What does this research mean for the field?

Load typology, rather than model sophistication, is the dominant factor determining the fit quality of synthetic load profile generation models, necessitating typology-specific model selection. Novelty: ClaimNovelty.NOVEL_FINDING. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

To evaluate and compare the cross-typology transferability of data-driven and bottom-up stochastic models for generating synthetic electrical load profiles in residential and commercial environments.

September 15, 2026ElectricityOpen Access

Synthetic Load Profile Generation for Residential and Commercial Loads: A Comparative Study of Stochastic Models

View Full Paper

Ask AI

Bookmark

Authors

JJJuan Jiménez RIRicardo Isaza-Ruget JGJavier Rosero García

Discussion

Member takes

Overview

Comparative study reveals that load typology dictates stochastic model accuracy across residential and commercial demand, indicating that model selection requires typology-specific trade-offs.

Key Points

To evaluate and compare the cross-typology transferability of data-driven and bottom-up stochastic models for generating synthetic electrical load profiles in residential and commercial environments.
Evaluated three stochastic models: a first-order autoregressive profile (AR(1)), a physically constrained ON/OFF event model, and a nonlinear-least-squares (NLS) calibrated model.
Analyzed Class A power-quality records at 10-minute resolution from a single-family home (13 days, N=1,860 samples, skewness 6.56) and an institutional commercial building (9 days, N=1,333 samples, skewness 3.47).
Benchmarked models using six distributional statistics (mean, standard deviation, p50, p90, p95, p99), Kolmogorov–Smirnov statistics, root mean square error, and hourly variance profiles.
For commercial loads, the simple AR(1) model reproduced all percentiles within 7%, while the NLS-calibrated variant achieved upper-tail errors under 2% at a computational cost roughly 1,000 times higher.
For residential loads, no tested model reproduced the distribution center, dispersion, and upper tail simultaneously within a 10% error margin.
Residential upper-tail demand showed a trade-off where AR(1) had minimal bias (p99 error: −5.8%) but high variation across realizations (CV = 15.3%), whereas the NLS model achieved precise p95 replication (−0.3% error).

Cite This Study

Jiménez et al. (2026) studied this question.

synapsesocial.com/papers/6aa913a29013453be30a1b8d https://doi.org/10.3390/electricity7030105

View Full Paper

Ask AI

Bookmark

September 15, 2026ElectricityOpen Access

Synthetic Load Profile Generation for Residential and Commercial Loads: A Comparative Study of Stochastic Models

View Full Paper

Ask AI

Bookmark

Authors

JJJuan Jiménez RIRicardo Isaza-Ruget JGJavier Rosero García

Discussion

Member takes

Overview

Comparative study reveals that load typology dictates stochastic model accuracy across residential and commercial demand, indicating that model selection requires typology-specific trade-offs.

Key Points

To evaluate and compare the cross-typology transferability of data-driven and bottom-up stochastic models for generating synthetic electrical load profiles in residential and commercial environments.
Evaluated three stochastic models: a first-order autoregressive profile (AR(1)), a physically constrained ON/OFF event model, and a nonlinear-least-squares (NLS) calibrated model.
Analyzed Class A power-quality records at 10-minute resolution from a single-family home (13 days, N=1,860 samples, skewness 6.56) and an institutional commercial building (9 days, N=1,333 samples, skewness 3.47).
Benchmarked models using six distributional statistics (mean, standard deviation, p50, p90, p95, p99), Kolmogorov–Smirnov statistics, root mean square error, and hourly variance profiles.
For commercial loads, the simple AR(1) model reproduced all percentiles within 7%, while the NLS-calibrated variant achieved upper-tail errors under 2% at a computational cost roughly 1,000 times higher.
For residential loads, no tested model reproduced the distribution center, dispersion, and upper tail simultaneously within a 10% error margin.
Residential upper-tail demand showed a trade-off where AR(1) had minimal bias (p99 error: −5.8%) but high variation across realizations (CV = 15.3%), whereas the NLS model achieved precise p95 replication (−0.3% error).

Cite This Study

Jiménez et al. (2026) studied this question.

synapsesocial.com/papers/6aa913a29013453be30a1b8d https://doi.org/10.3390/electricity7030105

View Full Paper

Ask AI

Bookmark