Source-linked AI summary
DER-Fault: A Simulation-Based Fault Dataset for DER-Integrated Distribution Systems Under Diverse Operating Conditions
Fathima Razeeya Mohamed Razick, Petr Musilek
TL;DR
Fault diagnostics in DER-integrated distribution systems need to be developed and evaluated across variable operating conditions, but diverse datasets are needed to represent that variability. DER-Fault is a simulation-based dataset built from representative loading conditions and PV, EV, and V2G scenarios. Its validation reports phase-specific fault responses and reduced voltage-phasor disturbance at higher fault resistance.
Problem
Fault-diagnostic measurements vary with operating conditions, while data-driven methods need sufficiently large, diverse datasets that represent different network operating and fault conditions.
Source: §2 · p. 2Method
DER-Fault simulates normal and faulted operation on the Iowa 240-bus network using representative annual load conditions, DER scenarios, and recorded pre- and post-fault phasors.
Source: §2 · p. 2, §4.3 · p. 5, §4.5 · p. 6Results
Median Dmax decreased from 57.3% for the 0.05–1 Ω range to 9.1% for the 15–25 Ω range as fault resistance increased.
Source: §4.6.4 · p. 9Takeaways & Limitations
The dataset provides a reusable resource for developing, evaluating, and comparing data-driven fault-diagnostic approaches under varying loading and DER conditions.
Source: §1 · p. 1, §2 · p. 2The dataset uses one simulated distribution-system model and steady-state pre- and post-fault phasors, so applying it to other networks or conditions requires care.
Source: §5 · p. 9
Abstract
from arXiv · showhide
The growth of distributed energy resource (DER) integration introduces greater variability in distribution system operating conditions, creating a need for comprehensive datasets for developing and evaluating data-driven fault diagnostic methods. This paper presents \textbf{DER-Fault}, a simulation-based dataset generated in OpenDSS using the Iowa 240-bus distribution system with photovoltaic (PV) generation, electric vehicle (EV) charging, and vehicle-to-grid (V2G) operation. To capture annual load variability while maintaining a manageable number of simulation cases, 60 representative operating conditions are selected from 8,760 hourly load profiles, covering peak, off-peak, transition, typical weekday, and typical weekend conditions. Six DER scenarios with varying PV and EV penetration levels and V2G participation are considered. The dataset includes normal operation and ten short-circuit fault classes, with varying fault resistance. Pre- and post-fault voltage and line-current phasor measurements are recorded at selected measurement locations, resulting in 19,800 simulation cases. Technical validation demonstrates broad coverage of annual loading and DER operating conditions, distinct phase-specific responses corresponding to the simulated fault types, and a consistent reduction in voltage-phasor disturbance with increasing fault resistance. The resulting dataset provides a structured resource for developing and evaluating machine-learning and other data-driven methods for fault detection, classification, and localization in DER-integrated distribution systems.
1 Value of the data
DER-Fault provides a reusable resource for fault-diagnostic research under changing distribution-system loading and DER conditions. Its varied measurements and scenarios support development, comparison, and robustness studies across data-driven methods.
1 Value of the data
- The dataset supports development and evaluation of fault detection, classification, and localization methods under changing loading and DER conditions.It includes multiple fault types, locations, and conditions, with pre- and post-fault voltage and current phasors and associated metadata.Source: §1 · p. 1
- PV generation, EV charging, and V2G variation supports investigation of how operating conditions affect fault characteristics and diagnostic performance.Source: §1 · p. 1
- The simulation resource supports method development, comparative studies, and reproducible research across protection, monitoring, DER integration, and data-driven fault analytics.Source: §1 · p. 2
2 Background
DER integration makes distribution-system operating conditions more variable, and fault measurements depend on both fault conditions and the network state. DER-Fault addresses the need for diverse datasets to develop and evaluate diagnostic approaches across those conditions.
2 Background
- Fault-induced voltage and current variations depend on both the fault and the operating condition, making evaluation across varied conditions important.Source: §2 · p. 2
- Data-driven diagnostic methods depend on sufficiently large, diverse datasets; limited loading or fixed DER penetration may not capture DER-system variability.Source: §2 · p. 2
- System operators face challenges in obtaining large, systematically labeled fault datasets because real events are uncommon, measurements can be limited, and controlled fault generation is impractical.Source: §2 · p. 2
- DER-Fault uses the Iowa 240-bus network to provide simulated normal and faulted cases with representative annual operating conditions, PV, EV, and V2G scenarios, and phasor measurements with metadata.The dataset is intended for development, evaluation, and comparison of data-driven fault-diagnostic approaches.Source: §2 · p. 2
3 Data Description
The dataset comprises simulation cases spanning loading, DER, and fault conditions, with electrical measurements linked to labels and operating metadata. Its measurements and labels are organized to identify each case, measurement location, phase, quantity, and operating state.
