Performance Limits Of Thermostatically Controlled Loads Under Probabilistic Switching

This paper establishes performance limits for probabilistic control schemes that coordinate populations of thermostatically controlled loads (TCLs) for the purpose of providing power system services, such as regulation and load following. In the literature, a common strategy for dispatching TCLs is to send probabilistic on/off commands, i.e. "switch on with 20% probability"; this is what we refer to as probabilistic switching. There is ongoing research to improve the tracking performance of TCL control schemes that utilize probabilistic switching the purpose of this paper is to provide a theoretically based performance limit for these control schemes. We first show how to analytically obtain this performance limit when TCLs are sent a uniform switching probability. We then extend it to the non-uniform case where different probabilities are sent for TCLs at different temperature ranges. We demonstrate the effects of both population size and the magnitude of the switching probabilities on a system's performance. Based on this knowledge, we develop non -uniform probabilistic control schemes that minimize the variance of the error due to probabilistic switching, and can avoid short cycling of TCLs. (C) 2017, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.