Electrochemical energy storage system health state assessment method and device, computer equipment, medium and product

Abstract

The invention relates to a health state assessment method and device of an electrochemical energy storage system, computer equipment, a medium and a product. The method comprises the steps of determining target evaluation indexes according to state parameters of the electrochemical energy storage system, obtaining sample data corresponding to each target evaluation index, obtaining a sample data set, training a convolutional neural network model based on the sample data set, obtaining a candidate weak predictor set, and obtaining a candidate weak predictor set; and screening from the candidate weak predictor set to obtain a target weak predictor, combining all the target weak predictors to obtain a target strong predictor, and performing health state prediction on the electrochemical energy storage system through the target strong predictor to obtain a numerical health state prediction value. By adopting the method, the complex relationship between the state parameters and the health state of the electrochemical energy storage system can be more accurately captured, so that the health state of the energy storage system can be accurately evaluated.

Description

Technical Field

[0001] The present application relates to the technical field of reliability assessment, and in particular to a method, device, computer equipment, medium and product for assessing the health status of an electrochemical energy storage system. Background Art

[0002] Electrochemical energy storage systems are complex in structure and are often used in important occasions. Currently, the reliability of electrochemical energy storage systems mainly depends on the sensors and monitoring systems installed therein, such as BMS (battery management system) and EMS (energy management system). Although these systems work continuously, most of them only issue alarms when the system fails, and the data collected in other cases cannot be fully utilized. Combining the characteristics of the energy storage system and reliability theory analysis, the system usually has signs before a failure occurs. Therefore, based on the changing characteristics of the system data, the health status of the electrochemical energy storage system can be effectively predicted, and preventive measures can be formulated in a timely manner to ensure the safety of the system.

[0003] In traditional methods, the safety assessment and status assessment of battery energy storage systems usually start from the common faults of electrochemical energy storage power stations, find system fault-related variables and influencing factors, and then determine the system fault monitoring indicators. According to different fault types and areas, appropriate monitoring and diagnosis methods are selected to minimize the harm caused by system failures.

[0004] However, this assessment method usually focuses on the fault analysis level, or only conducts health status assessment for a specific problem. Since the reliability and health status of the electrochemical energy storage system involve multiple types of components, the sensitive stress and failure mechanism of each component may be different. It is difficult to accurately assess the health status of the entire system through a single fault analysis model. Summary of the invention

[0005] Based on this, it is necessary to provide a health status assessment method, device, computer equipment, medium and product for an electrochemical energy storage system that can accurately assess the health status of the system in response to the above technical problems.

[0006] In a first aspect, the present application provides a method for predicting the health status of an electrochemical energy storage system, comprising:

[0007] According to the state parameters of the electrochemical energy storage system, the target evaluation index is determined, and the sample data corresponding to each target evaluation index is obtained to obtain a sample data set;

[0008] The convolutional neural network model is trained based on the sample data set to obtain a set of candidate weak predictors;

[0009] A target weak predictor is obtained by screening from a set of candidate weak predictors, and all target weak predictors are combined to obtain a target strong predictor;

[0010] The health status of the electrochemical energy storage system is predicted through the target strong predictor to obtain a numerical health status prediction value.

[0011] In one embodiment, the step of determining the target evaluation index according to the state parameters of the electrochemical energy storage system includes:

[0012] Determining candidate indicators that affect the health state of the electrochemical energy storage system according to the state parameters of the electrochemical energy storage system;

[0013] Perform factor analysis on the candidate indicators to obtain the principal component eigenvalues ​​corresponding to each candidate indicator;

[0014] The candidate indicators corresponding to the principal component eigenvalues ​​within the preset range are used as target evaluation indicators; the target evaluation indicators include equilibrium state, battery capacity, discharge depth, charging rate, charge and discharge cycle life, total battery voltage, total battery current, battery ambient temperature, ambient humidity, and main circuit state.

[0015] In one embodiment, the step of training a convolutional neural network model based on a sample data set to obtain a set of candidate weak predictors includes:

[0016] Select a preset group of sample data from the sample data set as training data;

[0017] Determine the network structure of the convolutional neural network model according to the input and output dimensions of the training data, and initialize the model parameters of the convolutional neural network model;

[0018] In the process of training the convolutional neural network model for multiple rounds based on the training data, the model parameters corresponding to each round of training are obtained;

[0019] The convolutional neural network model corresponding to the model parameters is used as the weak predictor obtained in the corresponding round, and the weak predictors obtained in all rounds are used as a set of candidate weak predictors.

[0020] In one embodiment, the process of performing multiple rounds of training on a convolutional neural network model based on training data includes:

[0021] Input the training data into the convolutional neural network model for the current round of training, obtain the output result of the current round, and obtain the prediction error of the current round according to the output result;

[0022] The weight of the training data is adjusted according to the prediction error, and the adjusted training data is input into the convolutional neural network model for the next round of training.

[0023] In one embodiment, the step of adjusting the weight of the training data according to the prediction error includes:

[0024] The proportion by which the weight of the training data is increased when the prediction error is greater than the error threshold.

