下一代智能电网是基于巨量”主动能量体(AEA)”互联互动的复杂超网络。AEA是具有自主智能的源/荷/储设备，也是理性的市场参与者；可主动感知态势、参与交易、执行电能的发/用/储动作。在区块链(分布式存储、智能合约等)关键技术的赋能作用下，众多AEA自发结成去中心化的虚拟电力社群(VPC)并开展P2P交易。电力市场的去中心化发展趋势，要求建立与之相适应的组织和运行模式。本项目依次研究VPC成员、共识、互动：1)建立描述AEA主动能量行为的约束最优化模型和深度学习预测模型；2)设计P2P电力交易规则，包括订单匹配算法和订单校核-裁剪算法；3)给出巨量AEA的博弈策略，分析由此涌现的VPC复杂动力学特性并揭示其规律，为互动规则和集中监管方案的制定提供决策参考。设计关键算法的GPU并行加速方案，解决巨量AEA导致的数值计算难题。力图在下一代智能电网的行为建模、规则设计和复杂性分析上取得开拓性成果。

Next generation smart grid is a complex hypernetwork based on the interconnection and interaction among huge numbers of “active energy agents (AEAs)”. Not only are AEAs various source, load, energy storage facilities with autonomous intelligence, but also are rational players of electricity market. AEAs are capable of active perception, active transaction, and active implementation of producing, consuming, and charging/discharging electricity. Enabled by key technologies of blockchain (distributed data storage, smart contract, etc.), numerous AEAs spontaneously form a decentralized organization, virtual power community (VPC), and perform the P2P transaction of electricity. The decentralization trend of electricity market requires novel modes of organization and operation. As a result, this program will investigate VPC in terms of members, consensus, and interactions: 1) Constrained optimization models and deep learning prediction models will be established to describe the active energy behaviors of AEAs; 2) P2P transaction rules will be designed, which include the algorithms of order matchmaking and check-tailoring; 3) game strategies of huge numbers of AEAs will be derived, the emergent dynamics of VPC system will be analyzed, and potential laws will be revealed, so as to aid the decision-making of interaction rules and centralized supervision. GPUs will be adopted for the parallel acceleration of key algorithms to deal with the numeric computational problems due to huge numbers of AEAs. Innovative results are expected on the aspects of behavior modeling, rule design, and complexity analysis of next generation smart grid.