Research On The Prediction Of PV Output Based On The Uncertainty Theory

With the development of global economy, environmental pollution and energy crisishave become two major problems, so the development and utilization of new energysources is inevitable. With its advantages, Photovoltaic power generation has attractedwidespread attention, large-scale photovoltaic grid connected power generation has gooddevelopment potential. However, photovoltaic is influenced by many meteorological andenvironmental factors, the variability of which will bring inconvenience for powersystem dispatch and overall planning, so the prediction of PV output is of greatimportance.Solar radiation and photovoltaic battery temperature are the two most importantfactors that influence the PV output. In the previous prediction methods, the results aremostly determined values. While this paper proposes several new forecasting models, inwhich the cloud cover and cloud cover index are processed by the uncertainty theory, andultimately get the predictive PV output values and prediction intervals meeting differentconfidence levels.Solar radiation is the most influential factor. Changes in the shape, height, thicknessand other characteristics of the cloud influence the solar radiation values that thephotovoltaic cells received, so the cloud must be considered. Cloud cover is not only avague concept, but also random, meanwhile, the distribution of cloud cover index is alsorandom, so we can use the uncertainty theory to analysis both of them, and then correctthe solar radiation values by REST model in cloudless weather with cloud cover index.We can use BP neural network model to predict the photovoltaic battery temperature byusing the battery temperature and solar radiation values, which are the most two relevantfactors, as the input data.By combining the solar radiation model and PV battery temperature model, theforecasts of PV output can be realized. By using the BMS photovoltaic power plant datain America, the validity of the models can be verified. Prediction results show that themodel can provide much wealthier information for the power system dispatchers and hasa good reference value.