SMART GRID OPTIMIZATION FOR TELECOMMUTING USING MACHINE LEARNING AND STATISTICAL METHODS: A POWER SYSTEMS ENGINEERING PERSPECTIVE

By

YABANI GELWASA GALADIM¹   and PEACE IFEOLUWA IKUFORIJ²

¹Department of Elect/Elect Engineering Abdullahi Fodio University of Science and Technology, Aliero, Nigeria.

²Department of Mathematics, Abdullahi Fodio University of Science and Technology, Aliero, Nigeria.

CORRESSPONDING AUTHOR

Yabani Gelwasa Galadim¹

Department of Elect/Elect Engineering

Abdullahi Fodio University of Science and Technology,

Aliero, Nigeria.

yabani.galadima@ksusta.edu.ng

https://orcid.org/0009-0006-0525-4044

ABSTRACT

The emergence of telecommuting has redefined global work dynamics, demanding resilient, reliable, and efficient energy infrastructures. Smart grid technologies, enhanced by machine learning (ML) and statistical methods, provide transformative solutions for optimizing power delivery, demand response, and sustainability in the context of remote work. This paper presents a comprehensive engineering perspective on smart grid optimization for telecommuting, using Abdullahi Fodio University of Science and Technology, Aliero, as a reference case. The study explores load forecasting, renewable energy integration, voltage stability, and demand-side management through hybrid ML-statistical models. Limited mathematical formulations are used to illustrate optimization frameworks. Results indicate that integrating ML and statistical approaches enhances forecasting accuracy, reduces energy wastage, and ensures sustainable power for telecommuting environments. Contributions to knowledge include a hybrid optimization model for smart grid resilience, a statistical-ML synergy framework for demand-side response, and context-specific applications to Nigerian power networks.

Keywords: Smart Grid, Telecommuting, Machine Learning, Statistical Methods, Power Systems Optimization, Demand Response.

Introduction

The COVID-19 pandemic accelerated the adoption of telecommuting globally, reshaping electricity consumption patterns and intensifying the demand for resilient smart grids[1], [2]. As employee’s transition from centralized offices to distributed home-based workspaces, energy consumption becomes more unpredictable, placing new challenges on conventional grids [3], [4]. Smart grids, characterized by bidirectional communication, advanced metering infrastructure (AMI), and intelligent control, have emerged as the backbone of modern telecommuting-based societies [5], [6]

DOI: https://doi.org/10.5281/zenodo.17675932

EXPLORING THE APPLICATION OF STATISTICAL LEARNING THEORY TO VOLTAGE STABILITY ANALYSIS ON THE BIRNIN KEBBI 330 KV POWER TRANSMISSION NETWORK: A COMPREHENSIVE REVIEW

BY

YABANI GELWASA GALADIM¹ and PEACE IFEOLUWA IKUFORIJ²

¹Department of Elect/Elect Engineering Abdullahi Fodio University of Science and Technology,Aliero, Nigeria.

²Department of Mathematics, Abdullahi Fodio University of Science and Technology, Aliero, Nigeria

CORRESSPONDING AUTHOR

Yabani Gelwasa Galadim¹

Department of Elect/Elect Engineering

Abdullahi Fodio University of Science and Technology,

Aliero, Nigeria.

yabani.galadima@ksusta.edu.ng

https://orcid.org/0009-0006-0525-4044

ABSTRACT

Voltage instability remains a persistent challenge in Nigeria’s high-voltage grid, including the Birnin Kebbi 330 kV corridor. Aging infrastructure, increasing load, and limited high-resolution telemetry hinder conventional analysis. This review explores how Statistical Learning Theory (SLT), with key concepts like empirical risk minimization, structural risk minimization, margin maximization, and capacity control, can enable robust voltage stability analysis amid data scarcity and noise. We relate SLT ideas to practical tasks such as regression of stability indices, binary and ordinal classification of stability states, and early-warning forecasting. The review covers model families (SVMs, kernel methods, regularized regression), feature extraction from SCADA-like signals, handling class imbalance, and methods for uncertainty quantification. It outlines evaluation protocols, using MSE as the main regression metric, proposes a Birnin Kebbi–specific workflow, and addresses open challenges like domain shift, missing data, and causal confounding. The conclusion presents a roadmap that combines SLT principles with modern techniques (e.g., GAN-generated data) to produce interpretable, capacity-controlled, offline monitoring suitable for under-instrumented networks.

