Research On Generating Power Forecasting Of Grid-connected PV Power Plant

In the society, there is a serious lack of conventional energy. At the same time, the natural ecological environment which mankind needs to survive is deteriorating. Solar energy as a new energy is playing an important part in people’s daily life for its abundance, cleanability and limitless in geography. So far, grid PV power has occupied 75 percent of the photovoltaic power generation. Photovoltaic power generation is not only vulnerable to uncertainties and solar intermittent, but also existing low conversion rate, accounting for a large area and adverse impact on power quality of a large grid. Therefore, making predictions to the amount of PV power can reduce the negative impact on the grid.In order to improve the generating power forecasting accuracy of PV power plants. Firstly, it elaborates the background and significance of research and research status of photovoltaic power plant generating capacity forecasting at inland and abroad. The principle and consist of photovoltaic power plants is briefly introduced. The characteristics of output power is analyzed. Meanwhile factors that affect the amount of photovoltaic power forecast are detailedly analyzed, including seasonal, day type, temperature and light intensity, relative humidity. Pearson similarity analysis method is used in influencing factors of the results. Secondly, it is a task to analysis the data processing of PV power plant, including data storage, data types and data processing. And then, the overall block diagram of the photovoltaic power plant generating capacity of prediction is given. And it is an important work to analysis the prediction of BP neural network topology, including learning algorithm, defects. For the existence of the defect, it combines the dynamics learning rate Ak and LM numerical optimization algorithm, and introduces genetic algorithm that optimizes weightings and thresholds of BP. Finally, the power generation forecasting mode of photovoltaic power plants is designed and simulation result is analyzed. Including BP neural network model and GA-BP prediction model, and the accuracy and speed of two prediction models are compared, the latter can be found in the prediction accuracy is increased, GA-BP prediction model has a certain value.