detection modules, and the outputs of multiple power parameter detection modules are respectively used as corresponding inputs of the LightGBM model-BiLSTM neural network model 4;
[0066] The outputs of the LightGBM model-BiLSTM neural network model 4 are used as the corresponding inputs of the LightGBM model-BiLSTM neural network model 6 and the ANFIS neural network classifier of the BiLSTM neural network model-PSO of triangular intuitionistic fuzzy numbers;
[0067] The output of the current sensor is used as the input of the corresponding power parameter detection module, and the output of the power parameter detection module is used as the corresponding input of the LightGBM model-BiLSTM neural network model 5, and the output of the LightGBM model-BiLSTM neural network model 5 is used as the corresponding input of the LightGBM model-BiLSTM neural network model 4, the corresponding input of the LightGBM model-BiLSTM neural network model 6 and the corresponding input of the ANFIS neural network classifier of the BiLSTM neural network model-PSO of triangular intuitionistic fuzzy numbers;
[0068] The outputs of multiple vibration sensors are used as inputs of the subtractive clustering classifier 2. The outputs of different types of vibration sensors output by the subtractive clustering classifier 2 are used as inputs of corresponding power parameter detection modules. The outputs of multiple power parameter detection modules are used as corresponding inputs of the LightGBM model-BiLSTM neural network model 6. The output of the LightGBM model-BiLSTM neural network model 6 is used as the corresponding input of the ANFIS neural network classifier of the BiLSTM neural network model-PSO of triangular intuitionistic fuzzy numbers.
[0069] The output of the ANFIS neural network classifier of the BiLSTM neural network model-PSO of triangular intuitionistic fuzzy numbers is used as the input of the TDL beat delay device, and the output of the TDL beat delay device is used as the corresponding input of the LightGBM model-BiLSTM neural network model 4, the LightGBM model-BiLSTM neural network model 5, the LightGBM model-BiLSTM neural network model 6 and the ANFIS neural network classifier of the BiLSTM neural network model-PSO of triangular intuitionistic fuzzy numbers;
[0070] The five parameters x, y, z, j and k output by the BiLSTM neural network model of triangular intuitionistic fuzzy numbers-PSO ANFIS neural network classifier constitute the triangular intuitionistic fuzzy numbers of the detected power equipment state:<x,y,z> ; j; k], x, y and z constitute the triangular fuzzy number of the state of the detected power equipment:<x,y,z> , j is the credibility of the triangular fuzzy number of the detected power equipment state, k is the unreliability of the triangular fuzzy number of the detected power equipment state, and the sum of j and k is less than or equal to 1;
[0071] The BiLSTM neural network model of triangular intuitionistic fuzzy numbers-the ANFIS neural network classifier output of PSO divides power equipment into five safety levels: very safe, relatively safe, safe, unsafe and very unsafe.
[0072] The design methods of each model are as follows:
[0073] 1. Design of Elman neural network model of denoising autoencoder-PSO
[0074] The Elman neural network model of denoising autoencoder-PSO is a denoising autoencoder and an Elman neural network model of PSO (particle swarm optimization algorithm) in series. The denoising autoencoder (DAE) is a dimensionality reduction method that converts the high-dimensional data of power equipment parameters into low-dimensional data. The denoising autoencoder is a typical three-layer neural network. There is an encoding process between the hidden layer and the input layer, and a decoding process between the output layer and the hidden layer. The autoencoder is encoded by encoding the input data of the power equipment parameters, and the reconstructed input data of the power equipment parameters is obtained by decoding the output of the hidden layer. The data of the hidden layer is the dimensionality reduction data. Then define the reconstruction error function to measure the learning effect of the autoencoder. Constraints can be added based on the error function to generate various types of autoencoders. The encoder and decoder and the loss function are as follows:
[0075] Encoder: h = δ (Wx + b) (1)
[0076] Decoder:
[0077] Loss function:
[0078] The training process of DAE is similar to that of BP neural network. W and W' are weight matrices, b and b' are biases, h is the output value of the hidden layer, and x is the input vector. is the output vector, δ is the excitation function, and the Sigmoid function or tanh function is generally used. The denoising autoencoder trains the sparse autoencoder network by adding noise data to the input data of the power equipment. Due to the effect of the noise data, the data features learned by the autoencoder network are more robust. The autoencoder network is divided into an encoding process and a decoding process. The input layer to the hidden layer is the encoding process of the power equipment parameters, and the hidden layer to the output layer is the decoding process of the power equipment parameters. The goal of the autoencoder network is to use the error function to make the input and output of the power equipment parameters as close as possible, and to minimize the error function through back propagation to obtain the optimal weights and bias of the autoencoder network, so as to prepare for the establishment of a deep autoencoder network model. In the denoising autoencoder network process, some values in the original data of the detection signal of the power equipment are set to 0 by random probability to obtain noisy data. According to the encoding and decoding principle of the autoencoder network, the noisy power equipment detection signal data is used to obtain the encoded data and the decoded data. Finally, the error function is constructed by the decoded data and the original data, and the error function is minimized by back propagation to obtain the optimal network weights and bias. The original data of the power equipment detection signal is corrupted by adding noise, and then the corrupted data is input into the neural network as the input layer.
