基于CNN-LSTM的电解铝变压器振动-温升耦合异常智能诊断技术研究OA

These problems make it difficult to adapt to the high reliability requirements of electrolytic aluminum power supply.This article takes the electrolytic aluminum power supply system of Guangxi Hualei New Materials Co.,Ltd.as the research object,and combines on-site operation and maintenance practice to construct an integrated technical solution of"electrolytic aluminum power supply scene adaptation data acquisition coupling feature extraction CNN-LSTM intelligent diagnosis":real-time and accurate collection of vibration and temperature rise signals is achieved through anti-interference and dust-proof sensor networking,multi-dimensional coupling features are extracted based on wavelet packet transform,and CNN-LSTM fusion model is used to achieve accurate identification and early warning of key component faults such as iron core,winding,and cooling system.Experiments and field applications have shown that the diagnostic accuracy of the system reaches 99.2%,with an abnormal response delay of≤80ms.It can warn faults 4～6 hours in advance and reduce downtime losses caused by electrolytic aluminum power supply interruptions by 1.2 million yuan in the past year.This provides an efficient technical paradigm for predictive maintenance of electrolytic aluminum power supply system transformers.