Prediction on moonlet power system data based on modified probability neural network

In this paper, an approach is proposed for time series prediction based on modified probability neural network (MPNN). Bayesian-statistics and decision-making theories and non-parameters density function estimation using Parzen-window function are applied to MPNN. The efficiency of the approach was demonstrated by a case study, an application for prediction on moonlet power system data, through comparison with other methods the linear ARMA model and the nonlinear widely used BP neural network. It was found that MPNN has the highest precision with the least time cost. Consequently, it verified and illustrated that large number of battery data can be predicted quickly and accurately using MPNN, moreover, it is valuable in the field of moonlet power system data prediction.