[0079] The reconstruction result of the denoising autoencoder should be similar to the original data of the power equipment detection signal. This method can eliminate the disturbance of the power equipment detection signal and obtain a stable structure. The original power equipment detection signal input data is obtained by adding noise to obtain interference input, and then input into the encoder to obtain feature expression, and then mapped to the output layer through the decoder. The Elman neural network model can be regarded as a forward neural network with local memory units and local feedback connections. In addition to the hidden layer, there is also a special association layer, which receives feedback signals from the hidden layer. Each hidden layer node has a corresponding association layer node connection. The association layer takes the hidden layer state of the previous moment together with the network input at the current moment as the input of the hidden layer, which is equivalent to state feedback. The transfer function of the hidden layer is generally a Sigmoid function, the output layer is a linear function, and the association layer is also a linear function. In order to effectively solve the approximation accuracy problem in parameter detection, the role of the association layer is enhanced. In this network, Z -1 represents a one-step delay operator. Assume that the number of input layers, output layers and hidden layers of the Elman neural network model are m, n and r respectively; w1, w2, w3 and w4 represent the connection weight matrices from the structural layer unit to the hidden layer, from the input layer to the hidden layer, from the hidden layer to the output layer, and from the structural layer to the output layer respectively.
[0080] Elman neural network of PSO This model includes three parts: Elman neural network structure determination, PSO algorithm optimization and determination of Elman neural network model. The basic steps are as follows:
[0081] Step 1-1, given M groups of input and output samples of the Elman neural network as training sets, and the original data are standardized;
[0082] Step 1-2, determine the Elman neural network structure according to the number of input and output parameters, thereby determining the length of the PSO algorithm particles;
[0083] Step 1-3, encode the indirect weights and thresholds between all neurons in the Elman neural network model structure into individuals represented by real digital numbers. If the Elman neural network model contains N optimized weights and thresholds, each individual will be represented by an N-dimensional vector composed of N weight and threshold parameters to initialize the particle swarm;
[0084] Step 1-4, take the absolute value and the sum of the prediction errors as the individual fitness value, and obtain the individual extreme value and the global extreme value according to the fitness value;
[0085] Step 1-5, determine whether the global extreme value meets the PSO end condition. If so, exit PSO optimization and go to Step 6; if not, update the speed and position of each particle and go to Step 4;
[0086] Step 1-6, decode the particles corresponding to the global extreme value, and use them as the initial weights and thresholds of the Elman neural network model;
[0087] Step 1-7, assign the optimal initial weights and thresholds obtained in Step 6 to the Elman neural network model, train and determine the network model.
[0088] 2. Design of NARX neural network model for PSO
[0089] NARX neural network model (Nonlinear Auto-Regression with External input neural network) is a dynamic feedforward neural network model. NARX neural network model is a nonlinear autoregressive network model with external input. It has a multi-step delay dynamic characteristic and connects several layers of the closed network through feedback. NARX neural network model is the most widely used dynamic neural network in nonlinear dynamic systems, and its performance is generally better than that of full regression neural network.
[0090] The NARX recurrent neural network model is mainly composed of an input layer, a hidden layer, an output layer, and input and output delays. Before application, the input and output delay orders and the number of hidden layer neurons must be determined in advance. The current output of the NARX neural network model depends not only on the past output y(tn), but also on the current input vector X(t) and the delay order of the input vector. The signal of the input power equipment parameter of the NARX neural network model is transmitted to the hidden layer through the delay layer. The hidden layer processes the input signal of the power equipment parameter and transmits it to the output layer. The output layer linearly weights the hidden layer output signal to obtain the final NARX neural network model output signal. The delay layer delays the feedback signal of the NARX neural network model and the signal of the power equipment parameter of the input layer, and then transmits it to the hidden layer.
