Pumped storage power station maintenance strategy evaluation method and device, computer device, readable storage medium and program product

Abstract

The application relates to a pumped storage equipment maintenance strategy evaluation method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: performing maintenance on a pumped storage equipment to be maintained according to a maintenance strategy, obtaining a plurality of index data of the pumped storage equipment to be maintained after maintenance; screening key index data from the plurality of index data; performing feature extraction processing on the key index data to obtain a feature vector corresponding to the key index data; inputting the feature vector into a plurality of pre-trained maintenance strategy evaluation models respectively to obtain a strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model; screening a target maintenance strategy evaluation model from each maintenance strategy evaluation model; and performing fusion processing on the strategy evaluation result of the maintenance strategy output by each target maintenance strategy evaluation model to obtain a target strategy evaluation result of the maintenance strategy. The method can improve the accuracy of maintenance strategy evaluation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computers, and in particular to a pumped storage power station maintenance strategy evaluation method and device, computer equipment, a computer readable storage medium and a computer program product. BACKGROUND

[0002] In order to maintain the pumped storage power station and ensure stable operation of the pumped storage power station, the pumped storage power station needs to be regularly maintained through a maintenance strategy, and the actual effect of the maintenance strategy is evaluated.

[0003] In the prior art, when evaluating the actual effect of the maintenance strategy, manual evaluation is generally used. However, the manual evaluation method is subjective and can result in low accuracy of the evaluation of the pumped storage power station maintenance strategy. SUMMARY

[0004] Therefore, it is necessary to provide a pumped storage power station maintenance strategy evaluation method and device, computer equipment, a computer readable storage medium and a computer program product that can improve the accuracy of the evaluation of the pumped storage power station maintenance strategy.

[0005] In a first aspect, the present application provides a pumped storage power station maintenance strategy evaluation method, comprising:

[0006] According to the maintenance strategy of the pumped storage power station to be maintained, the pumped storage power station to be maintained is maintained, and a plurality of index data of the pumped storage power station to be maintained after maintenance is obtained;

[0007] From the plurality of index data, key index data is selected;

[0008] The key index data is subjected to feature extraction processing to obtain a feature vector corresponding to the key index data;

[0009] The feature vector is input into a plurality of pre-trained maintenance strategy evaluation models, respectively, to obtain a strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model and a prediction probability corresponding to the strategy evaluation result;

[0010] From each maintenance strategy evaluation model, a target maintenance strategy evaluation model whose prediction probability corresponding to the strategy evaluation result of the maintenance strategy output is greater than a preset probability is selected;

[0011] According to the model weight of each target maintenance strategy evaluation model, the strategy evaluation result of the maintenance strategy output by each target maintenance strategy evaluation model is subjected to fusion processing to obtain a target strategy evaluation result of the maintenance strategy.

[0012] In one of the embodiments, the screening of the key indicator data from the plurality of indicator data comprises:

[0013] Each indicator data is respectively input into a pre-trained first importance prediction model to obtain a first importance corresponding to each indicator data;

[0014] Each indicator data is respectively input into a pre-trained second importance prediction model to obtain a second importance corresponding to each indicator data;

[0015] The first importance and the second importance corresponding to each indicator data are respectively fused to obtain a target importance corresponding to each indicator data;

[0016] The indicator data corresponding to the target importance greater than a preset importance is screened from each indicator data as the key indicator data.

[0017] In one of the embodiments, the feature extraction processing of the key indicator data to obtain a feature vector corresponding to the key indicator data comprises:

[0018] The data type of the key indicator data is identified;

[0019] According to the data type of the key indicator data, a corresponding relationship between the data type and a feature extraction model is queried to obtain a feature extraction model corresponding to the key indicator data;

[0020] The key indicator data is input into the corresponding feature extraction model for feature extraction processing to obtain a first feature vector corresponding to the key indicator data;

[0021] The first feature vector is subjected to re-feature extraction processing to obtain a second feature vector corresponding to the key indicator data;

[0022] The first feature vector and the second feature vector are fused to obtain the feature vector of the key indicator data.

[0023] In one of the embodiments, the input of the feature vector into a plurality of pre-trained maintenance strategy evaluation models to obtain the strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model and the prediction probability corresponding to the strategy evaluation result comprises:

[0024] The feature vector is respectively input into a plurality of pre-trained maintenance strategy evaluation models to obtain the prediction probability of the maintenance strategy under various preset evaluation results output by each maintenance strategy evaluation model;

[0025] The predicted probability of the maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results is corrected to obtain a corrected predicted probability of the maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results.

[0026] From various preset evaluation results of the maintenance strategy output by each maintenance strategy evaluation model, the preset evaluation result with the maximum corrected predicted probability is selected as the strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model.

[0027] In one of the embodiments, each pre-trained maintenance strategy evaluation model is trained in the following manner:

[0028] According to a sample maintenance strategy of a sample pumped storage power station, the sample pumped storage power station is maintained to obtain a plurality of sample index data of the sample pumped storage power station after maintenance;

[0029] From the plurality of sample index data, key sample index data is selected;

[0030] The key sample index data is subjected to feature extraction processing to obtain a sample feature vector corresponding to the key sample index data;

[0031] The sample feature vector is input into a maintenance strategy evaluation model to be trained to obtain a predicted strategy evaluation result of the sample maintenance strategy;

[0032] An actual strategy evaluation result of the sample maintenance strategy is obtained, and the maintenance strategy evaluation model to be trained is iteratively trained according to the difference between the predicted strategy evaluation result and the actual strategy evaluation result of the sample maintenance strategy to obtain the pre-trained maintenance strategy evaluation model.

[0033] In one of the embodiments, after the strategy evaluation result output by each target maintenance strategy evaluation model is fused according to the model weight of each target maintenance strategy evaluation model to obtain a target strategy evaluation result of the maintenance strategy, the method further includes:

[0034] According to the target strategy evaluation result, a corresponding relationship between a strategy evaluation result and a strategy evaluation level is queried to obtain a strategy evaluation level corresponding to the target strategy evaluation result;

[0035] A strategy update instruction corresponding to the strategy evaluation level is obtained;

[0036] According to the strategy update instruction, the maintenance strategy is updated to obtain an updated maintenance strategy.

[0037] In a second aspect, the application further provides a pumped storage equipment maintenance strategy evaluation device, comprising:

[0038] A data acquisition module is configured to perform maintenance on the pumped storage equipment to be maintained according to a maintenance strategy of the pumped storage equipment to be maintained, and obtain a plurality of index data of the pumped storage equipment to be maintained after maintenance.

[0039] A data screening module is configured to screen key index data from the plurality of index data.

[0040] A feature extraction module is configured to perform feature extraction processing on the key index data to obtain a feature vector corresponding to the key index data.

[0041] A strategy evaluation module is configured to input the feature vector into a plurality of pre-trained maintenance strategy evaluation models respectively, obtain a strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model, and a prediction probability corresponding to the strategy evaluation result.

[0042] A model screening module is configured to screen a target maintenance strategy evaluation model from the plurality of maintenance strategy evaluation models, wherein the prediction probability corresponding to the strategy evaluation result of the maintenance strategy output by the target maintenance strategy evaluation model is greater than a preset probability.

