In recent years, we have observed rapid changes in the customer demand along with shorter product life cycles. In addition, sustainability concerns about production systems are growing, especially due to energy supply fluctuations in terms of either availability or cost. Among these challenges, energy efficiency is of the utmost importance, and Reconfigurable Manufacturing Systems (RMS), most notably through their scalability feature, could represent a valuable solution: production resources can be reorganized promptly to adapt throughput to external factors, such as uncertain demand or Time-Of-Use prices. Although the aforementioned challenges concern day-to-day management, they should be anticipated at the design stage of the production system, whose behavior might otherwise not meet expectations and hinder the competitiveness of the company. One possibility is to consider the expected performance of such a system from the viewpoint of different productivity and energy-efficiency criteria, through line balancing and future production planning. This can be modeled as a bi-level optimization problem, in which the line balancing of the RMS is the upper level and the configuration planning is the lower level. We consider three criteria, namely the number of workstations, the expected service level and the expected energy cost, taking into account demand uncertainty through scenarios. A three-phase matheuristic is developed and its performances on instances derived from the literature are discussed. The results show that consistent energy cost savings can be achieved, even with very few configurations.