Abstract: Infrared temperature measurement is a crucial method for assessing the condition of composite insulators. To reduce the impact of infrared lens shooting distance and angle on infrared temperature measurement of composite insulators and improve temperature measurement accuracy, this paper proposes a prediction model based on a Particle Swarm Optimization-Back Propagation Artificial Neural Network (PSO-BP). First, through high-voltage temperature measurement tests on the FXBW4-110/70 model 110kV composite insulators under normal and polluted conditions, the surface temperatures of the composite insulators were collected using both infrared thermometers and optical fiber thermometers. The infrared temperature values, distance, and angle were used as input parameters for the PSO-BP model, while the optical fiber temperature values were used as output parameters. The PSO model was utilized to optimize the initial weights and thresholds of the BP model. And traditional statistical regression methods such as linear, non-linear, and exponential regression, machine learning models like ELM, RBF, and PLS, and deep learning algorithms like BP and GA-BP were employed to correct the infrared temperature of the composite insulators. The results indicated that the PSO-BP model outperformed traditional statistical regression methods, machine learning models, and deep learning algorithms in correcting infrared temperature measurement errors, reducing the average errors of the infrared thermometer and thermal imager from 17.06% and 9.07% to 1.36% and 0.97%.