This paper presents a new method for estimating the individual battery state-of-charge (SOC) of electric scooter (ES). The proposed method is to model ES batteries by using the fuzzy inference neural network system. A reduced form genetic algorithm (RGA) is employed to tune control point of the B-spline membership functions (BMFs) and the weightings of the fuzzy neural network (FNN). The proposed FNN with RGA (FNNRGA) optimization approach can achieve the faster learning rate and lower estimating error than the conventional gradient descent method. The validity of the SOC estimator is further verified by a constructed multiple input multiple output (MIMO) FNN structure for estimating the SOCs of battery powered ES. A fixed velocity discharging profiles of the ES batteries are investigated to train the FNN for precise estimating the SOCs of the battery strings. Furthermore, a testing data profile is used to demonstrate the superior robust and over-fitting suppressed performance of the proposed method. The estimated SOCs are directly compared with the actual SOCs under different FNN methods, verifying the accuracy and the effectiveness of the proposed intelligent modeling method.