Aggregated Water Heater System (AWHS) Optimization for Ancillary Services [report]

Manasseh Obi
2020 unpublished
In this dissertation, I present a two-stage optimization routine that schedules an Aggregated Water Heater System (AWHS) to concurrently provide three utility ancillary services, namely, frequency regulation, frequency response, and peak demand mitigation. Water heaters can be controlled to manage their energy take, the amount of energy a water heater can absorb upon command. The AWHS is a model aggregation of thousands of water heaters, the energy take and power characteristics of which are
more » ... ed on U.S Census household data and usage behavior patterns. The aggregate energy take available in the AWHS may be dispatched en masse for participation in utility ancillary service markets, while accommodating the unique characteristics of the AWHS resource. The optimization routine is performed in two stages. In stage one, the optimization routine anticipates future energy take, power, weather temperature, and market prices based on historical data. Upon convergence, the optimization algorithm shifts to the next five minute time interval and re-iterates the optimization with new projected allocations. This process repeats continuously until a full day's worth of projections are simulated. In stage two, the AWHS considers how dispatches affect the AWHS resource, and re-optimizes to maximize revenue for each ancillary service based on forecasted market prices. The AWHS algorithm relies on forecasts in order to mitigate the effects of overdispatching, which can result in prolonged energy take recovery times. Over-dispatching i can also lead to lost opportunity cost for the AWHS, and this is prevented by setting dispatch and reserve capacity limits for the ancillary services based on system requirements. The optimization routine addresses the challenge of dispatching this dynamic resource by assessing how the system recovery can be managed in a way that adequately positions the AWHS to participate in subsequent rounds of bidding. After every dispatch, the available energy take decreases and a new energy curve, the resource recovery curve, is re-calculated. Further, the energy dispatch constraints are dispatch-dependent and need to be recalculated for every selection of dispatch vectors. We propose a solution for calculating the recovery energy take available after a dispatch. This solution slides the entire 24-hour daily window ahead in five minute increments, causing the optimization solution to constantly change as new future data projections are considered. The primary contribution to knowledge is a solution to the problems caused by Renewable Energy Resources (RER) that uses a novel two-stage method to optimally dispatch the energy take available in the AWHS among the aforementioned three ancillary services in a way that maximizes revenue while minimizing over-dispatching, system recovery time and energy take forecasting errors. With special thanks to my advisor and dissertation committee chair, Dr. Robert Bass, who guided me through this doctoral journey. Without his support, technical insights and constant feedback, this PhD would not have been achievable. To the researchers at Portland State University's Power Lab, Thomas L. Clarke, Emily Barrett, Crystal Eppinger, Jaime Kolln, Rector Blake, and Tylor Slay, thank you for your pursuit of excellence in research. My very special gratitude goes to my research colleague and friend, Kevin Marnell, I consider you a trail blazer. I would also like to thank my dissertation committee members, Dr. Dan Hammerstrom, Dr. John Acken, and Dr. Raul Bayoan Cal, for all your carefully curated feedback. My deepest appreciation goes to my wife and children, for allowing me to work on my PhD while working a full time job, and on the many weeknights and weekends I spent on this research, away from home. I love you all.
doi:10.15760/etd.7301 fatcat:ui3osdc4szahpbria75bdv27fe