Evidence-Based Model Calibration for Efficient Building Energy Services

Energy services play a growing role in the control of energy consumption and the improvement of energy efficiency in non-residential buildings. Mos t f the energy use analyses involved in the energy efficiency service process require on-field measure ments and energy use analysis. Today, while detailed on-field measurements and energy counting stay generally expensive and time-consuming, energy simulations are increasingly cheaper due to the continuous improvement of computer speed. This work consists in the development of a simulati on-based approach dedicated to whole-building energy use analysis for use in the frame of an ener gy fficiency service process. Focus is given to th e development of a new simplified dynamic hourly buil ding energy simulation tool adapted to energy use analysis of existing buildings, its calibration by means of available energy use data and to the integration of the calibration process into the Ene rgy Service Process. The proposed evidence-based calibration methodology is deeply related to on-fie ld inspection and data collection issues and is developed to fit with the audit/inspection process. After calibration, the model can be used to suppor t the other steps of the Energy Services Process, suc h as ECOs selection and evaluation and continuous performance verification. The new systematic calibration methodology gives pr iority to the physical identification of the model’s parameters (i.e. to the direct measurement) and relies on the notion of hierarchy among the source of information (as a function of their relia bility) used to identify the value of the parameter s. The improved Morris’ sensitivity analysis method is used for “factor fixing” (i.e. distinction between non-influential model’s parameters) and “parameters screening” (i.e. classification of influential parameters by order of importance) in order to orie nt the data collection work and guide the parameters adjustment process. At the end of the ca libration process, the Latin Hypercube Monte Carlo sampling is used to quantify the uncertainty on the final outputs of the calibrated model. The developed simulation tool and the associated ca libration method are applied to a synthetic case (“Virtual Calibration Test Bed”) and to real case s tudy building located in Brussels, Belgium. Both applications (real and synthetic cases) allow highlighting the complexity and the limits of calibration as it is used today. Calibration remain s a highly underdetermined problem and a compromise has to be found between data collection and modeling efforts and model’s requirements in order to proceed to efficient energy use analysi s. At the end of these applications, it is believed that partially manual methods remain more efficient and the best quality assurance when proceeding to calibration of a building energy simulation model. The use of an evidence-based method ensures stickin g to reality and avoids bad representation and hazardous adjustment of the parameters. Moreover, i t is shown that the intensive use of a sensitivity analysis method is of a great help to orient data c olle tion and parameters adjustment processes. Defining confidence/uncertainty ranges for each par ameter, in addition to a “best-guess” value, also allowed quantifying the uncertainty on the final ou tputs of the model and helped the user in evaluatin g the quality of the calibrated model.