With the development of smart grid, residents have the opportunity to schedule their household appliances （HA） for the purpose of reducing electricity expenses and alleviating the pressure of the smart grid. In this...With the development of smart grid, residents have the opportunity to schedule their household appliances （HA） for the purpose of reducing electricity expenses and alleviating the pressure of the smart grid. In this paper, we introduce the structure of home energy management system （EMS） and then propose a power optimization strategy based on household load model and electric vehicle （EV） model for home power usage. In this strategy, the electric vehicles are charged when the price is low, and otherwise, are discharged. By adopting this combined system model under the time-of-use electricity price （TOUP）, the proposed scheduling strategy would effectively minimize the electricity cost and reduce the pressure of the smart grid at the same time. Finally, simulation experiments are carried out to show the feasibility of the proposed strategy. The results show that crossover genetic particle swarm optimization algorithm has better convergence properties than traditional particle swarm algorithm and better adaptability than genetic algorithm.展开更多
车网互动(vehicle to grid,V2G)技术利用调度模型生成的决策调度电动汽车(electric vehicle,EV)有序参与电网管理,可实现高效削峰填谷,采用联邦学习方式可以在充电站不愿共享原始数据的条件下完成调度模型训练,因此选定符合多方利益的...车网互动(vehicle to grid,V2G)技术利用调度模型生成的决策调度电动汽车(electric vehicle,EV)有序参与电网管理,可实现高效削峰填谷,采用联邦学习方式可以在充电站不愿共享原始数据的条件下完成调度模型训练,因此选定符合多方利益的训练标签和保证模型参数聚合结果的正确性对于V2G调度决策至关重要。为此,提出一种面向V2G调度的可信联邦学习方法。首先,构建V2G实时调度模型可信联邦学习架构,其包括标签生成模块、可验证联邦学习模块和实时调度模块3个部分;然后,综合考虑EV用户、运营商及电网侧负荷波动,提出一个计及电网多方主体利益的实时调度标签数据生成模型,并设计调度模型标签的动态更新方法;其次,提出模型参数聚合的安全存证与验证方法,确保联邦学习模型参数聚合的正确性;最后,对3种充电时段类型EV占主导生成的标签数据和所提出验证方法的时间开销、存储开销和Gas开销进行分析。算例结果表明,所提出的标签模型展示了EV用户、运营商以及电网侧负荷波动的最优值特征,构建聚合树的时间开销达到毫秒级,相比于传统验证方式,聚合验证智能合约的Gas开销显著降低。因此,所提出的可信联邦学习方法与电网中多方主体利益一致,并具有较好的性能。展开更多
基金supported by the National Basic Research Program of China(973 Program)under Grant No.2012CB215202the National Natural Science Foundation of China under Grant No.51205046 and No.61450010
文摘With the development of smart grid, residents have the opportunity to schedule their household appliances （HA） for the purpose of reducing electricity expenses and alleviating the pressure of the smart grid. In this paper, we introduce the structure of home energy management system （EMS） and then propose a power optimization strategy based on household load model and electric vehicle （EV） model for home power usage. In this strategy, the electric vehicles are charged when the price is low, and otherwise, are discharged. By adopting this combined system model under the time-of-use electricity price （TOUP）, the proposed scheduling strategy would effectively minimize the electricity cost and reduce the pressure of the smart grid at the same time. Finally, simulation experiments are carried out to show the feasibility of the proposed strategy. The results show that crossover genetic particle swarm optimization algorithm has better convergence properties than traditional particle swarm algorithm and better adaptability than genetic algorithm.
文摘车网互动(vehicle to grid,V2G)技术利用调度模型生成的决策调度电动汽车(electric vehicle,EV)有序参与电网管理,可实现高效削峰填谷,采用联邦学习方式可以在充电站不愿共享原始数据的条件下完成调度模型训练,因此选定符合多方利益的训练标签和保证模型参数聚合结果的正确性对于V2G调度决策至关重要。为此,提出一种面向V2G调度的可信联邦学习方法。首先,构建V2G实时调度模型可信联邦学习架构,其包括标签生成模块、可验证联邦学习模块和实时调度模块3个部分;然后,综合考虑EV用户、运营商及电网侧负荷波动,提出一个计及电网多方主体利益的实时调度标签数据生成模型,并设计调度模型标签的动态更新方法;其次,提出模型参数聚合的安全存证与验证方法,确保联邦学习模型参数聚合的正确性;最后,对3种充电时段类型EV占主导生成的标签数据和所提出验证方法的时间开销、存储开销和Gas开销进行分析。算例结果表明,所提出的标签模型展示了EV用户、运营商以及电网侧负荷波动的最优值特征,构建聚合树的时间开销达到毫秒级,相比于传统验证方式,聚合验证智能合约的Gas开销显著降低。因此,所提出的可信联邦学习方法与电网中多方主体利益一致,并具有较好的性能。
