随着全球社会日益关注可持续能源的实践过渡，建筑综合能源系统优化对提高能源消费的低碳性及经济性具有重要意义。为此对建筑能源运营商的调度策略和定价策略展开了研究。首先，同时考虑供给侧和需求侧的信息交互特性，以供给侧为领导者、需求侧为追随者建立一种基于主从博弈框架的建筑综合能源系统双侧优化模型。其次，针对主从博弈框架双侧的多次信息交互提出了一种基于自适应动作探索机制的深度确定性策略梯度算法以有效求解所建模型。自适应动作探索机制根据累计奖励的方差与critic网络平均损失值构建自适应探索系数改进算法的动作选择策略，保证算法的精度和稳定性。最后，通过实例验证了所提算法的有效性。实验结果表明，与其他深度强化学习算法相比，所提算法可以提高算法的收敛精度、稳定性和能源运营商的总收益，从而辅助能源供应商做出更好的决策。

As the global society is increasingly concerned about the transition to sustainable energy practices, building integrated energy systems optimization is significant in improving low-carbon and economic energy consumption. Therefore,research is conducted on the scheduling and pricing strategies of building energy operators. Firstly, the information interaction characteristics of both the supply side and the demand side are considered. A two-side optimization model of the building integrated energy system based on the Stackelberg game framework is established with the supply side as the leader and the demand side as the follower. Secondly, a deep deterministic strategy gradient algorithm is proposed based on the adaptive action exploration mechanism to solve the constructed model efficiently given the multiple information interactions between the two sides of the Stackelberg game framework. The adaptive action exploration mechanism constructs the action selection strategy of the adaptive exploration coefficient improvement algorithm based on the variance of the cumulative rewards and the average loss value of the critic network, ensuring the algorithm's accuracy and stability. Finally, the effectiveness of the proposed algorithm is verified by examples. The experimental results show that compared with other deep reinforcement learning algorithms, the proposed algorithm can improve the convergence accuracy and stability of the algorithm, as well as the total revenue of the energy operator, thus assisting the energy supply side in making better decisions.