Energy performance assessment: From static to dynamic benchmarking and cluster-based analysis in residential building

Benchmarking residential building energy consumption is a critical tool for evaluating performance and identifying anomalous energy usage. This research proposes a systematic methodology for developing dynamic energy benchmarks for individual residential buildings in semi-arid climates. The approach combines thermal load data from a TRNsys numerical model with in-situ data on occupancy, weather, and equipment consumption to ensure a realistic representation of building operation. The methodology involves three steps: (i) Benchmarking setup and data analysis; (ii) Cluster analysis and baseline identification for static benchmarking; and (iii) Integration of dynamic benchmarks for continuous monitoring. Using k-means clustering, the building's total energy demand is categorized into three clusters based on different data inputs that align with seasonal variability: EB 0 is designed for high cooling demand, EB 1 for high heating demand, and EB 2 for balanced energy needs (mid-season conditions). Static benchmarks set baselines within each cluster and identify potential areas for improvement. However, static benchmarks alone are insufficient for capturing fluctuations in household energy performance across time. To address this, the framework incorporates dynamic benchmarking through control charts, which continuously update with new data, enabling the detection of deviations, irregularities, or shifts in usage patterns. The proposed framework provides a transferable methodology that not only supports the assessment of prototype buildings but can also be applied to operational dwellings in a similar context. By integrating empirical datasets, machine learning techniques, and dynamic performance monitoring, the approach offers an adaptive tool to support energy management strategies while accounting for seasonal and behavioral variability. • A hybrid data-driven approach benchmarks residential energy performance. • Empirical and simulated data are integrated for dynamic benchmarking analysis. • k-means clustering identifies three seasonal energy consumption regimes. • Control charts enable anomaly detection and continuous energy performance tracking. • The framework enhances interpretability and scalability across semi-arid climates.