In June 2016, floods in Hubei Province, China, knocked out three of the seven collection facilities operated by Hubei NE Biofuel Company, disrupting the entire network and causing an estimated $7.7 million in losses. Events like this are the blind spot of conventional biomass supply chain planning, which assumes collection sites stay operational for their whole life cycle. Combined with the seasonality of agricultural feedstock — in Hubei, autumn yields are four times the winter yields — disruption risk directly threatens the economic viability of biomass-to-energy projects.

To address these challenges, the authors formulate a three-echelon, multi-period mixed-integer nonlinear optimization model that simultaneously determines optimal collection facility locations, biomass sourcing decisions, transportation flows, inventory levels, and backup facility strategies. The formulation explicitly captures seasonal feedstock availability and probabilistic facility disruptions through reliability-based assignments and chance constraints, enabling the design of reliable supply chain networks. 

The optimization model was implemented in AMPL and solved using Artelys Knitro with its default settings. Applied to a real-world biomass supply chain in Hubei Province, China — 35 supply zones, 20 candidate sites, 5 biorefineries and 4 seasons — with operational data provided by Hubei NE Biofuel Company, the optimized solution identified an eight-facility network with backup assignments that significantly improved resilience.  

Compared with traditional disruption-free planning approaches, this design reduced expected supply chain costs from $6.56 million to $5.65 million (-14%), while ensuring reliable biomass deliveries to biorefineries under facility disruption scenarios. The results also show that, once inventory and multi-echelon effects are accounted for, assigning suppliers to their nearest facilities can be suboptimal — a conclusion only reachable by solving the full nonconvex chance-constrained formulation with Artelys Knitro rather than an approximation of it.  

Authors: Zhixue Liu (Huazhong University of Science and Technology), Shukun Wang (Huazhong University of Science and Technology), Yanfeng Ouyang (University of Illinois at Urbana-Champaign).