Research On Short-Term Forecast Of Wind Power On A Large Scale Wind Farm

Chinese wind plants are large-scale plants and most of them are located in where the load is few. It's a challenge to the power system safely and stably operation that the wind's intermittence and random which would be an obstacle to the wind power development in the future. Taking a wind power plant in Nei Monggu as an example, the paper studied some kinds of forecasting models. The paper's main works are as follows:According the the established correlation coefficient matrix of wind turbines, turbines' operation datas were preprocessed. The forecasting parameters are focused by studing the impact mean impact value to power. Establishing each turbine's power curve in different time is reasonable.The time series model is simple and has acceptable result in short-term prediction, and it is the common linear prediction model. Neural network model can deal with the fluctuation in time series, has abundant theory and high accuracy and is common non-linear prediction model. SVM has higher convergence speed, good leaning and generalization ability. SVM can effectively predict the trend of wind speed even the wind speed fluctuation is severe.A hybrid model integrating time series analysis with wavelet decomposition can do well with non-stationary time series, and has good prediction accuracy. Hybrid algorithm integrating time series analysis with Kalman filter hardly increases the computational complexity and enhances the order of established time series, but model's accuracy was largely improved.The integrated model based on maximum entropy principle can effectively improve the robustness and prediction accuracy.