A novel hybrid machine learning approaches for prediction of greenhouse energy demand and production

Abstract

In response to the growing environmental complexities, the agricultural sector is actively integrating advanced technologies to fortify its adaptability and operational efficiency. A pivotal avenue of exploration centers on exploiting machine learning models for the prediction of greenhouse parameters. This study explores the prediction of greenhouse energy demand and production. For that, the study employs a new hybrid approaches (series and parallel), combining artificial neural networks and boosting trees, to predict air-conditioning electrical consumption and photovoltaic modules' electrical production. Model performance is evaluated based on statistical indicators, including the coefficient of correlation (R) and the normalized root mean square error (nRMSE). Results reveal that series and parallel hybrid models demonstrate acceptable to good performance ($10\%<nRMSE<30\%$), particularly during mid-August to mid-September, influenced principally by external temperature and solar radiation (models inputs). The hybrid model, including series and parallel approaches, exhibits variable performance compared to individual artificial neural networks and boosting trees methods.