الإصدار 6 - العدد 18 (2026-09-20)

Deep Learning for Electricity Load Consumption Forecasting in Jordan & Palestine

1- Dr: Bayan Assaf

2- Senior Lecturer : Mahmoud Khaseeb

1- PhD Candidate in Economic Sciences (ESCT) - University of Manouba - Tunisia; Manager of Studies and Research Department Ministry of Labor-Palestine

2-Senior Lecturer of Economics -Birzeit University-Palestine

This study dives into how advanced tools, like Long Short-Term Memory (LSTM) and Convolutional Neural Networks (CNNs), can forecast electricity use in Jordan and Palestine. We worked with data from homes, shops, and factories across cities and rural areas, tackling issues like scorching summers and shaky power grids. Our predictions were off by just 4% to 13%, with LSTM models beating older methods hands-down. By factoring in weather and time patterns, we slashed errors from 10.8% to under 5.5% for households. In one busy commercial zone, we even hit a 3.9% error rate. These results show how smart tech can help plan energy use, boosting Jordan’s solar and wind projects and easing Palestine’s power struggles.

To make these forecasts even more useful, we paired our models with a smart system to guide energy-saving habits. Picture getting a text nudging you to run your AC less during peak hours—this is what our Online Energy Reporting System (OERS) aims to do. By sorting users into groups based on their power habits, we can offer tailored tips, like shifting heavy appliance use to off-peak times. This approach not only reduces costs for families and businesses but also alleviates the strain on Jordan and Palestine’s grids, paving the way for the smoother integration of renewable energy sources, such as solar and wind.