In the scheduling of integrated energy systems, the dynamic supply and demand matching and multi-energy coupling problems in demand response scenarios need to be solved urgently. This study proposes a collaborative model of graph attention network (GAT) and SAC algorithm. First, GAT is used to extract energy node characteristics and construct energy network topological relationships, and the multi-head attention mechanism is used to quantify the coupling weights of multi-energy flows such as electricity, heat, and gas; then the SAC algorithm is used to construct a dynamic decision-making framework, and the entropy regularization strategy is used to achieve multi-objective collaborative optimization of supply and demand balance, cost optimization, and renewable energy consumption. A composite reward function is designed that includes energy cost, carbon emission penalty, and load fluctuation rate. The study confirms that the GAT-SAC model can effectively capture the spatiotemporal characteristics of the energy network, providing a new method for solving dynamic energy scheduling problems.