| Title |
Electricity Consumption Prediction using Generative Models |
| Authors |
김희상(Heesang Kim) ; 임현승(Hyeonseung Im) ; 문양세(Yang-Sae Moon) |
| DOI |
https://doi.org/10.5370/KIEE.2022.71.1.218 |
| Keywords |
Deep learning; Electricity consumption; Multi-step prediction; Generative model |
| Abstract |
Recently, advanced metering infrastructure (AMI) has been deployed for power demand distribution and energy saving, and correspondingly traditional watt-hour meters installed in apartments and industrial sites are also being replaced with AMIs.
Accordingly, power demand prediction using AMIs will become increasingly important to save electrical energy consumption. In this paper, we develop various deep learning-based electricity consumption prediction models using simple neural networks, convolutional neural networks, recurrent neural networks, and encoder-decoder-based generative models. To build prediction models, we use average power demand data collected from various home smart meters. Experimental results show that the generative model outperforms other deep learning-based models in terms of mean squared error, and we roughly explain why the generative model is better than other models by examining the activation layer output distributions. |