Power grid equipment health state prediction method and system based on multi-modal data fusion

Abstract

The invention provides a power grid equipment health state prediction method and system based on multi-modal data fusion, and the method comprises the steps: obtaining the multi-modal state associated data of target power grid equipment in real time; performing data feature extraction on the multi-modal state associated data according to a preset multi-modal feature extraction model to obtain a to-be-analyzed multi-modal state feature, and dynamically obtaining an optimal fusion weight of the to-be-analyzed multi-modal state feature according to the obtained historical cycle health index prediction error of the target power grid equipment and the equipment operation environment data; and according to the optimal fusion weight, carrying out feature fusion on the to-be-analyzed multi-modal state features to obtain a corresponding equipment health index prediction value. According to the method, rich equipment operation state characteristics are captured in real time through multi-source data, and a multi-modal characteristic fusion weight dynamic optimization strategy is combined, so that the timeliness, the high efficiency and the accuracy of equipment health assessment are effectively improved, the power production loss and the maintenance cost are reduced, and reliable guarantee is provided for stable and safe operation of a power system.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid equipment status monitoring, and in particular to a power grid equipment health status prediction method and system based on multi-modal data fusion. Background Art

[0002] As the scale of the power system continues to expand, the number of power grid devices increases, which also brings challenges to the safe operation and management of the power system. The operational stability of the power system is directly related to social production and life, and the reliability and health status of power grid equipment are the key to ensuring stable power supply. Therefore, how to reliably and accurately evaluate the health status of power grid equipment is an important guarantee for maintaining the stable and safe operation of the power system.

[0003] In order to monitor the health status of power grid equipment, the industry has developed many physical model monitoring methods and artificial intelligence fault prediction methods, but they all have certain application defects and are difficult to effectively meet the needs of intelligent, efficient and reliable operation status evaluation of power grid equipment: physical model monitoring methods mainly rely on regular equipment detection and subsequent analysis, based on physical and chemical tests, combined with the experience and judgment of electricians. For example, chromatographic analysis of transformer oil is used to detect potential faults, or infrared thermal imaging technology is used to detect equipment overheating problems. Although it can directly and specifically judge the specific state of a certain type of power grid equipment at a specific time, its shutdown detection will interfere with the continuous operation of the power system. The evaluation process has a lag and it is difficult to ensure the timeliness and preventability of equipment fault discovery. The subjectivity of human evaluation makes it difficult to ensure the reliability of state analysis. Although artificial intelligence fault prediction methods can realize fault prediction by analyzing the operation data of equipment through machine learning models, existing methods often ignore the multimodality of power grid equipment health status analysis, and only analyze and process a single type of data. In addition, the evaluation algorithm lacks sufficient adaptive adjustment mechanism to cope with the ever-changing power grid equipment and environmental conditions. It is difficult to adjust the model in real time to reflect the latest changes in equipment status, and it is difficult to ensure the accuracy and reliability of equipment status prediction results. Summary of the invention

[0004] The purpose of the present invention is to provide a method for predicting the health status of power grid equipment based on multimodal data fusion. By collecting multi-source data, rich equipment operation status characteristics are captured in real time, and a multimodal feature fusion weight dynamic optimization strategy is combined to perform a real-time online scientific and reasonable evaluation of the equipment health status. This solves the application defect that the existing power grid equipment status prediction is difficult to ensure real-time and accuracy. It can effectively improve the timeliness, efficiency and accuracy of equipment health assessment, reduce power production losses and maintenance costs, and provide reliable guarantee for the stable and safe operation of the power system.

[0005] In order to achieve the above objectives, it is necessary to provide a method and system for predicting the health status of power grid equipment based on multimodal data fusion in response to the above technical problems.

[0006] In a first aspect, an embodiment of the present invention provides a method for predicting the health status of power grid equipment based on multimodal data fusion, the method comprising the following steps: Acquire multimodal state-related data of target power grid equipment in real time; the multimodal state-related data includes equipment image data, equipment operation state time series data and equipment operation log data; According to a preset multimodal feature extraction model, data feature extraction is performed on the multimodal state association data to obtain corresponding multimodal state features to be analyzed; the multimodal state features to be analyzed include image features, state time series features and log text features; Obtaining the historical cycle health index prediction error and equipment operating environment data of the target power grid equipment, and dynamically obtaining the optimal fusion weight of the multimodal state feature to be analyzed based on the historical cycle health index prediction error and the equipment operating environment data; the historical cycle health index prediction error is the prediction error generated by predicting the health index of the target power grid equipment in the previous evaluation period; According to the optimal fusion weight, feature fusion is performed on the multimodal state features to be analyzed to obtain a corresponding device health index prediction value.

[0007] Furthermore, the preset multimodal feature extraction model includes a parallel preset neural network model, a long short-term memory network and a preset natural language processing model; The step of extracting data features from the multimodal state association data according to a preset multimodal feature extraction model to obtain corresponding multimodal state features comprises: Extracting features from the device image data according to the preset neural network model to obtain corresponding image features; Extracting features from the equipment operation status time series data according to the long short-term memory network to obtain corresponding status time series features; Feature extraction is performed on the device operation log data according to the preset natural language processing model to obtain corresponding log text features.

[0008] Furthermore, the step of obtaining the historical cycle health index prediction error of the target power grid device includes: According to the preset multimodal feature extraction model, feature extraction is performed on the multimodal state association data of the target power grid device in the previous historical period to obtain the corresponding multimodal features of the historical period; According to the fusion weight of the previous historical period, feature fusion is performed on the multimodal features of the historical period to obtain the corresponding historical period equipment health index prediction value; The true value of the equipment health index of the target power grid equipment in the previous historical period is obtained, and the corresponding historical period health index prediction error is obtained according to the true value of the equipment health index of the previous historical period and the predicted value of the equipment health index of the historical period.

[0009] Furthermore, the step of dynamically acquiring the optimal fusion weight of the multimodal state feature to be analyzed according to the historical cycle health index prediction error and the equipment operating environment data includes: According to the historical cycle health index prediction error and the equipment operating environment data, based on the genetic algorithm combined with reinforcement learning, the previous historical cycle fusion weight corresponding to the historical cycle health index prediction error is dynamically adjusted to obtain the optimal fusion weight; the fusion weight is expressed as: in, and They represent the fusion weights of the i-th target power grid equipment in the t+1-th and t-th prediction cycles respectively; represents the historical cycle health index prediction error of the i-th target power grid equipment in the t-th prediction period; Represents the equipment operating environment data of the i-th target power grid equipment; Represents the joint execution function consisting of reinforcement learning and genetic algorithm.

