Dealing with storage without forecasts in smart grids: problem transformation and online scheduling algorithm

Abstract

Renewable and distributed energy sources are today possible but these technologies bring benefits as well as challenges, such as their intermittent nature, that leads to utilization problems for the power grid. On the other hand, upcoming storage technologies, such as electric vehicles, hold the potential to store and utilize this intermittent supply at a later time but bring challenges of their own, for example efficient storage utilization and intermittent energy demand. In this paper we propose a novel modelling of the problem of unforecasted energy dispatch with storage as an online scheduling problem of tasks on machines, by transforming time constraints of energy requests into equivalent machine constraints as well as by modelling energy storage through the extension of existing online scheduling techniques with the concept of load credit. Based on this transformation, we also present an algorithm that dispatches load and utilizes efficiently any storage capabilities in order to mitigate the effect of unreliable or non-existent demand forecasts, and we prove that the resulting solution's competitive ratio is within a logarithmic factor of the optimal offline solution. Finally, we provide an extensive simulation study for a variety of scenarios based on data from a large network of consumers, showing that the presented algorithm is highly competitive even to methods that assume exact knowledge about the demand requests.