BP神经网络以及衍生方法对电池SOC预测的研究进展
Advances in Research on BP Neural Network and Derivative Method on Battery SOC Prediction

电池荷电状态(SOC)的估算是电池管理系统衡量电池寿命的重要指标。精确的SOC估计在防止过充放电、提高电池利用率、保障电动汽车电池系统的安全性和稳定性方面具有重大的意义。SOC的估算主要有安时积分法，卡尔曼滤波法，开路电压法，神经网络等。安时积分法需要定期修改荷电状态，误差较大；卡尔曼滤波法是依据均方最小的误差原则，较高地依赖型的精确度；开路电压法电池须长时间的静置，而实际工况中电流值上下波动大，测量误差较大；神经网络具有强大的非线性映射能力，无需特别精准的数学模型，在实际的SOC估算具有非常重要的优势。本文主要针对BP神经网络及其衍生的方法做了概述，实践证明BP神经网络与其他算法结合是最有前景的，不仅能缩小误差，提高准确性，还能突出它强大的非线性拟合能力。
The Estimation of the SOC for batteries is an important indicator of the battery management system to measure the life of the batteries. Accurate SOC estimates are significant in preventing over charge and discharge, improving battery utilization, and ensuring the safety and stability of electric vehicle battery systems. The estimation of the SOC mainly has Ah integral method, Kalman filtering method, open circuit voltage method, neural network, etc. The Ah integral method needs to regu-larly modify the charged state, the error is large; the Karman filtering method is based on the minimum error principle, higher dependence; the open circuit voltage method must stand for a long time, and the actual work The current value is large, the measurement error is large; the neu-ral network has powerful nonlinear mapping capabilities, no special accurate mathematical model, is very important in the actual SOC estimation. This paper mainly provides an overview of the BP neural network and its derivatives, and practice of the BP neural network and other algorithms is the most promising, not only reduces the error, and improves accuracy, but also highlights its pow-erful nonlinear fitting capacity.