FIELD OF THE INVENTION

The present invention relates to a system and method for optimisation. More particularly, but not exclusively, the present invention relates to a system and method of optimising the lifetime cost of a hydrogen, ammonia, iron and/or steel production and processing facility powered predominantly by intermittent energy sources.

BACKGROUND OF THE INVENTION

The applicant has identified that there exists a problem in that it is desirable for more processes to be powered by renewable energy sources, however, many processes rely on constant input power supplies whereas renewable energy sources such as solar energy and wind energy are typically intermittent. The applicant has determined that it would be beneficial to have a system and method to facilitate the adaptation of a system or method such that a renewable energy source is able to be used to power such a process, including the planning of the system location, the planning of the system components and/or the sizing of the system components.

SUMMARY OF THE INVENTION

In accordance with one aspect of the present invention, there is provided a method of planning a plant for powering a process from an intermittent energy source, wherein the method includes the steps of: using a computer to model the plant; inputting data to the model relating to historical availability of the intermittent energy source to simulate the plant; inputting technological information to the model to simulate the plant, including a plurality of values relating to configuration of the plant; inputting financial information to the model; selecting one or more of the values to be variables; selecting an optimization criteria; and optimizing the criteria by varying the variables using an algorithm to identify a set of optimum values.

Preferably, the intermittent energy source includes a plurality of renewable energy sources.

In a preferred form, the process is for forming green hydrogen, green ammonia, green iron and/or green steel.

It is preferred that the optimizing step is conducted according to an algorithm which allows for simultaneous optimum sizing of multiple components of the plant for fixed energy input from the intermittent energy source. More preferably, the optimization algorithm is a Bayesian search algorithm. The applicant has identified that a Bayesian search algorithm may be particularly advantageous, also being a non-obvious choice involving difficulties in being applied to this specific domain. More specifically, the applicant identified difficulties in choosing priors to be applied for this application of a Bayesian search algorithm and, with inventive input, was able to arrive at a solution by finding priors broad enough to encompass optimal solutions but narrow enough to reduce computational search requirements within the boundaries of the optimal solution.

Preferably, the optimization simultaneously determines the size of multiple components of the plant.

Preferably, the model models flows of inputs and outputs between the plant components.

In accordance with another aspect of the present invention, there is provided a system for planning a plant for powering a process from an intermittent energy source, wherein the system includes a computer-implemented model of the plant, wherein the computer- implemented model is adapted to receive data relating to historical availability of the intermittent energy source to simulate the plant, the computer-implemented model is adapted to receive technological information to simulate the plant, the technological information including a plurality of values relating to configuration of the plant, the system is adapted to receive financial information, the system enables selection by a user of one or more of said values to be variables, the system enables selection by a user of an optimization criteria, and the system automatically optimizes said optimization criteria by varying the variables using an algorithm to identify a set of optimum values.

Preferably, the intermittent energy source includes a plurality of renewable energy sources.

The process may be for forming green hydrogen, green, green ammonia, green iron and/or green steel.

Preferably, the optimization is conducted according to an algorithm which allows for simultaneous optimum sizing of multiple components of the plant for fixed energy input from the intermittent energy source. More preferably, the optimization algorithm is a Bayesian search algorithm.

In a preferred form, the optimization simultaneously determines the size of multiple components of the plant.

Preferably, the model models flows of inputs and outputs between the plant components.

There is also provided a method for powering a process from an intermittent energy source, wherein the method includes the steps of: using a computer to model the plant; inputting granular time series data into the model relating to historical availability of the intermittent energy sources as a proxy for future availability; inputting technical information to the model to simulate the components of the plant, including a plurality of values relating to the configuration of the plant; inputting financial information to the model; selecting several of these input data as one or more of the values to be variables; selecting optimization criteria; and optimizing the criteria by varying the variables using an algorithm to identify a set of optimum values. Preferably, the optimization jointly determines the size of multiple components of the plant.

Preferably, the model simulates the flow of inputs and outputs between the plant components. BRIEF DESCRIPTION OF THE DRAWINGS

Preferred embodiments of the invention will be described, by way of a non-limiting example only, with reference to the accompanying drawings in which:

Figure 1 shows example annual output plots and commodity flow;

Figure 2 shows an example of levelized cost for hydrogen at a particular geographic location;

Figure 3 shows a flow chart representing a technical model for production of hydrogen;

Figure 4 shows minimum available energy (Min() function example);

Figure 5 shows a flow chart representing a technical model for hydrogen generation in a centralized (star-like) energy distribution arrangement;

Figure 6 shows a flow chart representing a technical model for production of ammonia;

Figure 7 shows a flow chart representing a technical model for production of green iron; Figure 8 represents an optimization algorithm;

Figure 9 shows a generic model flow chart;

Figure 10 shows levelized cost of hydrogen for a range of potential geographic locations;

Figure 11 shows an example output table of Appendix A; and Figure 12 shows a diagrammatic representation of the optimisation system.

