An actor critic approach for real time pricing based demand response program in microgrid with optimal allocation of DG

Abstract

The rapid increase in energy demand in distribution systems presents several challenges for utility operators in effectively planning and managing load demand. The optimal coordination of distributed generators (DGs) and demand response (DR) programs can help address these challenges by improving demand scheduling and enhancing overall system performance. In this study, an artificial intelligence-based actor-critic algorithm is applied to solve the demand scheduling problem. The proposed DR framework aims to minimize total system power loss and the cost of grid energy procurement while maintaining the comfort levels of individual customers. The framework is implemented on a standard IEEE 33-bus distribution network integrated with two different sets of DGs. The optimal DG locations are identified using the Harris Hawk Optimization algorithm. Each customer is modelled as an independent reinforcement learning agent located at a specific bus and adjusting its load according to its preferences and operational characteristics. The analysis is carried out under two control scenarios: customer-operated energy management and grid-operated energy management. The results show that the actor-critic algorithm effectively manages load scheduling among customer agents while satisfying load curtailment constraints and minimizing customer discomfort. During intraday operation, the proposed approach achieves an 11.39% reduction in total power loss, a 9.78% reduction in overall power demand, and a 14.08% reduction in peak power demand. Furthermore, the cost of grid power procurement is reduced by 3.5% in Case 1 and 9.84% in Case 2. These results demonstrate that grid-coordinated energy management provides significantly greater economic benefits compared with customer-operated energy management.