The Optimization Research Of Photovoltaic Battery Remaining Capacity Prediction Based On Fuzzy Neural Network

Solar photovoltaic system has advantages of pollution-friendly, security, convenient and not subjected by regional restriction and unlimited reservation, etc. Lead-acid battery is the primary energy storage device in solar photovoltaic system, therefore the study of the higher accuricy prediction of battery’s residual capacity is quite significance. The residual capacity of implementation of accurate prediction not only beneficial to improve the working efficiency of the battery and prolong the working life, but also prevent the improper charge and discharge. However the prediction problem of lead-acid battery’s residual capacity with its internal complex electrochemical characteristics, meanwhile the battery industry has subtle difficult to conquer this hard nut which lead to research on capacity prediction has no centralized standard. Through a large number of relevant literature and academic journals while considering the existing research based on school laboratory, this paper applies a improved model to the residual capacity prediction of solar photovoltaic lighting system which on the basis of the existing fuzzy neural network.Firstly this paper describes the primary characteristics of lead-acid battery parameters,such as theoretical knowledge, mainly including the battery voltage, battery capacity, state of charge, battery internal resistance. And secondly detailed analysis of the discharge character of lead-acid battery and the four major factors affecting the service life of lead-acid battery,and discusses the working principle of storage battery and equivalent circuit model. Thirdly Method for the forecast of the residual capacity of lead-acid battery at home and abroad are summarized in detail, and the comparative analysis the advantages and disadvantages of various methods. Fourthly, considering the characteristics of lead-acid battery, solar photovoltaic lighting system selection based on fuzzy neural network algorithm to improve forecasts of the residual capacity of battery model. The model have 5 layers which are input layer, hidden layer, reasoning, summation layer and output layer. Its main characteristic is to select seven variables to define the fuzzy data collection, selection of subordinate degree function like Gaussian function, reasoning layer adopts two fuzzy rules for product and sum.And the improved self-learning algorithm, compared with other three kinds of commonly used algorithm experiment analysis, is based on the improved model in prediction accuracy and compatibility optimization effect.Finally, via simulation in MATLAB simulation experiment, numerical two root was obtained from the sample of numerical simulation and experimental curve, we find that the improved FNN model and algorithm can get better accuracy and stability without affect any regular working moment of battery, which has a certain application value.