A Soft Measurement Model of SF6 Gas Leakage State Based on Neural Network

The gas leakage detection of SF6 gas density relay is a critical aspect in ensuring the safe operation of electrical equipment. Although the current application of SF6 density relay soft measurement models and their implementation technologies can achieve maintenance-free real-time calibration, the calibration process requires complex numerical calculations. Due to the good linear characteristics of SF6 gas pressure and temperature measurement when density remains constant, this paper integrates neural networks and pattern recognition theory to propose a SF6 gas leakage state soft measurement model based on BP neural network. Through training with simulated datasets, the model generated by this method can directly determine whether the gas is leaking based on SF6 gas pressure and temperature without the need for complex analytical calculations. Test results show that compared to traditional diagnostic methods, this approach offers higher accuracy and computational speed, significantly enhancing real-time performance.