This study aimed to evaluate the performance of various time series forecasting algorithms for predicting Total Reduced Sulfur (TRS) emissions in the pulp and paper industry. The analysis was conducted using a real time series dataset, where 25 samples autocorrelation function was computed to identify the level of correlation among the variables lags. The results showed that the last four previous TRS values had a total autocorrelation above 0.8. Additionally, the partial correlation function indicated that only the last previous TRS value had a correlation value above 0.3. The comparison of the performance of different time series algorithms, including XGBRegressor, LSTM, CNN, MLP-MHA, MLPRegressor, and ARIAMA, was made based on two different horizons (1 and 8). The evaluation metrics used were MAE, MSE, R2-score, and execution time. The results indicated that ARIAMA outperformed the other algorithms on horizon 1, with an MAE of 0.0565 and R2-score of 0.7046. On the other hand, LSTM had the best performance on horizon 8, with an MAE of 0.0853 and R2-score of 0.5241. These findings suggest that advanced time series prediction algorithms can provide more accurate models for predicting TRS emissions in the pulp and paper industry, contributing to environmental and health mitigation efforts. Finally, the use of these models can help the industry comply with regulatory standards and avoid penalties while also supporting its economic sustainability.