Efficient Heuristics for Power Constrained Planning of Thermostatically Controlled Loads

Thermostatic loads are promising energy buffers that could be unlocked through Demand Response. Existing research has shown that an aggregation of thermostatic loads can be made to follow a power curve using reactive control. In this paper we investigate the use of planning to proactively control an aggregation of Thermostatic Loads as storage to overcome temporary dips in power availability. We first present a formal problem definition of the planning problem under consideration, and apply Dynamic Programming to it to obtain an optimal solution. However, because we can prove the problem to be NP-complete, we need efficient heuristic solutions to solve practical instances. Therefore, we then extend the Dynamic Programming solution with approximations of the state and action space to obtain efficient heuristic solutions. We evaluate our proposed heuristics through simulation on instances with a range of power capacity constraints, and show that they are able to obtain a factor two improvement over the reactive approach under the hardest power constraints.