A robust load profiling framework using Mahalanobis distance-dependent Chinese restaurant process

Read full text Summarise with VIRAT
Authors: Siqi Zhang, Xinpeng Jiang, Yilin Ge, Zhihong Feng, Pei Zhou, Donglian Qi
Cited by: 0
References: 18

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Abstract Load profiling plays an essential role in demand-side management, yet the high dimensionality and strong temporal dependence of smart meter data make accurate customer clustering a challenging task. Many existing methods struggle to capture the intrinsic structure of daily consumption patterns and therefore produce suboptimal segmentation results. To overcome these limitations, we develop a Mahalanobis Distance-Dependent Chinese Restaurant Process (MDDCRP), a nonparametric Bayesian approach that explicitly incorporates temporal covariance into the clustering process. By introducing a regularized Mahalanobis distance, the proposed model effectively accounts for temporal correlations and enables more reliable geometric separation among load profiles. A complete theoretical formulation is provided to ensure efficient and tractable inference. Experimental evaluations conducted on multiple real-world datasets show that MDDCRP consistently delivers higher segmentation accuracy than conventional approaches, including K-means, highlighting its effectiveness and robustness for practical utility analytics.

Fields of Study & Subjects

Smart Grid Energy Management Bayesian Methods and Mixture Models Energy Load and Power Forecasting

Publication details

Journal / source
Journal of Physics Conference Series
Volume
3192
Issue
1
Pages
012015–012015
Publication date
1 March 2026
Language
EN