Development of LSTM-Attention-Based Microelectronic systems for Enhanced Greenhouse Temperature Prediction

Integrating Agriculture, a vital sector for global food security, is increasingly challenged by resource scarcity, climate variability, and rising operational costs. Efficient management of resources, especially water and energy, has become crucial for sustainable farming. This paper investigates integrating microelectronic systems, including microcontrollers and sensors, into greenhouse environments to enhance agricultural practices. Additionally, it proposes a predictive model for greenhouse air temperature using a Long Short-Term Memory (LSTM) neural network combined with an attention mechanism (LSTM-AT). This hybrid model addresses the limitations of traditional LSTM models in handling long-term data improving prediction accuracy. The LSTM-AT model was validated against multiple models, such as GRU and RNN, under varying prediction horizons and weather conditions. Results show that the LSTM-AT model outperforms the alternatives, achieving a minimum R2 of 0.95, a maximum RMSE of 1.35°C, and a maximum MAPE of 12.01%. These findings highlight the potential of microelectronic systems and advanced prediction models to optimize greenhouse environments, reducing energy consumption and increasing agricultural productivity.