An Intelligent Data-Driven Learning Approach to Enhance Online Probabilistic Voltage Stability Margin Prediction
Heng-Yi Su, Hsu-Hui Hong
Abstract
This letter presents a self-adaptive data-driven learning method for enhanced probabilistic prediction of voltage stability margin (VSM). An online probabilistic extreme learning machine (ELM) algorithm based on the power transformation technique is developed. The prediction interval (PI) estimation for VSM is formulated as a Box-Cox transformation (BT) model to take into account uncertainties associated with predictions. The parameters in the transformed model are determined by the maximum likelihood estimator. The proposed PI-based VSM estimation method is applied to power grids with high proliferation of renewable energy generation. It enables to update the prediction model online and adapt to changing operating conditions. Numerical studies along with comparative results demonstrate the efficacy and robustness of the proposed method.