Predicting Energy Consumption Data Using Deep Learning: An LSTM Approach

• Julio Macedo Chaves LAPISCO / IFCE
• Elene Firmeza Ohata LAPISCO / UFC
• Matheus Araujo dos Santos LAPISCO / UFC
• José Daniel de Alencar Santos LAPISCO / IFCE
• Matheus Jardim Bernardes PUCRS
• Daniel Seleme Dora PUCRS
• Pedro Pedrosa Rebouças Filho LAPISCO / IFCE / UFC

Resumo

Nowadays, most of the energy produced globally comes from fossil fuels. However, this type of energy generation harms the environment by emitting various toxic residues in natural bodies. In recent years, an area that has gained strength is producing and consuming clean energy. One that stands out is solar energy production because it is easy to implement and relatively cheap compared to other clean energy production. Therefore, several projects were created focusing on generating energy from sunlight, and using it. Due to this growing number of enterprises in the area, the creation of applications to help manage production and its use has increased. The use of Deep Learning techniques to help this industry has also gained strength; predictive models for energy consumption have been widely studied to help enterprises in future planning. In this work, we developed a deep learning model using Long Short-Term Memory (LSTM) capable of predicting energy consumption from solar power plant data, using the data to train the model and make inferences in the future. We explored configuration combinations such as data filtering with smoothing techniques, model hyperparameters, number of layers, number of neurons, and optimal prediction horizon. The achieved results demonstrate the validity and effectiveness of the implemented methodology.