Power-Peak-Curbing Switching Schedule for a Multiagent System

抄録

This paper introduces a method to determine the switching schedule in a multiagent system with a constraint on the peak power. The proposed system consists of a single shared power resource and multiple agents. Each agent has a local controller and a local switch. Some agents are connected with signal lines so that they can exchange information about their state. The switching schedules for agent input are preliminarily determined by a model based on agent behavior and the power constraint. When running the system, each agent selects one of the prescheduled switching patterns to determine their state and future target patterns. The concept of ON cards is introduced to avoid excess demands from the agents. The ON cards are agent-specific and state the privileges possessed by that agent to execute its primary prescheduled switching patterns. An exchange of these cards is negotiated among the network of agents connected by signal lines; this occurs once for each control cycle. We performed a simulation for a DC motor defined in a linear time-invariant (LTI) system with unique velocity reference schedules; the results show the effectiveness of the proposed multiagent system. Another simulation demonstrates that even during operation, the proposed system allows the addition or removal of participants. In conclusion, this study shows that power peaks in a closed multiagent system can be curbed by applying a negotiated predetermined switching schedule.