Multi-objective pareto ant colony system based algorithm for generator maintenance scheduling
Existing multi-objective Generator Maintenance Scheduling (GMS) models have considered unit commitment problem together with unit maintenance problem based on a periodic maintenance strategy. These models are inefficient because unit commitment does not undergo maintenance and periodic strategy cann...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | eng eng |
| Published: |
2022
|
| Subjects: | |
| Online Access: | https://etd.uum.edu.my/10224/1/s902471_01.pdf https://etd.uum.edu.my/10224/2/s902471_02.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Existing multi-objective Generator Maintenance Scheduling (GMS) models have considered unit commitment problem together with unit maintenance problem based on a periodic maintenance strategy. These models are inefficient because unit commitment does not undergo maintenance and periodic strategy cannot be applied on different types of generators. Present graph models cannot generate schedule for the multi-objective GMS models while existing Pareto Ant Colony System (PACS) algorithms were not able to consider the two problems separately. A multi-objective PACS algorithm based on sequential strategy which considers unit commitment and GMS problem separately is proposed to obtain solution for a proposed GMS model. A graph model is developed to generate the units’ maintenance schedule. The Taguchi and Grey Relational Analysis methods are proposed to tune the PACS’s parameters. The IEEE RTS 26, 32 and 36-unit dataset systems were used in the performance evaluation of the PACS algorithm. The performance of PACS algorithm was compared against four benchmark multi-objective algorithms including the Nondominated Sorting Genetic, Strength Pareto Evolutionary, Simulated Annealing, and Particle Swarm Optimization using the metrics grey relational grade (GRG), coverage, distance to Pareto front, Pareto spread, and number of non-dominated solutions. Friedman test was performed to determine the significance of the results. The multiobjective GMS model is superior than the benchmark model in producing the GMS schedule in terms of reliability, and violation objective functions with an average improvement between 2.68% and 92.44%. Friedman test using GRG metric shows significant better performance (p-values<0.05) for PACS algorithm compared to benchmark algorithms. The proposed models and algorithm can be used to solve the multi-objective GMS problem while the new parameters’ values can be used to obtain optimal or near optimal maintenance scheduling of generators. The proposed models and algorithm can be applied on different types of generating units to minimize the interruptions of energy and extend their lifespan. |
|---|