Multiscale Adaptive Agent-Based Management of Storage-Enabled Photovoltaic Facilities

We propose a multiscale algorithm for autonomous agents to adaptively manage the operation of storage-enabled photovoltaic (PV) facilities based upon sequential decisions in a partially observable environment. The stochastic environment is learned and modeled by an approach called an e-Machine, which operates on a set of a priori determined temporal scales, to give the agent an additional degree of freedom when optimizing its control decisions. We compare the performance of the proposed scheme with those of (1) control decisions based on a heuristic environment model rather than systematic learning, and (2) decisions made on a pre-determined scale. We argue that the systematic environment learning on multiple temporal scales makes the agent highly adaptable and, as a result, able to demonstrate a superior capability in managing the PV-storage facility according to a predetermined objective. The particular application focused upon in the paper is in managing distributed PV facilities as potential replacement of conventional peaking power plants.