Research And Application Of Short-Term Wind Power Forecasting Based On Supervised Learning Method

Wind energy is a kind of renewable energy,which has been paid much attention by many countries.With the development of wind power industry,large-capacity wind power merge into the grid.But due to wind randomness,volatility and sudden changes,the instability of wind power and low wind energy utilization rate remains to be improved.Therefore,short-term wind power prediction high precision is still the focus of the study.Short-term wind power forecasting refers to the prediction of the future power generation of 0?72h which has a very important significance for power system maintenance.Based on above,the short-term wind power forecasting methods are studied from the aspects of data processing,the influence of wind speed characteristics on the actual power generation,the combined forecasting model and the power prediction method based on the wind speed climbing.Firstly,the historical data were processed by the Bean wind power standard curve.Because the general method cannot describe the mapping relationship between wind speed and power,this paper analysis the influence factors of the actual wind turbine power generation,when the wind speed of continuous rise and fall,will cause serious impact on the wind power machine.In order to improve the forecasting accuracy of short-term wind power,incorporating wind speed characteristics into the Short-term wind power forecasting method.The wind speed will be labeled by the features.As the single model may appear large local error and instability,the original training set is divided into several subsets by using Bagging algorithm,the predicted results would be combined Effectively.Wind power forecasting process is dynamic,need to constantly update the training set,so the Online sequential extreme learning machine(OS-ELM)would be used,the experiments show that the combined method indeed improves the accuracy of wind power prediction,and the prediction efficiency is greatly improved.According to the characteristics of wind speed,the wind power data would be divided into three categories.then Establish LSSVM prediction model for them.Aiming at the disadvantages of the LSSVM kernel function parameters and the regularization parameters,the gravitational search algorithm(GSA)was introduced to select the parameters of LSSVM model intelligently.In order to make GSA more accurate for searching the optimal solution region,the linear function is used to improve the traditional gravitational exponential function,the initial gravity coefficient was set directly into a dynamic selection.Finally,all kinds of power prediction algorithms were integrated into the power prediction system,and the effectiveness of the algorithm was verified in the actual production environment.