[0025] In one embodiment, the step of screening a target weak predictor from a set of candidate weak predictors and combining all the target weak predictors to obtain a target strong predictor comprises:

[0026] The weak predictor of the round corresponding to the prediction error that meets the preset conditions is used as the target weak predictor;

[0027] The weight of the target weak predictor is determined according to the output error corresponding to the target weak predictor, and the target weak predictors are weighted combined according to the weight to obtain the target strong predictor.

[0028] In a second aspect, the present application also provides a health status prediction device for an electrochemical energy storage system, comprising:

[0029] An indicator determination module is used to determine the target evaluation indicator according to the state parameters of the electrochemical energy storage system, and obtain sample data corresponding to each target evaluation indicator to obtain a sample data set;

[0030] A model training module is used to train the convolutional neural network model based on the sample data set to obtain a set of candidate weak predictors;

[0031] A predictor combination module, used for screening a target weak predictor from a set of candidate weak predictors, and combining all the target weak predictors to obtain a target strong predictor;

[0032] The state prediction module is used to predict the health state of the electrochemical energy storage system through a target strong predictor to obtain a numerical health state prediction value.

[0033] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method steps of any one of the first aspects are implemented.

[0034] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements any one of the method steps in the first aspect when the computer program is executed by a processor.

[0035] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, which implements any one of the method steps in the first aspect when executed by a processor.

[0036] The above-mentioned electrochemical energy storage system health status assessment method, device, computer equipment, medium and product integrate the advantages of multiple weak predictors to form a target strong predictor. Since the target strong predictor can adapt to different types of electrochemical energy storage systems and changes in various operating conditions, it can more accurately capture the complex relationship between the state parameters and the health status of the electrochemical energy storage system, thereby accurately assessing the health status of the energy storage system. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0038] FIG1 is an application environment diagram of a method for evaluating the health status of an electrochemical energy storage system in one embodiment;

[0039] FIG2 is a schematic flow chart of a method for evaluating the health status of an electrochemical energy storage system in one embodiment;

[0040] FIG3 is a comparison diagram of prediction errors of a weak predictor and a strong predictor in one embodiment;

[0041] FIG4 is a schematic flow chart of a method for evaluating the health status of an electrochemical energy storage system in another embodiment;

[0042] FIG5 is a structural block diagram of a device for evaluating the health status of an electrochemical energy storage system in one embodiment;

[0043] FIG. 6 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0045] The method for predicting the health status of an electrochemical energy storage system provided in an embodiment of the present application can be applied in an application environment as shown in FIG. 1. The terminal 102 communicates with the server 104 through a network. The data storage system can store data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on a cloud or other network server. The terminal 102 is used to obtain the state parameters of the electrochemical energy storage system through the server 104, determine the target evaluation index according to the state parameters of the electrochemical energy storage system, and obtain the sample data corresponding to each target evaluation index to obtain a sample data set, train the convolutional neural network model based on the sample data set, obtain a candidate weak predictor set, screen the target weak predictor from the candidate weak predictor set, and combine all the target weak predictors to obtain a target strong predictor, and predict the health status of the electrochemical energy storage system through the target strong predictor to obtain a numerical health status prediction value. The terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices, and the Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car devices, projection devices, etc. The portable wearable device may be a smart watch, a smart bracelet, a head-mounted device, etc. The head-mounted device may be a virtual reality (VR) device, an augmented reality (AR) device, smart glasses, etc. The server 104 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0046] In an exemplary embodiment, as shown in FIG2 , a method for predicting the health status of an electrochemical energy storage system is provided, and the method is applied to the terminal 102 in FIG1 as an example for description, including the following steps 202 to 208. Among them:

[0047] S202: Determine target evaluation indicators according to state parameters of the electrochemical energy storage system, and obtain sample data corresponding to each target evaluation indicator to obtain a sample data set.

[0048] Optionally, since the electrochemical energy storage system involves multiple factors such as the state of the energy storage battery, the charging and discharging process, temperature, environment, and electrical parameters, and each factor is divided into several small indicators, such as the battery state can be divided into the state of charge (SOC), depth of discharge (DOD), battery health state (SOH) and self-discharge rate. When used for the health status and testing of the specific electrochemical energy storage system, if all indicators are evaluated and integrated, the model will be too complicated, and the correlation between the indicators is strong, so it is necessary to screen the indicators before the model is established. By analyzing the state parameters of the electrochemical energy storage system in a healthy state and an unhealthy state, the target evaluation indicators that can reflect the health status of the energy storage system are determined. Afterwards, a large number of actual data corresponding to the target evaluation indicators of the electrochemical energy storage system under different operating conditions are collected to form a sample data set.

[0049] S204: Training the convolutional neural network model based on the sample data set to obtain a set of candidate weak predictors.

[0050] Optionally, the convolutional neural network model can automatically learn effective feature representations from the input data, and can make predictions or forecasts by learning the relationship between the input (target evaluation index of the sample data) and the output (such as different health status categories of the energy storage system). It should be noted that the health status prediction method of the electrochemical energy storage system provided in the embodiment of the present application can also be used for other algorithm models that use gradient-based optimization, such as BP (Back Propagation) neural network models. During the training process, the neural network continuously adjusts its internal weights and biases to minimize the prediction error. After multiple iterative training, multiple different convolutional neural network models are obtained. Each convolutional neural network model has different model parameters, which together constitute a set of candidate weak predictors.

