论文部分内容阅读
为建立日光温室中短期气温预报模型,以2个冬季生产季的日光温室实时气温观测资料为基础,利用BP神经网络建模和曲线拟合的方法,对日光温室1~7d气温预报模型进行了研究。结果表明:1)以室外气温为输入要素的温室气温预报模型,最高气温预报值与观测值的符合度指数(D)为0.68~0.93,均方根误差(RMSE)为3.1~6.3℃;2)最低气温预报值与观测值的符合度指数(D)为0.81~0.95,均方根误差(RMSE)1.5~2.2℃;3)日光温室内最低气温预报绝对误差小于2℃的预报准确率Rate(≤2℃)为78%~95%;4)逐时气温预报模型预报值与实测值的符合度指数(D)为0.95~0.99,均方根误差(RMSE)为1.0~2.8℃,逐时气温预报模型预测准确率较高。结合目前气象台站“周预报”结果,模型可较准确地预报温室内1~7d最低气温,并模拟日光温室内气温的逐时变化,可为冬季日光温室低温灾害预警及室内气温调控提供有益参考。
In order to establish the medium and short-term temperature forecast model in sunlight greenhouse, based on the observation data of real-time temperature in solar greenhouse in two winter production seasons, BP neural network modeling and curve fitting method were used to forecast the air temperature in 1 ~ 7 days the study. The results showed that: 1) The model of greenhouse temperature forecasting based on the outdoor temperature was 0.68-0.93, the root mean square error (RMSE) was 3.1-6.3 ℃; 2 ) The coincidence index (D) between the minimum temperature forecast value and the observed value is 0.81-0.95, and the root mean square error (RMSE) is 1.5-2.2 ℃. 3) The forecast accuracy rate of the minimum temperature forecast absolute value less than 2 ℃ in sunlight greenhouse Rate (≤2 ℃) was 78% -95%. 4) The coincidence index (D) between forecast value and measured value of hourly temperature forecasting model was 0.95-0.99, and the root mean square error (RMSE) was 1.0-2.8 ℃ When the temperature prediction model prediction accuracy is high. Combining with the current weather station “weekly forecast ” results, the model can predict the minimum temperature in 1 ~ 7 days in greenhouse and simulate the time-dependent changes of air temperature in solar greenhouse, which can be used for early warning of low temperature disasters in winter solar greenhouse and indoor air temperature regulation Useful reference.