Combining agent-based residential demand modeling with design optimization for integrated energy systems planning and operation. (1st April 2020)

• Record Type: Journal Article
• Authors: Zhang, Zhihui; Jing, Rui; Lin, Jian; Wang, Xiaonan; van Dam, Koen H.; Wang, Meng; Meng, Chao; Xie, Shan; Zhao, Yingru
• Is Part Of: Applied energy. Volume 263(2020)
• Publication Date: 2020-04-01
• Subjects: Energy demand prediction -- Agent-based modeling -- Integrated energy systems -- System optimization -- Stochastic programming; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2020.114623
• Languages: English
• ISSNs: 0306-2619

Abstract

Graphical abstract: Highlights: Propose a holistic approach combining demand prediction with system optimization. Community energy demands are simulated by agent-based modeling approach. K-means clustering is used to generate representative stochastic demand scenarios. A stochastic programming approach is applied for system optimal design and dispatch. More than 36% cost savings can be achieved after optimization for case study. Abstract: When optimizing the design of integrated energy systems, their energy demands are usually assumed as given input. However, the demand of energy systems may not always be available information or it can be uncertain, particularly for projects at the planning stage. Moreover, the overall energy demand is dependent on the behavior of many individuals, which might change as a result of technical, economic or policy interventions. Therefore, this study proposes a holistic approach to combine the demand modeling with design and dispatch optimization for integrated energy systems. The approach can be decomposed into two stages. Firstly, at demand simulation stage, Agent Based Modeling is adopted to generate uncertain demand scenarios for a case study community and the energy-consuming activities for various types of households living in different kinds of apartments are simulated based on probability models and demographic information. Dozens of demand scenarios are obtained via iterative simulation, and k-means clustering approach is further applied to generate representative stochastic scenarios. Secondly, at system optimization stage, the uncertain demand scenarios are used as input of an established stochastic Mixed Integer Linear Programming model, by which the system design and dispatch strategy can be optimized simultaneously. The case study shows that the obtained optimal solutions can save 36% of annual total cost compared to the business-as-usual baseline scenario.