Identification Of HV Cable Metal Sheath Grounding System Defects Based On Deep Learning

In recent years,the metal sheath of high voltage(HV)cables has been widely used in urban transmission and distribution networks by commonly adopting cross-bonded ground8 ing methods,which can effectively suppress the sheath induced voltage.The influence of external environmental factors leads to excessive current in the metal sheath,causing various problems such as increased metal sheath loss,reduced transmission efficiency and shortened cable life,which seriously affects the safe and stable operation of the power system.The cable sheath current,as one of the key condition quantities inspected by operation and maintenance workers,can effectively reflect the operational status of HV cables.However,the current diagnostic criteria based on sheath current are relatively simple,and the relation between sheath current characteristics and different defect types has not yet been established.In this context,this paper proposes a method for identification of HV cable metal sheath grounding system defects based on deep learning.Firstly,this paper takes a three-stage single-core HV cable cross-bonded grounding system as the research object,analyses the formation mechanism of the sheath current and establishes a theoretical model of the sheath current of the HV cable grounding system.The sheath current is composed of the induction current of electromagnetic induction and the leakage current of electrostatic induction,of which the induction current is the main part of the sheath current.Secondly,the causes of the four types of defects-sheath circuit open circuit,cable joint breakdown,cross-bonded box water ingress and sheath grounding-are described and the theoretical model of the sheath current following each fault is analysed in detail.Then,a HV cable metal sheath grounding system model was built by PSCAD/EMTDC software,and simulations were carried out for normal conditions,sheath circuit open circuit,cable joint breakdown,cross-bonded box water ingress and sheath grounding fault respectively,and the variation of sheath current waveforms in the same sheath circuit and same grounding box under different defects were analysed,and the amplitude ratio and phase angle difference of sheath currents under different sheath circuits were analysed.The results show that the amplitude ratio of the sheath currents at the first and end of the sheath circuit is close to 1 and the phase angle difference is less than 1° under normal conditions,while the amplitude ratio and phase angle difference of the sheath currents under defects have a large difference.Finally,14 feature vectors were constructed based on the sheath current amplitude ratio and phase angle difference to reflect the operation status of the grounding system,and a HV cable grounding metal sheath grounding system defect database containing 18 operation statuses was constructed.The deep learning algorithm long and short-term memory(LSTM)is used to identify the defects in the metal sheath grounding system of cables.The simulation results show that the proposed method has an accuracy of 99.72%,and can accurately and effectively identify different defects in the metal sheath of cables.