Research On Wind Speed Forecasting Of Regional Wind Farm Group Based On Spatio-temporal Correlation

Wind power development exhibits a remarkable characteristic which developsintensively，whose intermittency and volatility is a enormous challenge to the safe andstable operation of the grid connected to. Because wind power is closely related to windspeed, accurate real-time wind speed prediction becomes critical, that will be conduciveto the timely adjustment of grid scheduling program to mitigate the effects of intermittentof wind power on the grid.To improve accuracy of real-time wind speed prediction, this paper presentedSTCP-ANN combination prediction methods based on spatio-temporal correlation (STC)and artificial neural network (ANN). Firstly the method got established STCP model topredict in unequal intervals based on the physical characteristics wind speed evolutionand wind speed and direction information of a number of points in the neighborhood ofthe target forecast point. And then we established ANN model to finish real-time forecastwith wind speed time series of target forecast site. Finally, STCP results with a slidingtime window and ANN forecasting results were combined to get final forecasting windsped through non-optimal combination of positive rights. Among ANN prediction model,typical BP and Elman neural network were used for STCP-BP and STCP-Elmancombination forecasting method respectively.Based on the measured data of wind farm groups in a region of northern Hebei,simulation tests are conducted in Matlab for two combination prediction modelmentioned herein and BP network, Elman network analysis methods are compared.Prediction accuracy of STCP-BP combination forecasting method in advance24-stepimproves22.74%compared to BP. Under the same kind of situation, STCP-ElmanElman combination forecasting improved19.22%compared to Elman. Predictionaccuracy of STCP-Elman respectively improves6.91%,11.76%,7.85%,8.88%in6step,12-step,18-step,24-step ahead prediction compared to STCP-BP. Results shows thatSTCP-ANN combination forecasting method proposed effectively improve the accuracyof real-time prediction and huge advantage of Elman network in the dynamicspatio-temporal modeling,testing the accuracy and validity of STCP-ANN.