Ant Colony Optimisation for Power Plant Maintenance Scheduling


Maintenance of power plants is aimed at extending the life and reducing the risk of sudden breakdown of power generating units. Traditionally, power generating units have been scheduled for maintenance in periods to ensure that the demand of the system is fully met and the reliability of the system is maximized. However, in a deregulated power industry, the pressure of maintaining generating units is also driven by the potential revenue received by participating in the electricity market. Ideally, hydropower generating units are required to operate during periods when electricity prices are high and to be able to be taken offline for maintenance when the price is low. Therefore, determination of the optimum time periods for maintenance of generating units in a power system has become an important task from both a system reliability and an economic point of view. Due to the extremely large number of potential maintenance schedules, a systematic approach is required to ensure that optimal or near-optimal maintenance schedules are obtained within an acceptable timeframe. Metaheustics are high-level algorithmic frameworks that aim to solve combinatorial optimisation problems with a large search space in a reasonable computational run time. Inspired by the foraging behavior of ant colonies, Ant Colony Optimisation (ACO) is a relatively new metaheuristic for combinatorial optimisation. The application of ACO to a number of different applications has provided encouraging results when applied to scheduling, including the job-shop, flow-shop, machine tardiness and resource-constrained project scheduling problems. In this thesis, a formulation is developed that enables ACO to be applied to the generalized power plant maintenance scheduling optimisation (PPMSO) problem. The formulation caters for all constraints generally encountered as part of real-world PPMSO problems, including system demands and reliability levels, precedence rules between maintenance tasks, public holidays and minimum outage durations in the case of shortening of maintenance tasks. As part of the formulation, a new heuristic and a new local search strategy have been developed. The new ACO-PPMSO formulation has been tested extensively on two benchmark PPMSO problems from the literature, including a 21-unit and a 22-unit problem. It was found that the ACOPPMSO formulation resulted in significant improvements in performance for both case studies compared with the results obtained in previous studies. In addition, the new heuristic formulation was found to be useful in finding maintenance schedules that result in more evenly spread reserve capacity and resource allocations. When tested using a modified version of the 21-unit and the 22-unit problems, the new local search strategy specifically designed for duration shortening was found to be effective in searching locally for maintenance schedules that require minimal shortening of outage duration. The ACO-PPMSO formulation was also successfully able to cater for all constraints as specified in both original and the modified versions of the two benchmark case studies. In order to further test the ACO-PPMSO formulation developed, it was first applied to a scaled-down version of the Hydro Tasmania hydropower system (five power stations) and then to the full system (55 generating units). As part of the studies, the ACO-PPMSO formulation was linked with the simulation model used by Hydro Tasmania to assess the impact of various maintenance schedules on the total energy in storage of the system at the end of the planning horizon, the total thermal generation, the total number of days where the reliability level is not met, as well as the total unserved energy throughout the planning horizon. A number of constraints were considered, including the anticipated system demands, a 30% capacity reliability level, the minimum and maximum durations between related maintenance tasks, the precedence constraints and the minimum outage duration of each task in the case of shortening of maintenance tasks. The maintenance schedule was optimised for the maximum end-of-horizon total energy in storage, the minimum thermal generation and the minimum total outage durations shortened and deferred, under 77 different inflow conditions. The optimal maintenance schedule obtained compared favourably with that obtained by Hydro Tasmania over many years based on experience. Specifically, the ACO-PPMSO schedule results in higher end-ofhorizon total energy in storage and satisfies both hard and soft constraints, which overall equates to over $0.5 million dollars of savings when compared to the schedule obtained using the practitioners’ experience and engineering judgment. The ACO-PPMSO algorithm was also shown to be a useful decision-making tool for scheduling maintenance under different circumstances when tested with four scenarios commonly encountered in practical maintenance scheduling problems. In conclusion, the ACO-PPMSO formulation developed, tested and applied as part of this thesis research provides a powerful and flexible means of obtaining optimal or near-optimal maintenance schedules for power plants.