Abstract:Power grid is probably the most complex engineering system in the world. The dominated centralized schemes for power system require complicated communication network to collect global operating condition and a powerful central controller to process the huge amount of data. In addition, centralized schemes lack flexibility and cannot easily adapt to intentional/unintentional network topology changes. Since distributed control schemes are more flexible, reliable, and cheaper to implement, they are promising choice for smart grid upgrade. To address the needs of power systems and problems with existing solutions, the author proposes several fully distributed multi-agent system (MAS) based algorithms based on the most recent developments of control and optimization theory, which have been successfully applied to load management, optimal reactive power dispatch (ORPD), cooperation of multiple renewable generators (RGs), and optimal resource allocation. For load management, a fully distributed consensus based global information discovery algorithm is proposed to discover the global net active power, loading level and load connection information. The information discovery algorithm can guarantee convergence for power systems of any size and topology and is robust against certain types of faults. For OPRD, a distributed MAS based Q-learning algorithm is utilized to coordinate the control settings of reactive control devices to minimize the active power loss while satisfying certain operation constraints. The proposed algorithm is model free and can adapt to changes of system operating conditions automatically. For the coordination of multiple RGs in a microgrid, a distributed subgradient-based control scheme is proposed to synchronize the utilization levels of RGs based on local frequency measurement and available generation prediction. The proposed control scheme can be applied to the power systems with different types of generators under both excessive and insufficient renewable generations. For optimal resource allocation, a fully distributed pricing strategy is proposed to maximize the social welfare, while respecting system constraints. The effectiveness of the all proposed algorithms has been demonstrated through simulations. The corresponding work has produced five papers in the IEEE transactions on Power Systems, Smart Grid and Systems, Man, and Cybernetics.

Distributed Reactive Power Control

A Distributed Strategy For Flexible Load As Spinning Reserves In Power System

Decentralized Control for Reactive Power Sharing Using Adaptive Virtual Impedance

Multi-agent Deep Reinforcement Learning Algorithm for Distributed Economic Dispatch in Smart Grid.

Distributed Real-Time Multi-Objective Control of a Virtual Power Plant in DC Distribution Systems

DDPG-based Reactive Power Optimization Strategy for Active Distribution Network

Target-Value-Competition-Based Multi-Agent Deep Reinforcement Learning Algorithm for Distributed Nonconvex Economic Dispatch

Multiagent-Based Reinforcement Learning for Optimal Reactive Power Dispatch.

Distributed Optimal Reactive Power Control Of Power Systems

Deep Reinforcement Learning Based Coordinated Voltage Control in Smart Distribution Network

A Statistical Learning Approach to Reactive Power Control in Distribution Systems

Distributed Control Algorithm for Optimal Reactive Power Control in Power Grids

The Cooperative Control of Distributed Generation Based on Multi-agent System

Distributed Voltage Regulation of Active Distribution System Based on Enhanced Multi-agent Deep Reinforcement Learning

An adaptive active power rolling dispatch strategy for high proportion of renewable energy based on distributed deep reinforcement learning

Multi-Agent Deep Reinforcement Learning for Voltage Control with Coordinated Active and Reactive Power Optimization

Active power correction strategies based on deep reinforcement learning — part II: A distributed solution for adaptability

Distributed Reactive Power Optimization for Energy Internet via Multiagent Deep Reinforcement Learning With Graph Attention Networks

Coordinated Electric Vehicle Active and Reactive Power Control for Active Distribution Networks

Novel distributed algorithms for power systems operations

Real-time Decentralized and Robust Voltage Control in Distribution Networks