Research On Short-term Power Forecasting Of Grid-connected PV System

At present, grid-connected photovoltaic system has been rapid development, however, photovoltaic power because of will be affected by the change of meteorological factors influence, discontinuity and uncertainty, power fluctuations affect the normal operation of the grid, so accurate power prediction technology for electric power dispatching, power load, with conventional energy generation planning and photovoltaic power generation planning has reference significance.Because China has not yet established a radiation prediction system perfect, so this paper respectively in the absence of radiation prediction and there are two kinds of conditions of the radiation prediction, with the United States University of Oregon solar monitoring laboratory data as the research object, the prediction models are established and forecast analysis using PSO-BP and LS-SVM. This paper mainly do the following work:1. Based on understanding the photovoltaic battery characteristics, analyzes the relationship between the type of day, season, weather factors and photovoltaic power generation, photovoltaic power main influencing factors of periodic and stochastic fluctuation was determined.2. Research the BP neural network and particle swarm optimization algorithm, in the absence of radiation forecast conditions, seasonal and establishes three kinds of typical day type PSO-BP prediction model. And using the established model and BP neural network is used to predict the effect compares.3. Study the statistical learning theory and least squares support vector machine, and use gray correlation analysis to determine the similar day forecast, enhanced the relationship between sample data and predicted data. Using similar day training sample data of LS-SVM prediction model, the prediction of radiation condition, the model and the BP neural network model are predicted and compared the prediction results.4. To build photovoltaic power do needs analysis, process analysis and structure design of prediction system. And based on the previous work, the realization of a B/S photovoltaic power forecasting system, laid the foundation for forecasting information release.By comparison, the two prediction model established in this paper has good prediction effect in two kinds of conditions. Establish the forecasting model and forecast system will help to improve the photovoltaic power prediction accuracy, reduces the influence of power fluctuation of power grid operation, promote the development of photovoltaic grid-connected power system.