Investigating solutions for solar farms connected to congested grids A master Thesis TOM

(1)

Investigating solutions for solar farms connected to congested grids

A master Thesis TOM

(2)

Abstract

(3)

Introduction

Throughout the last century, the growing usage of finite fossil fuels as an energy source has had a negative effect on the climate and geopolitical relations. Since the 1970s there has been a push for more sustainable energy production, and with the global Paris accord, the 2008 EU Climate Change Package and the 2019 Dutch climate accord, there has been a continued international effort that continues to this day to transition to a sustainable energy system. As a result renewable energy sources (RES) such as solar and wind have already become common in our society and produce a significant amount of energy on our electricity grid (REN21, 2019). As an alternative for fossil fuels, it is the goal that through these sources energy can become sustainable, free of carbon emissions and ready for the long future (Letcher, 2016). It is expected that the contribution of RES will increase in the coming years, as their deployment is promoted by the Dutch government (Ministerie van Economische Zaken en Klimaat, 2019) in an effort to reduce carbon emissions by 49% by 2030. Unfortunately, at this moment it is not possible to use all the energy produced by RES. This problem is a result of the variability of their energy supply (Letcher, 2016). The production of energy by means (Denholm and Hand, 2011) of solar and wind has a large dependence on the weather, and for solar, the time of day and the time of year (Badwawi et al., 2015); (Denholm and Hand, 2011)) . This variability introduces problems for the balance of the electric power system. In the electric power system, the supply of electricity needs to be balanced with the electricity demand and with network losses at all times to be safe, dependable and stable. To keep the electricity power system balanced flexibility is required (Denholm and Hand, 2011); (Group Technology and Research DNV GL, 2017) . In fossil fuel-based power systems the only source of variability is on the demand side, where the demand for energy tends to be higher in the morning and the evening, following consumer patterns ((Group Technology and Research DNV GL, 2017)). With the expected increase of variable renewable energy sources additional production side variation will be introduced to this existing variation on the demand side. This means an increase in flexibility is necessary to balance this additional variation. A lack of flexibility will not only make the system less stable, but it will also halt progress in the deployment of renewable energy sources as they will not be allowed access to the electricity grid due to the variation they would introduce. The lack of flexibility sources already becomes apparent in some parts of the Netherlands, like Oost-Groningen and Drenthe, where new production sites for renewable energy are not allowed to access the net for this very reason (Enexis, 2019) . In a large portion of the rural area in the Northern Netherlands there is limited or no capacity for new RES connections, see figure 1.1. This problem is expected to become very relevant, as 1 GW of extra solar production, divided over

(6)

Figure 1.1: Grid availability in the Northern Netherlands (Enexis, 2019)

(7)

Figure 1.2: Overview of physical and institutional layout of the Dutch Electricity Market (De Vries, 2013) (Enexis, 2019) . This will be elaborated on in the theoretical background. This means that it is up to RES power plant operators in these congested areas to implement some method to reduce variability on a plant level and make sure their plant can access the grid. In this research we will be looking into reducing the variability of solar power plants by means of energy storage to allow them to deliver energy to a congested power grid. In this research the focus will be on solar farms in the Northern Netherlands as a specific case, as in that area there is around 1 GW of solar production planned that mostly cannot be connected to the grid because of the variability they would introduce. This research will be performed in cooperation with Repowered, a company that aims to provide solutions to RES production plants with limited grid access. They work together with Solar Fields, a company that builds, operates and maintains solar plants in the Netherlands. This research will be performed using real-life data from different solar plants of Solar fields and be aided by expertise from employees of Repowered. The research will consist of a literature overview followed by a simulation model, investigating different ways by which to increase flexibility of solar farms. First different policies an energy storage system can operate by will be modeled and analyzed. Thereafter a subset of different storage methods will be compared using the simulation model, an overview of existing methods to analyze flexibility in power systems is already given in the theoretical background, together with justification for our specific subset of methods. In the end the economic viability of the energy storage systems will be analyzed. This research will be conducted along the lines of the following research question: How can energy storage systems increase flexibility of solar farms in the Northern Netherlands and allow them access to a congested grid? This is subdivided in three subquestions:

1. By means of what operational policies can energy storage systems increase flexibility of solar farms?

2. What are the advantages and disadvantages of using different energy storage systems to introduce flexibility to solar farms?

(8)

Chapter 2

Theoretical Background

In this theoretical background we will provide an overview of the state-of-the-art research concerning reducing variability introduced by RES power plants. We will also investigate different types of energy storage systems that could facilitate this variability reduction and different methods in which energy storage systems are used in this way. This will help by giving different operational policy options for energy storage operation as well as indicating the most promising energy storage methods. It will also provide examples of other simulation models and give a direction for building the simulation model used in this research. In our literature search we will focus on a specific type of RES, solar farms, as these are under investigation in this paper.

