极大极小值交替优化算法实现换热网络全局热集成

Mini-max alternate optimization algorithm for global heat integration of heat exchanger networks

针对确定性方法应用于换热网络全局热集成时易陷入局部极值的问题,采用乘子法建立辅助函数,将原来的换热网络有约束问题转化为无约束问题,在此基础上提出了基于牛顿法的极大、极小值交替优化算法应用于换热网络优化。该算法通过优化进程中极大值、极小值的交替计算,不断跳出当前的局部极小值并继续通过局部优化方法求解新一轮的局部极小值,从而实现换热网络的全局热集成。同时提出防止"回跳"策略,避免该算法计算过程中在某个区域重复优化的问题。将算法应用于两个经典换热网络实例,取得了较好的结果,验证了极大、极小值交替优化算法能够有效地改善确定性方法易陷入局部极值的问题,具有较强的全局搜索能力,使优化质量较文献得到了进一步提升。

Deterministic methods are easily trapped into local optimal when applied in global heat integration of heat exchanger network（HEN）resulting from the serious nonconvex and nonlinear characteristics of HEN problem. An auxiliary function was established on the basis of multiplier method to turn the primal constrained problem into an unconstrained one. Then,a mini-max alternate optimization algorithm based on Newton method was put forward to apply in the optimization of HEN,where the maximal and minimal optima were alternately solved to jump out of the current local optimum,continue to obtain the new local minimum value with the local optimization method,and realize the global heat integration of HEN. Moreover,a strategy to prevent the ＂rebound＂ phenomenon was proposed to avoid the repeated optimization in the certain region. Two HEN cases were used to verify the feasibility of the algorithm. The results compared with the literature data showed that mini-max alternate optimization algorithm is effective to improve the performance of deterministic method in the optimization of HEN and has strong global search ability.