Battery state parameter prediction method and apparatus, and electronic device and storage medium

Abstract

A battery state parameter prediction method and apparatus, and an electronic device and a storage medium. The method comprises: respectively performing data processing on historical discharge data and real-time monitoring data, so as to determine historical SOH-associated data and battery feature data of a battery to be predicted (S102); using the historical SOH-associated data to determine, by means of an autoregressive model in a dual-layer prediction framework, an SOH prediction sequence of said battery within a specific prediction time period (S103); and on the basis of the battery feature data and the SOH prediction sequence, using a machine learning prediction model in the dual-layer prediction framework to determine predicted SOH data of said battery at a specific prediction time point (S104). Thus, by means of a dual-layer prediction framework integrating an autoregressive model and a machine learning prediction model, the prediction of predicted SOH data of a battery to be predicted at a specific prediction time point can be realized by capturing the characteristics of battery performance changing dynamically over time, thereby further improving the accuracy of a prediction result.

Description

Method and device for predicting battery state parameter, electronic device and storage medium

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] The present disclosure claims priority to the Chinese patent application No. 2024107902397, entitled "Method and device for predicting battery state parameter, electronic device and storage medium", filed on June 19, 2024, with the State Intellectual Property Office of China, the content of which is incorporated herein by reference in its entirety. TECHNICAL FIELD

[0003] The present disclosure relates to the technical field of data processing, and in particular to a method and device for predicting battery state parameter, an electronic device and a storage medium. BACKGROUND

[0004] SOH (State of health) data is a battery health state, commonly used to represent the remaining life of the battery, and is related to the safety of the battery; in particular in the field of electric vehicles and renewable energy storage, accurate prediction of battery state health data (i.e., SOH data) is crucial for the battery management system (BMS); accurate prediction of SOH data not only affects the maintenance and replacement plan of the battery, but also relates to the safety and economy of the battery-related system.

[0005] In the implementation of SOH data prediction of the battery, the prediction model used often has a fixed algorithm framework (for example, a pure physical model or an empirical model), and relies on a single data source (for example, battery charge and discharge cycle data), ignoring the complex changes of battery performance over time; such a model using a fixed algorithm framework cannot be applied to the prediction of SOH data of different battery types, reducing the accuracy of the prediction result.

[0006] SUMMARY

[0007] Therefore, the purpose of the present disclosure is to provide a method and device for predicting battery state parameter, an electronic device and a storage medium, which utilize a double-layer prediction framework to accurately predict the predicted SOH data of a battery to be predicted at a specific prediction time point by capturing the dynamic changes of battery performance over time.

[0008] An optional embodiment of the present disclosure provides a method for predicting battery state parameter, the method comprising:

[0009] obtaining historical discharge data and real-time monitoring data of a battery to be predicted;

[0010] respectively, to determine historical SOH correlation data and battery characteristic data of the battery to be predicted;

[0011] SOH prediction sequence of the battery to be predicted in a specific prediction time period is determined by using the historical SOH correlation data, through an autoregressive model in a double-layer prediction framework, and combining performance change characteristics of the battery to be predicted;

[0012] Based on the battery characteristic data and the SOH prediction sequence, a machine learning prediction model in the double-layer prediction framework is used to determine the predicted SOH data of the battery to be predicted at a specific prediction time point.

[0013] In an optional implementation, the data processing of the historical discharge data and the real-time monitoring data respectively to determine the historical SOH correlation data and the battery characteristic data of the battery to be predicted comprises:

[0014] The historical discharge data and the real-time monitoring data are respectively subjected to data cleaning processing to filter out abnormal data in the historical discharge data and the real-time monitoring data, to obtain filtered historical discharge data and filtered real-time monitoring data;

[0015] The filtered historical discharge data and the filtered real-time monitoring data are respectively subjected to data normalization processing to obtain normalized historical discharge data and normalized real-time monitoring data;

[0016] A plurality of historical voltage data, a plurality of historical current data, and historical SOH correlation data of the battery to be predicted in a specific monitoring time period are determined by analyzing the normalized historical discharge data;

[0017] Based on the plurality of historical voltage data and the plurality of historical current data, internal resistance change data of the battery to be predicted in a specific monitoring time period is determined;

[0018] Battery attribute data of the battery to be predicted is determined by analyzing the normalized real-time monitoring data;

[0019] Based on the internal resistance change data and the battery attribute data, battery characteristic data of the battery to be predicted is determined.

[0020] In an optional implementation, the battery attribute data comprises one or more of battery temperature data, real-time current data, real-time SOC data, and battery mileage data.

[0021] In an optional implementation, the determination of the battery characteristic data of the battery to be predicted based on the internal resistance change data and the battery attribute data comprises:

[0022] determine a first cross-derivative variable of the battery to be predicted by analyzing a first correlation between the battery temperature data and the real-time SOC data;

[0023] determine a second cross-derivative variable of the battery to be predicted by analyzing a second correlation between the real-time current data and the battery mileage data;

[0024] determine the internal resistance change data, the battery temperature data, the real-time current data, the real-time SOC data, the battery mileage data, the first cross-derivative variable and the second cross-derivative variable as battery characteristic data of the battery to be predicted.

[0025] In an optional embodiment, the SOH prediction sequence of the battery to be predicted in a specific prediction time period is determined by using the historical SOH correlation data, through an autoregressive model in a double-layer prediction framework, and combining the performance change characteristics of the battery to be predicted, including:

[0026] determine an SOH difference sequence of the battery to be predicted at each first historical prediction time point by using historical SOH data in the historical SOH correlation data, through a difference layer in the autoregressive model;

[0027] determine AR data of the battery to be predicted by using the SOH difference sequence of the battery to be predicted at each first historical prediction time point, through an autoregressive layer in the autoregressive model;

[0028] determine MA data of the battery to be predicted by using historical error data of the battery to be predicted at each second historical prediction time point in the historical SOH correlation data, through a moving average layer in the autoregressive model;

[0029] determine an SOH prediction sequence of the battery to be predicted in a specific prediction time period based on the AR data and the MA data.

