Datasets Comparison

This dataset contains simulated fault events for fault detection, classification, and localization in 110 kV power transmission lines. Three fault types were simulated: single-phase-to-ground fault (SPGF), two-phase short circuit (2PSC), and three-phase short circuit (3PSC). Each fault type was simulated with the fault location varying from the beginning to the end of the line in 5% increments, ensuring comprehensive coverage of fault positions along the transmission lines. A no-fault scenario was also included to represent normal operating conditions, where the system’s load was varied incrementally by 0.1 MW to simulate typical fluctuations in power demand. For each scenario, voltage and current data were extracted, including line-to-ground voltage magnitudes at all busbars and the initial short-circuit current magnitudes at the beginning of each transmission line. These features, totaling 24 input variables, are used to describe the state of the system during faults and normal operations. The output includes the fault type (SPGF, 2PSC, 3PSC, or no fault), fault line (Line1 to Line5, or None), and fault position along the line (0, 5, ..., 100). The dataset consists of 618 data points, representing various fault and normal scenarios along the transmission lines. A Python script was developed to automate the fault simulation, data extraction, and dataset generation process.

To reproduce the dataset, install DIgSILENT PowerFactory with a valid license. Set up a power transmission line model including components such as buses, generators, transmission lines, transformers, and loads, and define system parameters like line impedances and operating conditions. Develop a Python script to automate the simulation of faults in power transmission lines. The script should simulate various fault types (e.g., single-line-to-ground, double-line-to-ground, line-to-line, and three-phase faults) by configuring parameters like fault location, fault resistance, and fault duration. The script will collect key electrical parameters such as phase voltages and currents during the faults. The recorded data is then exported into a suitable format (e.g., CSV or Excel) for further analysis, with optional preprocessing to label fault types and locations for easy classification.

Power Engineering, Machine Learning, Power System Protection, Deep Learning

Creative Commons Attribution 4.0 International

This dataset contains simulated fault events for fault detection, classification, and localization in 110 kV power transmission lines. The simulations were conducted using DIgSILENT PowerFactory, with a Python script automating fault generation, data extraction, and dataset compilation. The simulations include three fault types: single-phase-to-ground fault (spgf), two-phase short circuit (2psc), and three-phase short circuit (3psc). Each fault type was simulated with fault locations varying from the beginning to the end of the transmission line in 5% increments, ensuring comprehensive spatial coverage along the lines. Additionally, a no-fault scenario was included to represent normal operating conditions, where system loads were incrementally varied by 0.1 MW to simulate typical fluctuations in power demand. The dataset contains 24 input variables that describe the electrical state of the system under both fault and no-fault conditions. Features Voltage Magnitudes: Line-to-ground voltage magnitudes for phases A, B, and C at busbars: Ua1, Ub1, Uc1 - Bus 1 Ua2, Ub2, Uc2 - Bus 2 Ua3, Ub3, Uc3 - Bus 3 Short-Circuit Current Magnitudes: Initial (subtransient) short-circuit current magnitudes for phases A, B, and C at the beginning of transmission lines: Ia1, Ib1, Ic1 - Line 1 Ia2, Ib2, Ic2 - Line 2 Ia3, Ib3, Ic3 - Line 3 Ia4, Ib4, Ic4 - Line 4 Ia5, Ib5, Ic5 - Line 5 Targets The target variables for classification and regression tasks are: Fault: Type of fault (spgf, 2psc, 3psc, or no-fault) Line: Faulted line (Line1, Line2, Line3, Line4, Line5, or NONE in case of no-fault) Position: Fault position along the line (0%, 5%, ..., 100%, or NONE in case of no-fault) The dataset consists of 618 samples, covering various fault and normal operating scenarios. For each sample: Fault events are labeled by fault type, faulted line, and fault position. No-fault scenarios are included to improve model robustness in distinguishing between fault and normal operating conditions. The dataset provides a comprehensive and systematic representation of electrical behaviors under faults and normal conditions, facilitating fault detection, classification, and localization tasks. The integration of DIgSILENT PowerFactory simulations with Python ensures accuracy and reproducibility in data generation.

Additional Provided Files The dataset is accompanied by supporting files to aid in understanding and replicating the fault simulation process: diagram.pdf: A visual representation of the 110 kV power transmission system model used in the simulations. This diagram illustrates the configuration of busbars, transmission lines, and other critical components within the power grid model. 110kV_Grid.pfd: The DIgSILENT PowerFactory project file containing the detailed 110 kV transmission system model. ldf_calc.py: A Python script for performing load flow (LDF) calculations using the DIgSILENT PowerFactory API. This script is designed to analyze normal operating conditions by simulating various load scenarios and extracting relevant electrical measurements. shc_calc.py: A Python script for performing short-circuit (SHC) calculations using the DIgSILENT PowerFactory API. This script automates the simulation of faults, including SPGF, 2PSC, and 3PSC, at varying positions along the transmission lines, and extracts voltage and current data for each scenario.

Power Engineering, Machine Learning, Power System Protection, Deep Learning

Creative Commons Attribution 4.0 International