2.1 Flexibility in electricity grids

(9)

(10)

of energy delivered to the grid can be significantly reduced. An example would be a solar plant connected to a cooled storage facility, where the cooling policy could be adjusted to cool exactly when solar energy production is highest, reducing net variability. However, this is only possible if such flexible demand exists in the local area. Grid flexibility is also facilitated by means of economic incentives on the energy market. In the Netherlands the main electricity market is the day-ahead market, where supply and demand are traded a day in advance. Another market is used in the Netherlands is the imbalance market where on a minute-based time interval prices are adjusted to unforeseen imbalances on the net (Tennet, 2019). For a more in-depth review of the players in the Dutch Energy Market, see appendix B. Energy storage systems could buy on the day-ahead market when prices are low and sell the energy when prices are high (Breeze, 2018e) . In this way energy storage systems contribute to grid flexibility by following the economic incentives. Battery storage systems are capable of responding within seconds and can also provide short-term flexibility, that is necessary for frequency control or imbalance market trading, as well (Hsieh and Anderson, 2017); (Letcher, 2016) . Following economic incentives, like responding to prices on the imbalance market instead of only the day-ahead market could also be a way of increasing the economic viability of battery storage systems (Breeze, 2018e) . Another option to increase flexibility of solar plants, and the one investigated in this research, is by means of energy storage systems. An energy storage system is a system that can store energy for an amount of time, and release it later. This can be particularly helpful for solar production, in that it can reduce peak production (in the early afternoon) by storing energy, and release this energy in the evening when the solar plant does not produce anymore, a technique called peak shaving. A lot of different energy storage systems exist and they have different characteristics. In figure 2.1 different storage systems are compared by their maximum power output and the duration that they can maintain that production, parameters also deemed most important by Ela (Ela et al., 2016) . Since long-term flexibility is the focus of this research only systems capable of providing energy over a period longer than 15 minutes are considered. This makes battery systems, like LA, Li-Ion, NaS and FRB, suitable, as well as Compressed Air Energy Storage (CAES), Pumped Hydro Storage (PHS) and power to gas (P2G). An overview of energy storage systems and their applicability to the case of the Northern Netherlands, together with a thorough explanation of figure 2.1 is provided in appendix C.

(11)

In this research the storage systems modeled will be battery storage systems and power-to-gas (P2G) storage systems as these are capable of providing energy over longer periods of time (several hours for battery systems, and even longer for P2G systems). These were chosen mainly for their availability, their efficiency and their costs, as described in appendix C.

2.2 Providing flexibility using energy storage systems

There is a lot of research in using energy storage as a way of increasing system flexibility. For example (Ho et al. (2012); Kavadias et al. (2018)) investigate energy storage as a way of reducing variability in the area of micro grids, smaller local electricity grids, where there exists electricity demand within the micro grid and there is no connection to a main national grids. Their research focuses on improving self-consumption of energy produced by household PV- systems or small solar farms, that is using more of this locally produced energy by shifting the energy supply to meet the demand. (Clegg and Mancarella, 2015) (Hovsapian, 2017) (Krishan and Suhag, 2018) and others research the use of energy storage as a way of providing utility and ancillary services to the electricity power system. They take do not from the perspective of a RES power plant, but address variability existing on a full grid. These studies also do not investigate solar plants that are connected to the main grid but still need to address their variability, like is the case in the Northern Netherlands. Some research in similar areas does exist and can be used however. Shaw-Williams et al. (2019) did a case study in Queensland, Australia, where they investigated using battery storage systems to store privately produced solar power for self-consumption. They used a power flow model to simulate the situation, simulating demand and weather patterns using Monte-Carlo-simulation on a 30 minute basis. It was found that introducing battery storage systems could increase solar penetration of the local grid from a theoretical maximum of 20-40% determined by Gaunt et al. (2017) to 50-75% due to the reduced variability. Resch et al and Moretón et al also investigated the use of battery storage systems to improve flexibility of the local grid (Moretón et al., 2013) (Resch et al., 2019). Resch et al modelled a battery with different operational policies, both as community energy storage, mainly mitigating local variability from PV generation and domestic demand, and as primary control reserve, providing short-term flexibility to the grid based on economic incentives such as imbalance pricing. They used a simulation model, using real imbalance data for the primary control reserve case and stochastic PV-production with stochastic demand for the community energy storage case. Moretón et al used a power flow model using DIgSILENT PowerFactory. Similarly Clegg et al investigated P2G solutions using a power flow model in the UK (Clegg and Mancarella, 2015) , in tandem with a gas flow model, to find the impact of combining both grids. Kim et al investigated using P2G to mitigate variability in a grid consisting of a nuclear base load and RES (Kim et al., 2018). They used a mathematical model based on Modelica to model the situation, using control mechanisms such as Proportional Integral controllers to respond to stochastic energy production. Due to their focus on response rate they modeled on the timescale of seconds. This means that they aim to use P2G as frequency control, or for imbalance trading. They found that P2G can effectively respond to grid incentives on that timescale and increase overall flexibility. However, such a short-term approach is unsuitable for our specific case, as long term reduction in variability is required to meet the variability of solar energy production. Also they used a grid approach to flexibility, dealing with variability on a grid, whereas in the case of the Northern Netherlands an increase in plant flexibility is required to be allowed access to the grid in the first place.

