Research On Short-term Load Forecasting Considering Photovoltaic System

Short-Term Lord Forecasting (STLF) is very necessary in power system, and the result of forecasting is highly related to the safety and economical efficiency of the operation of power system. Therefore, how to increase the accuracy of load forecasting has always been a popular field concerned by researchers both at home and abroad.It has been shown by amount of research that the meteorological factors have been played a significant part in power load among many factors. So investigation the relation ship between meteorological factors and power load is an effective way to increase the precision of load forecasting. In this thesis, load characteristic in Hangzhou region is analyzed, including change law by year,month and week. Then the relationship curves are plotted. Based on the theory proposed above, three integrated meteorology indexes are put forward to handle meteorological factors.Then the application of artificial neural network in short-term lord forecasting is introduced, especially the theory of Back Propagation. Besides, an improved BP algorithm is brought forth based on the theory above, and the method is verified by the example of Hangzhou region.In connection with the short point of BP algorithm, an adaptive artificial neural network model is proposed and applied to the short-term forecasting. Regarding to the huge amount of training data, a dynamic adaptive method is used to process them. The method proposed increases the efficiency of data processing and shortens time of forecasting. Then a new model considering noise interference is put forward. At last the dynamic adaptive artificial neural network proposed above is verified by the example of Hangzhou region.The influence on the load of addition of photovoltaic grid-connected power system is analyzed, and a small capacity photovoltaic is set up in Hangzhou region. Then the influence factors on the photovoltaic system are analyzed. At last the power load of whole system is forecasted using the method of artificial neural network.