Enhancing Electricity Tariff Classification with CNNs-Based and Time Series Data Augmentation

Understanding and classifying electricity tariffs accurately is the key factor to optimizing energy consumption and planning. However, the limited and imbalanced time series data collected often facade challenges for the deep learning models. In this study, we explore the usage of data augmentation, specifically in magnitude warping and scaling techniques to enhance the performance of a Convolutional Neural Network (CNN) models in classifying electricity tariffs. We compare the CNNs performance on the original data versus augmented datasets to evaluate improvements in accuracy, stability, and generalization. The findings show that incorporating time series data augmentation significantly enhances classification performance by increasing ResNet-18 from 96.61 % to 100% and improving ResNet-50 from 94.92% to 100% after applying magnitude warping and scaling techniques. These improvements highlight the effectiveness of augmentation in increasing the classification performance, improving generalization and enhancing model robustness in real-world energy applications.