Equipment state prediction and risk assessment method based on artificial intelligence

Abstract

The invention provides an equipment state prediction method and device based on artificial intelligence, and the method comprises the steps: obtaining an operation log of new energy equipment, and carrying out the feature extraction and marking of the denoised equipment operation signal data through a principal component analysis model; expanding the characteristic value data with the label based on the generative adversarial network, and checking the expanded data by using a non-parameter checking method to obtain training data; training and utilizing a long-short-term memory network LS < TM > based on an attention mechanism to establish a new energy equipment operation state intelligent prediction model, and obtaining a trained operation state intelligent prediction model; and finally, determining a model output value, realizing accurate prediction of the state of the electrical equipment of the transformer substation, and performing risk assessment. The number of evaluation indexes is reduced by using a principal component method, and the accuracy and efficiency of health degree evaluation of the device are improved by using an attention mechanism-based long-short-term memory network LS < TM > model; the generalization ability and the prediction performance of the model are enhanced; and the stability and reliability of the model are improved.

Description

Technical Field

[0001] The present application relates to the field of power maintenance technology, and in particular to an artificial intelligence-based equipment status prediction and risk assessment method. Background Art

[0002] With the rapid development of the power industry, the operating environment of power facilities has become increasingly complex, and equipment failures and defects have posed a severe challenge to the stability and safety of power facilities. In particular, key facilities such as high-voltage transmission lines, substations, and power generation equipment, once they fail, will not only affect the stable operation of the power system, but may also cause widespread power outages, and even cause casualties and property losses. Therefore, how to promptly detect potential failures and defects during equipment operation has become an urgent problem to be solved in the power industry.

[0003] At present, traditional power facility inspection methods often follow a fixed schedule or are based on the operator's experience, relying mostly on manual inspections and regular inspections. This method is not only labor-intensive, but also limited by the frequency and efficiency of manual inspections, making it difficult to detect minor defects in a timely manner. At the same time, the problem with this method is that it cannot accurately reflect the actual use and wear of the equipment, which may lead to premature or late maintenance, thereby increasing unnecessary maintenance costs or the risk of unexpected failures. In addition, the disadvantage of traditional methods is the lack of continuous monitoring and trend prediction of equipment operating status, making it difficult to cope with the increasingly complex management needs of power facilities.

[0004] With the development of industrial Internet and big data technology, data-based equipment health assessment has gradually become possible. By collecting and analyzing the large amount of data generated by the device during operation, its health status can be more accurately assessed, predictive maintenance can be achieved, resource allocation can be optimized, unexpected downtime can be reduced, and overall production efficiency can be improved.

[0005] However, the existing methods for predicting the health status of power facilities have technical problems such as insufficient feature extraction of operating data, incomplete and inefficient data features, and easy loss of some important information, which affects the accuracy and reliability of the prediction of the health status of power facilities; there are also technical problems such as inaccurate prediction of the health status of power facilities, leading to improper inspection management and increased safety hazards.

[0006] Therefore, it is particularly important to invent an artificial intelligence-based equipment status prediction and risk assessment method to accurately predict the equipment status. Summary of the invention

[0007] In view of this, it is necessary to provide an artificial intelligence-based equipment status prediction and risk assessment method that can overcome at least one of the above defects.

[0008] The present invention provides an equipment state prediction method and device based on artificial intelligence, including: obtaining the operation log of new energy equipment, using the principal component analysis model to extract and annotate the denoised equipment operation signal data; expanding the labeled feature value data based on the generative adversarial network, and using the non-parametric verification method to verify the expanded data to obtain training data; training to establish an intelligent prediction model for the operation state of new energy equipment using the long short-term memory network LSTM based on the attention mechanism, and obtaining a trained intelligent prediction model for the operation state; finally determining the model output value, realizing accurate prediction of the state of substation electrical equipment, and performing risk assessment. The principal component method is used to reduce the number of evaluation indicators by dimensionality reduction, and the long short-term memory network LSTM model based on the attention mechanism is used to improve the accuracy and efficiency of the device health assessment; the generalization ability and prediction performance of the model are enhanced; and the stability and reliability of the model are improved.

[0009] In a first aspect, an embodiment of the present application provides a device state prediction and risk assessment method based on artificial intelligence, the method comprising:

[0010] The operation log of the new energy equipment is obtained, the operation log is preprocessed, and the feature extraction and labeling of the denoised equipment operation signal data are performed using the principal component analysis model to obtain the labeled feature value data representing the equipment operation status and health status.

[0011] The labeled feature value data is expanded based on the generative adversarial network, the expanded data is tested using a non-parametric verification method, and the expanded data and the original data are selected as training data.

[0012] An intelligent prediction model for the operating status of new energy equipment is established using the long short-term memory network (LSTM) based on the attention mechanism. The training data is input into the model as the model input vector. The model output value is determined through model learning and training to obtain a trained intelligent prediction model for the operating status.

[0013] The features extracted in real time are input into the model as model input vectors to determine the model output values, thus achieving accurate prediction of the status of substation electrical equipment and conducting risk assessment.

