A Fault Location Method Based On Voltage Sags Monitoring Data

With the rapid increase of power quality monitoring data and the improvement of data analytics,it becomes a hot point in power quality research field to transform the monitoring data into useful information.The data analytics will combine the massive monitoring data with the operation demand.The value of the power quality data will be explored to satisfy the operation demand.To explore some applications of the power quality monitoring data,this thesis presented a fault location method based on voltage sag monitoring data.The relationship between residual voltages on monitors and fault locations is learned by a backstep propagation neural network(BPNN).The main contents were given as follows.1.The voltage sag monitoring data in a city grid from 2016 to 2017 were clustered by twostep clustering method.Four types of voltage sags were obtained.According to the residual voltages in different type voltage sags,the potential relationship between voltage sag monitoring data and fault types and locations was shown.2.A fault location method based on BPNN was proposed,including the designation of BPNN structure,definitions of the variables on input and output layers,the parameters and the training algorithm.A comprehensive method to distinguish bus and line fault with the BPNN output value was proposed to determine the fault location detail.3.An “offline training & online application” mode of fault location was proposed and the procedure was designed.Firstly,the data format conversion was realized by compiling the interface program of BPA and MATPOWER and the voltage sag data were achieved by automatic simulations.Secondly,BPNN was trained with the simulation data offline to get the fault location model.Finally,the model can be applied online with voltage sag data from the power quality monitoring system to obtain the fault locations.4.The proposed method was validated by the transmission and distribution standard network model and one actual urban power network model.In the actual urban power network,the proposed method to identify the detail fault locations was verified.Pearson correlation analysis proved that there was a strong correlation between the discriminant index and the fault locations.In the application in the distribution network,the impact of the number of monitors on the accuracy was focused on.The proposed method was based on the “big data”.With “offline training” with “big” voltage sag data,the topologies of the grid and the impedances of the lines can be taken no consideration during “online application” which will increase the speed of fault location with the satisfied accuracy.This is an important attempt in the application of power quality monitoring data.