[0043] A result determination module is configured to perform fusion processing on the strategy evaluation result of the maintenance strategy output by each target maintenance strategy evaluation model according to a model weight of each target maintenance strategy evaluation model, and obtain a target strategy evaluation result of the maintenance strategy.

[0044] In a third aspect, the application further provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0045] Perform maintenance on the pumped storage equipment to be maintained according to a maintenance strategy of the pumped storage equipment to be maintained, and obtain a plurality of index data of the pumped storage equipment to be maintained after maintenance.

[0046] Screen key index data from the plurality of index data.

[0047] Perform feature extraction processing on the key index data to obtain a feature vector corresponding to the key index data.

[0048] Input the feature vector into a plurality of pre-trained maintenance strategy evaluation models respectively, obtain a strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model, and a prediction probability corresponding to the strategy evaluation result.

[0049] filter, from the each maintenance strategy evaluation model, a target maintenance strategy evaluation model with a predicted probability corresponding to a strategy evaluation result of the maintenance strategy output by the target maintenance strategy evaluation model being greater than a preset probability;

[0050] fuse, according to a model weight of each target maintenance strategy evaluation model, the strategy evaluation result of the maintenance strategy output by the each target maintenance strategy evaluation model to obtain a target strategy evaluation result of the maintenance strategy.

[0051] In a fourth aspect, the present application also provides a computer readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the following steps:

[0052] maintain the to-be-maintained pumped storage equipment according to a maintenance strategy of the to-be-maintained pumped storage equipment to obtain a plurality of index data of the to-be-maintained pumped storage equipment after maintenance;

[0053] filter, from the plurality of index data, key index data;

[0054] perform feature extraction processing on the key index data to obtain a feature vector corresponding to the key index data;

[0055] input the feature vector into a plurality of pre-trained maintenance strategy evaluation models respectively to obtain a strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model and a predicted probability corresponding to the strategy evaluation result;

[0056] filter, from the each maintenance strategy evaluation model, a target maintenance strategy evaluation model with a predicted probability corresponding to a strategy evaluation result of the maintenance strategy output by the target maintenance strategy evaluation model being greater than a preset probability;

[0057] fuse, according to a model weight of each target maintenance strategy evaluation model, the strategy evaluation result of the maintenance strategy output by the each target maintenance strategy evaluation model to obtain a target strategy evaluation result of the maintenance strategy.

[0058] In a fifth aspect, the present application also provides a computer program product comprising a computer program, the computer program being executed by a processor to implement the following steps:

[0059] maintain the to-be-maintained pumped storage equipment according to a maintenance strategy of the to-be-maintained pumped storage equipment to obtain a plurality of index data of the to-be-maintained pumped storage equipment after maintenance;

[0060] filter, from the plurality of index data, key index data;

[0061] perform feature extraction processing on the key index data to obtain a feature vector corresponding to the key index data;

[0062] input the feature vectors into a plurality of pre-trained maintenance strategy evaluation models respectively, to obtain a strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model, and a prediction probability corresponding to the strategy evaluation result;

[0063] from each maintenance strategy evaluation model, a target maintenance strategy evaluation model is screened out, for which the prediction probability corresponding to the strategy evaluation result of the maintenance strategy output is greater than a preset probability;

[0064] the strategy evaluation results of the maintenance strategy output by each target maintenance strategy evaluation model are fused according to the model weights of each target maintenance strategy evaluation model, to obtain a target strategy evaluation result of the maintenance strategy.

[0065] The above pump storage device maintenance strategy evaluation method, device, computer equipment, computer readable storage medium and computer program product, first, according to the maintenance strategy of the pump storage device to be maintained, the pump storage device to be maintained is maintained, a plurality of index data of the pump storage device to be maintained after maintenance is obtained, then from the plurality of index data, the key index data is screened out, and the key index data is processed by feature extraction, to obtain the feature vector corresponding to the key index data, then the feature vectors are input into a plurality of pre-trained maintenance strategy evaluation models respectively, to obtain a strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model, and a prediction probability corresponding to the strategy evaluation result, and from each maintenance strategy evaluation model, a target maintenance strategy evaluation model is screened out, for which the prediction probability corresponding to the strategy evaluation result of the maintenance strategy output is greater than a preset probability, and finally, the strategy evaluation results of the maintenance strategy output by each target maintenance strategy evaluation model are fused according to the model weights of each target maintenance strategy evaluation model, to obtain a target strategy evaluation result of the maintenance strategy. In this way, the key index data of the pump storage device to be maintained after maintenance is comprehensively considered, and a plurality of pre-trained maintenance strategy evaluation models are considered, and the strategy evaluation results of the maintenance strategy output by each target maintenance strategy evaluation model are fused according to the model weights of each target maintenance strategy evaluation model, so that the target strategy evaluation result obtained finally is more accurate, thereby improving the evaluation accuracy of the pump storage device maintenance strategy, and avoiding the defect that the evaluation accuracy of the pump storage device maintenance strategy is low due to manual evaluation of the maintenance strategy. BRIEF DESCRIPTION OF DRAWINGS

[0066] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed to be used in the description of the embodiments of the present application or the related art will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0067] Figure 1 is a flowchart of a pumped storage power station maintenance strategy evaluation method in one embodiment;

[0068] Figure 2 is a flowchart of a step of screening out key index data in one embodiment;

[0069] Figure 3 is a flowchart of a pumped storage power station maintenance strategy evaluation method in another embodiment;

[0070] Figure 4 is a structural block diagram of a pumped storage power station maintenance strategy evaluation device in one embodiment;

[0071] Figure 5 is an internal structure diagram of a computer device in one embodiment. DETAILED DESCRIPTION

[0072] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0073] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0074] Figure 1 is a flowchart of a pumped storage power station maintenance strategy evaluation method according to an exemplary embodiment, as shown in Figure 1As shown, the pumped storage power station maintenance strategy evaluation method is applied to a server. It can be understood that the method can also be applied to a terminal, and can also be applied to a system including a terminal and a server, and is realized through interaction between the terminal and the server. The terminal can be, but is not limited to, various personal computers, notebook computers, smart phones and tablet computers; the server can be a stand-alone physical server, or a server cluster or distributed system formed by multiple physical servers, or a cloud server providing cloud computing services. In the example embodiment, the method includes the following steps S101 to S106. Among them:

[0075] In step S101, the pumped storage power station to be maintained is maintained according to the maintenance strategy of the pumped storage power station to be maintained, and a plurality of index data of the pumped storage power station to be maintained after maintenance are obtained.

[0076] Among them, the pumped storage power station to be maintained refers to the pumped storage power station that needs to be maintained.

[0077] Among them, the maintenance strategy refers to the strategy for maintaining the pumped storage power station to be maintained, and specifically includes maintenance type, maintenance item, maintenance time, maintenance period, maintenance requirement, etc.

