Using Temporal Constraint Networks for Smart Grid Scheduling

Abstract

Smart Grid scheduling problems are characterized by quickly changing situations and multiple external factors that cannot be controlled. Most smart grid research applies stochastic models over the total power consumption of a household or system to find a schedule that achieves an optimization, a balancing, or constraint satisfaction. While these solutions are able to efficiently capture the unpredictability in the problem, they only make use of a limited amount of information. These models are not yet able to make flawless decisions, so we try to optimize this by using a model that is able to use more information. A different scheduling model that is able to do so is the Temporal Constraint Satisfaction Problem (TCSP). However, the information that this model needs requires substantial effort to obtain in smart grid environments, especially for academic purposes. For this reason, research into the benefits of doing so is scarce.

In this paper we attempt to use this model with detailed environmental information, specifically the activation of electric devices in households, to better optimize a motivating smart grid scheduling problem. We find that problems with characteristics typical in smart grids are difficult to express as TCSP. We discuss these characteristics and provide the concept of non-binary conditions and conditional preference to solve these difficulties and provide more expressive conditional reasoning about optimality in TCSP.