A new AI model for predicting energy consumption could help managers plan ahead as well as open up opportunities for cutting carbon emissions, according to research in the International Journal of Information and Communication Technology.
The system uses a transformer-based generative adversarial network (GAN) to learn complex patterns in energy use. GANs are machine-learning systems in which two neural networks work against each other to produce realistic data. By adding Bayesian statistical optimisation, the team could set the system for best performance.
They tested the model on 2000 hourly energy and emissions records drawn from smart meters, building management systems, and industrial grids. They removed references to missing data and deleted outliers to normalise the input. The model made predictions that closely matched the observed energy data while taking only a few seconds to train and run.
The researchers explain that their computer-generated scenarios will allow operators to test demand-shifting measures, integrate renewable generation, and identify inefficiencies ahead of actual power generation and emission formation.
The approach addresses several weaknesses in conventional predictive models, which are known to struggle with non-linear consumption patterns, sudden spikes and troughs, and multiple external factors. This carbon-aware energy optimisation could be used across sites with internal power generation, such as industrial sites and in wider energy systems.
Liu, Y. and Li, B. (2026) ‘Modelling and predicting energy consumption patterns using generative adversarial networks for effective carbon management’, Int. J. Information and Communication Technology, Vol. 27, No. 92, pp.79–110.
News media may use this press release as source material, in whole or in part, provided the content is not materially misrepresented. A link back to the original article is appreciated.
No comments:
Post a Comment