[0091] NARX neural network has the characteristics of nonlinear mapping ability, good robustness and adaptability, and is suitable for predicting power equipment parameters. x(t) represents the input of power equipment parameters, m represents the delay order of power equipment parameter input; y(t) is the output of NARX neural network model, that is, the predicted value of power equipment parameters in the next period; n is the output delay order; s is the number of neurons in the hidden layer; thus, the output of the jth hidden unit can be obtained as:
[0092]
[0093] In the above formula, w ji is the connection weight between the ith input and the jth hidden neuron, b j is the bias value of the jth hidden neuron, and W is the connection weight of the NARX neural network model.
[0094] The design process of the NARX neural network model of PSO is as follows:
[0095] Step 2-1, initialize the PSO, including setting the group size, number of iterations, randomly giving initial particles and initial particle speed;
[0096] Step 2-2, the parameters corresponding to the particle vector determine a NARX neural network model to predict the parameters of the power equipment, and the model is used to calculate the fitness value of the individual through the autoregressive integrated moving average model, the NARX neural network model and the fitness function f(x);
[0097] Step 2-3, compare the calculated fitness function value with its own optimal value fPBest. If it is less than fPBest, replace the previous round of optimization solution with the new fitness value, and replace the previous round of particles with the new particles;
[0098] Step 2-4, compare the best fitness value of each particle with the optimal fitness value fGBest of all particles. If it is less than fGBest, replace the original global best fitness value with the best fitness value of the particle, and save the current state of the particle;
[0099] Step 2-5, determine whether the fitness value meets the requirements. If not, perform a new round of calculations, move the particles, generate new particles, and return to step B. If the fitness value meets the requirements, the calculation ends.
[0100] 3. LightGBM model-BiLSTM neural network model design
[0101] The LightGBM model-BiLSTM neural network model is a series connection of the LightGBM model and the BiLSTM neural network model. The LightGBM model is an improved model based on the gradient boosting tree (grandient boosting decision tree, GBDT), which adopts a gradient boosting strategy. The boosting strategy can not only integrate the classification information obtained by multiple single classifiers to improve the classification accuracy, but also promote weak learners to strong learners. Light GBM is improved on the basis of GBDT. One of the characteristics of its algorithm that is different from other tree boosting algorithms is the use of a leaf-wise growth strategy. The model mainly uses iterative training of decision trees to obtain the optimal model. The calculation formula of the LightGBM model is as follows:
[0102]
[0103] Among them: F g (X) is the prediction model formed by multiple decision trees; X is the independent variable; f k (X) is the Kth decision tree. In order to make the prediction result close to the real result, the model seeks the minimum value of the objective function in each iteration. The objective function includes two parts: the loss function and the regularization term. The calculation formula of the loss function is:
[0104]
[0105] L(y q ,y' q ) is the loss function, y q is the identification type of the sample, A is the input sample size, P q is the sample recognition probability. The regularization term reduces the overfitting of the LightGBM model by controlling the splitting of leaf nodes. The objective function of the model is:
[0106]
[0107] O is the objective function, Z is the number of leaf nodes, W v is the output value of the vth leaf node, and γ and λ are the set parameters. The LightGBM model is an improved algorithm based on the gradient boosting decision tree (GBDT), which uses the boosting idea, a series of classification and regression trees (CART) models as weak learners, and uses an arbitrary loss function iteration method to generate a strong learner. LightGBM has the advantages of high training efficiency, low memory usage, high accuracy and support for parallel learning. It adopts the Gradient-based one-side sampling (GOSS) algorithm. The idea is to retain sample points with larger gradients when downsampling samples, and randomly sample sample points with smaller gradients in proportion, thereby reducing the number of samples and reducing computational costs; it adopts the Exclusive feature bundling (EFB) strategy. High-dimensional data often have multiple eigenvalues. Feature bundling can reduce the number of features and redundant information between eigenvalues, thereby reducing computational complexity; it adopts a leaf-wise decision tree growth strategy with depth restriction, which has higher prediction accuracy than traditional strategies, and prevents overfitting while ensuring high efficiency.
