Autonomous optimal control for demand side management with resistive domestic hot water heaters using linear optimization

• We consider domestic hot water heaters for autonomous demand side management.
• We propose linear optimization based on one-way communicated incentives.
• A kNN-based data mining algorithm is used to estimate future hot water demand.
• Simulations show savings of up to 12% relative to conventional operation.

Electric domestic hot water heaters are well suited for demand side management, as they possess high nominal power ratings combined with large thermal buffer capacities. In this paper, their potential for demand side management via pseudo cost functions is studied. A fully mixed thermal model of the water heater is used to formulate the optimization problem as a binary integer program, whereas a multi-layer model is used to simulate actual system behavior. The current approach requires only one-way communication of a pseudo cost function, which may depend upon factors such as expected electricity prices, local grid load, and expected renewable electricity production. An optimal heating strategy is then determined on-site based on the expected demand and the pseudo cost function. The expected demand is found via a nearest-neighbor data-mining algorithm using historic data. The optimization algorithm is used in conjunction with long-term simulations assuming different user behavior patterns and optimization strategies. The current approach compares favorably with conventional night tariff-switched operation. Assuming typical user behavior and using real spot market electricity prices as the pseudo cost function leads to cost savings of approximately 12% and energy savings of approximately 4%. Higher energy savings of approximately 12% can be attained by setting the pseudo cost function constant resulting in energy-driven optimization.