Grouped Gaussian Processes for Solar Power Prediction
This work addresses scalable forecasting for renewable energy generation, which is incremental as it builds on existing Gaussian process methods with spatial coupling.
The paper tackled the problem of forecasting distributed solar and wind power generation by proposing a multi-task regression model with coupled Gaussian process priors to exploit spatial dependence, resulting in maintained or improved point-prediction accuracy relative to benchmarks and better uncertainty quantification.
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.