CN-122017581-A - Construction method of state prediction model of battery system and related device

CN122017581ACN 122017581 ACN122017581 ACN 122017581ACN-122017581-A

Abstract

The application provides a method and a related device for constructing a state prediction model of a battery system, wherein a first prediction model is constructed according to accelerated degradation experimental data of the battery system and battery system data, the first prediction model is used for predicting internal state parameters of the battery system, a second prediction model is constructed according to charging characteristic data, time sequence characteristic data and internal state characteristic data of the battery system, the internal state characteristic data is determined according to the internal state parameters, the second prediction model is used for predicting a health index sequence of the battery system and an uncertainty range, the health index sequence is used for reflecting life attenuation trend of the battery system, and the uncertainty range is used for reflecting standard deviation of each health index in the health index sequence. The thermodynamic physical model can be used as a priori condition, and the constraint condition is embedded into the neural network model, so that the state prediction accuracy of the battery system is greatly improved.

Inventors

LI LONG
PENG YANQIU
YUAN DINGDING

Assignees

武汉亿纬储能有限公司

Dates

Publication Date: 20260512
Application Date: 20260128

Claims (10)

1. A method for constructing a state prediction model of a battery system, the method comprising: Constructing a first prediction model according to the accelerated degradation experimental data of the battery system and the battery system data, wherein the first prediction model is used for predicting internal state parameters of the battery system; And constructing a second prediction model according to the charging characteristic data, the time sequence characteristic data and the internal state characteristic data of the battery system, wherein the internal state characteristic data is determined according to the internal state parameters, the second prediction model is used for predicting a health index sequence of the battery system and an uncertainty range, the health index sequence is used for reflecting the life decay trend of the battery system, and the uncertainty range is used for reflecting the standard deviation of each health index in the health index sequence.
2. The method of claim 1, wherein constructing a first predictive model from the accelerated degradation experimental data of the battery system and the battery system data comprises: Acquiring observed degradation data of the battery system under a plurality of environmental stresses; fitting the observed degradation amount data by adopting a parameterized model to obtain pseudo life data when the battery system is degraded to a preset failure threshold value; Constructing an equivalent life model according to the pseudo life data and the plurality of environmental stresses, wherein the equivalent life model is used for reflecting the mapping relation between the pseudo life data and the plurality of environmental stresses; and updating parameters of the equivalent life model according to the battery system data and a filtering algorithm to obtain the first prediction model.
3. The method of claim 2, wherein said constructing an equivalent life model from said pseudo life data and said plurality of environmental stresses comprises: Determining a characteristic lifetime from the pseudo lifetime data; Converting the characteristic life into a position parameter; Converting the reference environmental stress into a reference parameter, and converting the environmental stresses into a plurality of influence parameters, wherein each influence parameter comprises an influence weight, and each influence weight is obtained by fitting according to the accelerated degradation experiment data; And taking the position parameter as a result value, and summing the reference parameter and the influence parameters to construct the equivalent life model.
4. The method of any of claims 1-3, wherein the constructing a second predictive model from the charging characteristic data, the time series characteristic data, and the internal state characteristic data of the battery system comprises: Determining the charging characteristic data according to the charging curve data of the battery system; determining the time sequence characteristic data according to the time sequence data of the battery system; determining the internal state characteristic data according to the internal state parameters; determining feature vector data from the charging feature data, the time sequence feature data, and the internal state feature data; And inputting the feature vector data into a preset prediction model, and training the preset prediction model by adopting a mixing loss function to obtain the second prediction model.
5. The method of claim 4, wherein the hybrid loss function comprises a first loss function, a second loss function, and a third loss function, the method further comprising: Determining a first loss function according to first error data of output data and actual observation data of the preset prediction model; determining a second loss function according to second error data of output data of the preset prediction model and derived data, wherein the derived data is determined according to internal state parameters output by the first prediction model; Carrying out normalization processing on the weight regularization loss function to obtain a third loss function; And carrying out weighted summation on the first loss function, the second loss function and the third loss function to obtain the mixed loss function.
6. The method of any one of claims 1-5, wherein after the constructing a second predictive model from the charging characteristic data, the time series characteristic data, and the internal state characteristic data of the battery system, the method further comprises: If the health state data of the battery system is higher than a first health state threshold value but lower than a second health state threshold value, the health index sequence and the uncertainty range are used as priori data, and a wiener process degradation model is constructed; And performing Kalman filtering and Bayesian updating on the wiener process degradation model according to the battery system data so as to adjust the wiener process degradation model in real time.
7. The method of claim 6, wherein constructing a wiener process degradation model using the health index sequence and the uncertainty range as a priori data comprises: Obtaining a wiener process equation according to an integral term of the time-varying drift rate, an initial health index, a health index at a corresponding moment, a random fluctuation term of the wiener process and an observation noise term; discretizing the wiener process equation to obtain a state equation and an observation equation.
8. The method of claim 6, wherein after performing a kalman filter and a bayesian update on the wiener process degradation model based on the battery system data to adjust the wiener process degradation model in real time, the method further comprises: Determining the predicted residual life, the first degradation speed and the first uncertainty corresponding to the current moment through the output of the second prediction model; Determining an inverse Gaussian distribution probability density function corresponding to the predicted remaining life according to the first degradation speed and the first uncertainty; determining a second degradation rate and a second uncertainty from an output of the wiener process degradation model; Determining a posterior distribution probability density function corresponding to the predicted remaining life according to the second degradation speed and the second uncertainty; And determining a probability density function corresponding to the predicted residual life according to the inverse Gaussian distribution probability density function and the posterior distribution probability density function.
9. An electronic device comprising a processor, a memory, and one or more programs stored in the memory and configured to be executed by the processor, the programs comprising instructions for performing the steps in the method of any of claims 1-8.
10. A computer storage medium storing a computer program comprising program instructions which, when executed by a processor, cause the processor to perform the method of any of claims 1-8.

