Adaptive smart home energy management with forecast-enhanced user centric proximal policy optimization

Efficient energy optimisation in smart homes is increasingly important due to dynamic electricity pricing, fluctuating renewable generation, and rising demand for flexible energy control. However, many existing Smart Home Energy Management Systems (SHEMS) rely on static scheduling or rule-based logic, offering limited adaptability and user involvement. This paper presents a novel user-centric optimisation framework based on Proximal Policy Optimisation (PPO) to intelligently manage appliance scheduling and energy storage operations in residential environments. The model incorporates short-term forecasts of electricity demand, renewable energy (RE) generation, and time-varying prices as auxiliary inputs to enhance decision-making under uncertainty. A flexible reward structure allows users to adjust the trade-off between cost savings and comfort in real time. Simulation results show that the proposed system achieves average daily cost savings of 23% while maintaining comfort satisfaction above 60% across cost-priority, comfort-priority, and balanced modes. These findings confirm the effectiveness of combining deep reinforcement learning with forecast-aware and adaptive decision-making for intelligent, user-driven smart home energy management.