Research On Lithium-Ion Power Battery Capacity Estimation Based On Ultrasonic Detection And Neural Network

Due to various factors such as energy and environment,the development of new energy vehicles has become an inevitable trend.With the development of new energy vehicles,lithium-ion power batterie which is the core energy supply components has been extensively researched.How to accurately measure the remaining power of lithium-ion power batteries is of great significance for the safe driving and timely charging of electric vehicles.Deep learning has a very good effect on the approximation of complex data relationships.Combining it with ultrasonic detection of batteries can effectively calculate the remaining battery power.In this paper,lithium-ion power batteries are used as experimental objects.During the continuous charging and discharging process,real-time current,real-time voltage,and current power are recorded.At the same time,the lithium ion power battery is subjected to ultrasonic testing and an ultrasonic feedback signal is obtained.After analyzing these data,it is found that there is a strong correlation between the peak-to-peak value of the ultrasonic feedback signal and the battery capacity.The voltage and current are also closely related to the lithium-ion power battery capacity.Then this article does research on the obtained ultrasound data:(1)The BP neural network is trained on the data of the two processes of charging and discharging,and using particle swarm optimization algorithm to initialize the weight value and threshold of the network and then the model is used to estimate the capacity of the battery.The experimental results show that for the charging process,the average error of the lithium battery capacity estimated by obtained model is 1.7%;while for the discharging process is 1.1%.(2)Linear regression and polynomial regression methods are used to fit the obtained ultrasound data and estimate the battery capacity.The experimental results show that for the charging process,the average error of the lithium battery capacity estimated by linear fitting is 6.2%,while by polynomial regression is 2.2%;for the discharge process,the average error of the lithium battery capacity estimated by linear fitting is 1.5%,while by polynomial regression is 1.3%.(3)the BP neural network is also trained for the voltage and current data during the charging and discharging processes,and then the model is used to estimate the capacity of the lithium-ion power battery.The experimental results show that for the charging process,average error of the lithium battery capacity estimated by obtained model is 3.5%;while for the discharging process is 3.4%.Finally,this article comprehensively compares and analyzes all the above methods.It was found that when using the model obtained by BP neural network which was trained on the peak-to-peak value of the ultrasonic feedback signal and lithium battery power data during the charging and discharging of the lithium battery to estimate the capacity of lithium battery,the error is the smallest and the result is the best.At the same time,the ultrasonic signal data is analyzed.After processing the abnormal data,the BP neural network is retrained.Before the BP neural network training,the particle swarm optimization algorithm is also used to initialize the network’s weight value and threshold.The final experimental results show that for the charging process,the error of the final neural network model to estimate the lithium battery capacity is only 1%,and for the discharge process,the error of the final neural network model to estimate the lithium battery capacity is even lower than 1 %.This paper proposes a new method for estimating the remaining power of a battery based on the ultrasonic detection of a lithium battery and a neural network,which can accurately measure the amount of remaining power,and provides a new idea for the capacity estimation of a lithium-ion power battery.It will play a certain role in battery life and quality assessment.