Artificial Intelligence in Assessing Electricity and Water Demand in Oilseed Processing

The aim of this study was to explore the use of neural networks as a decision-support tool for sustainable oilseed processing. The investigation focused on how different production profiles (crude vegetable oil, refined oil, hydrogenated oil and margarine) affect electricity and water use in selected Polish processing plants. The collected data were first grouped with cluster analysis to identify similar operational cases. The clusters were then visualized with a Self-Organizing Map (SOM), producing a two-dimensional topological feature map. This analysis indicated a subset of data for which it was appropriate to build predictive models of electricity and water consumption. Multi-layer perceptron (MLP) neural networks yielded highly accurate predictions of electricity (R2 = 0.967 on the test set) and water (R2 = 0.967 on the test set) use in oilseed processing. The resulting models can assist in selecting the most energy- and water-efficient processing configuration.