[0078] Among them, the index data of the pumped storage power station to be maintained after maintenance refers to the index data related to the pumped storage power station to be maintained collected after the maintenance of the pumped storage power station to be maintained by using the maintenance strategy of the pumped storage power station to be maintained, which is used to measure the actual effect of the implementation of the maintenance strategy, and specifically includes reliability index, maintenance cost, equipment state, accident event, etc. The reliability index includes equivalent availability coefficient, forced outage rate, unit startup success rate, unplanned outage time, etc.; the maintenance cost includes ten-thousand-yuan asset maintenance fee, single-unit multi-year average maintenance and maintenance fee, etc.; the equipment state includes running state, performance index, health degree adjustment information, etc.; the accident event includes accident event level, accident event quantity, etc.

[0079] Illustratively, the server generates a maintenance instruction of the pumped storage power station to be maintained according to the maintenance strategy of the pumped storage power station to be maintained, and controls the maintenance equipment to maintain the pumped storage power station to be maintained according to the maintenance instruction of the pumped storage power station to be maintained, and collects a plurality of index data of the pumped storage power station to be maintained after maintenance.

[0080] For example, the server sends the maintenance instruction of the pumped storage power station to be maintained to the maintenance equipment, the maintenance equipment maintains the pumped storage power station to be maintained according to the maintenance instruction, and collects a plurality of index data of the pumped storage power station to be maintained after maintenance after the maintenance of the pumped storage power station to be maintained is completed, and sends the plurality of index data to the server.

[0081] Step S102, screening key indicator data from the plurality of indicator data.

[0082] The key indicator data refers to indicator data corresponding to an importance greater than a preset importance, and specifically refers to some important indicator data.

[0083] For example, the server respectively inputs each indicator data into a pre-trained importance prediction model (such as a neural network model, a deep learning model, etc.), to obtain an importance corresponding to each indicator data; and screens, from the plurality of indicator data, indicator data corresponding to an importance greater than a preset importance, as key indicator data.

[0084] Step S103, performing feature extraction processing on the key indicator data to obtain a feature vector corresponding to the key indicator data.

[0085] The feature vector is used to represent the feature information of the key indicator data.

[0086] For example, the server pre-processes the key indicator data to obtain pre-processed key indicator data, and inputs the pre-processed key indicator data into a pre-trained feature extraction model to perform feature extraction processing, to obtain a feature vector as a feature vector corresponding to the key indicator data.

[0087] Step S104, respectively inputting the feature vectors into a plurality of pre-trained maintenance strategy evaluation models to obtain a strategy evaluation result of a maintenance strategy output by each maintenance strategy evaluation model, and a prediction probability corresponding to the strategy evaluation result.

[0088] The maintenance strategy evaluation model is a model for outputting a strategy evaluation result and a prediction probability of a maintenance strategy, such as a large language model, a neural network model, a deep learning model, etc.

[0089] Different pre-trained maintenance strategy evaluation models are obtained by different model training, such as a maintenance strategy evaluation model A obtained by a large language model, a maintenance strategy evaluation model B obtained by a neural network model, and a maintenance strategy evaluation model C obtained by a deep learning model.

[0090] The strategy evaluation result refers to a predicted evaluation result of the maintenance strategy, such as an evaluation score, an evaluation level, etc.

[0091] The prediction probability is used to measure the possibility corresponding to the strategy evaluation result, such as 0.7, 0.8, etc.

[0092] Exemplarily, the server respectively inputs the feature vector into a plurality of pre-trained maintenance strategy evaluation models, performs prediction processing on the feature vector by each maintenance strategy evaluation model, and obtains a prediction probability of each maintenance strategy evaluation model outputted maintenance strategy under various preset evaluation results; filters out the preset evaluation result with the maximum prediction probability from the various preset evaluation results of the maintenance strategy outputted by each maintenance strategy evaluation model as the strategy evaluation result of the maintenance strategy outputted by each maintenance strategy evaluation model, and takes the prediction probability corresponding to the preset evaluation result with the maximum prediction probability as the prediction probability corresponding to the strategy evaluation result of the maintenance strategy outputted by each maintenance strategy evaluation model.

[0093] For example, assuming that the prediction probabilities of the maintenance strategy outputted by the maintenance strategy evaluation model under the preset evaluation results a, b and c are 0.5, 0.3 and 0.2 respectively, the strategy evaluation result of the maintenance strategy outputted by the maintenance strategy evaluation model is the preset evaluation result a, and the prediction probability corresponding to the strategy evaluation result is 0.5.

[0094] Step S105, filtering out the target maintenance strategy evaluation model with the prediction probability corresponding to the strategy evaluation result of the outputted maintenance strategy greater than the preset probability from each maintenance strategy evaluation model.

[0095] The preset probability refers to a probability value set in advance, such as 0.5, 0.6, etc.

[0096] The target maintenance strategy evaluation model refers to the maintenance strategy evaluation model with the prediction probability corresponding to the strategy evaluation result of the outputted maintenance strategy greater than the preset probability.

[0097] Exemplarily, the server filters out the maintenance strategy evaluation model with the prediction probability corresponding to the strategy evaluation result of the outputted maintenance strategy greater than the preset probability from each maintenance strategy evaluation model as the target maintenance strategy evaluation model.

[0098] For example, assuming that the prediction probability corresponding to the strategy evaluation result a of the maintenance strategy outputted by the maintenance strategy evaluation model A is 0.9, the prediction probability corresponding to the strategy evaluation result b of the maintenance strategy outputted by the maintenance strategy evaluation model B is 0.8, the prediction probability corresponding to the strategy evaluation result b of the maintenance strategy outputted by the maintenance strategy evaluation model C is 0.5, the prediction probability corresponding to the strategy evaluation result a of the maintenance strategy outputted by the maintenance strategy evaluation model D is 0.7, and the preset probability is 0.6, the filtered target maintenance strategy evaluation models are the maintenance strategy evaluation model A, the maintenance strategy evaluation model B and the maintenance strategy evaluation model D.

[0099] Step S106, according to the model weight of each target maintenance strategy evaluation model, the strategy evaluation result of the maintenance strategy output by each target maintenance strategy evaluation model is fused to obtain the target strategy evaluation result of the maintenance strategy.

[0100] Wherein, the model weight is used to measure the importance of the strategy evaluation result of the maintenance strategy output by the target maintenance strategy evaluation model, such as 0.7, 0.6, etc., which is determined according to the prediction accuracy of the target maintenance strategy evaluation model.

[0101] Wherein, the fusion processing refers to weighted summation processing.

[0102] Wherein, the target strategy evaluation result refers to the final strategy evaluation result, such as evaluation score, evaluation level, etc., which is used to measure the actual effect of the implementation of the maintenance strategy, so as to facilitate the subsequent iteration and improvement of the maintenance strategy. For example, the actual effect of the implementation of the maintenance strategy is evaluated by reliability index, maintenance cost, equipment state, etc., the reliability is improved, and the maintenance strategy is iterated and improved.

[0103] Exemplarily, the server queries the corresponding relationship between the prediction accuracy and the model weight according to the prediction accuracy of each target maintenance strategy evaluation model to obtain the model weight of each target maintenance strategy evaluation model; then, according to the model weight of each target maintenance strategy evaluation model, the strategy evaluation result of the maintenance strategy output by each target maintenance strategy evaluation model is fused to obtain the fusion evaluation result as the target strategy evaluation result of the maintenance strategy.

