Lithium ion battery residual life online prediction method based on deep learning algorithm

Abstract

The invention discloses a lithium ion battery residual life on-line prediction method based on a deep learning algorithm, and the method comprises the steps: training a deep learning network through a large amount of offline battery data sets, taking a discharge curve segment collected at an online stage as data input, and carrying out the online prediction of the residual life of a lithium ion battery. Based on the deep learning network optimized in the off-line stage, the remaining life of the battery is obtained by combining a signal processing method and a transfer learning technology, and then the health state of the battery in online use can be deduced and battery management is carried out; different from traditional pure theoretical calculation and empirical model prediction, the residual life prediction model is based on the deep learning network, dependence on physical and chemical models and mathematical models in the battery is avoided, the robustness and universality of the prediction method are improved, and meanwhile the calculation cost is effectively reduced.

Description

technical field [0001] The invention belongs to the lithium ion battery technology, and relates to an online prediction method for the remaining life of a lithium ion battery based on a deep learning algorithm. Background technique [0002] In the actual application process of batteries, the complete charging and discharging process is usually not easy to achieve. The lithium-ion battery online data sampling system can only collect fragment discharge information (including voltage, capacity changes, etc.) of the battery, while the current round of charging and discharging of the battery The maximum discharge capacity of the process is often uncertain, so its remaining lifetime can only be estimated by collecting partial discharge curve segments. The existing battery remaining life prediction is mainly based on the complete charging and discharging process, that is, the maximum discharge capacity of each charging and discharging cycle needs to be known, which limits the pract...

AI Technical Summary

Problems solved by technology

[0002] In the actual application process of the battery, the complete charging and discharging process is usually not easy to realize. The lithium-ion battery online data sampling system can only collect the fragmented discharge information of the battery (including voltage, capacity change, etc.), and the current round of charging and discharging of the battery The maximum discharge capacity of a process is often uncertain, so its remaining lifetime can only be estimated by collecting partial discharge curve segments

The existing remaining battery life prediction is mainly based on the complete charge and discharge process, that is, the maximum discharge capacity of each charge and discharge cycle needs to be known, which limits the practical application of battery state estimation

At the same time, in the process of battery capacity fading, due to different operating conditions, there is an uncontrollable capacity regeneration phenomenon, and these noises will cause the battery management system to be unable to more effectively grasp the main trend of battery capacity fading

[0003] In addition, building a deep learning model from scratch to train and predict the online battery that needs to be evaluated will undoubtedly increase the calculation cost and the amount of data collected online, and reduce the prediction accuracy

Images