Intelligent BEMS design using detailed thermal simulation models and surrogate-based stochastic optimization

The topic of optimized building operation has attracted considerable interest in the research community: in this context model-based supervisory control design approaches have been shown to yield effective/optimized operation with regards to energy performance or other related operational parameters. A hindrance towards the adoption of such methodologies is the need for a mathematical model tailored to each building which is capable of capturing all pertinent dynamics. Developing and tuning such a model can be a time-consuming and costly proposition, and is the main reason why such approaches have found little applicability beyond the research space. The utilization of models constructed in the building design phases - for the reason of estimating energy performance - properly adapted for the task at hand can be a viable methodology to overcome this problem. We present in this paper, an online process where a stochastic optimization algorithm utilizing a detailed thermal simulation model of the building along with historical sensor measurements and weather and occupancy forecasts, is used to design effective control strategies for a predefined period. A detailed description of the methodology is provided and the proposed approach is evaluated on a heating experiment conducted in a real building located in Greece.