[0051] S206: Screening a target weak predictor from the candidate weak predictor set, and combining all target weak predictors to obtain a target strong predictor.

[0052] Optionally, a weak predictor with better performance is screened out from the candidate weak predictor set as a target weak predictor. For example, based on the prediction error of the weak predictor, a weak predictor whose prediction error is within a specified range is selected as the target weak predictor. Thereafter, these target weak predictors are combined to obtain a target strong predictor with stronger performance.

[0053] S208: Predicting the health status of the electrochemical energy storage system through the target strong predictor to obtain a numerical health status prediction value.

[0054] Optionally, it should be noted that the result obtained by training the traditional convolutional neural network model is a weak classifier. In the weak classifier, the weight of the sample is usually set according to the category prediction result of the sample. If the sample is misclassified, its weight will be increased so that more attention can be paid to these samples that are difficult to classify correctly in the subsequent training process. In the health status prediction method provided in the embodiment of the present application, in order to provide a more refined and accurate evaluation result, when training the convolutional neural network model, the weight of the sample with the wrong category predicted in the original weak classifier is changed to the weight of the sample with an error exceeding the threshold in the weak predictor, so that the convolutional neural network model no longer only focuses on the misclassified samples, but pays more attention to those samples whose prediction results in the weak predictor have a large error with the actual results. Then, on the basis of multiple weak predictors, a target strong predictor that can output a numerical value in a certain interval is obtained, and the health status of the electrochemical energy storage system is quantitatively represented by this numerical value.

[0055] In addition, in practical applications, multiple weak classifiers obtained by training the convolutional neural network model can be directly combined to form a strong classifier. On the basis of the strong classifier, the weights of the wrongly predicted class samples set in the original strong classifier are changed to the weights of the samples whose errors exceed the threshold set in the strong predictor, thereby achieving the effect of outputting numerical results.

[0056] For example, the BP neural network model is used as an example to illustrate that the BP neural network model is trained multiple times to obtain multiple weak classifiers, and the process of combining them to finally obtain the target strong predictor includes:

[0057] (1) Randomly select m groups of training data from the sample space and initialize the distribution weights of the test data , determine the neural network structure according to the sample input and output dimensions, and initialize the convolutional neural network weights and thresholds: ,in, .

[0058] (2) When training the pth weak classifier, use the training data to train the BP neural network and predict the training data output to obtain the prediction sequence The sum of the prediction errors , The calculation formula is: ,in, .

[0059] (3) According to the prediction sequence The prediction error To calculate the weight of the sequence , The calculation formula is: .

[0060] (4) According to the prediction sequence weight , adjust the weight of the next round of training samples, the adjustment formula is: .

[0061] (5) After P rounds of training, we get P groups of weak classification functions , composed of P groups of weak classification functions The combination obtains a strong classification function , whose expression is: After that, based on the strong classifier, the weight setting method is changed, and finally a strong predictor that can output numerical results is obtained.

[0062] In the above-mentioned health status prediction method of the electrochemical energy storage system, a target strong predictor is formed by integrating the advantages of multiple weak predictors. Since the target strong predictor can adapt to different types of electrochemical energy storage systems and changes in various operating conditions, it can more accurately capture the complex relationship between the state parameters and the health status of the electrochemical energy storage system, thereby accurately evaluating the health status of the energy storage system.

[0063] In an exemplary embodiment, the step of determining the target evaluation index according to the state parameters of the electrochemical energy storage system includes: determining the candidate index affecting the health state of the electrochemical energy storage system according to the state parameters of the electrochemical energy storage system; performing factor analysis on the candidate index to obtain the principal component eigenvalues ​​corresponding to each candidate index; taking the candidate index corresponding to the principal component eigenvalue within a preset range as the target evaluation index; the target evaluation index includes equilibrium state, battery capacity, discharge depth, charging rate, charge and discharge cycle life, total voltage of the battery, total current of the battery, battery ambient temperature, ambient humidity, and main circuit state.

[0064] Optionally, the state parameters of the electrochemical energy storage system include voltage, current, temperature, humidity, etc. First, a series of candidate indicators that may affect the health status of the electrochemical energy storage system are determined based on these state parameters. These candidate indicators are preliminarily screened out and have a high correlation with the health status of the energy storage system. Afterwards, factor analysis is performed on the candidate indicators. In this process, the principal component eigenvalues ​​corresponding to each candidate indicator are calculated. The principal component eigenvalues ​​reflect the principal component's ability to interpret the original data. The candidate indicators corresponding to the principal component eigenvalues ​​within a preset range are used as target evaluation indicators. The preset range is usually determined based on experience or data analysis. The purpose is to screen out those indicators that have an important impact on the health status of the energy storage system and have a strong explanatory power. For example, if the principal component eigenvalue is large, it means that the indicator has a high importance in the comprehensive evaluation.

[0065] In this embodiment, through factor analysis and screening, the target evaluation indicators determined are more targeted and can accurately reflect the health status of the electrochemical energy storage system, avoiding the use of too many irrelevant or secondary indicators, and improving the efficiency and accuracy of the evaluation.

