The use of micro-algae for wastewater treatment is a promising technique that contributes to CO2 capture and nutrient recovery. However, the lack of effective forecasting models limits the scalability of this technique. This study aims to develop a time-series-based forecasting model to predict the growth curve of microalgal biomass under environmental conditions similar to those found in wastewater. Data collected on microalgal growth was used to train six time-series models: Long Short-Term Memory (LSTM), Extreme Gradient Boosting (XGBoost), Auto-Regressive Integrated Moving Average (ARIMA), Random vector functional link (RVFL), Physics-informed neural networks (PINN) and Prophet. The model performance metrics were compared, and the best model was identified. The results demonstrated that the RVFL was the most accurate model, with minimal prediction errors ( < 0.01). Residual analysis confirmed a normal distribution of errors without outliers, supporting the model's reliability. These findings suggest that the proposed RVFL model can effectively forecast microalgal growth, potentially reducing the need for costly and labour-intensive laboratory trials and advancing microalgae-based wastewater treatment. © 2025 The Institution of Chemical Engineers
