Prediction of greenhouse temperature and relative humidity is extremely important as it allows for the prediction of environmental parameters for manual intervention ahead of time.
However, temperature and relative humidity prediction systems face two important limitations. Inconsistent time resolution in data collection and the lack of standardized protocols for environmental data collection. These issues collectively lead to heterogeneous control strategies that undermine system interoperability in agricultural applications. This study predicted temperature and relative humidity at different time intervals in greenhouses in South China using BPPSO, LSSVM, and RBF models, demonstrating the advantages of temperature and relative humidity prediction. The results showed that the R² of the temperature and relative humidity predictions gradually increased with shorter time intervals, with a 15-min interval achieving the maximum value. The R² for the temperature predictions from the three models were 0.923, 0.923, and 0.912, while the R² for the relative humidity predictions were 0.948, 0.952, and 0.948, respectively. The accuracy of the relative humidity prediction was higher than the accuracy of the temperature prediction. All three models can be used to predict temperature and relative humidity in greenhouses in southern China, with the LSSVM showing higher R² than the other two models. When the time interval was 15 min, the temperatures MAE, MAPE, and RMSE were 0.574, 1.941, and 0.867, respectively, while for relative humidity they were 2.747, 3.383, and 3.907, respectively. In this study, we concluded that LSSVM models with a time interval of 15 minutes are suitable for predicting temperature and relative humidity in greenhouses in South China.
This study provides a reference for early interventions in greenhouse temperature and relative humidity control.
Source: Nature Magazine
