Research On The Output Power Forecasting Of Grid-Connected PV Plant

Along with our country paying more attention to new energy resources and implementation of new energy development strategies in recent years, photovoltaic power generation is playing an increasing part in the power industry. With the rapid increase of photovoltaic power capacity, the grid-connection of large scale photovoltaic power generations will bring great challenges to power system management. Accurate prediction of the photovoltaic power generation will be conducive to the reasonable scheduling by the power sector, it also has great significance on the safe and stable operation of the power grid and efficient utilization of solar energy. To improve the prediction precision of the photovoltaic power station output power, a method uses seeker optimization algorithm and support vector machine to predict the solar irradiance is put forward, then takes solar irradiance as input variable, and gets the output power by using methodology based on the ensemble empirical mode decomposition and extreme learning model.After analyzing each input factors that influence the photovoltaic power generating capacity, the solar irradiance is selected as the main factor. A prediction model of solar irradiance is built based on support vector machine which is trained with the historical sequence of solar irradiance and its parameter optimization is searched by seeker optimization algorithm, finally, the solar irradiance is predicted with the trained model.Then on the basis of the solar irradiance, to predicate the output of photovoltaic power plant. A prediction model is built basing on the ensemble empirical mode decomposition and extreme learning model, which can decompose the historical time series of output and get a series of relatively stable component variables. Next the extreme learning model is used to make prediction on these component variables, the forecasting results of extreme learning model are summed to obtain the final forecasting result.According to the measured data gets from a grid-connected photovoltaic power station in Gansu province, MATLAB is used to process the simulation analysis of the predicted solar irradiance and power output. At the same time, by comparing with the physical statistical method and the traditional back propagation neural network method, it can be conclude that, the solar irradiance forecast based on seeker optimization algorithm and support vector machine and the output power prediction method based on the ensemble empirical mode decomposition and extreme learning model has a relatively higher prediction precision, therefore the prediction method is feasible.