2.3 Contribution to literature

(12)

(13)

Chapter 3

Methodology

(14)

Figure 3.1: Naive storage policy

model will be to maximize profit. Profit, taking into account reduced grid connection costs because of the reduced flexibility, will be an operationalized version of flexibility, describing the overall effectiveness of the storage system at enabling the solar farm to deliver its power to the grid. Flexibility in our model, which we interpret as reducing the maximum (peak) power output whilst maintaining the average power output, thus smoothing the production and requiring a lower tier grid connection, will be the main focus. Since the energy storage systems store power to discharge it later, meaning it will be discharged a different price, price optimization is also part of the operational policy, increasing profit. A full mathematical model of this is included in appendix D. A large amount of modelling tools to investigate grids with high penetration levels of RES already exist (Ringkjøb et al., 2018). These tools could be of aid in producing the simulation model.

Some data, like the PV*SOL data and some solar plant specific data used in this research will be provided by the company Repowered, a partner in this research, who investigate ways for solar farms to produce on congested networks. They mainly investigate this for plants by Solar Fields, a Dutch company constructing and operating solar plants in the Netherlands, many of which are in congested areas. The naive policy will follow the design by (Shaw-Williams et al., 2019) and is described in figure 3.1. The policy only takes into account the production limit set by the grid connection, the current energy level in the storage system and a set price threshold to determine when prices are right to buy or sell. This decision can be based on daily price patterns, since prices tend to be highest around in the mornings and the evenings (Shaw-Williams et al., 2019). This naive policy can be extended upon by including for example:

- buying electricity from the grid when prices are low - Including imbalance pricing

(15)

Chapter 4

Results

4.1 Scenario design

This research investigates two different scenarios for solar farms connected to a congested grid. The first scenario is the most basic case possible where a solar farm is connected to a congested grid. As such it will shed the most light on the impact of storage systems and restricted grid access on the economic performance of a solar farm, without having other factors at play. To analyze the impact of different grid restrictions, multiple restriction levels are investigated, as well as different storage system parameters such as efficiency, capacity and type. This scenario is based on a real case from Repowered and Solar Fields, where the question from the company is whether the solar farm can still be profitable at a restricted grid connection. The second scenario is a more complicated case, where besides solar production on a congested grid there also exists flexible and inflexible demand. The inflexible demand will change the overall profile of the net energy production, and the flexible demand will be modeled to improve flexibility and will be compared with battery storage and P2G. This scenario will shed light on the impact of demand patterns on the performance of flexibility measures, as well as the effect of demand-response policy. Scenario 2 is also based on a real case from Repowered, where the question from the company is whether or not demand-response is an effective policy, and if storage systems can be combined with a demand-response policy. Scenario 2 can also be compared to scenario 1, to see the difference a demand pattern makes on the effectiveness of storage systems.

4.2 Scenario 1: A standalone solar farm scenario

4.2.1 Input data analysis

(16)

1.5 MW 3 MW 6 MW 10 MW System revenue e 250.600 e 390.300 e 508.700 e 527.900

Curtailment % 53% 27% 4% 0%

Table 4.1: Revenue and curtailment at different cutoff scenario’s

11, as well as 18 and 21. At night, between 0 and 5 prices are lowest, and they are moderate at the other hours. It becomes clear that when mapping the daily solar production on this price curve that daily peak production occurs just between the two daily price peaks.

For this particular park the effects of a grid connection limit of 6 MW, 3MW and of 1.5 MW are investigated. A lower grid connection reduces the maximum amount of kW that can be delivered to the grid. However, it also generates lower capital costs and operational costs, payable to the grid operator. These costs were disregarded in this analysis, as they differ from grid operator to grid operator, location to location and connection capacity to connection capacity. The connection limits chosen can also be seen in figure 4.1. These specific limits were chosen to represent a range of different curtailment ratios. Without any extra measures, a limit of 6 MW would lead to about 4% curtailment of power, or 380 MWh, representing a loss of revenue of about 4%, or e 20.000. A cutoff at 3 MW results in about 27% curtailment of power, or 2.7 GWh, representing a loss of revenue of about 26% or e 137.000. More details can be seen in table 4.1.

