摘要
采用阵列式分布的测温电缆检测粮仓温度变化情况,利用机器学习技术来预测粮食温度,用粮仓1年监测数据来预测粮堆未来27 d温度。传统的BP、RBF、RF、SVR单模型对粮堆温度进行预测存在误差大、泛化能力差等缺点,提出一种基于Bagging集成的鲸鱼算法优化支持向量回归模型(Bagging-WOA-SVR),并与灰狼算法优化支持向量回归模型作比较。将影响粮堆温度的多种因素做灰色关联分析,选取粮仓内温度、粮仓内湿度、粮仓外温度、粮仓外湿度、粮仓平均温度、地表温度作为神经网络的输入,粮堆平均温度作为预测输出,选取3个指标为评判标准,对比分析模型预测精度。结果表明:提出的Bagging-WOA-SVR模型相比之下有着较好的稳定性,均方误差为0.24,相关系数为0.9892。
The array distributed temperature measurement cable was used to detect the temperature change of the granary,the machine learning technology was used to predict the grain temperature,and the 1-year monitoring data of the granary was used to predict the temperature of the granary in the next 27 days.The traditional single model of BP,RBF,RF and SVR had the disadvantages of large error and poor generalization ability in predicting the temperature of grain pile.A support vector regression model optimized by whale algorithm based on Bagging ensemble(Bagging-WOA-SVR)was proposed and compared with the support vector regression model optimized by grey wolf algorithm.A variety of factors affecting the temperature of the granary were analyzed by grey correlation analysis.The temperature inside the granary,the humidity inside the granary,the temperature outside the granary,the humidity outside the granary,the average temperature of the granary,and the surface temperature were selected as the input of the neural network.The average temperature of the granary was used as the prediction output,and three indicators were selected as the evaluation criteria to compare and analyze the prediction accuracy of the model.The results indicated that the proposed Bagging-WOA-SVR model had better stability in comparison,with a mean square error of 0.24 and a correlation coefficient of 0.9892.
作者
韩建军
张梦琪
赵道松
郭妍妍
杨雅冰
Han Jianjun;Zhang Mengqi;Zhao Daosong;Guo Yanyan;Yang Yabing(College of Civil Engineering,Henan University of Technology,Zhengzhou 450001)
出处
《中国粮油学报》
CAS
CSCD
北大核心
2024年第6期7-12,共6页
Journal of the Chinese Cereals and Oils Association
