基于邻域粗糙集的气象因子选择在虾塘水温预测中的应用

Application of meteorological factor selection based on neighborhood rough set in shrimp pond water temperature prediction

基于邻域粗糙集对影响虾塘水温变化的气象因子进行选择,并选取模型预测虾塘水温,为南美白对虾养殖趋利避害提供科学参考。首先,将平均气温、最高气温、最低气温、降水量、气压、2 min风速、10 min风速和瞬时风速等8个气象因子组合输入SFNN模型(单隐层前馈神经网络模型)、高斯回归模型和岭回归模型进行虾塘水温预测,选取预测效果最好的SFNN模型为本研究预测模型。然后,运用邻域粗糙集和熵理论,考虑气象因子和虾塘水温之间的相关性、冗余性和交互性,选出影响虾塘水温变化的主要气象因子。最后,利用选出的主要气象因子和SFNN模型实现虾塘水温预测。将基于邻域粗糙集选出的5个气象因子组合与8个气象因子组合,以及8个气象单因子分别输入SFNN模型,预测结果表明:邻域粗糙集选出的5个气象因子组合预测结果最好,其预测均方根误差、均方误差、平均绝对误差最小,分别为1.1211、1.2569和0.8938,决定系数(R2)为0.7916;在气象单因子中,气压对虾塘水温的预测结果较好。因此,基于邻域粗糙集选出的5个气象因子组合,通过SFNN模型进行虾塘水温预测结果最好,此方法在南美白对虾养殖趋利避害、防灾减灾中具有一定的实用价值。

Based on the neighborhood rough set,the meteorological factors affecting the change of shrimp pond water temperature were selected.A model was selected to predict the water temperature of shrimp pond,which provided a scientific reference for shrimp culture to seek advantages and avoid disadvantages.Firstly,eight meteorological factors,including average temperature,highest temperature,lowest temperature,amount of precipitation,barometric pressure,two-minute wind speed,ten-minute wind speed,and instantaneous wind speed,were combined into the single hidden layer feed forward neural network(SFNN)model,Gaussian regression model,and bridge regression model to predict the water temperature of shrimp ponds.SFNN model with the best prediction effect was selected as the prediction model.Then considering the correlation,redundancy and interaction between meteorological factors and shrimp pond water temperature,the main meteorological factors affecting the shrimp pond water temperature were selected by using neighborhood rough set and entropy theory.Finally,the water temperature of the shrimp pond was predicted by using the selected meteorological factors and SFNN model.Combination of five meteorological factors selected based on neighborhood rough set,the combination of eight meteorological factors,and eight single meteorological factors were input into the SFNN model,respectively.The prediction results showed that,the combination of five meteorological factors selected by neighborhood rough set had the best prediction results,and its root mean square error,mean square error and mean absolute error were the smallest,which were 1.1211,1.2569 and 0.8938,respectively.And the determination coefficient was 0.7916.Among single meteorological factors,atmospheric pressure predicted best on shrimp pond water temperature.Therefore,the combination of five meteorological factors selected based on the neighborhood rough set had the best prediction in shrimp pond water temperature by the SFNN model.This method had practical value in seeking advantages and avoiding disadvantages,and disaster prevention and reduction in prawn breeding.