DETAILED DESCRIPTION

Recently, green hydrogen and green ammonia have become a popular topic of investigation for many companies around the globe. Hydrogen is heralded as a potential green fuel of the future, while ammonia allows for vectoring and stable storage of hydrogen. In addition, ammonia, being the second most produced synthetic chemical globally and used predominantly in fertilizer production, attracts a lot of interest from a decarbonization perspective. Currently, green ammonia does not exist. Ammonia production relies on natural gas, creating significant amounts of carbon dioxide. There is also an interest in production of green steel. Currently, the steel industry emits approximately 6% of total carbon dioxide pollution.

Transition to green industries would require reliance on various renewable energy sources, all of which are highly intermittent (with the exception of hydropower). Due to this high and unpredictable intermittency in power supply, as well as technical limitations of various processes along the value chain (i.e., ramp up and ramp down of the Haber-Bosch plant), finding the best configuration for a complex multi-variate system is a challenge. To reduce the levelized cost of each of the commodities would require the process to run at full capacity all the time. Achieving this is, however, impossible because of energy intermittency. Energy storage solutions are still expensive. It remains a challenge to find the most optimal solution of all the components in the green industries landscape, while making sure that the levelized cost of produced commodities remains as low as possible.

The current invention relates to the modelling and optimization framework where each of the system subcomponents can be independently scaled up and/or down affecting its generation capacity and associated investment cost. All components are independent from each other in a sense that only commodity flow and their availability (water, energy, hydrogen, nitrogen, etc.) affect the next component along the value chain. The cascading build of the system and unique interaction of the components through commodity streams makes this modelling framework completely flexible and organic. In comparison, all other modelling approaches used hard calculations that directly compute values, and as such, put limitation on the scale-up of single components. This prevents running optimization algorithms and finding the best configuration of the system.

The applicant's approach combines a technical and financial model in one framework (including computation of the entire energy generation based on location and simple environmental parameters such as wind speed, solar irradiation, ambient temperature, etc.). This allows for optimization of the system in a single loop where adjusting one parameter will have an effect on the final outcome.

Accordingly, with reference to Figures 1 to 12 of the accompanying drawings, an example of the present invention provides a method of planning a plant for powering a process predominantly or at least partly from an intermittent energy source. More specifically, Figure 1 shows example annual output plots and commodity flow 10 for each hour of an entire year which may be used as data for the modelling. Figure 2 shows the levelised cost of hydrogen (LCOH) as a total (see bar 96), as well as separated into total capital expenditure 92 and total operational expenditure 94. Further, Figure 2 also separates total capital expenditure into separate components 44 to 62, and also separates total operational expenditure 94 into separate components 64 to 86.

The plant for powering a process may take different forms and, by way of non limiting example, may be represented in the form of a cascading model for hydrogen production as shown in Figure 3, a centralised architecture model for hydrogen production as shown in Figure 5, a cascading model for ammonia production as shown in Figure 6, or a cascading model for iron production as shown in Figure 7.

Turning to Figure 8 and Figure 9, the method may include the steps of generating a mathematical model 446 of the plant; inputting data 432 to the model relating to historical availability of the intermittent energy source; inputting technical information 450 to the model to simulate the plant, including a plurality of values relating to configuration of the plant; inputting financial information 452 to the model; selecting one or more of the values to be variables; selecting an optimization criteria; and optimizing the criteria in an optimisation loop 442 by varying the selected variables using an algorithm 440 to identify a set of optimum output values 492. The search using the optimisation loop 442 may end after a fixed number of steps or where convergence is detected, at which point the final results are provided.

The optimum output values 492 may include quantities such as levelised cost of hydrogen (LCOH) and/or other quantities as listed at reference numeral 488 in Figure 9. Levelised cost of hydrogen ($/kg) may be shown graphically for a number of locations, as depicted in Figure 10. In addition, the optimum values 492 may be provided in the form of an optimisation output table 540 as shown in Figure 11. Different weather data sets may be used for a variety of locations so as to provide a separate set of optimum values (544, 546, 548, 550) for each location, to assist with choosing of a suitable location for the plant.

Figure 12 shows a diagrammatic representation of the optimisation system 620, illustrating the system inputs (622 to 636) which are used to form models and simulations (638, 640, 644, 646) with an optimisation algorithm 642 being used in the optimisation loop so as to provide optimisation output 648. The optimisation output may be in the form of, for example, levelised cost of hydrogen (LCOH) 650, size and utilisation of every component in the process 652, and inputs and outputs for each component 654.

The applicant has identified that performing optimization on a multi -variate, multi dimensional and time dependent system is intrinsically a difficult and challenging task. In this invention, we have «-subsystem components along the value chain that can be independently scaled up or scaled down to achieve the best configuration. Having a system with n degrees of freedom, or in other words a system that operates in multi-dimensional space filled with many local extrema (minima and maxima), cannot be optimized by manual trial and error. Finding the global minimum or maximum (in this particular case the global minimal levelized cost of commodity) is extremely difficult as optimization algorithms often tend to converge on a local minimum. The optimization method deployed to perform optimization on this unique model uses a surrogate function, distributed and asynchronous approach to search the hyperparameter space to quickly and efficiently find the best configuration and avoid local extrema in a given amount of time and/or runs.

