Research On Uncertainty Analysis Method For Wind Power Prediction And Its Application

Wind power prediction is one of effective ways of circumventing adverse effects from large-scale wind power integrating into electrical system. Due to fluctuations of wind power and accuracy of power prediction, a single point estimate triggers a great risk to dispatching planning and wind farm operation decision making. Uncertainty analysis of wind power prediction can provide possible fluctuation range of power prediction under certain confidence level which largely improves risk-resistance ability. In the thesis, research on uncertainty analysis of wind power prediction and its application has been undertaken and following works were conducted:1. Research on influencing factors of wind power predictionTwo factors have been considered:numerical weather prediction (NWP) and forecasting model. Wind speed, wind direction, temperature, pressure and relative humidity from NWP and mast measurement for each month have been analyzed to obtain their characteristics of statistic and error distribution. Then, the way how training samples influence prediction accuracy has been studied for BP neural network optimized by Genetic Algorithm through refinement of samples' time unit. As for support vector machine, a triple selection mechanism of training samples is presented in order to make the error analysis on prediction.2. Quantile regression (QR)-based uncertainty analysis of wind power predictionA risk index of wind power prediction termed as PaR (Predict at Risk) has been firstly defined, which is a quantitative description of maximum forecasting error under a confidence level. Secondly, the source of uncertain factors for two power prediction methods has been studied, including historical data-based prediction method and numerical weather prediction data-based method. By this means, it is to establish according QR based-uncertainty analysis models. Results validate model simplicity, computation efficiency and practicality.3. Development of wind power prediction systemA wind power prediction system has been developed following specific requirements of both electrical system and wind farm owners in northern China. On the basis of established models including uncertainty analysis models for short-term prediction and ultra short-term prediction, the system could provide results of forecasting and its uncertainty analysis, statistic analysis of historical data, error analysis and so forth. The system is characterized in user-friendly graphical interface, safe and stable operation, and support for remote login which meets the national regulations on wind power prediction.