基于遗传神经网的自适应电池荷电态预估模型

电池荷电态(SOC)是放电电流、端电压、温度等多种因素的复杂的非线性函数,而且不同类型的电池具有很大的差异,不能建立统一的模型。因此要对其做出精确的预估是一件很困难的事情,需要耗费很多的人力和时间对特定类型的电池进行大量试验然后建模。为克服这些缺点,提出一种基于遗传神经网的自适应SOC预估模型,通过遗传算法对神经网络结构及其学习算法进行优化,在较短的时间内寻找到适合特定类型电池的神经网络模型,大大缩短了人工建模需要的时间,提高了模型对SOC预估的性能。对于三种不同类型电池的数据进行建模的仿真试验结果验证了本方法的有效性。

Adaptive genetic algorithm based on artificial neural network model for estimation of SOC of battery

State of charge (SOC) is a complex non-linear function concerned with discharge current, terminal voltage, temperature and so on. And for the reason that the battery performance varies greatly from one type to another, it is difficult to construct a unified model of SOC. The most popular method need a lot of experiments to find the proper model of the SOC. To overcome the shortcomings of the traditional method, a novel design of adaptive genetic algorithm based on artificial neural network model is proposed to model the SOC of different type of battery. This adaptive model can find the proper artificial neural network model for a special type of battery, within a short time. The simulation results of different types of battery verified the validity of this adaptive modeling method.