Other optimisation algorithms have been tested and are applicable to be used with the framework. A non-exclusive list of such optimisation algorithms includes evolutionary algorithms, machine learning algorithms and brute-force. This framework is not limited to green hydrogen, green ammonia or green steel production. The model and optimization strategy can be applied to optimize any value chain and design a working facility with the lowest levelized cost of produced commodity. This invention can be applied in the mining industry, mineral processing and other chemical industries, other than ammonia. Effectively this invention can be applied in any value chain optimization process.

Several different pathways for optimization of green hydrogen and green ammonia value chains from intermittent energy sources have previously been developed.

The majority of the previous work has used averaged values for renewable energy generation to perform simple calculations and to size the system components. That simplified approach has limitations because it does not account for intermittent behaviour of the system, and, as such, the outcomes are unrealistic. Performing multi -variate, time dependent optimization is a complex and challenging task. As such, some other approaches attempted to simplify the optimization process by focusing solely on finding the optimal mix of power generation components (i.e., wind and solar). Again, that is not an optimal solution as it does not account for behaviour of other components along the value chain.

In one instance, a techno-economic approach has been developed to optimize green hydrogen and ammonia production from intermittent energy sources. The model presented in that study is simplified and includes only optimization of the most essential system components (wind and solar generation, electrolyser, hydrogen storage, air separation unit, and Haber-Bosch process for ammonia generation). Assumptions have been made that ultrapure water is available at all times and that hydrogen compression is not necessary. Because of technical limitations of the Haber-Bosch process, its long ramp up and ramp down times, the modelling uses fossil fuel powered energy generation to support the Haber- Bosch process resulting in the generated ammonia not being entirely green. The model uses various other simplifications and the way each of the components interact with each other is different to the work done by the applicant.

Julien Armijo, C0dric Philibert (2020) Flexible production of green hydrogen and ammonia from variable solar and wind energy: Case study of Chile and Argentina. International Journal of Hydrogen Energy, 45 (3), 1541-1558, https://doi.Org/10.1016/j.ijhydene.2019.l 1.028

Another attempt to optimize the entire process for green hydrogen and ammonia generation looks at power generation (wind and solar energy generation) to apply statistical modelling and find the most suitable wind and solar combination. In addition, that optimisation attempt accounts for stability of solar and wind resources over a period of time. In contrast to that attempt, the applicant has identified that it would be advantageous to have a method and system which enables the flexibility to optimise a proposed plant based on a small or large number of variables. Moreover, the applicant has identified that it would be advantageous to have a method and system which, rather than simply seeking to deliver the best combination of lowest cost of electricity (LCOE) and maximum power output stability, enables the flexibility to hold fixed values relating to the energy source mix and to optimise a proposed plant based on other variables to be optimised (e.g. for a given wind energy and solar energy mix).

Another study looked at the impact of intermittency on plant sizing and levelized cost of ammonia (LCOA), however, the approach presented in that paper has limitations.

Richard Nayak-Luke, Rem Bahares-Alcdntara and Ian Wilkinson (2018) “Green” Ammonia: Impact of Renewable Energy Intermittency on Plant Sizing and Levelized Cost of Ammonia. Ind. Eng. Chem. Res. 2018, 57, 43, 14607-14616, https ://doi. org/10.1021/acs. iecr.8b02447

The optimization process developed by the applicant is the most sophisticated approach to date for green hydrogen and green ammonia optimization.

There are other frameworks that use basic approaches to find the best system configuration, for example, commercially available HOMER Energy (https://www.homerenergy.com), but the approach utilized by this software is substantially different. Furthermore, the applicant has identified that the HOMER model is not suited for complex optimization. Example 1 (Green Ammonia Generation from Intermittent Resources):

When modelling ammonia generation (see Figure 6), a model comprising of intermittent power generation (i.e., wind 276 and solar 274), is connected in a cascading manner to a reverse osmosis plant 290 that purifies brackish water. This subsequently stores the water at storage 302 which is then used in the water deionizer plant 308 for further purification and metal ion removal. The water is then stored in a deionized water storage tank 318, and supplied to an electrolyser 322, where together with electric energy, it is converted into hydrogen 326. The produced hydrogen has a pressure in the range of 10 to 30 bar and is stored in low pressure hydrogen storage tanks 330 or by other means. From there, hydrogen is supplied to a hydrogen compressor 334 where with the help of electrical energy it is compressed to medium pressure (200 - 250 bar) which is required as an input for the Haber-Bosch process to generate ammonia. Cascading energy output from the hydrogen compressor 334 is then supplied to an Air Separation Unit (ASU) 332 where atmospheric air is liquefied to produce pure nitrogen 312. This nitrogen is piped to the low pressure storage 328 and a nitrogen compressor 320 pumps it into medium pressure storage 316 (200 - 250 bar). The final step is the Haber-Bosch plant 306 where, under high pressure and temperature, hydrogen and nitrogen feedstocks are converted into ammonia 304.

In this example, there are several high-level plant configurations which the model could consider:

• A natural gas power plant 294 can be included to support the operation of the Haber- Bosch process due to its long ramp up and ramp down times, which makes it hard to sustain given high intermittency energy sources (such as wind and solar).

• The model can choose between two types of electrolysers. A deionizer step and subsequent storage modelling are only required when the proton exchange membrane (PEM) electrolyser is used. However, if an alkaline electrolyser is used, water from reverse osmosis is sufficient.

