基于GA优化的支持向量机模型在青椒作物需水量预测中的应用

Application of Support Vector Machine Model Based on GA Optimization in Water Consumption Prediction of Green Peppers

为节约灌溉用水,采用垄沟集雨覆盖种植技术与滴灌技术相结合(MFR-DI),并对使用该技术种植的青椒进行作物需水量预测。根据多年气象资料、青椒冠层温度以及逐日作物需水量资料,构建了以冠层温度、气象因素为输入因子的预测MFR-DI种植模式下青椒作物需水量的GA-SVM模型,使用2017年的数据对模型进行了测试,结果表明:在输入相同气象因子时,GA-SVM1(RMSE=0.9010 mm/d,MAE=0.6735 mm/d,NS=0.9718)比SVM(RMSE=0.9607 mm/d,MAE=0.7691 mm/d,NS=0.9680)预测模型具有更高的精度性能。此外,在输入相同数量的因子时,将冠层温度作为GA-SVM的输入因子之一,比仅输入气象因子的GA-SVM模型预测精度更高,其RMSE,MAE,NS值分别为0.7817 mm/d,0.5838 mm/d,0.9788。结果说明采用GA优化的SVM预测模型,提高了模型的收敛速度,使模型的精确度更高。另外,在作物需水量预测模型中引入冠层温度,可以提高模型的预测准确性,为实现高效智能节水提供参考。

In order to save irrigation water,the planting technology of furrow rainwater harvesting combined with drip irrigation(MFR-DI)was adopted,and the crop water consumption of green peppers in this technology was predicted.Based on years of meteorological data,canopy temperature and daily crop water consumption of green peppers,the GA-SVM model for predicting daily crop water consumption of green peppers in MFR-DI planting mode was constructed with canopy temperature and meteorological factors as input factors.The model was tested with data of 2017.The results showed that when the same meteorological factors were input,GA-SVM1(RMSE=0.9010 mm/d,MAE=0.6735 mm/d,NS=0.9718)model had higher precision performance than SVM(RMSE=0.9607 mm/d,MAE=0.7691 mm/d,NS=0.9680)model.In addition,under the same number of input factors,when canopy temperature was introduced as one of the input factors of GA-SVM,the prediction accuracy was higher than that of GA-SVM model with only meteorological factor.The RMSE,MAE and NS of GA-SVM with canopy temperature being introduced were 0.7817 mm/d,0.5838 mm/d and 0.9788,respectively.The results show that GA can improve the convergence speed of SVM model and make the prediction model more accurate.In addition,introducing canopy temperature into crop water consumption prediction model can improve the prediction accuracy of the model and provide reference for realizing efficient and intelligentwater saving.