[0066] In an exemplary embodiment, the step of training a convolutional neural network model based on a sample data set to obtain a set of candidate weak predictors includes: selecting a preset group of sample data from the sample data set as training data; determining the network structure of the convolutional neural network model according to the input and output dimensions of the training data, and initializing the model parameters of the convolutional neural network model; in the process of training the convolutional neural network model for multiple rounds based on the training data, obtaining the model parameters corresponding to each round of training; using the convolutional neural network model corresponding to the model parameters as the weak predictor obtained in the corresponding round, and using the weak predictors obtained in all rounds as the set of candidate weak predictors.

[0067] Optionally, a preset group of sample data is selected from the sample data set as training data. This step is to determine the specific instance used to train the convolutional neural network model. The number and selection method of the preset groups can be adjusted according to the specific situation, such as randomly selecting a certain proportion of sample data. The network structure of the convolutional neural network model is determined according to the input and output dimensions of the training data. The input dimension is usually determined by the number of target evaluation indicators, and the output dimension is determined according to the specific prediction task, such as predicting the health status category or value of the energy storage system. After that, the model parameters of the convolutional neural network model are initialized, including the connection weights and biases between the neurons in each layer.

[0068] Furthermore, in the process of training the convolutional neural network model for multiple rounds based on the training data, in each round, the training data is input into the network, the error between the output result and the actual result is calculated, and then the model parameters of the network are adjusted according to the error back propagation algorithm to reduce the error. After each round of training, the model parameters corresponding to the round are obtained. The convolutional neural network model corresponding to the model parameters obtained in each round of training is used as the weak predictor obtained in the corresponding round. After multiple rounds of training, multiple different weak predictors are obtained, and the weak predictors obtained in all rounds are used as a candidate weak predictor set.

[0069] In this embodiment, multiple different weak predictors are obtained through multiple rounds of training, which increases the diversity of the model. By constructing a set of candidate weak predictors, the risk of overfitting can be reduced. Each weak predictor may learn different features and patterns during the training process. The multiple weak predictors in the set work together to improve the generalization ability of the model.

[0070] In an exemplary embodiment, the process of performing multiple rounds of training on a convolutional neural network model based on training data includes: inputting the training data into the convolutional neural network model for the current round of training, obtaining the output result of the current round, and obtaining the prediction error of the current round based on the output result; adjusting the weight of the training data based on the prediction error, and inputting the adjusted training data into the convolutional neural network model for the next round of training.

[0071] Optionally, the training data is input into the convolutional neural network model for the current round of training. The convolutional neural network processes the input data through each layer of neurons through forward propagation calculation, and finally obtains the output result, which is the model's predicted value for the input data. Afterwards, the prediction error of the current round is obtained according to the output result, and the weight of the training data is adjusted according to the prediction error. Among them, the prediction error usually refers to the difference between the model output result and the actual value, which can be calculated by a specific error function, such as mean square error, cross entropy, etc. Specifically, the error is back-propagated from the output layer to the input layer. During the propagation process, the connection weights between the neurons in each layer are adjusted according to the size of the error and the principle of gradient descent. The purpose of weight adjustment is to enable the model to better fit the training data in subsequent training and reduce the prediction error. For samples with larger errors, the corresponding weights will be increased, so that the model pays more attention to these samples in subsequent training, while for samples with smaller errors, the weights will be relatively reduced. Afterwards, the adjusted training data is input into the convolutional neural network model for the next round of training. The above process is repeated, and rounds of iterative training are continuously performed until the preset stopping conditions are met, such as reaching the maximum number of training rounds, the error is less than a certain threshold, etc.

[0072] In this embodiment, by continuously adjusting the weights to reduce the prediction error, the model can gradually approach the true input-output relationship, thereby improving the degree of fit to the training data and the accuracy of prediction of new data. In addition, through the process of adjusting the weights, the model can perform adaptive learning according to the difficulty and importance of the samples, pay more attention to difficult-to-predict samples, and improve the model's ability to handle complex samples.

[0073] In an exemplary embodiment, the step of adjusting the weight of the training data according to the prediction error includes: when the prediction error is greater than an error threshold, increasing the proportion of the weight of the training data.

[0074] Optionally, when the convolutional neural network model is trained for multiple rounds based on the training data, after calculating the prediction error of the current round, the prediction error is compared with the preset error threshold. If the prediction error is greater than the error threshold, it means that the current model does not predict this part of the training data well. At this time, increase the weight ratio of the training data. Specifically, it is to increase the importance of these data in subsequent training. Originally, in the training, the contribution of each sample data to the adjustment of the model parameters was relatively balanced, but now these sample data with larger prediction errors will be given higher weights. In this way, in subsequent training, the model will pay more attention to these samples that are difficult to accurately predict, and adjust the model parameters to a greater extent through the back propagation algorithm to adapt to the characteristics and patterns of these data.

[0075] In this embodiment, by focusing on sample data with larger errors for weight adjustment, the model is made to work harder to fit these difficult-to-process data in subsequent training, which can improve the model's adaptability to various situations, thereby improving the overall prediction accuracy.

