*2.1. Methods Used in Biofuels Upstream Supply Chain Planning*

The biofuel supply chain planning and management are influenced by various factors, among which are included biomass availability, choice of biomass cultivation type, harvesting, transport mode of the biomass, pre-treatment facility, biomass conversion technology type, conversion facility location and capacity, product storage facility, product distribution, routing and inventory. There are many decisions, of major and minor, to be made across the whole supply chain planning starting from crop selection to final product as covered in a review conducted by De Meyer et al. [12] and Atashbar et al. [13], in this case, biofuel consumption. There are three primary decision-making levels in supply chain planning and management: strategic, tactical and operational level (Figure 2). The strategic decision-making level refers to a decision where long-term investment is involved [12], for instance, location, capacity and type of storage, pre-processing, conversion facility, transportation mode (investment in ownership of the vehicle), crop selection. Tactical decision-making level refers to medium-term decision decisions, usually monthly or weekly and are within the constraint of a strategic decision [12]. Some examples of tactical decisions are harvest planning, inventory planning, transport routing and shipment capacity. Operational decision-making level usually refers to a decision over a short time frame, ranging from hourly to weekly that is within the limit of a tactical decision [12]. For instance, inventory planning and transport scheduling that ensures undisrupted and efficient operations of plants and processes at another facility through the supply chain.

Under each decision-making level, three main approaches are applied in solving the supply chain problem: mathematical programming, heuristic approaches, multicriteria decision analysis. Mathematical programming refers to mathematical models that represent real-world problems. The model is solved by optimising the objective function. For instance:


While the mathematical programming method aims to identify the optimal solution by solving the equations simultaneously, the heuristic approach looks for an optimal point to a defined problem through a stepwise approach. According to a review conducted by De Meyer [12], three different heuristics algorithms are observed: genetic algorithm, particle swarm optimisation and binary honeybee foraging for upstream biomass planning and management optimisation. Multicriteria decision analysis is also one of the methods used for decision making in supply chain generally. According to Roy [14], multicriteria decision analysis is defined as 'a decision aid and a mathematical tool allowing the comparison of different alternatives or scenarios according to many criteria, often conflicting, in order to guide the decision-maker towards a judicious choice.

Thirty-two scientific publications could be found between 2014–2019 reviews the methods used for decision making at strategic and tactical levels under respective problem statement of the biofuel supply chain, with some exceptions for publications before 2014 for its method development related to recent publications. It shows a summary of recent scientific publications on decision making of biofuel supply chain (Table 3) It is observed that most biofuel supply chain planning and management problems are formulated as a Mixed Integer Linear Programming (MILP) problem, covering 21 papers out of 32 articles. Aguayo et al. [15] present a MILP model to address a static and dynamic corn-stover harvest scheduling problem in cellulosic ethanol production by minimising system cost. Ahn et al. [16] developed a mathematical programming model for strategic planning and design of microalgae biomass-to-biodiesel supply chain which minimise the total cost of the supply chain by taking resource constraints, demand constraints and technology into accounts. The model is then applied in a case study of the biodiesel market in South Korea.

**Figure 2.** Decision-making levels of biofuel supply chain.

Cuˇ ˇ cek et al. [17] presented a multi-period MILP model for regional bioenergy supply network optimisation with sources of biomass from first, second and third generations of biofuels. The model enables strategic decision on raw material sourcing, conversion technology selection by maximising economic performance. Foo et al. [18] developed a linear programming (LP) model to identify optimal feedstock allocation of a direct biomass source-sink allocation. The model is extended to a MILP model for pragmatic decision planning in biomass supply chain logistics, where a minimum threshold quantity is met in an optimal network. Miret et al. [19] developed a multi-period MILP model that proposes optimal bioethanol supply chain design with the account to facility location, process selection and inventory policy. This study compared the economic, environment and social aspect of bioethanol production from first and second-generation biomass. Ng and Maravelias [20] presented a multi-period MILP model for biofuel supply chain design, taking a case study in Wisconsin. Ng et al. [21] developed a MILP model which maximises biomass utilisation of rubber seed oil to produce biodiesel that considers centralised and decentralised rubber seed processing facilities. Rabbani et al. [22] developed a MILP model that selects optimal biomass pre-processing plant, biofuel plant and biofuel storage warehouse. Santibañez et al. [23] presented a multi-objective, multi-period MILP model that seeks to optimise the biorefinery supply chain in fulfilling ethanol and biodiesel demands in Mexico with consideration of economic, environmental and social criteria.

