A Review of Unit Commitment and Economic Dispatch in Power Systems

Abstract

In the current work, we study computational models and methods concerning Unit Commitment and Economic Dispatch (UCED), with a focus on Unit Commitment (UC) in centrally dispatched power systems. The optimization of UCED is an essential requirement for the reliability, cost-effectiveness, environmental sustainability, and energy justice of power systems. A total of 148 strictly peer-reviewed literature articles are examined. A group thereof is used to outline the historical evolution of UCED, computational tools, and related solvers. A subgroup of 62 review articles published since 2015 is used for a targeted evaluation based on 15 criteria. These include Demand Response (DR), flexibility parameters such as distributed resources optimization through Virtual Power Plants, and security-constrained UC parameters such as ramp up/down limits. Five categories of optimization methods, 22 models, and tools, as well as nature-inspired algorithms (NIAs), are appraised. This research adds a new dimension to the existing literature by including both CSOs and market participants’ interests under one view, while describing standardized prevailing UC solving methods and summarizing future requirements. Mainly Linear Programming (LP), Mixed Integer LP (MILP), and Mixed Integer non-LP (MINLP) remain dominant in real-system applications due to their robustness and compatibility with commercial solvers. Since 2019, reviews increasingly emphasize uncertainty management, flexibility, energy storage, policy implications as well as market participants deregulated and price-based environments. Deterministic UC models prevail, but stochastic and robust approaches are expanding. Realistic system applications on a variety of UCED-related issues, especially at (sub)hourly resolution, remain limited and are concentrated in the US, Brazil, and EU regions. Flexibility, energy storage, and uncertainty are common concerns for both CSOs and market participants. The existing modelling framework still needs to be advanced to transfer the application from IEEE systems to real systems while meeting uncertainty concerns. Three multiscale models are depicted for further examination.