BiLSTM networks for solar radiation forecasting in greenhouse environments

This paper investigated an innovative prediction approach using the bidirectional long short-term memory (BiLSTM) model for the forecasting of solar radiation in the smart farm in Naju, South Jeolla Province. Accurate solar radiation prediction holds substantial significance in facilitating efficient solar power generation, significantly impacting agricultural sustainability and resource utilization. This study rigorously assessed the effectiveness of various predictive models for forecasting solar radiation, a crucial factor in optimizing solar energy utilization in greenhouse settings. The variety of statistical and machine learning models aims to improve the accuracy of solar radiation prediction, which is crucial for precise energy management and cultivation practices. An hour-ahead mean absolute error (MAE) of 13.53 demonstrated the proposed BiLSTM model's impressive forecasting capabilities. The comparative analysis among individual models then revealed that the LSTM model achieves an MAE of 15.23, extreme gradient boosting (XGBoost) at 24.64, and autoregressive integrated moving average (ARIMA) at 28.52. This exploration underscored the important role of accurate solar radiation forecasting, specifically highlighting the effectiveness of the BiLSTM model in optimizing solar power generation within greenhouse environments. Its importance extended to the promotion of sustainable agricultural practices and the improvement of resource-efficient energy control strategies.