Energy Performance Methodologies

Evaluating energy performance in non-domestic buildings: A review

Authors EH Borgstein, R Lamberts, and JLM Hensen, published in Energy and Buildings, 2016.

This paper is available in PDF form at researchgate.net.

The Abstract gives the topic as the methods used in evaluating the energy performance of buildings, and notes the fundamental importance of such evaluation in the context of efforts to reduce worldwide energy use.

A comprehensive review is claimed; there are more than two hundred references. Five categories of methodology are reviewed: Engineering calculation methodologies; Simulation; Statistical methods; Machine learning; Other building evaluation techniques. This is followed by two further sections; Public evaluation systems and standards, and Discussion.

Engineering calculations methodologies typically involve the implementation of a simplified mathematical equation based on the physics of buildings. Often they use steady state models that average variables over a long period and assume fixed building parameters. Such models may be used early in the design process, and for estimating energy performance. Approaches with varying levels of sophistication are discussed together with their uses and limitations.

Simulation of building energy performance allows dynamic thermal modelling through the use of computer models. Building energy simulation software can be applied to new buildings for compliance evaluation, but can also be used to model operational performance. The evaluation of building performance through simulation may involve representing geometry and material properties, aspects of the external environment such as orientation and shading, proposed usage patterns and schedules of operation. Dynamic performance simulation may cover periods of up to a year, use measured climatic data, and may require multiple iterations. A number of building simulation software packages are discussed.

Statistical methods predict and evaluate energy performance based on existing datasets of multiple buildings, and are often used to form the basis of benchmarks and evaluation systems. Such evaluation is often based on consumption per unit of floor area, since this information is easy to obtain for many buildings. The principal statistical methods for benchmark development and evaluation of building energy performance are described.

Machine learning is one of the big data techniques made possible through advanced building management systems, remote metering, and the regulated publishing of energy data on buildings. Artificial Neural Networks, Clustering Analyses, and other machine learning methods are discussed.

Other building evaluation techniques include dynamic methods and real-time analysis, load-curve analysis and energy bill disaggregation, energy audits, post-occupancy and user satisfaction evaluations.

Public evaluation systems and standards describes the practical application of the technical benchmarking methodologies reviewed earlier, through voluntary and mandatory evaluation systems and standards.

The Discussion attempts to compare the methodologies listed above in terms of their inputs, accuracy, applications and restrictions: “statistical benchmarking is effective for identifying the energy performance level of a building, but does not give a detailed understanding of the underlying mechanisms and reasons for the performance. Machine learning methods also operate without understanding of the physical characteristics of buildings, which may limit their effectiveness in identifying energy performance improvements. Engineering calculation methodologies are effective for identifying the improvement potential from specific retrofit measures, but do not have high levels of reliability. Simulation models can be highly accurate, but this requires extensive calibration and large amounts of building detail, which limits their application.”