Analyzing the electricity consumption of a hypermarket at quarter-hour intervals is a serious challenge in itself, but what happens if we want to increase the accuracy of the forecast? The study by Csilla Obádovics and Levente Szabados seeks to answer exactly this question, presenting a methodological experiment that focuses on the comparison of traditional forecasting methods and state-of-the-art artificial intelligence-based models.
Background of the research
The authors examined the detailed consumption data of a Hungarian hypermarket between September 2007 and July 2008, which were recorded in quarter-hour breakdowns. The aim was to forecast the electricity consumption of industrial consumers for 48 hours in advance, taking into account weather data, the number of visitors and other variables. One of the main objectives of the research is to lay the foundations of a standardizable energy management system.
Economic significance of the forecast
Accurate energy consumption forecasts play a key role in optimizing electricity procurement. Forecasting models not only contribute to cost reduction but also increase market efficiency and grid stability. In addition, the application of AI allows energy companies to predict equipment failures, which reduces downtime and optimizes maintenance costs.
Surprising correlations in energy consumption
The research revealed several interesting facts:
- Energy consumption is higher during the day, especially in the case of a hypermarket.
- In summer, consumption increases significantly due to cooling, while in winter the fluctuation is smaller.
- The days of the month and the days of the week also affect the turnover of stores and thus the energy demand.
- The effect of temperature does not appear immediately, but after a delay of about 2 hours.
- Temperatures above 14 degrees Celsius have a significant impact on energy use, while below them there is no strong correlation.
Traditional and AI-based models
The researchers used a variety of models, including linear regression and state-of-the-art deep learning solutions such as the TiDE and Lag-Llama models. Classical models have the advantage of stability and transparency, while deep learning systems excel in modeling nonlinear time series and handling complex interactions.
The tests showed that the TiDE model adapts more efficiently to changing energy demand. At the same time, the pre-trained version of Lag-Llama performed outstandingly, surpassing both traditional methods and the TiDE model.
The research points out that the use of artificial intelligence in the field of energy management is not only a technological advance but can also result in significant savings from an economic point of view.
Are you interested in how AI can be used to make energy management more efficient? The study, published in the latest issue of the Statistical Review, uses detailed diagrams to show how traditional methods and state-of-the-art deep learning models can be used to predict electricity consumption. The article offers practical solutions to reduce energy costs and increase the stability of the grid. Discover how AI can help address climate challenges while also creating economic benefits.
You can access the full article at this link:
https://www.ksh.hu/statszemle_archive/all/2025/2025_01/2025_01_021.pdf
