An Intelligent Online Diagnosis Method for Surge Arresters Status Based on ZnO Varistors Impulse Aging Test

The implementation of online monitoring and intelligent diagnosis method for the surge arrester can enhance the supervision of its full lifecycle status, thereby further improving the safe and reliable operation of the power grid. This paper proposes an intelligent online diagnosis method for surge arrester status. Firstly, an impulse aging test of ZnO varistors used in surge arresters is conducted to establish an experimental dataset of impulse test data of ZnO varistors. The electrical parameters are measured and then analyzed. Considering the economic cost and difficulty of on-site testing, it is proposed to establish a relationship model between the online monitoring parameters such as the power-frequency total leakage current <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">I</i><sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x</sub>, as well as its harmonics, and the offline parameter DC reference voltage <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">U</i><sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1mA</sub> by machine learning(ML). Subsequently, correlation analysis is performed for feature selection to verify the equivalence of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">I</i><sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x</sub> and the resistive component <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">I</i><sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">xr</sub> as features in ML. In addition, various hyperparameter optimization algorithms and objective functions are compared to evaluate different ML models for predicting <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">U</i><sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1mA</sub>. An optimal model is proposed and validated using the new core and the new varistor datasets. The maximum relative errors are 2.20% and 1.78%, with 81% and 91% of the samples having relative errors under 1%, respectively. Finally, an intelligent online diagnosis method is proposed for surge arrester status. Its diagnostic effectiveness is validated, and its generalizability is discussed.