[Paper Review] Real-time Fault Localization in Power Grids With Convolutional Neural Networks.

Wenting Li, Deepjyoti Deka+2 more|arXiv (Cornell University)|Oct 11, 2018

Power Systems Fault Detection22 references4 citations

TL;DR

This paper proposes a data-driven fault localization method in power grids using a Convolutional Neural Network (CNN) that leverages physically interpretable bus voltage features. Despite low observability (7% of buses), the CNN achieves high-accuracy fault location within a small neighborhood, outperforming existing machine learning methods and enabling real-time operation under complex fault conditions.

ABSTRACT

Diverse fault types, fast re-closures and complicated transient states after a fault event make real-time fault location in power grids challenging. Existing localization techniques in this area rely on simplistic assumptions, such as static loads, or require much higher sampling rates or total measurement availability. This paper proposes a data-driven localization method based on a Convolutional Neural Network (CNN) classifier using bus voltages. Unlike prior data-driven methods, the proposed classifier is based on features with physical interpretations that are described in details. The accuracy of our CNN based localization tool is demonstrably superior to other machine learning classifiers in the literature. To further improve the location performance, a novel phasor measurement units (PMU) placement strategy is proposed and validated against other methods. A significant aspect of our methodology is that under very low observability (7% of buses), the algorithm is still able to localize the faulted line to a small neighborhood with high probability. The performance of our scheme is validated through simulations of faults of various types in the IEEE 68-bus power system under varying load conditions, system observability and measurement quality.

Motivation & Objective

• Address the challenge of real-time fault localization in power grids under complex fault dynamics and low measurement availability.
• Overcome limitations of existing methods that rely on static load assumptions or require high sampling rates and full observability.
• Develop a data-driven approach with physically interpretable features to improve model transparency and performance.
• Propose a novel PMU placement strategy to enhance fault location accuracy and system observability.
• Demonstrate robust fault localization performance under varying load conditions, measurement quality, and system observability levels.

Proposed method

• The method employs a Convolutional Neural Network (CNN) classifier trained on bus voltage measurements to detect and localize faults in real time.
• Features used in the CNN are derived from voltage magnitudes and angles at buses, selected for their physical interpretability in power system fault analysis.
• A novel PMU placement strategy is introduced to maximize fault localization accuracy under limited measurement availability.
• The approach is validated through extensive simulations on the IEEE 68-bus system under diverse fault types, load conditions, and measurement quality levels.
• The CNN model is trained and tested under low observability (7% of buses equipped with PMUs), simulating realistic deployment constraints.
• The method leverages temporal patterns in voltage signals to capture transient fault dynamics, enabling real-time localization.

Experimental results

Research questions

• RQ1 Can a CNN-based classifier using physically interpretable voltage features achieve superior fault localization accuracy compared to existing machine learning methods in power grids?
• RQ2 How does the proposed method perform under low system observability (e.g., 7% of buses equipped with PMUs) and varying measurement quality?
• RQ3 To what extent does the proposed PMU placement strategy improve fault localization performance compared to conventional or random placement methods?
• RQ4 Can the method localize faults accurately across diverse fault types and load conditions without relying on static load assumptions?
• RQ5 What is the real-time feasibility and robustness of the CNN-based localization framework under transient fault dynamics?

Key findings

• The proposed CNN-based fault localization method achieves significantly higher accuracy than other machine learning classifiers reported in the literature.
• Even with only 7% of buses equipped with PMUs, the method localizes the faulted line to a small neighborhood with high probability.
• The use of physically interpretable features enhances model transparency and improves generalization across varying system conditions.
• The novel PMU placement strategy outperforms existing methods in terms of fault localization accuracy and observability efficiency.
• The method demonstrates robust performance across various fault types and under different load conditions in IEEE 68-bus system simulations.
• The approach maintains real-time capability despite complex transient states and low sampling rates, making it suitable for practical deployment.