The integrated energy system (IES) can make multiple energy sources couple and complement each other, which further improves the operational economy. However, the uncertainty of renewable energy resources and the demand response program’s complexity will aggravate the optimal scheduling difficulty. In this paper, a weight twin delayed deep deterministic policy gradient (WTD3) deep reinforcement learning-based approach is proposed for addressing the IES scheduling problem considering the uncertainties of multiple energy and the integrated demand response. The proposed method can reduce the deviation predicted through weight adjustment of the neural network, and improves the decision-making ability of optimal scheduling. The simulation results confirmed that our proposed WD3-based dispatch strategy outperforms the obtained by other DRL methods and traditional approaches.