Data selection of a compact GRNN for Ni-Cd batteries fast charging

This paper presents a data selection technique for training the neural network controller in order to archive a compact network and to decrease battery charging time. A fast-charging device for nickel-cadmium (Ni-Cd) batteries is designed through the generalized regression neural network (GRNN) and implemented with the MATLAB/SIMULINK for testing and operating on real system. The input-output data for training neural networks were collected from rigorous experimentation. The suitable data were selected to establish GRNN comprising only 13 processing elements. Experimental with real time implementation clearly show that the proposed technique not only requires less neural processing units but also yields less MSE than ANFIS and RBF technique.