Research On Short-term Wind Power Prediction Based On The Classification Of Weather Types

The wind power has the characteristics of volatility and randomness.The large-scale integration of wind energy into power grid has exerted great influence on the safety and stability operation of the power system.The short-term wind power forecasting can reduce the adverse impact of wind power instability on the power grid,improve peaking capacity of the system,and enlarge assimilative capacity of grid to absorb more wind power.Based on the weather types,the short-term wind power forecasting technology is proposed in the thesis,and the main research involves the following aspects:Firstly,based on WF and MEEMD method,the filter of wind power data is studied.The research result shows that the filter effect of MEEMD method on wind power data is better than that of WF method.Secondly,the research on identification and characteristic analysis of wind power ramp event is carried out.In the paper,the standard swing door trending algorithm is modified by adding feedback link,and the improved swing door trending algorithm is proposed.Standard swing door trending algorithm and improved swing door trending algorithm are both applied to identify the ramp events from wind power data.The result shows that the improved swing door trending algorithm has higher identification rate for the wind power ramp events from wind power data than that of standard swing door trending algorithm.Based on improved swing door trending algorithm,ramp events are identified from wind power data of three wind farms in Shandong Province,and the characteristics of the ramp events are studied.Finally,the dynamic adaptation short-term wind power prediction method based on the classification of weather types is carried out.Firstly,based on the wind speed fluctuation characteristics,different weather types are classified,and the wind fluctuation process with different time scale under different weather types are transformed into the fluctuation process with same time scale based on polynomial fitting and sampling.Secondly,under different weather types,wind power is predicted by ARIMA,ELM and LSSVM models respectively,and the optimal prediction model is selected according to the prediction effect.Finally,based on different weather types,the optimal prediction models are dynamically and adaptively selected for short-term wind power forecasting.The result shows that compared with LSSVM,ELM and ARIMA methods,the root mean square errors of the dynamic adaptation short-term wind power prediction method based on the classification of weather type are reduced by 2.8%,4% and 6.2% respectively in short-term wind power forecasting with a time scale of 3 days.