Multi-timescale scheduling of multi-energy complementary system for desert-gobi-wasteland region considering source-load uncertainties

The high-efficiency integration of renewable energy into multi-energy complementary systems (MECS) is imperative for sustainable power generation. However, it continues to be challenging due to the inherent uncertainties of renewable sources and multi-energy loads. To tackle this issue, this study develops a novel multi-timescale scheduling framework that integrates Copula theory to model photovoltaic (PV)-wind power correlations. To enhance load-side uncertainty management, an improved wild horse optimizer (IWHO) combined with a deep echo state network (DESN) is developed for electric and heat load forecasting. The forecasting results are applied to a three-layer scheduling model including day-ahead, rolling, and real-time scheduling for MECS in “Desert-gobi-wasteland” regions, analyzing how flexible load responses and energy storage improve economic efficiency. Comparative analyses of typical spring and summer days show significant differences in energy balancing dynamics. Using a typical spring day as an example, rolling and real-time scheduling achieved cost reductions of 16.51 % and 19.96 %, respectively, compared to day-ahead scheduling, with wind and solar curtailment costs decreasing by 47.1 % and 69.73 %. Energy storage optimizes renewable energy utilization, further reducing cost. The integration of reducible and transferable load strategies into day-ahead scheduling resulted in additional cost reductions of 1.83 % and 0.99 %, respectively. This research demonstrates that multi-timescale scheduling and flexible load responses are pivotal for enhancing the operational efficiency of MECS in modern power systems, providing actionable strategies for renewable integration and uncertainty mitigation. ● A multi-timescale scheduling approach including day-ahead, rolling, and real-time enhances system performance. ● Copula theory models renewable energy output correlations, while improved wild horse optimizer improves load forecasting. ● Rolling and real-time scheduling cut costs by 16.51 % and 19.96 %. ● Day-ahead costs by 1.83 % and 0.99 % by reducible and transferable load strategies.