[0076] In an exemplary embodiment, the step of screening a target weak predictor from a set of candidate weak predictors and combining all the target weak predictors to obtain a target strong predictor includes: taking a weak predictor of a round corresponding to a prediction error that meets a preset condition as the target weak predictor; determining a weight of the target weak predictor according to an output error corresponding to the target weak predictor, and performing weighted combination of the target weak predictors according to the weight to obtain a target strong predictor.

[0077] Optionally, first, screening is performed from a set of candidate weak predictors. The preset conditions are usually related to the prediction error. For example, weak predictors with prediction errors within a certain range or less than a specific threshold can be screened out as target weak predictors, which means that these weak predictors perform relatively well during the training process and predict the training data more accurately. For the screened target weak predictors, the weight is determined according to their corresponding output errors. Generally speaking, the weak predictor with smaller output errors is given a larger weight. This is because a small output error represents a strong prediction ability of the weak predictor, and a greater contribution should be made when it is combined into a strong predictor. The target weak predictors are weighted and combined according to the determined weights. In this way, multiple target weak predictors are integrated into a target strong predictor.

[0078] For example, assume that 2000 sets of energy storage system operation and health status parameter data are collected, and the input of each set of data is 10 variables, corresponding to the 10 evaluation indicators mentioned above, and the output result is a value between 0 and 1. Among them, 1 represents that the energy storage system is in excellent health, and when the output result is 0, it means that the energy storage system is in extremely poor health. 1900 sets of data are randomly selected from the 2000 sets of values ​​as training data, and 100 sets of data are used as test data. The structure of the convolutional neural network is formulated according to the data dimension of the input and output. A total of 10 convolutional neural network weak predictor functions are trained and generated. Finally, 10 weak predictors are used to form a strong predictor to evaluate the health of the energy storage system. Among them, it is assumed that the number of iterations of the convolutional neural network is 20, the learning rate is set to 0.1, and the error threshold is set to 0.1, that is, the test samples with a prediction error exceeding 0.1 are samples that need to be strengthened.

[0079] Furthermore, 10 weak predictors are used to form a strong predictor and a weak predictor to evaluate the health of the energy storage system. The comparison of the prediction errors of the weak predictor and the strong predictor is shown in Figure 3. As shown in Figure 3, the error volatility of the weak predictor composed of the convolutional neural network is large, and the error value is large. Even when it is relatively stable, the error is maintained at around 0.2. Although the prediction error of the strong predictor fluctuates from time to time, this is because the model encounters more accidental state data that was not encountered during training and produces occasional large errors. However, under normal circumstances, the error will be controlled around 0.1, which is lower than the prediction error of the weak predictor.

[0080] In this embodiment, the target strong predictor obtained by screening and combining the target weak predictors has better generalization ability, can adapt to different data sets and actual application scenarios, and more accurately evaluate the health status of the energy storage system.

[0081] In an exemplary embodiment, as shown in FIG4 , a method for predicting the health status of an electrochemical energy storage system is provided, the method comprising the following steps:

[0082] Determine candidate indicators that affect the health status of the electrochemical energy storage system based on the state parameters of the electrochemical energy storage system; perform factor analysis on the candidate indicators to obtain the principal component eigenvalues ​​corresponding to each candidate indicator; use the candidate indicators corresponding to the principal component eigenvalues ​​within a preset range as target evaluation indicators; the target evaluation indicators include equilibrium state, battery capacity, discharge depth, charging rate, charge and discharge cycle life, total battery voltage, total battery current, battery ambient temperature, ambient humidity, and main circuit state.

[0083] Obtain sample data corresponding to each target evaluation indicator to obtain a sample data set.

[0084] A preset group of sample data is selected from the sample data set as training data; the network structure of the convolutional neural network model is determined according to the input and output dimensions of the training data, and the model parameters of the convolutional neural network model are initialized.

[0085] The training data is input into the convolutional neural network model for the current round of training to obtain the output result of the current round, and the prediction error of the current round is obtained according to the output result; when the prediction error is greater than the error threshold, the weight ratio of the training data is increased, and the adjusted training data is input into the convolutional neural network model for the next round of training.

[0086] In the process of performing multiple rounds of training on the convolutional neural network model based on the training data, the model parameters corresponding to each round of training are obtained; the convolutional neural network model corresponding to the model parameters is used as the weak predictor obtained in the corresponding round, and the weak predictors obtained in all rounds are used as a candidate weak predictor set.

[0087] The weak predictor of the round corresponding to the prediction error that meets the preset conditions is used as the target weak predictor; the weight of the target weak predictor is determined according to the output error corresponding to the target weak predictor, and the target weak predictor is weightedly combined according to the weight to obtain the target strong predictor.

[0088] The health status of the electrochemical energy storage system is predicted through the target strong predictor to obtain a numerical health status prediction value.

[0089] In this embodiment, a target strong predictor is formed by integrating the advantages of multiple weak predictors. Since the target strong predictor can adapt to different types of electrochemical energy storage systems and changes in various operating conditions, it can more accurately capture the complex relationship between the state parameters and the health status of the electrochemical energy storage system, thereby accurately evaluating the health status of the energy storage system.

