模拟退火法用于连续变量问题全局优化初探

A SIMULATED ANNEALING METHOD FOR GLOBAL OPTIMIZATION TO THE CONTINUOUS VARIABLE PROBLEMS

针对过程系统连续变量优化问题中普遍存在的多峰现象，初步探索了应用模拟退火法求解全局最优解的问题，针对变量只有上下限不等式约束的问题，根据连续变量问题的特性，提出了一种相邻状态的产生函数，并分析了模拟退火过程的起始温度、终止温度以及降温速度对优化计算的影响，给出了这些参数的适宜区域，通过三个例题的计算，将模拟退火法与传统优化方法─基于梯度的方法进行了对比分析，结果表明该法能够有效地解决传统的确定型优化方法所不能奏效的全局最优解问题。

A simulated annealing approach is introduced in this paper to search the global optimal solution, aiming at the multipeak phenomenon which .generally exists in the process system optimal problems with continuous variables. For the Problems which only have upper and lower limits to all variables, a function of producing adjacent state is proposed. The effects of initial and end temperature, as well as temperature descent in simulated annealing process to the optimal calculation are analysed, their suitable scopes are given based on the property of the continuous variable. The results of three examples show that the simulated approach can efficiently solve the global optimal problem which is difficult to deal with by the traditional deterministic optimal methods.