Optimising the operation of wind powered electrolysers

Integrated wind power – hydrogen systems may make a useful contribution to achieving climate targets. Both centralised wind powered electrolysers and multi-MW decentralised solutions are likely to see multiple electrolyser units/systems deployed together. Since the efficiency of electrolysers is a function of their load factor, there is a possibility of optimising operation of the individual units in order to maximise the overall plant efficiency. Here we outline optimal strategies for electrolysis facilities with two, three and four independent units, and find that using these optimised strategies can increase annual wind powered hydrogen production by 2.3 - 3.8%, depending on the specific set up. By quantifying the increase in hydrogen production through using optimised control strategies, this research can help industry to identify the best overall control scheme for electrolyser plants composed of multiple units.


Introduction
In the 2015 Paris Agreement, 196 Parties adopted a legally binding international treaty on climate change.Its overarching goal is to hold the increase in the global average temperature to well below 2°C above pre-industrial levels and pursue efforts to limit the temperature increase to 1.5°C above preindustrial levels [1].An important part of this goal is reducing the environmental footprint of energy production; in 2016, this accounted for 73.2% of greenhouse gas emissions [2].The International Energy Agency (IEA) has provided a roadmap on how to achieve a net zero energy system by 2050 [3].Among others, the IEA found that wind power and electrolysis are key technologies; by 2050, their core scenario required capacities of 8.3 TW and 3.6 TW, respectively.
In some cases, there may be advantages to combining wind and hydrogen production technologies, into electrically integrated turbine-electrolyser systems.Siemens Gamesa Renewable Energy sees advantages of deploying such integrated systems in offshore contexts, citing benefits of capex reduction due to replacing high voltage infrastructure with pipes, reduced electrical losses, and increased plant load factor, on the grounds that electrolysers are more flexible than electrical networks [4].Major GW scale wind projects are also considering hydrogen export over electrical export, including the Flotta Hydrogen Hub [5] and the Aquaventus initiative [6].
Integrated turbine -electrolyser systems raise new engineering questions, including around how to operate the system in the most efficient way.As centralised wind powered electrolysers and multi-MW decentralised solutions are likely to see multiple electrolyser systems deployed together, there is a possibility of optimising operation of the individual units to maximise the efficiency of the plant overall.In order to explore this, we need to understand what drives electrolyser efficiency.This is explored in the rest of this introduction.
The efficiency of an electrolyser depends on several factors.One of these is the electrolyser's load factor, which we define as the ratio of incoming power, in a given instant, to nominal power capacity.
Electrolyser efficiency as a function of load factor has been explored in literature, such as a white paper from Siemens Energy [7] and Ferguson's doctoral thesis [8].They break down the overall system efficiency into two main components: the electrolysis stack, and the supporting equipment, commonly known as the balance of plant, which includes gas dryers, water purifiers, and power converters.
Generally, the electrolysis stack efficiency decreases linearly with load factor.The exception is at low load factors, when hydrogen losses, due to parasitic currents, membrane-cross over and recombination with oxygen, are relatively high.However, if we set a relatively high minimum power demand, such as 20% of nominal capacity, then we will only operate in conditions where hydrogen losses are so low as to be negligible [8].This leaves us with a near linear decrease in stack efficiency with load factor.
The efficiency of the balance of plant increases with load factor.This is because, for a given power demand, the generation of hydrogen increases.This means less energy is required per unit mass of hydrogen.The Siemens Energy white paper [7] and Ferguson's doctoral thesis [8] both agree that this trend follows a hyperbolic curve.
This behaviour was captured in an equation in the Ferguson doctoral thesis.The equation used specific energy requirement (kWh/kg) instead of efficiency, as this avoids the nuance around the different ways to measure the energy content of hydrogen (higher vs lower heating value).Combining the contribution of the linear and hyperbolic components gave equation (1).
where  is the specific energy requirement, kWh/kg,  is load factor and is dimensionless,  represents the change in stack efficiency with changing load factor,  approximately represents the thermodynamic minimum energy requirement of the reaction, and  represents the energy requirement of the supporting systems.At the operational temperature of the machine studied in the thesis, the values of ,  and  were 16.18 kWh/kg, 38.46 kWh/kg and 4.26 kWh/kg respectively.With these values, the contribution of the linear and hyperbolic parts of equation ( 1) are shown in Figure 1, while combining these aspects gives the curve shown in Figure 2.  The result is that the overall system has a low efficiency at low load, reaches peak efficiency at around a medium load (approximately 52%), and then efficiency decreases as the load is increased to nominal capacity.In addition to the Siemens Energy white paper and the Ferguson doctoral thesis, this trend can be found in a journal paper from the Energiepark Mainz project, which showed real efficiency data from a 6 MW electrolyser [9].
Having introduced electrolyser efficiency, and the fact that integrated turbine -electrolyser systems are likely to deploy multiple electrolyser units working together, this work explores how overall operational efficiency could be maximised, and the resulting increase in hydrogen production compared to non-optimised operation.

