Journal of Engineering Research
Innovation and Scientific Development

Mid-Term Electrical Load Forecasting Using Convolutional Neural Networks (CNNS)

Document Type : Research Paper

Authors
1Elijah Adebayo Olajuyin, 
  1. 1  Bamidele Olumilua University of Education, Science and Technology,Ikere-Ekiti, Ekiti State. Electrical and Electronic Engineering Department, School of Engineering.
Abstract

This research presents a novel approach to mid-term electrical load forecasting using Convolutional Neural Networks (CNN) for the Ado-Ekiti 11kV Distribution Network in Nigeria. The study utilizes five years of historical load data (2020-2024) from five feeders to develop and validate a CNN-based forecasting model. The research methodology incorporates comprehensive data preprocessing, exploratory data analysis, and hyperparameter tuning to optimize model performance. The hyperparameter-tuned model achieved exceptional accuracy with an R² value of 0.9420, RMSE of 0.0490, and MAPE of 10.0535%, demonstrating significant improvement over baseline models. The model successfully generated one-year- ahead load forecasts for 2025, revealing important trends and patterns in load distribution across the network. Strong correlations were identified among feeders (correlation coefficients 0.75-0.92), indicating synchronized load behaviour. The findings provide valuable insights for distribution network planning, infrastructure optimization, and load management strategies. This research contributes to the growing body of knowledge on deep learning applications in power distribution systems and offers practical recommendations for improving network operations.

Graphic Abstract
JERISD PUBLICATION LOGO
Vol 3, Number 1
February 2025
Pages 14-22
Files
Download: PDF
History
  • Received: 12/01/2025

  • Revised: 17/02/2025

  • Accepted: 22/02/2025

  • Published: 27/02/2025
Statistics