[0030] In an optional embodiment, the autoregressive model is trained by the following steps:

[0031] obtain training SOH data and SOH data labels of each training sample battery;

[0032] obtain training SOH data and SOH data labels of each training sample battery;

[0033] predict an SOH difference sequence of a test sample battery by using the autoregressive model to be detected;

[0034] An autocorrelation plot of the SOH difference sequence of the test sample battery is drawn using an ACF test function;

[0035] By analyzing the autocorrelation plot, it is determined whether the SOH difference sequence of the test sample battery is stationary;

[0036] If yes, it is determined that the autoregressive model to be detected is a trained autoregressive model, and the model coefficients of the autoregressive model are determined.

[0037] In an optional embodiment, the machine learning prediction model is any one of a random forest model and a gradient boosting machine.

[0038] In an optional embodiment, the machine learning prediction model is trained by the following steps:

[0039] The key model parameters of the machine learning prediction model are obtained using a grid search method;

[0040] According to the key model parameters, a machine learning prediction model to be trained is constructed;

[0041] Using the obtained each training sample battery and the SOH data label of each training sample battery, the machine learning prediction model to be trained is trained to obtain a trained machine learning prediction model.

[0042] In an optional embodiment, the key model parameters include one or more of the number of trees, the maximum depth of the tree, and the learning rate.

[0043] An optional embodiment of the present disclosure also provides a battery state parameter prediction device, which comprises:

[0044] A data acquisition module is configured to acquire historical discharge data and real-time monitoring data of a battery to be predicted;

[0045] A data processing module is configured to process the historical discharge data and the real-time monitoring data respectively to determine historical SOH correlation data and battery characteristic data of the battery to be predicted;

[0046] A sequence prediction module is configured to determine an SOH prediction sequence of the battery to be predicted in a specific prediction time period by an autoregressive model in a double-layer prediction framework, in combination with performance change characteristics of the battery to be predicted, using the historical SOH correlation data.

[0047] An SOH prediction module is configured to determine a predicted SOH data of the battery to be predicted at a specific prediction time point by a machine learning prediction model in the double-layer prediction framework, based on the battery characteristic data and the SOH prediction sequence.

[0048] In an optional implementation, when the data processing module is configured to determine the historical SOH correlation data and the battery characteristic data of the battery to be predicted by processing the historical discharge data and the real-time monitoring data respectively, the data processing module is configured to:

[0049] perform data cleaning processing on the historical discharge data and the real-time monitoring data respectively, filter out abnormal data in the historical discharge data and the real-time monitoring data, and obtain filtered historical discharge data and filtered real-time monitoring data;

[0050] perform data normalization processing on the filtered historical discharge data and the filtered real-time monitoring data respectively, and obtain normalized historical discharge data and normalized real-time monitoring data;

[0051] determine, by analyzing the normalized historical discharge data, a plurality of historical voltage data, a plurality of historical current data, and historical SOH correlation data of the battery to be predicted in a specific monitoring time period;

[0052] determine, based on the plurality of historical voltage data and the plurality of historical current data, internal resistance change data of the battery to be predicted in the specific monitoring time period;

[0053] determine, by analyzing the normalized real-time monitoring data, battery attribute data of the battery to be predicted;

[0054] determine, based on the internal resistance change data and the battery attribute data, battery characteristic data of the battery to be predicted.

[0055] In an optional implementation, the battery attribute data includes one or more of battery temperature data, real-time current data, real-time SOC data, and battery mileage data.

[0056] In an optional implementation, when the data processing module is configured to determine the battery characteristic data of the battery to be predicted based on the internal resistance change data and the battery attribute data, the data processing module is configured to:

[0057] determine, by analyzing a first correlation between the battery temperature data and the real-time SOC data, a first cross-derivative variable of the battery to be predicted;

[0058] determine, by analyzing a second correlation between the real-time current data and the battery mileage data, a second cross-derivative variable of the battery to be predicted;

[0059] The internal resistance change data, the battery temperature data, the real-time current data, the real-time SOC data, the battery mileage data, the first cross-derivative variable, and the second cross-derivative variable are determined as the battery characteristic data of the battery to be predicted.

[0060] In an optional embodiment, when the sequence prediction module is used to determine the SOH prediction sequence of the battery to be predicted in a specific prediction time period by using the historical SOH correlation data and an autoregressive model in a double-layer prediction framework in combination with the performance change characteristics of the battery to be predicted, the sequence prediction module is used to:

[0061] determine, by using historical SOH data in the historical SOH correlation data and a difference layer in the autoregressive model, an SOH difference sequence of the battery to be predicted at each first historical prediction time point;

[0062] determine, by using the SOH difference sequence of the battery to be predicted at each first historical prediction time point and an autoregressive layer in the autoregressive model, AR data of the battery to be predicted;

[0063] determine, by using historical error data of the battery to be predicted at each second historical prediction time point in the historical SOH correlation data and a moving average layer in the autoregressive model, MA data of the battery to be predicted;

[0064] determine the SOH prediction sequence of the battery to be predicted in a specific prediction time period based on the AR data and the MA data.

[0065] In an optional embodiment, the prediction device further comprises a model training module, and the model training module is used to train the autoregressive model by the following steps:

[0066] obtain training SOH data and SOH data labels of each training sample battery;

[0067] obtain the SOH difference sequence of the test sample battery by using the training SOH data and the SOH data labels of each training sample battery and the to-be-detected autoregressive model;

[0068] obtain the SOH difference sequence of the test sample battery by using the training SOH data and the SOH data labels of each training sample battery and the to-be-detected autoregressive model;

[0069] draw an autocorrelation graph of the SOH difference sequence of the test sample battery by using an ACF test function;

[0070] determine whether the SOH difference sequence of the test sample battery is stationary by analyzing the autocorrelation graph;

[0071] If yes, it is determined that the to-be-detected autoregressive model is a trained autoregressive model, and a model coefficient of the autoregressive model is determined.

[0072] In an optional embodiment, the machine learning prediction model is any one of a random forest model and a gradient boosting machine.

[0073] In an optional embodiment, the model training module is configured to train the machine learning prediction model by the following steps:

[0074] The grid search method is used to obtain key model parameters of the machine learning prediction model.

[0075] According to the key model parameters, a to-be-trained machine learning prediction model is constructed.

[0076] Each training sample battery and an SOH data label of each training sample battery are used to train the to-be-trained machine learning prediction model, so as to obtain a trained machine learning prediction model.

[0077] In an optional embodiment, the key model parameters include one or more of a number of trees, a maximum depth of a tree, and a learning rate.