[0104] For example, assuming that the target maintenance strategy evaluation models are maintenance strategy evaluation model A, maintenance strategy evaluation model B and maintenance strategy evaluation model D, and the corresponding model weights are A1, B1 and D1 respectively, then the target strategy evaluation result of the maintenance strategy = the strategy evaluation result of the maintenance strategy output by the maintenance strategy evaluation model A x A1 + the strategy evaluation result of the maintenance strategy output by the maintenance strategy evaluation model B x B1 + the strategy evaluation result of the maintenance strategy output by the maintenance strategy evaluation model D x D1.

[0105] In the above pump storage equipment maintenance strategy evaluation method, first, the pump storage equipment to be maintained is maintained according to the maintenance strategy of the pump storage equipment to be maintained, to obtain a plurality of index data of the pump storage equipment to be maintained after maintenance. Then, key index data is screened from the plurality of index data, and feature extraction processing is performed on the key index data to obtain a feature vector corresponding to the key index data. Then, the feature vector is respectively input into a plurality of pre-trained maintenance strategy evaluation models to obtain a strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model and a prediction probability corresponding to the strategy evaluation result. Then, a target maintenance strategy evaluation model is screened from each maintenance strategy evaluation model, in which the prediction probability corresponding to the strategy evaluation result of the maintenance strategy output by the target maintenance strategy evaluation model is greater than a preset probability. Finally, the strategy evaluation result of the maintenance strategy output by each target maintenance strategy evaluation model is fused according to the model weight of each target maintenance strategy evaluation model to obtain a target strategy evaluation result of the maintenance strategy. In this way, the key index data of the pump storage equipment to be maintained after maintenance and the plurality of pre-trained maintenance strategy evaluation models are comprehensively considered, and the strategy evaluation result of the maintenance strategy output by each target maintenance strategy evaluation model is fused according to the model weight of each target maintenance strategy evaluation model, so that the target strategy evaluation result obtained finally is more accurate, thereby improving the evaluation accuracy of the pump storage equipment maintenance strategy and avoiding the defect that the evaluation accuracy of the pump storage equipment maintenance strategy is low due to manual evaluation of the maintenance strategy. In one exemplary embodiment, as shown in FIG. 1, the step S102 of screening the key index data from the plurality of index data includes the following steps S201 to S204.

[0106] Figure 2 In one exemplary embodiment, as shown in FIG. 1, the step S102 of screening the key index data from the plurality of index data includes the following steps S201 to S204. Wherein:

[0107] Step S201: input each index data into a pre-trained first importance prediction model respectively to obtain a first importance corresponding to each index data.

[0108] Step S202: input each index data into a pre-trained second importance prediction model respectively to obtain a second importance corresponding to each index data.

[0109] Step S203: fuse the first importance and the second importance corresponding to each index data respectively to obtain a target importance corresponding to each index data.

[0110] Step S204: screen the index data corresponding to the target importance greater than a preset importance from each index data as the key index data.

[0111] ​The first importance, the second importance and the target importance are all used to measure the importance of the index data. The preset importance is a pre-set importance, such as 0.6.

[0112] The first importance prediction model is used to output the first importance corresponding to the index data, and can be a neural network model, a deep learning model, etc. The second importance prediction model is used to output the second importance corresponding to the index data, and can be a neural network model, a deep learning model, etc. It should be noted that the first importance prediction model and the second importance prediction model refer to different importance prediction models, such as the first importance prediction model is a neural network model, and the second importance prediction model is a deep learning model.

[0113] The key index data refers to the index data corresponding to the target importance greater than the preset importance.

[0114] Exemplarily, the server respectively inputs each index data into the pre-trained first importance prediction model for importance prediction to obtain the first importance corresponding to each index data; respectively inputs each index data into the pre-trained second importance prediction model for importance prediction to obtain the second importance corresponding to each index data, then respectively fuses the first importance and the second importance corresponding to each index data to obtain the fusion importance corresponding to each index data as the target importance corresponding to each index data, such as target importance = first importance + second importance, or target importance = first importance * weight of first importance + second importance * weight of second importance; finally, from each index data, the index data corresponding to the target importance greater than the preset importance is screened out as the key index data.

[0115] In this embodiment, the key index data is screened out from the plurality of index data, which is beneficial to accurately predict the strategy evaluation result of the maintenance strategy based on the key index data, avoids the influence of redundant data on the strategy evaluation result, and thus improves the evaluation accuracy of the pumped storage power station equipment maintenance strategy.

[0116] In an example embodiment, the step S103 of performing feature extraction processing on the key indicator data to obtain the feature vector corresponding to the key indicator data specifically includes the following contents: identifying the data type of the key indicator data; querying the corresponding relationship between the data type and the feature extraction model according to the data type of the key indicator data to obtain the feature extraction model corresponding to the key indicator data; inputting the key indicator data into the corresponding feature extraction model to perform feature extraction processing to obtain the first feature vector corresponding to the key indicator data; performing re-feature extraction processing on the first feature vector to obtain the second feature vector corresponding to the key indicator data; and performing fusion processing on the first feature vector and the second feature vector to obtain the feature vector of the key indicator data.

[0117] wherein different key indicator data corresponds to different data types.

[0118] wherein the feature extraction model can be a convolutional neural network model, a deep learning model, etc.

[0119] wherein the first feature vector refers to the feature vector corresponding to the shallow feature of the key indicator data; and the second feature vector refers to the feature vector corresponding to the deep feature of the key indicator data.

[0120] wherein the feature vector of the key indicator data is obtained through fusion processing of the first feature vector and the second feature vector.

[0121] For example, the server identifies the data type of the key indicator data according to the data type identification instruction; queries the corresponding relationship between the data type and the feature extraction model according to the data type of the key indicator data to obtain the feature extraction model corresponding to the data type of the key indicator data as the feature extraction model corresponding to the key indicator data; then inputs the key indicator data into the corresponding feature extraction model to perform feature extraction processing to obtain the first feature vector corresponding to the key indicator data; performs re-feature extraction processing on the first feature vector through the feature extraction model to obtain the second feature vector corresponding to the key indicator data; and finally performs fusion processing on the first feature vector and the second feature vector to obtain the fusion feature vector as the feature vector of the key indicator data.

[0122] In this embodiment, the feature extraction model corresponding to the key indicator data is determined first, and then the key indicator data is input into the corresponding feature extraction model to perform multiple times of feature extraction processing, which is beneficial to improve the determination accuracy of the feature vector of the key indicator data.

[0123] In an exemplary embodiment, the step S104 of inputting the feature vector into the plurality of pre-trained maintenance strategy evaluation models respectively to obtain the strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model and the prediction probability corresponding to the strategy evaluation result comprises the following contents: inputting the feature vector into the plurality of pre-trained maintenance strategy evaluation models respectively to obtain the prediction probability of the maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results; correcting the prediction probability of the maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results to obtain the corrected prediction probability of the maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results; and selecting the preset evaluation result with the maximum corrected prediction probability from the various preset evaluation results of the maintenance strategy output by each maintenance strategy evaluation model as the strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model.