[0090] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages 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 carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0091] Based on the same inventive concept, the embodiment of the present application also provides a device for predicting the health status of an electrochemical energy storage system for implementing the method for predicting the health status of an electrochemical energy storage system involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the embodiments of the device for predicting the health status of one or more electrochemical energy storage systems provided below can refer to the limitations of the method for predicting the health status of an electrochemical energy storage system above, and will not be repeated here.

[0092] In an exemplary embodiment, as shown in FIG5 , a health status prediction device for an electrochemical energy storage system is provided, comprising: an indicator determination module 10, a model training module 20, a predictor combination module 30 and a status prediction module 40, wherein:

[0093] The indicator determination module 10 is used to determine the target evaluation indicator according to the state parameters of the electrochemical energy storage system, and obtain sample data corresponding to each target evaluation indicator to obtain a sample data set.

[0094] The model training module 20 is used to train the convolutional neural network model based on the sample data set to obtain a set of candidate weak predictors.

[0095] The predictor combination module 30 is used to screen a target weak predictor from the candidate weak predictor set, and combine all the target weak predictors to obtain a target strong predictor.

[0096] The state prediction module 40 is used to predict the health state of the electrochemical energy storage system through a target strong predictor to obtain a numerical health state prediction value.

[0097] In an exemplary embodiment, the indicator determination module 10 is also used to determine candidate indicators that affect the health status of the electrochemical energy storage system according to the state parameters of the electrochemical energy storage system; perform factor analysis on the candidate indicators to obtain the principal component eigenvalues ​​corresponding to each candidate indicator; and use the candidate indicators corresponding to the principal component eigenvalues ​​within a preset range as target evaluation indicators; the target evaluation indicators include equilibrium state, battery capacity, discharge depth, charging rate, charge and discharge cycle life, total battery voltage, total battery current, battery ambient temperature, ambient humidity, and main circuit state.

[0098] In an exemplary embodiment, the model training module 20 is also used to select a preset group of sample data from the sample data set as training data; determine the network structure of the convolutional neural network model according to the input and output dimensions of the training data, and initialize the model parameters of the convolutional neural network model; in the process of performing multiple rounds of training on the convolutional neural network model based on the training data, obtain the model parameters corresponding to each round of training; use the convolutional neural network model corresponding to the model parameters as the weak predictor obtained in the corresponding round, and use the weak predictors obtained in all rounds as a candidate weak predictor set.

[0099] In an exemplary embodiment, the model training module 20 is also used to input the training data into the convolutional neural network model for the current round of training, obtain the output result of the current round, and obtain the prediction error of the current round based on the output result; adjust the weight of the training data according to the prediction error, and input the adjusted training data into the convolutional neural network model for the next round of training.

[0100] In an exemplary embodiment, the model training module 20 is further configured to increase the weight ratio of the training data when the prediction error is greater than an error threshold.

[0101] In an exemplary embodiment, the predictor combination module 30 is further configured to use the weak predictor of the round corresponding to the prediction error that meets the preset condition as the target weak predictor; determine the weight of the target weak predictor according to the output error corresponding to the target weak predictor, and perform weighted combination on the target weak predictors according to the weights to obtain the target strong predictor.

[0102] Each module in the above-mentioned electrochemical energy storage system health status prediction device can be implemented in whole or in part by software, hardware and their combination. 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.

[0103] In an exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be shown in FIG6. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. The processor, the memory, and the input / output interface are connected via a system bus, and the communication interface, the display unit, and the input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. 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 and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be implemented through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a health status prediction method of an electrochemical energy storage system is implemented. The display unit of the computer device is used to form a visually visible picture, which may be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0104] Those skilled in the art will understand that the structure shown in FIG. 6 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0105] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the following steps when executing the computer program: determining a target evaluation index according to a state parameter of an electrochemical energy storage system, and acquiring sample data corresponding to each target evaluation index to obtain a sample data set; training a convolutional neural network model based on the sample data set to obtain a set of candidate weak predictors; screening a target weak predictor from the set of candidate weak predictors, and combining all target weak predictors to obtain a target strong predictor; and predicting the health state of the electrochemical energy storage system by using the target strong predictor to obtain a numerical health state prediction value.

[0106] In one embodiment, the processor determines the target evaluation index according to the state parameters of the electrochemical energy storage system when executing the computer program, including: determining the candidate index that affects the health state of the electrochemical energy storage system according to the state parameters of the electrochemical energy storage system; performing factor analysis on the candidate index to obtain the principal component eigenvalues ​​corresponding to each candidate index; taking the candidate index corresponding to the principal component eigenvalue within a preset range as the target evaluation index; the target evaluation index includes equilibrium state, battery capacity, discharge depth, charging rate, charge and discharge cycle life, total voltage of the battery, total current of the battery, battery ambient temperature, ambient humidity, and main circuit state.

[0107] In one embodiment, the processor executes a computer program involving training a convolutional neural network model based on a sample data set to obtain a set of candidate weak predictors, including: selecting a preset group of sample data from the sample data set as training data; determining the network structure of the convolutional neural network model according to the input and output dimensions of the training data, and initializing the model parameters of the convolutional neural network model; in the process of training the convolutional neural network model for multiple rounds based on the training data, obtaining the model parameters corresponding to each round of training; using the convolutional neural network model corresponding to the model parameters as the weak predictor obtained in the corresponding round, and using the weak predictors obtained in all rounds as the candidate weak predictor set.

