“…In this approach at each iteration only a subset of discrete variables which are sufficiently close to a discrete value are rounded-off, the remaining variables (treated as continuous) being then reoptimized. It is largely agreed that the round-off technique is generally suitable for discrete variables with small steps (e.g., load tap changer (LTC) transformer ratio and phase shifter angle) but requires some caution for discrete variables with larger steps (e.g., shunt compensation banks, network switching) [12], [14], [16]. However, the round-off approaches act "blindly" since they do not look at the discretization effect on either the objective or the inequality constraints, suffering in consequence from two drawbacks: (i) the solution feasibility is not guaranteed while no method to restore feasibility is proposed, and (ii) the objective value may be unacceptably deteriorated.…”