To enable rapid online identification of abnormal heating in induction motor windings

a winding heat source inversion strategy based on reduced-order modeling and particle swarm optimization was proposed. Based on an analysis of the motor structure and the forward and inverse heat transfer processes

a full-order model(FOM)of the induction motor temperature rise was constructed using the finite element method. Sampling points under various operating and environmental conditions were obtained through Latin hypercube sampling

and FOM simulations were performed at these points to construct a dataset for model order reduction(ROM). The proper orthogonal decomposition(POD)and response surface method(RSM)were employed to extract the dominant modes and predict their corresponding coefficients

respectively

thereby establishing a reduced-order surrogate thermal model for the induction motor. Using temperatures measured at a limited number of points on the housing as inputs

the established reduced-order thermal surrogate model and particle swarm optimization(PSO)algorithm enable rapid reconstruction and identification of the winding heat source. Experimental validation was conducted by introducing inter-turn short circuits to simulate abnormal winding heating. The results show that the maximum errors between the calculated temperature rises of the winding and housing using the developed ROM and the measured values are within 3.7% and 4.2%

respectively

while the computational speed is increased by a factor of up to 1 667 compared with the FOM. The proposed inversion model accurately diagnoses abnormal winding heat sources caused by faults

with a single inversion taking only 25 seconds

providing an effective method for online monitoring of early abnormal heating in motor windings.