[0108] In one embodiment, the process of performing multiple rounds of training on a convolutional neural network model based on training data when a processor executes a computer program includes: inputting the training data into the convolutional neural network model for a current round of training, obtaining an output result of the current round, and obtaining a prediction error of the current round based on the output result; adjusting the weight of the training data based on the prediction error, and inputting the adjusted training data into the convolutional neural network model for the next round of training.

[0109] In one embodiment, the processor executing the computer program involves adjusting the weight of the training data according to the prediction error, including: when the prediction error is greater than the error threshold, increasing the proportion of the weight of the training data.

[0110] In one embodiment, the processor executes the computer program involving screening a target weak predictor from a set of candidate weak predictors, and combining all the target weak predictors to obtain a target strong predictor, including: using a weak predictor of a round corresponding to a prediction error that meets a preset condition as the target weak predictor; determining a weight of the target weak predictor according to an output error corresponding to the target weak predictor, and performing weighted combination of the target weak predictors according to the weight to obtain the target strong predictor.

[0111] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: according to the state parameters of the electrochemical energy storage system, a target evaluation index is determined, and sample data corresponding to each target evaluation index is obtained to obtain a sample data set; a convolutional neural network model is trained based on the sample data set to obtain a candidate weak predictor set; a target weak predictor is screened from the candidate weak predictor set, and all target weak predictors are combined to obtain a target strong predictor; and a health state prediction of the electrochemical energy storage system is performed using the target strong predictor to obtain a numerical health state prediction value.

[0112] In one embodiment, when a computer program is executed by a processor, the method involves determining target evaluation indicators based on state parameters of an electrochemical energy storage system, including: determining candidate indicators that affect the health state of the electrochemical energy storage system based on the state parameters of the electrochemical energy storage system; performing factor analysis on the candidate indicators to obtain principal component eigenvalues ​​corresponding to each candidate indicator; taking the candidate indicators corresponding to the principal component eigenvalues ​​within a preset range as target evaluation indicators; the target evaluation indicators include equilibrium state, battery capacity, depth of discharge, charging rate, charge-discharge cycle life, total voltage of the battery, total current of the battery, battery ambient temperature, ambient humidity, and main circuit state.

[0113] In one embodiment, when the computer program is executed by a processor, the computer program involves training a convolutional neural network model based on a sample data set to obtain a set of candidate weak predictors, including: selecting a preset group of sample data from the sample data set as training data; determining the network structure of the convolutional neural network model according to the input and output dimensions of the training data, and initializing the model parameters of the convolutional neural network model; in the process of training the convolutional neural network model for multiple rounds based on the training data, obtaining the model parameters corresponding to each round of training; using the convolutional neural network model corresponding to the model parameters as the weak predictor obtained in the corresponding round, and using the weak predictors obtained in all rounds as the candidate weak predictor set.

[0114] In one embodiment, the process of performing multiple rounds of training on a convolutional neural network model based on training data when a computer program is executed by a processor includes: inputting the training data into the convolutional neural network model for a current round of training, obtaining an output result of the current round, and obtaining a prediction error of the current round based on the output result; adjusting the weight of the training data based on the prediction error, and inputting the adjusted training data into the convolutional neural network model for the next round of training.

[0115] In one embodiment, the method of adjusting the weight of the training data according to the prediction error when the computer program is executed by the processor includes: when the prediction error is greater than the error threshold, increasing the proportion of the weight of the training data.

[0116] In one embodiment, when the computer program is executed by a processor, the steps of screening a target weak predictor from a set of candidate weak predictors and combining all the target weak predictors to obtain a target strong predictor include: taking a weak predictor of a round corresponding to a prediction error that meets a preset condition as the target weak predictor; determining a weight of the target weak predictor according to an output error corresponding to the target weak predictor, and performing weighted combination of the target weak predictors according to the weight to obtain the target strong predictor.

[0117] In one embodiment, a computer program product is provided, including a computer program, which implements the following steps when executed by a processor: determining a target evaluation index according to a state parameter of an electrochemical energy storage system, and obtaining sample data corresponding to each target evaluation index to obtain a sample data set; training a convolutional neural network model based on the sample data set to obtain a candidate weak predictor set; screening a target weak predictor from the candidate weak predictor set, and combining all the target weak predictors to obtain a target strong predictor; and predicting the health state of the electrochemical energy storage system using the target strong predictor to obtain a numerical health state prediction value.

[0118] In one embodiment, when a computer program is executed by a processor, the method involves determining target evaluation indicators based on state parameters of an electrochemical energy storage system, including: determining candidate indicators that affect the health state of the electrochemical energy storage system based on the state parameters of the electrochemical energy storage system; performing factor analysis on the candidate indicators to obtain principal component eigenvalues ​​corresponding to each candidate indicator; taking the candidate indicators corresponding to the principal component eigenvalues ​​within a preset range as target evaluation indicators; the target evaluation indicators include equilibrium state, battery capacity, depth of discharge, charging rate, charge-discharge cycle life, total voltage of the battery, total current of the battery, battery ambient temperature, ambient humidity, and main circuit state.

