Performance Assessment of Steady-State Voltage Stability Indices for Parameter Validation Using ANFIS

Precision and consistency in monitoring the voltage stability conditions of power system networks remain vital aspects of modern power system operation. Thus, this study presents a machine learning approach for comparative performance evaluation of some voltage stability indices (VSIs). Six essential power system parameters from the transmission line data and load flow solution at different real and reactive load levels are used as input information for implementing adaptive neuro-fuzzy inference system (ANFIS) models based on the subtractive clustering rules. The performance of the considered VSIs is evaluated using the mean absolute percentage error (MAPE), the Percentage Relative Root-Mean-Square Error (RRMSEp), and the linear regression based on correlation coefficient R. The IEEE 30-bus system is used as a test, and the predictive analysis results are obtained for several simulation runs for parameter tuning and model regularity. Considering the MAPE analysis, FVSI with average MAPE = 4.7429% and maximum MAPE = 6.6574% outperformed the other VSIs; average MAPE values are 5.6165% for LMN, 12.4684% for NLSI and 14.7167% for LQP. The consistency of FVSI for precise monitoring of power system voltage stability condition, compared to the other VSIs, is verified by the RRMSEp and R analysis results.