Short-term wind forecasting using spatio-temporal covariance models

Speaker: Tianxia (Tylar) Jia

Date: Wed, Jul 26, 2023

Location: PIMS, University of British Columbia, Online

Conference: PIMS-FACTS Workshop on Forecasting and Mathematical modeling for Renewable Energy

Subject: Mathematics, Applied Mathematics, Atmospheric and Oceanic Physics: Climate Modelling

Class: Scientific


This talk introduces a methodology for improving short-term wind speed forecasting in Alberta. Regime-switching spatio-temporal covariance models are applied using two datasets: (1) large-scale reanalysis dataset containing large scale atmospheric information for atmospheric clustering using k-means and hidden Markov models; (2) wind speed data from 131 weather stations across Alberta are used to train and test the covariance models. The predictive performance is assessed for different models and clustering methods.