Predição da emissão de gases de efeito estufa em uma estação de tratamento de águas residuais

Abstract

This work investigates the application of different predictive methods to improve understanding of greenhouse gas (CO2 and N2O) emissions at a wastewater treatment plant, with data obtained from a plant located in Helsinki, Finland. Time series data were collected over a period of two months, with hourly measurements. The analysis of this data involved several steps, including cleaning and preparing the data, as well as exploratory analysis to identify relevant patterns and behaviors. Based on this analysis, four predictive models were applied: linear regression, polynomial regression, simple neural network, and long short-term memory network. The results showed that the more complex models performed better in predicting and identifying possible ways to reduce greenhouse gas emissions at the treatment plant.