Robust Scheduling in a Self-Organizing Hierarchy of Autonomous Virtual Power Plants

In order to operate power management systems (PMSs) in a stable and efficient way, a crucial task is to stipulate the future output of dispatchable power plants in the form of schedules. Solving the scheduling problem is NP-hard in the number of power plants schedules are created for. Future decentralized PMSs, which will be characterized by a vast amount of small dispatchable generators and devices, thus necessitate scalable mechanisms based on the principles of autonomy and self-organization. While such systems can adapt to changing conditions, they also must be able to cope with uncertainties originating from an increasing number of weather-dependent power plants and new consumer types such as electric vehicles. In this paper, we take on this challenge and formalize the scheduling problem on the basis of so-called trust based scenarios in the context of a self-organizing PMS featuring a hierarchical system structure. Our model allows the agents to cope with uncertainties by creating schedules for different developments of the future output and consumption. Based on these insights, we compare two different approaches to create robust power plant schedules. These mainly differ in the way in which system level uncertainties are identified and dealt with. As our evaluation shows, our approach enables the agents to come to a compromise between risk, economic efficiency, and scheduling times.