Low-Cost Dynamic Model Capture For Buildings

A fundamental problem in the design of closed-loop Cyber-Physical Systems (CPS) is in accurately capturing the dynamics of the underlying physical system. In the context of modeling buildings to improve their energy efficiency, the modeling difficulty arises due to the fact that each building is designed and used in a different way and therefore, it has to be uniquely modeled (Heterogeniety). Furthermore, each building system consists of a large number of interconnected Heating Ventilation and Air conditioning (HVAC) subsystems which interact in a complex manner and are subjected to time varying environmental conditions (Complexity).
Small and medium sized buildings constitute more than 90% of the commercial buildings stock in the United States, but only about 10% of such buildings are equipped with a building automation systems. The market penetration of intelligent building control will only increase if the cost of capturing the building models is reduced.

IMpACT is a methodology and a toolbox for analysis of uncertainty propagation for building inverse modeling and controls. Given a plant model and real input data, IMpACT automatically evaluates the effect of the uncertainty propagation from sensor data to model accuracy and control performance. It also utilizes non parametric statisticals method to quantify the uncertainty in the sensor measurements and to determine near optimal sensor placement and density for accurate signal measurements.

The information content in the data used for model training may vary because of the measurement process and system's operation. Data with high information content with respect to parameters to be estimated should yield more accurate estimates and result in a better model. Moreover, we want to avoid running experiments/functional tests in a building longer than necessary. The goal of this work is to successfully automate the model training procedure​ and find the optimal input signal trajectory which maximizes the information about the model parameters subject to operational constraints.​

Observing the actual building rather than a simulation of the building will move the advanced HVAC system automation much closer to reality much faster, and at lower cost. This means starting with a generic description of the building, then adjusting the model behavior through customizable, adjustable parameters. Using re-enforecment learning methods, the predictions that are generated by the model become closer and closer to matching reality.