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

TL;DR

This paper presents a machine learning-based method that significantly improves fault location accuracy in wind farm collector networks by incorporating a correction model and extensive simulation validation.

Contribution

It introduces a novel Gated Residual Network-based correction model to enhance deterministic fault distance estimators in wind farms.

Findings

76% reduction in fault location error

Robustness validated across diverse fault scenarios

Scalable and adaptable to system changes

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.