CN-121978573-A - SOH online estimation method and device, electronic equipment and storage medium

CN121978573ACN 121978573 ACN121978573 ACN 121978573ACN-121978573-A

Abstract

The embodiment of the invention discloses an SOH online estimation method, an SOH online estimation device, electronic equipment and a storage medium, and relates to the technical field of battery health state SOH online estimation. And matching and screening the rest point combination through a cross-day state continuation mechanism, calculating the SOC variation, and deriving capacity fading based on physical rules (such as ampere-hour integral) to generate an SOH estimated value and a confidence level assessment. And extracting statistical features from the historical data, training a random forest model to predict SOH values, and estimating and quantifying uncertainty through outside bags. And according to the confidence and uncertainty of the physical rule and the machine learning output, adopting self-adaptive weighted fusion or attenuation weight filtering to carry out moving average processing on the historical data so as to obtain an SOH value. The invention effectively solves the problems of low data utilization rate and poor reliability in the prior art.

Inventors

YANG SHU
YAO YUCHENG
GUO JIWEI

Assignees

深圳织算科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260407

Claims (10)

1. An SOH online estimation method, comprising: Collecting operation data of a battery in real time, preprocessing, and intelligently detecting SOC (system on chip) standing points according to the operation data through a multi-mode standing point identification mechanism to obtain a standing point set, wherein the operation data comprises voltage, current, SOC, temperature and time stamp; Dynamically calibrating the standing point set by using a pre-established OCV-SOC relation curve, and matching and screening the standing points by combining a set cross-day state continuation mechanism to obtain a matched standing point combination and an SOC variation, wherein the standing point information comprises type, time and SOC; Calculating capacity attenuation according to the SOC variation through a physical rule to obtain an SOH estimated value and evaluate confidence coefficient, and simultaneously predicting the SOH predicted value through training a random forest model and estimating an uncertainty index predicted by a quantization model outside a bag; And carrying out weighted fusion on the SOH estimated value and the SOH predicted value according to the confidence coefficient and the uncertainty index through self-adaptive weighted fusion to obtain the SOH value, or carrying out moving average processing on historical data through attenuation weight filtering to obtain the SOH value.
2. The SOH online estimation method as set forth in claim 1, wherein the collecting and preprocessing operation data of the battery in real time includes: Continuously monitoring each key operation parameter of the battery in real time by utilizing a high-precision sensor, and denoising the operation parameter by adopting a filtering technology, wherein the filtering technology comprises moving average filtering or Kalman filtering; filling the missing value in the operation parameters by using an interpolation method, aligning data according to the time stamp of the operation parameters, and smoothing the operation parameters by low-pass filtering.
3. The SOH online estimation method according to claim 1, wherein the intelligent detection of SOC resting points according to the operation data by the multimode resting point identification mechanism to obtain a set of resting points includes: Traversing the operation data, and when the battery is suddenly changed from a high-current working state to zero current and continuously kept standing to reach a stable threshold value, identifying a high-SOC standing point if the SOC value is higher than a set percentage at the moment; and if the SOC value is lower than the set percentage, identifying a low SOC standing point, and screening and verifying the high SOC standing point and the low SOC standing point to obtain a standing point set, wherein the high current working state means that the absolute value of the current is larger than the set value.
4. The SOH online estimation method according to claim 1, wherein dynamically calibrating the set of standing points by using a pre-established OCV-SOC relationship curve, and matching and screening the standing points by combining a set cross-day state continuation mechanism to obtain a paired standing point combination and an SOC variation, includes: dynamically calibrating each standing point in the standing point set by utilizing an OCV-SOC relation curve established in advance through experiments, and inquiring in the OCV-SOC relation table through an interpolation algorithm to obtain the SOC value of each standing point; And (3) designing a cross-day state continuation mechanism to store the information of the rest points, matching the high SOC rest points with the low SOC rest points by adopting a bidirectional pairing strategy to obtain a plurality of pairs of rest point combinations, calculating the SOC difference value between the rest point combinations, and calculating the SOC variation of the combination of which the SOC difference value meets the set condition.
