Solar greenhouse environment prediction model based on SSA-LSTM-Attention

An accurate and rapid prediction of the greenhouse environment is essential to promote the growth and development of crops. The greenhouse microclimates can be characterized by temporality, nonlinearity, and strong coupling among various environmental factors. In this study, a prediction model was introduced for the solar greenhouse environment, according to the integration of SSA-LSTM-Attention (sparrow search algorithm-long short-term memory-attention mechanism). A novel model was developed to reliably predict the key environmental parameters in greenhouses. Thereby data collection was realized in more informed and effective environmental practice. An IoT (internet of things) data acquisition was utilized to gather extensive environmental data from both inside and outside the greenhouse. Firstly, Pearson correlation analysis was applied to the vast amount of data. The most significantly correlated factors were identified with the greatest impact on the greenhouse environment. Thereby the efficiency of the model was improved using variables. Then, a time series dataset of environmental features was constructed to encapsulate the temporal dynamics and interdependencies within the data. This dataset served as the input for the SSA-LSTM-Attention model. The complex patterns and relationships were captured using the dataset. Three components were selected: the sparrow search algorithm (SSA) for the hyperparameter optimization, the Long short-term memory (LSTM) network for the temporal dependencies, and the attention mechanism for the most relevant information within the input sequence. Four environmental factors were predicted in the solar greenhouses: indoor temperature, indoor humidity, light intensity, and soil moisture content. The SSA-LSTM-Attention model was achieved in an exceptional average fitting index of 97.9%. This outstanding performance was achieved the best, compared with the benchmark models, including BP (back propagation neural network), GRU (gate recurrent unit), LSTM, and LSTM-Attention (long short-term memory-attention mechanism). Specifically, the SSA-LSTM-Attention model also outperformed the BP, GRU, LSTM, and LSTM-Attention by 8.1, 4.1, 3.5 and 3.0 percentage points, respectively, in terms of prediction accuracy. Furthermore, a remarkably low average absolute percentage error of 2.6%, which was significantly lower than the errors recorded by the four control models, with reductions of 6.5, 3.2, 2.8 and 2.5 percentage points, respectively. The experimental results demonstrated that the sparrow search algorithm was optimized to improve the LSTM prediction accuracy by the attention mechanism. In conclusion, the sparrow search algorithm was integrated with the automatic hyperparameter optimization of the LSTM-Attention model. Advanced machine learning can be expected to serve as a complex agricultural challenge. The findings can provide invaluable data support for the environmental regulation in solar greenhouses. The model can hold the promise to more precisely control and optimize the critical environmental influencing factors on crop growth. The resource use efficiency significantly enhanced the crop yields. Ultimately, a great contribution can be gained in the sustainable greenhouse.