Due to the stochastic nature of equipment failures, the accurate assessment of power system reliability is a complex task. Consequently, the optimal selection of new network infrastructure to improve reliability is even harder. In this paper, an optimization via simulation approach is proposed to find the optimal set of network assets to improve system reliability. Particularly, an Industrial Strength COMPASS algorithm is implemented to find the optimal set of new transmission lines that maximizes system reliability subject to a budget constraint. This algorithm iteratively proposes, in a first stage, a set of new transmission lines that are then tested, in a second stage, via simulation of the system operation, including impact of various network failures. In the second stage, the sequence day-ahead unit commitment plus real-time operation is modeled along with a sequential Monte Carlo simulation to determine highly detailed system operation under network outages and thus calculate the associated expected energy not supplied.