Operational optimisation of smart grids

© 2011 Dr. Kent Charles Barton SteerThe pressures of population growth, environmental damage, resource depletion and technological development are evolving our energy networks into smart grids. Our world is becoming more instrumented and more monitored; we are capturing more information than ever before. As the quest for renewable energy drives an increase in distributed generation, and with the ubiquity of communication infrastructure, the number and the diversity of choices available to network operators is exploding. Utilising the available information to make good decisions about how to operate energy networks is, perhaps, the activity which most warrants the ‘smart’ label. Smart grid operational optimisation problems are characterised by abundant data, far-reaching control and persuasively large potential benefits. However, due to time constraints, these problems tend to exceed the tractability limits of exact solution methods. In this work we investigate both analytically and experimentally the trade-off between solution timeliness and solution accuracy in the context of a receding horizon controller with a metaheuristic subroutine. We demonstrate that for a given system, forecast horizon and metaheuristic algorithm, there exists an optimal control period which balances the theoretical best solution with the practical limitations imposed by the available time in the online setting. Following our findings in this area, we proceed to develop a smart grid controller which encapsulates the knowledge obtained through offline optimisation in an online-suitable decision tree ensemble. This controller is then applied to operational optimisation in a district heating network—an area which has not previously received due attention in the smart grid literature. We finish with a study of plug-in hybrid electric vehicle charge scheduling strategies, including two new strategies introduced here. As the number of such vehicles becomes large, their combined impact on the grid will become significant. It is shown that a simple charge-start delay can improve the overall grid load profile and avoid the need for additional generation capacity