Temperature and humidity prediction model based on VMD-LSTM in the edible fungi greenhouse

The greenhouse environment of edible fungi has nonlinear, multi-coupling and time-varying properties, which is important for the cultivation of edible fungi exacting prediction of temperature and humidity changes in the greenhouse environment of edible fungi. In this paper, Spearman is used to analyze the environmental data inside and outside the greenhouse and the switch quantity data of environmental control equipment. Using the time-frequency analysis method VMD (Variational Mode Decomposition) decomposition technology can effectively filter the high-frequency and low-frequency noise in the data signal and decompose the environmental history data of the greenhouse into a series of different sub-modes. It can reduce the complexity of data and extract the essential features of data signals. First, the data is preprocessed by minimum-maximum normalization. Secondly, Recurrent Neural Network (RNN) model, Long Short-Term Memory (LSTM) model, VMD-RNN and VMD-LSTM models are built for predicting the environmental changes of temperature and humidity in the greenhouse, respectively. At last, RMSE, R2 and MAPE are selected to evaluate the above model. The accumulated error analysis of each model is carried out by multi-step prediction method to further validate the robustness of the models that are built. The experimental results show that the temperature and humidity prediction model based on the combination of VMD technique and LSTM neural network model has higher prediction accuracy than the traditional RNN and LSTM neural network. The R2, RMSE and MAPE of the mushroom greenhouse are 0.9407, 1.1651 and 0.0073 for humidity and 0.9891, 0.0984 and 0.0024 for temperature. The RMSE, R2, and MAPE metrics of the VMD-LSTM model are optimized. In the analysis of the model based on the multi-step prediction method each model has the problem of error accumulation but the prediction performance of the VMD-LSTM model is better than the other three models. The results can provide a stable and suitable greenhouse environment for the growth of mushroom house which can improve the productivity and promote the quality of mushroom growth.