An artificial intelligence-based two-stage framework for dynamic pricing and demand-side management in smart grids

Abstract

Rapid urbanization, the rapid growth of electricity demand, and the widespread adoption of electric vehicles have increased the complexity of managing modern power systems. Conventional fixed pricing mechanisms are increasingly insufficient to ensure efficient energy utilization, peak load mitigation, and system resilience in smart grids. In this context, this study proposes a two-stage artificial intelligence-based framework for real-time demand-side management. In the first stage, long short-term memory models forecast short-term electricity demand across residential, commercial, and industrial sectors. In the second stage, a reinforcement learning-based mechanism dynamically adjusts hourly electricity tariffs by integrating forecasted demand with real-time grid conditions. The framework is validated through a case study. The results obtained from the case study demonstrate the effectiveness of the proposed framework when compared with the conventional Time-of-Use pricing scheme currently implemented in the city. Specifically, the proposed real-time pricing approach reduces the Peak-to-Average Ratio by 6.05% in the residential sector, 9.21% in the commercial sector, and 8.36% in the industrial sector, while achieving peak load reductions of 8.50%, 14.16%, and 10%, respectively. In addition, the proposed pricing strategy increases electricity provider revenue by 24.5%, 10.6%, and 14.2% across the residential, commercial, and industrial sectors. These results demonstrate that the proposed artificial intelligence-driven framework effectively smooths load profiles, mitigates peak demand, and improves economic efficiency, providing a practical pathway toward intelligent and sustainable energy management in smart grids.