In order to provide a reliable service and supply the demand most of the time, all generators in a power grid should be subjected to an effective maintenance plan. The smarter the maintenance performed could result in a better performance of the system. However, a challenge is to minimise maintenance costs that do not compromise the benefits. Considering these facts, this study presents a reliability-based smart-maintenance approach of generators to compute the net-maximum economic benefit. The approach is derived from Kijima model type I to characterise the impact of maintenance over the component's virtual age, and Markov chains to model the component's lifetime. To achieve a more realistic model, generators' failure and repair rates are considered time-dependent variables. Then, the optimum preventive maintenance schedule is obtained by using an advanced algorithm named accelerated quantum particle swarm optimisation in combination with sequential Monte Carlo simulation. The effectiveness of the approach is investigated through a case study with four different scenarios: (i) no preventive maintenance plan, (ii) yearly periodic preventive maintenance, (iii) reliability-centred maintenance and (iv) smart maintenance. The results suggest that the approach is convenient for power system generators and delivers a significant knowledge contribution in the area of maintenance.