algorithm keeps running the SQP program to find a better solution or terminates when no further
improvements are found.
Third, Mansoornejad’s approach has another problem of switch timing from SQP to GA. If the SQP
step size is too small, the algorithm shifts to the GA. This switching strategy may raise two additional
problems: (1) How should we determine the “large enough” step size for proceeding in SQP; and (2) The
intermediate termination of SQP may not generate a useful base for the following GA. To solve these two
problems, we switch to GA only if we reach a local optimun in SQP.
In order to exploit the major advantages of both GA and SQP and improve the defects of the
algorithms, a hybrid GA – SQP method is developed and applied in case studies. In general, the proposed
hybrid approach first implements global search with GA and then runs SQP to reach the final solutions.
Through this algorithm, the GA can converge very quickly at beginning and provide a good initial
solution for SQP, which then searches until no further improvements are found.