Real-Time Incentive Design Under Unknown Demand Response Curves via a Proportional–Integral Control Frameworks

Abstract

Implementing voluntary Demand Response (DR) in real-time or close to real-time is technically feasible because of the deployment of smart grid infrastructures such as smart meters, remote control units, and mobile communication devices. However, most cases require estimating/predicting load demand curves and typically operate from an electricity market clearing perspective. In this work, two frameworks of voluntary incentive-based DR to enhance the resilience and reliability of the system are designed, including a single Proportional-Integral controller based real-time DR framework and a multiple PI controllers cooperation based DR framework, which can be operated in real-time without the requirement of estimating or predicting the load demand curves. Under the proposed frameworks, the overall system can converge to the optimal tracking point to improve resilience/reliability via the cooperation of the different DR devices. Additionally, the corresponding incentive payments can be precisely determined. In particular, both frameworks are proven to be stable. In addition, deep reinforcement learning is adopted to enhance the performance of the proposed frameworks. Because of the generality of the proposed frameworks and the simplicity of the PI controllers, our proposed DR frameworks are promising for practical implementation. Finally, numerical studies verify the effectiveness of the proposed frameworks.