基于主成分分析和可解释置信规则库的粮堆温度预测方法

Grain Stack Temperature Prediction Method Based on Principal Component Analysis and Interpretable Belief Rule Base

摘要: 可靠的粮堆温度预测对于粮食安全存储影响重大，以可解释的方法对粮堆的温度进行预测可以提高预测结果的可靠性。由于在建模过程中，BRB (Belief Rule Base)的可解释性可能被削弱或丧失。因此，提出了一种新的基于主成分分析和可解释性置信规则库粮堆温度预测模型(Principal Components Analysis, Interpretable Belief Rule Base, PCA-IBRB)。首先，通过PCA将影响粮堆温度的主要指标筛选出来，根据筛选出的指标并结合粮堆温度特征对模型的可解释性建模准则进行定义。其次，利用ER (Evidential Reasoning)对模型的结果进行推理。然后，提出了一种新的带有可解释性约束的投影协方差矩阵自适应进化策略(Projection Covariance Matrix Adaptation Evolutionary Strategies, P-CMA-ES)算法，来保证优化过程的可解释性；最后，通过对吉林省某粮仓的实测数据进行了粮堆温度预测实验研究，平均MSE值达到了0.0044，验证了模型在粮堆温度预测中的有效性。

Abstract: The reliable temperature prediction of grain heap has great influence on safe storage of grain. Grain heap temperature prediction by interpretable method can improve the reliability of prediction results. Since the interpretability of Belief Rule Base may be weakened or lost during the modeling process. Therefore, a new reservoir temperature prediction model based on Principal Components Analysis and Interpretable Belief Rule Base (PCA-IBRB) is proposed. First, the main indicators affecting the temperature of the grain reactor were selected by PCA, and the interpretability modeling criteria of the model were defined according to the selected indicators and the grain reactor temperature characteristics. Secondly, the Evidential Reasoning (ER) method, as a transparent reasoning engine, ensures the interpretability of the reasoning process. Then, a new Projection Covariance Matrix Adaptation Evolutionary Strategies (P-CMA-ES) algorithm with interpretability constraint is proposed to ensure the interpretability of the optimization process. Finally, the temperature prediction of grain reactor was studied by the measured data, and the average MSE value reached 0.0044, which verified the effectiveness of the model in the temperature prediction of grain reactor.