[0010] Furthermore, the step of dynamically adjusting the fusion weight of the previous historical period corresponding to the historical period health index prediction error based on the genetic algorithm combined with reinforcement learning to obtain the optimal fusion weight includes: According to the historical cycle health index prediction error and the equipment operating environment data, a corresponding fitness function is established, and the goal is to minimize the fitness function value. The fusion weight of the previous historical cycle is optimized by the genetic algorithm to obtain the initial optimized fusion weight; the fitness function is expressed as: in, represents the fitness function; represents the historical cycle health index prediction error of the i-th target power grid equipment in the t-th prediction period; Represents the equipment operating environment data of the i-th target power grid equipment; According to the historical cycle health index prediction error, a corresponding reward function is constructed, and the fitness function is used as the action value function, and the initial optimized fusion weight is optimized by reinforcement learning according to the reward function and the action value function to obtain the optimal fusion weight; the reward function is expressed as: in, represents the reward function; e represents the natural exponent; It represents the historical cycle health index prediction error of the i-th target power grid equipment in the t-th prediction period.

[0011] Furthermore, the step of performing reinforcement learning optimization on the initial optimized fusion weight according to the reward function and the action value function to obtain the optimal fusion weight includes: The equipment operating environment data and fusion weight are used as the input state and output action of the agent respectively, and the fusion weight adjustment amount is used as the Q value. The initial optimized fusion weight is updated based on Q learning; the update rule of the initial optimized fusion weight is expressed as: In the formula, in, and Respectively represent the fusion weights of the t+1th and tth prediction periods; represents the fusion weight adjustment required to update the fusion weight of the t-th prediction period; represents the learning rate; is the discount factor; Represents the gradient of the reward function with respect to the fusion weight; Represents the action value function value corresponding to the device operating environment data and fusion weight in the t-th prediction period.

[0012] Furthermore, the method further comprises: The historical health index sequence of the power system is obtained, and the historical health index sequence of the power system and the equipment health index sequence of each target power grid equipment in the power system are input into a preset system health status prediction model for operating status evaluation to obtain the corresponding system health index prediction value.

[0013] Furthermore, the preset system health status prediction model is a long short-term memory network; the system health index prediction value is expressed as: in, express The system health index prediction value output at all times; and are the weights and biases of the LSTM output layer, is the last unit of the long short-term memory network at time point The hidden state of .

[0014] Furthermore, the method further comprises: According to a preset self-evolution strategy, the preset multimodal feature extraction model and the preset system health status prediction model are updated through online learning.

[0015] In a second aspect, an embodiment of the present invention provides a power grid equipment health status prediction system based on multimodal data fusion, the system comprising: A status data acquisition module is used to acquire multimodal status associated data of target power grid equipment in real time; the multimodal status associated data includes equipment image data, equipment operation status time series data and equipment operation log data; A multimodal feature extraction module, used to extract data features from the multimodal state association data according to a preset multimodal feature extraction model to obtain corresponding multimodal state features to be analyzed; the multimodal state features to be analyzed include image features, state time series features and log text features; A fusion weight acquisition module is used to obtain the historical cycle health index prediction error and equipment operating environment data of the target power grid equipment, and dynamically obtain the optimal fusion weight of the multimodal state feature to be analyzed based on the historical cycle health index prediction error and the equipment operating environment data; the historical cycle health index prediction error is the prediction error generated by the health index prediction of the target power grid equipment in the previous evaluation period; The equipment health assessment module is used to perform feature fusion on the multimodal state features to be analyzed according to the optimal fusion weight to obtain the corresponding equipment health index prediction value.

[0016] The above-mentioned application provides a method and system for predicting the health status of power grid equipment based on multimodal data fusion. The method realizes real-time acquisition of multimodal status-related data of the target power grid equipment, including equipment image data, equipment operation status time series data and equipment operation log data. According to the preset multimodal feature extraction model, data feature extraction is performed on the multimodal status-related data to obtain the corresponding multimodal status features to be analyzed, including image features, status time series features and log text features. Then, the historical cycle health index prediction error and equipment operation environment data of the target power grid equipment are obtained, and the optimal fusion weight of the multimodal status features to be analyzed is dynamically obtained based on the historical cycle health index prediction error and the equipment operation environment data. According to the optimal fusion weight, the multimodal status features to be analyzed are feature fused to obtain the corresponding equipment health index prediction value. Technical solution. Compared with the existing technology, this power grid equipment health status prediction method based on multimodal data fusion captures rich equipment operation status characteristics in real time by collecting multi-source data, and combines the multimodal feature fusion weight dynamic optimization strategy to conduct real-time online scientific and reasonable evaluation of equipment health status. It can effectively improve the timeliness, efficiency and accuracy of equipment health assessment, reduce power production losses and maintenance costs, and provide reliable guarantee for the stable and safe operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of a method for predicting the health status of power grid equipment based on multimodal data fusion in an embodiment of the present invention; Figure 2 is another flow chart of a method for predicting the health status of power grid equipment based on multimodal data fusion in an embodiment of the present invention; Figure 3 is another flow chart of a method for predicting the health status of power grid equipment based on multimodal data fusion in an embodiment of the present invention; Figure 4 It is a structural schematic diagram of a power grid equipment health status prediction system based on multimodal data fusion in an embodiment of the present invention; Figure 5 is another structural schematic diagram of a power grid equipment health status prediction system based on multimodal data fusion in an embodiment of the present invention; Figure 6 It is another structural schematic diagram of the power grid equipment health status prediction system based on multimodal data fusion in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical scheme and beneficial effects of the present application clearer, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. Obviously, the embodiments described below are part of the embodiments of the present invention and are only used to illustrate the present invention, but are not used to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] The power grid equipment health status prediction method based on multimodal data fusion provided by the present invention can be understood as a method for extracting and integrating the features of multiple data sources of power grid equipment, and combining the multimodal feature fusion weight dynamic optimization strategy based on genetic algorithm and reinforcement learning to construct an equipment health index evaluation system, so as to realize real-time, scientific and comprehensive monitoring and evaluation of the equipment health status. The method can be applied to a terminal or a server. Among them, the terminal can be but not limited to various personal computers, laptops, smart phones, tablet computers and portable wearable devices, and the server can be implemented with an independent server or a server cluster composed of multiple servers. The server can use the power grid equipment health status prediction method based on multimodal data fusion provided by the present invention to perform real-time, efficient and accurate power grid equipment health index prediction according to actual application requirements, and use the obtained power grid equipment health index prediction value for subsequent system health index prediction and other research on the server, or transmit it to the terminal for terminal users to view and analyze; the following embodiment will describe the power grid equipment health status prediction method based on multimodal data fusion of the present invention in detail.

