An Approach to Runtime Synthesis of Energy Management Policies

"Interpreting software models to control cyberphysical systems facilitate efficient behavior changes at runtime. Energy management within the Microgrid is one such domain in which this approach has been successfully applied. Energy management models are augmented with policies which are able to describe behavior of plant elements based on responses to external events. The current solution requires policies to be hand crafted with the help of domain experts. This limits granular behavioral specification especially in the context of future weather related events. In this paper we present our approach which utilizes demand and production energy forecasts based on cloud services to automate the generation of low level Event-Condition-Action (ECA) policies. These policies describe economically optimized behavior over a short-term forecast horizon. Our approach also addresses the volatile nature of demand-side load forecasting by utilizing kalman filters for profile correction."