Deep Learning-Based Analysis on Monthly Household Consumption for Different Electricity Contracts

Abstract

In predicting electricity consumption, deep learning models based on various neural network architectures are widely used. Since many factors affect electricity consumption in reality, it is difficult to deal with statistical approaches, while deep learning models can be trained using enough data in practice. In this paper, we analyze the characteristics of the electricity consumption data according to the contract type and measure the performance of the future electricity consumption prediction by applying the deep learning model. The experimental data show different trends according to the contract types, and it is expected that these differences may affect the learning performance of prediction models. Through the experiment, we check the difference of performance depends on the complexity and configuration of models by contract types.