Prediction Of Power Generation Capacity Of Photovoltatic Systems Base On Artificial Neural Network

With the enhancement of modern society’s demand for energy sources, the energy crisis and environmental pollution problems have become increasingly severe. Solar photovoltaic energy because of its renewable, clean, environmental protection has attracted worldwide attention. However, photovoltaic power generation system has its own characteristics, such as randomness, periodicity and instability. Power system is a real-time balancing system. According to power generation characteristics of photovoltaic power generation system, power system which can be incorporated into a large power grid would have great impact on the power grid and bring a lot of unnecessary problems to the stable operation of large power grid. For large-scale network of photovoltaic power generation system, accurate photovoltaic power prediction network is an important prerequisite for effective mitigation of adverse effects. So it has very important practical significance and application value on the prediction of photovoltaic power generation system’s generation.Firstly, the working principle of power generation photovoltaic cells are elaborated; Secondly, the paper introduces the classification of photovoltaic power generation system and a variety of basic components of photovoltaic power generation system; Meanwhile the basic structure of photovoltaic power generation system of Anhui Engineering University of 110KW photovoltaic system is introduced; and then analyzes the power output characteristics of photovoltaic power generation system, the solar radiation intensity, environment temperature and weather types can be regarded as the main influencing factors of photovoltaic power generation system’s power generation. It can be preparing for the establishment of prediction model.This paper has put forward two kinds of prediction model based on BP neural network. Wavelet neural network model is firstly proposed; And then,additional momentum term and adaptive change learning rate of the improved BP algorithm are proposed because the convergence speed of BP algorithm is slow and easy to fall into local extremum detects; According to the BP neural network’s initial weights, threshold blind selection andrandomness,the paper proposes a prediction model of genetic algorithm and simulated annealing particle swarm algorithm optimization based on BP neural network in order to overcomeslowness of convergence rate and defects of falling into local extremum easily.Finally, based on the data of the Anhui Polytechnic University photovoltaic micro grid system, the forecast model was simulated in the MATLAB programming environment. The prediction results show that the proposed model and algorithm have higher prediction and convergence rate.