[0078] An optional embodiment of the present disclosure further provides an electronic device, which comprises a processor, a memory and a bus, the memory stores machine readable instructions executable by the processor, and the processor and the memory communicate through the bus when the electronic device is running, and the machine readable instructions are executed by the processor to perform the steps of the battery state parameter prediction method as described above.

[0079] An optional embodiment of the present disclosure further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to perform the steps of the battery state parameter prediction method as described above.

[0080] The method, device, electronic device and storage medium provided by the embodiments of the present disclosure can predict the state parameters of the battery by using the historical discharge data and the real-time monitoring data of the battery to be predicted, and then can improve the accuracy of the prediction results by enriching the data referred to in the prediction and capturing the dynamic change characteristics of the battery performance over time.

[0081] In order to make the above objectives, features and advantages of the present disclosure more apparent, the following will describe preferred embodiments in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0082] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the present disclosure, and therefore should not be considered as a limitation to the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0083] FIG. 1 is a flowchart of a method for predicting battery state parameters according to an optional embodiment of the present disclosure;

[0084] FIG. 2 is a schematic structural diagram of a device for predicting battery state parameters according to an optional embodiment of the present disclosure;

[0085] FIG. 3 is a schematic structural diagram of a device for predicting battery state parameters according to an optional embodiment of the present disclosure;

[0086] FIG. 4 is a schematic structural diagram of an electronic device according to an optional embodiment of the present disclosure. DETAILED DESCRIPTION

[0087] To make the purposes, technical solutions, and advantages of the embodiments of the present disclosure clearer, the following will be combined with the accompanying drawings of the embodiments of the present disclosure to make a clear and complete description of the technical solutions in the embodiments of the present disclosure. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all the embodiments. The components of the embodiments of the present disclosure generally described and shown in the accompanying drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present disclosure provided in the accompanying drawings is not intended to limit the scope of the claimed present disclosure, but only represents selected embodiments of the present disclosure. Based on the embodiments of the present disclosure, every other embodiment obtained by a person skilled in the art without creative work falls within the scope of protection of the present disclosure.

[0088] It is found through research that SOH (State of health) data is battery health status, commonly used to represent the remaining life of the battery, which relates to the safety of the battery; especially in the field of electric vehicles and renewable energy storage, accurate prediction of battery state health data (i.e., SOH data) is crucial for the battery management system (BMS); accurate prediction of SOH data not only affects the maintenance and replacement plan of the battery, but also relates to the safety and economy of the battery-related system.

[0089] In the implementation of the prediction of SOH data of the battery, the prediction model used often has a fixed algorithm framework (for example, a pure physical model or an empirical model), and relies on a single data source (for example, battery charge and discharge cycle data), ignoring the complex changes of battery performance over time; such a model with a fixed algorithm framework cannot be applied to the prediction of SOH data of different battery types, reducing the accuracy of the prediction results.

[0090] Based on this, the embodiments of the present disclosure provide a prediction method of battery state parameters, which can predict the predicted SOH data of the battery to be predicted by means of historical discharge data and real-time monitoring data of the battery to be predicted, and then the accuracy of the predicted SOH data can be realized by enriching the data referred to in the prediction; and by means of the double-layer prediction framework integrating the autoregressive model and the machine learning prediction model, the prediction of the predicted SOH data of the battery to be predicted at a specific prediction time point is realized by capturing the characteristics of the dynamic changes of the battery performance over time, so as to further improve the accuracy of the prediction results.

[0091] Please refer to FIG. 1, which is a flowchart of a prediction method of battery state parameters provided by an optional embodiment of the present disclosure. As shown in FIG. 1, the prediction method of battery state parameters provided by an optional embodiment of the present disclosure comprises:

[0092] S101, obtaining historical discharge data and real-time monitoring data of a battery to be predicted.

[0093] S102, respectively, the historical discharge data and the real-time monitoring data are processed to determine the historical SOH correlation data and the battery characteristic data of the battery to be predicted.

[0094] S103, using the historical SOH correlation data, through the autoregressive model in the double-layer prediction framework, combined with the performance change characteristics of the battery to be predicted, the SOH prediction sequence of the battery to be predicted in a specific prediction time period is determined.

[0095] S104, based on the battery characteristic data and the SOH prediction sequence, using the machine learning prediction model in the double-layer prediction framework, the predicted SOH data of the battery to be predicted at a specific prediction time point is determined.

[0096] The prediction method of the battery state parameter provided by the embodiment of the present disclosure considers that the information contained in the "single" data is not rich enough, so when predicting the predicted SOH data of the battery to be predicted, the historical discharge data and real-time monitoring data of the battery to be predicted and other multi-aspect data are used to realize accurate prediction of the predicted SOH data of the battery to be predicted. On this basis, in order to ensure the accuracy of the obtained historical discharge data and real-time monitoring data, the historical SOH correlation data and battery characteristic data are extracted from the historical discharge data and real-time monitoring data by data cleaning, normalization and other data processing of the historical discharge data and real-time monitoring data, to ensure the quality of the predicted SOH data. Then, by means of the double-layer prediction framework integrating the autoregressive model and the machine learning prediction model, the prediction of the predicted SOH data of the battery to be predicted at a specific prediction time point is realized by capturing the characteristics of the dynamic change of the battery performance with time, so as to improve the accuracy of the prediction result.

[0097] In step S101, in order to ensure the comprehensiveness of the characteristic data used for predicting the predicted SOH data, the historical discharge data and real-time monitoring data of the battery to be predicted are respectively obtained from the battery management system (BMS system).

[0098] The historical discharge data includes historical charge and discharge information (such as historical charge times, historical charge time, historical current, historical temperature, cumulative mileage and historical SOH correlation data) of the battery to be predicted; the real-time monitoring data includes real-time current, real-time temperature, state of charge data (SOC) and mileage data.

[0099] Here, for the use of real-time monitoring data of the battery to be predicted, not only can the feature data available for predicting the SOH data prediction be enriched, but also the predicted SOH data obtained by the prediction can reflect the current use state and environment of the battery to be predicted in real time, and thus the data in the BMS system can be fully utilized to achieve accurate prediction of the predicted SOH data.

[0100] In order to ensure the accuracy of the feature data used for predicting the SOH data prediction, after the historical discharge data and the real-time monitoring data are obtained, the historical discharge data and the real-time monitoring data are respectively subjected to multiple data processing to filter out abnormal data in the above data and eliminate the dimension influence.

