top of page
Search

🔍 ANN-Based Fault Detection, Classification, and Location in IEEE 13-Node Test Feeder Using MATLAB

IntroductionIn power distribution systems, fast and accurate fault identification is essential for reliability and protection. This blog explores the application of Artificial Neural Networks (ANN) to detect, classify, and locate faults in the IEEE 13-node test feeder, using MATLAB and Simulink.

⚙️ Overview of the IEEE 13-Node Test Feeder

The IEEE 13-node test feeder is a standard benchmark model used for analyzing power distribution system behavior under fault conditions. It includes:

  • 12 bus points (numbered 632, 633, ..., up to 680)

  • Balanced and unbalanced load scenarios

  • Simulation environment available in MATLAB’s demo models

The simulation setup involves injecting faults at different nodes to generate training data for the ANN model.

💥 Fault Creation and Data Collection

Various types of faults are simulated at different nodes, including:

  • Single Line-to-Ground (SLG) Fault

  • Line-to-Line Fault

  • Double Line-to-Ground Fault

For each fault scenario at each node:

  • Phase voltages and currents (ABC) are collected

  • The data is labeled in binary format to form the target vector for supervised learning

  • All datasets are stored in .mat files for future training use

Each fault scenario is modeled separately, and data is stored in a matrix structure containing inputs (features) and targets (labels).

🧠 Training the Artificial Neural Network (ANN)

Once the data is collected from all nodes and fault scenarios, it is compiled to create:

  • Input dataset – representing measurements (voltage/current signals)

  • Target dataset – representing fault classification (fault type or fault location)

Using MATLAB’s Neural Network Toolbox (nntool), the training process includes:

  1. Loading the input and target datasets

  2. Selecting the ANN architecture and training options

  3. Running the training algorithm

  4. Viewing performance plots (training, validation, testing) and regression graphs

Separate ANN models are trained for:

  • Fault classification (determining the type of fault)

  • Fault location (identifying the faulted bus/node)

🧪 Exporting and Implementing Trained ANN Models

After training:

  • The ANN models are exported to Simulink using the toolbox export feature

  • These models are integrated with the test feeder simulation for real-time detection and classification

The implementation includes:

  • Feeding real-time system input (ABC data) into the trained ANN

  • Using one model for fault detection/classification

  • Using another model for fault location identification

📍 Fault Detection, Classification, and Location Results

The trained models are tested by placing faults at different locations. For each test:

  • The fault is detected (whether a fault exists)

  • The type of fault is classified (e.g., BCG, AB, AG)

  • The location is identified, typically close to the actual faulted bus (e.g., near Bus 671 or 675)

Multiple test cases are shown where the ANN model accurately identifies fault type and location.

✅ Conclusion

This ANN-based approach effectively detects, classifies, and locates faults in the IEEE 13-node feeder using MATLAB and Simulink. The results demonstrate the reliability and accuracy of ANN in smart grid and fault analysis applications.

Key Benefits:

  • High accuracy in fault classification and location

  • Real-time integration with Simulink

  • Data-driven approach using supervised learning

Comments

bottom of page