A New Decomposition Algorithm on Non-separable Optimization Problems

Considered is the non-separable optimization problem for a class of large-scale systems, where the overall objective function is not of an additive form with respect to each decision variable. A three-level optimization algorithm is proposed. It first converts the original problem into separable parametric optimization problem. The optimal solution of the original problem is then selected from the set of solutions of parametric optimization problem. Theoretical base of the algorithm is established. Simulation result for one example shows that the algorithm is effective.