A novel wind model downscaling with statistical regression and forecast for the cleaner energy

Abstract

This study proposes an innovative statistical downscaling approach for meteorological wind models, enhancing cleaner energy applications. This method efficiently and swiftly incorporates topographic and meteorological data into wind park modelling by integrating regression-diagnosis-feature selection with statistical regression and forecasting techniques. The novel method in a de facto reduce predictive errors by approximately an average of 23.21%, lower errors downscaling and forecasting by 5.94% compared to existing methods, and optimise the balance between computational efficiency and forecasting accuracy. In an Arctic wind park case study, we systematically analysed meteorologically simulated wind speeds at various geographic points surrounding the park. The method enabled effective meteorological statistical mapping and accurate wind speed forecasting. Additionally, using the Granger causality test, the model identified a significant influence of surrounding terrain features on wind patterns. This research contributes by 1. introducing a novel downscaling technique surpassing current methods, 2. providing a resource-efficient approach for managing wind energy in complex terrains, and 3. offering a versatile model applicable to broader scenarios. The work marks an advancement in the field, promoting more reliable and efficient wind power and highlighting the importance of statistical downscaling in meteorological contexts for cleaner applications.