基于极限学习机的磷酸铁锂电池SOC估算研究

Research on SOC estimation of LiFePO_4 batteries based on ELM

SOC(state of charge)的准确估算是电池管理系统的重要目标之一。针对传统神经网络方法在磷酸铁锂电池SOC估算中存在计算复杂、学习时间过长的问题,提出了一种新的基于ELM(extreme learning machine)的电池SOC估算方法。利用电池充放电系统完成磷酸铁锂电池在不同电流下的放电实验,获得实时测量的电压、电流。运用实验获得的数据对模型进行训练和预测,将预测效果与BP(back propagation)神经网络和SVM(support vector machine)进行对比,研究ELM在SOC预测中的可行性和优势。经分析可知,基于ELM的磷酸铁锂电池荷电状态估算模型的精度更高,并且网络训练速度得到大幅提升。

Accurate SOC estimation（state of charge） is one of the most important goals of battery management system. Considering the complex computation and long learning time of traditional neural network method to estimate the state of charge of LiFePO4 battery, a new method based on ELM（extreme learning machine） was proposed. The battery charging and discharging system was adopted to discharge the battery at different currents and the real-time measurement of voltage and current was obtained. The experimental data was applied to train the model, and by compared with BP（back propagation） neural network and SVM（support vector machine）, its feasibility and advantages in SOC prediction were analyzed. The research results illustrate that the model of SOC estimation based on ELM is more accurate and the network＇s training speed is greatly improved.