Study On Short-Term Prediction Of Photovoltaic Output Power Based On Optimized Neural Network

Based on the basic situation and social development of our country at this stage,we can see that the development of national science and technology and the progress of society are inseparable from the good natural environment,so environmental protection work needs to be given enough attention.As one of the clean energy sources,solar energy is widely and infinitely available,and has the characteristics of clean and pollution-free.It is gradually becoming one of the most powerful energy sources in the 21st century.Distributed photovoltaic power generation is a major and effective use of solar energy.China's distributed PV compared with the Western countries,although the late start but the rapid development.The total installed capacity in 2015 reached 43GW and which includes the new installed capacity of 16.5GW,in a global leader,distributed photovoltaic power generation has a huge development prospects and research value.The output of distributed photovoltaic power generation is influenced by its own characteristics and structure.It has the characteristics of intermittence,periodicity,volatility and randomness.When it is integrated into the grid,it will affect the safety and stability of the power grid and increase the difficulty of power dispatching.Master the output characteristics of photovoltaic power generation and a reasonable and effective prediction to help better understand and use of photovoltaic,to help the relevant power sector to develop power generation plans and scheduling plans and at the same time reduce the stability of photovoltaic grid,this is of great significance to the development of photovoltaic power generation.There are many problems in the research on the output power prediction method of photovoltaic power generation at home and abroad,such as cumbersome design method and long training time.In this paper,it puts forward a short-term prediction method of photovoltaic power generation output with improved input and optimal prediction model.Firstly,the single correlation analysis and the compound correlation analysis of the main meteorological factors influencing the photovoltaic output are carried out,and the influence factors are divided into the direct influence level and the indirect influence level by the path analysis method.The result of this analysis is more informative and practical.Secondly,use the dimensionality reduction principle of principal component analysis to reduce the number of original variables that are related to each other to a few independent variables that are independent of each other as input to the neural network.At the same time,the prediction model of PV output is optimized by using the global search characteristic of genetic algorithm.The specific way is to optimize the weights and thresholds of commonly used BP neural networks to compensate for their ease of trapping into local minima and the longer duration of convergence.Finally,through the analysis of the prediction results under the stable and abrupt weather of a distributed PV demonstration base in Liaoning Province,the conclusion is drawn that the proposed method has higher accuracy and faster convergence training speed.The photovoltaic output prediction model can be used to predict output power of Liaoning region which under its unique climatic conditions and geographical location,and to better promote the development of distributed photovoltaic power generation in Liaoning Province,then provides better basis for the development of power generation plan and dispatch plan for the relevant power sector.