• Low pressure 330 and medium pressure hydrogen storage 338 can be omitted as hydrogen produced in the electrolyser 322 can be supplied directly to the hydrogen compressor 334 and later to the Haber-Bosch plant 306. • Low 328 and medium pressure 316 nitrogen storages can also be omitted if energy quality is high.

• It is also possible that the pressure supplied by the air separation unit 332 will negate the need for a nitrogen compressor 320.

In this instance, modelling deals with five commodities or materials, including energy, water (reverse osmosis water and deionized water), hydrogen (low and medium pressure), nitrogen (low and medium pressure) and ammonia.

The model uses one of the available and embedded optimization algorithms to work in a loop and to find the lowest levelized cost of ammonia (LCOA), by scaling up, scaling down all or removing the optional subsystem components. The optimization relies on the input and interaction with the financial component of the model, where these value chain components are assigned weights (financial weights). At each step, the optimisation algorithm will sample from the priors over all the optional system components. These priors are defined as part of the system setup for this task but are further refined by the user as the system's results are analysed. The surrogate function can be any chosen regression function which estimates the value of the objective function for points in the search space not yet sampled. The optimiser can use the surrogate function's estimates to select points in the search space that are more likely to yield the best LCOA. It is using these priors and surrogate functions by which the system can avoid local minima while not exhaustively searching the space.

Figure 6 - Please refer to attached Figure 6 Flow diagram (Green Ammonia).

Example 2 (Green Steel Generation in Direct Reduced Iron (DRI) process from Intermittent Resources):

Green steel generation (see Figure 7) involves the same steps as listed in Example 1, to produce hydrogen. This includes power generation (i.e. wind 366 and solar 364), which is supplied in a cascading manner to a reverse osmosis plant 378 to pre-treat water and store this water in a storage tank 388. Subsequently, electrical power and reversed osmosis water is supplied to the deionizer plant 392 and the ultrapure water is stored in storage tanks 400. This water is then piped and the cascading power is supplied to the electrolyser 404. The produced low pressure hydrogen 412 is then transferred from the electrolyser 404 to a low pressure hydrogen storage tank 412. This is then supplied to a direct reduced iron furnace 390 where cascading electrical power, hydrogen and iron ore feedstock are fed into a furnace and converted into steel or pig iron.

In this example, there are several high-level plant configurations which the model could consider:

• A natural gas power plant 402 may be included to support the operation of the direct reduced iron process due to its high operational temperature and long ramp up and ramp down times, which make it hard to sustain given high energy intermittency (such as wind 366 and solar 364).

• The model can choose between two types of electrolysers. A deionizer step and subsequent storage modelling are only required when a proton exchange membrane (PEM) electrolyser is used. However, if an alkaline electrolyser is used, reverse osmosis water is sufficient.

• The model can choose to either use an energy storage facility 380 (such as a battery) or emit it to smooth out energy intermittency in the process.

In this example, modelling deals with four commodities including energy, water (reverse osmosis water and deionized water), hydrogen (low and medium pressure), and ammonia.

The model uses one of the available and embedded optimization algorithms to work in a loop and find the lowest levelized cost of steel (LCOS) produced in a direct reduced iron process. The optimisation is done by independently scaling up, scaling down all or removing the optional subsystem components. The optimization relies on input and interaction with the financial component of the model, where these value chain components are assigned weights (financial weights).

Figure 7 - Please refer to attached Figure 7 Flow diagram (Green Iron). Example 3 (Green Hydrogen Generation from Intermittent Resources):

Green hydrogen generation (see Figure 3) involves some of the same steps listed in Example 1 and Example 2. Green hydrogen production revolves around power generation. This power is supplied in a cascading manner to a reverse osmosis plant 114 to pre-treat water and store reversed osmosis water in a storage tank 126. The subsequent power and reversed osmosis water supply is sent to the deionizer plant 128. The ultrapure water is stored in a storage tank 134. This water is then supplied together with cascading power supply, to an electrolyser 140. Produced low pressure hydrogen can subsequently be used in one of the two processes hydrogen delivery mechanism.

In this example, there are several high-level plant configurations and priors which the model could consider:

• The model can choose to either use an energy storage facility 120 (such as a battery) or emit it to smooth out intermittency in the process.

• For hydrogen delivery, one process pumps low pressure hydrogen to a hydrogen compressor 154 between 350 or 700 bar by supplying generated cascading electrical power and is distributed directly to end customers. Alternatively, low pressure hydrogen can be liquefied to generate liquid hydrogen with superior energy density (energy per volume) compared to compressed hydrogen. Hydrogen liquefaction is a cryogenic process that requires continuous power supply.

• A natural gas power plant 138 may be included to support cryogenic hydrogen liquefaction, as hydrogen will boil if it is not kept at cryogenic temperatures, which makes it hard to sustain given high intermittency (such as wind 106 and solar 104).

• The model can choose between two types of electrolysers. A deionizer step and subsequent storage modelling are only required when a proton exchange membrane (PEM) electrolyser is used. However, if an alkaline electrolyser is used, reverse osmosis water is sufficient.

