摘要
采用RBF网络与BP网络的方法,利用MATLAB工具箱并结合气象资料中的相对湿度、平均气温和太阳日辐射量,建立了预测核桃作物需水量的神经网络预测模型。两种预测模型通过实例证实了预测的准确性,并且将这两种网络模型进行了比较分析。RBF神经网络预测作物需水量的绝对误差平均值为0.254 7mm/d、相对误差平均值为5.47%,BP神经网络预测作物需水量的绝对误差平均值为0.320 6mm/d、相对误差平均值为6.97%,由此可见,RBF网络预测的精度比BP网络高。并且,通过程序记时显示RBF网络训练用时0.063 0s,比BP网络训练所需的时间要短的多,因此RBF神经网络具有较好的实用价值,实现了精度与实用性的统一。
Using the method of RBF network, BP network and MATLAB toolbox, combined with the meteorological data ot the rela-tive humidity, average temperature, and the daily solar radiation level, this paper has established the neural network model for fore- casting the water consumption of walnut crop. Both forecast model has proved their accuracy through the practical examples and the comparison is conducted between the two models. The mean value of the absolute error of RBF neural network forecast model is 0. 254 7 mm/d with the mean value of relative error of 5.47~. BP neural network forecast model presents a mean value of absolute error of 0. 320 6 mm/d with the mean value of relative error of 6.97~, which suggests that RBF network shows a higher level of accuracy than BP network. In addition, the programming clock shows the RBF model present a training time of 0. 063 0 s, which is much shorter than that of BP network. Therefore, it is suggested that RBF network has more practical value because that it realizes the integration of accuracy and practicability.
出处
《节水灌溉》
北大核心
2013年第3期16-19,23,共5页
Water Saving Irrigation
基金
国家高技术研究发展计划("863"计划)资助项目"新疆特色果树微灌节水增效技术研究与示范"(201130103-1)
新疆自治区高技术研究与发展计划项目(200712111)