[0108] The BiLSTM neural network model of the loss function triangular intuitionistic fuzzy number is a bidirectional long short-term memory recurrent neural network. The LSTM neural network model changes the weight of the self-loop by adding input gates, forget gates and output gates, which makes up for the defects of gradient vanishing and gradient explosion in the traditional recurrent neural network model. The BiLSTM neural network model consists of two LSTM neural network models with the same structure and opposite directions. After the input power equipment data is processed by the LSTM neural network models in the positive and reverse directions, the output is the splicing of the processing results of the LSTM neural network models in the two directions. The output of the BiLSTM neural network model contains the complete past and future information of the input power equipment data, and the weights are shared during the training process. While not increasing the amount of data power equipment parameters, the generalization ability of the BiLSTM neural network model is improved. The LSTM neural network model introduces the mechanism of memory cell and hidden layer state to control the information transmission of power equipment parameters between hidden layers. There are three gates (Gates) calculation structures in the memory cell of an LSTM neural network model, namely input gate, forget gate and output gate. Among them, the input gate can control the addition or filtering of new power equipment parameter information; the forget gate can forget the information of power equipment parameters that need to be discarded and retain the useful information in the past; the output gate can make the memory unit only output the information of power equipment parameters related to the current time step. These three gate structures perform operations such as matrix multiplication and nonlinear summation in the memory unit, so that the memory power equipment parameters will not decay in continuous iterations. The long short-term memory unit (LSTM) structural unit consists of a cell, an input gate, an output gate, and a forget gate. The LSTM neural network model is a model that can maintain short-term memory for a long time and is suitable for predicting the changes in power equipment parameter disturbances in time series. The BiLSTM neural network model effectively prevents the gradient from disappearing during RNN training. The long short-term memory (LSTM) network is a special RNN. The LSTM neural network model can learn the long-term dependency information of power equipment parameters while avoiding the gradient disappearance problem. LSTM adds a structure called memory cell to the neural nodes of the hidden layer of the RNN in the internal structure of the neuron to memorize the dynamic change information of the power equipment parameters in the past, and adds three gate (Input, Forget, Output) structures to control the use of historical information of the input power equipment parameters. The BiLSTM neural network model takes into account the time series and nonlinearity of the time series values of the input power equipment parameters, and has a higher prediction accuracy of the time series values of the power equipment parameters.
[0109] 4. BiLSTM neural network model--metabolism grey GM (1, 1) model design
[0110] BiLSTM neural network model--metabolism gray GM (1, 1) model is a BiLSTM neural network model and a metabolism gray GM (1, 1) model connected in series. The present invention adopts the metabolism gray GM (1, 1) model to predict the time span of power equipment parameters. The metabolism gray GM (1, 1) model can predict the power equipment parameter value at a future moment according to the predicted value of the power equipment parameter. After the power equipment parameters are predicted by the above method, the power equipment predicted value is added to the original sequence of the power equipment parameters, and a data model is removed at the beginning of the sequence accordingly, and then the prediction of the power equipment parameters is predicted. And so on, the power equipment parameters are predicted. This method is called the metabolism gray GM (1, 1) model, which can realize the prediction of power equipment parameters for a long time. The changing trend of the power equipment parameters can be grasped more accurately, and the power equipment status safety preparation can be effectively made. The metabolic grey GM (1, 1) model has many advantages over the traditional statistical prediction method. It does not need to determine whether the predicted variable obeys the normal distribution, does not require a large number of sample statistics of power equipment parameters, and does not need to change the prediction model at any time according to the changes in the input variables of power equipment parameters. Through the generation technology of cumulative power equipment parameters, a differential equation model of power equipment parameters is established. The prediction results are obtained after the original values of the accumulated power equipment parameters are restored. The differential equation model has a higher prediction accuracy of power equipment parameters. The essence of establishing the metabolic grey GM (1, 1) model is to perform a cumulative generation of the original data of the power equipment parameters so that the generated series presents a certain regularity. By establishing the differential equation model, the fitting curve is obtained to predict the power equipment parameters.
