Improved nonlinear mapping network for wind power forecasting in renewable energy power system dispatch

Abstract

Wind power forecast (WPF) plays vital role in renewable energy power system dispatch. Inaccurate WPF caused by the stochastics and fluctuation of wind energy would lead to frequently modify power system dispatch scheduling and even endanger power system stability operation. Time-series analysis method can find hidden patters of historical wind power time-series but neglects the influence of meteorological factor on WPF accuracy, so it cannot deal with unexpected scenarios such as no wind scenario. Numerical weather forecast (NWF) method cannot accurately forecast wind power in wind curtailment scenario. To overcome these drawbacks mentioned above, a new WPF model based on nonlinear mapping network (NMN–WPF) is proposed to forecast wind power. By combining the advantages of time-series analysis and NWF methods, the proposed NMN–WPF method not only models hidden pattern of wind power time-series, but also adopts meteorological data from NWF system as external factor. Hence, the proposed method can effectively deal with the problem of wind curtailment and no wind scenarios because it balances both interior factor of wind power time-series data itself and external factor of meteorological data. Test cases are performed for practical wind plants. WPF results show that the proposed method can accurately forecast wind power and it is very suitable for renewable energy power system dispatch.