Research On Combination Of TS & SQP Optimization Methods

This paper proposes two new algorithms about tabu search. One algorithm of TS which combined with a new nested constraint handling method is proposed to solve global nonlinear constrained optimization problems. Another algorithm of tabu search which also combined with SQP is to solve mixed integer nonlinear optimization problems. Its application on heat exchange network synthesis (HENS ) is also studied. The main results can be summarized as follows:(1) A new constraint handling method for TS is proposed. Up to now, most of continuous tabu search algorithms could only solve constraint-free continuous nonlinear optimization problems. However, many problems are constrained. Most constraint handling methods are based on penalty function, which is inefficient for TS as it is hardly to keep the search within the feasible region by purely adding the penalty to the objective function. A nested method is thus proposed to deal with constrained optimizations based on a strategy that is capable of punishing the candidates out of the feasible region, leading the search back into the feasible region, and keeping the search within the feasible region. It combines SQP and continuous tabu search in a nested inner and outer loop way. The SQP algorithm is used to fast solve a small scale NLP optimization problem for constraints in the inner loop, and the continuous tabu search is used to seek the global optimal in the outer loop. Efficiency is demonstrated by simulations.(2) A new tabu search algorithm combined with SQP to solve mixed integer nonlinear optimization problem is also presented. Considering that when MINLP is given its integer part variables, it can be transformed into common nonlinear optimization problems. Based on this, a new combined algorithm of TS & SQP is proposed to deal with the MINLP in a way with outer and inner loops. TS algorithm is used to deal with the integer variables in the outer loop to search the best integer value, and SQP is used to solve the sub-NLP in the inner loop to quickly obtain the best continuous value. The simulation shows the goodperformance of global convergence in solving the mixed integer nonlinear optimization problems.(3) The TS & SQP algorithm is also applied to deal with the heat exchange network synthesis (HENS). Considering of the particularity of HENS, an automatic modeling method is proposed between the inner and outer loop of the previous method. TS is still used to search optimal stream match in the outer loop, and the SQP is used to solve the relevant heat exchange optimization model which is automatically modeled adpting to the structure obtained by the outer loop TS method. The simulation shows our method is efficient to solve the HENS withgood performance in reaching the global optimal solution.