Just like in the previous examples, the developed model uses one of the available and embedded optimization algorithms to work in a loop and find lowest levelized cost of hydrogen (LCOH) produced in direct reduced iron process. This optimization is done by independently scaling up, scaling down all or removing the optional subsystem components. The optimization relies on input and interaction with the financial component of the model, where these value chain components are assigned weights (financial weights).

Figure 3 - Please refer to attached Figure 3 Flow diagram (Green Hydrogen).

Minimum Available Energy (Min Function Example)

Instead of directly calculating how much commodity can be produced, the model relies on flexible connection between each of the system components. The model uses a minimum function to ensure the most optimal operation. The minimum function evaluates available feedstock commodities that are necessary in a given process and converts them into energy equivalent for fair comparison. By doing so the process not only operates at the most optimal performance, but also makes sure that output commodities are produced only when sufficient feedstock commodities are available.

One example is shown in Figure 4. The electrolyser component 176 relies on supply of deionized or reverse osmosis water 172 from respective water storage tank and availability of power supply 174, but also the electrolyser should not produce more hydrogen 180 if the next value chain component, low pressure hydrogen storage, is full. In addition, another obvious limitation is that the electrolyser 176 will not be able to process more water and power than its given rated capacity.

In this case, minimum function ( Min( )) compares different commodities converted into energy to find the smallest one and drive the process based on minimum available commodity. The Mini) function compares available power, the electrolyser’ s rated capacity, ratio of water available to water required to produce given amount of hydrogen (i.e., 11 1 of water to produce 1 kg of hydrogen) and multiplies this by power consumption of the electrolyser (power required to produce given amount of hydrogen, i.e., 54 kWh per kg of ¾). This allows conversion of available water to be expressed in energy terms. Finally, this subsystem checks the size of low pressure hydrogen storage and the current amount of hydrogen stored in it. By subtracting one from the other, this establishes how much more hydrogen can be stored. This value is then multiplied by the power consumption of the electrolyser (power required to produce given amount of hydrogen, i.e., 54 kWh per kg of ¾) in order to convert available hydrogen storage into energy. Th eMin() function compares these values in terms of energy to make sure that only feasible and allowable amounts of output feedstock are produced.

Figure 4 - Please refer to attached Figure 4 Minimum available energy (Min() function example).

UNIQUE FEATURES

1. The method and process minimizes the levelized cost of producing hydrogen using a cascading or parallel arrangement of system subcomponents with their associated purchase, construction and operating costs, and flows of electricity and other inputs to the components, under the constraint of producing a determined volume of hydrogen. The optimization process simultaneously determines the size and utilization of every component and the flows of inputs and outputs to and from each of them. The system is currently strictly cascading for the use of electricity. The applicant has discussed a possible further optimization by allowing every component of the system to tap into the "pool" of electricity sources directly instead of relying on the remainder amount of electricity after use from the previous step. It is proposed that such a system will be implemented to minimise the cost of hydrogen.

2. The system subcomponents are linked together, meaning that changing the operating characteristics of the size of one component has an effect on the entire system. System subcomponents can be scaled independently affecting the total cost of producing hydrogen as well as the commodity flows between components (energy, water, hydrogen, nitrogen, ammonia, etc.).

3. The size of the system components is completely independent, one to each other. However, their production capacity depends on the availability and flow of energy and other commodities within the system. 4. The model simulates the performance of a complex multi-component system run on highly intermittent energy and material flow within the system. Complex optimization is performed on given intermittent energy and material profiles with selected data interval granularity, i.e., hourly interval. 5. The modelling framework consists of technical and financial components, collectively forming a complete techno-economic model. For finding the best configuration, every system subcomponent and its associated financial calculations are optimized all together and in a loop until convergence is reached and the lowest possible cost of producing hydrogen is determined at the end of the process. 6. The optimization uses electricity inputs, either in the form of hourly energy generated by each source of intermittent power or in the form of weather data such as solar irradiation, ambient temperature, wind speed processed, to produce hourly power generation profiles.

7. Energy storage components, for example a battery, is/are incorporated in the model as an option to decrease the intermittency of power generation by storing energy and using it when solar or wind energy is insufficient. This also increases the utilization rate of the components that need energy and smooths their energy consumption profile over short time periods, thereby reducing the cost of producing hydrogen.

8. There is room for further expansion of the system by adding and testing other sources of electricity production or storage such as concentrated solar power (CSP), pumped hydro or hydropower generation (with or without inter-seasonal storage).

9. The system is currently used to optimize the cost of producing hydrogen. The same methodology may be used to minimize the cost of production of green steel from highly intermittent energy sources. Other multi-component industrial and mining processes can also benefit from this current invention.

10. The method is completely automated and can run with various energy datasets for different locations. This enables the user to select the best location and the best configuration according to the input data. 11. The methods of optimisation use a surrogate function, distributed and asynchronous approach to search the hyperparameter space to quickly and efficiently find the best configuration and avoid local extrema (in this particular case avoiding local minima).

12. Other optimisation algorithms have been tested and are applicable to be used with the framework. A non-exclusive list includes evolutionary algorithms, machine learning algorithms, brute-force, etc.

13. Finding global minima is a difficult and sometimes impossible task. The model computes a convergence factor to show how a number of runs reduces the levelized costs of commodities. If levelized cost does not change anymore, the simulation can be terminated to save computing time and assuming that the solution it converged to is the best configuration.