3 Data Description
- The dataset contains 60 representative operating conditions, six PV/EV/V2G scenarios, and 19,800 cases covering normal operation and ten fault classes.Pre- and post-fault voltage and line-current phasors are recorded at selected measurement locations.Source: §3 · p. 2
3.1 Measurement Data
- Each measurement row represents one simulation case identified by SampleID, with voltage features followed by line-current measurements.Voltage features include pre- and post-fault magnitude and phase angle; current magnitudes and angles are recorded by line terminal and phase in both states.Source: §3.1 · pp. 2–3
3.2 Labels and Metadata
- A linked labels and metadata file records each case’s operating condition, timestamp, DER scenario, and fault information across 11 classes.Label 0 denotes normal operation and labels 1–10 denote the ten simulated fault types; dataset specifications are summarized in Table 1.Source: §3 · p. 2, §3.2 · p. 3, Table 1 · p. 3
4 Experimental Design, Materials and Methods
The dataset-generation framework combines system modeling, measurement configuration, representative-load selection, DER scenario modeling, fault simulation, and validation. These stages are presented as the procedure for capturing fault behavior across loading and DER conditions.
4 Experimental Design, Materials and Methods
- The framework includes distribution-system modeling and measurement configuration to establish the simulated network and recorded measurements.Source: §4 · p. 3
- Representative load selection and DER scenario modeling are stages of the framework for capturing varied operating conditions.Source: §4 · p. 3
- Fault simulation and dataset validation complete the generation procedure.Source: §4 · p. 3
4.1 Distribution System and Simulation Setup
The Iowa 240-bus distribution system was modeled in OpenDSS, with annual load data used to set baseline conditions. Each simulation solved the network before and after fault application to produce voltage and line-current phasor measurements.
4.1 Distribution System and Simulation Setup
- The Iowa 240-bus system was implemented in OpenDSS using the original network model and annual load data [6].The network includes three-phase and single-phase connections.Source: §4.1 · p. 3
- Table 2 describes the dataset's labels and metadata fields.Source: Table 2 · p. 4
- Hourly bus-level active and reactive demand over one year established the baseline loading conditions.The annual profile contains 8,760 operating conditions.Source: §4.1 · p. 4
- Figure 1 presents the test system's single-line diagram.Source: Figure 1 · p. 4
- For each case, the specified operating condition was modified for its DER scenario, then the network was solved before and after fault application.The resulting quantities were used to construct the measurement dataset.Source: §4.1 · pp. 4–5
4.2 Measurement Configuration
The dataset uses 93 PMUs in a full-observability configuration, with voltage and monitored-line current phasors collected before and after faults.
4.2 Measurement Configuration
- A total of 93 PMUs defines the study's full-observability measurement configuration, using a one-hop criterion for bus observability.A bus is observable when a PMU is at that bus or a directly connected neighbor.Source: §4.2 · p. 5
- At each PMU-equipped bus, voltage magnitude and phase angle were recorded for available phases under pre-fault and post-fault conditions.Source: §4.2 · p. 5
- Current magnitude and phase angle were recorded for each monitored line phase under pre-fault and post-fault conditions.The measurement functions retrieve voltage magnitude and angle at selected buses and current magnitude and angle for monitored line phases.Source: §4.2 · p. 5
4.3 Load Selection
Sixty actual hourly conditions were selected from the annual load profiles to represent variability while limiting simulation cases, retaining each timestamp's bus-level demand values.
4.3 Load Selection
- 60 representative operating conditions were selected from the 8,760 hourly conditions in the annual load profiles.Five actual hourly conditions were retained from each month: peak, off-peak, transition, typical weekday, and typical weekend.Source: §4.3 · p. 5
- Monthly peak and off-peak conditions use maximum and minimum feeder demand; transition selects the largest remaining one-hour demand change.Typical weekday and weekend points are closest to their respective monthly mean feeder demands.Source: §4.3 · p. 5
- Each category contributes 12 conditions, and selected timestamps retain original bus-level active and reactive demands.This preserves the spatial loading distribution and active–reactive power pairing.Source: §4.3 · p. 5
4.4 DER Scenario Modeling
Six DER scenarios vary PV and EV penetration and V2G participation. The model assigns DER capacity and demand using feeder or household assumptions, time profiles, and bus-level net-load calculations.
