Spatial Optimization of Photovoltaic-Based Hydrogen-Electricity Supply Chain through an Integrated Geographical Information System and Mathematical Modelling Approach

Hydrogen is a potential energy carrier for renewables that has a clean emission during the point of use. To implement hydrogen energy system in large-scale, a comprehensive hydrogen supply network should be built to supply the hydrogen with optimal infrastructure arrangement. Although the optimization of hydrogen supply chain has been extensively studied, the investigation of an integrated hydrogen-electricity supply chain is still lacking. Considering the interconvertibility of hydrogen and electricity, this study presents a spatial optimization framework that integrates geographical information system with mathematical modelling for the design and optimization of a photovoltaic-based hydrogen-electricity supply chain. The proposed framework allows the concurrent targeting of vehicle fuel and electricity demands as well as the identification of suitable locations for supply chain infrastructures. Case study results showed that the minimum cost of hydrogen-electricity supply chain is about 14.9 billion USD/y assuming two days of autonomy, and the cost of battery constitutes 43% of the total supply chain cost. When the days of autonomy is 8 and above, electricity storage in the form of hydrogen and reconversion through fuel cell is preferred.


Nomenclature Sets
Set of hydrogen in different storage and transportation forms Set of study regions Set of study regions, identical to set Parameters

Capacity of solar panel at region j kWp
Annualized capital cost of supply chain USD/y Annual operating cost of supply chain USD/y 1A9 Total cost of conventional energy use in a year USD/y ^\1 Total cost of supply chain in a year USD/y Annualized replacement cost of supply chain USD/y

Introduction
Hydrogen energy could play a major role in addressing the energy transition as it provides great storage and flexibility capacities that cannot be achieved with electricity. It is a versatile fuel that can be synthesized from various energy sources such as renewables and fossils. In addition, hydrogen can be stored in large quantities for long-term, and the stored energy can be released through combustion or electrochemical conversion (Martin et al. 2020). It is superior to most conventional fuels with its high heating value and clean emission (Acar and Dincer 2018).
As hydrogen only emits water at the point of use, synthesizing hydrogen from renewable energy sources would make the entire energy system clean and sustainable. While fossil fuels have been dominating global hydrogen production, IEA (2019a) reported that the declining costs in solar PV and wind electricity have grown the interest in electrolytic hydrogen, and there have been several demonstration projects in recent years. Hydrogen is deemed to decarbonize global energy use through various applications, for instance, as an alternative fuel for automobiles or temporary energy storage for renewables (Hydrogen Council 2017). The excess electricity produced from renewables can be converted to hydrogen via electrolysis and the produced hydrogen can serve as backup power for the energy system.
Before hydrogen can be applied in the energy sector, it has to be produced and processed from raw materials. A typical hydrogen supply chain (HSC) consists of energy sources, production technologies, storage and transportation, as well as the final utilization of hydrogen energy. The optimization of HSC is crucial to ensure smooth product logistics and a balanced supply-demand with minimum investment cost.

