Claim Missing Document
Check
Articles

Found 1 Documents
Search

Mathematical Modeling of Household Electricity Use Based On Time Variation and Appliance-Use Intensity in Minasa Upa, Makassar, Indonesia Auliah Rahmah; Fitriyanty Dwi Lestary; Aldiansyah Nur Alam
Journal of Electrical Engineering and Informatics Vol. 4 No. 1 (2026): Journal of Electrical Engineering and Informatics
Publisher : Fakultas Teknik Universitas Negeri Makassar

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59562/jeeni.v4i1.15068

Abstract

This study develops a parsimonious and physically interpretable framework for describing household electricity use by combining appliance operating intensity, dominant temporal-use patterns, and an empirical consumption model. A quantitative survey was conducted among 93 households. Descriptive analysis characterized appliance operating intensity and dominant use periods, while the primary empirical model used 37 observations with explicitly recorded monthly electricity consumption in kWh. Candidate models were assessed using residual diagnostics, heteroskedasticity and normality tests, multicollinearity assessment, and five-fold cross-validation. The selected model was ln(E) = -5,3721 + 1,5946 ln(D), where E is monthly electricity consumption and D is installed power. The regression explained 57,23% of variance on the logarithmic scale (adjusted R² = 0,560). After smearing correction and back-transformation, the observed–predicted R² was approximately 0,679. The model showed no statistical evidence of heteroskedasticity (Breusch-Pagan p = 0,960) or non-normal residuals (Jarque-Bera p = 0,935), with VIF = 1,00. Five-fold cross-validation produced R² = 0,560, MAE = 195,22 kWh, and RMSE = 248,85 kWh. The study contributes a validation-oriented two-layer framework that separates a physical appliance-intensity formulation from an empirical household consumption model. This design preserves interpretability while avoiding unverified reconstruction of kWh from billing values and demonstrates how parsimonious modeling can remain useful when field data are heterogeneous and high-resolution smart-meter records are unavailable.