Robust Learning Control for Building Energy Systems

Colorado State University, Fort Collins CO

Investigators

Abstract

This project will develop of new control strategies for heating, ventilating, and air-conditioning (HVAC) systems that combine artificial intelligence and robust control theory. A primary objective of the work is to develop learning controllers, which provide (a-priori) guaranteed closed-loop stability even while the controller trains online. It is a follow-on project to NSF project 9804757, i8Robust Learning Control for Heating, Ventilating and Air-Conditioning Systems, it where several significant advancements in the state of the art were achieved. The primary objectives of the proposed research are: New analysis tools and algorithms, which allow for new theoretical approaches for learning including: Continuous learning algorithms; recurrent dynamic networks; and learned dynamic models. New theoretical tools, including enhanced Integral Quadratic Constraints (IQCs) for improved speed and accuracy the more general classes of networks, and advanced systematic algorithms for tunneling, to facilitate navigation of the learning around regions of instability. Design and test of more advanced robust reinforcement learning algorithms; Extensive experimental study of the new tools being developed here for multi-input, multi-output (MIMO) robust reinforcement learning control, utilizing a recently developed experimental platform for heating, ventilating, and air-conditioning (HVAC) system control. Disseminate in through conference and journal publications and by assisting colleagues at Korea Institute for Energy Research in conducting their own tests of these methods. To accomplish these objectives, an interdisciplinary team has been formed consisting of a specialist in robust control from the Electrical & Computer Engineering Department, a specialist in reinforcement learning for neural networks from the Department of Computer Science, and a specialist in design, modeling and control of HVAC systems from the Mechanical Engineering Department.