Multi-objective optimization method of blast furnace operation performance indicators based on data and knowledge

During blast furnace (BF) ironmaking, hot metal quality and energy consumption directly affect industrial economic benefits. Field engineers usually adjust the production state based on their experience, which makes control optimization uncertain. A data and knowledge-based optimization framework for BF operation performance indicators is proposed in this paper. Firstly, design the Attention Long Short Term Memory (Attention-LSTM) neural network to predict the Silicon content and coke ratio, which is used as fitness value function of the multi-objective optimization model. To initialize the population, knowledge based non-dominated sorting genetic algorithm (KB-NSGA) uses a knowledge base of historical optimal solutions. Finally, an optimization solution set that meets the actual production requirements is obtained using the TOPSIS evaluation method. The KB-NSGA successfully achieves the goal compared with other genetic algorithms. The effectiveness of the proposed method is verified by long-term running experiments.