论文部分内容阅读
为建立更准确、稳定的病虫害预测预报模型,减少农作物病虫害损失、提高农作物产量与质量,运用主成分分析法从42个基础气象因子中整合形成8个新的自变量输入模型,采用试凑法对网络关键参数进行筛选,用2002—2011年数据进行网络训练,建立了以Morlet小波函数为传递函数的小波神经网络模型,并与以Sigmoid函数为传递函数的BP神经网络模型进行了比较。在小波神经网络训练过程中,有6年拟合精度在90%以上,平均拟合精度为89%,预测结果 MAPE值为4.1939,MSE值为5.9764;在BP神经网络的训练过程中,有4年拟合精度超过90%,平均拟合精度仅为81.07%,预测结果中MAPE值为6.4694,MSE值为8.2457。从训练结果看,小波神经网络更能准确描述麦蚜发生期的变化规律,其拟合能力较BP神经网络好;从预测精度和模型的稳定性来看,小波神经网络好于BP神经网络。
In order to establish a more accurate and stable prediction model of pests and diseases, reduce the loss of plant diseases and insect pests and improve the yield and quality of crops, eight new input models of independent variables were integrated from 42 basic meteorological factors by principal component analysis, The key parameters of the network were screened, and the network training was carried out with the data of 2002-2011. The wavelet neural network model with Morlet wavelet function as the transfer function was established and compared with the BP neural network model with Sigmoid function as the transfer function. In the training process of wavelet neural network, the accuracy of 6-year fitting is over 90% and the average fitting accuracy is 89%. The MAPE value of prediction result is 4.1939 and the MSE value is 5.9764. In the process of BP neural network training, there are 4 The annual fitting accuracy was more than 90%, and the average fitting precision was only 81.07%. The predicted value of MAPE was 6.4694 and the MSE value was 8.2457. From the training results, the wavelet neural network can describe the variation regularity of the wheat aphid accurately, and its fitting ability is better than that of the BP neural network. From the prediction accuracy and stability of the model, the wavelet neural network is better than the BP neural network.