[0020] In one embodiment, Figure 1 As shown, a method for predicting the health status of power grid equipment based on multimodal data fusion is provided, comprising the following steps: S11. Acquire multimodal state association data of target power grid equipment in real time; wherein, target power grid equipment can be understood as various distribution equipment used in the power system, such as transformers, relays and circuit breakers, etc., which are not specifically limited here; the corresponding multimodal state association data can be understood as the collected multi-source data related to the health status of the target power grid equipment, considering the physical state of the power grid equipment, such as equipment wear, damage or other fault signs, etc., which can be reflected in the equipment image, and the voltage, current, fault code and other related data that change with time during the operation of the power grid equipment, as well as the operation log information recorded during the operation of the equipment, etc. can directly or indirectly reflect the health status of the power grid equipment. In order to ensure the comprehensiveness and reliability of the equipment status analysis, in this embodiment, preferably, in each health status assessment cycle of the target power grid equipment, multimodal state association data including equipment image data, equipment operation status time series data and equipment operation log data are collected as basic data for operation status analysis; it should be noted that the data content involved in the specific multimodal state association data varies depending on the type of the target power grid equipment, which is not described in detail here.

[0021] S12. According to a preset multimodal feature extraction model, data feature extraction is performed on the multimodal state associated data to obtain corresponding multimodal state features to be analyzed; the multimodal state features to be analyzed include image features, state time series features and log text features; wherein, the preset multimodal feature extraction model can be understood as a machine learning model that can effectively extract features from various source data in the collected multimodal state associated data. In order to ensure the efficiency and reliability of feature extraction of various source data, this embodiment preferably uses a preset multimodal feature extraction model including a parallel preset neural network model, a long short-term memory network and a preset natural language processing model to extract features from multimodal data.

[0022] Specifically, the step of extracting data features from the multimodal state association data according to a preset multimodal feature extraction model to obtain corresponding multimodal state features includes: According to the preset neural network model, feature extraction is performed on the device image data to obtain corresponding image features; wherein the preset neural network model can be understood as a neural network capable of extracting image features, but in order to ensure the efficiency and reliability of image feature extraction, the present embodiment preferably adopts a convolutional neural network model (Convolutional Neural Networks, CNN) as the preset neural network model, and the specific structure and parameters of the convolutional neural network model can be adjusted and set through training optimization according to actual application requirements; Assume that the grid equipment image set obtained by image acquisition of the target grid equipment in the power system is , where n represents the total number of power grid devices in the power system, The device image data representing the i-th power grid device may describe the physical state of a transformer, relay, circuit breaker or other power distribution facility, and To pass the The feature map after layer convolutional network processing, Represents the total number of convolutional layers of the convolutional neural network model. The device image data is processed by the convolutional neural network model. The specific process of image feature extraction is as follows: Each convolutional layer includes a convolution module, an activation module, and a pooling module in sequence. For: 1) The convolution module performs convolution operations: using a set of convolution kernels For the input feature map Perform convolution operation (for , That is, the original image), the mathematical representation of convolution is: Among them, * represents the convolution operation, For the Tier The feature map corresponding to the convolution kernel; 2) The activation module performs the activation operation: activation function (such as ReLU activation function) Perform nonlinear transformation to increase the expressiveness of the model. The activation output of each convolution kernel is: 3) Pooling module performs pooling operation: Pooling (such as maximum pooling) is performed to reduce the feature dimension and increase the robustness of the model. The pooling result is: Then the above three modules are processed in sequence. Output of the layer It is composed of all the feature maps after pooling; the final image features can be obtained after processing each convolution layer in the convolutional neural network model .

[0023] According to the long short-term memory network, the device operation status time series data is feature extracted to obtain the corresponding state time series features; wherein, the long short-term memory network can be understood as directly using the existing LSTM (Long Short-Term Memory) network or an improved long short-term memory network constructed by training the LSTM (Long Short-Term Memory) network as the basic network, which only needs to be able to accurately and reliably extract the time series features of the voltage, current, fault code and other time-varying data collected during the operation of the power grid equipment, and is not specifically limited here; assuming that the collected device operation status time series data of the target power grid equipment in the power system is , where n represents the total number of power grid devices in the power system, Indicates Power grid equipment The equipment operation status time series data at each moment is extracted by using the LSTM network. The sequence in The processing process of the LSTM unit at time can be expressed as: 1) Forget gate, which determines which information will be discarded: 2) Input gate, which determines what new information will be stored in the cell state middle: 3) Update the cell state, combining the information of the forget gate and the input gate to update the cell state: 4) Output gate, determines the output value: in, represents the sigmoid function, represents element-wise multiplication, and are the LSTM network parameters.

[0024] The device operation log data is feature extracted according to the preset natural language processing model to obtain corresponding log text features; wherein, the preset natural language processing model can in principle adopt any model that can realize text data feature extraction. In order to ensure the efficiency and timeliness of the equipment health status assessment, this embodiment preferably adopts TF-IDF (term frequency–inverse document frequency) to reliably extract the log file features; assuming that the collected device operation log data set of the target power grid equipment in the power system is , where n represents the total number of power grid devices in the power system, Indicates The equipment operation log data of each power grid device is file data, so it is necessary to perform text preprocessing including word segmentation and removal of useless words, and then use TF-IDF to extract text features. Specifically, TF-IDF is used to extract specific words. In the log The weight calculation in can be expressed as: in, is the word frequency, It is the inverse document frequency, which calculates the importance of each word in all log documents; the required log text features can be obtained by combining the weights of each word or performing preset optimization processing on the combined vector according to actual application requirements.

