A Bayesian Networks Structure Learning And Reasoning Based LDG Scheduling In Steel Industry

It is very crucial for the byproduct gas system in steel industry to perform an accurate and effective scheduling, which enables to reasonably utilize the energy resources, effectively reduce the production cost of enterprises, improve production efficiency and reduce environmental pollution. In order to guarantee the normal operation of the energy system, the actual production mode is currently to predict the gas flow real time series, and then to rely on the personal experience of the energy scheduling operators to realize the balance scheduling for the byproduct system. Although, to some extent it provides some useful guidance for the energy scheduling, it is still lack of scientific and reasonable operations so as to implement few measures on automation and intelligence in steel industry. Therefore, the study on gas generation and consumption and its recovery and reuse fully can provide scientific guidance for efficient utilization of the byproduct energy and improve the level to save energy in steel enterprise.In this paper, a novel data-driven based dynamic scheduling method is proposed for real time gas scheduling a steel plant, in which a probability relationship used the Bayesian network modeling is presented to determine the adjustable gas users that impact on the changes of the gas tanks level and provide their scheduling amounts by adopting the maximal posterior probability of gas consumption status on-line. In the modeling stage, a constrained genetic algorithm is designed to optimize the topologic sequence of the network input nodes. For the practical applicability, the proposed scheduling solution is verified by a recurrent neural network.To verify the effectiveness of the proposed data-driven based method, the real gas flow coming from a steel plant in China are employed, considering the actual industrial situations that may appear, such as gas tank level shortage, gas tank level surplus and changing the gas tank boundaries. The experimental results indicate that the proposed method can provide real time and scientific gas scheduling solution for the energy system of steel industry.