Data Predictive Control for building energy management

Abstract

Decisions on how to best optimize energy systems operations are becoming ever so complex and conflicting, that model-based predictive control (MPC) algorithms must play an important role. However, a key factor prohibiting the widespread adoption of MPC in buildings, is the cost, time, and effort associated with learning first-principles based dynamical models of the underlying physical system. This paper introduces an alternative approach for implementing finite-time receding horizon control using control-oriented data-driven models. We call this approach Data Predictive Control (DPC). Specifically, by utilizing separation of variables, two novel algorithms for implementing DPC using a single regression tree and with regression trees ensembles (random forest) are presented. The data predictive controller enables the building operator to trade off energy consumption against thermal comfort without having to learn white/grey box models of the systems dynamics. We present a comprehensive numerical study which compares the performance of DPC with an MPC based energy management strategy, using a single zone building model. Our results demonstrate that performance of DPC is comparable to an MPC controller, with only 3.8% additional cost in terms of optimal objective function and within 95% in terms of R2 score, thereby making it an alluring alternative to MPC, whenever the associated cost of learning the model is high. Abstract-Decisions on how to best optimize energy systems operations are becoming ever so complex and conflicting, that model-based predictive control (MPC) algorithms must play an important role. However, a key factor prohibiting the widespread adoption of MPC in buildings, is the cost, time, and effort associated with learning first-principles based dynamical models of the underlying physical system. This paper introduces an alternative approach for implementing finite-time receding horizon control using control-oriented data-driven models. We call this approach Data Predictive Control (DPC). Specifically, by utilizing separation of variables, two novel algorithms for implementing DPC using a single regression tree and with regression trees ensembles (random forest) are presented. The data predictive controller enables the building operator to trade off energy consumption against thermal comfort without having to learn white/grey box models of the systems dynamics. We present a comprehensive numerical study which compares the performance of DPC with an MPC based energy management strategy, using a single zone building model. Our simulations demonstrate that performance of DPC is comparable to an MPC controller, with only 3.8% additional cost in terms of optimal objective function and within 95% in terms of R 2 score, thereby making it an alluring alternative to MPC, whenever the associated cost of learning the model is high.

Smart power consumption abnormality detection in buildings using micromoments and improved K‐nearest neighbors

“…Finally, it is worth noting that a great interest has been devoted recently to modeling and simulating energy saving in buildings using novel ML models, which helps tremendously in cutting down the model development time to days. This is in contrast to its competing techniques, such as physics‐based models, which generally require months to be constructed 62,63 …”

A data-centric predictive control approach for nonlinear chemical processes

“…Another issue is that this method cannot provide better control than what has already been implemented in the existing control system. Potential approaches [14][15][16] on combining machine learning algorithms (e.g., Regression Trees and Random Forests) with predictive control were recently developed for building energy management. The big operational data are converted into black-box input-output static models, which are further applied to the receding horizon control.…”

“…However, as a consequence of using static input-output models, this approach does not consider the internal state evolution and may lose the information of the past inputs applied to the system over prediction horizons, which could lead to a loss of control performance [17]. Furthermore, the existing DPC schemes (e.g., [14,15]) do not ensure closed-loop stability. Applications of these approaches to mission-critical chemical processes without stability guarantees may have significant risks.…”

Modeling and optimization of complex building energy systems with deep neural networks

“…These data along with future weather forecasts can be utilized for data-driven real-time optimization approaches. In [7], the authors developed a data predictive control method to replace the traditional MPC controller by using data to build a regression tree that represent the dynamical model for a building. However, regression trees still results in a linear model that can be far away from the true dynamics of building systems.…”