[0101] In step S102, the historical discharge data and the real-time monitoring data are respectively subjected to normalization, data cleaning and other data processing steps to extract the historical SOH associated data and the battery feature data of the battery to be predicted from the historical discharge data and the real-time monitoring data.

[0102] In an optional embodiment, step S102 includes:

[0103] S1021, respectively, the historical discharge data and the real-time monitoring data are subjected to data cleaning processing to filter out abnormal data in the historical discharge data and the real-time monitoring data, to obtain filtered historical discharge data and filtered real-time monitoring data.

[0104] In this step, the historical discharge data and the real-time monitoring data are respectively subjected to data cleaning processing, and an abnormal value detection processing mechanism is adopted to filter out abnormal data in the historical discharge data and the real-time monitoring data, so as to ensure the accuracy of the feature data used in the SOH data prediction process, to obtain filtered historical discharge data and filtered real-time monitoring data.

[0105] Specifically, a detection threshold can be set in advance for the historical discharge data and the real-time detection data, and the abnormal data is determined from the historical discharge data and the real-time detection data by comparing the historical discharge data and the real-time detection data with the detection threshold, so as to realize the filtering of the abnormal data and ensure the accuracy of the feature data.

[0106] For example, a data detection threshold, i.e., a charging time threshold and a current threshold, is set in advance for the historical charging time in the historical discharge data and the real-time current in the real-time detection data; the historical charging time is compared with the charging time threshold, and in the case that the historical charging time is greater than the charging time threshold, the historical charging time is determined as abnormal data and is filtered out; the real-time current is compared with the current threshold, and in the case that the real-time current is greater than the current threshold, the real-time current is determined as abnormal data and is filtered out.

[0107] S1022, respectively, on the filtered historical discharge data and the filtered real-time monitoring data data normalization processing, to obtain normalized historical discharge data and normalized real-time monitoring data.

[0108] In this step, in order to further eliminate the dimensional influence in the historical discharge data and real-time monitoring data, on the basis of the filtered historical discharge data and the filtered real-time monitoring data, the filtered historical discharge data and the filtered real-time monitoring data are normalized respectively, to obtain the normalized historical discharge data and the normalized real-time monitoring data.

[0109] S1023, by analyzing the normalized historical discharge data, determining a plurality of historical voltage data, a plurality of historical current data and historical SOH associated data of the battery to be predicted in a specific monitoring time period.

[0110] In this step, by analyzing the normalized historical discharge data, a plurality of historical voltage data, a plurality of historical current data and historical SOH associated data of the battery to be predicted in a specific monitoring time period are extracted from the normalized historical discharge data; in this way, the plurality of historical voltage data and the plurality of historical current data of the battery to be predicted in a specific monitoring time period can be used to analyze the internal resistance change data of the battery to be predicted in a specific monitoring time period.

[0111] S1024, based on the plurality of historical voltage data and the plurality of historical current data, determining the internal resistance change data of the battery to be predicted in a specific monitoring time period.

[0112] In this step, combined with the correlation between voltage, current and resistance, the plurality of historical voltage data and the plurality of historical current data are used to determine the internal resistance change data R internal .

[0113] S1025, by analyzing the normalized real-time monitoring data, determining the battery attribute data of the battery to be predicted.

[0114] Among them, the battery attribute data includes one or more of battery temperature data Temp t , real-time current data Current t , real-time SOC data SOC t and battery mileage data Mileage t .

[0115] S1026, based on the internal resistance change data and the battery attribute data, determining the battery characteristic data of the battery to be predicted.

[0116] In this step, by capturing the derived variables between various battery attribute data, integrating the internal resistance change data and battery attribute data, the battery characteristic data that can be used to predict the SOH data of the battery to be predicted is determined. The battery characteristic data is used as the input features of the two-layer prediction framework to achieve accurate prediction of the SOH data.

[0117] Here, considering the complex interactions between various battery attribute data, this embodiment of the disclosure also determines the derived variables that exist between two battery attribute data when determining battery feature data.

[0118] In one optional implementation, step S1026 includes:

[0119] S10261. By analyzing the first correlation between the battery temperature data and the real-time SOC data, the first cross-derived variable of the battery to be predicted is determined.

[0120] In this step, battery temperature data and real-time SOC data are used to analyze the first correlation between real-time SOC data and battery temperature data, and the first cross-derived variable Temp of the battery to be predicted is determined. t ×SOC t That is, the first cross-derived variable Temp. t ×SOC t Real-time SOC data used to indicate the state of charge (SOC) of the battery to be predicted. t Battery temperature data Temp t The patterns of change under influence.

[0121] Specifically, the first cross-derived variable, Temp. t ×SOC t For real-time SOC data SOC t With battery temperature data Temp t The product between them.

[0122] S10262. By analyzing the second correlation between the real-time current data and the battery mileage data, the second cross-derived variable of the battery to be predicted is determined.

[0123] In this step, real-time current data and battery mileage data are used to analyze the second correlation between the changes in real-time current data and battery mileage data, thereby determining the second cross-derived variable Current for the battery to be predicted. t ×Mileage t That is, the second cross-derived variable Current t ×Mileage t Current data used to indicate the real-time current of the battery to be predicted tMileage t under the influence of the battery temperature data Temp

[0124] Specifically, the second cross-derivative variable Current t ×Mileage t is a product between the real-time current data Current t and the battery mileage data Mileage t .

[0125] S10263、the internal resistance change data R internal , the battery temperature data Temp t , the real-time current data Current t , the real-time SOC data SOC t , the battery mileage data Mileage t , the first cross-derivative variable Temp t ×SOC t and the second cross-derivative variable Current t ×Mileage t are determined as the battery characteristic data of the battery to be predicted.

[0126] In step S103, the determined historical SOH correlation data of the battery to be predicted is input into an autoregressive model in the double-layer prediction framework, so as to capture the dynamic change characteristics (i.e., performance change characteristics) of the prediction SOH data of the battery to be predicted with time through the autoregressive model, and to predict the SOH prediction sequence of the battery to be predicted in a specific prediction time period.

[0127] The historical SOH correlation data includes historical SOH data, historical error data, etc., and the historical error data refers to the difference between the actual SOH data and the prediction SOH data of the battery to be predicted at a historical prediction time point.

[0128] Here, the autoregressive model (AR) is divided into three parts: a difference layer, an autoregressive layer, and a moving average (MA) layer; wherein the difference layer is used to smooth the SOH prediction sequence obtained by prediction; the autoregressive layer is used to describe the correlation between the current SOH data and the historical SOH correlation data; and the moving average layer is used to describe the linear combination between the current error term and the historical error term.

