摘要

This paper proposes a novel residual electrical endurance (REE) prediction modeling method for low-voltage electromagnetic alternating current (ac) contactors based on conditional density estimation (CDE). For a type of ac contactor, the current REE value is defined by the ratio of the current operation number against its maximum operation number. Voltage and current signals between contacts during breaking processes are acquired as the initial data. An arcing characteristic set of every breaking process is computed from the voltage/current signals, and the characteristic sets are grouped and averaged to achieve training and testing data sets. Four data schemes are applied and compared. CDE modeling with cross-validation is performed on the training data set to build REE prediction models, which are then applied to estimate the REEs of testing observations. The experiment proves that the modeling method is able to reach relatively precise predictions.