Improving Predictive Reliability and Automation of Smart Grids Using the StarNet Ensemble Model ⚡

Enhancing Predictive Reliability and Automation for Smart Grids with Machine Learning

---

📝 Overview

StarNet Ensemble introduces a framework featuring an integrated machine-learning-driven GUI for predicting smart grid stability. This system leverages a stacking-based ensemble learning approach to ensure reliable and effective electricity distribution in smart grids.

---

🧑‍🔬 Abstract

The increasing complexity of modern power networks demands advanced predictive models to ensure grid stability and autonomous operation. The purpose of this research is to improve the predictive reliability and automation of smart grids through the development of the StarNet Ensemble Model, a stacking-based machine learning framework designed to enhance stability forecasting and decision-making in dynamic energy environments. The proposed method integrates multiple base learners in a hierarchical ensemble structure, enabling robust analysis and comparison across predictive models. A synthetic dataset was generated from a 4-node star network and extended to include consumer node variations, comprising 60,000 data points with 12 predictive features and two dependent variables representing system stability states. Parameters such as reaction times, nominal power values, and price elasticity coefficients were analyzed to train and validate the model. Experimental evaluation demonstrates that the StarNet Ensemble Model achieved a prediction accuracy of 99.43%, significantly outperforming conventional baseline models. The model enables real-time stability prediction and supports automated decision- making for load management, demand response, and fault prevention. The findings reveal that the StarNet Ensemble Model provides a scalable and intelligent solution for predictive maintenance and autonomous control in smart grids. Its high accuracy and automation capability contribute to enhanced grid resilience, reduced operational costs, and improved reliability of future energy systems.

---

📁 Repository Structure

Dataset/    # Synthetic smart grid datasets
Code/       # Source code, notebooks, and web-based GUI

---

🚀 Getting Started

1. Clone the repository

   git clone https://github.com/your-username/StarNet-Ensemble.git
   cd StarNet-Ensemble

2. Install dependencies

   # Example for Python projects
   pip install -r Code/requirements.txt

3. Run the provided Jupyter notebook

   • Open Code/smart-grid-research-paper.ipynb in Jupyter Notebook, JupyterLab, or VS Code.
   • Update the dataset path in the notebook to point to your local copy of the dataset (e.g., replace /kaggle/input/smart-grid-stability/smart_grid_stability_augmented.csv with the correct path).
   • Run all cells sequentially to reproduce the results and figures.

   # Example: launch Jupyter in the Code directory
   cd Code
   jupyter notebook

---

📊 Dataset Description

• Source: Synthetic, generated from a 4-node star network with consumer node variations.
• Size: 60,000 samples
• Features: 12 main predictive features (e.g., reaction times, nominal power, price elasticity)
• Targets: 2 dependent variables (system stability status)
• Purpose: Supports predictive modeling and evaluation within the web-based framework.

---

🏆 Results Summary

• StarNet stacking ensemble model achieved:
  • Accuracy: 99.43%
  • Robustness: Outperformed traditional and advanced ML models
• Key Advantages:
  • Real-time, web-based monitoring and prediction
  • Automated grid services (load shedding, demand response)
  • Predictive maintenance and reduced operational costs
  • Interactive GUI for stakeholders

---

👥 Authors

Name	Affiliation	Email
Amit Chhabra†‡	CSE, Chandigarh College of Engineering and Technology, Chandigarh	amitchhabra@ccet.ac.in
Sunil K. Singh‡	CSE, Chandigarh College of Engineering and Technology, Chandigarh	sksingh@ccet.ac.in
Sudhakar Kumar‡	CSE, Chandigarh College of Engineering and Technology, Chandigarh	sudhakar@ccet.ac.in
Manraj Singh‡	CSE, Chandigarh College of Engineering and Technology, Chandigarh	mannmanraj239@gmail.com
Utkarsh Chauhan‡	CSE, Chandigarh College of Engineering and Technology, Chandigarh	co21364@ccet.ac.in
Saksham Arora‡	CSE, Chandigarh College of Engineering and Technology, Chandigarh	mco22390@ccet.ac.in
Brij B. Gupta*	Dept. of Computer Science & Info. Engg., Asia University, Taiwan	bbgupta@asia.eu.tw
Wadee Alhalabi	Dept. of Computer Science, King Abdulaziz University, Saudi Arabia	wsalhalabi@kau.edu.sa
Varsha Arya	Dept. of Business Administration, Asia University, Taiwan	varshaarya2108@gmail.com
Bassma Saleh Alsulami	Dept. of Computer Science, King Abdulaziz University, Saudi Arabia	balsulami@kau.edu.sa
Ching-Hsien Hsu	Dept. of Computer Science & Info. Engg., Asia University, Taiwan	robertchh@asia.edu.tw
† Corresponding author ‡ Equal contribution * Principal Investigator

---

📚 Citation

Note: This paper is currently unpublished and subject to change.
Please contact the authors before citing or referencing this work.

---

📬 Contact

For questions or collaboration, please contact the corresponding author:
Dr. Sudhakar Kumar – sudhakar@ccet.ac.in

---

📝 License

This project is licensed under the MIT License.
See LICENSE for details.

---