Research On SOC And SOH Prediction Of Lithium Battery Based On Machine Learning Algorithm

In recent years,people have paid more and more attention to environmental pollution and energy regeneration,non-renewable energy such as fossil fuels has caused serious pollution to the earth’s environment,and has a huge negative impact on climate change and people’s daily life.In order to solve this problem,people began to find and use clean energy to reduce environmental pollution and dependence on non-renewable energy.In this context,the new energy industry is developing rapidly.More and more people are beginning to realize that the use of new energy to replace traditional nonrenewable energy has become a mainstream trend,especially in the new energy vehicle industry,major automobile manufacturers are increasing their investment in research and development of new energy vehicles.The development of new energy industry cannot be separated from energy storage technology.Among them,lithium-ion battery as a clean secondary energy storage battery is favored by the new energy industry.Compared with other secondary batteries,lithium-ion batteries have many advantages.For example,lithium-ion batteries with the same size and volume have a higher energy density and excellent cycle performance.Even after repeated charging and discharging,they can still maintain good battery health.Therefore,lithium-ion batteries are widely used in aerospace,military equipment,automobiles and household appliances.However,due to the continuous occurrence of internal electrochemical reaction,lithium-ion batteries will age during the long-term cycle of use,which will lead to the battery capacity degradation and eventually lose its use value,and even lead to dangerous situations such as fire caused by the leakage of battery electrolyte.Therefore,it is of great safety significance to predict the state of charge(SOC)and state of health(SOH)of lithium ion batteries.During the use of lithium-ion batteries,a series of chemical changes occurred inside the batteries are very complex.The common model-based SOC and SOH prediction methods are difficult to reflect the nonlinear characteristics of the internal reactions of the batteries.This paper uses the stochastic forest algorithm in machine learning methods to establish the SOC and SOH prediction model of lithium-ion batteries.This method is a data-driven modeling method.The prediction model based on random forest has a good effect in solving the problem of nonlinear high-latitude pattern recognition,and is the main focus of research on this kind of problem at present.Because the manufacturing process of the battery itself is different,and the operation status of the battery in the actual use process cannot be unified,in order to ensure the reliability of the experimental sample data,this paper obtains the lithium ion battery aging data through the method of cyclic charging and discharging experiments on the18650 lithium iron phosphate battery in the laboratory,which is used as the training sample and test sample for machine learning.This paper compares the performance of the traditional random forest algorithm and the improved random forest algorithm based on particle swarm optimization in the prediction of lithium ion battery SOC and SOH.The experimental results show that the traditional random forest algorithm model performs well in the prediction of lithium battery SOC,but does not perform well in the prediction of SOH,and the prediction results have no reference value.Therefore,this paper proposes an improved random forest model based on particle swarm optimization algorithm,which improves the prediction accuracy of SOC to a higher level,and also improves the prediction accuracy of SOH and reduces the error of the model.The stochastic algorithm model based on particle swarm optimization has high prediction accuracy and good robustness,which can provide an important reference for battery management and maintenance.