A scheduling method for blast furnace gas system in steel industry based on a modified generative adversarial network

Abstract

Blast furnace gas (BFG) is a crucial secondary energy for the manufacturing process of the steel industry and its efficient utilization is the cornerstone of energy savings and carbon emissions reduction. However, current energy scheduling operation mostly relies on human experience, which is lacking of effectiveness and economy for unknown scenarios (working conditions). In this study, a scheduling method based on a modified generative adversarial network (GAN) is proposed. The state characteristics, which are represented by equal-length information granules, are extracted from the time series of the influenced variables that consisting of a scheduling scenario, and the experience-based scheduling rules are constructed to identify the appropriate scheduling moment. Then, a novel generation framework is built, in which the generator is established by employing multiple GANs to avoid the negative influence of different distribution features on a single discriminator. Finally, a scheduling solution for the corresponding moment is calculated by matching the generated scenarios with the actual one. A series of experiments by using the data coming from a steel enterprise are carried out for verification, and the results show that the scenarios generated via the proposed method are consistent with the actual ones, and the solutions are effective when facing with such kinds of scheduling problems under unknown scenarios.