Answers are grounded in this paper and cite their sources.
Wind–solar–coal systems require flexible generation and storage operation to accommodate renewable variability while meeting regional demand and scheduled electricity exports. Coordinating coal-fired units with molten salt thermal energy storage can improve renewable integration and operating economy. However, day-ahead schedules depend on uncertain net load forecasts, and higher forecast accuracy does not necessarily yield lower operating costs. This paper proposes a cost sensitivity-assisted forecast correction framework to improve the economic value of net load predictions. Analysis of the scheduling model links forecast errors to asymmetric balancing costs, equipment state changes, and inter-temporal storage constraints. Starting from a temporal convolutional network forecast, the framework learns cost responses to signed forecast perturbations evaluated through the complete scheduling model. These responses supervise action value differences in the critics of a soft actor–critic agent, guiding joint correction of hourly forecasts to minimize realized daily cost. Tests on the Reliability Test System–Grid Modernization Laboratory Consortium benchmark show a 7.47% reduction in mean realized cost relative to the uncorrected temporal forecast, despite higher forecast error. With identical cost sensitivity inputs, auxiliary critic supervision reduces mean cost by 2.01% compared with ordinary critic training.
No takes yet. Share an insight, caveat, or question.
Yan et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: