Building energy modeling (BEM) is increasingly used in building industries for a variety of purposes, such as code compliance evaluation for new construction, continuous commissioning for existing buildings, and quantification of savings from energy efficiency projects. Building energy models without proper validation and calibration will lead to significant discrepancy between projected and actual building energy consumption. Deficiencies in predictive accuracy and consistency of BEM were identified as a barrier to increase BEM use. Supporting development and use of methods for model input calibration is identified as an initiative to address this problem. The goal of the proposed project is to improve the predictive accuracy and consistency of building energy models by providing an educated and automated multi-stage calibration tool. Based on sensitivity analysis, meta-modeling, and gradient-based optimization, the proposed tool will identify the subset of model inputs that has the greatest influence on the model output and optimize this subset by minimizing the error between model output and monthly utility data. Compared with the existing model calibration tools and processes, the proposed tool has a higher level of automation while enabling human-computer interaction for model input screening, measured data screening and post-calibration sanity check. This process will add physical insight and reduce the number of parameters and uncertainty bounds in the calibration process. The complexity reduction through meta- modeling in the proposed tool can also make the calibration process less computationally expensive. In Phase I, algorithms for the multi-stage automated calibration process will be developed. The algorithms will then be validated using sub-metered field data from an institutional building with 93,600 ft2 area, and finally, the algorithms will be integrated as a package offered as a single tool. The proposed tool will calibrate building energy models with less time and effort. The calibrated model will be crucial for analysis of potential energy savings in deep energy retrofits of existing commercial buildings. The calibrated model can also provide information on the operation of building sub-systems, which can be used for early operating fault detection and identification of potential energy conservation measures. In next phases, the methodology and tool will be improved for further commercialization and adoption in building industries. In addition, other markets, such as building automated management and utility companies, will be explored, where accurate prediction of electric demand by using calibrated models is needed for building-grid interactive controls.
