A two-stage method for predicting and scheduling energy in an oxygen/nitrogen system of the steel industry

As essential energy resources in steel industry, oxygen and nitrogen are massively utilized in many production procedures, such as iron-making by blast furnaces, steel-making by converters, etc. The trends of the energy generation/consumption flows along with the related scheduling works play a pivotal role on the energy management of steel enterprises. Aiming at an oxygen/nitrogen system of a steel plant in China, a two-stage predictive scheduling method is proposed in this study for resolving the optimal energy decision-making problem. Given the high cost of time consuming on the load change of air separation units (ASU) of the oxygen/nitrogen system, a Granular-Computing (GrC)-based prediction model is firstly established at the stage of prediction, which extends the predicting length to even a day based on data segment rather than generic point-wise mode. At the stage of optimal scheduling, a mixed-integer program model is constructed on the basis of constraining the number of adjustable energy units, which considers not only the actual capacity of the energy devices, but the practical energy conversion procedure as well. The experiments employing the real data coming from this plant also involve two stages, the long-term prediction and the energy scheduling, and the experimental results exhibit both satisfactory accuracy and practicability. Furthermore, the results of system application also indicate the effectiveness of the proposed method.