[0014] In one embodiment, the preprocessing of the operation log includes:

[0015] Remove non-character tokens and stop words from the run log, keep only key characters, and split compound tokens into single words according to the camel case principle.

[0016] In one embodiment, the use of the principal component analysis model to extract and label features of the denoised device operation signal data includes:

[0017] Obtain the preprocessed operation logs and map the operation log data set to a high-dimensional space;

[0018] In the high-dimensional space, the high-dimensional data is projected into a low-dimensional space with the same dimension as the original dimension by using PCA transformation, and the dimension of the operation log data set is reduced, and finally a reconstructed data set is obtained.

[0019] The degree of abnormality of each sample is determined by calculating the reconstruction error. The reconstruction error E s (x) is calculated as:

[0020]

[0021] Among them, d is the preset dimension, represents the center point of the i-th cluster after clustering, i and j represent the number of clusters and dimension counts respectively, C i represents the i-th family, and k represents the number of clusters.

[0022] In one embodiment, the method of expanding the labeled feature value data based on a generative adversarial network and verifying the expanded data using a non-parametric verification method includes:

[0023] Generative adversarial networks are used to expand and enhance log data;

[0024] The KS verification method is used to screen and expand the data. The KS verification method calculates the method D iss as follows:

[0025]

[0026] Among them, F n (X) represents the empirical distribution function of sample X, F m (Y) represents the empirical distribution function of sample Y, n and m represent the number of samples X and Y respectively;

[0027] When D iss If the data is within the preset threshold, the qualified extended data is retained, otherwise it is discarded.

[0028] In one embodiment, the generative adversarial network is constructed in the following manner:

[0029] The generator deceives the discriminator by generating defects similar to the actual sample data;

[0030] The discriminator extracts abnormal data features by introducing a convolutional layer with a self-attention mechanism and classifies the abnormal data features.

[0031] The discriminator loss function is obtained by judging the supervised loss function of the discriminator through cross entropy.

[0032] Use the generative adversarial method between the generator and the discriminator to train the network model and optimize the loss function; improve the generative adversarial network model.

[0033] In one embodiment, the method of establishing an intelligent prediction model for the operation status of new energy equipment using a long short-term memory network LSTM based on an attention mechanism includes:

[0034] S1, input layer: standardize the labeled feature value data according to the preset dimension data to obtain input data of multiple dimensions;

[0035] S2, word embedding layer: uses a vector machine model to convert input data into a word embedding sequence;

[0036] S3, the first LSTM layer: converts the word vector sequence into high-dimensional features, calculates the candidate memory information of the word vector sequence at the current moment through the input gate, obtains the hidden state through the calculation results of the forget gate and the input gate, and outputs the word vector sequence feature information;

[0037] S4, sentence vector layer: the word vector sequence feature information is weighted by global information to obtain the sentence vector sequence;

[0038] S5, the second LSTM layer: converts the sentence vector sequence into high-dimensional features, calculates the candidate memory information of the sentence vector sequence at the current moment through the input gate, obtains the hidden state through the calculation results of the forget gate and the input gate, and outputs the sentence vector sequence feature information;

[0039] S6, attention layer: perform self-attention calculation on the sentence vector feature sequence to obtain self-attention features;

[0040] S7, fusion feature layer: the self-attention features are concatenated and fused to obtain enhanced features;

[0041] S8, output layer: The enhanced features are input into the fully connected layer. The fully connected layer integrates the enhanced feature information, maps the multi-dimensional feature information into two-dimensional feature input, and obtains the prediction result of the electrical equipment status of the substation at the next moment.

[0042] In one embodiment, the standardization of the labeled feature value data according to the preset dimension data includes:

[0043] The step of normalizing the labeled feature value data according to the preset dimension data includes:

[0044]

[0045] Where X is the original data, The mean value of the original data, σ is the standard deviation of the original data.

[0046] In one embodiment, in step S6, the loss function L com for:

[0047] L com =λ 1 l 1 +λ 2 l 2 ,

[0048]

[0049] in is the model output value, y is the measured value, t and T represent the number of real samples, and λ 1 , 2 Represents the weight coefficient.

[0050] In one embodiment, in step S7, the steps of the fusion operation are:

[0051] Use the fully connected layer to reduce the channel dimension and then increase the dimension;

[0052] Multiply the shallow features by channel, fuse with the deep features to get channel alignment, and get the aligned fused feature F m ;

[0053] The fused feature F m The calculation formula is:

[0054]

[0055] Among them, A v () indicates calculating the pooled average value, M a () indicates calculating the global pooling maximum value, w 1 、w 2 Represent the weights of adjacent connection layers, F h Represents the deep features, F 1 represents the initial features, and δ represents the activation function.

[0056] In a second aspect, an embodiment of the present application provides an artificial intelligence-based device state prediction and risk assessment device, which is applied to the artificial intelligence-based device state prediction and risk assessment method as described in the first aspect, and the device includes:

[0057] The data acquisition module is used to obtain the operation log of the new energy equipment, pre-process the operation log, use the principal component analysis model to extract and annotate the features of the denoised equipment operation signal data, and obtain labeled feature value data that characterizes the equipment operation status and health status.

