Research Of Generation Forecasting For Photo-Voltaic Based On Artificial Neural Networks

With the increasing of human demand for energy,the use of solar energy is more and more extensive concerned,due to the non polluting,photo-voltaic power generation as a kind of sustainable energy,is extensive concerned by experts in recent years.Photo-voltaic power generation is intermittent,random,Photo-voltaic power fluctuations will affect the power grid stability and electric power quality.Therefore,the photo-voltaic power generation prediction technology is a urgent need further research work,the accurate prediction of photo-voltaic power has better effect on the grid scheduling,power quality.Firstly by Huaneng Yingkou power plant photo-voltaic observation station data photo-voltaic power fluctuations characteristics and its influencing factors are analyzed,the analysis result shows that,the seasons and weather types has an important influence on photo-voltaic power fluctuation characteristics.Based on the traditional particle swarm optimization algorithm,particle "premature" problem,this paper delete the inertia weight and add the random factor,improve the global convergence of the particle swarm algorithm.Using particle swarm optimization algorithm to optimize the neural network photo-voltaic power generation forecasting model,through the comparison between the predicted data and the measured data,the validity of the proposed method is verified.Based on the above theoretical research,this paper designs a photo-voltaic power generation system,which can be applied the proposed algorithm to the practical application.According to the provisions of the two power system security,the design of photo-voltaic power generation forecast system equipped with a reverse physical isolation devices to ensure the safety of data transmission.In addition,the software structure and hardware structure of the system are designed,and the function of the system is demonstrated.According to the above research,we can conclude that: season,type of weather has greater impact on photo-voltaic power generation.Through the analysis of this paper,the photo-voltaic power of the similar weather in the same season should be the input layer of neural network model;remove the inertia weight and add the random search factor can improve the particle swarm algorithm with global search ability,so as to obtain a better neural network model to predict accurately the photo-voltaic power;sunny forecast effect to is obviously superior to the sky and snow days of similar seasonal patterns of weather,so the sky and rain and snow weather photo-voltaic power fluctuation rules law remains to be further study.