Past Studies on Hydrogen Supply Chain Optimization
Optimization method is the most popular choice for the design and modelling of HSC to determine the best configuration that fulfil the predefined economic, safety, or environmental criteria (Dagdougui 2012). As mathematical optimization allows the modelling of complex HSC subjecting to a series of design variables (Maryam 2017), many studies have employed it for the optimization of hydrogen supply chain (Won et al. 2017, Li et al. 2020, Hwangbo et al. 2017. Unlike mathematical optimization model, a geographic information system (GIS) based approach is not generic but dependent on the spatial conditions at a national or regional scale such as transportation network, land use and population. According to Dagdougui (2012), both optimization and GIS-based approaches have their strengths and weakness: optimization model allows complex modelling of HSC but the spatial conditions are neglected, while a GIS-based model considers the spatial conditions but is only applicable for simple problems. Therefore, an integrated mathematical optimization and GIS model can complement the strengths of both methods.
In the past, there have been several studies that coupled GIS with mathematical optimization model. Johnson and Ogden (2012) used GIS to develop a hydrogen demand model which is based on population data, fuel cell vehicle efficiency, average distance travelled and the per-capita vehicle ownership. The candidate pipeline network is defined following the existing pipeline. Almaraz et al. (2015) employed GIS to locate the existing hydrogen infrastructure and compute the delivery distance using the road network. The result obtained in mathematical optimization (selected energy sources, production and storage capacities, selected refuelling stations) will be displayed on the map using ArcGIS. Samsatli et al. (2016) developed an energy network model known as STeMES to optimize the design and operation of integrated wind-hydrogen-electricity networks for decarbonizing the domestic transport sector of Great Britain. The land availability for onshore wind turbine siting is decided by geographical factors such as wind speed, slope of land, accessibility to road network, connectivity to national grid, protected areas, as well as distance to human activities and wildlife. Welder et al. (2018) conducted a spatial-temporal optimization of a future energy system in Germany for power-to-hydrogen application in mobility and industry. The land available for onshore wind turbine placement is assessed by eliminating water bodies as well as setting up buffer zones around water bodies, residential areas, commercial areas, industrial areas, and critical infrastructures such as power lines, major roads, and railways. In addition, the protected land, land that is far from road networks, and the region with low wind speed are excluded during the land eligibility analysis. The pipeline routes are specified to be built along high pressure compressed natural gas, highways, or railway routes. Through the literature review, several research gaps have been identified: • The lack of integration of hydrogen supply chain with other supply chains. As hydrogen supply system is not standalone, its integration with other supply chains is an important research direction, and the critical factor for such integration is to identify the appropriate "insertion points" through which the supply chains are connected (Li et al. 2019).
• Most site suitability analyses were focusing on wind turbine placement but the determination of suitable areas for photovoltaic (PV) system and supply chain infrastructure placement is yet to be investigated.
• Most hydrogen supply chain studies only consider conventional hydrogen storage and transportation methods (compressed gas or cryogenic liquid), but not the other possible options such as liquid organic hydrogen carrier (LOHC).
• Most hydrogen supply chain studies only consider fixed days of autonomy/inventory but the effect of increasing and decreasing the storage requirement is not being evaluated.
To address the identified research gaps, this study aims to develop a comprehensive framework for the spatial optimization of PV-based hydrogen-electricity supply chain (HESC). In the proposed framework, site suitability analysis is conducted using GIS to determine feasible areas for the placement of supply chain components. The optimal siting and capacities of supply chain infrastructures for concurrent targeting of vehicle fuel and electricity demands are then determined through a mathematical optimization model. In mathematical modelling, the storage and transportation of hydrogen in the form of compressed gas, cryogenic liquid or LOHC will be considered. Moreover, the influence of days of autonomy assumption on the optimal modes of production, storage and transportation of hydrogen and electricity is investigated.

