Deriving a new lower bound in stochastic programming problem considering variance

One approach in stochastic programming involves minimizing the expectation of cost, which includes a recourse function to represent the penalties for constraint violation. A known solution for such problems uses the L-shaped method based on the Benders decomposition. Stochastic programming problems that take the variance of the recourse function into account are non-convex problems, and an exact solution based on the branch-and-bound method has previously been demonstrated. However, since there are cases where the lower bound for the variance of the recourse function used in the branch-and-bound method is not effective, we present a new lower bound and demonstrate the effectiveness of the resulting solution by a numerical experiment.