Research On Fault Location Of Distribution Network With DG Based On Multi-criterion Fusion

According to statistics,apart from the factor of insufficient power generation,95% of power outages suffered by power users are caused by distribution network faults,and most interphase faults are evolved from single-phase ground faults.Among them,single-phase grounding faults account for over 80% of the total number of faults in the distribution network.Therefore,research on fault location mainly focuses on solving the problem of locating single-phase to ground fault sections.At this stage,distributed generation technology has many advantages such as cleanness,efficiency,environmental friendliness,energy diversification,etc.,which is in line with the national sustainable development strategy.With the increasingly tense world energy situation,distributed generation technology has developed rapidly in recent years.However,as more and more distributed power sources are connected,the fault current in the distribution network is limited,and the fault characteristics become unclear and harmonic complex,which has a serious impact on the accuracy of fault diagnosis.Considering the fault location and identification of distribution networks with distributed power sources has practical significance in improving the economic benefits of distribution networks and reducing operational risks.The integration of distributed power sources also increases the requirements for fault location technology in distribution networks.The main research contents are as follows:(1)Considering the structural characteristics of distribution networks with distributed power sources,especially for the asymmetry of line parameters in medium and low-voltage distribution networks and the presence of electromagnetic coupling between phases in three-phase systems,it is difficult to directly analyze transient processes in phase domain systems.Therefore,a traditional fault location method has been proposed with the aim of quickly locating the fault point.Firstly,using real-time data provided by micro-synchronous phasor measurement units,combined with the commonly used Karrenbauer transform,the phase domain system is transformed into a module domain system without electromagnetic coupling for analysis.The influence of Karrenbauer transform on the phase angle difference before and after faults in a low-voltage active distribution network sections is studied,and the criteria for the occurrence of faults and fault types in a certain section are obtained,and preliminary verification is obtained through simulation.Ultimately,it reduces both the amount of fault diagnosis and the influencing factors.(2)In response to the fact that a single-phase mode transformation modulus in the above methods cannot reflect the characteristics of all faults,and the current complex structure of intelligent distribution networks,a fault feature sample library is proposed based on the Karrenbauer mode current phase angle difference data,and a support vector machine regression proxy prediction model based on machine learning is established to further determine the location and classification of section faults.The mean square error and correlation coefficient are simultaneously used as the prediction accuracy evaluation index of the surrogate model to analyze the fault characteristics of distribution network with distributed generation,and a high prediction result is obtained,indicating that the prediction performance is good.(3)In response to the problems faced by distribution systems,including waste caused by overlapping fault information between advanced applications,insufficient computing power to process large amounts of data and information,and difficulties in obtaining sample datasets under existing machine learning methods,this thesis proposes a multi criterion distribution network localization method based on the combination of traditional transformation algorithms and modern artificial intelligence.Finally,compared with other methods in the field of fault detection,it is concluded that the method proposed in this paper establishes a targeted training sample set in the feature extraction stage by combining fault mechanisms.It fully utilizes the characteristics of distribution network fault mechanisms,is not affected by the size of transition resistance,and can be applied to distribution networks with distributed power sources that change line parameters,coexist with multiple distributed power sources,and operate in modern intelligent distribution networks.