Methodology
In this study, a optimization framework comprising GIS and mathematical modelling is proposed for the optimization of HESC that caters for vehicle fuel and electricity demands. Figure 1 shows the superstructure of the HESC which involves electricity production from solar radiation through PV system. The electricity produced can be exported to other regions via power cables or converted into hydrogen via electrolyzer. Hydrogen produced can be stored in gaseous form (GH2), liquid form (LH2) or bonded with LOHC upon processing, and transported to other regions through trucks or pipelines. There are four types of refuelling stations based on the mode of hydrogen transportation, either in the form of GH2 trailer, LH2 trailer, LOHC trailer or pipeline. The electricity demand of a region can be fulfilled by the PV electricity, imported electricity or electricity produced from hydrogen. Hydrogen reacts with oxygen in fuel cell to produce electricity.
Meanwhile, the hydrogen demands in refuelling stations are satisfied by the hydrogen produced from PV electricity or imported electricity. Figure 1: Superstructure of PV-based hydrogen and electricity production system Figure 2 illustrates the workflow of the proposed framework, which comprises a two-stage approach for the spatial optimization of HESC. In the first stage, GIS is employed for spatial data processing and site suitability analysis. Spatial information such as land use, road network, elevation and slope are being used to determine the potential areas for infrastructure placement.
Meanwhile, spatial population data is used to estimate regional vehicle fuel and electricity demands. The transportation distance between study regions will be determined using network analysis tool in ArcGIS. Second stage of the framework involves the formulation of a mathematical optimization model to identify the optimal HESC configuration for targeted fuel and electricity demands.
Based on the population data, the potential electricity and hydrogen fuel demands in each region can be estimated through Eqs (1) and (2), and the required parameters are given in Table 1.   (Shabadin et al. 2014) and projected to year 2020 assuming 5% annual vehicle growth rate The spatial analysis is performed using ArcMap 10.5. As shown in Figure 5(a) to (d), the feasible areas for supply chain infrastructure placement, average solar irradiation, fuel and electricity demands in all regions can be determined through spatial analysis. Based on Figure 5, region 4 receives the highest solar irradiation while region 10 has the highest fuel and electricity demands. Table 2 displays the transportation distance between the regions. With the processed spatial data, an optimization model is used to determine the optimal arrangement of supply chain infrastructures to fulfil regional energy demands using electricity produced from solar PV.  1  0  205  52  79  109  96  145  189  149  171  2  205  0  205  198  96  149  130  88  174  120  3  52  205  0  27  109  77  145  189  150  171  4  79  198  27  0  102  52  139  183  143  165  5  109  96  109  102  0  53  70  114  78  100  6  96  149  77  52  53  0  89  133  93  115  7  145  130  145  139  70  89  0  44  54  31  8  189  88  189  183  114  133  44  0  88  35  9  149  174  150  143  78  93  54  88  0  53  10  171  120  171  165  100  115  31  Mathematical optimization is used to determine the least-cost HESC network. This section describes mathematical equations for the objective function, mass and energy balance, as well as the generic costing of equipment. As the costing of specific supply chain components is lengthy and dependent on the techno-economic parameters provided, the detailed calculations are not displayed here but in the Supplementary Material.

Objective Function
The objective function of this model is to minimize the cost of hydrogen-electricity supply chain, which is given by the summation of total capital, operating and replacement costs of the supply chain components in a year: The generic formulas for the annualized capital and replacement costs of equipment are extracted from Huang et al. (2019). Eq (4) shows the calculation for annualized capital cost, where the capital recovery factor is given by Eq (5). Meanwhile, the annualized replacement cost can be calculated using Eq (6). The annual operating cost of equipment is given by Eq (7).

Electricity Generation, Storage and Transmission
Eq (8) computes the electricity generation from solar radiation. The peak power of solar panel can be determined using Eq (9), while Eq (10) bounds the land-use to be within the available area in a region. The constant R(9M is used to scale the land use of entire system based on the area occupied by solar panels. Eq (11) indicates that the electricity produced in solar panel can be utilized in the form of electricity or converted into hydrogen. The electricity available is either used to fulfil the electrical demand or export to other regions, as indicated in Eq (12). Considering the mismatch between electricity production and demand, the efficiency losses in energy storage, <== is taken into account. Eq (13) shows the net amount of electricity delivered to each region. The electricity imported by a region can be used for electricity or hydrogen demands, as defined in Eq (14).
Eq (15) illustrates that the electricity demand of a region will be satisfied by the imported electricity, locally produced PV electricity, and the electricity produced from hydrogen in fuel cell.
Note that the electricity produced in fuel cell should be converted into AC form as the electrical load is in AC form. The electricity produced in fuel cell is given by Eq (16) Eq (17) constrains that a region should not be producing and importing electricity at the same time, where the binaries can be defined using Eqs (18) For this study, the electricity storage requirement is estimated based on the days of autonomy concept. The energy storage will be installed at the electricity-producing site as shown in Eq (20).

Hydrogen Production, Storage, and Transportation
The amount of hydrogen produced using locally produced electricity can be computed using Eq (21). On the other hand, the electricity imported from other regions can also be used to produce hydrogen as illustrated in Eq (22). Note that the imported electricity has to be converted from AC into DC before inputting it to electrolyzer.
Hydrogen can be stored in several forms such as compressed gas, cryogenic liquid or liquid organic hydrogen carrier. Thus, the hydrogen produced in electrolyzer has to be converted into suitable forms prior to storage. Eq (23) defines the sum of locally produced hydrogen being converted to other forms. Considering the losses when converting hydrogen from one form to another, Eq (24) gives the net amount of hydrogen remaining upon conversion. The same applies to the hydrogen produced from imported electricity, as shown in Eqs (25) and (26).
Eq (27) shows that the processed hydrogen can either be utilized locally or transported to other regions. Eq (28) Eqs (30) Hydrogen can be transported from one region to another using truck or pipeline. Eqs (36) and (37) define the amount of hydrogen exported from/imported to a region respectively. The binaries