To test for the sensitivity of the model to different datasets, several other datasets were used. The sensitivity to price differences was tested with a synthesized price dataset, which was a moving average of the original price data from 2018 with step size 3 (MA3) as well as with another price dataset from 2016. Both an inflation corrected version of the 2016 price data, as well as the original 2016 price data have been used. The correction for inflation was performed to make the revenue comparable with the 2018 data. To achieve this the average price of the 2016 price data has been set equal to the average price in 2018, applying linear transformation to the data. The original price data was 60% below the 2018 data. As can be seen in figure 4.5 and 4.6, the MA3 dataset slightly smooths out the price differences. The sensitivity for the variance in solar production is tested by using another dataset generated in PV*SOL, at a different location in the Netherlands. This alternate production (AP) data has a slightly larger average production, a difference of 13%.

(17)

Figure 4.2: Average daily production pattern of the solar farm

(18)

Figure 4.4: Average daily price of electricity in 2018

(19)

(20)

Figure 4.7: Storage level of P2G storage system at different capacities with 55% efficiency at a grid restriction of 3000kW

4.2.2 Linear programming approach

(21)

(22)

(23)

Figure 4.10: Average daily charge and discharge behavior of a P2G storage system with 500 kW capacity and 55% efficiency

To compare storage policy behavior of P2G storage and battery storage in the LP solutions a specific case was chosen, that of a grid restriction of 3000kW, a 500kW P2G system with 55% efficiency and a 1000kWh battery system with 90 % efficiency. These systems both have a charge and discharge capacity of 500kW. From the figures 4.10 and 4.11 it can be seen that the P2G storage charges more during the day within a broad range, whilst the battery storage system charges in a smaller range, mainly at peak solar production hours. The battery storage system also discharges more in peaks. For both systems, the charge and discharge policies follow the price pattern and production pattern observed in the data (see figure 4.4 and figure 4.2) but battery systems have more extremes in their policy. It appears that the strict capacity limitation, which limits the amount of hours it can charge and discharge per day, makes the system favor certain specific hours. This behavior has been observed in the other cases as well.

(24)

(25)

When we analyze a similar situation with a battery storage system see figure 4.13, at a grid restriction of 1500kW and an efficiency of 90%, we see a much smaller decrease in curtailment ratio, with a reduction from 53% to around 49% at a capacity of 1000kWh and finally to around 47% at a capacity of 3000kWh for the base case data. The ROI is also observed to decrease with increasing capacity. When comparing the MA3 data to the normal price data it can be observed that the ROI for the battery storage system decreases , similarly to the P2G system. However, at a smoother price distribution the curtailment ratio is clearly half a percent to one percent lower than in the case with the original data, whereas in the P2G storage curtailment was independent of this price difference. It appears that with less extreme price differences, it is more optimal for the battery to focus more on reducing curtailment. At different grid restrictions similar results were obtained, with the battery system curtailment ratio decreasing at smoother prices, and the P2G curtailment ratio staying the same at smoother prices. The graphs of ROI and curtailment ratio at different grid restrictions can be seen in Appendix E.

Figure 4.13: Curtailment ratio and ROI values obtained using the normal and the MA3 price data at a grid restriction of 1500kW, using a battery storage system with an efficiency of 90%

(26)

(27)

Figure 4.15: Storage level of the naive P2G model at 1500kW grid restriction, at 55% efficiency

4.2.3 Naive model

In this segment the results of the naive model are analyzed. Similar parameters to the LP problems were used, meaning around 200 scenarios were analyzed. The naive model uses historical price data to determine optimal times to charge and discharge energy. The average daily price per hour is determined, and the optimal hours for charging and discharging are determined this way. For the naive battery storage the system always charges between 14:00 and 16:59, as well as when production exceeds the grid restriction. For the naive P2G storage, the naive system only charges when production exceeds the grid restriction, as on average the efficiency loss of P2G exceeds the price difference so no designated hours for storage will be feasible. Both naive models discharge in the intervals 8:00-10:59 and 18:00-20:59. In practice the battery system will not discharge in the morning as it discharges completely within two hours from full capacity, so only the evening interval will be used. At high grid restrictions, like 1500kW and 3000kW, the production exceeding this restriction is larger than the amount of energy able to be fed back to the grid during the designated hours. This means that for the naive model for P2G systems the storage level will increase seasonally, similarly to the LP-solution described in figure 4.7. The naive model and the LP model show similar results. This can be seen in figure 4.15 and 4.16. It seems that all feature increases in ROI with efficiency, which is expected. However, it can also be observed that ROI of the naive model is within 1.5% of the LP-policy most of the time. This signifies that quite a simple policy can perform close to the optimal policy.

