Skip to main content
eScholarship
Open Access Publications from the University of California

Laboratory field studies performance feedback

  • Author(s): Federspiel, Clifford C
  • Zhang, Qiang
  • Arens, Edward
  • et al.
Abstract

This report describes the development and testing of a new method of benchmarking whole-building energy use in laboratory buildings. The energy intensity of laboratory buildings is four to five times higher than that of other kinds of commercial buildings such as office buildings. This fact coupled with the importance of high-tech industries in California makes energy-efficiency of laboratory buildings an important issue in California.

The most common method of benchmarking energy use in buildings is to compare the energy use of the building under consideration with the energy use of a population of like buildings. Usually there is some empirical compensation for features and factors that affect energy use such as the size of the building and the weather conditions. Two fundamental limitations of this approach are: 1) only similar kinds of buildings can be compared, and 2) the entire population may be inefficient, which would cause many inefficient buildings to be rated as efficient. The first limitation is important when benchmarking laboratory buildings because there is no public database of energy use and building features that can be used to construct empirical benchmarks for laboratories. The second limitation is also important because there is evidence that energy-consuming processes in laboratory buildings, especially HVAC systems, are inefficient because of highly conservative design practices and the need for risk avoidance.

The benchmarking method described in this report is fundamentally different than the method described above. The principle is to construct a benchmark that represents the minimum amount of energy required to meet a set of basic functional requirements of the building. These requirements include code-compliant environmental controls, adequate lighting, etc. The benchmark is computed based on idealized models of equipment and system performance. Using idealized models produces a benchmark that is independent of design and easy to compute.

Once the benchmark has been computed for a single building, an effectiveness metric is computed by dividing the model-based benchmark by the actual consumption. This metric, or its inverse, can be compared with the metrics of other buildings. Since functional requirements have been incorporated into the benchmark, it is possible to compare the performance of dissimilar buildings, or buildings that have rare or unique functional requirements.

A benchmarking tool was developed that implements the benchmarking method described above. The tool is an MS Access database with calculation methods for implementing the model-based calculations. The database produces reports that allow a user to view historical performance trends as well as relative performance compared to other buildings in the database.

The performance of the model-based benchmarking method was compared with two alternative methods based on the ability to predict actual energy use. Using building energy data from the UC Berkeley campus, it was shown that the model-based benchmarking method was more accurate when a combination of laboratory and non-laboratory buildings were analyzed.

In addition to demonstrating the efficacy of model-based benchmarking, several other lessons were learned about building energy analysis. By constructing a model that represents the best possible performance, errors in the input data regarding schedule of operation and recorded energy use were detected because of effectiveness metrics that were significantly greater than unity. This demonstrates that recorded energy-related data may be unreliable and that the model-based benchmarking method may be able to detect errors of this kind in addition to detecting unsatisfactory energy-use performance. From analyzing building data that included a class 100 cleanroom it was clear that improvements in the fan power model could yield further improvements in the benchmarking accuracy.

Main Content
Current View