[0124] The preset evaluation result can be a preset evaluation score a, a preset evaluation score b, a preset evaluation score c, etc., or can be a preset evaluation grade X1, a preset evaluation grade X2, a preset evaluation grade X3, a preset evaluation grade X4, etc.

[0125] Each maintenance strategy evaluation model can output the prediction probability of the maintenance strategy under various preset evaluation results, such as (0.1, 0.2, 0.3, 0.4), which respectively represent the prediction probability of the maintenance strategy under the preset evaluation grade X1, the preset evaluation grade X2, the preset evaluation grade X3, and the preset evaluation grade X4.

[0126] The correction of the prediction probability of the maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results is mainly to adjust the prediction probability to avoid too large deviation between the output prediction probability and the actual situation, such as adjusting the prediction probability (0.01, 0.02, 0.97) to (0.1, 0.1, 0.8).

[0127] Exemplarily, the server respectively inputs the feature vector into a plurality of pre-trained maintenance strategy evaluation models, performs prediction processing on the feature vector by each maintenance strategy evaluation model, and obtains a prediction probability of a maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results; then, the prediction probability of the maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results is corrected by a pre-trained probability correction model (such as a deep learning model), to obtain a corrected prediction probability of the maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results; finally, the preset evaluation result with the maximum corrected prediction probability is selected from various preset evaluation results of the maintenance strategy output by each maintenance strategy evaluation model as a strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model, and the corrected prediction probability corresponding to the preset evaluation result with the maximum corrected prediction probability is taken as a prediction probability corresponding to the strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model.

[0128] For example, assuming that the prediction probabilities of a maintenance strategy output by a certain maintenance strategy evaluation model under preset evaluation results a, b and c are 0.01, 0.02 and 0.97 respectively, and after correction, the prediction probabilities of the maintenance strategy output by the maintenance strategy evaluation model under the preset evaluation results a, b and c are 0.1, 0.1 and 0.8 respectively, then the strategy evaluation result of the maintenance strategy output by the maintenance strategy evaluation model is the preset evaluation result c, and the prediction probability corresponding to the strategy evaluation result (i.e., the preset evaluation result c) is 0.8.

[0129] In this embodiment, the prediction probability of the maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results is output, the prediction probability is corrected, and then the strategy evaluation result and the prediction probability of the maintenance strategy output by each maintenance strategy evaluation model are determined based on the corrected prediction probability, which is beneficial to improve the accuracy of the strategy evaluation result and the prediction probability of the maintenance strategy output by each maintenance strategy evaluation model, and further improves the accuracy of the finally obtained target strategy evaluation result.

[0130] In one exemplary embodiment, the pumped storage power station maintenance strategy evaluation method provided by the application further comprises a training step of the maintenance strategy evaluation model, specifically comprising the following contents: according to the sample maintenance strategy of the sample pumped storage power station, the sample pumped storage power station is maintained to obtain a plurality of sample index data of the sample pumped storage power station after maintenance; from the plurality of sample index data, the key sample index data is screened out; the key sample index data is subjected to feature extraction processing to obtain a sample feature vector corresponding to the key sample index data; the sample feature vector is input into the maintenance strategy evaluation model to be trained to obtain a predicted strategy evaluation result of the sample maintenance strategy; the actual strategy evaluation result of the sample maintenance strategy is obtained, and according to the difference between the predicted strategy evaluation result and the actual strategy evaluation result of the sample maintenance strategy, the maintenance strategy evaluation model to be trained is iteratively trained to obtain a pre-trained maintenance strategy evaluation model.

[0131] Among them, the maintenance strategy evaluation model to be trained can be a large language model, a neural network model, a deep learning model, etc.

[0132] Among them, each pre-trained maintenance strategy evaluation model is obtained by separate training.

[0133] Exemplarily, the server generates sample maintenance instructions of the sample pumped storage power station according to the sample maintenance strategy of the sample pumped storage power station, and controls the maintenance equipment to maintain the sample pumped storage power station according to the sample maintenance instructions of the sample pumped storage power station, and collects a plurality of sample index data of the sample pumped storage power station after maintenance; then, each sample index data is input into the pre-trained importance prediction model to obtain an importance corresponding to each sample index data; from the plurality of sample index data, the sample index data with an importance greater than a preset importance is screened out as the key sample index data; then, the key sample index data is preprocessed to obtain preprocessed key sample index data, and the preprocessed key sample index data is input into the pre-trained feature extraction model for feature extraction processing to obtain a feature vector as a sample feature vector corresponding to the key sample index data; then, the sample feature vector is input into the maintenance strategy evaluation model to be trained to obtain a predicted strategy evaluation result of the sample maintenance strategy; finally, the actual strategy evaluation result of the sample maintenance strategy is obtained, and according to the difference between the predicted strategy evaluation result and the actual strategy evaluation result of the sample maintenance strategy, a loss value is calculated, and the maintenance strategy evaluation model to be trained is iteratively trained according to the loss value until the loss value obtained by the predicted strategy evaluation result output by the trained maintenance strategy evaluation model is less than a preset threshold, then the training is stopped, and the trained maintenance strategy evaluation model is used as the pre-trained maintenance strategy evaluation model.

[0134] In this embodiment, by iteratively training the to-be-trained maintenance strategy evaluation model, the accuracy of the strategy evaluation result output by the maintenance strategy evaluation model can be improved, thereby improving the accuracy of the target strategy evaluation result obtained finally, and further improving the evaluation accuracy of the pumped storage power station equipment maintenance strategy.

[0135] In one exemplary embodiment, after the step S106, the strategy evaluation result output by each target maintenance strategy evaluation model is fused according to the model weight of each target maintenance strategy evaluation model to obtain the target strategy evaluation result of the maintenance strategy, the following contents are further included: querying the corresponding relationship between the strategy evaluation result and the strategy evaluation level according to the target strategy evaluation result to obtain the strategy evaluation level corresponding to the target strategy evaluation result; obtaining the strategy update instruction corresponding to the strategy evaluation level; and updating the maintenance strategy according to the strategy update instruction to obtain the updated maintenance strategy.

[0136] The strategy evaluation level can be high, medium, and low, or first grade, second grade, third grade, fourth grade, etc.

[0137] Different strategy evaluation levels correspond to different strategy update instructions.

[0138] The strategy update instruction is mainly used to update the maintenance strategy, such as iteratively improving the maintenance strategy to update the maintenance type, maintenance item, maintenance time, maintenance period, and maintenance requirement in the maintenance strategy.

[0139] For example, the server queries the corresponding relationship between the strategy evaluation result and the strategy evaluation level according to the target strategy evaluation result to obtain the strategy evaluation level corresponding to the target strategy evaluation result; then, the server queries the corresponding relationship between the strategy evaluation level and the strategy update instruction according to the strategy evaluation level to obtain the strategy update instruction corresponding to the strategy evaluation level; finally, the server performs a series of update processes on the maintenance strategy according to the strategy update instruction to obtain the updated maintenance strategy.