(28)

Figure 4.16: Storage level of the naive P2G model at 3000kW grid restriction, at 55% efficiency

(29)

(30)

4.2.4 Conclusion of scenario 1

From the data analysis several conclusions can be drawn. The ROI of the storage models decreases almost linearly with increasing storage capacity and grid restriction, and increased with increasing efficiency. The P2G models, both in the LP and the naive model are the best at alleviating curtailment, as can be seen in decrease in the curtailment rate achieved in P2G models versus the battery storage models. The optimal policy at 55% P2G efficiency could even reach a ROI of 8.8%, reducing curtailment from 53% to 43% at a grid restriction of 1500kW. Because of this, P2G systems perform very well in scenario’s with strict grid restriction and very poorly in scenarios with little grid restriction. The battery storage systems decrease in ROI with increasing curtailment as well, but less so, as they seemingly profit mainly from price arbitration during the day, possible due to the high roundtrip-efficiency and cannot store as much energy that would otherwise be curtailed as P2G can. A significant difference between the LP solution and the naive P2G model can be seen in figures 4.16 and 4.7. The naive model discharges every day at the designated hours, whereas the LP solution waits for all the high prices, occurring mainly later in the year. The maximum storage level reached is therefore very different for the two models, with a difference of around factor 7. The LP-solution results in ROI values of respectively 5.7%,5.3% and 4.9% at capacities of 500kW, 1000kW and 1500kW, with 55% efficiency and a 3000kW grid restriction whilst the naive P2G model results in a ROI of respectively 4%, 3.6% and 3.3% in the same scenario.

As expected, the LP solutions yielded better results, but the naive P2G model achieved ROI values of above 50% of the LP ROI values achieved in similar scenarios. This indicates a very significant utility of the storage system stems from reducing curtailment and time shifting of energy production, as the naive model implements both these techniques and the LP solutions show similar behavior. Since reducing curtailment does not improve much when using accurate forecasting, as it can be performed in real-time, only the time-shifting of production used in the naive model can be further improved upon by implementing improved production and price forecasting. As seen in figure 4.17, the naive battery model approaches the LP-solution more in terms of ROI, coming close to 20% lower ROI than the optimal solution. This suggests there is little room for improvement for battery systems policies within the parameters used in this analysis. Also, the optimal solution in the optimal scenario only achieved a ROI of around 6%. With the low lifetime of current batteries and the higher prices of batteries with longer lifetimes, this ROI is not enough to pay back the batteries within their lifetime at the moment. New technologies, like the VRFB battery have a lifetime of 20000 cycles, which is sufficient to pay back the investment. However, these technologies are too expensive at the moment, see appendix C. Lower prices also influence the ROI, as in this analysis a cost ofe 300 per kWh is used, which is an optimistic estimate when comparing it with the values obtained from literature in appendix C. In contrast, the P2G storage policy could achieve 8.8% ROI at 55% efficiency in the optimal scenario with the optimal policy, and 7.5% with the naive policy in the same conditions. At the current lifetime of P2G technologies, with a maximum of around 90,000 operational hours this ROI seems sufficient to pay back the investment. However, this ROI occurs at the 1500kW grid restriction. At the 3000kW grid restriction this ROI drops to 5.8% for the optimal model and 4% for the naive model, and the ROI drops to around 2-1% for a 6000kW grid restriction. This means that a P2G system is only feasible when the grid access is really limited.

(31)

4.3 Scenario 2: Rooftop PV with local demand

4.3.1 Input data

In this scenario 0.8 MWp of solar production, mounted on the roof of a cooled warehouse facility is analyzed. These panels are scheduled to be in production within the 2020. The configuration is East-West, with half of the panels facing the East, and the other half of the panels facing the west. The production data are simulated using PV*SOL (201, 2019) This particular location has limited grid access, with an upper limit of 50 kW with the possible relaxation to 150kW. The location also has local demand, which is plotted together with the power supplied by the solar panels in figure 4.19. This demand is partially flexible, as the coolers can be switched on or off at any moment during the day and at current operation are switched on at night and during the afternoon break. Since there is more demand than production in this facility, the revenue turns out negative. The solar production in this case reduces costs, and the storage and demand response policies will be aimed at reducing costs. The difference between case without any policy and the case with policy will be analyzed.

