Research On Transformer Fault Diagnosis Based On Cloud-fog Collaboration

Electricity is the most important and direct energy source for maintaining social operation,while transformers are the key equipment for transmitting electricity.Once a fault occurs,it may lead to regional power outages and even cause significant losses to people’s production and life.How to ensure the safe operation of transformers has become an important research topic.The Internet of Things,Big data and other technologies provide powerful technical support for transformer fault diagnosis to ensure power system security.This thesis is based on the cloud fog collaborative framework and takes power transformers as the object to study the representation of transformer fault information and the transformer fault diagnosis method based on deep learning,which has important theoretical significance and engineering application value.The fault diagnosis framework and fault diagnosis process for power transformers based on cloud and fog synergy are established,and key technologies such as data acquisition,transmission,storage and pre-processing based on the fog end are studied to obtain the relationship between the fault state of power transformers and the amount of gas in the oil,which provides a basis for fault characterization.A fault diagnosis model based on the combination of self-organized mapping neural network and two-way long-and short-term memory network is established.The self-organizing mapping neural network is used to extract the gas content data in transformer oil as the fault quantity and characterization parameters,and the bidirectional long-and short-term memory network is used to extract the data temporal characteristics,and the differential evolutionary optimization algorithm is used to optimize the network weights,while the model hyperparameters are tuned to improve the model training speed.The superiority of the model is verified through experiments and comparative analysis with other models.To address the problem that certain types of faults in real scenarios do not have training data and traditional learning classification algorithms cannot correctly perform fault diagnosis processing,a description method of transforming fault types into oil and gas content attributes through semantic attributes is proposed,using the K nearest neighbor algorithm as a binary classifier and convolutional neural network as a feature extractor to obtain semantic feature vectors.The feasibility and effectiveness of the model are experimentally verified by attribute correction and similarity comparison to reduce the domain offset.Based on the OpenFog cloud-fog collaboration architecture,a prototype transformer fault diagnosis system is developed to realize transformer condition monitoring and real-time fault diagnosis.The correct validity of the data pre-processing method based on the fog end and the fault diagnosis model trained based on the cloud is verified,which provides an effective method and technical support for realizing power transformer fault diagnosis.