A Lithium Battery Online Monitoring System And The Prediction Of The Remaining Capacity

The characteristics of the lithium battery are small size,high voltage,high specific capacity,high security,good environmental protection and no pollution,which make it widely used in various fields.But the battery's performance and price has become the main "bottleneck" of these products.In order to make the use of battery more effective and scientific,it is urgent to carry on the on-line monitoring of the battery and the accurate prediction of its remaining capacity.Based on this reality,this paper developed a set of real-time monitoring system for the lithium battery's parameters,which is based on STC12C5A60S2 MCU and LabVIEW software platform.Lithium battery monitoring system acquires the lithium battery's voltage,the discharge current and many other parameters,and then get the lithium battery's remaining capacity by calculating.We can display and storage these parameters in LabVIEW user interface.In this paper,the composition and working principle of lithium battery are described,and the charge and discharge characteristics in the working process are analyzed in detail.In this paper,several methods for predicting the residual capacity of batteries are summarized.And on this basis,combined with the battery's real work process in our life,the prediction method is presented which uses different prediction methods in the battery's different working stages.The open circuit voltage method,ampere-hour integral method and back propagation neural network method are used to forecast the battery's SOC in the battery's different working stages.For simulating the battery's using process,we have the discharge experiment on the batteries which have the same type and different recency,and then we use the data to practice the BP neural network.Because BP neural network has some deficiencies,we should to improve the BP neural network to make the network has faster convergence speed and higher accuracy.Then the prediction model of battery's residual capacity is established on MATLAB,the BP neural network through continuous learning make the estimation results gradually approaching the actual value.Experimental results show that the levenberg-marquardt algorithm is better than the additional momentum method to improve the BP neural network on lithium battery's SOC prediction;the system can real-time display the battery's parameters,realize the storage of data and can carry on the dynamic forecast SOC of the battery;The prediction method of this paper realizes the high nonlinear mapping between the input and output of the battery,which has high accuracy and feasibility.