Research On Multi Parameter Online Monitoring System Of Transformer Based On BP Neural Network

The electric power system contains a lot of electrical equipment,and the electrical equipment plays an important role in the power system.Transformer online monitoring is conducive to the working state of the transformer's real-time knowledge and understanding,when the transformer is out of work can be found and treated in time,to avoid a harmful impact on the whole system,at the same time,it is also conducive to overhaul the fault timely on the transformer,which will lest cause greater damage,so as to reduce the occurrence of the fault caused by the the loss to the stable operation of power system.In this paper,it briefly introduces the influence of transformer online monitoring on the entire power industry,and discusses the research background and significance of online monitoring of transformer,with analyzing the development situation in recent years;secondly,it introduces several commonly used wireless sensor network technology,analyzes and compares their advantages and disadvantages,and selects the ZigBee technology to collect and transmit the information of transformer online monitoring parameters,and makes a detailed introduction to the ZigBee technology;Then,this paper discusses the design and overall structure of online monitoring system of transformer,it also carries out the hardware circuit design on the structure of the research on acquisition and transmission of wireless sensor network data.The parameters of the transformer are monitored by the local discharge and the dissolved gas in the oil.The sensor is selected and the pulse current sensor and ultrasonic sensor are used to detect the partial discharge signal of the transformer,and the semiconductor gas sensor is used for the detection of dissolved gas in the oil;Finally,the software design of wireless sensor network data collection and transmission is introduced.The working principle of BP neural network is briefly introduced.The input and output vectors and the number of hidden nodes are selected respectively to determine the training of learning samples.The multi-parameter online monitoring model of transformer based on BP neural network was constructed,and four kinds of fault types of transformer were determined.