Irrigation Controller Validation

I helped to run in-field experiments for the Predictive Optimal Water and Energy Irrigation (POWEIr) Controller in Morocco, Jordan, and Kenya. This work was featured in an MIT news article and the MIT MechE Video titled “No Drop to Spare.”

For this work, I first conducted sensitivity analyses through simulations to see the effect of weather and farmer input error on the POWEIr controller’s optimal schedules and crop yield. Then I helped to build and set up the controller hardware on research stations and farms in the Middle East and North Africa. I used model predictive control to optimize daily irrigation schedules based on machine learning predictions of local weather and solar power, all software that I implemented in Python. I compared the pumping energy, water use, and crop yield of the system with the POWEIr controller directly to measurements from farmers growing the same crops in the same area as a reference. The results of this work showed that the POWEIr controller could save up to 44% of water and 43% of pumping energy for similar crop yields compared to typical practices.

Stay tuned for updates and publications!