Check the current and voltage on charge or discharge, use fuzzy control principle and ah approach to estimate the soc, and correct the actual charge to get the exact result, in this work, a multi-layer feed forward neural paradigm is used for soc fuzzy estimator design, and a rapid learning algorithm that is realized model structure identification is proposed also, and find the fuzzy principles make use of their massively parallel architecture and learning capacity 构造一多层前馈神经网络来进行模糊推理,提出了一种快速学习算法用于网络的学习过程,利用神经网络的并行计算、自学习的特性,根据样本数据发现模糊推理规则,并在应用过程中对模糊规则自动修正,提高预测准确性。这种方法不需要对蓄电池有很多的先验知识,不需要建立复杂的数学模型。