Integration of machine learning and mathematical programming methods into the biomass feedstock supplier selection process

Recent concerns over the use of and reliance on fossil fuels have stimulated research efforts in identifying, developing, and selecting alternative energy sources. Biofuels represent a promising replacement for conventional fuels for heating and mobility applications, however, variability in the quality and availability of biomass feedstocks greatly affect the utility of biofuels due to the impact on cost and life cycle environmental performance. Thus, methods for mitigating these potential impacts are needed when selecting biomass feedstock suppliers. In the research herein, the selection of the best supplier is investigated for a biomass supply chain (BSC) network by including both qualitative and quantitative factors. Most existing supplier-selection methods consider four steps: (1) Problem formulation, where Decision-Tree Analysis is applied as a qualitative method for defining the type of biomass feedstock materials for biofuel production, (2) Criteria definition, (3) Preevaluation of qualified suppliers, which employs the Support Vector Machine (SVM) method, and (4) Final selection. Integration of machine learning (ML) techniques and a mathematical programming model is undertaken with this method to select the most appropriate feedstock suppliers. It is shown that integrating ML and mathematical programming methods offers a promising approach to supplementing existing supplier selection methods for biomass-to-biofuel supply chains.