Battery storage software for households (2013-2015)
Assoc Prof Peter Vamplew is collaborating with Dr Adam Berry and Tim Moore (CSIRO Energy Technology) and Assoc Prof Doug Creighton (Deakin University) on a project to develop intelligent software to control local battery storage for household solar power systems. An inherent limitation of solar power is that production is highest during daytime hours, which may not coincide with the time when the household is actually consuming power.
Local storage in batteries effectively allows the power to be ‘time-shifted’ so that it is available when it is required. However maximising the benefits of such storage requires sophisticated decision making to balance multiple, potentially conflicting requirements as the household will want to minimise their dependency on grid power, and maximise the financial benefit of their solar system, but also minimise the amount of wear on the batteries.
This project applies Assoc Prof Vamplew’s expertise in multi-objective reinforcement learning to create a software system which learns to balance all of these objectives within a simulation of the system before being applied to the actual solar power system. In particular the team is investigating whether the costs of installing dedicated battery storage can be reduced by making use of battery storage associated with an electric vehicle.
Assoc Prof Peter Vamplew