An exact solution method and a genetic algorithm-based approach for the unit commitment problem in conventional power generation systems

The unit commitment problem (UCP) is one of the fundamental problems in power systems planning and operations that comprises two decisions: commitment and dispatching of conventional generating units. The objective is to minimize total operating costs -fuel and start-up costs- while satisfying several operational and technical constraints. The UCP is characterized as a highly constrained mixed-integer nonlinear NP-hard problem, which makes it difficult to develop a rigorous optimization method for real-size systems. Hence, we devise an efficient mixed-integer quadratic programming formulation as an exact method with brand-new linear representations for each of the three crucial constraint sets, namely minimum uptime/downtime, start-up and rampup/down constraints. Furthermore, to be able to solve a large-scale UCP and to deal with its complexities, we propose a Genetic Algorithm-based matheuristic approach that can provide optimal/near-optimal solutions quickly, thanks to its unique binary-integer coding scheme and several problem-specific operators. During the genetic evolution, commitment and dispatching schedules are determined by combining genetic operations and the Improved Lambda Iteration Method reinforced by the incorporation of average fuel cost optimization and ramp rate limits. The final dispatching schedule is then determined via a start-up adjustment procedure and an efficient quadratic programming model. The computational experiments show that both proposed exact approach and GA-based matheuristic can provide satisfactorily good schedules even for large-scale conventional power systems in quite a reasonable computation time.

Automated summary

The unit commitment problem involves the decisions of commitment and dispatching of conventional generating units, aiming to minimize total operating costs while satisfying operational and technical constraints. It is a highly constrained mixed-integer nonlinear NP-hard problem, posing challenges for optimization methods. We propose an efficient mixed-integer quadratic programming formulation and a Genetic Algorithm-based matheuristic approach as solutions. Computational experiments demonstrate their effectiveness in providing satisfactory schedules for large-scale power systems.