Answers are grounded in this paper and cite their sources.
Forecasting performance improves for renewable energy sites using advanced data techniques, suggesting better reliability in power generation.
Accurate and efficient forecasts can help address uncertainties posed by distributed renewable power generation (DRPG). Forecasting for DRPG sites requires considering several critical factors, such as large quantity, strong spatio‐temporal correlation and scarce historical measurements. To this end, a framework for jointly forecasting the outputs of multiple wind and photovoltaic (PV) power generation sites is proposed, particularly for cases where not all sites have sufficient historical measurements. Firstly, the Kriging method is used to interpolate the data for sites without any measurement and the measure correlate–predict (MCP) method to derive long‐term data for sites with only short‐term measurements. Then three joint forecasting architectures are applied and compared to forecast outputs of wind and PV sites so as to improve the modelling efficiency by aggregating different forecasting tasks into one multi‐task learning model and improve forecasting accuracy by exploiting spatial correlation among different sites. The framework is tested using data from 15 wind sites and 15 PV sites. Results demonstrate that the Kriging and MCP methods are effective for estimating data for DRPG sites without enough history measurements to build forecasting models. The multi‐task regression model based on the least absolute shrinkage and selection operator achieves the highest accuracy among the three studied architectures.
No takes yet. Share an insight, caveat, or question.
Zhao et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: