Abdul Mujiburrohman Luthfi, Iswanto, Alfian Maarif, Gilang Nugraha Putu Pratama
Power consumption prediction has an important role in efficient energy management, enabling better utilization of resources and cost savings. Conventional methods have often faced challenges in accurately predicting power consumption due to the complexities inherent in real-world energy systems. In order to tackle the problem, an approach utilizing a CNN-BiLSTM model in combination with Raspberry Pi 4, Arduino Uno R3, and PZEM 004T is presented. In this study, a seven-day experiment is conducted to implement the CNN-BiLSTM model for power consumption prediction. The model is trained on the collected data and subsequently tested on unseen samples. The results obtained demonstrate the effectiveness of the CNN-BiLSTM model, with a Mean Absolute Error (MAE) of 3.0906, which indicates low prediction errors. © 2023 IEEE.
Universitas Negeri Yogyakarta, Yogyakarta, Indonesia; Universitas Muhammadiyah Yogyakarta, Yogyakarta, Indonesia; Universitas Ahmad Dahlan, Yogyakarta, Indonesia
