Biofuel, an efficient alternative to fossil fuels, has gained considerable attention as a potential source to satisfy energy demands. Biomass collection and distribution typically incur a significant portion of the biofuel production cost. Thus, it is imperative to design a biofuel supply chain network that not only aims to minimize the delivery cost but also incorporates biomass quality properties that make this raw material so unique yet challenging. This paper proposes a novel two-stage stochastic programming model that captures different time- and weather-dependent biomass quality parameters (e.g., the moisture content, ash content, and dry matter loss) and their impact on the overall supply chain design. To efficiently solve this optimization model, we propose a parallelized hybrid decomposition algorithm that combines the sample average approximation with an enhanced progressive hedging algorithm. The proposed mathematical model and solutions are validated with a real-life case study. The numerical experiments reveal that the biomass quality variability impacts the supply chain design by requiring additional depots, and therefore, it increases the capital investment. The storage of unprocessed biomass at depots and biorefineries decreased by 88.5% and 97.9%, respectively, and the densified biomass inventory at biorefineries increased 17-fold when baseline quality considerations were taken into account.

This study presents a mathematical model that designs a reliable multi-modal transportation network for a biofuel supply chain system while site-dependent facility failure and congestion are taken into consideration. The proposed model locates the multi-modal facilities and biorefineries and determines the optimal production, storage, and routing plans in such a way that the overall system cost is minimized. We propose a hybrid Constraint generation-based Rolling horizon algorithm to solve this challenging $\mathcal{NP}$-hard problem. The performance of this algorithm is tested in a example case study with numerical analysis showing that the hybrid algorithm can find near-optimal solutions to large-scale problem instances in a reasonable amount of time. Results indicate that the effect of congestion reduces the usage of multi-modal facilities in the biofuel supply chain network while bio-refineries and multi-modal facilities tend to move away from coastal areas when disruption probabilities are taken into consideration.