top of page
Search

Fault Classification, Location, and Detection in Power Systems using Neural Networks

Fault Classification, Location, and Detection in Power Systems using Neural Networks

Introduction

We explore the intricacies of fault detection, classification, and location using neural networks in a three-bus power system setup. Neural networks offer advanced capabilities to accurately identify faults and pinpoint their locations, enhancing the reliability and efficiency of power distribution systems.

System Overview The power system comprises three buses:

• Bus 1 and Bus 2: Source feeders

• Bus 3: Acts as a load mass Between Bus 1 and Bus 3 lies a transmission line where faults can occur, impacting system operations.

Fault Detection and Classification Models Different fault scenarios are simulated using specialized models:

• AG Ground Fault: Fault between Line A and Ground

• AB Double Line-to-Ground Fault: Fault between Line A and Line B

• BC Line-to-Line Fault: Fault between Line B and Line C

• ABC Triple Line Fault: Fault affecting all three lines

Data Collection for Neural Network Training Data collection is crucial for training the neural network models:

• Input Data: RMS voltage and current of Bus 1, zero sequence voltage and current from Bus 1 during fault conditions.

• Target Data: Binary classification data indicating fault presence (1) or absence (0) and fault type classification.

Training the Neural Network Using MATLAB's Neural Network Toolbox:

1. Data collected during fault conditions is used to train the neural network.

2. The neural network is optimized to match collected data with high accuracy (e.g., R-squared value close to 1).

Model Deployment in Simulink The trained neural network models are deployed in Simulink for real-time fault detection, classification, and location finding:

• Fault Detection Model: Outputs 1 if a fault is detected, 0 otherwise.

• Classification Model: Outputs a binary sequence identifying the type of fault (e.g., AG Ground Fault).

• Location Finding Model: Determines the exact kilometer location of the fault along the transmission line.

Simulation and Results Simulations are conducted using different fault scenarios and line lengths:

• Results show the neural network effectively detects faults (changes output from 0 to 1) and classifies them accurately based on predefined fault types.

• The location finding model provides precise fault locations, aiding in swift corrective actions.

Conclusion Utilizing neural networks for fault detection, classification, and location in power systems enhances operational reliability and minimizes downtime. The integration of advanced AI techniques ensures proactive fault management, crucial for maintaining efficient power distribution networks.