This paper presents a novel design methodology for power systems. A superstructure-based modelling technique has been applied to identify the cost-effective match between available power generation equipment and energy consumers. Multi-period design is conducted to ensure accurate equipment performance estimation. The proposed MILP (mixed integer linear programming) optimization model is able to reflect the machinery performance variations affected by the environmental conditions, and to estimate the deteriorated machinery performance due to part-load operation. To maintain the linear nature of the overall mathematical model, machinery performance is linearized with reasonable accuracy. Moreover, the multi-period methodology is able to conduct synthesis of power systems for processes with non-constant energy demands. Case studies are illustrated to demonstrate the importance of considering the effect of ambient conditions and part-load operation on machinery performance. With the ability to satisfy varying energy demands, and more accurate description of the machinery performance, the optimal design yielded from the improved model would exhibit better flexibility and reliability.