Joint Spatio-Temporal Simulation of Gridded Wind-Solar Fields

Abstract:

The inherent variability in wind and solar power availability can expose electricity systems to risks of under-supply. These risks are poised to become larger as renewable energy sources make up higher proportions of the power supply. Estimating the grid-level risks posed by wind and solar variability is difficult given limited data records and often requires simulation models for scenario generation. Surface wind and solar fields are correlated in space and time, and with one-another. Appropriately estimating the risk of under/over-supply relative to demand requires simulating the spatio-temporal structure of these fields, as well as the probability distribution of the appropriately aggregated (to account for the spatial distribution of the installed capacity) total energy that can be produced for each time horizon. We introduce a novel non-parametric K-Nearest Neighbors spatio-temporal bootstrap that captures the spatio-temporal dynamics of wind and solar fields, their joint dependence, and the probability distribution of the aggregated energy over a domain for each time step. Daily and monthly models are considered for simulation and near term forecasting using the Texas Interconnect as a case study. Simulations reproduce a variety of chosen statistics of the historical data with an appropriate uncertainty distribution. The application of the method for system operation and design considering wind and solar droughts is discussed.