[0025] Through the above method steps, data features can be extracted from the multimodal state correlation data of each target power grid device to be evaluated, and the corresponding multimodal state features to be analyzed can be obtained. ,in, , and Respectively represent image features, state time series features and log text features of each power grid device; in principle, by directly splicing and fusing the multimodal state features corresponding to each power grid device, an indicator that can characterize the health status of the power grid equipment can be obtained, but considering that there are certain differences in the descriptions of the operating status of the power grid equipment by different types of data features, direct fusion may result in the final health index not being in line with reality and having no specific application value. In order to further ensure that the health index constructed based on the fusion of multimodal data features is closer to reality, more accurate and reliable, this embodiment preferably introduces the fusion weight of multimodal features to control the importance of different features to the modeling of the power grid equipment state, and adopts the following recent historical health index evaluation deviation based on the target power grid equipment combined with the actual operating environment of the power grid equipment to dynamically optimize and update the optimal fusion weight of the multimodal features; for the specific method of obtaining the optimal fusion weight of the multimodal features used in each health status evaluation cycle, refer to the following steps.

[0026] S13, obtaining the historical cycle health index prediction error and the equipment operating environment data of the target power grid equipment, and dynamically obtaining the optimal fusion weight of the multimodal state feature to be analyzed based on the historical cycle health index prediction error and the equipment operating environment data; wherein the historical cycle health index prediction error can be understood as the prediction error generated by predicting the health state of the target power grid equipment using the same health state prediction method in the previous evaluation cycle; specifically, the step of obtaining the historical cycle health index prediction error of the target power grid equipment includes: According to the preset multimodal feature extraction model, feature extraction is performed on the multimodal state association data of the previous historical period of the target power grid equipment to obtain the corresponding historical period multimodal features; wherein, the specific extraction process of the historical period multimodal features can refer to the acquisition process of the multimodal state features to be analyzed in the previous text, which will not be repeated here; According to the fusion weight of the previous historical cycle, the multimodal features of the historical cycle are feature fused to obtain the corresponding historical cycle equipment health index prediction value; wherein, the fusion weight of the previous historical cycle can be understood as the multimodal feature fusion weight used for equipment health status prediction in the previous cycle of the current evaluation cycle. If the previous historical cycle is the first evaluation cycle, the corresponding fusion weight can be set according to actual application requirements according to human experience or other methods. If it corresponds to the second and subsequent evaluation cycles, the optimal fusion weight acquisition method given in this embodiment can be used to obtain and update the corresponding fusion weight, which will not be described in detail here; correspondingly, the historical cycle equipment health index prediction value can be understood as the fusion feature obtained by weighted fusion of the multimodal features of each historical cycle according to the fusion weight of the previous historical cycle; The true value of the equipment health index of the target power grid equipment in the previous historical period is obtained, and the corresponding historical period health index prediction error is obtained according to the true value of the equipment health index of the previous historical period and the predicted value of the equipment health index of the historical period.

[0027] In order to ensure the scientificity and accuracy of the multimodal feature fusion weight setting, this embodiment, on the basis of considering the preset error of the historical period, also introduces the equipment operating environment data for comprehensive analysis; the equipment operating environment data in this embodiment can be understood as other environmental factor data that will affect the equipment operating state, for example, external variables of the power grid equipment state and performance, such as humidity, temperature, whether there is vibration, electromagnetic interference and other data, can be set according to actual application requirements, and are not specifically limited here; specifically, the step of dynamically obtaining the optimal fusion weight of the multimodal state feature to be analyzed based on the historical period health index prediction error and the equipment operating environment data includes: According to the historical cycle health index prediction error and the equipment operating environment data, based on the genetic algorithm combined with reinforcement learning, the previous historical cycle fusion weight corresponding to the historical cycle health index prediction error is dynamically adjusted to obtain the optimal fusion weight; the fusion weight is expressed as: In the formula, in, and They represent the fusion weights of the i-th target power grid equipment in the t+1-th and t-th prediction cycles respectively; represents the historical cycle health index prediction error of the i-th target power grid equipment in the t-th prediction period; and They represent the predicted value of the equipment health index of the i-th target power grid equipment in the historical period in the t-th prediction period and the true value of the equipment health index in the previous historical period respectively; , and Respectively represent Image features, state time series features, and log text features of each power grid device; Represents the concatenation operation of features; is the dot product operation; represents the i-th power grid device , and The weight coefficients of the three features; Represents the equipment operating environment data of the i-th target power grid equipment; It represents a joint execution function composed of reinforcement learning and genetic algorithm. Preferably, the genetic algorithm is responsible for initially obtaining the optimization weights under the influence of environmental factors, and then the reinforcement learning algorithm is used to perform secondary optimization on the fusion weights obtained by the genetic algorithm, and finally the required optimal fusion weights are obtained through a preset number of cyclic optimizations.

[0028] Specifically, the step of dynamically adjusting the fusion weight of the previous historical period corresponding to the historical period health index prediction error based on the genetic algorithm combined with reinforcement learning to obtain the optimal fusion weight includes: According to the historical cycle health index prediction error and the equipment operating environment data, a corresponding fitness function is established, and the goal is to minimize the fitness function value. The fusion weight of the previous historical cycle is optimized by the genetic algorithm to obtain the initial optimized fusion weight; the fitness function is expressed as: in, represents the fitness function; represents the historical cycle health index prediction error of the i-th target power grid equipment in the t-th prediction period; Represents the equipment operating environment data of the i-th target power grid equipment; In practical applications, the genetic algorithm is used to optimize the fusion weights of the previous historical period, and the process of obtaining the initial optimized fusion weights is as follows: 1) Establish the optimization goal by minimizing the fitness function: in, Indicates that the fitness function value is the smallest when it corresponds to the i-th power grid device , and The weight coefficients of the three features; 2) Initialize and define feature weights The initial population of ; each individual represents a set of possible weight configurations; 3) Genetic algorithm stage: Select: Apply the fitness function to select individuals with higher fitness. High fitness means that the device health index under the weight configuration is closer to the target health index, and these individuals are more likely to be selected into the next generation. Crossover: Randomly select a pair of individuals for crossover operation, that is, exchange some of their genes (weights) at one or more points to generate new offspring. New weight configuration combinations can be introduced to increase the diversity of the search solution space. Mutation: Modify some genes (weights) of individuals with a certain probability to further increase the diversity of the population, which helps the algorithm to jump out of the local optimal solution and explore a wider solution space; Update and iteration: Repeat the above steps until a certain number of iterations or performance criteria are reached. At this time, a set of better weight configurations that have been preliminarily screened by the genetic algorithm will be obtained, that is, the corresponding initial optimized fusion weights will be obtained.

