Fault Location Method For Small Current Grounding System Based On Transient Information

Small current grounding is widely employed in medium and high voltage distribution network in China.In this situation,the steady-state zero-sequence current is small and fault characteristics are weak after a single-phase grounding fault occurs..Compared with the transmission network,the distribution network has shorter lines and more branches.In addition,cable lines and hybrid are applied in distribution network.All of this bring great challenges to the fault location in distribution networks.The traditional fault location methods used in transmission network have some limitations when used in distribution network directly.The methods using fault steady state information such as impedance and voltage are affected by weak steady-state signals.Especially in the resonant grounding system,the problem is more serious after the line capacitive current is compensated.Traveling wave head signal is difficult to capture accurately because of the complex topology of distribution network,which makes the reliability of location method based on traveling wave need to be improved.There are abundant transient information after fault occurs in smalll current grounding distribution network,Fully extracting the fault information contained in it can effectively solve the problem of fault location in complex distribution network.Therefore,this thesis studies the fault location method of small current grounding system based on transient process information after fault.The specific work are as follows:Firstly,the transient process of single-phase-to-ground fault in small current grounding system is analyzed.The transient current is composed of a plurality of high-frequency components.Among these frequencies,there are several frequencies that make the equivalent loop resonate.The frequency corresponding to the largest amplitude is called the main resonant frequency.Theoretical analysis shows that there is a certain connection between the main resonance frequency and the corresponding component's amplitude with the fault distance.It is proposed that the main resonance frequency can be used as a transient fault feature to locate the fault,which lays a theoretical foundation for fault location method based on transient information.Secondly,the wavelet multi-resolution analysis is used to extract the transient fault features,and the wavelet energy in the frequency band of the main resonant frequency is obtained.Considering that the main resonance frequency and its components are affected not only by fault distance,but also by fault resistance and fault initial phase angle,in order to avoid the influence of fault resistance and fault initial phase angle on the wavelet energy,the double-ended wavelet energy ratio at the corresponding scale of the main resonance frequency is selected as the fault feature.The relationship between wavelet energy ratio and fault distance is fitted by neural network,and the fault location method based on double-ended wavelet energy ratio and BP neural network is obtained.The applicability of the proposed method in the hybrid line and branch line is analyzed and discussed.By using PSCAD and MATLAB simulation examples,the feasibility and accuracy of this method for locating various line structures(overhead line,cable line,overhead line-cable hybrid line and branch line)in 10kV radial distribution network system are verified.Finally,due to the fitting effect of different neural networks is different,RBF network and SVM network are used to construct the location model,and the results are compared with those of BP network,the positioning results of BP network,RBF network and SVM network are combined.A weighted combination positioning model with minimum mean square error as the objective and a variable weight combination model with adaptive fuzzy are constructed.The root mean square error,average absolute error and maximum absolute error are set as the performance evaluation indexes of the positioning model,and the positioning results of the single network model are improved.The simulation results are compared with the location results of a single network model to verify the effectiveness and reliability of the fault location model based on the combination of multiple neural networks.