5. The SOH online estimation method as set forth in claim 1, wherein said calculating capacity fade from said SOC variation amount by physical rule, obtaining an SOH estimation value and evaluating a confidence level, comprises: calculating capacity attenuation of the battery based on a physical rule and the SOC variation combined with the rest point, and deriving a current SOH estimated value by combining state data of the battery, wherein the state data comprises initial capacity and charge and discharge efficiency; And setting a confidence coefficient evaluation system according to the confidence coefficient factors, carrying out comprehensive confidence coefficient evaluation on the SOH estimated value according to the confidence coefficient evaluation system, and introducing a scaling factor and a mapping function to obtain the confidence coefficient, wherein the confidence coefficient factors comprise measurement errors, model uncertainty indexes and data quality.
6. The SOH online estimation method as set forth in claim 1, wherein the predicting the SOH prediction value by training a random forest model and estimating the uncertainty index predicted by the quantization model outside the bag comprises: Extracting statistical characteristics closely related to battery aging from historical operation data of the battery to construct a characteristic vector, wherein the statistical characteristics comprise voltage characteristics, current characteristics, SOC characteristics, temperature characteristics and energy characteristics; Using SOH values meeting set standards in the historical operation data as training labels, training a random forest regression model by adopting a machine learning algorithm, and optimizing model super-parameters through grid search and cross verification; and inputting the feature vector into the trained random forest regression model to obtain an SOH predicted value, and predicting an uncertainty index by using an out-of-bag estimation built in a random forest or a prediction variance quantization model between trees.
7. The SOH online estimation method according to claim 1, wherein the step of obtaining the SOH value by performing weighted fusion on the SOH estimation value and the SOH prediction value according to the confidence level and the uncertainty index through adaptive weighted fusion, or obtaining the SOH value by performing moving average processing on the historical data through attenuation weight filtering comprises: If only the SOH estimated value is valid, the SOH estimated value is used as an SOH value, and if only the SOH predicted value is valid, the SOH predicted value is used as an SOH value; If the SOH estimated value and the SOH predicted value are both effective, carrying out dynamic weighted fusion on the SOH estimated value and the SOH predicted value according to the confidence coefficient and the uncertainty index to obtain an SOH value; And if the SOH estimated value and the SOH predicted value are invalid, carrying out exponential decay moving average processing on the weight of the historical data according to the time sequence by adopting decay weight filtering to obtain the SOH value.
8. An SOH online estimation apparatus, comprising: The system comprises a data acquisition preprocessing module, a multi-mode static point identification mechanism, a data processing module and a data processing module, wherein the data acquisition preprocessing module is used for acquiring and preprocessing operation data of a battery in real time, and intelligently detecting the SOC static points according to the operation data to obtain a static point set; the static point calibration matching module is used for dynamically calibrating the static point set by utilizing a pre-established OCV-SOC relation curve, and matching and screening the static points by combining a set cross-day state continuation mechanism to obtain a matched static point combination and an SOC variation; The rule model bi-prediction module is used for calculating capacity attenuation according to the SOC variation through a physical rule to obtain an SOH estimated value and evaluate the confidence coefficient, and simultaneously predicting the SOH predicted value through training a random forest model and estimating an uncertainty index predicted by a quantization model outside a bag; And the SOH online estimation module is used for carrying out weighted fusion on the SOH estimation value and the SOH predicted value according to the confidence coefficient and the uncertainty index through self-adaptive weighted fusion to obtain an SOH value, or carrying out moving average processing on historical data through attenuation weight filtering to obtain the SOH value.
9. An electronic device comprising at least one processor and at least one memory, wherein, The memory has computer readable instructions stored thereon; The computer readable instructions are executed by one or more of the processors to cause an electronic device to implement the SOH online estimation method of any one of claims 1 to 7.
10. A storage medium having stored thereon computer readable instructions, the computer readable instructions being executable by one or more processors to implement the SOH online estimation method of any one of claims 1 to 7.