[0029] According to the historical cycle health index prediction error, a corresponding reward function is constructed, and the fitness function is used as the action value function, and the initial optimized fusion weight is optimized by reinforcement learning according to the reward function and the action value function to obtain the optimal fusion weight; wherein, reinforcement learning optimization can be understood as defining the performance of the reward function to quantify the weight configuration, and using Q learning to fine-tune the weight configuration based on reward feedback; this embodiment defines the reward function as the closer the predicted value of the device health index is to the true value of the device health index, the greater the value of the reward function, and vice versa, so that the optimization process will be more inclined to find those weight configurations that can make the predicted value of the device health index close to the true value of the device group health. Preferably, the reward function is defined as a negative exponential function of the opposite number of the difference between the two (historical cycle health index prediction error), expressed as: Where, e represents the natural index; represents the historical cycle health index prediction error of the i-th target power grid equipment in the t-th prediction period; represents the reward function, when The smaller, The closer it is to 1, the more negative the reward is, and then the greater the reward is, and vice versa. The reward function design adopted in this embodiment can make the predicted value of the equipment health index close to the true value of the equipment health index through the optimization process incentive model, that is, it tends to select those weight configurations that make the equipment operation status healthier.

[0030] Q-Learning is a well-known model-free learning algorithm in reinforcement learning. It aims to find the expected utility of taking an action in a given state, which is the so-called Q value (Quality value). The Q-learning algorithm does not need to model the dynamics of the environment. It directly learns the optimal strategy through interaction with the environment. In this embodiment, the weight adjustment amount is preferably integrated. Set as Q value, equipment operating environment data Set as the environmental data obtained by the agent, fusion weight The action that the agent needs to take, that is, combined with Q learning, weight The update rule is as follows. Time t is used to distinguish the current value from the value at the next moment. Define The calculation expression is the fitness function defined by the above genetic algorithm: in, and Respectively represent the fusion weights of the t+1th and tth prediction periods; represents the fusion weight adjustment required to update the fusion weight of the t-th prediction period; is the learning rate, the initial value is set to 0.00001; is the discount factor, set to 0.4; Represents the reward function with respect to weights The gradient of Represents the action value function value corresponding to the device operating environment data and fusion weight in the t-th prediction period.

[0031] The above steps for optimizing the fusion weights can actually be understood as continuously executing the genetic algorithm and the reinforcement learning algorithm, adjusting the weight configuration according to the historical cycle health index prediction error and the equipment operating environment data feedback until the algorithm converges (weight The weights used to predict the health status of the target power grid equipment in the current cycle are finally obtained. ; This embodiment adopts a weight dynamic optimization strategy that combines genetic algorithm and reinforcement learning. It can use genetic algorithm to search for reasonable weight configuration on a global scale, and then use reinforcement learning algorithm to perform fine-tuning and local optimization on this basis to achieve effective optimization of multimodal feature fusion weights. While providing reliable guarantee for obtaining more accurate equipment health status prediction results, it can also effectively improve the optimization efficiency of fusion weights and enhance the robustness and adaptability of the fusion weight dynamic optimization algorithm.

[0032] S14. According to the optimal fusion weight, the multimodal state features to be analyzed are subjected to feature fusion to obtain the corresponding equipment health index prediction value; wherein the equipment health index prediction value can be understood as a multimodal fusion feature obtained by weighted fusion of the multimodal state features to be analyzed in the current period based on the optimal fusion weight obtained through the above method steps. For specific calculation, please refer to the above description on The specific expression of is not repeated here.

[0033] This embodiment acquires multimodal state association data of the target power grid equipment including equipment image data, equipment operation state time series data and equipment operation log data in real time, extracts data features of the multimodal state association data according to a preset multimodal feature extraction model, obtains corresponding multimodal state features to be analyzed including image features, state time series features and log text features, acquires historical cycle health index prediction errors and equipment operation environment data of the target power grid equipment, dynamically acquires optimal fusion weights of the multimodal state features to be analyzed according to the historical cycle health index prediction errors and equipment operation environment data, and performs feature fusion on the multimodal state features to be analyzed according to the optimal fusion weights to obtain the corresponding equipment health index prediction value. The technical solution can capture rich equipment operation state features in real time by collecting multi-source data, and conducts real-time online scientific and reasonable evaluation of equipment health status in combination with the dynamic optimization strategy of multimodal feature fusion weights, which can effectively improve the timeliness, efficiency and accuracy of equipment health evaluation, facilitate timely perception and prevention of power grid equipment failures, and thus effectively reduce power production losses and maintenance costs.

[0034] At the same time, considering that the health status of each power grid device in the power system is directly related to the normal operation of the entire power system, this embodiment preferably further performs real-time and efficient operation status prediction and analysis of the entire power system on the basis of realizing reliable health status assessment of each power grid device through the above method steps, thereby providing reliable data support for the formulation of stable operation strategies of the power system; specifically, Figure 2 As shown, a method for predicting the health status of power grid equipment based on multimodal data fusion is provided, and the method also includes: S15. Obtain the historical health index sequence of the power system, and input the historical health index sequence of the power system and the equipment health index sequence of each target power grid device in the power system into a preset system health status prediction model to perform operation status evaluation, and obtain the corresponding system health index prediction value; wherein, the historical health index sequence of the power system can be understood as the historical time series data of the power system operation health index, and the specific acquisition method can refer to the corresponding existing technology implementation; correspondingly, the equipment health index sequence of each target power grid device can be understood as a data sequence composed of the corresponding equipment health index of each historical period (true value of the health index) and the equipment health index prediction value of the current period obtained by the aforementioned step S14, which are spliced ​​in time series. ,in, represents the predicted value of the equipment health index of the i-th target power grid equipment in the t-th prediction period; It represents the true value of the health index of the ith target power grid equipment in the tmth historical cycle, and m represents the time step (the total number of historical evaluation cycles).

