Combined application of life cycle assessment and neural network model for modeling energy and environmental emissions in sugar beet production

Abstract

• The life cycle and energy of sugar beet production are studied in this study. • Environmental impact categories are identified for sugar beet production using the CML and USEtox methods. • Output energy and environmental impacts are predicted by ANN. • ANN results are more acceptable in modeling. Current agricultural energy management methods do not sufficiently account for the environmental impacts related to farm size. This gap limits the ability to accurately predict the ecological effects of sugar beet cultivation, particularly regarding climate change, eutrophication, acidification, human toxicity, and terrestrial and aquatic ecotoxicity. Moreover, optimizing agricultural productivity and integrating artificial intelligence (AI) are necessary to improve energy management and anticipate future consumption patterns, considering current constraints. Our innovative solution combines LCA with AI to better forecast the ecological impacts of sugar beet cultivation across farms of different sizes. By using historical data and advanced neural models, this method anticipates future energy consumption while integrating environmental and energy constraints. The study highlights that medium-sized farms (MF) show the lowest environmental impacts, according to the CML and USEtox methodologies, unlike small (SF) and large farms (LF), which do not significantly differ in their contribution to pollutants. The average energy consumption for beet production was 1,092,000 MJ/ha on small farms, 1,126,100 MJ/ha on medium farms, and 1,107,080 MJ/ha on large farms. The key critical points identified were related to the use of diesel, machinery, seeds, and pesticides. In terms of energy distribution, chemical fertilizers accounted for 41% of the total energy used, followed by fuel (24%), pesticides (18%), and machinery (7%). Neural models revealed correlation coefficients (R 2 ) ranging from 0.634 to 0.945 for testing, 0.861 to 0.967 for validation, and 0.944 to 0.982 for training. These results suggest good predictive accuracy for our AI-based approach. Finally, to improve energy efficiency, the study proposes modernizing agricultural equipment and developing more sustainable technologies, such as hybrid tractors, robotic and automated weeding systems, and cultivation techniques that reduce carbon emissions.