This paper addresses the real time by product gas scheduling in an integrated iron- and steel-making industry with uncertainty in by-product gas flows from a rolling horizon algorithm. Adaptive time-series models determined from real data performs forecast for each producer and consumer of by-product gases in main units of the steel-making plant. The individual consumptions of the blast furnace and coke oven gases are modelled using the seasonal Holt-Winters method with smoothing constants estimated via genetic algorithm, whereas the individual productions of the blast furnace and coke oven are identified from autoregressive and integrated moving-average. LDG gas production is forecasted using a heuristic method that leverages the operational information. The model's parameters are updated periodically due to the nonlinearities present in the time series. After the forecasting phase, the algorithm performs short-term decisions using a MILP optimization model, that minimizes the imbalance between the random dynamics of by-product fuel generation and consumption and maximizes the energy efficiency. Through computational simulations, we show that the operational stability of gas holders and the electrical energy production increase, whereas the waste of gas in flare stack decreases, when the control horizon of the rolling horizon algorithm is reduced. (C) 2019 Elsevier Ltd. All rights reserved.