Image recognition-based transformer temperature controller automatic calibration method

Abstract

This invention provides an automatic calibration method for transformer temperature controllers based on image recognition, belonging to the field of metrological calibration technology using computer vision. The method includes: injecting a pseudo-random current sequence into the temperature controller under test to identify thermal response characteristic parameters; extracting the visual approximation vector of the instrument pointer in real time using machine vision; using a vision-coupled thermal impedance feedforward observer to fuse and calculate the state residual between the effective temperature of the sensing bulb and the visual reading, generating a dual-source collaborative calibration control command; responding to this command, adjusting the external temperature field's temperature change rate and the internal fine-tuning heat flow in parallel, driving the pointer to approach the calibration point along a preset trajectory. Through dual-source collaborative control and dynamic thermal impedance modeling, the contradiction between heat conduction hysteresis and visual feedback delay in traditional calibration is effectively resolved, achieving high-precision and high-efficiency automated calibration.

Images