[0035] In principle, the preset system health status prediction model in this embodiment can adopt any machine learning model that can make predictions based on time series data. However, in order to ensure the real-time and high efficiency of the power system health status assessment, this embodiment preferably adopts a long short-term memory (LSTM) network that integrates convolutional structure and time series analysis and is committed to continuous learning and updating, so as to quickly adapt to new data input and use the existing equipment health index data to make efficient and dynamic predictions of future system health status; assuming that the health data of N target devices with m time steps have been collected; for the system health index, an input feature vector can be defined Indicates time point Contains the health index sequence of all devices; the mathematical model of the LSTM unit used for system health status prediction is as follows: 1) Forget Gate: in, represents the sigmoid activation function, and represents weights and biases, It is the concatenation of the hidden state at the previous moment and the current input; 2) Input Gate: in, is a candidate cell state, and are the weights and biases of the input gate; 3) Cell status update: Here, multiplication operations are used to implement the process of forgetting and adding new information; 4) Output gate: in, is the hidden state at the current time point, and are the weight and bias of the output gate; In practical applications, the entire long short-term memory (LSTM) network is composed of multiple LSTM units mentioned above, each unit At the point in time Generate a hidden state , the final network output is the final system health index It is expressed as: in, and are the weights and biases of the output layer, is the last unit of the LSTM network at time point The hidden state of .

[0036] This embodiment acquires multimodal state association data of target power grid equipment in real time, extracts data features of the multimodal state association data according to a preset multimodal feature extraction model, obtains corresponding multimodal state features to be analyzed, and dynamically acquires the optimal fusion weight of the multimodal state features to be analyzed according to the historical cycle health index prediction error and equipment operation environment data of the acquired target power grid equipment, and performs feature fusion on the multimodal state features to be analyzed according to the optimal fusion weight to obtain the corresponding equipment health index prediction value, and inputs the preset system health state prediction model according to the acquired power system historical health index sequence and the equipment health index sequence of each target power grid equipment in the power system to perform operation state evaluation to obtain the corresponding system health index prediction value. The technical solution can not only effectively improve the timeliness, efficiency and accuracy of equipment health evaluation, facilitate timely perception and prevention of power grid equipment failures, and effectively reduce power production losses and maintenance costs, but also can perform real-time and efficient operation state prediction and analysis of the entire power system based on the health state prediction results of each power grid equipment, thereby providing reliable data support for the formulation of stable operation strategies of the power system.

[0037] In addition, considering that in actual applications, the operation and operating environment of power grid equipment will continue to change, in order to ensure the long-term effectiveness and robustness of the power grid equipment health status prediction model and the power system health status prediction model, and then ensure the prediction accuracy and efficiency of the power grid equipment health status and the power system health status in each evaluation cycle, this embodiment preferably adopts an online learning method to adaptively update and adjust the parameters of the health status evaluation and prediction model in real time to ensure that each model can continuously adapt and optimize its own prediction ability; specifically, Figure 3 As shown, a method for predicting the health status of power grid equipment based on multimodal data fusion is provided, and the method also includes: S16. Performing online learning and updating on the preset multimodal feature extraction model and the preset system health status prediction model according to a preset self-evolution strategy; The preset self-evolution strategy in this embodiment can be understood as an online learning strategy, which can automatically improve the operating performance of the preset multimodal feature extraction model and the preset system health status prediction model by updating the model in real time based on the new data gradually received during operation, so as to create an intelligent agent system that can quickly adapt to new situations and continuously improve. In the process of interacting with the environment, the intelligent agent can update its strategy in real time to adapt to changes in the environment; the specific process of online learning and updating based on the preset self-evolution strategy is as follows: At every point in time , assuming a corresponding set of model parameters Contains the weights and biases of all models involved in the evaluation and prediction process; as new data is received as input to each model in real time, the model parameters need to be constantly modified , so that each model can better adapt to new situations and improve prediction accuracy; Set up a self-evolution strategy Based on the current model parameters (the preset multimodal feature extraction model and the preset system health status prediction model involved in steps S11-S15) and the newly collected data (device image data of each target power grid device, device operation status time series data, device operation log data and device operation environment data, etc.), and a factor called "learning rate" (This factor can control the update rate (update amplitude) of the parameters. It determines the size of the step size during the gradient descent process. It must be just right. Too large a step may lead to unstable convergence, while too small a step may lead to slow convergence) to help update the parameters of the model; the specific formula is expressed as: In the formula, in, and Respectively Moment and The self-evolution strategy parameters at time +1 correspond to and +1 model parameters include all weights and biases of the preset multimodal feature extraction model and the preset system health status prediction model, that is, the update objects that need to be continuously updated and optimized based on the newly collected data to make the model more adaptable and have better prediction capabilities; Function uses the power system historical health index and current parameters to predict the system health status at the next moment and update the model parameters accordingly; Represents the loss function, which is used to measure the model In the data On performance, Relative to the parameter The loss gradient of It is at the moment The number of newly collected samples; and The model parameters are Next, corresponding to The predicted value of the system health index and the actual measured value of the system health index for samples.

[0038] In practical applications, suppose that the health status of an important power grid transformer is being monitored. The data collection device will periodically transmit the collected data back to the central system; when new data is received, In time When transmitted to the central system, the parameters of the evaluation and prediction models are updated as follows: 1) Use the current model and current parameters In the new data Perform forward calculations on the device to obtain the predicted value of the current device health status; 2) Calculate the loss function The value of , and use back propagation to calculate the gradient ; 3) Use the above Function update parameters: , and use the new parameters Replace the old parameters and update the model to the new version for continued use.

