Machine Learning Framework for Energy Consumption Optimization using the TabPFNRegressor Algorithm

The optimization of energy consumption has the potential to reduce the costs of wastewater treatment operations while maintaining water quality and complying with environmental regulations. An energy optimization system enables managers and operators to develop wastewater treatment plants (WWTPs) to control their energy and make decisions that could improve energy efficiency, productivity, and business sustainability while maintaining water quality. This paper studies the potential of applying a machine learning solution based on TabPFNRegressor, which combines the tabular prior data fitted network (TabPFN) and feature selection techniques to optimize energy consumption in the WWTP. Recursive feature elimination is used to select the most critical features of the Melbourne water treatment plant and airport weather station dataset. Furthermore, the TabPFNRegressor is compared with different ML models, including decision tree regressor, random forest regressor, support vector regressor, Lasso, Ridge, and K-neighbors regressor, using both full features and optimized features. The results showed that TabPFNRegressor achieves the best results compared with six models, with the lowest error rates of 34.6433 for RMSE and 28.1260 for MAE.