Study On Photovoltaic Power Short-Term Forecast Based On Improved Neural Network

Recent years, with the rapid development of energy and economy, the supply of traditional fossil fuels is increased, which leads to global warming and the survival and development of mankind has caused serious damage. In order to effectively deal with the increasing environmental pressures, the alternative strategy to clean energy alternative to traditional fossil fuels has become an inevitable trend in the development of global energy. Solar photovoltaic power generation has the advantages of wide distribution, easy exploitation, clean non pollution, high energy utilization rate and obvious advantages, which has gained extensive attention and has achieved rapid development. But the output of photovoltaic has obvious intermittence, uncertainty and uncontrollability, large-scale integration will inevitably impact on the safe and stable operation of power grid, power grid scheduling and planning adds complexity of operation, to some extent limit the development of large-scale photovoltaic power generation. The short-term prediction can effectively reduce the uncertainty factor of power grid photovoltaic power output, improve the reliability level of power grid operation, therefore, this paper focuses on photovoltaic power prediction technologies are studied, the main contents are as follows:(1) A brief overview of the development prospects of domestic and international photovoltaic power generation, analysis the prediction of the meaning and purpose of the security and stability of the grid photovoltaic power, and the domestic photovoltaic power prediction methods are summarized in detail.(2) A photovoltaic power generation system simulation model is builded in the comprehensive analysis of PSCAD software. In order to stabilize the output power, and an improved voltage perturbation method is proposed for photovoltaic power output of photovoltaic power generation system to achieve coordinated optimization. On the basis of the analysis of the influence of the type of sunshine, radiation intensity, temperature and relative humidity on the photovoltaic power output, Which can provide a basis for the follow-up of photovoltaic power prediction.(3) Optimization problem belongs to the typical nonlinear, high latitude photovoltaic power prediction, starting from the prediction model and the evaluation index of photovoltaic power generation, the basic principle of nonlinear prediction method and prediction method of the traditional advantages and disadvantages are analyzed, finally, put forward relevant index for prediction evaluation prediction scheme precision.(4) Aiming at the defects of the traditional BP neural network prediction model and iterative number, long convergence time, put forward a kind of improved BP neural network predictive model based on particle swarm algorithm for problem analysis, simulation environment and the practical case of the proposed PV system prediction model is trained and tested based on the efficiencv of the verification the proposed method.