Methodology
We modelled the hydrogen system using equation (1), which is simplified model of an electrolyser plant deployed in the Orkney islands in 2017 [8] [10].
We considered that the response, start up, heating and maintenance times were negligible, such that the relationship shown in Figure 2 is always valid.This is justified on the grounds that electrolyser response times are usually on a seconds timescale and start-up is usually on a minutes time scale, which is less than the 10 minute timestep over which data was available (described in Section 2.2. ) [8].For the 0.6 MW system studied in [8], heating from ambient to operational temperature at high load usually took only tens of minutes; so would have a minimal effect on the results.Maintenance periods could take longer, but these would affect all the scenarios in a similar way, so if we focus on the difference between the scenarios then this would not greatly affect the results.
The minimum turndown ratio (i.e. the minimum capacity as a percentage of nominal capacity) of the electrolyser was set to 20% [8].We use this value in all the examples given in this paper.

Optimising the operation of multiple electrolyser systems
As electrolyser systems are generally inefficient at low load factor, we should aim to minimise their time operating in this zone.We can do that by operating a subset of electrolyser units, rather than all available units, when there is low power availability.By concentrating the low incoming power to a subset of electrolyser units, we will increase their load factor and, correspondingly, their efficiency.
Another advantage of concentrating power to a subset of the electrolyser units is that the effective minimum load factor is reduced, and the operational envelope of the system increases.In the context of our model with a minimum turndown ratio of 20%, splitting the capacity into multiple units means we now only need 20% of the capacity of the split unit.To give an example, using a minimum turndown ratio of 20%, a single 5 MW electrolyser would need at least 1 MW to turn on.If this capacity is split into two 2.5 MW electrolysers, one unit could operate with 0.5 MW.The increase in operational envelope is shown graphically for a 5 MW system in Figure 3. Overlaying the operational envelope of a 5 MW electrolyser with a 20% minimum load factor over a 5 MW wind turbine power curve [11].The highlighted sections identify increased access to power when capacity is split across multiple units."4 units" represents four subsets of 1.25 MW, "3 units" represents three subsets of 1.67 MW, "2 units" represents two subsets of 2.5 MW, and 1 unit represents a single 5 MW unit.
If we use the above strategy of concentrating low incoming power to a single unit and the incoming power increases, we will reach and then surpass the optimal efficiency point of the single unit.If the load factor continues to increase, the efficiency of the single unit will start to decrease.At some point, it becomes favourable to turn on an additional electrolyser unit, as this will split the incoming power between the units, reduce their individual load factors, and be more efficient.
This concept is shown graphically for an electrolyser site with two independent units in Figure 4.The operational profile of a single unit is shown in the blue dashed line, the operational profile of both units working together is shown in the red dashed line, and the optimised operational strategy is shown by the yellow line.
Figure 4 shows how turning on one of the units (half of the site capacity) allows the system to operate with a lower site load factor (0.1 instead of 0.2).It also shows how it is possible to achieve a lower specific energy requirement for low incoming power, achieving a specific energy requirement of 55 kWh/kg at a load factor of around 0.25.Finally, it shows the point at which the second electrolyser unit should be turned on; in this case, it is at 36% of overall load factor.
Figure 5 and Figure 6 show the equivalent information when there are three and then four independent units.As before, adding further additional units allows the site to make hydrogen with an overall lower load factor, and allow the incoming energy to be used more efficiently.The figures also show the load factors at which to turn on additional units: 24% and 42% when there are three units, and 18%, 32% and 44% when there are four units.
If we superimpose the optimised strategies from Figure 4, Figure 5 and Figure 6, we get Figure 7.This confirms the above arguments: splitting the site capacity across independent units allows us to

Calculating hydrogen production
In this study, we modelled the power supply to the electrolysers using one year of wind speed data on a 10 minute timestep from an offshore meteorological station owned by the Offshore Renewable Energy (ORE) Catapult.This was converted to a power output using a power curve representing ORE Catapult's 7 MW demonstration wind turbine [12].We used the estimated 7 MW wind turbine power production as the power input to electrolysers with a single unit, two independent units, three independent units, and four independent units.These had the specific energy curves shown in Figure 7.We did this for electrolysers with capacities of 100%, 90% and 80% of the turbine, because there is an economic case for using an electrolyser capacity that is lower than the turbine capacity [13].
To briefly provide some additional information, this economic case is based on the trade-off between wind turbine capex and capacity factor, and electrolyser capex and capacity factor; having an electrolyser with a lower capacity than the turbine means less is spent on electrolyser capex and that its capacity factor will be greater than that of the turbine.This means that the contribution of electrolyser capex to the levelised cost of hydrogen can be reduced, although it increases the cost contribution of the turbine.Still, overall, this approach can somewhat reduce the levelised cost of hydrogen.This subject is explored in techno-economic literature [13].

