Rolling horizon control architecture for distributed agents of thermostatically controlled loads enabling long-term grid-level ancillary services

Abstract

This paper proposes novel modeling and control frameworks to evaluate the underlying capability of Thermostatically Controlled Loads (TCLs). The modeling architecture is based on distributed agents responsible for monitoring and controlling a small group of TCLs. Agents are mathematically formulated as a tracking problem in a mixed-integer linear programming format. Control actions are obtained by solving the agents sequentially in a rolling horizon control architecture. A step required to avoid the curse of dimensionality dictated by the number of devices and the duration of the planning horizon. The proposed architectures provide more practical evaluations in quantifying TCLs’ capability. In essence, individual devices are described by distinct parameters and different operating conditions imitating real-life devices found in actual distribution systems. In contrast to aggregate models developed in the literature, this work inherently allows devices’ heterogeneity, captures the outside temperature variations, and provides control access for both set-points and devices’ status.In this paper, the scope of TCLs evaluation is providing long-term services such as load-shifting and sustained load-reductions. Simulation results have shown that 1000 heterogeneous TCLs represented by air-conditioners are able to achieve both services. However, consequences on customers’ comfort are expected proportional to the extracted service. For instance, a sustained average power reduction of 319 KW causes an average set-point increase of 3.8 °F.