The authors address the determination of the set of active constraints in optimal power flows based on the Newton approach. An adaptive method for handling movement penalties is proposed. The method uses information about the size of the pivots in the factorization of the Hessian matrix to selectively introduce movement penalties whenever it is strictly necessary. Thus, because movement penalties are only used when required, the number of state variables whose movement is penalized is greatly reduced. The main advantage presented by the method is that it ensures positive definitiveness without negatively affecting the convergence characteristics. No tuning is necessary, which is convenient for real-time applications. Test results carried out on two network models of actual power systems (100 and 1650 buses) are reported.71334342