Burnable off-gases generated in engineering process plants are regularly utilised as energy sources. A common use is for steam production, where excess steam is allocated to power generation turbines. Fluctuating off-gas productions may, however, result in power generation losses from turbine trips, due to insufficient steam. Numerous power co-generation investment models exist, which are typically based on cost minimisations or meeting energy demands. These models do not, however, incorporate plant specific historic steam profiles and typically use average-based patterns for decision making. This paper presents a novel stochastic mixed integer linear programming model that utilises historic steam profiles to determine optimal turbine investments in terms of the net present value. A further advantage is the ability to investigate the investment and procurement of a, typically very expensive, supplementary energy resource to assist during low off-gas flow periods. The proposed model is solved to optimise over 10 years for an engineering factory seeking to invest into an energy recovery plant. Optimal results demonstrate how natural gas in a fluctuating off-gas environment can increase power generation profits and should be invested in, together with a 30 MW turbine. Furthermore, an average based approach yields sub-optimal investments and overestimates the net present value beyond 22%. (C) 2018 Elsevier Ltd. All rights reserved.