[0058] The data processing module is used to expand the labeled feature value data based on the generative adversarial network, verify the expanded data using a non-parametric verification method, and select the expanded data and the original data as training data.

[0059] The model building module is used to establish an intelligent prediction model for the operating status of new energy equipment using the long short-term memory network LSTM based on the attention mechanism, input the training data into the model as the model input vector, determine the model output value through model learning and training, and obtain a trained intelligent prediction model for the operating status.

[0060] The state prediction module is used to input the real-time extracted features into the model as model input vectors, determine the model output value, achieve accurate prediction of the state of substation electrical equipment, and conduct risk assessment.

[0061] In a third aspect, an embodiment of the present application provides an electronic device, including:

[0062] processor;

[0063] a memory for storing processor-executable instructions;

[0064] Among them, the processor is configured to implement the artificial intelligence-based equipment status prediction and risk assessment method as described in the first aspect when executing the instructions.

[0065] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions, wherein the instructions instruct a device to execute the artificial intelligence-based device state prediction and risk assessment method as described in the first aspect.

[0066] The present invention provides an equipment state prediction method and device based on artificial intelligence, including: obtaining the operation log of new energy equipment, using the principal component analysis model to extract and annotate the denoised equipment operation signal data; expanding the labeled feature value data based on the generative adversarial network, and using the non-parametric verification method to verify the expanded data to obtain training data; training to establish an intelligent prediction model for the operation state of new energy equipment using the long short-term memory network LSTM based on the attention mechanism, and obtaining a trained intelligent prediction model for the operation state; finally determining the model output value, realizing accurate prediction of the state of substation electrical equipment, and performing risk assessment. The principal component method is used to reduce the number of evaluation indicators by dimensionality reduction, and the long short-term memory network LSTM model based on the attention mechanism is used to improve the accuracy and efficiency of the device health assessment; the generalization ability and prediction performance of the model are enhanced; and the stability and reliability of the model are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1A flowchart of an artificial intelligence-based device status prediction and risk assessment method provided in one embodiment of the present application.

[0068] Figure 2 A schematic diagram of the structure of a long short-term memory network LSTM model based on an attention mechanism is provided for another embodiment of the present application.

[0069] Figure 3 A schematic diagram of the execution flow of a long short-term memory network LSTM model based on an attention mechanism provided in another embodiment of the present application.

[0070] Figure 4 A schematic diagram of a module of an equipment status prediction and risk assessment device based on artificial intelligence provided in one embodiment of the present application.

[0071] Figure 5 A schematic diagram of an electronic device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0072] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments.

[0073] It should be noted that, in the embodiments of the present application, "at least one" refers to one or more, and "more" refers to two or more. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art in the present application. The terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application.

[0074] It should be noted that, in the embodiments of the present application, words such as "first" and "second" are only used for the purpose of distinguishing descriptions, and cannot be understood as indicating or implying relative importance, nor can they be understood as indicating or implying order. Features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way.

[0075] Based on the implementations in this application, all other implementations obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0076] With the rapid development of the power industry, the operating environment of power facilities is becoming increasingly complex, and equipment failures and defects pose a severe challenge to the stability and safety of power facilities. In particular, key facilities such as high-voltage transmission lines, substations, and power generation equipment, once they fail, will not only affect the stable operation of the power system, but may also cause widespread power outages, and even cause casualties and property losses. Therefore, it is particularly important to invent an artificial intelligence-based equipment status prediction and risk assessment method to accurately predict the equipment status.

[0077] At present, traditional power facility inspection methods mostly rely on manual inspections and regular inspections. This method is not only labor-intensive, but also limited by the frequency and efficiency of manual inspections, making it difficult to detect minor defects in a timely manner. In addition, the disadvantage of traditional methods is the lack of continuous monitoring and trend prediction of equipment operating status, making it difficult to cope with the increasingly complex power facility management needs.

[0078] In view of this, the present application provides an artificial intelligence-based equipment status prediction and risk assessment method that can accurately predict equipment status and perform risk assessment.

[0079] Figure 1 This is a flow chart of a method for predicting device status and assessing risk based on artificial intelligence provided by an embodiment of the present application. Figure 1 The artificial intelligence-based equipment status prediction and risk assessment method shown includes at least the following steps:

[0080] S110, obtaining the operation log of the new energy equipment, preprocessing the operation log, using the principal component analysis model to extract and annotate the features of the denoised equipment operation signal data, and obtaining labeled feature value data representing the equipment operation status and health status.

[0081] In the embodiment of the present application, in step S110, the operation log is preprocessed, including: removing non-character tags and stop words in the operation log, retaining only key characters, and splitting the complex tags into single words according to the principle of camel case nomenclature.

[0082] Specifically, the purpose of log analysis data preprocessing can also include filtering non-compliant data, cleaning, cleaning out meaningless data, and format conversion and regularization. According to subsequent needs, filter and separate basic data of different topics. Data preprocessing can be achieved through MR programs. Specifically, for example, the clickstream model is used to generate pageviews and visit information tables to facilitate subsequent statistical analysis.

