面向主动需求侧管理的多智能体优化方法研究

Optimization control integrating active demand side management plays the key role in booming the efficiency of distributed energy systems. This research adopts advanced theory and systematic approaches in multi-agent coordination, distributed optimization, game theory and graph theory. Concentrating on the inherent natures of building group oriented distributed energy systems (namely, high penetration and random volatility of renewable energy, variety of loads and dynamic accessibility, and non-cooperative property of users), this project investigates multi-agent optimization towards active demand side management. In particular, the following three key problems are considered: (1) distributed multi-agent optimization for the systems with various types of loads and controlled devices, (2) distributed model predictive control for the systems integrating distributed energy sources, energy storage devices as well as deferable loads, and (3) multi-agent non-cooperative game for the systems with time-varying strategy spaces. The goal of this project is to develop valid multi-agent optimization algorithms, providing important theoretical basis and solutions for optimization control of building group oriented distributed energy systems integrating active demand side management. This developed work is valuable for booming efficient usages of renewable energy and for reducing energy consumption.

结合主动需求侧管理的分布式能源产储用一体化优化控制是分布式能源系统发挥最大优势的关键技术。本项目采用和借鉴当前先进的多智能体协调控制理论、分布式优化理论、博弈论、图论等系统化方法，围绕建筑群分布式能源系统的三个本质特征（可再生能源的高渗透率和强随机波动性；负荷种类、用能规律、负荷曲线的多样性和动态接入特性；系统各组成单元的非协作性），开展面向主动需求侧管理的多智能体优化方法的研究。重点研究：（1）融合多种负荷类型、多种可调节设备的多智能体分布式协作优化；（2）结合分布式能源装置、储能设备和可延时负荷的分布式模型预测控制；（3）基于时变策略空间的多智能体非合作博弈策略。通过本项目的研究，在多智能体优化方法上取得突破，为建筑群分布式能源系统产储用能一体化优化控制提供理论支撑和解决办法。本项目的开展对于提高可再生能源的综合能效、应对急剧增长的能源需求和能耗黑洞具有重要的意义和价值。

该项目主要围绕面向主动需求侧管理的智能优化调度与资源分配协调问题展开研究，在动态有向网络多智能体分布式优化方法上取得重要突破，包括完成了融合多种可调节设备的多智能体协调优化研究、结合分布式能源装置和可延时负荷的滚动时间窗分布式模型预测控制和协同优化研究、基于时变策略空间的多智能体非合作博弈和多层协作优化研究。给出了能够实现快速、全局收敛、完全分布式的多智能体优化算法，包括针对连续变量的多智能体分布式优化算法、以及针对离散变量、同步通信和异步通信下的多智能体分布优化算法。提出了针对异构多智能体网络、能克服随机波动性的滚动时域分布式优化调度算法，给出了结合分布式能源装置、储能设备和可延时负荷的分布式模型预测控制和优化协调方法。提出了基于时变策略空间的多智能体非合作博弈和多层协作优化设计的系统化方法，获得了包括单向可调节、双向可调节的多层非协作多智能体优化算法。本项目研究结果对于智能电网以及其他分布式网络和系统的优化调度都具有重要的理论意义和应用价值。