[0111] 5. Subtractive clustering classifier design
[0112] Compared with other clustering methods, the subtractive clustering classifier of power equipment parameters does not need to determine the number of clusters in advance. It can quickly determine the location of the cluster center and the number of clusters based on the sample data density of the power equipment parameters. It takes each sample data point of the power equipment parameter as a potential cluster center, so that the clustering result is independent of the dimension of the problem. Therefore, the subtractive clustering algorithm of power equipment parameters is a rule-based automatic extraction method suitable for data modeling. Set N data points (X1, X2, ... X N ), each power equipment parameter historical data point is a candidate for the cluster center, i = 1, 2, ..., N, data point X i The density function of is defined as:
[0113]
[0114] Where, radius r a is a positive number, r a Defines an influential neighborhood of the point. Data points outside the radius contribute very little to the density index of the point and are generally negligible. Calculate each point X i The density value with the highest density index D is selected c1 The data point is taken as the first cluster center X c1 ; Then correct the density value to eliminate the influence of the previous cluster center. Correct the density value according to the following formula:
[0115]
[0116] Among them, D c1 is the highest density value corresponding to the initial cluster center, and the corrected radius r b The purpose of setting is to avoid the second cluster center being too close to the previous center. After correcting the density index of each data point, when D ck With D c1 When the following formula is satisfied, the cluster center corresponding to the density index is the Kth cluster center. Repeat this process until the new cluster center X ck The corresponding density index D ck With D c1 Clustering is terminated when the following equation is satisfied:
[0117] D ck / D c1 <δ (10)
[0118] In the formula, δ is a threshold value set in advance according to the actual situation. The basic idea of the subtractive clustering of power equipment parameters in the present invention is as follows: if the distance from a point to the center of a group is less than the cluster radius r a , then the point belongs to this group; when new power equipment parameter data is obtained, the group and the center of the group are changed accordingly. As the input power equipment parameter space data continues to increase, the algorithm of the present invention obtains a better power equipment parameter input space division by dynamically adjusting the cluster center and the number of clusters in real time. The steps are as follows:
[0119] Step 5-1. Data normalization processing, input the clustering radius r of each dimension of power equipment parameters a And parameter settings such as threshold δ;
[0120] Step 5-2: Perform subtractive clustering on the historical training power equipment parameter data set to obtain c cluster centers and store v i (i=1,2,…,c) and its corresponding density value D(v i );
[0121] Step 5-3: When the kth data in the newly added online power equipment data set arrives, calculate x k (k=1,2,…,M) to i cluster centers v i The distance d ki =||x k -v i ||, if d ki >r a , go to Step5-4; if d ki ≤r a , go to Step5-5;
[0122] Step 5-4, calculate x by formula (10) k The density value D(x k ), and D(x k )>ε, then the data x k If it does not belong to any existing cluster, a new cluster is created, the number of clusters in the input space c=c+1, and return to Step 5-3;
[0123] Step 5-5, determine the data point x according to the minimum distance criterion k Belongs to the nearest cluster subset, further compare the new data x k The density value of the cluster center is the same as the density value of the cluster center. If D(x k )>D(v i ), then the data x k is closer to its nearest cluster center, x k Replace the original cluster center as the new cluster center of the subset; if D(x k )≤D(v i ), the clustering result remains unchanged, and it is determined whether the new power equipment data group is completed. If it is completed, go to Step 5-6; otherwise, return to Step 5-3;
[0124] Step 5-6, calculate the cluster center v i With v j The distance between them, if min||v i -v j ||≤(0.5-0.7)r a , and D(v i )>D(v j ), then the cluster subset v i With v j can be merged into a cluster with the cluster center v i ; otherwise, keep the clustering results unchanged.
[0125] 6. BiLSTM neural network model of triangular intuitionistic fuzzy numbers-ANFIS neural network classifier design based on PSO
[0126] The BiLSTM neural network model of triangular intuitionistic fuzzy numbers - the ANFIS neural network classifier of PSO is a BiLSTM neural network model connected in series with the ANFIS neural network classifier. The output of the BiLSTM neural network model is used as the input of the ANFIS neural network classifier. The ANFIS neural network classifier uses a fuzzy reasoning system based on an adaptive network with the same function as a fuzzy reasoning system. The first layer fuzzifies the input variables; the second layer implements rule operations and outputs the applicability of the rules. The rule operation adopts multiplication; the third layer normalizes the applicability of each rule;
[0127] The transfer function of each node in the 4th layer is a linear function; the 5th layer is used to calculate the output of the ANFIS neural network classifier; the ANFIS neural network classifier is trained by combining the least squares and gradient descent algorithms, and the parameters are adjusted by the linear least squares estimation algorithm. In each iteration, the input signal is forwarded along the ANFIS neural network classifier until the 4th layer, and the parameters are adjusted by the least squares estimation algorithm; the signal continues to be forwarded along the ANFIS neural network classifier until the output layer. The ANFIS neural network classifier implements: (1) normalizing the input variables for preprocessing, and the membership functions of the input variables are uniformly Gauss functions; (2) according to the characteristics of the input variables, the previous power equipment parameters are divided into 5 fuzzy levels; (3) the ANFIS neural network classifier is used to classify the power equipment status. The optimization process of the ANFIS neural network classifier of PSO refers to the optimization method of the NARX neural network model of PSO of the present invention.
[0128] The five parameters x, y, z, j and k output by the BiLSTM neural network model of triangular intuitionistic fuzzy numbers-PSO ANFIS neural network classifier constitute the triangular intuitionistic fuzzy numbers of the detected power equipment state:<x,y,z> ; j; k], x, y and z constitute the triangular fuzzy number of the state of the detected power equipment:<x,y,z> , j is the credibility of the triangular fuzzy number of the detected power equipment state, k is the unreliability of the triangular fuzzy number of the detected power equipment state, and the sum of j and k is less than or equal to 1; According to the engineering practice of detecting the safety of power equipment, the corresponding relationship between the 5 triangular intuitionistic fuzzy numbers output by the BiLSTM neural network model of triangular intuitionistic fuzzy numbers-PSO ANFIS neural network classifier and the safety of power equipment is shown in Table 1. The 5 levels of power equipment safety are very safe, relatively safe, safe, unsafe and very unsafe.
