A Logistic Regression Approach for Improved Safety of the Under-Frequency Load Shedding Scheme Owing to Feeder Machine Inertia

Under Frequency Load Shedding (UFLS) is the last resource to maintain the power system frequency after contingences. Load shedding can be performed by conventional, adaptive, computational or wide area moni-toring techniques. The most common is a conventional UFLS. Load shedding is accomplished when pre-determined frequency thresholds are reached. Although the conventional methods are simple, easily implemented and dependable, they may operate improperly. In the Brazilian electric power system, several actions of the UFLS scheme were found to be improper owing to the dynamics of induction motors and distri-bution generation during the main source outage of distribution feeders. In this way, when the main source is out of service, machine inertia maintains the frequency for some electric cycles, leading the frequency relay to undue tripping. In this paper, the negative sequence voltage and the rate of change of frequency are handled by a logistic regression method to identify the events of improper action. The proposed method was evaluated through the modified IEEE 9-bus test system, and several cases have been tested. The proposed method reached a level of performance higher than 99% of correct operation, and it needs a small training set.