Edas Based On Niche Technology In Two Phases In The Iron And Steel Production Scheduling Research

Estimation of Distribution Algorithms(EDAs) provides a general framework solving complex system optimization problems. It does not depend on specific fields of problems, and possesses strong self-organization, self-adaptation and self-learning features, so it has extensively applicative future in all kinds of fields, for example, in the optimization of functions, the combinatorial optimization, the productive scheduling, the process controlling, the image-processing and so on . With achieving a series of advance, Estimation of Distribution Algorithms recently has been an important orientation of researching in evolutionary computation.This paper proposes an improved algorithm based on Estimation of Distribution Algorithms and introducing niche technology into Estimation of Distribution Algorithms, and then applying it to optimization of problem of steel production scheduling.The main work in this paper is as follows:1. This paper systematically studies and elaborates the origin, the basic principles, mechanisms, features, and applications of Estimation of Distribution Algorithms.2. Because of decreasing of population diversity in the late evolution, algorithm presents local premature. With niche technology applied to Estimation of Distribution Algorithms, we propose an improved algorithm- two-stage Estimation of Distribution Algorithm base on niche technology, and using the approach that niche probability combines with global population probability generate new solutions. Simulation results show that proposing algorithm effectively prevents premature convergence and greatly improves global searching efficiency.3. EDAs is applied to optimization of rolling batch planning in steel industry. In the first, building mathematical model of steel rolling planning; next, converting mathematical model into a non-repeat integer sequence for solve; at last, using practical production data of a steel industry carries out simulation, and verifies the effectiveness of the model and algorithm.