[0129] In an optional embodiment, step S103 includes:

[0130] S1031, determining, by using the historical SOH data in the historical SOH correlation data, a SOH difference sequence of the battery to be predicted at each first historical prediction time point through a difference layer in the autoregressive model.

[0131] Specifically, the SOH difference sequence of the battery to be predicted at each first historical prediction time point is determined by the following formula:

[0132] Δ d SOH t-1 =Δ d-1 SOH t -Δ d-1 SOH t-n ;

[0133] Wherein, Δ d SOH t-1 is the d-order SOH difference sequence of the battery to be predicted at t-1 moment, Δ d-1 SOH t is the d-1 order SOH difference sequence of the battery to be predicted at t moment, and Δ d-1 SOH t-n is the d-1 order SOH difference sequence of the battery to be predicted at t-n moment.

[0134] S1032, determining, by using the SOH difference sequence of the battery to be predicted at each first historical prediction time point, AR data of the battery to be predicted through an autoregressive layer in the autoregressive model.

[0135] In this step, the SOH difference sequence of the battery to be predicted at each first historical prediction time point is input into the autoregressive layer of the autoregressive model, so as to determine the autoregressive data of the battery to be predicted (i.e., the AR data of the battery to be predicted) through the autoregressive layer. The autoregressive data of the battery to be predicted describes the linear relationship between the SOH difference sequences at each historical prediction time point (e.g., the first historical prediction time point).

[0136] Specifically, the AR data of the battery to be predicted is determined by the following formula:

[0137] Wherein, AR is the AR data of the battery to be predicted, Δ d SOH t-1 is the d-order SOH difference sequence of the battery to be predicted at t-1 moment, Δ d SOH t-2 is the d-order SOH difference sequence of the battery to be predicted at t-2 moment, and Δ d SOH t-p is the d-order SOH difference sequence of the battery to be predicted at t-p moment, is a preset parameter involved in the autoregressive layer.

[0138] is calculated iteratively from the first preset initial value in the training process of the autoregressive model.

[0139] S1033, using the historical error data of the to-be-predicted battery at each second historical prediction time point in the historical SOH correlation data of the to-be-predicted battery, determining the MA data of the to-be-predicted battery through the moving average layer in the autoregressive model.

[0140] In this step, the error value generated by the to-be-predicted battery at each second historical prediction time point (i.e., historical error data) is extracted from the historical SOH correlation data of the to-be-predicted battery, and the historical error data of the to-be-predicted battery at each second historical prediction time point is input to the moving average layer in the autoregressive model to determine the moving average value of the to-be-predicted battery (i.e., the MA data of the to-be-predicted battery) through the moving average layer.

[0141] Specifically, the MA data of the to-be-predicted battery is determined by the following formula:

[0142] MA = θ1ε t-1 + θ2ε t-2 +... + θ q ε t-q ;

[0143] where MA is the MA data of the to-be-predicted battery, θ1, θ2,..., θ q are preset model parameters involved in the moving average layer, ε t-1 is the historical error data of the to-be-predicted battery at t-1 (i.e., the difference between the actual SOH data and the predicted SOH data of the to-be-predicted battery at t-1), ε t-2 is the historical error data of the to-be-predicted battery at t-2 (i.e., the difference between the actual SOH data and the predicted SOH data of the to-be-predicted battery at t-2), and ε t-q is the historical error data of the to-be-predicted battery at t-q (i.e., the difference between the actual SOH data and the predicted SOH data of the to-be-predicted battery at t-q).

[0144] θ1, θ2,..., θ q are calculated iteratively from the second preset initial value in the training process of the autoregressive model.

[0145] S1034, based on the AR data and the MA data, determining the SOH prediction sequence of the to-be-predicted battery within a specific prediction time period.

[0146] In this step, the three parts of the difference layer, the autoregressive layer and the moving average layer are combined to utilize the AR data and the MA data of the battery to be predicted to determine the SOH prediction sequence SOH of the battery to be predicted in a specific prediction time period t,ARIMA .

[0147] Specifically, the SOH prediction sequence SOH of the battery to be predicted in a specific prediction time period is determined by the following formula t,ARIMA :

[0148] wherein, is a constant term, and ε i is an error term of the autoregressive model; and ε i is an error value determined in the training process of the autoregressive model.

[0149] ε i is obtained by iterative calculation from the third preset initial value in the training process of the autoregressive model.

[0150] Here, in order to ensure that the autoregressive model with the best prediction effect is obtained in the training process, various information criteria (such as AIC criterion, BIC criterion) or autocorrelation diagrams (partial autocorrelation diagrams) are used to determine the best autoregressive model.

[0151] In an optional embodiment, the autoregressive model is trained by the following steps:

[0152] Step a, obtaining the training SOH data and the SOH data label of each training sample battery.

[0153] Step b, using the training SOH data and the SOH data label of each training sample battery, training the pre-constructed autoregressive model to obtain a to-be-detected autoregressive model.

[0154] In this step, the training SOH data of each training sample battery is used as an input feature, and the SOH data label of each training sample battery is used as an output feature, and the pre-constructed autoregressive model is trained to obtain a to-be-detected autoregressive model for stationarity detection.

[0155] Step c, using the to-be-detected autoregressive model to predict the SOH difference sequence of the test sample battery.

[0156] Step d, using the ACF test function to draw the autocorrelation diagram (or partial autocorrelation diagram) of the SOH difference sequence of the test sample battery.

[0157] The ACF test function is an autocorrelation function (Autocorrelation Function).

[0158] Step e, determining whether the SOH difference sequence of the test sample battery is stationary by analyzing the autocorrelation graph (or partial autocorrelation graph).

[0159] Step f, if yes, determining that the to-be-detected autoregressive model is a trained autoregressive model, and determining the model coefficients of the autoregressive model.

[0160] Step g, if no, continue training the autoregressive model.

[0161] In step S104, the determined battery feature data and SOH prediction sequence of the to-be-predicted battery are input into the machine learning prediction model of the double-layer prediction framework, so as to integrate the battery feature data and the SOH prediction sequence by using the machine learning prediction model, and obtain the predicted SOH data of the to-be-predicted battery at a specific prediction time point.

