Evolving and Comparing Greenhouse Control Strategies using Model-Based Multi-Objective Optimization

Optimal control of greenhouse environments is a difficult problem to solve, mainly due to the presence of many user-adjustable control settings and complex microclimate-crop interactions. Although maximizing profit is a typical primary objective, narrowing down a grower's needs into a single objective may not always be realistic, depending on other secondary objectives such as minimizing energy costs or maximizing yield. These challenges make obtaining suitable solutions elusive because growers cannot rely solely on experience to determine the exact effects on the crop of changing any of the controller's settings. We use NSGA-II, a popular multi-objective evolutionary algorithm, along with a combined microclimate-crop-yield model, to evolve microclimate control setpoints based on a classical control strategy, using the value of the crop yield and the variable costs as objectives. The results show that the evolved environmental setpoints can provide the grower a variety of better solutions, providing greater profitability compared to not using evolved setpoints. In addition, the evolved classical control strategy is compared with an enhanced version that has different setpoints based on the time of day. When the hypervolumes of their Paretooptimal fronts are used as a performance metric, the results show that the enhanced controller has superior performance.