A Machine Learning-Based Integrated Energy and Water Management System for a Greenhouse

The scarcity of water supply and the high price of electricity are among the impediments to high agricultural productivity and affordable crop production costs. A promising solution is to maximize solar energy production in the area to power the water pumping system of the greenhouse while minimizing the system's grid dependence by strategically scheduling water storage refills. Several studies have attempted to implement a management system for solar and water resources, employing varied approaches from simple rule-based to complex optimization-based techniques. However, existing implementations have focused on off-grid systems in open-field farms and have not explored the implementation of grid-tied systems in a semi-closed environment, aimed at minimizing the grid-dependence and maintaining the reliability of its water supply system. In this paper, an integrated energy and water management system (EWMS) was developed using an extreme learning machine (ELM) to forecast the day-ahead solar PV production and monthly crop water requirement (CWR) for a greenhouse. An experimental setup was established to verify the performance of the management system, and a model of the greenhouse farm's irrigation system was created using MATLAB/Simulink to explore different scenarios. The developed EWMS was able to increase the solar PV consumption up to 94.3% (2.34kWh higher) and reduce grid energy consumption by 3.49kWh (3.5-fold reduction) while maintaining a 99.2% water reliability index (WRI) compared to the base case operating without EWMS. The system outperformed the base setup, even with a larger storage tank capacity, in terms of its grid-dependence and self-consumption.