This paper presents a digital twin-enabled demand-side management framework for single-family houses. The aim is to develop an integrated platform that can systematically coordinate heterogeneous residential energy resources and support real-time optimization of home energy management by responding to dynamic electricity pricing. The digital twin environment (DTE) includes a multi-zone building thermal model, photovoltaic (PV) generation, battery storage system (BSS), electric-vehicle (EV) model, and schedulable electric appliances. The Soft Actor-Critic (SAC) algorithm is adopted as the embedded reinforcement learning optimizer in the DTE to determine optimal control actions for these resources by minimizing cost and comfort deviation. The framework is linked to a Modelica-based high-fidelity building emulator for data generation and subsequent validation. Compared with a rule-based strategy, the results demonstrate that the proposed RL-digital twin framework can effectively manage flexible resources while saving the electricity cost. In addition, the RL-digital twin efficiently satisfies occupants’ needs on thermal comfort and domestic hot water, which are significantly compromised under the rule-based strategy. The obtained results demonstrate the RL-digital twin framework potential for real-world deployment in residential demand-side management.