Results -what difference does an optimised strategy make?
Table 1 shows the hydrogen produced in each of the scenarios, which have a range of ratios of electrolyser capacity to turbine capacity, and have between one and four independent electrolyser units.It shows that, as the ratio of electrolyser to turbine capacity decreases, annual hydrogen production decreases.It also shows that, as the electrolyser capacity is split across several units which are operated using an optimised strategy, hydrogen production increases.
In Figure 8, the information from Table 1 has been used to find the resulting percentage increase in hydrogen production, where the simple strategy with a single unit is the reference (baseline) result for each capacity ratio.This makes it easier to see trends in the data.
Figure 8 shows that, where the ratio of electrolyser and turbine capacities is 100%, the percentage increase in hydrogen production with additional units is greatest, ranging from about 2.9 to 3.8% in the scenarios considered here.The percentage increase with capacity ratios of 90% and 80% is less, ranging from about 2.4 to 3.2% and 2.2 to 2.8%, respectively.

Discussion
The most important inference from these results is that splitting electrolysis capacity among multiple units and using an optimised strategy can increase the quantity of hydrogen produced from a wind powered electrolysis facility, in the region of 2.2 to 3.8%, in the scenarios shown in Figure 8.This is important information for project developers and equipment manufacturers, and outlines the potential benefit of splitting a given capacity among multiple units.Figure 8 also shows other, more subtle trends that are described in Section 4.1.and Section 4.2.

The effect of increasing the number of independent units
As discussed, Figure 8 shows there is a trend of increased hydrogen production with increasing independent units.However, the benefit of additional units decreases with more units (i.e., the increase in hydrogen production becomes less with each additional unit).There are two potential causes of this.The first is that the increased "access" to power becomes more marginal with increasing units.In the example of a 5 MW electrolyser with a 20% minimum turndown ratio, moving from one to two units decreases the minimum power consumption from 1 MW to 0.5 MW; the increased access to power is 500 kW.However, going from two to three units decreases the minimum power consumption from 0.5 MW to 0.33 MW; the increased access to power is about 167 kW.
The second is that the quantity of power that the electrolyser plant is gaining access to is relatively low.In Figure 3, going from one, to two, to three, to four independent units means the minimum plant power consumption reduces from 1 MW, to 0.50 MW, to about 0.33 MW, to 0.25 MW.Compare this to the nominal plant capacity of 5 MW.At these low power levels, the actual increase of energy into the electrolyser over a year will be relatively modest compared to total consumption.

The effect of changing electrolyser capacity compared to changing the number of independent units
Figure 8 shows that, as the electrolyser -turbine capacity ratio decreases, the impact of additional units decreases.For example, with a ratio of 100%, splitting the capacity over two units increases hydrogen production by about 2.9%; with a ratio of 90%, the increase is just 2.4%.
This was unexpected; we thought that, because a smaller electrolyser would mean less hydrogen production, the additional hydrogen produced from more repeating units during low wind speeds would become more important, not less.There could be three explanations for this trend.
One explanation is that, as the electrolyser capacity reduces and has a lower minimum load factor, the increased "access" to power through additional repeating units decreases.For example, as shown in Figure 3, a 5 MW, 20% minimum load factor electrolyser, with a single unit, could use down to 1 MW.With four units rated at 1.25 MW each, it could use down to 0.25 MW.This gives a difference of 0.75 MW.By contrast, going from one to four repeating units with a 4.5 MW electrolyser capacity drops the minimum power from 0.9 MW to 0.225 MW, giving a difference of 0.675 MW.Finally, with a 4 MW unit the minimum power drop would be from 0.8 MW to 0.2 MW, a difference of 0.6 MW.As the increased access to power decreases with decreasing electrolyser capacity, the effect of additional repeating units has a smaller effect on production.
A second, related explanation is that the change in electrolyser capacity (e.g. from 100% to 90% of turbine capacity) is more important than the change in number of repeating units.For example, changing electrolyser capacity from 5 MW to 4.5 MW (i.e., from 100% to 90% of a 5 MW wind turbine capacity), and if we have a single unit, the minimum operating power consumption reduces from 1 MW to 0.9 MW, a difference of 100 kW.In contrast, if we split the 5 MW and 4.5 MW capacities across four units, the minimum operation power consumption reduces from 0.25 MW to 0.225 MW, a difference of 25 kW.Thus, the change in electrolyser capacity appears to be more important than the change in number of repeating units.
A third explanation, which would need to be further explored, is the change in average efficiency of the electrolysers with changing capacity.As the electrolyser capacity changes, the average load factor, and therefore efficiency, will change.To add technical detail, the proportion of time exposed to each power level (e.g., 0.5 -1 MW, 1 -1.5 MW, etc.), will change with electrolyser capacity.This could impact average load factor and therefore average efficiency.In turn, this could have an influence on the importance of additional power at low levels.