[0162] Specifically, in the machine learning prediction model, the predicted SOH data of the to-be-predicted battery at a specific prediction time point is determined by the following formula:

[0163] SOH t = β0+ β1SOH t，ARIMA + β2R internal + β3Temp t + β4Curremt t + β5SOC t + β6Mileage t + β9(Temp t × SOC t ) + β 10 (Current t × Mileage t ) + ∈;

[0164] Wherein, SOH t is the predicted SOH data of the to-be-predicted battery at time t, SOH t,ARIMA is the SOH prediction sequence obtained by using the autoregressive model, β0, β1, β2, β3, β4, β5, β6, β9, β 10 are preset model parameters involved in the machine learning prediction model, and ∈ is an error term.

[0165] Here, the error term ∈ is the sum of the error term determined in the training process of the machine learning prediction model and the error term ε i of the autoregressive model; β0, β1, β2, β3, β4, β5, β6, β9, β 10 are obtained by iterative calculation from the fourth preset initial value in the training process of the machine learning prediction model.

[0166] The machine learning prediction model is any one of a random forest model and a gradient boosting machine.

[0167] In an alternative embodiment, the machine learning prediction model is trained by the following steps:

[0168] Step h, obtaining the key model parameters of the machine learning prediction model by using a grid search method.

[0169] Step i, constructing a machine learning prediction model to be trained according to the key model parameters.

[0170] Step j, training the machine learning prediction model to be trained by using the obtained each training sample battery and the SOH data label of each training sample battery, to obtain a trained machine learning prediction model.

[0171] The key model parameters include one or more of the number of trees, the maximum depth of the tree, and the learning rate.

[0172] For example, when the machine learning prediction model is a random forest model, the random forest model is trained by using the SOH data label of each training sample battery, and the parameter space of the random forest model is systematically searched by using a grid search method combined with a cross-validation method to obtain the optimal key model parameters. The random forest model makes a final decision by constructing multiple decision trees and averaging or majority voting their prediction results, which effectively improves the generalization ability of the model and reduces the overfitting problem of a single decision tree. The random forest model can provide an importance evaluation of each feature on the prediction result, which helps to further understand the driving factors of the battery SOH change.

[0173] The key parameters of the random forest model include the number of trees (n_estimators), the maximum depth of the tree (max_depth), the minimum number of samples required for splitting a node (min_samples_split), etc.

[0174] For example, when the machine learning prediction model is a gradient boosting machine, the gradient boosting machine (GBM) minimizes the prediction error by iteratively training decision trees, and each tree learns the direction of the prediction residual of the previous tree to gradually improve the prediction ability of the model.

[0175] In this embodiment, the GBM model is used to integrate and analyze the battery feature data extracted from the historical discharge data and the real-time monitoring data and the SOH prediction sequence predicted by the autoregressive model (ARIMA model).

[0176] To obtain the best prediction performance, the key parameters of the GBM model are optimized, including the learning rate, the maximum depth of the tree, the minimum number of leaf nodes per tree, etc. The parameter space of the random forest model is systematically searched by the grid search method combined with the cross-validation method to obtain the optimal key model parameters.

[0177] The SOH data label of the selected training sample battery is used to train the constructed GBM model with the optimized parameter configuration. In the training process, the prediction error is minimized by gradually adding decision trees and adjusting the weights until a predetermined number of iterations is reached or the model performance no longer improves significantly.

[0178] The prediction method for the battery state parameter provided by the embodiments of the present disclosure comprises the following steps: obtaining historical discharge data and real-time monitoring data of a battery to be predicted; performing data processing on the historical discharge data and the real-time monitoring data respectively to determine historical SOH correlation data and battery characteristic data of the battery to be predicted; determining an SOH prediction sequence of the battery to be predicted in a specific prediction time period by using the historical SOH correlation data, through an autoregressive model in a double-layer prediction framework, and combining the performance change characteristics of the battery to be predicted; and determining predicted SOH data of the battery to be predicted at a specific prediction time point by using a machine learning prediction model in the double-layer prediction framework based on the battery characteristic data and the SOH prediction sequence. In this way, the predicted SOH data of the battery to be predicted can be predicted by means of the historical discharge data and the real-time monitoring data of the battery to be predicted, and the accuracy of the predicted SOH data can be improved by enriching the data referred to in the prediction. In addition, by means of the double-layer prediction framework integrating the autoregressive model and the machine learning prediction model, the prediction of the predicted SOH data of the battery to be predicted at the specific prediction time point can be realized by capturing the characteristics of the dynamic change of the battery performance over time, so as to further improve the accuracy of the prediction result.

[0179] Please refer to FIG. 2 and FIG. 3, FIG. 2 is a structural schematic diagram of a prediction device for a battery state parameter according to an optional embodiment of the present disclosure, and FIG. 3 is a structural schematic diagram of a prediction device for a battery state parameter according to another optional embodiment of the present disclosure. As shown in FIG. 2, the prediction device 200 comprises:

[0180] A data acquisition module 210 is configured to acquire historical discharge data and real-time monitoring data of a battery to be predicted.

[0181] A data processing module 220 is configured to perform data processing on the historical discharge data and the real-time monitoring data respectively to determine historical SOH correlation data and battery characteristic data of the battery to be predicted.

[0182] The sequence prediction module 230 is configured to determine, by using the historical SOH correlation data, the SOH prediction sequence of the battery to be predicted in a specific prediction time period through an autoregressive model in a double-layer prediction framework, and in combination with the performance change characteristics of the battery to be predicted.

[0183] The SOH prediction module 240 is configured to determine, based on the battery characteristic data and the SOH prediction sequence, the predicted SOH data of the battery to be predicted at a specific prediction time point by using a machine learning prediction model in the double-layer prediction framework.

[0184] Further, when the data processing module 220 is configured to determine the historical SOH correlation data and the battery characteristic data of the battery to be predicted by processing the historical discharge data and the real-time monitoring data respectively, the data processing module 220 is configured to:

[0185] perform data cleaning processing on the historical discharge data and the real-time monitoring data respectively, filter out abnormal data in the historical discharge data and the real-time monitoring data, and obtain filtered historical discharge data and filtered real-time monitoring data;

[0186] perform data normalization processing on the filtered historical discharge data and the filtered real-time monitoring data respectively, and obtain normalized historical discharge data and normalized real-time monitoring data;

[0187] determine a plurality of historical voltage data, a plurality of historical current data, and historical SOH correlation data of the battery to be predicted in a specific monitoring time period by analyzing the normalized historical discharge data;

[0188] determine the internal resistance change data of the battery to be predicted in a specific monitoring time period based on the plurality of historical voltage data and the plurality of historical current data;

[0189] determine the battery attribute data of the battery to be predicted by analyzing the normalized real-time monitoring data;

[0190] determine the battery characteristic data of the battery to be predicted based on the internal resistance change data and the battery attribute data.

