Comparative Research Of Photovoltaic Maximum Power Point Tracking Based On BP And RBF Neural Network

As the result of lacking traditional resource(such as coal,oil,natural gas,etc.),people begin to pay more and more their attention to clean, renewable solar energy. so in the future, the application of photovoltaic cells have good development of prospects. However, a low conversion efficiency of photovoltaic cells, and the higher prices have been a serious obstacle to the promotion and application of PV systems, There is a methods to maximize using the power generated by PV cells: making PV cells output maximum power to reduce the circuit loss. So in this article the method of maximum power point tracking makes the PV cells working at the maximum power point to obtain the maximum output power,which will be focused research.For the nonlinear characteristics of PV systems and maximum power point tracking characteristics, In this paper, BP neural network technology applied to photovoltaic arrays maximum power point tracking, the variety common of BP neural network learning algorithm and training algorithm in the MATLAB7 .0 have be simulated, by comparative analysis of the simulation results and finally the Levenberg-Marquadt of back-propagation learning algorithm, trainlm training function in the neural network toolbox is used in photovoltaic arrays maximum power tracking, and analysised the simulation of results.Radial Based Function Neural Network in the generalization ability, the ability and learning speed of approximation are better than BP neural network, RBF neural network is used in this article study of PV systems, and a detailed analysis of its variety of training algorithms, and achieved different training algorithms in the MATLAB7.0 , compared to the simulation results of different algorithms,randomly selected (direct calculation) is applied to the RBF neural network of training, the last Neural Network Toolbox functions is applicated to establishing a RBF neural network, simulation results show this approach. Is the feasibility and effectiveness.Finally, two kinds of neural network structure: the hidden nodes, the number of trai- ning steps, approximation error, generalization ability, photovoltaic arrays maximum power point tracking efficiency of KPM five areas are discussed and compared in this paper. Integrated comparing of the results: RBF neural networks have better generalization ability, and can be faster and more accurate in tracking of photovoltaic arrays maximum power point.