A0892
Title: Modeling large-scale high-resolution spatial-temporal data with deep ESN and SPDE
Authors: Kesen Wang - King Abdullah University of Science and Technology (Saudi Arabia) [presenting]
Abstract: In the past decades, clean and renewable energy has gained increasing attention due to a global effort to reduce carbon footprints. Wind, a primary form of renewable energy, has been heavily invested as a substitute and replenishment for the existing energy portfolio. However, wind possesses a highly spatially non-linear dynamic nature and varies rather capriciously in time, rendering wind modelling quite challenging. Given the complex dynamics of the wind, it is challenging for the existing statistical models to fully grasp the underlying spatial and temporal dependence structures. Hence, there is a need for a non-linear dynamic model of high temporal resolution, which precisely captures the spatial dependence and accurately describes the fast-evolving wind dynamics in time. A combined model based on machine learning and stochastic partial differential equation (SPDE) is proposed to address the complex wind dynamics in time and preserve dependence structure in space.