Other thanMILP, some problems have been formulated as a mixed-integer non-linear programming (MINLP) model and then linearised to form the MILP model. López et al. [24] developed an MINLP optimisation model for biorefinery system design while taking the interactions of the biorefinery system with the surrounding watershed into account. The non-linear term came from the exponent to represent the economy of scale of biorefinery plant. The model is then solved as MILP by fixing the exponent as one. How et al. [25] developed a MINLP model that solves biomass supply chain synthesis problem by maximising overall profit and minimising CO2 emission through introduced CO2 penalty. Santibañez et al. [26] developed an MINLP model to identify the optimal configuration

of a distributed biorefineries system, where the model can decide utilised raw materials, processing technologies, processing facilities and manufactured products.

It is observed that most supply chain optimisation problems are solely focused on optimising the economic benefits. Environmental, energetic and social factors deserve more attention in optimisation as these factors weight significantly when it comes to strategic decision making; this is especially true for the biofuel policymaker. To address this limitation, some authors used Pareto optimisation approach to obtain an optimal solution under the multi-objective situation. Liu et al. [27] present a multi-objective MILP model of biofuel conversion pathways with accounts to economic, energy and environmental criteria. Economic criteria are measured by total annual profit, energetic criteria are measured by fossil energy input per MJ biofuel, while environmental criteria are measured by GHG emissions per MJ biofuel. Pareto optimal surfaced is obtained to study the trade-offs between the three criteria. Miret et al. [19] applied the same approach through the epsilon-constraint method to study the multi-objective bioethanol supply chain model with a case study in France. Orjuela et al. [28] developed a multi-objective LP model that analyses biodiesel supply chain taking into consideration of economic, environment and social aspect, where the social aspect is taken from the perspective of food security concern. The epsilon-constraint method is then used to study the trade-off between multiple criteria. Osmani and Zhang [29] used an augmented epsilon-constraint method to solve multiple objective models with more than two objectives. Pareto optimal solution is also obtained in Santibañez et al. [23] multi-objective study.

It is undeniable that in the biofuel supply chain, which has a complex upstream supply chain, there are many uncertainties that could affect the performance of the supply chain, among which are included crop yield, potential disruption situations like pest attacks, floods or droughts, and biofuel price uncertainties. In order to address these uncertainties, a number of publications are found to have taken these factors into account. Azadeh et al. [30] present a stochastic multi-period MILP model that maximise the profit of a biofuel supply chain with prices of biofuels assumed to be stochastic. A case study is then conducted for the biofuel supply chain in Iran. Bairamzadeh et al. [31] proposes a hybrid robust MILP model for bioethanol supply chain design and planning with considerations of different types of uncertainties which included randomness, epistemic and deep uncertainties. Ghelichi et al. [32] developed a two-stage multi-period stochastic MILP model for biodiesel supply chain design with Jatropha curcas as feedstock, under consideration of feedstock supply and product demand uncertainties. Maheshwari et al. [33] developed biofuel supply chain resiliency optimisation model that consider no disruption and disruption scenarios during the flood, drought, pest attack, equipment failure, each weighted by their probability of occurrence. Mohseni and Pishvaee [34] present a robust supply chain optimisation model that manages complexities in strategic and tactical planning of microalgae-based biofuel production. The supply chain is designed based on batch and continuous production system. Optimal scale for the batch system is determined by a trade-off between the cost of biofuel production, transportation and risk mitigation. Osmani and Zhang [29] presents a multi-objective, multi-period optimisation model of a second-generation biofuel supply chain under switchgrass yield, bioethanol demand and bioethanol sale price uncertainties. Santibañez et al. [35] presented a stochastic multi-period optimisation model that identifies optimal biorefinery supply chain planning under raw material price uncertainty considering environmental and economic aspects. A multi-scale multi-period MILP model is developed by Sharifzadeh et al. [36] to identify the optimal supply chain design of biofuel production using fast pyrolysis under consumer demands and biomass availability uncertainty.

Some authors integrated graphical approaches with mathematical programming in solving a biomass supply chain problem. Fan et al. [37] developed a novel graphical decision-making tool that allows the selection of transportation mode with lower environmental burden and energy consumption. Lam et al. [38] proposed a two-level graphical strategy for optimal regional-level biomass energy supply chain networks synthesis that minimises total carbon emissions footprint followed by optimal

biomass supply chain network synthesis within the region. The graphical approach was then applied to a case study at the Central European region.

