Predator-Prey Genetic Algorithm And Its Application In Neural Network Modeling In Blast Furnace

The blast furnace smelting process is a very highly complex process, whose operation mechanism has the feature of nonlinearity, large delay, great noise and distribution parameters characteristics, etc. an effective prediction and control of the blast furnace temperature is the necessary condition to guarantee the smooth process of smelting blast furnace. Since Silicon content of molten iron is the representation of blast furnace temperature, many scholars have made a large amount of research on silicon content prediction model of blast furnace. This paper mainly studies a novel evolution algorithm to solve the multi-objective optimization problem, namely Predator-Prey genetic algorithm, which is applied to BP neural network modeling of molten iron silicon content in blast furnace. And we conclude Pareto optimal solution set of the BP neural network model, among which the proper network model is chosen for prediction.This paper mainly divided into five chapters, telling about the following five aspects respectively.The first part introduced blast furnace iron making process and the research of blast furnace temperature prediction. The concept of hit rate of silicon content, time series prediction model, neural network model and genetic algorithm were briefly introduced; The second part mainly introduced the multi-objective optimization problems, because of the essential difference between the multi-objective optimization and single objective optimization, in this section, we introduced Pareto front and nondominated solution, and also discussed the traditional algorithms and evolutionary algorithms to solve the multi-objective optimization problem, giving a detailed introduction to NSGA-Ⅱ; The third part studied Predator-Prey genetic algorithm (PPGA). In this section, the author studied the algorithm strategy of PPGA, and carried on the experimental comparison, and it is concluded that the design the randomness and the certainty theory, and we put forward the improved Predator-Prey genetic algorithm, MPPGA algorithm; The fourth part, we made use of MPPGA to analysis BP neural network modeling of Bao Tou Steel blast furnace data, and draw the conclusion that based on complexity and precision of models. Among the Pareto optimal solutions we elected neural network model to predict silicon content; the fifth part is the summary and outlook. It summarizes the research results and points out some deficiencies in this study and subsequent research directions.