Optimal HESC Network
The optimization model was configured for various case studies to evaluate the economic performance of the proposed HESC network under different considerations, which include the following: • Base case scenario to determine the baseline cost of the proposed HESC network.
• Optimal cost and configuration of HESC with reduced/increased days of autonomy to identify the impact of energy storage requirement on the supply chain.
• Optimal cost of HESC when subjected to various electricity and hydrogen fuel penetration rates to determine suitable product charges.
The mixed-integer linear programming (MILP) model is formulated in GAMS and solved using CPLEX solver. The software is run on HP Elitebook 850 G5 with Intel Core i5-8250U (1.60 GHz) processor and 8 GB RAM. The time required to obtain a solution is about 52 mins and the model statistics are listed in Table 3.

Base Case
Through mathematical optimization, the minimum cost of HESC for the base case scenario is determined as 14.9 billion USD/y. Based on the cost breakdown displayed in Figure 6, battery constitutes the highest proportion of cost, followed by solar panel, electrolyzer, and converter. hydrogen. According to Figure 7(b), hydrogen is being transported from regions 2 and 4 to the other regions, which also indicates that the hydrogen produced in regions 1 and 9 is only used for local demand. As most of the hydrogen is transported using GH2 trailers, most regions are having refuelling stations that receive hydrogen from GH2 trailers. Meanwhile, hydrogen is transported from region 4 to 10 through pipeline. This is likely due to the high transportation load and long transportation distance that makes pipeline transportation more feasible than truck transportation.
Overall, it is found that hydrogen is mainly produced in regions with high solar irradiation and transported to other regions. Table 4 displays the capacities of major equipment in each region, where region 4 has the highest solar panel requirement.
While the base scenario considers two days of autonomy to allow the energy storage for supplying two days of loads in the absence of energy sources, the cost of optimal HESC network with fewer and more days of autonomy is inspected in Section 6.2. On the other hand, Section 6.3 discusses the suitable fuel and electricity charges for the HESC network to be cost-competitive in comparison to conventional energy system.

Influence of Days of Autonomy towards Optimal HESC Configuration
Days of autonomy represent the number of days an energy storage can supply the loads without energy generation. For this study, days of autonomy means the number of no-sun days an energy storage can support. Figure 8 shows the optimal cost of HESC subjected to various days of autonomy. Overall, the HESC cost increases with the days of autonomy. As the days of autonomy increases from 5 to 8, a significant cost reduction has been observed in battery and converter, while the costs of solar panel, electrolyzer, fuel cell, and hydrogen storage have increased sharply. This indicates that more electricity is produced from hydrogen and hydrogen storage is preferred as the days of autonomy increases. Based on Figure 9, it is observed that the battery system is no longer employed when the days of autonomy is targeted as 8 or above. This is accompanied by a sharp rise in GH2 storage for pipeline transportation. In addition, LOHC storage is employed when the days of autonomy is 5 or above. Figure 10 presents the proportion of hydrogen storage with different days of autonomy. With 1 or 2 autonomy days, the hydrogen is mainly stored in gaseous form to be transported via pipelines and trailers. Meanwhile, LOHC storage is dominant when the days of autonomy is set as 5. When the days of autonomy is 8 and beyond, the gaseous storage for pipeline and fuel cell becomes significant.

