Nonconvex generalized Benders decomposition for stochastic separable mixed-integer nonlinear programs (Q662860)

The paper develops a decomposition type algorithm for the solution of scenario-based mixed-integer nonlinear two-stage stochastic programs of a special structure: the first-stage decision variables are integer, the second-stage variables are continuous and all functions are separable in integer and continuous variables. Moreover, it is assumed that at least one of the constraints is nonconvex. Convexification of the problem leads to a lower and upper bounding scheme which is progressively improved to obtain an \(\varepsilon\)-optimal solution in finite time. The computational advantage of the new decomposition algorithm over general global optimizers for large-scale engineering problems is reported.