A novel integrated method for short-term wind power forecasting based on fluctuation clustering and history matching

Abstract

The intermittent and randomness of wind power poses a great challenge to the safe and stable operation of power systems. Accurate wind power forecasting is a key to reduce the negative impact of uncertainty on the entire system. To this end, a novel integrated method for short-term wind power forecasting is proposed based on frequency analysis, fluctuation clustering and history matching. Through variational modal decomposition (VMD) and Pearson correlation coefficient between wind resource and wind power, the wind power sequence is decomposed into a trend component and a fluctuation component. Then both the two components are sequentially divided into time series segments with equal lengths, which are set as the variation period determined by fast Fourier transform (FFT). Time series segments of the trend component are clustered by feature extraction and fuzzy C-means (FCM) clustering method, and accordingly the segments of fluctuation component are clustered as well. The correlation consistency between trend components and fluctuation components is further studied. Finally, the trend component is forecasted by random forest (RF) model, and the fluctuation component is forecasted by matching its counterpart trend component from historical similar segments database based on the correlation consistency. The performance and effectiveness of the proposed method are verified by using four wind farms in Jilin Province, China. Results indicate that the forecasting accuracy of the proposed method is greatly improved compared with the conventional forecasting model, which proves the validity.