Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Data-Driven Reduction of Fault Location Errors in Onshore Wind Farm Collectors

A. J. Alves Junior, M. J. B. B. Davi, R. A. S. Fernandes, M. Oleskovicz, D. V. Coury

TL;DR

This work tackles fault-location accuracy in onshore wind-farm collectors under increasing inverter-based resource penetration by augmenting a phasor-based baseline with a data-driven GRN-based error-correction layer (GRN-MM). It presents a comprehensive feature engineering strategy, a GRN-MM architecture, and a rigorous evaluation via Optuna-driven hyperparameter tuning and large-scale statistical validation on a PSCAD wind-farm model. The results show a substantial reduction in distance errors, with GRN-MM achieving an overall 76% improvement over state-of-the-art methods and demonstrating strong robustness across fault types and generation levels. The framework is demonstrated to be versatile (extendable to other fault locators) and compatible with existing measurement infrastructure, suggesting meaningful practical impact for faster restoration in modern wind-energy systems.

Abstract

Accurate fault location is essential for operational reliability and fast restoration in wind farm collector networks. However, the growing integration of inverter-based resources changes the current and voltage behavior during faults, challenging the effectiveness of traditional phasor-based diagnostic methods. In this context, the present paper introduces an advanced machine-learning solution that enhances a deterministic fault distance estimator by incorporating a correction model driven by a Gated Residual Network, specifically designed to minimize residual fault location errors. Through comprehensive feature engineering and selection processes, an improved predictor was developed and trained on a diverse set of fault scenarios simulated in a PSCAD-based real-world wind farm model, including variations in fault type, resistance, location, inception angle, and generation penetration. Hyperparameter optimization was performed using the Optuna framework, and the robustness of the method was statistically validated. Results show a significant improvement in accuracy, with a 76% overall decrease in fault location error compared to state-of-the-art approaches. The proposed method demonstrates strong scalability and adaptability to topological and operational changes. This approach advances the deployment of data-driven fault location frameworks for modern power systems.

Data-Driven Reduction of Fault Location Errors in Onshore Wind Farm Collectors

TL;DR

This work tackles fault-location accuracy in onshore wind-farm collectors under increasing inverter-based resource penetration by augmenting a phasor-based baseline with a data-driven GRN-based error-correction layer (GRN-MM). It presents a comprehensive feature engineering strategy, a GRN-MM architecture, and a rigorous evaluation via Optuna-driven hyperparameter tuning and large-scale statistical validation on a PSCAD wind-farm model. The results show a substantial reduction in distance errors, with GRN-MM achieving an overall 76% improvement over state-of-the-art methods and demonstrating strong robustness across fault types and generation levels. The framework is demonstrated to be versatile (extendable to other fault locators) and compatible with existing measurement infrastructure, suggesting meaningful practical impact for faster restoration in modern wind-energy systems.

Abstract

Accurate fault location is essential for operational reliability and fast restoration in wind farm collector networks. However, the growing integration of inverter-based resources changes the current and voltage behavior during faults, challenging the effectiveness of traditional phasor-based diagnostic methods. In this context, the present paper introduces an advanced machine-learning solution that enhances a deterministic fault distance estimator by incorporating a correction model driven by a Gated Residual Network, specifically designed to minimize residual fault location errors. Through comprehensive feature engineering and selection processes, an improved predictor was developed and trained on a diverse set of fault scenarios simulated in a PSCAD-based real-world wind farm model, including variations in fault type, resistance, location, inception angle, and generation penetration. Hyperparameter optimization was performed using the Optuna framework, and the robustness of the method was statistically validated. Results show a significant improvement in accuracy, with a 76% overall decrease in fault location error compared to state-of-the-art approaches. The proposed method demonstrates strong scalability and adaptability to topological and operational changes. This approach advances the deployment of data-driven fault location frameworks for modern power systems.

Paper Structure

This paper contains 22 sections, 5 equations, 10 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Data-Driven Approach for Fault Location Error Reduction
  3. Objective
  4. Strategy for Feature Engineering
  5. Feature Extraction
  6. Feature Selection
  7. Feature Preprocessing Strategy
  8. Gated Residual Network Architecture for Error Reduction
  9. Model Training Parameters
  10. Methodology for Assessing the GRN-MM Performance
  11. Hyperparameter Optimization Approach
  12. Statistics Evaluation Approach
  13. Test System
  14. Results for Assessing GRN-MM in Fault Location
  15. Results of the Feature Engineering
  16. ...and 7 more sections

Figures (10)

  • Figure 1: Proposed methodology for data-driven reduction of fault distance estimation errors.
  • Figure 2: Overview of the Feature Engineering Process.
  • Figure 3: Architecture of the GRN model for Error Reduction.
  • Figure 4: Test system parameters and single-line diagram.
  • Figure 5: Convergence of the objective function over 2,000 Optuna trials.
  • ...and 5 more figures