[0119] In one embodiment, when the computer program is executed by a processor, the computer program involves training a convolutional neural network model based on a sample data set to obtain a set of candidate weak predictors, including: selecting a preset group of sample data from the sample data set as training data; determining the network structure of the convolutional neural network model according to the input and output dimensions of the training data, and initializing the model parameters of the convolutional neural network model; in the process of training the convolutional neural network model for multiple rounds based on the training data, obtaining the model parameters corresponding to each round of training; using the convolutional neural network model corresponding to the model parameters as the weak predictor obtained in the corresponding round, and using the weak predictors obtained in all rounds as the candidate weak predictor set.

[0120] In one embodiment, the process of performing multiple rounds of training on a convolutional neural network model based on training data when a computer program is executed by a processor includes: inputting the training data into the convolutional neural network model for a current round of training, obtaining an output result of the current round, and obtaining a prediction error of the current round based on the output result; adjusting the weight of the training data based on the prediction error, and inputting the adjusted training data into the convolutional neural network model for the next round of training.

[0121] In one embodiment, the method of adjusting the weight of the training data according to the prediction error when the computer program is executed by the processor includes: when the prediction error is greater than the error threshold, increasing the proportion of the weight of the training data.

[0122] In one embodiment, when the computer program is executed by a processor, the steps of screening a target weak predictor from a set of candidate weak predictors and combining all the target weak predictors to obtain a target strong predictor include: taking a weak predictor of a round corresponding to a prediction error that meets a preset condition as the target weak predictor; determining a weight of the target weak predictor according to an output error corresponding to the target weak predictor, and performing weighted combination of the target weak predictors according to the weight to obtain the target strong predictor.

[0123] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the 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. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0124] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above 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 application.

[0125] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for predicting the health status of an electrochemical energy storage system, characterized in that: The method comprises: Determine target evaluation indicators according to the state parameters of the electrochemical energy storage system, and obtain sample data corresponding to each of the target evaluation indicators to obtain a sample data set; Training a convolutional neural network model based on the sample data set to obtain a set of candidate weak predictors; Screening a target weak predictor from the candidate weak predictor set, and combining all the target weak predictors to obtain a target strong predictor; The health state of the electrochemical energy storage system is predicted by the target strong predictor to obtain a numerical health state prediction value.

2. The method according to claim 1, characterized in that Determining the target evaluation index according to the state parameters of the electrochemical energy storage system includes: Determining candidate indicators that affect the health state of the electrochemical energy storage system according to the state parameters of the electrochemical energy storage system; Performing factor analysis on the candidate indicators to obtain the principal component eigenvalues ​​corresponding to the candidate indicators; The candidate indicators corresponding to the principal component eigenvalues ​​within a preset range are used as target evaluation indicators; the target evaluation indicators include equilibrium state, battery capacity, discharge depth, charging rate, charge and discharge cycle life, total battery voltage, total battery current, battery ambient temperature, ambient humidity, and main circuit state.

3. The method according to claim 1, characterized in that The convolutional neural network model is trained based on the sample data set to obtain a set of candidate weak predictors, including: Selecting a preset group of sample data from the sample data set as training data; Determine the network structure of the convolutional neural network model according to the input and output dimensions of the training data, and initialize the model parameters of the convolutional neural network model; In the process of performing multiple rounds of training on the convolutional neural network model based on the training data, obtaining model parameters corresponding to each round of training; The convolutional neural network model corresponding to the model parameters is used as the weak predictor obtained in the corresponding round, and the weak predictors obtained in all rounds are used as a candidate weak predictor set.

4. The method according to claim 3, characterized in that: The process of performing multiple rounds of training on the convolutional neural network model based on the training data includes: Inputting the training data into the neural network model for current round training to obtain the output result of the current round, and obtaining the prediction error of the current round according to the output result; The weight of the training data is adjusted according to the prediction error, and the adjusted training data is input into the convolutional neural network model for the next round of training.

5. The method according to claim 4, characterized in that The step of adjusting the weight of the training data according to the prediction error comprises: When the prediction error is greater than an error threshold, the weight ratio of the training data is increased.

6. The method according to claim 4, characterized in that The step of screening a target weak predictor from the candidate weak predictor set and combining all the target weak predictors to obtain a target strong predictor includes: The weak predictor of the round corresponding to the prediction error that meets the preset conditions is used as the target weak predictor; The weight of the target weak predictor is determined according to the output error corresponding to the target weak predictor, and the target weak predictors are weighted combined according to the weight to obtain a target strong predictor.

7. A health status prediction device for an electrochemical energy storage system, characterized in that: The device comprises: An indicator determination module is used to determine the target evaluation indicator according to the state parameters of the electrochemical energy storage system, and obtain sample data corresponding to each of the target evaluation indicators to obtain a sample data set; A model training module, used for training a convolutional neural network model based on the sample data set to obtain a set of candidate weak predictors; A predictor combination module, configured to screen a target weak predictor from the candidate weak predictor set, and combine all the target weak predictors to obtain a target strong predictor; The state prediction module is used to predict the health state of the electrochemical energy storage system through the target strong predictor to obtain a numerical health state prediction value.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images