Development of a fuzzy PID control system with neural networks for temperature control of potato (Solanum tuberosum) cultivation in greenhouses

Abstract

Precise temperature control is crucial for potato (Solanum tuberosum) cultivation in greenhouses, but conventional PID controllers are ineffective in the face of nonlinear dynamics and system disturbances. This work developed a fuzzy PID control system assisted by neural networks (ANFIS) to regulate the temperature in a simulated greenhouse, considering the conditions for potato cultivation. Following the VDI 2206 methodology and using MATLAB/Simulink, a first-order model of the greenhouse, was derived, and three control strategies were designed: classical PID, fuzzy PID, and ANFIS, the latter trained with data from the fuzzy PID model. System validation compared performance using standard metrics in step response and disturbance response. The results demonstrated that the ANFIS controller was superior, achieving a settling time of 126.34 s and maintaining the maximum deviation from external disturbances below 0.2 °C (specifically 0.15 °C), thus validating the main hypothesis. This performance significantly outperformed the classical PID controller (ts = 1053.3 s, Desv = 0.71 °C) and the fuzzy PID controller (ts = 224.94 s, Desv = 0.71 °C), showing a faster, more accurate, and more stable response. Although the neuro-fuzzy controller required greater computational effort, its performance demonstrated a considerable improvement in the speed and robustness of the system, establishing it as an effective solution for optimizing thermal stability in greenhouses used for potato cultivation.