**7. Conclusions**

A MINLP model for optimizing a biomass co-firing supply chain network has been developed, which integrates feedstock properties considerations while simultaneously minimizing economic costs and environmental emissions through goal programming. This model incorporates the impact of feedstock, transportation, and pre-treatment requirements. Changes in biomass properties as it moves through the network are accounted for, together with the impact of feedstock properties on conversion yield and equipment degradation. The inclusion of these in the model showed to be an important enhancement to traditional models because decisions on how much and when to source biomass and coal, and the use of pretreatment facilities, storage, and combustion in coal power plants were considerably affected by the said considerations.

Minimizing either the financial or environmental objective individually emphasized the conflicting nature of the two objectives. Simultaneously optimizing both objectives created a network which balanced performance on both objectives.

It was also shown that without considerations for feedstock properties, costs and emissions were artificially decreased, leading to the purchase of insu fficient fuel and combustion of inappropriate fuel which may result in damage or loss in e fficiency of the equipment. Hence, the model proposed in this study is a better fit to design and manage a biomass co-firing network.

Extensions on this research may consider biomass quality and availability as uncertain parameters in a robust multi-objective optimization model. Precise data regarding the quality of the feedstock is not readily available, and any error in estimation requires operational adjustments to be made. As feedstock quality has been proven to be an important inclusion in biomass co-firing networks, these networks should be made robust to such uncertainties. Additionally, this study assumes that the network's nodes are all functional and benign; however, in reality, the presence of faulty and uncooperative components must be considered [31]. Lastly, the parameters used in the validation of the proposed model may be considered too optimistic and di fficult to match in real market. Thus, the application of the model to real-world problems may be explored, along with e fficient solution strategies for the resulting large-scale problems.

**Author Contributions:** Conceptualization, C.L.S.; Data curation, J.L.G.S.J.; Formal analysis, J.L.G.S.J.; Supervision, C.L.S., K.B.A. and R.R.T.; Validation, J.L.G.S.J.; Writing—original draft, J.L.G.S.J.; Writing—review & editing, J.L.G.S.J., C.L.S., K.B.A. and R.R.T.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.
