Abstract:

Panax notoginseng has extremely high medicinal and economic value, likes temperature and humidity, temperature and humidity are important environment parameters affecting its growth. At present, the facility cultivation of Panax notoginseng is mainly based on artificial experience, and there is a serious lag in the regulation of the facility environment, which leads to Panax notoginseng being susceptible to diseases and insect pests resulting in yield reduction, and seriously hindering industry development. In this study, four deep machine learning algorithms, Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Long Short−Term Memory Neural Network (LSTM) and Informer model, were used to preliminarily optimize the temperature and humidity prediction model of Panax notoginseng, and an improved PCA−Informer model was constructed to enhance training efficiency and model performance. At the same time, environmental monitoring sensors were deployed for data collection was achieved through environmental monitoring sensors, and the PCA−Informer model was integrated into the platform software, and the main functional modules of the platform were realized by using the Django framework combined with Python technology to develop a platform for environmental monitoring, temperature and humidity prediction and early warning systems of Panax notoginseng facility cultivation. The results showed that: (1) the informer model had the highest prediction accuracy compared with the other three deep machine learning algorithms, the mean absolute error (MAE) of air temperature and humiditywas 0.860°C and 3.870pp, and the coefficients of determination (R²) was 0.959 and 0.964, respectively. (2) The PCA−Informer model constructed by adding Principal Component Analysis (PCA) algorithm to the Encoder layer of the Informer model could improved the training efficiency and performance of the temperature and humidity prediction model of facility cultivated Panax notoginseng. Compared with the Informer model, the MAE of the PCA−Informer model for predicting air temperature and humidity was reduced by 0.140°C and 0.621pp, respectively, and the R² was increased with 0.0100 and 0.0021, respectively. (3) The environmental monitoring, temperature and humidity prediction and early warning platform of Panax notoginseng facility cultivation realized the accurate prediction and early warning of temperature and humidity of Panax notoginseng in the next 3 days.

Key words: Panax notoginseng, Facility cultivation, Deep learning, Temperature and humidity prediction, PCA?Informer model, Early warning platform