基于深度学习和量子遗传算法的电池SoC估算方法研究

Estimation method of battery SoC based on deep learning and quantum genetic algorithm

储能电池的荷电状态是电池的重要特性,针对浅层学习算法的不足,提出了深度学习理论与量子遗传相结合的算法以提高估算结果的正确性。该算法能够自动从样本中提取更加抽象、更具表达能力的特征,实现输入和输出数据之间的复杂非线性映射;量子遗传算法自动寻优,得到每个RBM输出估算值的权值。通过对电池SoC训练样本和测试样本的估算,与BP训练网络估算结果对比,得出本文所提的DBN-QGA算法网络估计精度更高。

State of charge of storage battery is an important characteristic of battery. Considering to the weakness of shallow learning algorithm,this paper proposes deep learning theory combining with quantum genetic algorithm to raise the estimation accuracy. The algorithm can automatically extract features more abstract and more specific from the sample,realizing complex nonlinear mapping between the input and output data.Quantum genetic algorithm automatically optimizes and each RBM output estimation weights are achieved. By estimating the training samples and testing samples,compared with the BP network training estimation results,DBN-QGA algorithm network has higher estimation accuracy.