Other further work
The fact that the improvement reaches a plateau suggests there will be an optimal number of repeating units.This could be due to a compromise between additional units increasing production rates while also increasing complexity.This could be explored in future work.
Future work could explore the impact of changing the number of electrolyser units at a range of sites, with different wind regimes.In areas with high wind speeds, the benefit of accessing power produced at low wind speeds may be minimal.In contrast, in areas with low wind speeds, the benefit could be more significant.
Another aspect to be explored in future work is the other potential impacts of splitting a power supply between several independent electrolysers.
One potential impact is the economics of purchasing multiple small units (e.g. 4 x 1.25 MW) compared to a single large unit (e.g. 1 x 5 MW).Buying four smaller units may benefit from a discounted price.On the other hand, a single large unit may have a lower materials cost e.g.perhaps the 5 MW system can be housed in one or two containers, whereas each 1.25 MW system is likely to need its own container.A similar effect may be seen in the stacks and balance of plant -a 5 MW system may use less, though higher capacity, machinery than 4 x 1.25 MW systems.
Another potential impact is that having multiple independent units could offer greater reliability and flexibility than a single system.If one system among, for example, four, requires planned or unplanned maintenance, the overall site could continue to produce at up to 75% capacity.This would help to enable a rolling maintenance programme, instead of a full site shut down, provided that there no other limiting factors in the plant.
There may also be negative impacts.For example, installing electrical connections to several independent units could be more costly than to a single unit.Another possibility is that running the electrolysers at a lower load factor may degrade one electrolyser unit faster than the others.This could be mitigated by altering the unit which is first turned on.
Going forward, a technoeconomic analysis is required to find the overall impact of splitting electrolysis capacity between independent units on the levelised cost of hydrogen.

Conclusion
This work has explored how electrolysis plants divided into repeating units (e.g., a 10 MW facility composed of 4 x 2.5 MW units) can be operated strategically to maximise the overall efficiency of the facility.The research started in this study will help industry to identify the best overall control scheme for electrolyser plants.Early results show that splitting electrolysis capacity across independent units and optimising the operating strategy can increase hydrogen production can be increased by about 2.2 -3.8%, depending on the specific set up.

Figure 1 .
Figure 1.Contribution of the stack efficiency, represented by the linear component ( 1 =  + ) of equation (1), on the left hand axis, and contribution of the balance of plant efficiency, represented by the hyperbolic component ( 2 =   ) of equation

Figure 2 .
Figure 2. Specific energy requirement at operational temperature for a single electrolyser, based on equation (1)

Figure 3 .
Figure 3.Overlaying the operational envelope of a 5 MW electrolyser with a 20% minimum load factor over a 5 MW wind turbine power curve[11].The highlighted sections identify increased access to power when capacity is split across multiple units."4 units" represents four subsets of 1.25 MW, "3 units" represents three subsets of 1.67 MW, "2 units" represents two subsets of 2.5 MW, and 1 unit represents a single 5 MW unit.

Figure 4 .
Figure 4. Specific energy requirement curves for an electrolyser facility with two units, when one unit is operated independently and when both units are operated simultaneously.The Optimised Strategy shows when to turn on the second electrolyser; around overall site load factor of about 36%.

Figure 5 .
Figure 5. Specific energy requirement curves for an electrolyser facility with three units, along with the Optimised Strategy.

Figure 6 .
Figure 6.Specific energy requirement curves for an electrolyser facility with four units, along with the Optimised Strategy.

Figure 7 .
Figure 7. Specific energy requirements of the site using the Simple Strategy and the Optimised Strategy for two, three and four repeating units.

Figure 8 .
Figure 8.The improvement in hydrogen production using an Optimised Strategy for various numbers of repeating units and ratios of electrolyser to turbine capacity.

Table 1 .
Tonnes of hydrogen produced in one year from a range of electrolyser capacities, operational strategies, and repeating units.