[0083] It is understandable that Camel-Case is a set of naming rules (conventions) when writing computer programs. As its name CamelCase indicates, it refers to the use of a mixture of uppercase and lowercase letters to form the names of variables and functions. In order to make it easier for programmers to communicate their codes among their peers, they often adopt a unified naming method that is more readable.

[0084] Specifically, the main steps of the traditional principal component analysis method are: select initial variables according to the research question; determine whether it is appropriate to conduct principal component analysis (KMO and Bartlett test); calculate the correlation coefficient matrix, calculate the eigenvalues ​​and eigenvectors, calculate the variance contribution rate and the cumulative variance contribution rate. Select the principal component. Explain the principal component. Conduct further analysis after the principal component analysis.

[0085] The above are the basic steps of principal component analysis. It should be noted that different software and tools may have different implementation details, but the concepts are similar.

[0086] It can be understood that the principal component analysis model used in this embodiment to extract and annotate features of the denoised device operation signal data includes:

[0087] Obtain the preprocessed operation logs and map the operation log data set to a high-dimensional space;

[0088] In the high-dimensional space, the high-dimensional data is projected into a low-dimensional space with the same dimension as the original dimension by using PCA transformation, and the dimension of the operation log data set is reduced, and finally a reconstructed data set is obtained.

[0089] The degree of abnormality of each sample is determined by calculating the reconstruction error. The reconstruction error E s (x) is calculated as:

[0090]

[0091] Among them, d is the preset dimension, represents the center point of the i-th cluster after clustering, i and j represent the number of clusters and dimension counts respectively, C i represents the i-th family, and k represents the number of clusters.

[0092] S112, based on the generative adversarial network, the labeled feature value data is expanded, the expanded data is verified using a non-parametric verification method, and the expanded data and the original data are selected as training data.

[0093] The main purpose of principal component analysis (PCA) is to combine many variables of the original data and reduce them into a small group of new variables that can express all the information of the original data, thereby effectively describing the overall internal structure of the original data.

[0094] The following are the effective results of principal component analysis:

[0095] (1) Data dimensionality reduction: PCA transforms the original data into a set of new variables, i.e., principal components, whose dimensions are linearly independent through linear transformation. These new variables can explain most of the variation in the original data, thus achieving data dimensionality reduction.

[0096] (2) Improve data representation: Obtain higher representation accuracy from the original variables, especially when the data volume is large.

[0097] (3) Identifying latent structures in the data: PCA can identify latent structures in the data and reduce the number of variables while retaining as much information as possible.

[0098] (4) Simplify the data analysis process: PCA simplifies the data analysis process by extracting the main components in the data.

[0099] Therefore, the effect of principal component analysis is to help us better understand the intrinsic structure of the data by reducing dimensionality, improving data representation, identifying potential structures in the data, and simplifying the data analysis process.

[0100] In the embodiment of the present application, the method of expanding the labeled feature value data based on the generative adversarial network and verifying the expanded data using a non-parametric verification method includes:

[0101] Generative adversarial networks are used to expand and enhance log data;

[0102] The KS verification method is used to screen and expand the data. The KS verification method calculates the method D iss as follows:

[0103]

[0104] Among them, F n (X) represents the empirical distribution function of sample X, F m (Y) represents the empirical distribution function of sample Y, n and m represent the number of samples X and Y respectively;

[0105] When D issWhen it is within the preset threshold range, qualified augmented data is retained, otherwise it is discarded. It can be understood that the Generative Adversarial Networks (GANs) model is a deep learning model. The core idea of ​​GAN is to generate data through adversarial training so that the generated samples are as close to the real samples as possible.

[0106] It can be understood that the Generator: The task of the Generator is to generate realistic data samples. It accepts random noise as input and tries to generate a sample that looks real (such as an image, text, etc.). The Discriminator: The task of the Discriminator is to distinguish whether the input sample is from a real data set or a fake sample generated by the Generator. It outputs a probability value indicating whether the input sample is real or generated. The training process of GAN is a game process in which the Generator and the Discriminator compete with each other.

[0107] It can be understood that the generator: tries to generate fake samples that can deceive the discriminator, making it difficult for the discriminator to distinguish between generated samples and real samples. The discriminator: tries to improve its ability to distinguish so that real samples and fake samples can be more accurately identified. Adversarial training: The training process of GAN is a zero-sum game, where the generator and the discriminator confront each other. The generator tries to maximize the discriminator's misjudgment rate of generated samples, while the discriminator tries to minimize this misjudgment rate.

[0108] Specifically, the generative adversarial network is constructed in the following way:

[0109] The generator deceives the discriminator by generating defects similar to the actual sample data;

[0110] The discriminator extracts abnormal data features by introducing a convolutional layer with a self-attention mechanism, and classifies the abnormal data features;

[0111] The discriminator loss function is obtained by judging the supervised loss function of the discriminator through cross entropy;

[0112] Use the generative adversarial method between the generator and the discriminator to train the network model and optimize the loss function;

[0113] Improving Generative Adversarial Network models.

