A method for monitoring insulation degradation of an electromagnetic coil based on a long short-term memory network

Abstract

The application relates to a long short-term memory network-based electromagnetic coil insulation degradation monitoring method, which comprises two parts of a sensitive feature extraction module and an intelligent prediction module; wherein the sensitive feature extraction module comprises a statistical feature extraction unit and a sensitive feature screening unit, and the intelligent prediction module comprises a feature normalization unit and a long short-term memory network. High-frequency signals are injected into the electromagnetic coil to be measured, the response signals (i.e. high-frequency electrical parameters) of the coil are analyzed to perform degradation monitoring, the sensitive feature extraction module is used to perform statistical feature extraction and screening, then the feature data of various degradation states are normalized, and then a corresponding health index (HI) curve is constructed through the long short-term memory network to evaluate the degradation state. The application is used for electromagnetic coil degradation monitoring, even under the condition that the coil working condition changes and the coil temperature changes, the method can still complete accurate insulation degradation monitoring, and finally ensures the safe and stable operation of the electromagnetic coil in the use process.

Images