In this scenario the assumption is that the demand can be varied by a fixed amount, Qmax, as long as the total

(32)

(33)

Figure 4.20: Resulting additional revenue of the demand response policy and the curtailment ratio achieved at a grid restriction (Pmax) of 50 or 150 kW

4.3.2 Linear programming approach

To determine the optimal case for demand shifting a linear programming problem was formulated in the PuLP pythonic library and solved with the CBC solver. The mathematical model in appendix D was chosen as the basis, with the following additions:

Demand shifting cannot shift demand higher or lower than the mean demand +/- the Qmax

Ai+ Qi≤ max(Amean+ Qmax, Ai) ∀ i (4.1)

Ai− Qi≥ min(Amean− Qmax, Ai) ∀ i (4.2)

Also the total demand shifting on a day (starting at hour 0) must total 0.

23

X

n=0

Qi+n= 0 ∀ i mod 24 = 0 (4.3)

Finally where in appendix D Ai was used, this was replaced with Ai+ Qi.

(34)

production. However, the ROI achieved was only around 6%. This is comparable to the ROI achieved in the 3MW restriction in scenario 1, a 30% restriction of peak production. Besides the ROI being comparable in these two scenarios, curtailment ratios were also comparable, around 20%. A difference between the scenarios is the demand, which changes the net production profile. However, even with this difference, it seems that when the curtailment rate is comparable, the ROI seems to be comparable as well. An example is that at a grid restriction of 3000kW in scenario 1, the P2G storage achieved 5.8% ROI with 20% curtailment. In scenario 2 the P2G storage achieves 6% at a 21% curtailment ratio. Similar results are found when observing the other cases. It can be concluded that curtailment ratio is a good indicator of P2G performance within these parameters, a conclusion that was drawn at scenario 1 as well, where curtailment ratio significantly impacted ROI.

When comparing P2G to the demand shifting, it can be seen that 25kW capacity of P2G (at a grid restriction of 50kW) can reduce curtailment from 25.7% to 21.1% and 50kW capacity can reduce this further to 17.1%. In comparison, the demand shifting of 30kW can reduce curtailment to 22.8% with 50kW demand shifting reducing this further to 20.1%. P2G is a lot better at reducing curtailment at similar capacities, but demand shifting still performs better than the battery storage systems tested at reducing curtailment, which could only achieve 21% curtailment with 150kWh capacity. For reference see figure 4.21 and figure 4.22.

(35)

(36)

(37)

Figure 4.23: Average daily demand for different policies at a 50kW grid restriction

(38)

Grid restriction Maximum shiftable load (kW) Additional revenue Curtailment ratio

50 30 e -630 22.4%

50 50 e -970 20.4%

150 30 e -170 8.6%

150 50 e -240 7.4%

Table 4.2: Results of the naive model

4.3.3 Naive model

Multiple attempts were made at a naive model, however all naive models resulted in negative returns. A final naive model was used, with the policy that nightly load (between the hours 2 and 8) is shifted to the day (between the hours 10 and 16). This model resulted in the (negative) results shown in table 4.2:

This discrepancy between the optimal solution and the naive model can be attributed to multiple factors. First of all, the demand data has a very high variance. The averages shown in figure 4.19 do not reflect a reoccurring pattern, and the standard deviation of the demand at the hours of the day from the average demand at those hours was 36 on average. This makes it difficult to set up a naive heuristic with good performance in every situation. As electricity prices are higher during the day, performing demand shifting without reducing curtailment results in a net loss. For an algorithm to be effective in this scenario a solar forecast and a production forecast are necessary to determine when demand shifting is feasible. Developing such an algorithm is beyond the scope of this research, but an example could include a linear programming approach, optimizing the demand response over a forecast period of around 3 days. When using forecasted data instead of actual data, the difference between the optimal model and such a model depends on the accuracy of the forecasts, which is the subject of a lot of research at this moment, like Alsharif et al. (2019) for example. Forecasting services are also commercially available and used in practice in similar situations.

4.3.4 Conclusion scenario 2

The combination of demand and solar production produced a different effective production profile. Demand re-sponse was found to be an effective policy to reduce variability in this scenario, able to decrease the curtailment of energy and provide extra revenue in the optimal scenarios. However, developing a naive policy that was effective at this against APX prices failed. The reason for this seems to be that demand response is more dependent on knowledge about the future. The LP solutions showed day-to-day differences in the demand shifting, a policy that requires more knowledge about the production and demand in the future to implement in simulation.