[0114] Specifically, the training process of the generative adversarial network is as follows: Initialization: the generator and the discriminator are initialized to random states; Generation: the generator generates a fake sample from random noise; Discriminator: the discriminator classifies the generated samples and real samples, calculates their prediction results, and provides feedback based on its accuracy of the generated samples; Optimization: the generator and the discriminator update their parameters through back propagation, the generator adjusts its generation strategy through the feedback of the discriminator, and the discriminator improves its accuracy by distinguishing between real samples and generated samples; Iteration: this process is repeated until the generator can generate very realistic samples and the discriminator has difficulty distinguishing between generated samples and real samples.

[0115] S114, using the long short-term memory network LSTM based on the attention mechanism to establish an intelligent prediction model for the operating status of new energy equipment, inputting the training data into the model as the model input vector, determining the model output value through model learning and training, and obtaining a trained intelligent prediction model for the operating status.

[0116] Understandably, it is well known that LSTM does not handle long sequences and the prominence of important information well, which leads to poor performance in some cases. The attention mechanism simulates the characteristics of human visual attention mechanism and can solve this problem well. LSTM (Long Short-Term Memory Network) and Attention Mechanism are two important technologies in deep learning, which can be combined with each other to improve the performance of the model, especially when processing sequence data.

[0117] Specifically, the attention mechanism determines which parts of the input sequence should be focused on through weight distribution, which allows the model to dynamically adjust its focus when generating outputs in order to better capture the key information in the input sequence. In this way, by combining the long-term dependency capture capability of LSTM and the dynamic focus adjustment capability of the attention mechanism, our model can more effectively handle various complex sequence processing tasks and be applied to more fields.

[0118] The core of LSTM is a cell state and three gating mechanisms (input gate, forget gate, output gate). These gates dynamically control the flow of information by screening information, allowing the model to selectively remember or forget certain information. Cell state: The cell state is the memory unit of LSTM, which is used to carry long-term contextual information. It avoids the rapid dissipation of gradients through linear transmission (a small number of weighted operations and gated updates). Gating mechanism: Each gate consists of a Sigmoid activation function and some point multiplication operations, and the output range is [0, 1], indicating the degree of passage of a certain part of information. Forget gate: Controls which information needs to be deleted from the cell state. The formula is: f tIs the output of the forget gate, which determines whether to keep the previous information in the cell state. Input gate: controls which new information needs to be added to the cell state. It is divided into two steps: Decide which information needs to be updated: Generate candidate information: Finally update the cell status: Output gate: determines the hidden state at the current moment (i.e. output): The hidden state is:

[0119] Among them, "*" represents convolution, Hadamard product, b f , b i , b c , b o They represent the bias items of the forget gate, input gate, candidate information, and output gate, respectively, and are the weight matrices of the forget gate, input gate, and output gate, respectively. Among them, W xf , W xi , W xc and W xo are the weight matrices of the forget gate, input gate, candidate information, and output gate, σ is the sigmoid activation function, and tanh is the hyperbolic tangent activation function. At the same time, in order to keep the output and input sizes consistent, zero padding is required before the convolution operation.

[0120] The core of LSTM is to dynamically control the flow of information through the gating mechanism, and realize memory and forgetting in conjunction with the cell state. It is achieved through the following steps: The forget gate determines what information needs to be forgotten. The input gate determines what new information needs to be added. The cell state is updated to store new memories. The output gate controls the output at the current moment.

[0121] Specifically, if Figure 2 As shown, it is a schematic diagram of the structure of a long short-term memory network LSTM model based on the attention mechanism provided by another embodiment of the present application. Figure 3 As shown, it is a schematic diagram of the execution flow of the LSTM model based on the attention mechanism provided by another embodiment of the present application. The model is divided into 8 layers, namely: input layer, word vector layer, first LSTM layer, sentence vector layer, second LSTM layer, attention layer, fusion feature layer, and output layer.

[0122] The structure and function of the intelligent prediction model for the operation status of new energy equipment is established by using the long short-term memory network LSTM based on the attention mechanism, including:

[0123] S1, input layer: standardize the labeled feature value data according to the preset dimension data to obtain input data of multiple dimensions;

[0124] S2, word embedding layer: uses a vector machine model to convert input data into a word embedding sequence;

[0125] S3, the first LSTM layer: converts the word vector sequence into high-dimensional features, calculates the candidate memory information of the word vector sequence at the current moment through the input gate, obtains the hidden state through the calculation results of the forget gate and the input gate, and outputs the word vector sequence feature information;

[0126] S4, sentence vector layer: the word vector sequence feature information is weighted by global information to obtain the sentence vector sequence;

[0127] S5, the second LSTM layer: converts the sentence vector sequence into high-dimensional features, calculates the candidate memory information of the sentence vector sequence at the current moment through the input gate, obtains the hidden state through the calculation results of the forget gate and the input gate, and outputs the sentence vector sequence feature information;

[0128] S6, attention layer: perform self-attention calculation on the sentence vector feature sequence to obtain self-attention features;

[0129] S7, fusion feature layer: the self-attention features are concatenated and fused to obtain enhanced features;

[0130] S8, output layer: The enhanced features are input into the fully connected layer. The fully connected layer integrates the enhanced feature information, maps the multi-dimensional feature information into two-dimensional feature input, and obtains the prediction result of the electrical equipment status of the substation at the next moment.