Description

Construction method of state prediction model of battery system and related device Technical Field The application relates to the technical field of battery state prediction, in particular to a method and a related device for constructing a state prediction model of a battery system. Background Currently, state predictions for battery systems include predictions of remaining useful life, which refers to the length of time a device or system maintains an intended function under a particular operating condition. The prediction method is mainly divided into two types of data driving and model driving, namely a deep learning model captures degradation characteristics by analyzing time sequence data of a sensor, and a statistical model establishes probability distribution based on degradation tracks. In the field of lithium batteries, correlation modeling of focus capacity attenuation and charging characteristics is studied, and prediction accuracy is improved by adopting methods such as improved support vector regression. The prior art generally faces challenges such as adaptability to complex working conditions, extraction efficiency of degradation features and the like. Disclosure of Invention In view of the above, the application provides a method for constructing a state prediction model of a battery system and a related device, wherein a thermodynamic physical model is used as a priori condition, and constraint conditions are embedded into a neural network model, so that the state prediction accuracy of the battery system is greatly improved. In a first aspect, an embodiment of the present application provides a method for constructing a state prediction model of a battery system, where the method includes: Constructing a first prediction model according to the accelerated degradation experimental data of the battery system and the battery system data, wherein the first prediction model is used for predicting internal state parameters of the battery system; And constructing a second prediction model according to the charging characteristic data, the time sequence characteristic data and the internal state characteristic data of the battery system, wherein the internal state characteristic data is determined according to the internal state parameters, the second prediction model is used for predicting a health index sequence of the battery system and an uncertainty range, the health index sequence is used for reflecting the life decay trend of the battery system, and the uncertainty range is used for reflecting the standard deviation of each health index in the health index sequence. In one possible embodiment, the constructing a first prediction model according to the accelerated degradation experiment data of the battery system and the battery system data includes: Acquiring observed degradation data of the battery system under a plurality of environmental stresses; fitting the observed degradation amount data by adopting a parameterized model to obtain pseudo life data when the battery system is degraded to a preset failure threshold value; Constructing an equivalent life model according to the pseudo life data and the plurality of environmental stresses, wherein the equivalent life model is used for reflecting the mapping relation between the pseudo life data and the plurality of environmental stresses; and updating parameters of the equivalent life model according to the battery system data and a filtering algorithm to obtain the first prediction model. In a possible embodiment, said constructing an equivalent lifetime model from said pseudo lifetime data and said plurality of environmental stresses comprises: Determining a characteristic lifetime from the pseudo lifetime data; Converting the characteristic life into a position parameter; Converting the reference environmental stress into a reference parameter, and converting the environmental stresses into a plurality of influence parameters, wherein each influence parameter comprises an influence weight, and each influence weight is obtained by fitting according to the accelerated degradation experiment data; And taking the position parameter as a result value, and summing the reference parameter and the influence parameters to construct the equivalent life model. In one possible embodiment, the constructing a second prediction model according to the charging characteristic data, the time series characteristic data and the internal state characteristic data of the battery system includes: Determining the charging characteristic data according to the charging curve data of the battery system; determining the time sequence characteristic data according to the time sequence data of the battery system; determining the internal state characteristic data according to the internal state parameters; determining feature vector data from the charging feature data, the time sequence feature data, and the internal state feature data; And inputting the feature ve