Author(s): Oliveira, Pedro ; Marcondes, Francisco Supino ; Duarte, Maria Salomé Lira ; Durães, Dalila ; Martins, Gilberto ; Novais, Paulo
Date: 2024
Persistent ID: https://hdl.handle.net/1822/91800
Origin: RepositóriUM - Universidade do Minho
Subject(s): Anaerobic digestion; Deep learning; Electricity; Time series; Wastewater treatment plants
Over the decades, we have faced escalating global energy consumption and its consequential environmental impacts, including climate change and pollution. This study explicitly evaluates the use of Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) models for predicting electricity production from the biogas produced in a Wastewater Treatment Plant (WWTP) in Portugal. WWTPs play an essential role regarding environmental sustainability, namely the potential of biogas in mitigating energy consumption's environmental impact. Also, the work details a comparison between the LSTM and GRU model's performance, applying a grid-search methodology for hyperparameter optimization. The study employs the Root Mean Squared Error (RMSE) as an evaluation metric and uses the sliding window method to transform the problem into a supervised one. After several experiments, the results demonstrate that the LSTM-based model outperforms GRU-based models, achieving an RMSE of 347.9 kWh.
EC -European Commission(2022.06822)
info:eu-repo/semantics/publishedVersion
