ABSTRACT Food security is an influential factor for the development and future of humanity. Stored grain temperature is a key physical variable for grain quality control and management in stored-grain ecosystem. In this study, a prediction model using Fourier series based on least-squares method is introduced to describe the average temperature movement of grain pile and the daily air temperature, naming hysteresis cycle model (HCM). Then the HCM based on the Fourier analysis is proposed that incorporating with the least-squares method, the Fourier series model is applied to forecast the grain pile temperature from the daily air temperature. The HCM is built by finding relationship called conversion coefficients between the air and grain pile temperature that is constructed on the basis of Fourier analysis, and the optimal order of the model is determined by Akaike information criteria. Our method can reflect the time lag of the grain pile temperature changes with the change of air temperature and requires only a single variable (i.e. air temperature). Experiments including model fitting, temperature prediction, and results comparison for a period of 425 days from a real-world granary, yield satisfied results agreeing well with actual observation values. As a result of model validation, the proposed method has the root mean square error (RMSE) values ranging from 1.6°C to 2.3°C with average RMSE of 1.9°C. In other words, the proposed model can predict grain pile temperature with a high accuracy once an acceptable relationship between air temperature and grain pile temperature has been constructed from previously temperature data. Result analysis indicates effectiveness and workability of the proposed model for stored grain temperature prediction.

中文翻译：

---

从气温分析预测粮食温度，保障粮食储存安全

摘要 粮食安全是影响人类发展和未来的重要因素。储粮温度是储粮生态系统中粮食质量控制和管理的关键物理变量。本研究引入基于最小二乘法的傅里叶级数预测模型来描述粮堆的平均温度运动和日气温，命名为滞后循环模型（HCM）。然后提出了基于傅里叶分析的HCM，结合最小二乘法，应用傅里叶级数模型从日气温预测粮堆温度。HCM 是通过在傅立叶分析的基础上找到空气和谷物堆温度之间称为转换系数的关系来构建的，模型的最优阶数由 Akaike 信息准则决定。我们的方法可以反映粮堆温度随气温变化的时间滞后，并且只需要一个变量（即气温）。模型拟合、温度预测、真实粮仓425天的结果对比等实验，得到了与实际观测值吻合良好的满意结果。作为模型验证的结果，所提出的方法的均方根误差 (RMSE) 值范围为 1.6°C 到 2.3°C，平均 RMSE 为 1.9°C。换句话说，一旦根据先前的温度数据构建了空气温度和粮堆温度之间可接受的关系，所提出的模型就可以高精度地预测粮堆温度。