Extreme learning Kalman filter for short-term wind speed prediction

• Published in: Frontiers in energy research Vol. 10
• Main Author: Wang, Hairong
• Format: Journal Article
• Language: English
• Published: Frontiers Media S.A 12.04.2023
• Subjects: extreme learning machine; Kalman filter; neural networks; uncertain dynamical systems; wind speed prediction
• ISSN: 2296-598X; 2296-598X
• DOI: 10.3389/fenrg.2022.1047381

Accurate prediction of wind speed is critical for realizing optimal operation of a wind farm in real-time. Prediction is challenging due to a high level of uncertainty surrounding wind speed. This article describes use of a novel Extreme Learning Kalman Filter (ELKF) that integrates the sigma-point Kalman filter with the extreme learning machine algorithm to accurately forecast wind speed sequence using an Artificial Neural Network (ANN)-based state-space model. In the proposed ELKF method, ANNs are used to construct the state equation of the state-space model. The sigma-point Kalman filter is used to address the recursive state estimation problem. Experimental data validations have been implemented to compare the proposed ELKF method with autoregressive (AR) neural networks and ANNs for short-term wind speed forecasting, and the results demonstrated better prediction performance with the proposed ELKF method.