A controlled benchmark of temporal dependence for day-ahead electricity-demand discrepancies in Great Britain - supporting data and MATLAB code

Supporting data and MATLAB code for “A controlled benchmark of temporal dependence for day-ahead electricity-demand discrepancies in Great Britain”. The retrospective benchmark compares independent, Gaussian-copula and empirical electricity-demand discrepancy ensembles using shared finite marginals and 999 ordinary training days. A separate conditional quantile model adds recorded forecast and calendar covariates. Evaluation covers 364 ordinary days in Great Britain in 2024. The package contains frozen National Energy System Operator (NESO) source snapshots, derived inputs, MATLAB analysis and verification programs, saved numerical results, figure-source arrays, fixed parameter settings, a data dictionary, file checksums and reproduction instructions. The idealized buffer is a diagnostic device and does not measure actual reserve procurement or system reliability. Historical operational availability of revised demand outturns is not established. All adverse and inconclusive findings are retained. The recorded software environment is MATLAB R2025b Update 1 with Statistics and Machine Learning Toolbox and Optimization Toolbox. Supported by National Energy SO Open Data. Provider data remain subject to the NESO Open Data Licence; file-specific rights and licensing information are documented in the package. The associated manuscript is unpublished and has no assigned publication DOI.