Transformer fault detection method based on multi- source sensing data correlation analysis

Transformer equipment plays an important role in the stable operation of power system, and for monitoring and troubleshooting the operation status of large transformer equipment, it can improve the efficiency of equipment maintenance and improve the reliability and stability of power distribution equipment operation. The amount of local and onesided information obtained by a single sensor on the state of the transformer is very limited, and each sensor receives its own quality, performance, and noise effects, the information collected is often imperfect, with great uncertainty. Multi-sensor information integrated system using multiple similar or heterogeneous sensors can obtain a large amount of information from multiple angles in full space, although the information obtained by some sensors is wrong, or there will be redundancy and contradiction between information, can side-react the operation of transformer state, if combined with other correlation sensing factors, can fully react to the operating state. Through point clustering and linear regression technology, the equipment sensing historical data is trained offline to detect the potential correlation between the sensing data. The test results prove that the transformer fault of multi-signal sensing is more persuasive and reliable under the factors of combining many other sensing signals.