[0129] Table 1 Correspondence between power equipment safety and triangular intuitionistic fuzzy numbers
[0130] Serial number Safety level of power equipment Triangular intuitionistic fuzzy numbers 1 Very safe [<0.7,0.8,0.9>;0.9;0.1] 2 Safer [<0.6,0.7,0.8>;0.8;0.1] 3 Safety [<0.5,0.6,0.7>;0.7;0.2] 4 Unsafe [<0.4,0.5,0.6>;0.6;0.2] 5 Very unsafe [<0.3,0.4,0.5>;0.5;0.3]
[0131] The power equipment parameter collection and intelligent monitoring system of the present invention comprises an intelligent monitoring model for power equipment, the power parameters of the power equipment are input into the intelligent monitoring model for power equipment, and the power equipment is classified into different safety levels according to the output result of the intelligent monitoring model for power equipment;
[0132] The electrical parameters include temperature parameters, current parameters and vibration parameters;
[0133] The intelligent monitoring model of electric power equipment includes an electric power parameter detection module, a first model, a second model, a third model and a fourth model; the electric power parameter detection module is used to calculate the triangular fuzzy number of the electric power parameter to obtain the triangular fuzzy number of temperature, the triangular fuzzy number of current and the triangular fuzzy number of vibration;
[0134] The temperature triangular fuzzy number, the current triangular fuzzy number and the vibration triangular fuzzy number are respectively input into the first model, the second model and the third model for data prediction, the output of the first model is used as the input of the third model and the fourth model, the output of the second model is used as the input of the first model, the third model and the fourth model, and the output of the third model is used as the input of the fourth model; the fourth model is used to classify the safety level of the power equipment according to the outputs of the first model, the second model and the third model; the output of the fourth model is input into the first model, the second model, the third model and the fourth model after passing through the TDL beat delay device;
[0135] The first model, the second model and the third model are all a LightGBM model connected in series with a BiLSTM neural network model, and the fourth model is a BiLSTM neural network model connected in series with an ANFIS neural network classifier.
[0136] The power equipment parameter collection and intelligent monitoring device described in the present invention is as follows: Figure 3 As shown, it includes power equipment parameter collection terminal, environment control terminal, power equipment field monitoring terminal, power equipment gateway node, power equipment cloud platform server, power equipment management terminal and power equipment monitoring mobile phone APP;
[0137] The power equipment parameter collection terminal is used to collect power equipment parameter information, and the environment control terminal is used to control the operating environment control device of the power equipment. The power equipment parameter collection terminal, the environment control terminal, the power equipment on-site monitoring terminal, the power equipment cloud platform server, the power equipment management terminal and the power equipment monitoring mobile phone APP are used to realize two-way communication through the power equipment gateway node;
[0138] The electric power equipment on-site monitoring terminal is used to implement the electric power equipment parameter collection and intelligent monitoring method described in the present invention;
[0139] The power equipment management terminal and the power equipment monitoring mobile phone APP access the power equipment cloud platform server through the 5G network to collect and monitor the power equipment parameters.
[0140] 1. Design of power equipment parameter collection terminal
[0141] The electric power equipment parameter acquisition terminal based on the 4G communication module is used as an electric power equipment parameter perception terminal. The electric power equipment parameter acquisition terminal realizes information interaction with the electric power equipment gateway node through the 4G communication network. The electric power equipment parameter acquisition terminal includes sensors for collecting the voltage, current, temperature, deformation, sound, micro-meteorology, vibration and video images of the electric power equipment and corresponding signal conditioning circuits, STM32 microprocessors and 4G communication modules; the software of the electric power equipment parameter acquisition terminal mainly realizes the acquisition and preprocessing of the 4G communication module and electric power equipment parameters. The software is designed in C language with high compatibility, which greatly improves the work efficiency of software design and development and enhances the reliability, readability and portability of the program code. The structure of the electric power equipment parameter acquisition terminal is shown in Figure 4 .
[0142] 2. Environmental control terminal design
[0143] The environment control terminal realizes information interaction with the power equipment gateway node through the 4G communication module. The environment control terminal includes 4 digital-to-analog conversion circuits corresponding to the control of external devices, an STM32 microprocessor, a 4G communication module and 4 external device controllers; the 4 external device controllers are respectively a temperature control device, an illumination control device, a humidity control device, and a wind speed control device to adjust the environmental parameters of the power equipment; the environment control terminal see Figure 5 .