14. Taking advantage of cloud computing to analyse multiple scenarios in parallel.

15. The system has a front-end allowing easy, what-if scenarios to be run by non-coding professionals. 16. The systems computational architecture is built to minimise cloud computing costs by encapsulating simulation code so that cloud computation is only run when requested.

17. The system is modelled at fine granularity taking into account time delays due to plant characteristics and implicit and explicit storage of processes. Examples of the present invention may provide advantages by virtue of how elements of the system communicate. In one example, the components in the model may be connected based on their occurrence along with the process flow. Further, logic functions may be applied to ensure that the current system along the process flow does not use more feedstock commodities than available and/or to ensure that a commodity produced by the system is not larger than the subsequent system can consume or storage can accommodate.

In one example, an electrolyser system, which sits between a deionised water storage and a low pressure hydrogen storage, consumes both deionised water and renewable electricity to produce low pressure (up to 30 bar) hydrogen. The electrolyser model utilises the Min() function to find which parameter is the smallest. It looks at available energy, current electrolyser system capacity expressed in terms of power (in kilowatts), available deionised water in the deionised water storage tank divided by water in litres required to produce one kilogram of hydrogen and finally multiplied by the power consumption of the electrolyser expressed in kilowatt-hours per kilogram of hydrogen. This calculation allows expression of available water in terms of energy and allows comparing this parameter with two previous parameters. Another parameter in this Min() function for the electrolyser is a difference between the current weight of hydrogen in low pressure hydrogen storage and a maximum weight of hydrogen that that storage can contain expressed in kilograms of hydrogen, multiplied by the power consumption of electrolyser expressed in kilowatt-hours per kilogram of hydrogen. The applicant has determined that this allows expression of available hydrogen storage capacity in terms of energy and the comparison of it with other parameters in this Min() function. The reason for using the Min() function is that, even if other parameters are allowing the production of more hydrogen, we are restrained by the parameter with the lowest value. For example, if our available energy allows us to produce 1,000 kg of hydrogen theoretically, but there is no deionised water, we will not achieve this production.

An example Min() function is outlined, as follows:

Electrolyser function = Min(available energy (kWh), electrolyser capacity (kWh/h), available deionised water in water storage (L) divided by water consumption of electrolyser to produce a kilogram of hydrogen (L/kg Eh) and multiplied by the power consumption of electrolyser (kWh/kg Fh), the storage capacity of low pressure hydrogen storage (kg Fh) minus the current weight of hydrogen in the low pressure hydrogen storage (kg Fh) multiplied by the power consumption of electrolyser (kWh/kg Fh))

The complex and powerful optimisation provided by examples of the present invention may include mapping of the variables into a three-dimensional surface (or indeed a model having greater than three dimensions - even 17 dimensions in one example) and the ability to avoid being trapped in local minima while also promoting efficiency of the algorithm. In addition, the complex optimisation provided by examples of the present invention allows modelling of far more components of the system than the Homer model. For example, the Haber-Bosch component, air separator and other plant components are not readily addressed by the Homer model. Also, examples of the present invention provide the option for cascading or non-cascading energy, with the ability to take into consideration how different plant components have different energy demands and priorities.

While various embodiments of the present invention have been described above, it should be understood that they have been presented by way of example only, and not by way of limitation. It will be apparent to a person skilled in the relevant art that various changes in form and detail can be made therein without departing from the spirit and scope of the invention. Thus, the present invention should not be limited by any of the above described exemplary embodiments.

The reference in this specification to any prior publication (or information derived from it), or to any matter which is known, is not, and should not be taken as an acknowledgment or admission or any form of suggestion that that prior publication (or information derived from it) or known matter forms part of the common general knowledge in the field of endeavour to which this specification relates.

LIST OF REFERENCE NUMERALS

Figure 1: Example annual output plots and commodity flow

10 Example annual output plots and commodity flow 12 Energy MWh 14 Wind

16 Solar

18 Total energy

20 Electrolyser

22 Energy storage 24 Haber-Bosch plant

26 Hour of year

Figure 2: Example levelized cost for hydrogen at selection location

40 Example levelized cost for hydrogen at selection location 42 Levelized cost of hydrogen (LCOH) at point of sale (USD/kg ¾) 44 Solar energy generation (capex)

46 Wind energy generation (capex)

48 Gas turbine (capex)

50 Energy storage (capex)

52 Land (capex) 54 Electrolyser plant (capex)

56 Haber-Bosch plant (capex)

58 Hydrogen and ammonia storage and port infrastructure (capex)

60 Ammonia reconversion and distribution infrastructure (capex)

62 Water generation and treatment plant (capex) 64 Solar energy generation (opex)

68 Wind energy generation (opex)

70 Gas turbine (opex)

72 Energy storage (opex)

74 Land (opex) 76 Water generation and treatment plant (opex) 78 Energy transmission (opex) 80 Electrolyser Plant (opex) 82 Haber-Bosch plant (opex) 84 Hydrogen and ammonia storage and port infrastructure (opex)

86 Ammonia reconversion and distribution infrastructure (opex) 88 Power sales 90 Carbon tax 92 Total capital expenditure 94 Total operational expenditure

96 Levelized cost of hydrogen (USD/kg ¾)

Figure 3: Flow diagram of cascading model for hydrogen production

100 Flow diagram of cascading model for hydrogen production

102 Hourly Environmental Weather Data Input 104 Solar Energy Production Simulation

106 Wind Energy Production Simulation 108 Other Energy Production Simulation (i.e. Hydropower, Concentrated Solar Power, etc.)

