Researches On Wind Speed And Wind Power Forecasting System For Wind Farm

Wind power, as a kind of clean renewable energy, nowadays has received more and moreattention from all over the world. However, wind speed presents strong randomness andfluctuation which will lead to unstable wind power output, alongwith the increasingpercentage of wind power in the electricity grid, wind power will bring some challenges anddifficulties for the operation of electric power system. Therefore, wind power forecastingplays a very important role. Accurate wind power and wind speed prediction can providenecessary technical support and guidance for dispatch of the power system incorporated withwind power, thus to lower or eliminate the adverse effect brought by incorporating windpower. It also can help wind farm workers implement wind turbine maintenance whileavoiding valid wind speed so that wind farm operation cost and loss can be reduced. Undersuch circumstance, we launched a profound research on the problem of wind speed and windpower forecasting.In this paper, firstly, we briefly introduce the study situation of wind speed and windpower prediction at home and abroad, the classification of wind power prediction, analyse thecharacteristics of wind speed and wind power, and further propose the basic flow of realizingwind power forecasting. Secondly, three frequently used forecasting methods(time series, BPneural network and Elman neural network) are introduced, and their basic principle andimplementation procedures have been elaborated. However, the aforementioned threealgorithms have many drawbacks, so the core method of this paper—wind speed and windpower forecasting based on wavelet decomposition and differential evolution optimizedsupport vector machine (WD&DE-SVM) is proposed. According to this method, we usedifferential evolution algorithm to searching the optimal parameters of SVM(DE-SVM), andwind speed series is decomposed into several scales by wavelet analysis, for each scale a onestep wind speed forecasting model is established based on DE-SVM, and the predicted valueof each model was summed as the forecasted wind speed of next time. The forecasting windspeed value of current time is added into the history database, and repeat the previousforecasting procedure based on the forecasting value to predict next-next wind speed value,through this circulation prediction method, the multi-steps wind speed forecasting is achieved.Then, according to the wind speed-output calculation formula, we can obtain the forecastedwind power ouput value.Simulations have been done in Matlab to test the proposed wind power multi-stepforecasting model. And the proposed model is also compared with other methods such asTime series, BP and Elman neural network through cases, the results demonstrate the superiority of our method. In the end, a practical wind power forecasting system is developedbased on ASP.Net technology, this system can provide function such as short-term wind speedand wind power prediction, produce curve based on predicted results and history data enquiryetc. The above researches have important theoretical and realistic significance to the powersystem incorporated with wind power, and can provide a small contribution to the Nationalnew energy technology development and the energy structure adjustment.