Here is an overview of the projects I am currently involved.
The Shafdan Soil Aquifer Treatment (SAT) system supplies over 140Mm3/y of effluent for irrigation of raw- consumed crops. Currently, it operates beyond capacity, discharging millions of cubic meters of effluent into the sea. This can be improved by optimizing effluent distribution across basins. We propose to integrate machine learning (ML) methodologies to forecast the infiltration rate to the basins, considering both the past and current ambient and operational conditions. The infiltration rate is the most important characteristic of infiltration basins. Remarkably, our preliminary research, analyzing over 42,000 flooding cycles spanning a decade of operation across 50 infiltration basins, revealed that the infiltration rate remains head-independent during the drainage phase. This discovery simplifies the infiltration rate prediction significantly. ML techniques together with Prediction Explainability methods will provide forecasts of the infiltration rate and insights into the underlying (mechanistic) factors driving these predictions. Strategies for the optimal distribution of the SAT inflow across multiple basins will be examined and validated through simulations based on historical conditions. Water losses across different operational scenarios will be compared with the actual losses under the same conditions. Field tests will validate the effectiveness of the proposed approach. Such modeling has never been proposed before and could be developed only through collaboration between a data science team and SAT experts.
Further links: