| Title |
Deep learning and Prediction of Individual Household Energy Bills Data considering Weather information |
| Authors |
박지수(Jisoo Park) ; 홍승우(Seungwoo Hong) ; 서일홍(Il-hong Suh) |
| DOI |
https://doi.org/10.5573/ieie.2020.57.4.53 |
| Keywords |
Deep learning; Long-Short Term Memory(LSTM); Household energy bills prediction |
| Abstract |
Energy bills are one of the household's fixed expenses. In particular, when energy consumption is soaring due to weather conditions, a progressive rate is applied, which raises household burden. Therefore, consumers need to be rational in their energy use, and for this, they must be able to predict energy expenditure and adjust the energy use accordingly. To this end, in this paper, we propose a prediction method of the monthly energy bill for individual households using a deep-learning-based model, considering the weather, which has an important effect on energy bill prediction. As weather information, minimum temperature, maximum temperature, precipitation probability, precipitation, humidity, wind speed, snow level, and cloud cover are used, which shows an experimentally significant performance improvement. Also, three representative deep learning models (Multilayer Perceptron, Convolution Neural Network, Long-Short Term Memory) are designed and implemented for the given problem, and the model based on Long-short term memory exhibits the lowest error. The proposed method based on these results is applied to the actual energy usage data of 2,234 households and weather data of the Korea Meteorological Administration. The experiment shows that the energy bills of individual households can be predicted with a small average error of 5,110 won (4.28 dollars when the Korean won to the U.S. dollar exchange rate is 1,194 won per dollar) using the proposed approach. |