(39)

(40)

Chapter 5

Discussion and conclusion

5.1 Discussion

In this research the performance of storage systems in grid restricted solar parks was analyzed from a multitude of angles. As expected, higher efficiencies increased performance, and higher capacities gave diminishing returns. However, this research excluded some factors that could impact the economic performance of these systems. First of all, the price data used was derived from the APX-market, as this is the most used market and it is quite predictable. However, it is possible to get higher prices on the imbalance market for example, at higher risk. Standalone battery systems operating on the imbalance market do exist and are commercially viable in certain situations (Lund, 2019). These do however require forecasting and more sophisticated algorithms, and it is unclear how effective a storage system can be by both targeting the imbalance market, as well as reducing curtailment from a solar park. The possibility for storage systems to charge from the grid was excluded as well. This could increase the performance of high efficiency storage systems, like battery storage, by performing arbitrage between low prices at night and high prices in the morning. The reason this was excluded in this research is that currently in the Netherlands this is not feasible due to the energy tax payable over bought electricity (Ministerie van Economische Zaken en Klimaat, 2019). However, if storage systems are excluded from this taxation, as is the case in for example gas distribution networks, this could be a significant source of revenue for battery storage systems. Also the cost reduction that comes with a lower grid connection was excluded in the analysis, but in certain cases this can be a significant benefit of the storage system, large enough to potentially make it viable (Enexis, 2019). All these factors could make storage systems more viable, as it was found that the main issue of these systems is the price that is obtained for their services compared to their capital cost.

(41)

Another factor that could increase the performance is the improvement of the naive storage system policies. The policies developed in this research are really simple and can easily be improved upon. For the naive battery storage the system always charges at certain hours, determined from the yearly average price, as well as when production exceeds the grid restriction. For the naive P2G storage, the naive system only charges when production exceeds the grid restriction, as on average the efficiency loss of P2G exceeds the price difference so arbitrage would result in a net loss. A more sophisticated policy could allow P2G systems, for example, to possibly profit of arbitrage in certain situations, even with the low efficiencies of P2G.

It is assumed that including price and production forecasting can improve the performance of a storage system policy, but how much this improvement can be is limited by the optimal solution found. Since in this research it was found that even the optimal solution was not feasible in most situations, it seems that policy improvements are less important to achieving feasibility than improving price or decreasing costs.

On of the many assumptions made in this research is that it investigated static grid restrictions only. The reason for this is that in current Dutch law only such grid restrictions are allowed and defined. However, it is theoretically possible to make grid restrictions dynamic. An example could be a very low limit during the hours that solar has the highest production and a higher limit during the other hours. This would encourage the use of energy storage systems, and would significantly reduce the impact of variable energy production on the grid. Enexis (2019) has stated that they are interested in such measures as a possible way to reduce grid congestion.

When assessing the demand-response policy it must be noted that a lot of assumptions were made about the nature of the flexible demand, as only the combined data from the flexible and the inflexible demand were avail-able. A more holistic and comprehensive approach is necessary to find out the effectiveness of demand response in different scenarios. Necessary is a clear difference between inflexible and flexible demand, as well as the limits within the flexible demand can operate, such as its maximum capacity. These were not provided in this research, so assumptions had to be made to cover these factors. Also, further efforts into the development of an effective and generally applicable demand-response policy for grid-restricted are necessary.

When comparing these results to other literature on this topic, it can be seen that the case for storage systems becomes a lot more feasible when it alleviates curtailment. In other literature, this is less researched as there is no focus on grid-restricted solar parks. When comparing the use of storage systems in this research with the use in research by Shaw-Williams et al. (2019), it can be seen that using storage for self-consumption, as they do, has a higher benefit than using storage for time-shifting the load to the grid. In general, it can be concluded that grid restrictions have an impact on solar production that is different from the problems with solar production that are researched in literature, where a grid based approach is taken.

5.1.1 Managerial implications

(42)

Whilst it was found that under current market conditions the large capital expenses associated with storage systems are not worth the investment, when faced with grid restriction, a demand response policy allows a lower curtailment rate and increased revenue for very low cost. When flexible demand is available, it should be looked at as a first option to reduce variability and making grid restricted solar energy feasible.

In general, the feasibility of storage systems, or demand-response policy depend for a large amount on the electricity market conditions, especially electricity price. This research shows the infeasibility of these systems at the current costs and prices, but the same approach can be followed in new market conditions to figure out the feasibility of the discussed options in new conditions. Especially when subsidies apply to the energy produced, which increase the price of the energy, or when other factors make the energy produced more valuable than other energy, there is the potential that the systems discussed in this research are viable. The close relation between the naive model and the LP model shows that an LP approach can give an approximate (over)-estimation of the feasibility of storage systems and can be a quick decision support tool to decide whether or not storage systems might be feasible in the new market conditions.

5.2 Conclusion

(43)

5.2.1 Directions for further research

As discussed in the discussion, a lot of potential influential effects have been excluded in this analysis. Therefore on major direction for future research is formulating a comprehensive model and policy that includes other pricing mechanisms, such as imbalance pricing, as well as including forecasting of demand and prices. Such a model could give more insight in the feasibility of storage systems in the scenarios of solar parks with grid restrictions, and provide the operational policy to make them feasible.

(44)

Bibliography

PV*SOL, 2019.

