Counterfactual Probabilistic Electricity Load Baseline Estimation for Demand Response Using Contemporaneous Non-Activated Customers
Measuring electricity demand response requires the electricity load that an activated customer would have exhibited without activation. Selective activation and shared disturbances complicate estimation of this unobserved baseline. This study develops Quantile Generalized Synthetic Control (QGSC), combining a historical conditional quantile predictor with residual information from contemporaneous non-activated customers. A shrinkage-controlled common location shift updates the predicted quantiles, followed by empirical interval calibration using held-out historical event controls. Evaluation combines measured commercial and industrial electricity load with simulated activation and response. Five overlapping 60-customer cohorts are drawn from a single-city pool of 102 customers. Relative to a CalTRACK-style regression, QGSC reduces the 19-quantile grid approximation to the continuous ranked probability score from 170.440 to 164.458 kW, or 3.51%. Nominal 90% interval coverage increases from 77.89% to 85.16%, accompanied by a 9.79% increase in normalized width. Point-measurement mean absolute error decreases by 3.33%. Matched ablations attribute most of the grid-score reduction to contemporaneous correction, while donor and subgroup diagnostics reveal residual bias and undercoverage. These descriptive results support a modest incremental benefit for ex-post measurement under the specified retrospective conditions. They do not establish nominal coverage, statistically significant main-experiment superiority, or performance under actual demand response operations.
Authors
- Liang Shen (ORCID: https://orcid.org/0000-0001-7976-668X)
- Yukun Bao (ORCID: https://orcid.org/0000-0001-5418-8799)
Institutions
- Huazhong University of Science and Technology (CN)
Publication Details
- Journal
- Energies
- Published
- 2026-11-12
- DOI
- https://doi.org/10.3390/en19225442
- Primary Topic
- Smart Grid Energy Management
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Natural Science Foundation of China