在热轧生产调度数学模型的基础上,提出一种改进的基于小生境技术的分布估计算法。采用Hebb学习规则来更新概率模型,增强了小生境之间以及整个群体间的相互学习能力。在对新种群采样时,引入了一种竞争选择机制,使得父代中的优秀个体在进化过程中得以保留,从而保证了进化持续有效的进行。为将该算法用于热轧生产调度,设计了一种新的解码方式,并采用某钢厂实际生产数据进行仿真实验,验证了算法的有效性。 Based on the mathematic model of hot rolling production scheduling, an improved distribution estimation algorithm based on niche technology is proposed. The use of Hebb learning rules to update probability models enhances mutual learning among niches and across groups. When sampling new populations, a competitive selection mechanism was introduced to allow elite individuals in the parent generation to be retained during evolution, thus ensuring the evolution of the continuum effectively. In order to apply this algorithm to hot-rolling production scheduling, a new decoding method was designed and simulated by the actual production data of a steel mill, which verified the effectiveness of the algorithm.