[0131] Specifically, the standardization of the labeled feature value data according to the preset dimension data includes:

[0132] The step of normalizing the labeled feature value data according to the preset dimension data includes:

[0133]

[0134] Where X is the original data, The average value of the original data, σ is the standard deviation of the original data

[0135] Specifically, in step S6, the loss function L com for:

[0136] L com =λ 1 l 1 +λ 2 l 2 ,

[0137]

[0138] in is the model output value, y is the measured value, t and T represent the number of real samples, and λ 1 , 2 Represents the weight coefficient.

[0139] Specifically, in step S7, the steps of the fusion operation are:

[0140] Use the fully connected layer to reduce the channel dimension and then increase the dimension;

[0141] Multiply the shallow features by channel, fuse with the deep features to get channel alignment, and get the aligned fused feature F m ;

[0142] The fused feature F m The calculation formula is:

[0143]

[0144] Among them, A v () indicates calculating the pooled average value, M a () indicates calculating the global pooling maximum value, w 1 、w 2 Represent the weights of adjacent connection layers, F h Represents the deep features, F 1 represents the initial features, and δ represents the activation function.

[0145] Fully connected networks have too many parameters and require too much computation. However, after convolution and pooling, the size of CNN has been greatly reduced, and finally one or two layers of full connection are used, so the computation is within an acceptable range.

[0146] S116 is used to input the features extracted in real time into the model as model input vectors, determine the model output values, achieve accurate prediction of the status of the substation electrical equipment, and perform risk assessment.

[0147] Specifically, it can realize real-time monitoring of equipment operation status, release of alarm and early warning information, and real-time grasp and judgment of equipment operation status; realize the collection and integration of equipment status-related status data, and deep mining and analysis of multi-dimensional big data; realize the summary and analysis of equipment operation problems, provide support for the formulation of equipment operation and maintenance strategies, formulate anti-accident measures, etc., provide decision-making support for production command, and effectively guide the development of equipment maintenance, technical transformation and other work.

[0148] In the embodiment of the present application, the device system aims to guide equipment status maintenance and full life cycle management, based on the full-dimensional status information of substation equipment, with the application model after big data analysis as the core, and multi-dimensional function expansion around defect / fault diagnosis and prediction, which has great guiding significance for the operation and maintenance of substation equipment. Targeted in-depth mining and application of equipment status data mainly include: big data → typical defect library; big data → fault prediction; big data → family defects; big data → equipment life prediction, etc.

[0149] It is understandable that the artificial intelligence-based equipment status prediction and risk assessment method provided in the embodiment of the present application can realize fully automated status prediction and risk assessment, which not only improves the efficiency of power facility maintenance, but also reduces equipment failures and downtime through early warning. In addition, the system integrates a variety of advanced artificial intelligence technologies and a variety of data processing methods, and can process and analyze various complex information of power facilities, greatly improving the accuracy and reliability of fault prediction, and reducing safety accidents and economic losses caused by equipment failures.

[0150] Figure 4 Schematic diagram of a device module for predicting equipment status and assessing risk based on artificial intelligence provided by another embodiment of the present application. Figure 1 The artificial intelligence-based equipment status prediction and risk assessment method shown in FIG. Figure 4 The device for predicting equipment status and risk assessment based on artificial intelligence shown in the figure comprises: a data acquisition module 1, a data processing module 2, a model building module 3, and a status prediction module 4. Its functions are similar to Figure 1 The parts shown in FIG. 1 are the same or similar and will not be described in detail here.

[0151] In an embodiment of the present application, the data acquisition module 1 is used to obtain the operation log of the new energy equipment, preprocess the operation log, use the principal component analysis model to extract and label the denoised equipment operation signal data, and obtain labeled feature value data that characterizes the equipment operation status and health status.

[0152] In an embodiment of the present application, the data processing module 2 is used to expand the labeled feature value data based on a generative adversarial network, verify the expanded data using a non-parametric verification method, and select the expanded data and the original data as training data.

[0153] In the embodiment of the present application, the model building module 3 is used to establish an intelligent prediction model for the operation status of new energy equipment using a long short-term memory network LSTM based on an attention mechanism, input the training data into the model as a model input vector, determine the model output value through model learning and training, and obtain a trained intelligent prediction model for the operation status. For specific acquisition methods, please refer to Figure 1-4 And the corresponding description thereof, this application will not repeat them here.

[0154] In the embodiment of the present application, the state prediction module 4 is used to input the real-time extracted features as model input vectors into the model, determine the model output value, and achieve accurate prediction of the state of the substation electrical equipment.

