Optimizing the use of flexibility, provided by e.g. batteries and electric vehicles, provides opportunities for various stakeholders. Examples are aggregators acting on energy markets, or energy cooperations willing to maximize their selfconsumption. However, with large numbers of devices that need to be scheduled, the underlying optimization problem becomes difficult. This paper investigates the scalability of a smart grid optimization approach called Profile Steering. This approach uses a hierarchical structure to perform distributed optimization. In this paper, the approach is extended with methods to accept multiple profiles at once and the possibility to prune children with little flexibility. Simulation studies with almost 20,000 households are carried out to evaluate the scalability of Profile Steering. The results show that, with the presented improvements, the required optimization time of Profile Steering scales linearly with the number of children and a speedup factor of 56 is achieved with 1000 households. Furthermore, the approach scales well across multiple computing processes.