Keywords: Voltage Stability; Statistical Learning Theory; PAC-Bayes & Rademacher Complexity; Data-Driven Assessment; Nigerian 330 kV Grid

I Introduction

Voltage stability is a critical aspect of power system reliability, ensuring that all bus voltages remain within acceptable limits during normal fluctuations and system disturbances. A loss of voltage stability often appears as a gradual voltage decline, which can lead to a voltage collapse and widespread blackouts [1], [2]. These events are not only technically disruptive but also have serious socio-economic impacts, as evidenced by notable historical cases like the 2003 North American blackout and the 2012 India grid failure[1]. In developing countries such as Nigeria, voltage instability is more frequent due to aging infrastructure, insufficient reactive power compensation, poor maintenance, and the increasing complexity of load demands[3], [4]. The Birnin Kebbi 330 kV transmission network, a crucial hub in Nigeria’s

DOI: https://doi.org/10.5281/zenodo.17675926

CYBERSECURITYAWARENESSAMONG COLLEGE STUDENTS:A CHATBOT BASED ENHANCEMENT APPROACH

By

ZAKIYA KOUSER

5^th Semester BCA Student, Department of Computer Applications, BET Sadathunnisa Degree College Bangalore, Karnataka, India.

Abstract

Cybersecurity threats are rapidly increasing as college students rely heavily on digital platforms for academics, communication, and social interaction. Despite being digitally active, students often demonstrate unsafe practices such as password reuse, lack of two- factor authentication, delayed updates, and insecure social media behavior, leaving them vulnerable to phishing, identity theft, and data breaches. Traditional awareness programs, including seminars and posters, provide only short-term impact and fail to ensure continuous engagement. This research investigates the current level of cybersecurity awareness among college students through a survey-based study and highlights critical gaps in their knowledge and practices. To address these challenges, a chatbot-based enhancement approach is proposed. The chatbot delivers real-time cybersecurity tips, interactive quizzes, phishing simulations, and curated resources through platforms such as Telegram or WhatsApp, ensuring accessibility and ongoing learning. Compared to conventional awareness methods, the chatbot provides a more interactive, cost-effective, and scalable solution to foster secure online behavior. This study contributes a practical framework for higher education institutions to strengthen cybersecurity culture, with future scope for expansion into mobile applications and gamified learning systems.

Index Terms - Cybersecurity Awareness, College Students, Chatbot, Phishing, Information Security, Online Safety, Cyber Threats, Gamification.

   INTRODUCTION

Cybersecurity has become one of the most critical concerns in the digital era, as individuals and institutions increasingly depend on online platforms for communication, education, and information sharing. College students, in particular, represent a highly active group of internet users who engage daily in academic activities, social networking, online banking, and e-learning platforms. However, their frequent use of digital technologies often comes with poor security practices, such as weak or reused passwords, oversharing on social media, and neglecting security updates. These behaviors significantly increase their vulnerability to threats

DOI: https://doi.org/10.5281/zenodo.17675910

ASSESSING THE IMPACT OF HISTORICAL AND SYNTHETIC DATA ON DEEP LEARNING-BASED VOLTAGE STABILITY MONITORING IN SCARCE DATA ENVIRONMENTS: A COMPARATIVE REVIEW

By

YABANI GELWASA GALADIM¹ and PEACE IFEOLUWA IKUFORIJ²

¹Department of Elect/Elect Engineering Abdullahi Fodio university of Science and Technology, Aliero, Nigeria.