[0191] Further, the battery attribute data includes one or more of battery temperature data, real-time current data, real-time SOC data, and battery mileage data.

[0192] Further, when the data processing module 220 is configured to determine the battery characteristic data of the battery to be predicted based on the internal resistance change data and the battery attribute data, the data processing module 220 is configured to:

[0193] determine a first cross-derivative variable of the battery to be predicted by analyzing a first correlation between the battery temperature data and the real-time SOC data;

[0194] determine a second cross-derivative variable of the battery to be predicted by analyzing a second correlation between the real-time current data and the battery mileage data;

[0195] determine the battery characteristic data of the battery to be predicted by the internal resistance change data, the battery temperature data, the real-time current data, the real-time SOC data, the battery mileage data, the first cross-derivative variable and the second cross-derivative variable.

[0196] Further, when the sequence prediction module 230 is used to determine the SOH prediction sequence of the battery to be predicted in a specific prediction time period by using the historical SOH correlation data, through the autoregressive model in the double-layer prediction framework, and combining the performance change characteristics of the battery to be predicted, the sequence prediction module 230 is used for:

[0197] determine the SOH difference sequence of the battery to be predicted at each first historical prediction time point by using the historical SOH data in the historical SOH correlation data, through the difference layer in the autoregressive model;

[0198] determine the AR data of the battery to be predicted by using the SOH difference sequence of the battery to be predicted at each first historical prediction time point, through the autoregressive layer in the autoregressive model;

[0199] determine the MA data of the battery to be predicted by using the historical error data of the battery to be predicted at each second historical prediction time point in the historical SOH correlation data, through the moving average layer in the autoregressive model;

[0200] determine the SOH prediction sequence of the battery to be predicted in a specific prediction time period based on the AR data and the MA data.

[0201] Further, as shown in FIG. 3, the prediction device 200 further comprises a model training module 250, and the model training module 250 is used to train the autoregressive model by the following steps:

[0202] obtain the training SOH data and the SOH data label of each training sample battery;

[0203] obtain the training SOH data and the SOH data label of each training sample battery;

[0204] predicting, by using the to-be-detected auto-regressive model, a SOH difference sequence of the test sample battery;

[0205] drawing an autocorrelation graph of the SOH difference sequence of the test sample battery by using an ACF test function;

[0206] determining, by analyzing the autocorrelation graph, whether the SOH difference sequence of the test sample battery is stationary;

[0207] if yes, determining that the to-be-detected auto-regressive model is a trained auto-regressive model, and determining model coefficients of the auto-regressive model.

[0208] Further, the machine learning prediction model is any one of a random forest model and a gradient boosting machine.

[0209] Further, the model training module 250 is configured to train the machine learning prediction model by the following steps:

[0210] obtaining key model parameters of the machine learning prediction model by using a grid search method;

[0211] constructing a to-be-trained machine learning prediction model according to the key model parameters;

[0212] training the to-be-trained machine learning prediction model by using each training sample battery and the SOH data label of each training sample battery, to obtain a trained machine learning prediction model.

[0213] Further, the key model parameters include one or more of the number of trees, the maximum depth of the tree, and the learning rate.

[0214] The prediction device for the battery state parameter provided in the embodiments of the present disclosure obtains historical discharge data and real-time monitoring data of a battery to be predicted; performs data processing on the historical discharge data and the real-time monitoring data respectively, and determines historical SOH correlation data and battery characteristic data of the battery to be predicted; determines an SOH prediction sequence of the battery to be predicted in a specific prediction time period by using the historical SOH correlation data, through an autoregressive model in a double-layer prediction framework, and in combination with performance change characteristics of the battery to be predicted; and determines predicted SOH data of the battery to be predicted at a specific prediction time point based on the battery characteristic data and the SOH prediction sequence, and by using a machine learning prediction model in the double-layer prediction framework. In this way, the predicted SOH data of the battery to be predicted can be predicted by means of the historical discharge data and the real-time monitoring data of the battery to be predicted, and then the accurate prediction of the predicted SOH data can be realized by enriching the data referred to in the prediction. In addition, by means of the double-layer prediction framework integrating the autoregressive model and the machine learning prediction model, the prediction of the predicted SOH data of the battery to be predicted at the specific prediction time point can be realized by capturing the characteristics of the dynamic change of the battery performance over time, so as to further improve the accuracy of the prediction result.

[0215] Please refer to FIG. 4, which is a structural schematic diagram of an electronic device provided in an optional embodiment of the present disclosure. As shown in FIG. 4, the electronic device 400 includes a processor 410, a memory 420 and a bus 430.

[0216] The memory 420 stores machine readable instructions executable by the processor 410. When the electronic device 400 is running, the processor 410 and the memory 420 communicate with each other through the bus 430. When the machine readable instructions are executed by the processor 410, the steps of the prediction method for the battery state parameter in the method embodiment shown in FIG. 1 can be performed. For specific implementation manners, please refer to the method embodiment, which will not be described here again.

[0217] The embodiments of the present disclosure further provide a computer readable storage medium, which stores a computer program. When the computer program is run by a processor, the steps of the prediction method for the battery state parameter in the method embodiment shown in FIG. 1 can be performed. For specific implementation manners, please refer to the method embodiment, which will not be described here again.

[0218] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here again.

[0219] In several embodiments provided by the present disclosure, it should be understood that the disclosed system, device and method can be implemented in other manners. The described device embodiments are merely illustrative, for example, the division of the units is only a logical function division, and there can be another division manner in actual implementation, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different units, or the among different units, can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0220] The units described as separated components can or can not be physically separated, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. In actual implementation, some or all of the units can be selected according to actual needs to achieve the purposes of the embodiments.

[0221] In addition, each functional unit in the various embodiments of the present disclosure can be integrated in one processing unit, or each unit can exist physically as a separate unit, or two or more units can be integrated in one unit.

[0222] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer readable storage medium of a processor. Based on such understanding, the technical solutions of the present disclosure, essentially or in part, or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present disclosure. The foregoing storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, and various other media that can store program codes.

