Grouped Gaussian Processes for Solar Power Prediction

TL;DR

This paper introduces a scalable multi-task Gaussian process model with coupled priors for improved solar and wind power forecasting, enhancing accuracy and uncertainty quantification over existing benchmarks.

Contribution

It proposes a novel coupled Gaussian process framework that exploits spatial dependencies for multi-site renewable power prediction, demonstrating superior accuracy and uncertainty estimates.

Findings

Improved point-prediction accuracy for solar and wind forecasts.

Enhanced quantification of predictive uncertainties.

Faster gains in forecast accuracy compared to models without coupled priors.

Abstract

We consider multi-task regression models where the observations are assumed to be a linear combination of several latent node functions and weight functions, which are both drawn from Gaussian process priors. Driven by the problem of developing scalable methods for forecasting distributed solar and other renewable power generation, we propose coupled priors over groups of (node or weight) processes to exploit spatial dependence between functions. We estimate forecast models for solar power at multiple distributed sites and ground wind speed at multiple proximate weather stations. Our results show that our approach maintains or improves point-prediction accuracy relative to competing solar benchmarks and improves over wind forecast benchmark models on all measures. Our approach consistently dominates the equivalent model without coupled priors, achieving faster gains in forecast accuracy. At the same time our approach provides better quantification of predictive uncertainties.