[0140] For example, the server improves the reliability from the aspects of design improvement, manufacturing improvement, and maintenance improvement according to the maintenance evaluation result (such as the target strategy evaluation result), and optimizes the maintenance process and decision-making mode to improve the maintenance strategy, including continuously optimizing the maintenance type, maintenance item, maintenance period, and maintenance requirement.

[0141] In this embodiment, the corresponding strategy update instruction is determined according to the strategy evaluation level corresponding to the target strategy evaluation result, and the maintenance strategy is updated according to the strategy update instruction, which is conducive to achieving the purpose of updating the maintenance strategy, so that the updated maintenance strategy is more suitable for the corresponding pumped storage power station equipment, thereby improving the implementation effect of the maintenance strategy.

[0142] In one exemplary embodiment, as shown in Figure 3 Another pump storage equipment maintenance strategy evaluation method is provided, which is applied to a server as an example and includes the following steps S301-S312. Among them:

[0143] Step S301, according to the maintenance strategy of the pump storage equipment to be maintained, the pump storage equipment to be maintained is maintained, and a plurality of index data of the pump storage equipment to be maintained after maintenance are obtained.

[0144] Step S302, each index data is input into a pre-trained first importance prediction model to obtain a first importance corresponding to each index data; and each index data is input into a pre-trained second importance prediction model to obtain a second importance corresponding to each index data.

[0145] Step S303, the first importance and the second importance corresponding to each index data are fused to obtain a target importance corresponding to each index data; and from each index data, an index data with a target importance greater than a preset importance is selected as a key index data.

[0146] Step S304, the data type of the key index data is identified; and according to the data type of the key index data, a corresponding relationship between the data type and a feature extraction model is queried to obtain a feature extraction model corresponding to the key index data.

[0147] Step S305, the key index data is input into the corresponding feature extraction model for feature extraction processing to obtain a first feature vector corresponding to the key index data; and the first feature vector is subjected to re-feature extraction processing to obtain a second feature vector corresponding to the key index data.

[0148] Step S306, the first feature vector and the second feature vector are fused to obtain a feature vector of the key index data.

[0149] Step S307, the feature vector is input into a plurality of pre-trained maintenance strategy evaluation models respectively to obtain a prediction probability of a maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results.

[0150] Step S308, the prediction probability of the maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results is corrected to obtain a corrected prediction probability of the maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results.

[0151] Step S309, from each maintenance strategy evaluation model output of the maintenance strategy of each preset evaluation result, the largest prediction probability of the modified preset evaluation result is selected as the strategy evaluation result of the maintenance strategy evaluation model output, and the prediction probability corresponding to the strategy evaluation result is obtained.

[0152] Step S310, from each maintenance strategy evaluation model, the prediction probability corresponding to the strategy evaluation result of the output maintenance strategy is greater than the preset probability.

[0153] Step S311, according to the model weight of each target maintenance strategy evaluation model, the strategy evaluation result of the maintenance strategy output by each target maintenance strategy evaluation model is fused to obtain the target strategy evaluation result of the maintenance strategy.

[0154] Step S312, according to the target strategy evaluation result, the corresponding relationship between the strategy evaluation result and the strategy evaluation level is queried to obtain the strategy evaluation level corresponding to the target strategy evaluation result; the strategy update instruction corresponding to the strategy evaluation level is obtained; the maintenance strategy is updated according to the strategy update instruction to obtain the updated maintenance strategy.

[0155] In the above pumped storage equipment maintenance strategy evaluation method, the key index data of the pumped storage equipment to be maintained after maintenance is considered comprehensively, and a plurality of pre-trained maintenance strategy evaluation models are considered, and the strategy evaluation result of the maintenance strategy output by each target maintenance strategy evaluation model is fused according to the model weight of each target maintenance strategy evaluation model, which can make the target strategy evaluation result obtained finally more accurate, thereby improving the evaluation accuracy of the pumped storage equipment maintenance strategy, and avoiding the defects that the evaluation accuracy of the pumped storage equipment maintenance strategy is low due to manual evaluation of the maintenance strategy.

[0156] In one exemplary embodiment, in order to more clearly illustrate the pumped storage equipment maintenance strategy evaluation method provided by the embodiments of the present application, the pumped storage equipment maintenance strategy evaluation method will be specifically described below with one specific embodiment. In one exemplary embodiment, the present application also provides another pumped storage equipment maintenance strategy evaluation method to evaluate the actual effect of the maintenance strategy implementation according to the reliability index, and iteratively improve the maintenance strategy, which specifically includes the following contents:

[0157] (1) According to the maintenance strategy, the pumped storage equipment to be maintained is maintained, and a plurality of index data of the pumped storage equipment after maintenance are obtained, the index data including reliability index, accident event, maintenance cost and equipment state; the reliability index includes equivalent availability coefficient, forced outage rate, unit startup success rate and unplanned outage time; the accident event includes accident event level and accident event quantity; the maintenance cost includes ten-thousand-yuan asset maintenance fee and single-unit multi-year average maintenance and repair fee; and the equipment state includes running state, performance index and health degree adjustment. For example, a maintenance evaluation task is generated according to user demand, and unit reliability data and key index data after maintenance are automatically calculated according to a maintenance evaluation algorithm.

[0158] (2) The index data are subjected to feature extraction processing, and index features are obtained.

[0159] (3) The index data are input into a plurality of maintenance strategy evaluation models, and maintenance strategy evaluation results output by each maintenance strategy evaluation model and corresponding prediction probabilities are obtained.

[0160] (4) From each maintenance strategy evaluation model, a target maintenance strategy evaluation model with a prediction probability greater than a preset probability is selected.

[0161] (5) According to the model weight of each target maintenance strategy evaluation model, the maintenance strategy evaluation results output by each target maintenance strategy evaluation model are subjected to comprehensive processing, and a target evaluation result of the maintenance strategy is obtained.

[0162] (6) According to the target evaluation result of the maintenance strategy, the reliability of the maintenance strategy is improved, and the maintenance strategy is iteratively improved. For example, according to the maintenance evaluation result, the reliability is improved from aspects of design improvement, manufacturing improvement and maintenance improvement, and the maintenance strategy is improved by optimizing the maintenance process and decision-making mode, including continuous optimization of maintenance type, maintenance project, maintenance cycle and requirements.

[0163] The above embodiment is advantageous in improving the evaluation accuracy of the pumped storage equipment maintenance strategy, and avoids the defect that the evaluation accuracy of the pumped storage equipment maintenance strategy is low due to manual evaluation of the maintenance strategy.

[0164] It should be understood that although the steps in the flowcharts involved in the embodiments described above are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least some of the steps in the flowcharts involved in the embodiments described above can include multiple steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be alternately or alternately executed with at least part of other steps or steps or stages in other steps.

[0165] Based on the same inventive concept, the embodiments of the present application also provide a pumped storage power station maintenance strategy evaluation device for implementing the above-mentioned pumped storage power station maintenance strategy evaluation method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more pumped storage power station maintenance strategy evaluation device embodiments provided below can refer to the limitations of the pumped storage power station maintenance strategy evaluation method described above, which will not be repeated here.