110 Available Energy 112 Energy (MWh) 114 Reverse Osmosis Plant 116 Brackish / untreated water (EbO) 118 Optional 120 Energy Storage 122 Water (H ₂ 0)

124 Storage level feedback loop 126 Reverse Osmosis Water Storage 128 Deionised Water Processing Plant

130 Optional, based on electrolyser election (alkaline versus proton exchange membrane electrolyser) 132 Carbon dioxide (CO ₂ ) 134 Deionised Water Storage 136 Optional, based on generated energy quality 138 Natural Gas Power Plant 140 Electrolyser

142 Oxygen (O ₂ ) 144 Low Pressure Hydrogen Storage 146 Hydrogen (¾) 148 Alternative I 150 Alternative II

152 Hydrogen Liquefaction Plant 154 Compressor 156 Liquid hydrogen (LH ₂ )

Figure 4: Minimum available energy (MinQ function example) 170 Minimum available energy (Min() function example)

172 Available Water

174 Available Energy

176 Electrolyser Module

178 Electrolyser Capacity (i.e., expressed in MW, to be optimized) 180 Produced Hydrogen 182 (To Low Pressure Storage)

184 Electrolyser Module Power Consumption = MinfAvailable Energy or Electrolyser Capacity or (Available Water / Water Required to Generate Given Amount of Hydrogen) ^» Energy Required to Generate Given Amount of Hydrogen or (Total Capacity of Low Pressure Hydrogen Storage - Current Hydrogen Amount in Low Pressure Hydrogen Storage) _* Energy Required to Generate Given Amount of Hydrogen) Figure 5 - Flow diagram of centralized (star-like) architecture model for hydrogen production

200 Flow diagram of centralized (star-like) architecture model for hydrogen production

202 Hourly Environmental / Weather Data Input 204 Solar Energy Production Simulation

206 Wind Energy Production Simulation 208 Other Energy Production Simulation (i.e. Hydropower, Concentrated Solar Power, etc.)

210 Energy (MWh) 212 Available Energy 214 Optional 216 Energy Storage 218 Remaining energy 220 Energy Distribution System 222 Reverse Osmosis Plant

224 Brackish / untreated water (H2O) 226 Water (H ₂ 0) 228 Storage level feedback loop 230 Reverse Osmosis Water Storage 232 Deionised Water Processing Plant

234 Deionised Water Storage 236 Optional, based on electrolyser selection (alkaline versus proton exchange membrane electrolyser)

238 Oxygen (O2) 240 Electrolyser

242 Hydrogen (¾) 244 Compressor

246 Low Pressure Hydrogen Storage Figure 6 - Flow diagram of cascading model for ammonia production

270 Flow diagram of cascading model for ammonia production

272 Hourly Environmental/ Weather Data Input 274 Solar Energy Production Simulation 276 Wind Energy Production Simulation

278 Other Energy Production Simulation (i.e. Hydropower, Concentrated Solar Power, etc.)

280 Available Energy 282 Carbon dioxide (CO2) 284 Energy (MWh)

286 Optional, based on generated energy quality 288 Optional 290 Reverse Osmosis Plant 292 Brackish / untreated water (H2O) 294 Natural Gas Power Plant

296 Energy Storage 298 Water (H ₂ 0) 300 Storage level feedback loop 302 Reverse Osmosis Water Storage 304 Ammonia (NH ₃ )

306 Haber-Bosch Plant 308 Deionised Water Processing Plant 310 Optional, based on electrolyser selection (alkaline versus proton exchange membrane electrolyser) 312 Nitrogen (N2)

314 Optional, based on generated energy quality and pressure of nitrogen provided by the air separation unit

316 Medium Pressure Nitrogen Storage 318 Deionised Water Storage 320 Nitrogen Compressor

322 Electrolyser 324 Oxygen (O2)

326 Hydrogen (¾)

328 Liquid Nitrogen Storage 330 Low Pressure Hydrogen Storage 332 Air Separation Unit

334 Compressor 336 Air

338 Medium Pressure Hydrogen Storage

340 Optional, based on generated energy and hydrogen production quality

Figure 7 - Flow diagram of cascading model for iron production

360 Flow diagram of cascading model for iron production 362 Hourly Environmental / Weather Data Input 364 Solar Energy Production Simulation

366 Wind Energy Production Simulation

368 Other Energy Production Simulation (i.e., Hydropower, Concentrated Solar Power, etc.)

370 Available Energy 372 Energy (MWh)

374 Brackish / untreated water (H2O)

376 Optional

378 Reverse Osmosis Plant 380 Energy Storage 382 Water (H2O)

384 Storage level feedback loop

386 Iron

388 Reverse Osmosis Water Storage 390 Direct Reduced Iron (DRI) Plant 392 Deionised Water Processing Plant

394 Optional, based on electrolyser selection (alkaline versus proton exchange membrane electrolyser) 396 Iron Ore 398 Carbon dioxide (CO2) 400 Deionised Water Storage 402 Natural Gas Power Plant 404 Electrolyser