[0144] 3. Design of power equipment gateway node
[0145] The power equipment gateway node includes a 4G communication module, an NB-IoT module, an STM32 microcontroller and a USB interface. The power equipment gateway node includes a 4G communication module to realize communication with the power equipment parameter acquisition terminal and the environmental control terminal, an NB-IoT module to realize two-way data interaction between the power equipment gateway node and the power equipment cloud platform server, and a USB interface to connect the power equipment on-site monitoring terminal to realize information interaction between the power equipment gateway node and the power equipment on-site monitoring terminal. Figure 6 .
[0146] 4. Software design for on-site monitoring of power equipment
[0147] The power equipment field monitoring terminal is an industrial control computer. It mainly collects and manages the power equipment parameters and realizes information exchange with the power equipment gateway node. The main functions of the power equipment field monitoring terminal are power equipment communication parameter setting, power equipment data analysis and power equipment intelligent monitoring model. The management software uses Microsoft Visual++6.0 as the development tool. The functions of the power equipment field monitoring terminal software are shown in Figure 7 .
[0148] The electronic device described in the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded into the processor, the method for collecting and intelligently monitoring electric power equipment parameters is implemented.
[0149] The computer-readable storage medium of the present invention stores a computer program, and when the computer program is executed by a processor, the method for collecting and intelligently monitoring electric power equipment parameters is implemented.
[0150] The computer-readable storage media may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, flash memory, or any other medium that can be used to store the desired program code in the form of instructions or data structures and that can be accessed by a computer.
[0151] The processor is used to execute the computer program stored in the memory to implement each step of the method involved in the above embodiment.
Claims
1. A method for collecting and intelligently monitoring power equipment parameters, characterized in that: The steps include: Collecting power parameters of power equipment and inputting them into an intelligent monitoring model of power equipment, and classifying the power equipment into different safety levels according to the output results of the intelligent monitoring model of power equipment; The electrical parameters include temperature parameters, current parameters and vibration parameters; The intelligent monitoring model of electric power equipment includes an electric power parameter detection module, a first model, a second model, a third model and a fourth model; the electric power parameter detection module is used to calculate the triangular fuzzy number of the electric power parameter to obtain the triangular fuzzy number of temperature, the triangular fuzzy number of current and the triangular fuzzy number of vibration; The temperature triangular fuzzy number, the current triangular fuzzy number and the vibration triangular fuzzy number are respectively input into the first model, the second model and the third model for data prediction, the output of the first model is used as the input of the third model and the fourth model, the output of the second model is used as the input of the first model, the third model and the fourth model, and the output of the third model is used as the input of the fourth model; the fourth model is used to classify the safety level of the power equipment according to the outputs of the first model, the second model and the third model; the output of the fourth model is input into the first model, the second model, the third model and the fourth model after passing through the TDL beat delay device; The first model, the second model and the third model are all a LightGBM model connected in series with a BiLSTM neural network model, and the fourth model is a BiLSTM neural network model connected in series with an ANFIS neural network classifier.
2. The method for collecting and intelligently monitoring electric power equipment parameters according to claim 1, characterized in that: The power parameter detection module includes a fifth model to a fourteenth model; The power parameters are input into the fifth model and the sixth model, the input of the fifth model is used as the input of the sixth model, the seventh model and the eighth model, and the output of the sixth model is used as the input of the seventh model, the eighth model and the ninth model; The difference between the outputs of the seventh model and the eighth model is used as the input of the ninth model, the tenth model and the eleventh model, and the output of the ninth model is used as the input of the tenth model, the eleventh model and the twelfth model; The difference between the outputs of the tenth model and the eleventh model is used as the input of the twelfth model and the thirteenth model, and the output of the twelfth model is used as the input of the thirteenth model; The outputs of the eighth model, the eleventh model and the thirteenth model are used as the inputs of the fourteenth model, and the output of the fourteenth model is used as the inputs of the eighth model, the eleventh model and the thirteenth model; The output values of the eighth model, the eleventh model, and the thirteenth model are a, b, and c respectively, and the triangular fuzzy numbers constituting the power parameters are:<a,b,c> ; The fifth model is a denoising autoencoder and an Elman neural network model connected in series; The sixth model, the ninth model and the twelfth model are all NARX neural network models; The seventh model, the tenth model, and the thirteenth model are all LightGBM models connected in series with the BiLSTM neural network model; The eighth model, the eleventh model and the fourteenth model are all a series connection of the BiLSTM neural network model and the metabolic grey GM (1, 1) model.
