Study On Power Forecasting For Photovoltaic System In Beijing

Photovoltaic power generation is one of the new energy industries that the most commercial developed prospect. The output of photovoltaic system is random, interval and uncontrolled, it will bring the impact on the public power grid, increase the difficulty of the load forecasting and power grid scheduling, influent the investment and the optimization design of the power grid, after its connecting to grid. So developing the prediction about the photovoltaic output is very important to reduce the bad influence of the grid, realize kind interconnection and economic scheduling, improve the power grids’ ability of accepting photovoltaic system.First of all, this passage selects the XuJi Company’s grid photovoltaic system as experimental platform, the external characteristics of photovoltaic cells are simulated, and light intensity and temperature’s effect on the output of the photovoltaic cells are analyzed at the same time. The output power data and the local meteorological observation data of this platform in the past25months are studied, the influence on the output of photovoltaic system by meteorological factors are researched in detail. Similar day theory is introduced, a kind of guidelines that determine the photovoltaic system’similar day in output is introduced as well, that is a calculation model about the evaluation forecast day and the historical similarity of meteorological conditions.This passage set up the photovoltaic system medium and short-term (30min) power generation forecasting model, predict the output of photovoltaic array within24h in three types of typical day, record the error with the actual power. The experimental results prove the effectiveness of introducing the theory of similar days, show the performance of RBF neural network better than the performance of BP neural network, the result of prediction is fit for the rule of the photovoltaic power station power prediction technical requirements.At last, according to the actual situation of operation, the energy management strategy about two kinds of states that grid and off-grid is introduced. To intervene in micro grid energy flow direction by the result of prediction, formulate scheduling decision about battery charging and so on. The combination of theory with practice about photovoltaic power generation prediction is realized; its practical utility in micro power grid operation is verified.