Wind speed forecasting in different seasons using ELM batch learning algorithm in Indian context

Abstract

Efficient wind speed forecasting is important for wind energy sector for better wind power integration. This paper focuses on developing seasonal wind speed forecasting models in Indian context. Wavelet transform (WT) technique has been used for denoising the data obtained from supervisory control and data acquisition (SCADA) of a 1.5 MW wind turbine located in central dry zone of Karnataka, to reduce the unnecessary fluctuations in the wind speed time series. Partial auto correlation function (PACF) has been used for selection of input parameters, which greatly influences the forecasting accuracy. Forecasting models have been developed using a fast and efficient extreme learning machine (ELM) algorithm. The results have been compared with conventional back propagation (BP) algorithm. The results show that the seasonal models developed using ELM have better forecasting performance compared to BP. © 2018 Authors.