Continuation-specific reliability assessment of receding-horizon wind–battery–reserve dispatch
Year: 2026
Type: preprint
Abstract: Reliability estimates for receding-horizon control are frequently evaluated under a planned action sequence, although only its first command is executed before replanning. We examine how this difference changes the shortfall event attributed to an initial wind–battery–reserve dispatch decision. In paired replays, the pre-action state, first command, realised wind and synthetic delivery obligation are held fixed while the subsequent control policy varies. Across 9,010 UK development origins and 2,586 Australian transfer origins, the deployed-minus-frozen-FULL shortfall differences for deterministic model predictive control were − 6.25 and − 12.34 percentage points; differences relative to its own fixed plan were + 5.58 and + 4.18 points. An exploratory 14-day UK extension (1,985 paired origins) yielded − 0.35 and + 6.80 points, respectively; the near-zero FULL contrast changed sign under origin thinning. Identical probability forecasts also produced different Brier scores when evaluated against deployed rather than planned events. Fixed-endpoint replanning and dependence-aware sensitivity analyses further delineated the sources and stability of these differences. Reliability comparisons therefore require an explicitly specified continuation and scoring target; neither the direction of the event shift nor a cross-target score improvement is universal.
Source: Research Square
Authors: Xin He¹, Yingshuo Qi¹, Xinrong Xie¹, Yongpeng Tong¹
Institution: Shihezi University
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7236768392
