Environmental Control for Edible Fungi Cultivation Based on Temporal Information and Deep Learning

ABSTRACT Currently, there are still prevalent issues in greenhouse environmental regulation, such as response lag, low control accuracy, and difficulty in coping with sudden environmental disturbances. To achieve high‐precision and dynamic control of the edible fungi cultivation environment, this study proposes an edible fungi environmental control method based on temporal information and deep learning. Firstly, this approach collects real‐time temporal information through various sensors and utilizes the RS485 bus and MODBUS‐RTU protocol for data transmission. Ultimately, by combining singular spectrum analysis, principal component analysis, and temporal convolutional networks, it achieves precise prediction of greenhouse environmental variables. Simulation test results demonstrate that the system can maintain various environmental parameters within the set target ranges during seven consecutive days of operation, achieving high‐precision dynamic control. Even under Gaussian noise interference (with a standard deviation of 2%), the stability indices for air temperature, humidity, CO 2 concentration, and light intensity remain within the range of 0.928 to 0.959. Furthermore, in simulated sudden disturbance experiments, such as a 50% drop in light intensity, rapid temperature changes of ±3°C, humidity fluctuations of ±5% RH, and a short‐term increase in CO 2 concentration by +200ppm, the recovery times for various environmental variables in the system are controlled within 0.7 h to 1.4 h, significantly shorter than those of traditional greenhouse systems. Therefore, the introduced approach can effectively capture the temporal characteristics of the greenhouse environment, enabling precise prediction and rapid response of environmental parameters. It holds significant reference value for smart agricultural greenhouse environmental control.