3. The method for collecting and intelligently monitoring electric power equipment parameters according to claim 2 is characterized in that: The PSO algorithm is used to optimize the ANFIS neural network classifier in the fourth model, the Elman neural network model in the fifth model, and the sixth model, the ninth model and the twelfth model.
4. The method for collecting and intelligently monitoring electric power equipment parameters according to claim 1, characterized in that: The fourth model outputs five parameters x, y, z, j and k to form the triangular intuitionistic fuzzy number of the power equipment state [<x,y,z> ; j; k], where x, y and z are triangular fuzzy numbers of the power equipment state<x,y,z> , j is the credibility of the triangular fuzzy number of the power equipment state, k is the unreliability of the triangular fuzzy number of the power equipment state, and the sum of j and k is not greater than 1.
5. The method for collecting and intelligently monitoring electric power equipment parameters according to claim 4, characterized in that: The triangular intuitionistic fuzzy number of the power equipment state [<x,y,z> ;j;k] and the corresponding relationship between the safety level of power equipment is as follows: The triangular intuitionistic fuzzy number [<0.7, 0.8, 0.9>; 0.9; 0.1], the safety level of power equipment is very safe; The triangular intuitionistic fuzzy number [<0.6, 0.7, 0.8>; 0.8; 0.1], the safety level of power equipment is relatively safe; Triangular intuitionistic fuzzy number [<0.5, 0.6, 0.7>; 0.7; 0.2], the safety level of power equipment is safe; The triangular intuitionistic fuzzy number [<0.4, 0.5, 0.6>; 0.6; 0.2], the safety level of power equipment is unsafe; The triangular intuitionistic fuzzy number [<0.3, 0.4, 0.5>; 0.5; 0.3], the safety level of power equipment is very unsafe.
6. The method for collecting and intelligently monitoring electric power equipment parameters according to claim 1, characterized in that: The temperature parameters and vibration parameters are input into the power parameter detection module after the clustering results are obtained by the subtraction clustering classifier.
7. A power equipment parameter collection and intelligent monitoring system, characterized in that: It includes an intelligent monitoring model for electric power equipment, the electric power parameters of the electric power equipment are input into the intelligent monitoring model for electric power equipment, and the electric power equipment is classified into different safety levels according to the output results of the intelligent monitoring model for electric power equipment; The electrical parameters include temperature parameters, current parameters and vibration parameters; The intelligent monitoring model of electric power equipment includes an electric power parameter detection module, a first model, a second model, a third model and a fourth model; the electric power parameter detection module is used to calculate the triangular fuzzy number of the electric power parameter to obtain the triangular fuzzy number of temperature, the triangular fuzzy number of current and the triangular fuzzy number of vibration; The temperature triangular fuzzy number, the current triangular fuzzy number and the vibration triangular fuzzy number are respectively input into the first model, the second model and the third model for data prediction, the output of the first model is used as the input of the third model and the fourth model, the output of the second model is used as the input of the first model, the third model and the fourth model, and the output of the third model is used as the input of the fourth model; the fourth model is used to classify the safety level of the power equipment according to the outputs of the first model, the second model and the third model; the output of the fourth model is input into the first model, the second model, the third model and the fourth model after passing through the TDL beat delay device; The first model, the second model and the third model are all a LightGBM model connected in series with a BiLSTM neural network model, and the fourth model is a BiLSTM neural network model connected in series with an ANFIS neural network classifier.
8. A power equipment parameter collection and intelligent monitoring device, characterized in that: Including power equipment parameter collection terminal, environmental control terminal, power equipment on-site monitoring terminal, power equipment gateway node, power equipment cloud platform server, power equipment management terminal and power equipment monitoring mobile APP; The power equipment parameter collection terminal is used to collect power equipment parameter information, and the environment control terminal is used to control the operating environment control device of the power equipment. The power equipment parameter collection terminal, the environment control terminal, the power equipment on-site monitoring terminal, the power equipment cloud platform server, the power equipment management terminal and the power equipment monitoring mobile phone APP are used to realize two-way communication through the power equipment gateway node; The power equipment on-site monitoring terminal is used to implement the power equipment parameter collection and intelligent monitoring method as described in any one of claims 1 to 6; The power equipment management terminal and the power equipment monitoring mobile phone APP access the power equipment cloud platform server through the 5G network to collect and monitor the power equipment parameters.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is loaded into the processor, the method for collecting and intelligently monitoring electric power equipment parameters according to any one of claims 1 to 6 is implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by the processor, the method for collecting and intelligently monitoring electric power equipment parameters according to any one of claims 1 to 6 is implemented.
