Ultra-short-term wind-speed bi-forecasting system via artificial intelligence and a double-forecasting scheme

Abstract

A key issue in the wind power industry is the accuracy and stability of wind-speed forecasting, which has an important impact on the power system management and market economy. Accurate prediction of wind speed points could guarantee regular work of smart grids, while interval prediction could evaluate the uncertainty of wind speed and further supply corresponding wind speed information. However, several studies concentrated on the deterministic point forecast of wind speed and seldom considered the interval forecast of wind speed, or performed the two separately, thereby making wind speed prediction inefficient and not comprehensive. To remedy this, we herein propose a wind speed bi-forecasting system that includes three parts: data preprocessing, combination forecasting, and evaluation, making up for the gaps in existing research and analysis. A decomposition algorithm is used to decompose the actual wind speed into a series of finite components as well as remove high-frequency noise components. The combination forecasting module adopts the weight combination mechanism based on a multi-objective multi-verse optimization algorithm to achieve a double output of prediction results. Finally, an evaluation module comprising hypothesis testing and evaluation standards was presented to comprehensively evaluate the system. The experimental results indicate that the system is excellent for predicting wind speed and analyzing the uncertainty of wind speed and thus could be practically applied as a wind power system programming technology.