[0166] In one exemplary embodiment, as shown in Figure 4 A pumped storage power station maintenance strategy evaluation device is provided, comprising: a data acquisition module 410, a data screening module 420, a data screening module 430, a strategy evaluation module 440, a model screening module 450, and a result determination module 460, wherein:

[0167] The data acquisition module 410 is configured to perform maintenance on the pumped storage power station to be maintained according to the maintenance strategy of the pumped storage power station to be maintained, and obtain a plurality of index data of the pumped storage power station to be maintained after maintenance.

[0168] The data screening module 420 is configured to screen key index data from the plurality of index data.

[0169] The feature extraction module 430 is configured to perform feature extraction processing on the key index data to obtain a feature vector corresponding to the key index data.

[0170] The strategy evaluation module 440 is configured to input the feature vector into a plurality of pre-trained maintenance strategy evaluation models respectively, and obtain a strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model and a prediction probability corresponding to the strategy evaluation result.

[0171] The model screening module 450 is configured to screen, from each of the maintenance strategy evaluation models, a target maintenance strategy evaluation model whose predicted probability of the strategy evaluation result of the output maintenance strategy is greater than a preset probability.

[0172] The result determination module 460 is configured to fuse the strategy evaluation result of the output maintenance strategy of each target maintenance strategy evaluation model according to the model weight of each target maintenance strategy evaluation model to obtain a target strategy evaluation result of the maintenance strategy.

[0173] In an example embodiment, the data screening module 420 is further configured to input each index data into a pre-trained first importance prediction model respectively to obtain a first importance corresponding to each index data; input each index data into a pre-trained second importance prediction model respectively to obtain a second importance corresponding to each index data; fuse the first importance and the second importance corresponding to each index data respectively to obtain a target importance corresponding to each index data; and screen, from each index data, an index data whose target importance is greater than a preset importance as a key index data.

[0174] In an example embodiment, the feature extraction module 430 is further configured to identify a data type of the key index data; query a corresponding relationship between the data type and a feature extraction model according to the data type of the key index data to obtain a feature extraction model corresponding to the key index data; input the key index data into the corresponding feature extraction model for feature extraction processing to obtain a first feature vector corresponding to the key index data; perform re-feature extraction processing on the first feature vector to obtain a second feature vector corresponding to the key index data; and fuse the first feature vector and the second feature vector to obtain a feature vector of the key index data.

[0175] In an example embodiment, the strategy evaluation module 440 is further configured to input the feature vector into a plurality of pre-trained maintenance strategy evaluation models respectively to obtain predicted probabilities of the output maintenance strategy of each maintenance strategy evaluation model under various preset evaluation results; correct the predicted probabilities of the output maintenance strategy of each maintenance strategy evaluation model under the various preset evaluation results to obtain corrected predicted probabilities of the output maintenance strategy of each maintenance strategy evaluation model under the various preset evaluation results; and screen, from the various preset evaluation results of the output maintenance strategy of each maintenance strategy evaluation model, a preset evaluation result with the maximum corrected predicted probability as the strategy evaluation result of the output maintenance strategy of each maintenance strategy evaluation model.

[0176] In an example embodiment, the pumped storage power station maintenance strategy evaluation device further comprises a model training device configured to: perform maintenance on the sample pumped storage power station according to the sample maintenance strategy of the sample pumped storage power station, to obtain a plurality of sample index data of the sample pumped storage power station after maintenance; screen key sample index data from the plurality of sample index data; perform feature extraction processing on the key sample index data to obtain a sample feature vector corresponding to the key sample index data; input the sample feature vector into the to-be-trained maintenance strategy evaluation model to obtain a predicted strategy evaluation result of the sample maintenance strategy; obtain an actual strategy evaluation result of the sample maintenance strategy, and perform iterative training on the to-be-trained maintenance strategy evaluation model according to a difference between the predicted strategy evaluation result and the actual strategy evaluation result of the sample maintenance strategy, to obtain the pre-trained maintenance strategy evaluation model.

[0177] In an example embodiment, the pumped storage power station maintenance strategy evaluation device further comprises a strategy updating module configured to: obtain a strategy evaluation grade corresponding to the target strategy evaluation result according to the target strategy evaluation result, by querying a corresponding relationship between strategy evaluation results and strategy evaluation grades; obtain a strategy updating instruction corresponding to the strategy evaluation grade; and update the maintenance strategy according to the strategy updating instruction to obtain an updated maintenance strategy.

[0178] The above-mentioned modules of the pumped storage power station maintenance strategy evaluation device can be realized by software, hardware, or a combination thereof. The above-mentioned modules can be embedded in or independent of a processor in a computer device in hardware form, or stored in a memory in a computer device in software form, so as to be called and executed by a processor to perform operations corresponding to the above-mentioned modules.

[0179] In an example embodiment, a computer device is provided, which can be a server, and an internal structure diagram of the computer device can be as shown in Figure 5As shown in the figure. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through the system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capability. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The database of the computer device is used to store index data, key index data and other data. The input / output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used to communicate with the terminal outside through the network connection. The computer program is executed by the processor to realize a pumped storage power station maintenance strategy evaluation method.

[0180] Those skilled in the art can understand that, Figure 5 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0181] In an exemplary embodiment, a computer device is also provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to realize the steps in each of the above method embodiments.

[0182] In an exemplary embodiment, a computer readable storage medium is provided, having a computer program stored thereon, the computer program being executed by a processor to realize the steps in each of the above method embodiments.

[0183] In an exemplary embodiment, a computer program product is provided, including a computer program, the computer program being executed by a processor to realize the steps in each of the above method embodiments.

[0184] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. The non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. The volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, the RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., without being limited thereto.

[0185] The technical features of the above embodiments can be combined arbitrarily. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combinations of the technical features do not exist contradictory, they should be considered as the scope of the present application.

[0186] The above-described embodiments are merely illustrative of several embodiments of the present application, and the description is relatively specific and detailed, but should not be understood as a limitation on the scope of the patent. It should be noted that for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of the present application. Therefore, the scope of protection of the present application should be subject to the appended claims.