[0039] The embodiment of the present application obtains multimodal state association data of target power grid equipment in real time, extracts data features from the multimodal state association data according to a preset multimodal feature extraction model to obtain corresponding multimodal state features to be analyzed, and fuses the multimodal state features to be analyzed according to the optimal fusion weights of the multimodal state features to be analyzed dynamically obtained based on the historical cycle health index prediction error of the acquired target power grid equipment and the equipment operating environment data to obtain the corresponding equipment health index prediction value, then inputs the acquired power system historical health index sequence and the equipment health index sequence of each target power grid equipment in the power system into a preset system health state prediction model to perform operating state evaluation to obtain the corresponding system health index prediction value, and according to a preset self-evolution strategy, the predicted The technical solution of setting up a multimodal feature extraction model and a preset system health status prediction model for online learning and updating can not only effectively improve the timeliness, efficiency and accuracy of equipment health assessment, facilitate timely perception and prevention of power grid equipment failures, and effectively reduce power production losses and maintenance costs, but also can conduct real-time and efficient operation status prediction and analysis of the entire power system based on the prediction results of the health status of each power grid equipment, thereby providing reliable data support for the formulation of stable operation strategies of the power system, and can also ensure that each model can continuously adapt to new data and optimize its own prediction ability, ensure the long-term effectiveness and robustness of the model, thereby ensuring the accuracy and efficiency of the prediction of the health status of power grid equipment and the health status of the power system, and providing reliable guarantee for the stable and safe operation of the entire power system.

[0040] It should be noted that although the steps in the above flowchart are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders.

[0041] In one embodiment, Figure 4 As shown, a power grid equipment health status prediction system based on multimodal data fusion is provided, and the system includes: The state data acquisition module 1 is used to acquire multi-modal state-related data of the target power grid equipment in real time; the multi-modal state-related data includes equipment image data, equipment operation state time series data and equipment operation log data; The multimodal feature extraction module 2 is used to extract data features from the multimodal state association data according to a preset multimodal feature extraction model to obtain corresponding multimodal state features to be analyzed; the multimodal state features to be analyzed include image features, state time series features and log text features; The fusion weight acquisition module 3 is used to obtain the historical cycle health index prediction error and the equipment operating environment data of the target power grid equipment, and dynamically obtain the optimal fusion weight of the multimodal state feature to be analyzed based on the historical cycle health index prediction error and the equipment operating environment data; the historical cycle health index prediction error is the prediction error generated by the health index prediction of the target power grid equipment in the previous evaluation period; The equipment health assessment module 4 is used to perform feature fusion on the multi-modal state features to be analyzed according to the optimal fusion weight to obtain the corresponding equipment health index prediction value.

[0042] In one embodiment, Figure 5 As shown, a power grid equipment health status prediction system based on multimodal data fusion is provided, and the system also includes: The system health assessment module 5 is used to obtain the historical health index sequence of the power system, and input the historical health index sequence of the power system and the equipment health index sequence of each target power grid equipment in the power system into a preset system health status prediction model for operating status evaluation to obtain the corresponding system health index prediction value.

[0043] In one embodiment, Figure 6 As shown, a power grid equipment health status prediction system based on multimodal data fusion is provided, and the system also includes: The model online updating module 6 is used to perform online learning and updating on the preset multimodal feature extraction model and the preset system health status prediction model according to a preset self-evolution strategy.

[0044] For the specific limitations of the power grid equipment health status prediction system based on multimodal data fusion, please refer to the limitations of the power grid equipment health status prediction method based on multimodal data fusion mentioned above. The corresponding technical effects can also be equivalently obtained, which will not be repeated here. Each module in the above-mentioned power grid equipment health status prediction system based on multimodal data fusion can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0045] In summary, an embodiment of the present invention provides a method and system for predicting the health status of power grid equipment based on multimodal data fusion. The method for predicting the health status of power grid equipment based on multimodal data fusion realizes real-time acquisition of multimodal state associated data of target power grid equipment, extracts data features from the multimodal state associated data according to a preset multimodal feature extraction model to obtain corresponding multimodal state features to be analyzed, and performs feature fusion on the multimodal state features to be analyzed according to the optimal fusion weights of the multimodal state features to be analyzed dynamically obtained based on the historical cycle health index prediction error of the acquired target power grid equipment and the equipment operating environment data to obtain the corresponding equipment health index prediction value, and then inputs the acquired power system historical health index sequence and the equipment health index sequence of each target power grid equipment in the power system into the preset system health status prediction model to evaluate the operating status and obtain The corresponding system health index prediction value, as well as the technical solution for online learning and updating of the preset multimodal feature extraction model and the preset system health status prediction model according to the preset self-evolution strategy, this method can not only effectively improve the timeliness, efficiency and accuracy of equipment health assessment, facilitate timely perception and prevention of power grid equipment failures, and effectively reduce power production losses and maintenance costs, but also can perform real-time and efficient operation status prediction and analysis of the entire power system based on the health status prediction results of each power grid equipment, thereby providing reliable data support for the formulation of stable operation strategies of the power system, and can also ensure that each model can continuously adapt to new data and optimize its own prediction ability, ensure the long-term effectiveness and robustness of the model, thereby ensuring the accuracy and efficiency of the prediction of the health status of power grid equipment and the health status of the power system, and provide reliable guarantee for the stable and safe operation of the entire power system.

[0046] Each embodiment in this specification is described in a progressive manner, and the same or similar parts of each embodiment can be directly referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. It should be noted that the technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0047] The above-mentioned embodiments only express several preferred implementation modes of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in the technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be based on the protection scope of the claims.

Claims

1. A method for predicting the health status of power grid equipment based on multimodal data fusion, characterized in that: The method comprises the following steps: Acquire multimodal state-related data of target power grid equipment in real time; the multimodal state-related data includes equipment image data, equipment operation state time series data and equipment operation log data; According to a preset multimodal feature extraction model, data feature extraction is performed on the multimodal state association data to obtain corresponding multimodal state features to be analyzed; the multimodal state features to be analyzed include image features, state time series features and log text features; Obtaining the historical cycle health index prediction error and equipment operating environment data of the target power grid equipment, and dynamically obtaining the optimal fusion weight of the multimodal state feature to be analyzed based on the historical cycle health index prediction error and the equipment operating environment data; the historical cycle health index prediction error is the prediction error generated by predicting the health index of the target power grid equipment in the previous evaluation period; According to the optimal fusion weight, feature fusion is performed on the multimodal state features to be analyzed to obtain a corresponding device health index prediction value.

