Research On The Identification Of HVDC Transmission Interference Based On Deep Learning Of Geomagnetic Observation Data

With a large number of HVDC lines in operation,the unbalanced currents generated during the transmission process have a serious impact on the geomagnetic observations of the surrounding stations.In order to detect the influence of HVDC transmission events in the geomagnetic observations,experts have used semi-manual identification methods to identify HVDC transmission events in the geomagnetic data.However,as the number of geomagnetic observation instruments constructed and the number of HVDC transmission lines increased,the size of the raw geomagnetic observations subject to HVDC transmission interference increased and the existing semi-manual identification of HVDC transmission interference events increased exponentially.In order to efficiently,accurately identify HVDC transmission interference events in geomagnetic observation data,this paper explores HVDC transmission interference event identification methods based on deep learning technology,and has the following research content:(1)This paper uses the statistical method for statistical processing,comprehensive analysis of the historical records of high-voltage transmission interference incidents accumulated in many years,summarizing the characteristics of HVDC transmission interference incidents,laying the foundation for the next production of high quality depth learning samples.(2)For HVDC transmission interference incidents,the long-lasting period is different,the amplitude change is uncertain,the shape difference is large,and the different measurements of the observation instrument affects the different characteristics of different levels,and the HVDC transmission interference event sample and normal sample Production technology and processes.(3)For the "step" morphological characteristics of HVDC transmission interference events and intended timings,a model of Interference Incidents Classification Model for High Voltage Direct Current Transmission based on CNN and LSTM(IICM-HVDCT-CNN-LSTM)are proposed.The model fuses the advantages of convolutional neural network and long short-term memory neural network.The former has strong morphological feature extraction capabilities,the latter can sufficient mining characteristics contained in the time series,using two network characteristics to extract geomagnetic observation data HVDC transmission interference event features,thereby achieving precision recognition.After the IICM-HVDCT-CNN-LSTM model,the recognition accuracy of the verification set reached 92.94%,and the accuracy of the test set identification reached92.86%.It indicates that the depth learning method has high accuracy on the HVDC transmission interference incident in the identification of geomagnetic observation data,which has reference significance for other types of interference events in the geomagnetic observation data.