.png){kind=icon}
.png){kind=icon}
Return
Disaggregation of Space-Heating load in buildings from hourly Smart-Meter data using a hybrid ICA–PLR approach
.png){kind=icon}
OA
AI
A
N
A
DOI：10.1016/j.enbuild.2026.117618.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Space-heating dominates electricity consumption in Norwegian public buildings, yet hourly space-heating load (SHL) is rarely sub-metered, limiting monitoring and demand response (DR). The objective of this study is to develop and evaluate a physics-guided, scalable method to disaggregate hourly SHL from whole-building hourly electric load using only weather and calendar information, without SHL sub-metering. Specifically, this study evaluates the performance of a hybrid ICA-PLR estimator for hourly SHL disaggregation against PLR, RFR, and FastICA benchmarks. This study presents a hybrid disaggregation framework that couples Independent Component Analysis (ICA) with a heating-degree-hour piecewise linear regression (HDH-PLR) under a temperature- and schedule-aware cold-hour mask. ICA extracts a heating-like latent component from cold-hour residuals across multiple buildings; HDH-PLR re-anchors scale and slope to a physics-consistent temperature signature; and mask gating with lag, seasonal, and weekend factors suppresses warm-weather false activation. The proposed framework requires only hourly smart-meter, weather, and calendar data. The method is applied to a portfolio of 22 public buildings (schools, kindergartens, offices, and institutions). All quantitative validation is performed on a sub-metered validation subset of 5 buildings using hourly SHL as ground truth. Across the validation subset, results indicate that hourly accuracy reaches R2 ≈ 0.91 with nMAE ≈ 13 % to 14 %, while estimated hourly SHL remains near zero during summer periods. At the building level in the validation subset, the method tracks morning recovery, night setbacks, and cold-spell amplitudes with fewer mild-weather false positives than the benchmark techniques. The approach yields interpretable parameters (balance temperature, lag, seasonal/weekend factors) that support weather-normalized monitoring and flexibility operations in public building portfolios. This work demonstrates that physics-guided disaggregation can extend operational visibility of public-building heating systems using only utility-grade data and support weather-normalized monitoring and flexibility planning across public-building portfolios.
Keywords:
Space-heating load
Load disaggregation
Independent Component Analysis (ICA)
FastICA
Piecewise Linear Regression (PLR)
heating degree-hours (HDH)
Smart-meter data
Random Forest Regression (RFR)
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
7.1
Papers:
1.5W
Citations:
6.8W
Organization
Cited Papers