Enhancing Agricultural Sustainability Through Intelligent Irrigation Using PVT Energy Applications: Implementing Hybrid Machine and Deep Learning Models

This research focuses on developing an intelligent irrigation solution for agricultural systems utilising solar photovoltaic-thermal (PVT) energy applications. This solution integrates PVT applications, prediction, modelling and forecasting as well as plants’ physiological characteristics. The primary objective is to enhance water management and irrigation efficiency through innovative digital techniques tailored to different climate zones. In the initial phase, the performance of PVT solutions was evaluated using ANSYS Fluent software R19.2, revealing that scaled PVT systems offer optimal efficiency for PV systems, thereby optimising electrical production. Subsequently, a comprehensive approach combining integral feature selection (IFS) with machine learning (ML) and deep learning (DL) models was applied for reference evapotranspiration (ETo) prediction and water needs forecasting. Through this process, 301 optimal combinations of predictors and best-performing linear models for ETo prediction were identified. Achieving R 2 values exceeding 0.97, alongside minimal indicators of dispersion, the results indicate the effectiveness and accuracy of the elaborated models in predicting the ETo. In addition, by employing a hybrid deep learning approach, 28 best models were developed for forecasting the next periods of ETo. Finally, an interface application was developed to house the identified models for predicting and forecasting the optimal water quantity required for specific plant or crop irrigation. This application serves as a user-friendly platform where users can input relevant predictors and obtain accurate predictions and forecasts based on the established models.