A trust-region algorithm based on global SQP for reactive power optimization

This paper presents a trust-region algorithm based on global sequential quadratic programming (SQP) for reactive power optimization. This method is not only reliable and accurate which is similar to SQP, but is also global convergent to trust-region search method. To guarantee feasible region of this subproblem is not null, inaccurate direction component decomposed method is adopted to compute trust-region subproblem. In order to avoid the Marotos effect, the penalty parameter is effectively regulated in merit function. This example of the computation shows that this algorithm has global convergence and is fast, accurate and reliable.