As a popular type of Redox Flow Batteries (RFBs), single flow Zinc Nickel Battery (ZNB) was proposed in the last decade without requiring an expensive and complex ionic membrane in the battery. In this paper, a Radial Basis Function (RBF) neural model is proposed for modelling the behaviours of ZNBs. Both the linear and non-linear parameters in the model are tuned through a new feedback-learning phase assisted Teaching-Learning-Based Optimization (TLBO) method. Besides, the fast recursive algorithm (FRA) is applied to select the proper inputs and network structure to reduce the modelling error and computational efforts. The experimental results confirm that the proposed methods are capable of producing ZNB models with desirable performance over both training and test data.