² Department of Mathematics Abdullahi Fodio University of Science and Technology, Aliero, Nigeria

CORRESSPONDING AUTHOR

Yabani Gelwasa Galadim¹

Department of Elect/Elect Engineering

Abdullahi Fodio University of Science and Technology,

Aliero, Nigeria.

yabani.galadima@ksusta.edu.ng

https://orcid.org/0009-0006-0525-4044

ABSTRACT

Voltage stability monitoring is essential for securing power system operation. In many grids, particularly in developing or remote regions, real-time measurement streams are rare, unreliable, or unavailable. Deep learning (DL) models carry major potential for predictive voltage stability assessment under these constraints, provided that data acquisition and preparation strategies are carefully designed. This paper reviews recent advances in DL-based voltage stability monitoring that function effectively without high-resolution real-time data. We discuss data acquisition pipelines drawing from historical operational logs, simulation-based datasets, and generative synthetic data. Techniques such as transfer learning, domain adaptation, physics-informed modeling, and data augmentation methods are assessed for their potential to compensate for limited live data. Voltage stability indices and margins are formalized, training objectives are defined, and evaluation protocols are summarized. Comparative workflows and a novel hybrid data acquisition pipeline are presented. This review examines critical challenges of label scarcity, model generalization, and clarity, and recommends hybrid dataset approaches for data-limited scenarios.

Keywords: Voltage stability, deep learning, scarce data, digital twins, domain adaptation, synthetic data generation.

INTRODUCTION

To ensure power system reliability, accurate and timely monitoring is vital to prevent widespread outages triggered by voltage instability[1], [2], [3]. While many modern grids leverage continuous sensor streams, in numerous contexts rural systems, developing nations, isolated micro grids real-time data is scarce due to cost, communication limits, or infrastructure gaps[4], [5], [6]. Conventional model-based methods such as continuation power flow (CPF) and static voltage stability indices (VSIs) remain useful but sometimes fail to capture nonlinear dynamics and

DOI: https://doi.org/10.5281/zenodo.17675924

DESIGN AND IMPLEMENTATION OF SMART DUSTBIN FOR AUTOMATED WET AND DRY WASTE SEGREGATION

BY

Sujan Banerjee^1*; Soumen Ghosh²; Aritra Bera³; Riju Das⁴; Sourav Pal⁵

^1,2,3,4,5Department of Electrical Engineering, Swami Vivekananda School of Diploma, Durgapur, India

CORRESPONDING AUTHOR

SUJAN BANERJEE

Department of Electrical Engineering,

Swami Vivekananda School of Diploma,

Durgapur, India

ABSTRACT

The increasing rate of urbanization has significantly contributed to the generation of mixed waste, posing serious environmental and health challenges. Manual segregation is inefficient, labor-intensive, and often hazardous. This paper presents the design and development of a smart dustbin system that automatically classifies waste into wet and dry categories using embedded sensors and microcontroller technology. The proposed system utilizes an Arduino UNO, ultrasonic sensor, soil moisture sensor, servo motor, and relay module to detect waste type and direct it to the appropriate compartment. The implementation aims to assist in effective waste management and promote a cleaner environment through automation.

Keywords: Smart Dustbin, Waste Segregation, Arduino UNO, Wet and Dry Waste, Soil Moisture Sensor, Automation.

               INTRODUCTION

Effective waste management is an increasingly critical issue, particularly in urban and semi-urban areas where rapid population growth and industrialization have led to significant waste generation. Improper disposal practices and the mishandling of waste materials contribute to several problems, including environmental pollution, health hazards, and reduced efficiency in recycling efforts. Conventional techniques for segregating waste typically rely heavily on manual labor, which can result in inaccuracies, delays, and higher operational costs.

To address these challenges, the integration of automation into waste management processes has emerged as a promising solution. Automated systems reduce human dependency, enhance accuracy, and ensure faster processing of waste materials. This project presents a smart dustbin prototype designed to automatically detect and classify waste based on its moisture level. The system employs a combination of sensors and an Arduino-based microcontroller to distinguish between wet and dry waste. By automating the segregation process, the system not only improves efficiency but also provides a cost-

DOI: https://doi.org/10.5281/zenodo.17675918