Multipoint-measurement multipoint-heating greenhouse temperature control with wooden pellet fuel using an adaptive model predictive control approach with a genetic algorithm

For the simplest case with a single point measurement and a single heater under operational constraint on heater, a model predictive control [1] was already proposed and applied to evaluate in a real field test. The obtained results indicated the effectiveness of the control although searching for an optimal solution was essentially based on the brute-force attack. It is not feasible to extend the same approach to determine an optimal solution in the case of multipoint temperature measurements and multipoint heaters for calculations in a real time operation. For his problem, a genetic algorithm was applied as a solution to regulate greenhouse temperature for the above-mentioned problem. Model predictive control performance strongly depended on the precision of the mathematical model of greenhouse temperature dynamics, and thus, a recursive least square identification strategy was also introduced to update parameters in the predictor on line. The control results in a laboratory demonstration greenhouse indicated that the proposed method performed well and showed an average decrease of 20-35% in the temperature error. Furthermore, the temperature difference also decreased by an average of 36%.