[0155] Figure 5 This is an electronic device provided by an embodiment of the present application. Figure 5 As shown, the electronic device includes at least the following parts: a processor 101 and a memory 100 , a communication interface 103 , and a bus 102 .

[0156] In the embodiment of the present application, the memory 100 is used to store instructions executable by the processor 101, and the processor 101 is configured to implement the following when executing the instructions: Figure 3 The artificial intelligence-based equipment status prediction and risk assessment method shown.

[0157] In an embodiment of the present application, a computer-readable storage medium includes instructions, and the instructions instruct a device to execute the method of the first aspect. For example, the instructions instruct the device to execute Figure 4 The artificial intelligence-based equipment status prediction and risk assessment method is shown in the process steps.

[0158] The program operating in the electronic device involved in one embodiment of the present application may be a program (a program that enables a computer to function) that controls a central processing unit (CPU) and the like to realize the functions of the above-mentioned implementation method involved in one scheme of the present invention. Then, the information processed by these devices is temporarily stored in a random access memory (RAM) during its processing, and then stored in various ROMs such as a read-only memory (Flash ROM) and a hard disk drive (HDD), and is read, corrected, and written by the CPU as needed.

[0159] It should be noted that a part of the electronic device of the above embodiment may also be implemented by a computer. In this case, a program for implementing the control function may be recorded in a computer-readable recording medium, and the program recorded in the recording medium may be read into a computer and executed.

[0160] It should be noted that the "computer" mentioned here refers to a computer built into an electronic device, and uses a computer including hardware such as an OS and peripheral devices. In addition, "computer-readable recording medium" refers to removable media such as floppy disks, magneto-optical disks, ROMs, CD-ROMs, and storage devices such as hard disks built into computers.

[0161] Furthermore, "computer-readable recording media" may include: media that dynamically store programs for a short period of time, such as communication lines when sending programs via networks such as the Internet or communication lines such as telephone lines; and media that store programs for a fixed period of time, such as volatile memories inside computers that serve as servers or clients in this case. In addition, the above-mentioned program may be a program for realizing a part of the above-mentioned functions, or a program that can realize the above-mentioned functions by combining with a program already recorded in a computer.

[0162] In addition, the electronic device in the above-mentioned embodiment can also be implemented as a collection (device group) composed of multiple devices. Each device constituting the device group can have a part or all of the functions or functional blocks of the electronic device in the above-mentioned embodiment. As a device group, it is sufficient to have all the functions or functional blocks of the electronic device.

[0163] It is understandable that the artificial intelligence-based equipment status prediction and risk assessment method, device, electronic device and storage medium provided in the embodiments of the present application can effectively improve the operation and maintenance level of power facilities and solve the problem that traditional manual inspections cannot identify facility defects in real time and comprehensively. Through the application of this system, potential hidden dangers of facility failures can be quickly discovered, and corresponding repair suggestions and optimization solutions can be provided through intelligent analysis, thereby minimizing equipment downtime and maintenance costs. At the same time, the system's adaptive ability enables it to continuously learn and optimize, improve performance based on more data collected in real time, and further enhance its adaptability to different types of power facilities and its ability to predict future defects.

[0164] Those skilled in the art should recognize that the above embodiments are only used to illustrate the present application and are not intended to be limiting of the present application. As long as they are within the spirit and scope of the present application, appropriate changes and modifications to the above embodiments are within the scope of protection claimed in the present application.

Claims

1. A method for predicting equipment status and risk assessment based on artificial intelligence, characterized in that: The method comprises: Obtaining the operation log of the new energy equipment, preprocessing the operation log, and using the principal component analysis model to extract and annotate the de-noised equipment operation signal data to obtain labeled feature value data representing the equipment operation status and health status; Based on the generative adversarial network, the labeled feature value data is expanded, the expanded data is tested using a non-parametric verification method, and the expanded data and the original data are selected as training data; The long short-term memory network LSTM based on the attention mechanism is used to establish an intelligent prediction model for the operation status of new energy equipment. The training data is input into the model as the model input vector. Through the learning and training of the model, the model output value is determined to obtain a trained intelligent prediction model for the operation status. The features extracted in real time are input into the model as model input vectors to determine the model output values, thus achieving accurate prediction of the status of substation electrical equipment and conducting risk assessment.

2. The method for predicting equipment status and assessing risk based on artificial intelligence according to claim 1, characterized in that: The preprocessing of the operation log includes: Remove non-character tokens and stop words from the run log, keep only key characters, and split compound tokens into single words according to the camel case principle.

3. The method for predicting equipment status and assessing risk based on artificial intelligence according to claim 1, characterized in that: The method of using the principal component analysis model to extract and label the features of the denoised equipment operation signal data includes: Obtain the preprocessed operation logs and map the operation log data set to a high-dimensional space; In the high-dimensional space, the high-dimensional data is projected into a low-dimensional space with the same dimension as the original dimension by using PCA transformation, and the dimension of the operation log data set is reduced, and finally a reconstructed data set is obtained; The degree of abnormality of each sample is determined by calculating the reconstruction error. The reconstruction error E s (x) is calculated as: Among them, d is the preset dimension, represents the center point of the i-th cluster after clustering, i and j represent the number of clusters and dimension counts respectively, C i represents the i-th family, and k represents the number of clusters.

