山西日光温室逐日极端气温预测模型研究

Forecast Model of Daily Extreme Temperature in Solar Greenhouse in Shanxi Province

利用山西省忻州市日光温室的室内小气候观测数据及气象站资料,用BP神经网络及逐步回归法建立多种输入变量不同天气条件下的日光温室内最高温度、最低温度的模型。结果表明:利用BP神经网络及逐步回归法建立的模型R2均在0.96以上,RMSE与AE大部分在2℃之下,模拟精度较高。利用逐步回归方法在模拟日光温室内晴天最高、最低温度和寡照的最高温度精度较高,利用BP神经网络模型在多云天气的最高、最低温度与寡照的最低温度模拟的精度较高。选择精度更好的模型对日光温室的极端气温做准确的预测,可为山西省设施农业的管理和调控及小气候预报提供支持。

Based on the meteorological data both inside and outside the solar greenhouse in Xinzhou, Shanxi Province, the minimum and maximum temperature forecast model was established with BP neural network and stepwise regression model. The results showed that R2 was higher than 0.96 and most of the root mean square error （RMSE） and absolute error （AE） was lower than 2℃ on BP neural network model. The precision of simulation was good. The precision of stepwise regression model was higher than that of BP neural network model in minimum and maximum temperature of clear day and maximum temperature of overcast day. The precision of BP neural network model was higher than that of stepwise regression model in minimum and maximum temperature of cloudy day and minimum temperature of overcast day. The model with more precision could predicate extreme temperature in solar greenhouse. The model could provide scientific basis for facility agriculture management and environment regulation and microclimate prediction in Shanxi Province.