Some authors integrated GIS functions to characterise the related biofuel supply chain when formulating the problem, especially in determining facility location. Harahap et al. [39] present a policy analysis using a spatially-explicit MILP model to optimise the overall palm oil supply chain—not just solely on palm oil biofuel supply chain in Sumatra, Indonesia. Hoo et al. [40] presented a spatial-economic optimisation MILP model to identify biomethane production plants with the aid of GIS network analysis. Zhang et al. [41] developed a GIS integrated optimisation model in designing a bioethanol feedstock supply chain, which allows the selection of facility location by minimising total system cost. The model is then applied on a case study in the northern part of Michigan's Lower Peninsula, in the US.

Other than mathematical programming, a heuristic approach is also used by some authors to solve complex biofuel supply chain models. Note that the heuristic approach looks for satisfactory solutions, not necessarily an optimal solution and often presents reduced runtimes in solving models [12]. Asadi et al. [42] developed a multi-objective metaheuristic algorithm for algae biofuel supply chain design with an integrated formulation of inventory, routing and location decision under demand uncertainties. Marufuzzaman et al. [43] developed a two-staged (strategic level decision making followed by tactical decision making) stochastic model that assists the design and management of biodiesel supply chain by taking feedstock and technology uncertainties into account. The problem is then solved using an algorithm that combines Lagrangian relaxation and L-shaped solution methods. Poudel et al. [44] present a hybrid decomposition algorithm in solving an optimisation problem when studying the impact of disruption and congestion at the facility of a biofuel supply chain. While for multicriteria decision analysis, Nana et al. present a spatial explicit biodiesel supply chain optimisation model that was solved using an analytical hierarchy process (AHP). How and Lam [45] proposed a multi-objective optimisation (MOO) solution of biomass supply chain management (SBSCM) through AHP that integrates both economic and environmental factors. Among which the environmental factors included abiotic depletion potential (ADP), acidification potential (AP), aquatic toxicity potential (ATP), global warming potential (GWP), land footprint, nutrification potential (NP), ozone depletion potential (ODP), photochemical ozone creation potential (POCP), terrestrial toxicity potential (TTP), water footprint. A general method is also observed in Ng and Maravelias [46] study on the biofuel supply chain. The authors applied a systematic method to compare and investigate the economic performance and energy efficiency of the biofuel supply chain under various configurations and transportation modes. The author introduced hybrid configurations that can potentially improve economic performance and energy efficiency of different supply chain configuration. The findings of the result form the basis for larger-scale biofuel supply chain optimisation model in a future study.

It is observed that there is increasing research and publications on operational biofuel supply chain methods in China and Iran, especially for second-generation and third-generation biofuel production. Waste cooking oil as a feedstock for biofuel production is found in China biofuel supply chain planning (Table 4). Waste cooking oil can be converted into biofuel through hydro-processed esters and fatty acids (HEFA) conversion process. Advanced biofuel produced from waste cooking oil through HEFA process is also considered as Carbon Offsetting and Reduction Scheme for International Aviation (CORSIA)-eligible fuels, which it can be used as sustainable aviation fuel for aircraft. A number of supply chain planning involving microalgae are also found in recent literature. While crop selection in different continents varies geographically, seasonality also affects feedstock supply, as indicated in the most multi-period mathematical models. This is especially true for the northern hemisphere. Despite being primary world producers of biofuel [1], there are minimal scientific publications on supply chain planning from countries in Southeast Asia region found in the literature, except for [38] who have done extensive operational supply chain research on biomass in Malaysia. However, publication solely focusing on biofuel supply chain with case study characterising biofuel production in this region is missing.




 region


Maheshwari et al. [33]

 S

 EC




Strategic:

 facility & biorefinery

MP - MILP

 Southern Illinois, USA

plant


 part of Mexico





*Energies* **2020** , *13*, 1799


**Table 4.** Feedstock categorisation.



### *Energies* **2020** , *13*, 1799

SF—sunflower;

 J—Jatropha;

SG—sorghum;

 WB—woody

 biomass;

SGS—switchgrass;

M—Miscanthus;

SFR—safflower;

MA—microalgae;

 NF—not specified.