4. The method for predicting equipment status and assessing risk based on artificial intelligence according to claim 2, characterized in that: The method of expanding the labeled feature value data based on the generative adversarial network and verifying the expanded data using a non-parametric verification method includes: Generative adversarial networks are used to expand and enhance log data; The KS verification method is used to screen and expand the data. The KS verification method calculates the method D iss as follows: Among them, F n (X) represents the empirical distribution function of sample X, F m (Y) represents the empirical distribution function of sample Y, n and m represent the number of samples X and Y respectively; When D iss If the data is within the preset threshold, the qualified extended data is retained, otherwise it is discarded.

5. The method for predicting equipment status and assessing risk based on artificial intelligence according to claim 4 is characterized in that: The generative adversarial network is constructed in the following way: The generator deceives the discriminator by generating defects similar to the actual sample data; The discriminator extracts abnormal data features by introducing a convolutional layer with a self-attention mechanism, and classifies the abnormal data features; The discriminator loss function is obtained by judging the supervised loss function of the discriminator through cross entropy; Use the generative adversarial method between the generator and the discriminator to train the network model and optimize the loss function; Improving Generative Adversarial Network models.

6. The method for predicting equipment status and assessing risk based on artificial intelligence according to claim 5 is characterized in that: The method of establishing an intelligent prediction model for the operation status of new energy equipment by using a long short-term memory network LSTM based on an attention mechanism includes: S1, input layer: standardize the labeled feature value data according to the preset dimension data to obtain input data of multiple dimensions; S2, word embedding layer: uses a vector machine model to convert input data into a word embedding sequence; S3, the first LSTM layer: converts the word vector sequence into high-dimensional features, calculates the candidate memory information of the word vector sequence at the current moment through the input gate, obtains the hidden state through the calculation results of the forget gate and the input gate, and outputs the word vector sequence feature information; S4, sentence vector layer: the word vector sequence feature information is weighted by global information to obtain the sentence vector sequence; S5, the second LSTM layer: converts the sentence vector sequence into high-dimensional features, calculates the candidate memory information of the sentence vector sequence at the current moment through the input gate, obtains the hidden state through the calculation results of the forget gate and the input gate, and outputs the sentence vector sequence feature information; S6, attention layer: perform self-attention calculation on the sentence vector feature sequence to obtain self-attention features; S7, fusion feature layer: the self-attention features are concatenated and fused to obtain enhanced features; S8, output layer: The enhanced features are input into the fully connected layer. The fully connected layer integrates the enhanced feature information, maps the multi-dimensional feature information into two-dimensional feature input, and obtains the prediction result of the electrical equipment status of the substation at the next moment.

7. The method for predicting equipment status and assessing risk based on artificial intelligence according to claim 6 is characterized in that: The step of normalizing the labeled feature value data according to the preset dimension data includes: The step of normalizing the labeled feature value data according to the preset dimension data includes: Where X is the original data, The mean value of the original data, σ is the standard deviation of the original data.

8. The method for predicting equipment status and assessing risk based on artificial intelligence according to claim 6, characterized in that: In step S6, the loss function L com for: L com =λ1l1+λ2l2, in is the model output value, y is the measured value, t and T represent the number of real samples, and λ1 and λ2 represent weight coefficients.

9. The method for predicting equipment status and assessing risk based on artificial intelligence according to claim 6, characterized in that: In step S7, the steps of the fusion operation are: Use the fully connected layer to reduce the channel dimension and then increase the dimension; Multiply the shallow features by channel, fuse with the deep features to get channel alignment, and get the aligned fused feature F m ; The fused feature F m The calculation formula is: Among them, A v () indicates calculating the pooled average value, M a () indicates the calculation of the global pooling maximum value, w1 and w2 represent the weights of adjacent connection layers, respectively. h represents the deep features, F1 represents the initial features, and δ represents the activation function.

10. An artificial intelligence-based device state prediction and risk assessment device, applied to the artificial intelligence-based device state prediction and risk assessment method according to any one of claims 1 to 9, characterized in that: The device comprises: A data acquisition module is used to obtain the operation log of the new energy equipment, pre-process the operation log, use the principal component analysis model to extract and annotate the de-noised equipment operation signal data, and obtain the labeled feature value data representing the equipment operation status and health status; A data processing module is used to expand the labeled feature value data based on the generative adversarial network, verify the expanded data using a non-parametric verification method, and select the expanded data and the original data as training data; The model building module is used to establish an intelligent prediction model for the operation status of new energy equipment using the long short-term memory network LSTM based on the attention mechanism, input the training data into the model as the model input vector, determine the model output value through model learning and training, and obtain a trained intelligent prediction model for the operation status; The state prediction module is used to input the real-time extracted features into the model as model input vectors, determine the model output value, achieve accurate prediction of the state of substation electrical equipment, and conduct risk assessment.

Images