Distributed Optimization for Electric Power Systems, Session 2: Performance Improvement and Path Forward (Slides)

Applications of Machine Learning to Improve Distributed Optimization Algorithms for Power Systems, Computing solutions for Distributed Optimization, Evaluating the Performance of Distributed Optimal Power Flow Algorithms with Nonideal Communications, Outlook for Distributed Optimization Algorithms. The rapid growth of distributed energy resources (e.g., solar PV, batteries, electric vehicle charging, etc.) motivate the transition from centralized to distributed computations to cooperatively control these devices such that network efficiency and reliability are maximized. Distributed optimization algorithms provide a number of potential advantages over centralized computations, including scalability via parallel computations, robustness to failure of individual computing agents, and data privacy. Accordingly, both academics and industry are looking towards distributed optimization algorithms to complement existing centralized computations through various use cases, such as volt/var control, solving optimal power flow, and managing retail electricity markets. To realize the potential advantages of distributed optimization, researchers have proposed various distributed optimization algorithms for these use cases, each with different advantages and disadvantages. This tutorial reviews various use cases for distributed optimization, presents illustrative examples of several solution algorithms, and discusses current limitations and future research needs.

Topics covered:
- Use Cases and Research Needs for Distributed Optimization in Power Systems
- Classifications of Distributed Optimization Algorithms
- Applications of Distributed Optimization to Volt/Var Control
Proximal Atomic Coordination Algorithms for Distributed Optimization

Learning Objectives: By the end of the tutorial session, the audience will understand:
- Use cases and motivations for distributed optimization in electric power systems.
- Recent progress in algorithms for solving distributed optimization problems, including both their theoretical underpinnings and practical implementation details.
-Research needs for state-of-the-art distributed optimization algorithms.