4.4 DER Scenario Modeling
- Six scenarios vary PV penetration, EV penetration, and V2G participation, with the first scenario representing the network without DERs.The remaining scenarios combine PV generation, EV charging, and V2G operation in different ways.Source: §4.4 · p. 5, Table 3 · p. 6
- PV penetration is defined relative to peak feeder active demand, and PV capacity is allocated across load buses in proportion to average active demand.The peak demand reference is the maximum total feeder demand among the selected operating conditions.Source: §4.4 · p. 5
- EV penetration is the share of 1,120 households equipped with one EV each, with rated charging power of 7.2 kW per EV.EV population is distributed among buses in proportion to average active-power demand.Source: §4.4 · p. 6
- V2G participation is the share of EVs able to supply power to the network, modeled at 5 kW discharge per participating EV.The V2G injection calculation uses the EV population and a normalized discharge profile.Source: §4.4 · p. 6
- Normalized 24-hour profiles scale PV generation, EV charging, and V2G injection at each selected operating condition.Source: §4.4 · p. 6
- For each condition, baseline bus-level demands are combined with PV generation, EV charging demand, and V2G injection to calculate net active power.Reactive power is determined using specified power factors, and resulting active and reactive values become the OpenDSS net bus loads.Source: §4.4 · p. 6
4.5 Fault Generation
Fault generation covers normal operation and ten short-circuit fault classes, with fault connections matched to each class and resistance varied across cases.
4.5 Fault Generation
- The dataset labels normal operation and ten fault classes, spanning single-line, line-to-line, double-line-to-ground, and three-phase-to-ground faults.Normal operation is label 0; labels 1–10 represent the fault classes.Source: §4.5 · p. 6
- Fault connections match each class: individual phases, phase pairs, phase pairs-to-ground, or all phases-to-ground.Fault resistance is independently sampled uniformly from 0.05 to 25 Ω for each faulted case.Source: §4.5 · p. 6
- Pre-fault voltage and current measurements precede fault application and post-fault acquisition across 60 loading conditions and six DER scenarios, yielding 19,800 cases.Five samples are generated for each of the 11 classes.Source: §4.5 · p. 6
4.6 Technical Validation of Dataset
The validation examines annual operating-condition coverage, DER power variation, phase-specific fault responses, and fault-resistance sensitivity. Overall, the results support technical consistency across the considered operating and fault conditions.
4.6.1 Operating condition coverage
- The 60 selected conditions span low, high, and intermediate loading levels across the annual profile, covering extremes and commonly occurring or transitional states.They preserve a broad range of annual loading while substantially reducing the operating points needed for fault simulation.Source: §4.6.1 · p. 7, Figure 2 · p. 7
4.6.2 DER scenario validation
- DER power varies across representative conditions; PV generation ranges from zero to 2470.7 kW in S4 and S5, while EV demand ranges from 1411.2 to 4032.0 kW in S6.The variation reflects time-dependent PV, EV charging, and V2G profiles rather than fixed DER power values.Source: §4.6.2 · pp. 7–8, Table 4 · p. 7
4.6.3 Phase-specific fault-characteristic validation
- Fault responses are phase-dependent: single-line faults predominantly affect their faulted phase, while line-to-line and double-line faults chiefly affect their two faulted phases.For LLLG, median phasor changes are 13.6%, 13.2%, and 13.4% for phases A, B, and C, respectively; the remaining phase changes are substantially smaller for two-phase faults.Source: §4.6.3 · p. 8, Figure 3 · p. 8
4.6.4 Fault-resistance sensitivity
- Median Dmax fell from 57.3% at 0.05–1 Ω to 9.1% at 15–25 Ω as fault resistance increased.Variation within resistance ranges also reflects fault type, location, loading condition, and DER scenario; the trend is consistent with expected disturbance reduction [11].Source: Figure 4 · p. 9, §4.6.4 · p. 9
4.6 Technical Validation of Dataset
- The validation finds that the dataset captures operating-condition and DER variation while preserving expected phase-specific fault characteristics and resistance effects.The authors describe these results as supporting technical consistency across the considered operating and fault conditions.Source: §4.6 · p. 7, §4.6.4 · p. 9
5 Limitations
DER-Fault is simulation-based on a single distribution-system model and has defined modeling and measurement boundaries.
5 Limitations
- The dataset uses one simulated distribution-system model, predefined DER profiles, and steady-state pre- and post-fault phasors, not high-resolution transient waveforms.It omits detailed inverter dynamics, battery state-of-charge behavior, and control strategies; simulations cover only the fault types and resistance range studied, so applications to other networks or conditions require care.Source: §5 · p. 9