Accelerating Convergence Method for Relaxation Cooperative Optimization

Abstract

To solve problems of higher computational cost and lower convergence speed in collaborative optimization, a new relaxation cooperative optimization method of accelerating convergence is presented. The optimizations in this method are divided into two stages. In the accelerating convergence stage, the calculation method of relaxation factors is improved, relaxation factors are constructed according to the inconsistent information between each disciplinary optimal solution and their average value; in the optimization solving stage, the optimal solution in the former stage is taken as the initial point, relaxation factors with consistent precision are collaboratively optimized to obtain the final optimal solution. Finally, this optimization method is tested by a typical numerical example. Experimental results show that this method can reduce the computational cost and accelerate the convergence speed greatly.