An Online Dual-Averaging Mixed-Integer Programming Approach for Power Management

Efficient scheduling of power generation, distribution, and usage are crucial for current and future smart grids. Improving the energy management in grid applications will benefit the wholesale electricity market as well as the consumers. This paper addresses the optimal power management problems in electric cooling systems based on appropriately constructed thermal dynamic models and cost profiles. In this venue, the dynamics and logical constraints for the cooling load are first formulated as mixed-integer linear programming models. We subsequently apply an online learning algorithm to adjust the weighting factor for customers’ satisfaction level considering the fluctuating prices and customers’ preference. The proposed approach is expected to save the electricity cost by adequately scheduling the operations of the cooling load without an adverse effect on the entire system. The effectiveness of temperature control and balance settings between electricity cost and customers’ satisfaction level is demonstrated via a simulation using the proposed approach.