Steam networks and their utilities should be dimensioned to satisfy day-to-day demand as well peaks. Peaks can be caused by unit start-ups, extreme weather conditions or combinations of high demand from process units. During maintenance operations or unexpected boiler shutdowns, periods of undercapacity may occur when steam demand surpasses the available steam production capacity. This can lead to boiler damage, network inoperability and production loss. Load shedding is a convenient way to prevent damage and reduce the impact of undercapacity. A model is developed with the goal of optimising the flows of steam from producers to consumers through a steam network’s headers, turbines and letdowns. The multi-period nature of the work ensures that nominal and peak demands for steam are properly taken into consideration. The model also investigates the trade-off between installing additional capacity versus production loss. A Mixed Integer Linear Programming (MILP) formulation is used to optimise the steam network operations, investment and load shedding decision- making. A multi-period case study based on anonymised data is carried out on an industrial cluster. Solutions to steam production undercapacity are investigated in the form of load shedding and infrastructure investments.