Dependability in multi-agent systems for smart grid applications

Abstract

Over the past decades, the spread of information and communication technologies has affected many different domains and industries and is known under equally many keywords. This ongoing digital transformation has also been affecting the electric grid and led to the notation of Smart Grids. With the shift towards decentralization caused by adopting distributed energy resources and increased power consumption partly driven by the popularity of electric vehicles, Smart Grid applications are becoming economically viable and even necessary to cope with increasing energy demands in existing power distribution networks. This thesis presents and discusses a systems engineering life cycle that incorporates dependability attributes in developing multi-agent systems for Smart Grid applications. It builds upon existing methodologies from Smart Grid, multi-agent system, and ontology development. The individual agents and ontologies are developed and refined during eleven activities structured into five main phases of the systems engineering life cycle. Each agent is equipped with a knowledge base, which allows the agent to consider the dependability attributes stored therein in its decision-making process. Switches and circuit breakers in urban power distribution networks allow to dynamically change the network configuration, e.g., to isolate faults. However, the switch states (open or closed) also affect the transformer and line losses caused by supplying the connected loads. This insight gives rise to the switching optimization problem, which serves as an ongoing use case throughout this thesis. It aims at reducing losses by dynamically restructuring the distribution network as loads shift between different areas, e.g., from residential homes to office buildings. An existing co-simulation framework is extended to support agents and ontologies that resulted from applying the systems engineering life cycle to the switching optimization problem. The evaluation results provide an indication of energy savings achievable by a distributed switching optimization approach. Furthermore, they show that incorporating dependability in the decision-making processes of agents can improve the performance of multi-agent systems in the Smart Grid domain.