Row and column-wise robust optimization model for biorefineries storing perishable biomass under weather uncertainty: Boosted by machine learning

Abstract

This study builds upon our earlier research (Razm et al., 2023). This study makes several contributions. First, we define three weather criteria (Rainfall, Temperature, Daylight hours) to incorporate weather conditions into the optimization model. We gather real weather data and conduct data preprocessing. Next, numerous calculations are performed based on the criteria to determine biomass availability ranges. Second, we immunize our system against uncertainties; however, uncertain parameters in our model possess specific features. Uncertainties exist both in the rows and columns. The traditional method cannot effectively address this issue. Therefore, we propose a row and column-wise robust optimization model to tackle weather and price uncertainties. Third, incorporating the aforementioned contributions into our previous model presents challenges. The new model is complex. Analyzing its behavior and interpreting results are challenging for this study. However, we conduct a series of numerical experiments and extract valuable managerial insights. Results show that despite incurring extra costs initially, the manager stands to gain more profit in the future, attribute to the system’s robustness. Finally, we enhance our model and increase system profitability by adopting data-driven robust optimization based on Machine Learning.