Research On Distributed Photovoltaic Power Forecasting Based On Statistical Up-scaling Method

At present,with the continuous development of photovoltaic power generation technology and the increasing installed capacity of distributed photovoltaic power generation,the randomness and intermittences of photovoltaic power generation have become an important factor restricting the development of photovoltaic power generation.The formation and application of large-scale distributed photovoltaic power generation grid-connected has posed a series of challenges to the safe and reliable operation of power system.If can achieve the precision of the distributed photovoltaic power,efficient prediction can help both scheduling mechanism reasonably modified scheduling plan in time,as far as possible to avoid the phenomenon of "discard power brownouts".Therefore,timely and accurate prediction of distributed photovoltaic power generation is of great importance,which can provide useful guidance for the dispatching center to carry out the work,so that the prediction deviation of distributed photovoltaic power generation can be effectively controlled.Based on the research status of distributed photovoltaic area power prediction,timely and interrelation between mining model output,the numerical weather prediction(NWP)data as sample,using the current hot spectral clustering method for the personage inside course of study and current application more widely circulating nerve network created to establish a scientific and reasonable,the strict specification model.In the first step,the reasonable classification of NWP information should be completed by combining the development law of irradiance with the spectral clustering method which is simple to operate and hot to be applied at present.The second step is to create a rigorous and easy to understand prediction model based on advanced and mature cyclic neural network technology for each group of data.Thirdly,in the process of prediction,the category of NWP data should be scientifically and reasonably determined first,and then the prediction analysis should be carried out through the established model to obtain the corresponding results.This method can realize the power prediction modeling of regional distributed photovoltaic power generation,and the accuracy of the prediction results can meet the engineering application requirements of power prediction.Compared with the traditional regional prediction accumulation method,the statistical upscaling method can save computing resources,shorten modeling time,and reduce the dependence of regional prediction model on the data completeness of a single photovoltaic power station,which has a very high engineering practical significance.