While China＇s tea consumption slows,globally tea lovers are sipping with increasing zest WHEN you think of China one of the first things that comes to mind is tea.It is a staple beverage intrinsically woven into the ...While China＇s tea consumption slows,globally tea lovers are sipping with increasing zest WHEN you think of China one of the first things that comes to mind is tea.It is a staple beverage intrinsically woven into the tapestry of Chinese lifestyle in a country that has the earliest records of tea drinking.China today shares its展开更多
THE first thing Cai Banghong does every morning is to look at the sales figure of various tea brands on his online store. Hmongling Tea Garden, before posting promotional information and collecting customers＇ feed- b...THE first thing Cai Banghong does every morning is to look at the sales figure of various tea brands on his online store. Hmongling Tea Garden, before posting promotional information and collecting customers＇ feed- back to help target his products Cai spent two decades growing his tea company from a small backroom operation into a modern tea company integrating planting, processing and sales.展开更多
在茶园水资源管理中,蒸散量(Evapotranspiration,ET)是评估作物水分需求的关键指标,由于茶园蒸散量预测具有时序性、不稳定性以及非线性耦合等特点,目前的茶园蒸散量预测模型存在预测精度较低的问题,针对此问题本文提出了一种新型的茶...在茶园水资源管理中,蒸散量(Evapotranspiration,ET)是评估作物水分需求的关键指标,由于茶园蒸散量预测具有时序性、不稳定性以及非线性耦合等特点,目前的茶园蒸散量预测模型存在预测精度较低的问题,针对此问题本文提出了一种新型的茶园蒸散量预测模型。首先使用互信息算法(Mutual information,MI)与主成分分析算法(Principal component analysis,PCA)相融合的数据处理算法(MIPCA),筛选强相关的特征并提取主成分;其次将时域卷积网络(Temporal convolutional network,TCN)与Transformer融合,利用灰狼算法(Grey wolf optimization,GWO)优化超参数,捕捉茶园数据的全局依赖关系;最后整合2个网络构建了MIPCA-TCN-GWO-Transformer模型,通过消融试验和对比试验验证了模型性能,并对模型在不同时间步长下的性能进行测试。结果表明,该模型平均绝对百分比误差(Mean absolute percentage error,MAPE)、均方根误差(Root mean square error,RMSE)和决定系数(Coefficient of determination,R^(2))3个评价指标分别为0.015 mm/d、0.312 mm/d和0.962,优于长短期记忆模型(Long short term memory,LSTM)等传统预测模型。在小时尺度、日尺度和月尺度下的R^(2)分别为0.986、0.978和0.946,在不同时间步长下展现了良好的适应性和准确性。本文构建的MIPCA-TCN-GWO-Transformer模型具有较高的预测精度和稳定性,可为茶园水资源优化管理和灌溉制度制定提供科学参考。展开更多
文摘While China＇s tea consumption slows,globally tea lovers are sipping with increasing zest WHEN you think of China one of the first things that comes to mind is tea.It is a staple beverage intrinsically woven into the tapestry of Chinese lifestyle in a country that has the earliest records of tea drinking.China today shares its
文摘THE first thing Cai Banghong does every morning is to look at the sales figure of various tea brands on his online store. Hmongling Tea Garden, before posting promotional information and collecting customers＇ feed- back to help target his products Cai spent two decades growing his tea company from a small backroom operation into a modern tea company integrating planting, processing and sales.
文摘在茶园水资源管理中,蒸散量(Evapotranspiration,ET)是评估作物水分需求的关键指标,由于茶园蒸散量预测具有时序性、不稳定性以及非线性耦合等特点,目前的茶园蒸散量预测模型存在预测精度较低的问题,针对此问题本文提出了一种新型的茶园蒸散量预测模型。首先使用互信息算法(Mutual information,MI)与主成分分析算法(Principal component analysis,PCA)相融合的数据处理算法(MIPCA),筛选强相关的特征并提取主成分;其次将时域卷积网络(Temporal convolutional network,TCN)与Transformer融合,利用灰狼算法(Grey wolf optimization,GWO)优化超参数,捕捉茶园数据的全局依赖关系;最后整合2个网络构建了MIPCA-TCN-GWO-Transformer模型,通过消融试验和对比试验验证了模型性能,并对模型在不同时间步长下的性能进行测试。结果表明,该模型平均绝对百分比误差(Mean absolute percentage error,MAPE)、均方根误差(Root mean square error,RMSE)和决定系数(Coefficient of determination,R^(2))3个评价指标分别为0.015 mm/d、0.312 mm/d和0.962,优于长短期记忆模型(Long short term memory,LSTM)等传统预测模型。在小时尺度、日尺度和月尺度下的R^(2)分别为0.986、0.978和0.946,在不同时间步长下展现了良好的适应性和准确性。本文构建的MIPCA-TCN-GWO-Transformer模型具有较高的预测精度和稳定性,可为茶园水资源优化管理和灌溉制度制定提供科学参考。
