Multi-agent control of community and utility using Lagrangian relaxation based dual decomposition

• Multi-agent control based decision is investigated for ac power flow constraints.
• Autonomous decision making with limited information exchange.
• Iterative decision making based on Lagrangian relaxation and dual decomposition.
• Subgradient method-based iterative decision making.
• A novel upper bound and lower bound switching iterative decision making.

In multi-agent based demand response program, communities and a utility make decisions independently and they interact with each other with limited information sharing. This paper presents the design of multi-agent based demand response program while considering ac network constraints. This project develops two types of information sharing and iterative decision making procedures for the utility and communities to reach Nash equilibrium. The distributed algorithms of decision making are based on Lagrangian relaxation, duality, and the concept of upper and lower bounds. The first algorithm is subgradient iteration based distributed decision making algorithm and the second algorithm is based on lower bound and upper bound switching. The two algorithms require different information flow between the utility and communities. With the adoption of distributed algorithms, the utility solves optimal power flow at each iteration while considering ac network constraints, and the communities also conduct optimization. Through information sharing, the utility and the communities update their decisions until convergence is reached. The decision making algorithms are tested against three test cases: a distribution network IEEE 399 system, two meshed networks (IEEE 30-bus system and IEEE 300-bus system). Fast convergence is observed in all three cases, which indicates the feasibility of the demand response design.