← Back to Publications List

Optimizing Energy Consumption Prediction Using Hybrid LightGBM and XGBoost: Integrating Heterogeneous Data for Smart Grid Management

Students & Supervisors

Student Authors
Hasanur Zaman Anonto
Bachelor of Science in Electrical & Electronic Engineering, FE
Md Ismail Hossain
Bachelor of Science in Electrical & Electronic Engineering, FE
Morzia Momo
Bachelor of Science in Computer Science & Engineering, FST
Supervisors
Abu Shufian
Lecturer, Faculty, FE

Abstract

The purpose of this study is to investigate energy consumption trends, as well as how environmental indicators of energy consumption, including temperature, humidity, and renewables impact energy consumption in a specified interval (15 January 2025 to 12 February 2025). This involves looking at time-series data, such as spikes January 2: 110 MWh. February 12: 120 MWh and typical usages 45 MWh to 60 MWh. Smoothened data suggests lockdown stability of energy; hourly data shows a steep increase. Temperature dependency is further explored in the study, where it is also shown that energy consumption is also dependent on temperature, where energy use at 20°C is 240 MWh and increase to be 360 MWh at 30°C on an annualized basis, further confirming the shaped dependency with increased cooling demands during warmer months. These boxplots show distribution in temperature, renewable energy, and energy consumption across the days of the week where Tuesday has the most waste energy consumption at 85 MWh while the least occurred on Sunday at 50 MWh. Furthermore, the cross-correlation analysis has shown that there is a strong positive correlation between temperature and energy consumption 0.8987 and between temperature and renewable energy 0.8711. Other factors, like humidity and square footage, exhibit weak correlations with energy use. By recognizing the trends in energy demand, as well as the effects of external factors, this study can shed light on energy management and optimization models, specifically pertinent to buildings and smart grids.

Keywords

Energy Consumption Renewable Energy Time Series Analysis Temperature Dependency.

Publication Details

  • Type of Publication: Conference 
  • Conference Name: IEEE Region 10 Symposium 2025 (TENSYMP2025)
  • Date of Conference: 07/07/2025 - 07/07/2025
  • Venue: University of Canterbury, Christchurch, New Zealand
  • Organizer: IEEE Region 10
View on Publisher