Automatic Calibration of a Building Energy Performance Model and Remote Fault Detection For Continuous Commissioning Using a Global Optimization Program

Abstract

Building simulation models have been used for many years to forecast and analyze a building's energy consumption. Over the past several decades, building energy simulation programs and input techniques have been developed significantly. This advance has also led to input variables becoming more complicated. Accordingly, calibrating these simulation-input variables so that the simulation output matches the measured cooling and heating data is also a very complicated and time-consuming process. In order to analyze the actual building's energy performance, the simulation model should be calibrated to generate output trends similar to the energy consumption of the building modeled both in magnitude and pattern. The simplified simulation program used in this thesis, which is a coding of the ASHRAE 'Simplified Energy Analysis Procedure' (Knebel, 1983) is introduced and tested. It is then successfully used to calibrate input variables automatically for a number of test cases using synthetic data, or data generated by the simulation program. The possibility of remote fault detection has also been tested successfully. The calibrations reported were performed by the nonlinear optimization method provided in Solver® in Microsoft Excel®. The procedure for automatic calibration is a global optimization process minimizing the error between synthetic "measured" and simulated energy consumption. The automatic calibration program uses the equations in the simulation program as constraints for calibration (optimization process). In other words, the calibration process is directly guided by the mathematic relationships among various variables and constants in the simulation-model. This ensures that the calibration steps are performed considering all the variables at the same time, understanding the characteristics of a simulation model. This automatic, accurate, and fast calibration method may also be used to facilitate remote fault detection. A remote fault-detection program would obtain HVAC-system-related parameters from calibrating a simulation to the measured energy consumption data. This program would then automatically provide information on the details of an operating fault in the HVAC system to the continuous commissioning engineer. This concept is tested in this thesis using synthetic data.