Multi-factor fungal growth prediction and optimal control during grain storage based on RF error compensation

Abstract

Grain storage security is primarily threatened by fungal growth, which produces mycotoxins that contaminate the stored grain and compromise its quality. Predicting the growth trends of fungi during storage is essential for ensuring the safety of grain reserves. Traditional fungal growth prediction models often rely on single-factor analysis, which leads to low accuracy. This paper proposes a multi-factor model for predicting fungal growth in grain storage, taking into account the effects of temperature and moisture. By adjusting current environmental conditions, we can predict and control the future growth rate of fungi, thereby enhancing grain storage security. However, while stricter regulation of environmental factors can reduce fungal growth, it also increases control costs. Therefore, it is economically prudent to minimize the extent of environmental adjustments, striking a balance between effectiveness and cost-efficiency. To address these challenges, we have refined the fungal growth model by incorporating the effects of temperature and moisture, transforming the model from colony diameter growth to fungal spore count, and utilizing a discrete model to predict both colony diameter and spore number. Finally, we employ a Random Forest (RF) model to correct prediction errors, optimizing control strategies by minimizing the tracking deviation of state variables and the cumulative increment of control variables.