Description

SOH online estimation method and device, electronic equipment and storage medium Technical Field The present invention relates to the field of online estimation of SOH of battery state of health, and in particular, to an online estimation method, apparatus, electronic device, and storage medium for SOH. Background Along with the large-scale participation of the energy storage power station in the power grid frequency modulation service, the battery is in a frequent and irregular interweaved charge-discharge state for a long time, and the complexity of the frequency modulation working condition provides a serious challenge for the accurate estimation of the state of health (SOH) of the battery. However, the traditional SOH estimation method exposes significant limitations when dealing with such complex scenes, namely, although a method based on a physical model (such as an ampere-hour integration method and an OCV-SOC curve method) has clear physical significance, under the frequency modulation working condition of discontinuous charge and discharge interweaving and complete circulation, effective data fragments are difficult to capture, so that the estimation failure rate is increased suddenly and the data utilization rate is low, while a method based on data driving (such as a neural network and a random forest) can mine nonlinear relations in massive data, but has poor interpretability due to lack of physical constraint, and is easy to deviate from physical common knowledge when the data quality is poor or a new working condition is encountered, so that the reliability is questionable. In addition, the traditional full capacity calibration method fails because the frequency modulation station is difficult to execute full charge and full discharge test, the error of the conventional ampere-hour integration method rapidly diverges along with high-frequency circulation, the single OCV method has insufficient precision, the ICA/DVA analysis cannot be applied online, and the Kalman filtering joint estimation method fails because of high-frequency drift of model parameters. The prior art either depends on ideal fragments or is limited by a static model, and the dynamic property and fragmentation characteristic of the frequency modulation working condition cannot be considered. Therefore, there is an urgent need for a highly reliable SOH online estimation method that can integrate physical rules with data-driven advantages and adapt to complex conditions. Disclosure of Invention The embodiment of the invention provides an SOH online estimation method, which aims to solve the problems that in the prior art, the data utilization rate is low, the estimation failure rate is high and the method cannot adapt to high-frequency dynamic working conditions in a frequency modulation fragmentation scene. The technical scheme is as follows: According to one aspect of the invention, the SOH online estimation method comprises the steps of collecting running data of a battery in real time, preprocessing, intelligently detecting SOC standing points according to the running data through a multi-mode standing point identification mechanism to obtain a standing point set, wherein the running data comprises voltage, current, SOC, temperature and time stamp, preprocessing comprises denoising, interpolation, alignment and filtering, dynamically calibrating the standing point set through a pre-established OCV-SOC relation curve, matching and screening the standing points through a set cross-day state continuation mechanism to obtain a matched standing point combination and SOC variation, calculating capacity attenuation according to the SOC variation through a physical rule to obtain an SOH estimation value and evaluating confidence, predicting the SOH prediction value through a random forest model, estimating an uncertainty index of the quantization model through the outside of the bag, integrating the SOH estimation value and the SOH prediction value through self-adaptive weighted fusion, and carrying out fusion on the SOH estimation value and the SOH prediction value through self-adaptive weighted fusion or the SOH attenuation value through the fuzzy fusion historical data. In one embodiment, the operation data of the battery are collected in real time and preprocessed by continuously monitoring each key operation parameter of the battery in real time by using a high-precision sensor, denoising the operation parameters by adopting a filtering technology, filling the missing value in the operation parameters by using an interpolation method, aligning the data according to the time stamp of the operation parameters, and smoothing the operation parameters by using a low-pass filter. In one embodiment, intelligent detection is performed on the SOC rest points according to the operation data through a multi-mode rest point identification mechanism, and the obtained rest point set is achieved through the steps that the operation data are