2. The method for predicting the health status of power grid equipment based on multimodal data fusion according to claim 1, characterized in that: The preset multimodal feature extraction model includes a parallel preset neural network model, a long short-term memory network and a preset natural language processing model; The step of extracting data features from the multimodal state association data according to a preset multimodal feature extraction model to obtain corresponding multimodal state features comprises: Extracting features from the device image data according to the preset neural network model to obtain corresponding image features; Extracting features from the equipment operation status time series data according to the long short-term memory network to obtain corresponding status time series features; Feature extraction is performed on the device operation log data according to the preset natural language processing model to obtain corresponding log text features.

3. The method for predicting the health status of power grid equipment based on multimodal data fusion according to claim 1, characterized in that: The step of obtaining the historical cycle health index prediction error of the target power grid equipment includes: According to the preset multimodal feature extraction model, feature extraction is performed on the multimodal state association data of the target power grid device in the previous historical period to obtain the corresponding multimodal features of the historical period; According to the fusion weight of the previous historical period, feature fusion is performed on the multimodal features of the historical period to obtain the corresponding historical period equipment health index prediction value; The true value of the equipment health index of the target power grid equipment in the previous historical period is obtained, and the corresponding historical period health index prediction error is obtained according to the true value of the equipment health index of the previous historical period and the predicted value of the equipment health index of the historical period.

4. The method for predicting the health status of power grid equipment based on multimodal data fusion according to claim 1, characterized in that: The step of dynamically acquiring the optimal fusion weight of the multimodal state feature to be analyzed according to the historical cycle health index prediction error and the equipment operating environment data comprises: According to the historical cycle health index prediction error and the equipment operating environment data, based on the genetic algorithm combined with reinforcement learning, the previous historical cycle fusion weight corresponding to the historical cycle health index prediction error is dynamically adjusted to obtain the optimal fusion weight; the fusion weight is expressed as: in, and They represent the fusion weights of the i-th target power grid equipment in the t+1-th and t-th prediction cycles respectively; represents the historical cycle health index prediction error of the i-th target power grid equipment in the t-th prediction period; Represents the equipment operating environment data of the i-th target power grid equipment; Represents the joint execution function consisting of reinforcement learning and genetic algorithm.

5. The method for predicting the health status of power grid equipment based on multimodal data fusion according to claim 4, characterized in that: The step of dynamically adjusting the fusion weight of the previous historical period corresponding to the historical period health index prediction error based on the genetic algorithm combined with reinforcement learning to obtain the optimal fusion weight includes: According to the historical cycle health index prediction error and the equipment operating environment data, a corresponding fitness function is established, and the goal is to minimize the fitness function value. The fusion weight of the previous historical cycle is optimized by the genetic algorithm to obtain the initial optimized fusion weight; the fitness function is expressed as: in, represents the fitness function; represents the historical cycle health index prediction error of the i-th target power grid equipment in the t-th prediction period; Represents the equipment operating environment data of the i-th target power grid equipment; According to the historical cycle health index prediction error, a corresponding reward function is constructed, and the fitness function is used as the action value function, and the initial optimized fusion weight is optimized by reinforcement learning according to the reward function and the action value function to obtain the optimal fusion weight; the reward function is expressed as: in, represents the reward function; e represents the natural exponent; It represents the historical cycle health index prediction error of the i-th target power grid equipment in the t-th prediction period.

6. The method for predicting the health status of power grid equipment based on multimodal data fusion according to claim 5, characterized in that: The step of performing reinforcement learning optimization on the initial optimized fusion weight according to the reward function and the action value function to obtain the optimal fusion weight comprises: The equipment operating environment data and fusion weight are used as the input state and output action of the agent respectively, and the fusion weight adjustment amount is used as the Q value. The initial optimized fusion weight is updated based on Q learning; the update rule of the initial optimized fusion weight is expressed as: In the formula, in, and Respectively represent the fusion weights of the t+1th and tth prediction periods; represents the fusion weight adjustment required to update the fusion weight of the t-th prediction period; represents the learning rate; is the discount factor; Represents the gradient of the reward function with respect to the fusion weight; Represents the action value function value corresponding to the device operating environment data and fusion weight in the t-th prediction period.

7. The method for predicting the health status of power grid equipment based on multimodal data fusion according to claim 1, characterized in that: The method further comprises: The historical health index sequence of the power system is obtained, and the historical health index sequence of the power system and the equipment health index sequence of each target power grid equipment in the power system are input into a preset system health status prediction model for operating status evaluation to obtain the corresponding system health index prediction value.

8. The method for predicting the health status of power grid equipment based on multimodal data fusion according to claim 7, characterized in that: The preset system health status prediction model is a long short-term memory network; the system health index prediction value is expressed as: in, express The system health index prediction value output at all times; and are the weights and biases of the LSTM output layer, is the last unit of the long short-term memory network at time point The hidden state of .

9. The method for predicting the health status of power grid equipment based on multimodal data fusion according to claim 7, characterized in that: The method further comprises: According to a preset self-evolution strategy, the preset multimodal feature extraction model and the preset system health status prediction model are updated through online learning.

10. A power grid equipment health status prediction system based on multimodal data fusion, characterized in that: The system comprises: A status data acquisition module is used to acquire multimodal status associated data of target power grid equipment in real time; the multimodal status associated data includes equipment image data, equipment operation status time series data and equipment operation log data; A multimodal feature extraction module, used to extract data features from the multimodal state association data according to a preset multimodal feature extraction model to obtain corresponding multimodal state features to be analyzed; the multimodal state features to be analyzed include image features, state time series features and log text features; A fusion weight acquisition module is used to obtain the historical cycle health index prediction error and equipment operating environment data of the target power grid equipment, and dynamically obtain the optimal fusion weight of the multimodal state feature to be analyzed based on the historical cycle health index prediction error and the equipment operating environment data; the historical cycle health index prediction error is the prediction error generated by the health index prediction of the target power grid equipment in the previous evaluation period; The equipment health assessment module is used to perform feature fusion on the multimodal state features to be analyzed according to the optimal fusion weight to obtain the corresponding equipment health index prediction value.

Images