Enhanced Predictive Modeling of Greenhouse Temperature Using Power Long Short Term Memory (PLSTM) Model

Abstract

It is important to have a precise and dynamic control of greenhouse environments in order to optimize crop growth and resource efficiency. Traditional forecasting methods often fall short of capturing the complexities of greenhouse systems, leading to inaccurate predictions. To address this challenge, we propose a novel approach utilizing Machine Learning (ML) algorithms to predict greenhouse internal temperature accurately. Our study evaluates the performance of Linear Regression (LR), Extreme Gradient Boost (XGBoost), and our proposed Power Long Short-Term Memory (PLSTM) model, using data from two distinct greenhouse databases. To develop predictive models capable of forecasting internal temperature for continuous and ef-fective climate management, we explore the correlations between various greenhouse-related climate parameters. The PLSTM model demonstrated higher performance, achieving coefficient of determination (R2) values of 0.9710 for Greenhouse 1 and 0.9999 for Greenhouse 2. These results underscore the potential of PLSTM for accurate greenhouse climate prediction,paving the way for advanced greenhouse management strategies that enhance crop yield and sustainability.