Cost-Competitiveness of HESC Network
Despite the high cost of HESC network, sensitivity analysis is conducted to identify the suitable charges for vehicle fuel and electricity for the proposed HESC network to be cost-competitive. Eq (38) is used to compute the total cost of conventional energy use based on the fuel and electricity demands defined in this study.
According to IEA (2019b), the cost of producing hydrogen from renewable electricity could fall 30% by 2030 due to the declining costs of renewables and scaling up of hydrogen production, where the fuel cells and electrolyzers can benefit from mass manufacturing. On the other hand, the cost reduction in matured, large-scale production of solar PV is inconceivable, but there is significant room to scale up further the manufacturing of battery (IEA 2020). Thus, two scenarios have been evaluated for the sensitivity analysis: • Scenario 1: The costs of supply chain infrastructures remain the same as in the base case scenario.
• Scenario 2: The cost of battery is reduced by 50%, and the costs of electrolyzer and fuel cell are 70% of the initial values. Meanwhile, the costs of other supply chain infrastructures remain the same as in the base case scenario For the sensitivity analysis, the cost of electricity will be expressed in USD/kWh, and the cost of fuel will be expressed in litre gasoline-equivalent, USD/L. The costs of fossil-based petrol and electricity are raised at the same rate until the total cost of conventional energy use,

1A9
is the same as the cost of the proposed HESC network, ^\1 . Figure 12 shows the result of sensitivity analysis and it can be observed that the costs of petrol and electricity need to elevate by about 500% in scenario 1 for the green HESC network to be costcompetitive. This means that the fuel cost should be at least 3 USD/L and the electricity cost should be 0.36 USD/kWh. For scenario 2, about 300% increment in the original petrol and electricity costs are required. This corresponds to a fuel cost of 2 USD/L and an electricity cost of 0.24 USD/kWh. Considering the cases where the penetration rate of the green HESC network is less than 100%, the petrol and electricity produced from fossil fuels are still being used for energy activities.
As demonstrated in Eq (39), with 10% penetration rate of HESC network, 90% of the energy is still produced using conventional energy system. Charging the use of conventional energy at higher rates than the business-as-usual scenario will result in extra income from petrol and electricity sales, as shown in Eqs (40) and (41). The extra income can be used to subsidize the green HESC network to reduce its investment cost, as displayed in Eq (42) Figure 13 and Figure 14 display the required fuel and electricity costs with various penetration rates of HESC network. Overall, with the extra income from petrol and electricity sales being used to subsidize the HESC network, the required costs are significantly lower than the unsubsidized scenario. At 100% penetration level, both the subsidized and unsubsidized scenarios are having the same outcome as the conventional energy is no longer used, so there is no additional income to subsidize the HESC network. From Figure 13 and Figure 14, it can be observed that when the HESC network is being subsidized, the required costs of fuel and electricity are close for scenarios 1 and 2 when the penetration rate is low. Nevertheless, as the penetration rate increases, the difference between scenarios 1 and 2 becomes more significant.

Conclusion
This study presents an integrated GIS and mathematical optimization framework for the spatial optimization of hydrogen-electricity supply network. Through this study, a few research gaps have been addressed: (i) the integration of hydrogen and electricity supply network in a single HESC optimization model, (ii) site suitability analysis using GIS to determine suitable areas for the installation of supply chain infrastructures, (iii) the storage and transportation of hydrogen in the form of LOHC has been considered in addition to the conventional gaseous and liquefied hydrogen storage, (iv) the impact of days of autonomy to the optimal supply chain configuration has been investigated.
The proposed methodology has been demonstrated through a case study in Johor, Malaysia.
Results showed that the least-cost HESC network in base case scenario would require an investment of 14.9 billion USD/y. By assuming two days of autonomy, the cost of battery is significant and contributes to 43% of the total cost. Besides, the hydrogen produced is mainly stored as compressed gas and transported via gas trailers and pipelines. When the days of autonomy is increased to 8 and above, electricity storage and transportation in the form of hydrogen is preferred, and the transported hydrogen is reconverted back to electricity at demand regions through fuel cells. On the other hand, for hydrogen demands at refuelling stations, hydrogen produced from electrolyzer is stored and transported to refuelling stations in the form of LOHC.
From the sensitivity analysis, it is shown that the HESC is unlikely to be fully replacing the conventional energy system in near-term due to the high cost of investment. In future, the HESC optimization framework can be further extended for multi-period modelling to model the stagewise development of hydrogen supply network in order to fulfil the fuel and electricity penetration targets at each time period. Moreover, environmental and safety indexes should be incorporated into the proposed framework for multi-objective optimization of HESC.