基于IGWO-SVR短期储藏小麦品质预测模型研究

储藏小麦品质具有复杂性、易变性、多耦合特性，导致难以准确预测其品质状况。为此，本研究从小麦多生理生化指标关联性研究角度提出了一种新的品质预测方法。利用柯西核函数和改进的线性核函数来构造支持向量回归机（SVR）混合核函数，并用改进的灰狼算法（IGWO）对混合核函数SVR参数寻优，由此建立IGWO-SVR模型用于短期储藏小麦的品质预测。选用周麦22对模型进行验证，结果显示：混合核函数IGWO-SVR模型的平均相对误差相比于线性核、多项式核和径向基核的模型分别下降了4.24%、2.56%和1.74%；IGWO-SVR各预测效果评价指标均优于GS-SVR、CS-SVR和GWO-SVR模型，模型整体预测精度和拟合效果显著提高。最后通过周麦22的发芽率作为品质评估指标和郑麦9023多指标数据分别对IGWO-SVR模型的有效性和适用性进行检验，得到平均绝对百分比误差MAPE分别为1.85%和3.87%，表明模型性能良好。试验结果表明了新建立模型在短期储藏小麦品质预测方面的可行性。

Research on prediction model of short-term stored wheat quality based on IGWO-SVR

The quality of the stored wheat has the characteristics of complexity, variability and multi-coupling, which makes it difficult to judge the current quality condition accurately. Therefore, a new prediction method based on the correlation between multiple physiological and biochemical indexes of the wheat is proposed in this study. The new hybrid kernel function used in SVR is constructed by cauchy kernel function and improved linear kernel function. Meanwhile, the key parameters of the SVR with hybrid kernel function are optimized by the IGWO, thus the IGWO-SVR model can be constructed to predict the quality of short-term stored wheat. The model was verified by Zhoumai 22, and test results show that the average relative error of the SVR model using hybrid kernel function has decreased by 4.24%, 2.56% and 1.74% compared with linear kernel function, polynomial kernel function and radial basis kernel function, respectively. Compared with the GS-SVR, CS-SVR and GWO-SVR models, the IGWO-SVR model has a significant improvement in terms of prediction accuracy, demonstrating that the overall prediction accuracy and fitting effect of the modified model are significantly improved. Finally, the validity and applicability of the IGWO-SVR model were tested by using the germination rate of Zhoumai 22 as the quality evaluation index and Zhengmai 9023 multi-index data, and the MAPE values were 1.85% and 3.87%, respectively, which indicate that the performance of the model is good. The results show that the new model is effective and feasible in the quality prediction of short-term stored wheat.