Claims

1. A method for evaluating maintenance strategies for pumped storage hydroelectric power plants, characterized in that, The method includes: According to the maintenance strategy of the pumped storage equipment to be inspected, the pumped storage equipment to be inspected is inspected and the equipment is inspected and the multiple indicator data of the pumped storage equipment after inspection are obtained; the maintenance strategy includes maintenance type, maintenance items, maintenance time, maintenance cycle and maintenance requirements; the indicator data includes reliability indicators, maintenance cost, equipment status and accident events. From the multiple indicator data, key indicator data is selected; The key indicator data is subjected to feature extraction processing to obtain the feature vector corresponding to the key indicator data; The feature vector is input into multiple pre-trained maintenance strategy evaluation models to obtain the strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model, and the prediction probability corresponding to the strategy evaluation result. From each maintenance strategy evaluation model, select the target maintenance strategy evaluation model whose predicted probability is greater than the preset probability corresponding to the output strategy evaluation result of the maintenance strategy. Based on the model weights of each target maintenance strategy evaluation model, the strategy evaluation results of the maintenance strategy output by each target maintenance strategy evaluation model are fused to obtain the target strategy evaluation result of the maintenance strategy. The step of performing feature extraction processing on the key indicator data to obtain the feature vector corresponding to the key indicator data includes: Identify the data type of the key indicator data; Based on the data type of the key indicator data, query the correspondence between the data type and the feature extraction model to obtain the feature extraction model corresponding to the key indicator data; The key indicator data is input into the corresponding feature extraction model for feature extraction processing to obtain the first feature vector corresponding to the key indicator data; the first feature vector refers to the feature vector corresponding to the shallow features of the key indicator data. The first feature vector is subjected to further feature extraction processing to obtain the second feature vector corresponding to the key indicator data; the second feature vector refers to the feature vector corresponding to the deep features of the key indicator data. The first feature vector and the second feature vector are fused to obtain the feature vector of the key indicator data. The step of inputting the feature vector into multiple pre-trained maintenance strategy evaluation models to obtain the strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model, and the predicted probability corresponding to the strategy evaluation result, includes: The feature vectors are input into multiple pre-trained maintenance strategy evaluation models to obtain the predicted probability of the maintenance strategy under various preset evaluation results output by each maintenance strategy evaluation model. The predicted probability of the maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results is corrected to obtain the corrected predicted probability of the maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results. From the various preset evaluation results of the maintenance strategy output by each maintenance strategy evaluation model, the preset evaluation result with the highest predicted probability after correction is selected as the strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model.

2. The method according to claim 1, characterized in that, The step of filtering key indicator data from the multiple indicator data includes: Each indicator data is input into a pre-trained first importance prediction model to obtain the first importance corresponding to each indicator data. Each indicator data is input into a pre-trained second importance prediction model to obtain the second importance corresponding to each indicator data; the first importance prediction model and the second importance prediction model refer to different importance prediction models; The first importance and the second importance corresponding to each indicator data are fused to obtain the target importance corresponding to each indicator data. From each of the indicator data, the indicator data with a target importance greater than a preset importance is selected as the key indicator data.

3. The method according to claim 1, characterized in that, Each pre-trained maintenance strategy evaluation model is trained in the following manner: According to the sample maintenance strategy of the sample pumped storage equipment, the sample pumped storage equipment is maintained, and multiple sample index data of the sample pumped storage equipment after maintenance are obtained. From the multiple sample indicator data, key sample indicator data are selected; The key sample indicator data is subjected to feature extraction processing to obtain the sample feature vector corresponding to the key sample indicator data. The sample feature vector is input into the maintenance strategy evaluation model to be trained to obtain the predicted strategy evaluation result of the sample maintenance strategy. Obtain the actual strategy evaluation results of the sample maintenance strategy, and based on the difference between the predicted strategy evaluation results and the actual strategy evaluation results of the sample maintenance strategy, iteratively train the maintenance strategy evaluation model to be trained to obtain the pre-trained maintenance strategy evaluation model.

4. The method according to any one of claims 1 to 3, characterized in that, After fusing the strategy evaluation results output by each target maintenance strategy evaluation model according to the model weights of each target maintenance strategy evaluation model to obtain the target strategy evaluation result of the maintenance strategy, the process further includes: Based on the target strategy evaluation results, the correspondence between the strategy evaluation results and the strategy evaluation levels is queried to obtain the strategy evaluation level corresponding to the target strategy evaluation results; Obtain the policy update instruction corresponding to the policy evaluation level; The maintenance strategy is updated according to the strategy update instruction to obtain the updated maintenance strategy.

5. A device for evaluating maintenance strategies for pumped storage hydroelectric power equipment, characterized in that, The device includes: The data acquisition module is used to perform maintenance on the pumped storage equipment to be maintained according to the maintenance strategy, and obtain multiple indicator data of the pumped storage equipment after maintenance; the maintenance strategy includes maintenance type, maintenance items, maintenance time, maintenance cycle, and maintenance requirements; the indicator data includes reliability indicators, maintenance costs, equipment status, and accident events. The data filtering module is used to filter out key indicator data from the multiple indicator data. The feature extraction module is used to perform feature extraction processing on the key indicator data to obtain the feature vector corresponding to the key indicator data. The strategy evaluation module is used to input the feature vector into multiple pre-trained maintenance strategy evaluation models respectively, and obtain the strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model, as well as the prediction probability corresponding to the strategy evaluation result. The model screening module is used to screen out target maintenance strategy evaluation models from each maintenance strategy evaluation model whose predicted probability corresponding to the output strategy evaluation result of the maintenance strategy is greater than a preset probability. The result determination module is used to fuse the strategy evaluation results of the maintenance strategy output by each target maintenance strategy evaluation model according to the model weight of each target maintenance strategy evaluation model, so as to obtain the target strategy evaluation result of the maintenance strategy. The feature extraction module is further configured to: identify the data type of the key indicator data; query the correspondence between the data type and the feature extraction model based on the data type of the key indicator data to obtain the feature extraction model corresponding to the key indicator data; input the key indicator data into the corresponding feature extraction model for feature extraction processing to obtain a first feature vector corresponding to the key indicator data; perform further feature extraction processing on the first feature vector to obtain a second feature vector corresponding to the key indicator data; and fuse the first feature vector and the second feature vector to obtain a feature vector of the key indicator data. The first feature vector refers to the feature vector corresponding to the shallow features of the key indicator data; the second feature vector refers to the feature vector corresponding to the deep features of the key indicator data. The strategy evaluation module is further configured to input the feature vector into multiple pre-trained maintenance strategy evaluation models to obtain the predicted probability of the maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results; correct the predicted probability of the maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results to obtain the corrected predicted probability of the maintenance strategy output by each maintenance strategy evaluation model under various preset evaluation results; and select the preset evaluation result with the highest corrected predicted probability from the various preset evaluation results of the maintenance strategy output by each maintenance strategy evaluation model as the strategy evaluation result of the maintenance strategy output by each maintenance strategy evaluation model.

6. The apparatus according to claim 5, characterized in that, The data filtering module is further configured to: input each indicator data into a pre-trained first importance prediction model to obtain the first importance of each indicator data; input each indicator data into a pre-trained second importance prediction model to obtain the second importance of each indicator data; fuse the first and second importance of each indicator data to obtain the target importance of each indicator data; and filter out indicator data whose target importance is greater than a preset importance from each indicator data as the key indicator data; wherein the first importance prediction model and the second importance prediction model refer to different importance prediction models.

7. The apparatus according to claim 5, characterized in that, The device further includes a model training module, used to perform maintenance on the sample pumped storage equipment according to the sample maintenance strategy of the sample pumped storage equipment, and obtain multiple sample index data after the maintenance of the sample pumped storage equipment; select key sample index data from the multiple sample index data; perform feature extraction processing on the key sample index data to obtain the sample feature vector corresponding to the key sample index data; and input the sample feature vector into the maintenance strategy evaluation model to be trained to obtain the prediction strategy evaluation result of the sample maintenance strategy. Obtain the actual strategy evaluation results of the sample maintenance strategy, and based on the difference between the predicted strategy evaluation results and the actual strategy evaluation results of the sample maintenance strategy, iteratively train the maintenance strategy evaluation model to be trained to obtain the pre-trained maintenance strategy evaluation model.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 4.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images