Battery Automatic Grouping Technology Research Based On Machine Learning

With the rapid development of modern society,people's demand for energy is growing.At the same time,the traditional resources are gradually decreasing.So electric vehicles become an important part of development of the car.As the core components of electric vehicles,power batteries gradually developed.But single power battery can not satisfy the power needed by electric vehicle.So the single battery is needed to connect into a battery pack.Due to the fact that the inconsistency between batterys has a great effect on the performance and life of the battery group,therefore,looking for a suitable battery grouping method is essential for the development of electric vehicles.This thesis introduces the related technology of battery grouping.By comparing various grouping technologies,the discharge curve of battery is selected as the standard of battery grouping,and transform the grouping problem into the clustering problem of time series.In this thesis,the existing time series clustering algorithm is compared,and time series clustering algorithm based on feature is selected.The mean,variance,kurtosis and time dimension of the discharge sequence is extracted,and used k-means clustering algorithm to complete the grouping.To compare with this method,the method based on self coding neural network is proposed.Firstly,comparing various activation functions and optimizers in the neural network,the activation function and optimizer which is smaller error are selected.The data of the battery discharge sequence is preprocessed and putted into the self coded neural network to train,the weight parameters are saved when the stable state is reached.Then the battery discharge sequence which is used to test is putted into the trained neural network models to get their characteristics.Finally,we use fast search and find density peak clustering algorithm to complete grouping.Finally,this thesis compared the two grouping method by contour coefficient,then choose the mothod whose contour coefficient is higher and multi parameter grouping method to do store energy and recycle charge and discharge experiments.Observing the self discharge rate of the battery group after a period of static time and the change of capacity after 200 cycles of charge-discharge.The experimental results show that the method this thesis proposed is superior to the multi parameter matching scheme in energy storage and capacity attenuation.