Distributed Optimization of Power System with Multiagent Technologies

Abstract

A prototype of the distributed control system was developed, which executes power loss optimization task for a specific part of the power grid. The control system calculates optimal states of reactive power compensators and voltage regulators, without using of a common data collection point. The distributed power loss optimization is achieved by dividing of the specific grid part into energy clusters (groups of energy facilities) according to geographical, economic or other principals. Every energy cluster is represented by an agent which receives measurements within its own area (energy cluster) and is able to request measurements from adjoin to the energy cluster substations of other clusters. Based on these measurements, the agent creates equivalents of the external grid and implements local (inside one energy cluster) optimization. Every agent executes this optimization task and then a negotiation process begins. During the negotiation process an optimal for all energy clusters mode is defined.The distributed system has a three-level architecture:–The upper level. The level contains energy cluster agents, which interact to each other and determine the optimal mode for every energy cluster and for the whole part of a grid.–The middle level. The level performs collection of equipment sates within the energy cluster and sends commands to equipment, which were defined during high level optimization process.–The lower level. The level contains local regulators of substations to keep determined power mode during diurnal cycle.The upper level optimizes the operation mode of the energy cluster group by agent communication. This process is a trading process among agents, in which the reactive power compensators states and voltages at the boundary substation of energy clusters are iteratively changed to determine a mode with minimal active power losses.The middle level represents an interaction of power equipment agents. The agents collect information about the current state and send it to upper level agent. When the state of the energy cluster has been determined during optimization process, voltage setpoints for every substation are sent back. The power equipment agents receive voltage setpoint and determine states of reactive power compensators and voltage regulators to implement the setpoint.The lower level of the system is aimed to maintain determined operation mode during diurnal cycle. It contains substation regulators, which controls installed reactive power compensators and voltage regulation equipment (tap-changers).The developed system has the following advantages:–Self-configuration of the system: the applied approach allows automatically detecting topology changes and rebuilding the operation of algorithms without additional human intervention.–Easy horizontal scaling: if a new energy cluster is added to the system, all its neighbors will automatically detect this agent and take it into account for the next optimization.–The system does not require a central point of data collection and storage. Thus, the communication channel requirements are reduced. Communication by GPS and GLONASS can be used for agent communication.