406 Oxygen (O2) 408 Hydrogen (¾) 410 Optional, based on generated energy quality 412 Low Pressure Hydrogen Storage 414 Compressor

416 Medium Pressure Hydrogen Storage 418 Optional, based on generated energy and hydrogen production quality

Figure 8 - Flow diagram of optimization algorithm

430 Flow diagram of optimization algorithm 432 Hourly Environmental / Weather Data Input

434 Apriori Belief Over Parameters or Range of Valid Parameters (4) 436 OPTIMISATION ENGINE 438 Available Energy 440 Space Search Algorithm (5) 442 Optimisation Loop (6)

444 Values to Optimise (3) 446 Plant Simulation 448 Economic Simulation 450 Technical Input (1) 452 Financial Input (2)

454 (1) e.g. electrolyser power consumption, Haber-Bosch plant ramp-up time, etc.; 456 (2) e.g., price per unit, discount rates, inflation rates, etc.; 458 (3) e.g., LCOE, LCOH, LCOA, LCOS, produced hydrogen, produced ammonia, carbon tax, etc.; 460 (4) e.g., size of reverse osmosis plant, solar farm capacity, wind farm capacity, energy storage capacity, etc.;

462 (5) chooses a new parameter value to search for best configuration;

464 (6) search ends after fixed number of steps or convergence detected, and final results are provided. Figure 9 - Flow diagram of a generic model

480 Flow diagram of a generic model

482 Hourly Environmental / Weather Data Input

484 (1) e.g., electrolyser power consumption, Haber-Bosch plant ramp-up time, etc.;

486 (2) e.g., price per unit, discount rates, inflation rates, etc.; 488 (3) e.g., LCOE, LCOH, LCOA, LCOS, produced hydrogen, produced ammonia, carbon tax, etc.

490 Available Energy 492 Output Values (3)

494 Plant Simulation 496 Economic Simulation

498 Technical Input (1)

500 Financial Input (2)

Figure 10 Example of levelized cost of hydrogen (LCOH) (USD/kg H2) for various geographic locations. Black bars correspond to the use of natural gas power plant to sustain continuous operation of the Haber-Bosch process, dotted bars correspond to green hydrogen generation. Detailed optimization outputs are available in Figure 11.

510 Example of levelized cost of hydrogen (LCOH) (USD/kg ¾) for various geographic locations. Black bars correspond to the use of natural gas power plant to sustain continuous operation of the Haber-Bosch process, dotted bars correspond to green hydrogen generation. Detailed optimization outputs are available in Figure 11.

512 Location 1

514 Location 2

516 Location 3 518 Location 4 520 Location 5

522 Location 6

524 Location 7

526 Levelized cost of hydrogen (LCOH) (USD/kg Lb) Figure 11 - Example optimization output table. Columns highlighted grey correspond to scenarios where natural gas power plant was used to support the process.

540 Example optimization output table. Columns highlighted grey correspond to scenarios where natural gas power plant was used to support the process Weather dataset 542 Weather dataset

544 Weather dataset 1 / Location 1

546 Weather dataset 2 / Location 2

548 Weather dataset 3 / Location 3

550 Weather dataset 4 / Location 4 552 Levelized cost of electricity (LCOE) [USD/MWh]

554 Levelized cost of hydrogen (LCOH) [USD/kg H2]

556 Levelized cost of ammonia (LCOA) [USD/tonnes NH3]

558 Gas power plant capacity [as % of Haber-Bosch plant capacity]

560 Gas energy generation to total energy generation [%] 562 Hydrogen produced [tonnes]

564 Ammonia produced [tonnes]

566 Carbon dioxide intensity (compared to hydrogen generation from methane steam reforming) [%]

568 Wind capacity factor [%] 570 Solar capacity factor [%]

572 Wind farm generation capacity [MW]

574 Solar farm generation capacity [MW]

576 Energy storage capacity [MWh]

578 Reverse osmosis plant capacity [MW] 580 Reverse osmosis water storage capacty [ML] 582 Deionizer plant capacity [MW]

584 Deionized water storage capacity [ML]

586 Low pressure hydrogen storage capacity [tonnes]

588 Hydrogen compressor capacity [MW] 590 Medium pressure hydrogen storage capacity (250 bar) [tonnes]

592 Air separation unit capacity [MW]

594 Liquid nitrogen storage capacity [tonnes]

596 Nitrogen compressor capacity [MW]

598 Medium pressure nitrogen storage capacity (250 bar) [tonnes] 600 Haber-Bosch plant capacity [MW]

602 Electrolyser plant [MW]

Figure 12 - Optimisation System

620 Optimisation System

622 Financial date & weightings 624 Energy simulations

626 Weather data

628 Plant configurations

630 Input materials

632 Commodities produced 634 Optimisation criteria

636 Other relevant raw data

638 Production (physics) Model

640 Plant simulations

642 Optimisation Algorithm 644 Economic model

646 Economic simulations

648 Optimisation Output

650 LCOH produced hydrogen

652 Size and utilization of every component in the process 654 Inputs & outputs for each component