[0223] Finally, it should be noted that the above-described embodiments are merely specific embodiments of the present disclosure, used to illustrate the technical solutions of the present disclosure, and are not intended to limit the present disclosure. The protection scope of the present disclosure is not limited thereto. Although the present disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can make modifications or easy changes to the technical solutions described in the foregoing embodiments, or easily think of changes or equivalent replacements for some of the technical features; and these modifications, changes or replacements do not cause the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should be covered within the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be subject to the protection scope of the claims.

Claims

1. A method for predicting a battery state parameter, the method comprising: obtaining historical discharge data and real-time monitoring data of a battery to be predicted; respectively processing the historical discharge data and the real-time monitoring data to determine historical SOH correlation data and battery characteristic data of the battery to be predicted; using the historical SOH correlation data, determining an SOH prediction sequence of the battery to be predicted within a specific prediction time period through an autoregressive model in a double-layer prediction framework, in combination with performance change characteristics of the battery to be predicted; based on the battery characteristic data and the SOH prediction sequence, using a machine learning prediction model in the double-layer prediction framework to determine predicted SOH data of the battery to be predicted at a specific prediction time point.

2. The prediction method of claim 1, wherein, the respectively processing the historical discharge data and the real-time monitoring data to determine historical SOH correlation data and battery characteristic data of the battery to be predicted comprises: respectively performing data cleaning processing on the historical discharge data and the real-time monitoring data to filter out abnormal data in the historical discharge data and the real-time monitoring data, to obtain filtered historical discharge data and filtered real-time monitoring data; respectively performing data normalization processing on the filtered historical discharge data and the filtered real-time monitoring data to obtain normalized historical discharge data and normalized real-time monitoring data; determining a plurality of historical voltage data, a plurality of historical current data, and historical SOH correlation data of the battery to be predicted within a specific monitoring time period by analyzing the normalized historical discharge data; based on the plurality of historical voltage data and the plurality of historical current data, determining internal resistance change data of the battery to be predicted within a specific monitoring time period; determining battery attribute data of the battery to be predicted by analyzing the normalized real-time monitoring data; based on the internal resistance change data and the battery attribute data, determining battery characteristic data of the battery to be predicted.

3. The prediction method of claim 2, wherein, The battery attribute data includes one or more of battery temperature data, real-time current data, real-time SOC data, and battery mileage data.

4. The prediction method of claim 3, wherein, The determining battery characteristic data of the battery to be predicted based on the internal resistance change data and the battery attribute data comprises: determining a first cross-derivative variable of the battery to be predicted by analyzing a first correlation between the battery temperature data and the real-time SOC data; determining a second cross-derivative variable of the battery to be predicted by analyzing a second correlation between the real-time current data and the battery mileage data; determining the internal resistance change data, the battery temperature data, the real-time current data, the real-time SOC data, the battery mileage data, the first cross-derivative variable, and the second cross-derivative variable as the battery characteristic data of the battery to be predicted.

5. The prediction method of claim 1, wherein, the using the historical SOH correlation data, determining an SOH prediction sequence of the battery to be predicted within a specific prediction time period through an autoregressive model in a double-layer prediction framework, in combination with performance change characteristics of the battery to be predicted comprises: determining, by using the historical SOH data in the historical SOH correlation data, a SOH difference sequence of the battery to be predicted at each first historical prediction time point through a difference layer in the autoregressive model; determining, by using the SOH difference sequence of the battery to be predicted at each first historical prediction time point, AR data of the battery to be predicted through an autoregressive layer in the autoregressive model; determining, by using the historical error data of the battery to be predicted at each second historical prediction time point in the historical SOH correlation data, MA data of the battery to be predicted through a moving average layer in the autoregressive model; determining a SOH prediction sequence of the battery to be predicted within a specific prediction time period based on the AR data and the MA data.

6. The prediction method of claim 1, wherein, The autoregressive model is trained by the following steps: obtaining training SOH data and SOH data labels of each training sample battery; obtaining the autoregressive model to be detected by using the training SOH data and the SOH data labels of each training sample battery to train the pre-constructed autoregressive model; predicting a SOH difference sequence of a test sample battery by using the autoregressive model to be detected; drawing an autocorrelation graph of the SOH difference sequence of the test sample battery by using an ACF test function; determining whether the SOH difference sequence of the test sample battery is stationary by analyzing the autocorrelation graph; if yes, determining that the autoregressive model to be detected is a trained autoregressive model, and determining model coefficients of the autoregressive model.

7. The prediction method of claim 1, wherein, The machine learning prediction model is any one of a random forest model and a gradient boosting machine.

8. The prediction method according to claim 1 or 7, wherein, The machine learning prediction model is trained by the following steps: obtaining key model parameters of the machine learning prediction model by using a grid search method; constructing a machine learning prediction model to be trained according to the key model parameters; obtaining a trained machine learning prediction model by training the machine learning prediction model to be trained by using the obtained SOH data of each training sample battery and the SOH data labels of each training sample battery.

9. The prediction method of claim 8, wherein, The key model parameters include one or more of the number of trees, the maximum depth of the tree, and the learning rate.

10. A battery state parameter prediction device, the prediction device comprising: a data acquisition module configured to acquire historical discharge data and real-time monitoring data of a battery to be predicted; a data processing module configured to process the historical discharge data and the real-time monitoring data respectively to determine historical SOH correlation data and battery characteristic data of the battery to be predicted; a sequence prediction module configured to determine a SOH prediction sequence of the battery to be predicted within a specific prediction time period by using the historical SOH correlation data, through an autoregressive model in a double-layer prediction framework, and combining performance change characteristics of the battery to be predicted; a SOH prediction module configured to determine predicted SOH data of the battery to be predicted at a specific prediction time point by using a machine learning prediction model in the double-layer prediction framework based on the battery characteristic data and the SOH prediction sequence. ​ 11. An electronic device comprising: A processor, a memory, and a bus, the memory storing machine readable instructions executable by the processor, the processor in communication with the memory via the bus when the electronic device is running, the machine readable instructions, when executed by the processor, performing the steps of the method for predicting a battery state parameter of any of claims 1-9.

12. A computer readable storage medium having stored thereon a computer program, the computer program, when executed by a processor, performing the steps of the method for predicting a battery state parameter of any of claims 1-9.

Images