Daniel Akinyele, Juri Belikov, and Yoash Levron. Battery storage technologies for electrical applications: Impact in stand-alone photovoltaic systems, 11 2017. ISSN 19961073.

M. I. Alizadeh, M. Parsa Moghaddam, N. Amjady, P. Siano, and M. K. Sheikh-El-Eslami. Flexibility in future power systems with high renewable penetration: A review. Renewable and Sustainable Energy Reviews, 57:1186–1193, 5 2016. ISSN 18790690. doi: 10.1016/j.rser.2015.12.200.

Mohammed H. Alsharif, Mohammad K. Younes, and Jeong Kim. Time series ARIMA model for prediction of daily and monthly average global solar radiation: The case study of Seoul, South Korea. Symmetry, 11(2), 2 2019. ISSN 20738994. doi: 10.3390/sym11020240.

Rashid Al Badwawi, Mohammad Abusara, and Tapas Mallick. A Review of Hybrid Solar PV and Wind Energy System. Smart Science, 3(3):127–138, 1 2015. ISSN 2308-0477. doi: 10.1080/23080477.2015.11665647. URL http://www.tandfonline.com/doi/full/10.1080/23080477.2015.11665647.

Paul Breeze. An Introduction to Energy Storage Technologies. Elsevier, 2018a. doi: 10.1016/b978-0-12-812902-9. 00001-8.

Paul Breeze. Hydrogen Energy Storage. In Power System Energy Storage Technologies, pages 69–77. Elsevier, 2018b. doi: 10.1016/b978-0-12-812902-9.00008-0.

Paul Breeze. Large-Scale Batteries. In Power System Energy Storage Technologies, pages 33–45. Elsevier, 2018c. doi: 10.1016/b978-0-12-812902-9.00004-3.

Paul Breeze. Pumped Storage Hydropower. In Power System Energy Storage Technologies, pages 13–22. Elsevier, 2018d. doi: 10.1016/B978-0-12-812902-9.00002-X. URL https://linkinghub.elsevier.com/retrieve/pii/ B978012812902900002X.

Paul Breeze. The Environmental Impact of Energy Storage Technologies. In Power System Energy Storage Tech-nologies, pages 79–84. Elsevier, 2018e. doi: 10.1016/b978-0-12-812902-9.00009-2.

Stephen Clegg and Pierluigi Mancarella. Integrated Modeling and Assessment of the Operational Impact of Power-to-Gas (P2G) on Electrical and Gas Transmission Networks. IEEE Transactions on Sustainable Energy, 6(4): 1234–1244, 10 2015. ISSN 19493029. doi: 10.1109/TSTE.2015.2424885.

(45)

Paul Denholm and Maureen Hand. Grid flexibility and storage required to achieve very high penetration of variable renewable electricity. Energy Policy, 39(3):1817–1830, 3 2011. ISSN 03014215. doi: 10.1016/j.enpol.2011.01.019. Anarghya Dinesh, Sharon Olivera, Krishna Venkatesh, Mysore Sridhar Santosh, Murugesan Geetha Priya, Ina-muddin, Abdullah M. Asiri, and Handanahally Basavarajaiah Muralidhara. Iron-based flow batteries to store renewable energies, 9 2018. ISSN 16103661.

E. Ela, M. Milligan, A. Bloom, A. Botterud, A. Townsend, T. Levin, and B. A. Frew. Wholesale electricity market design with increasing levels of renewable generation: Incentivizing flexibility in system operations. Electricity Journal, 29(4):51–60, 2016. ISSN 10406190. doi: 10.1016/j.tej.2016.05.001. URL http://dx.doi.org/10.1016/ j.tej.2016.05.001.

Elestor. HBr flow batteries, cost and effectiveness, 2019. URL https://www.elestor.nl/wp-content/uploads/ 2019/05/Elestorbrochure.pdf.

Enexis. Gebieden met schaarste, 2019. URL https://www.enexis.nl/zakelijk/duurzaam/ beperkte-capaciteit/gebieden-met-schaarste.

C. T. Gaunt, E. Namanya, and R. Herman. Voltage modelling of LV feeders with dispersed generation: Limits of penetration of randomly connected photovoltaic generation. Electric Power Systems Research, 143:1–6, 2017. ISSN 03787796. doi: 10.1016/j.epsr.2016.08.042. URL http://dx.doi.org/10.1016/j.epsr.2016.08.042. Group Technology and Research DNV GL. Flexibility in the Power System. Technical report, Group Technology

and Research DNV GL, 2017.

Wai Shin Ho, Haslenda Hashim, and Ho Wai Shin. Integrated design of renewable energy decentralized power plant comprising energy storage for off-grid eco village. Global Journal of Technology and Optimization, 3, 2012. URL www.pcoglobal.com/gjto.htmRE-P21/GJTO.

Hannele Holttinen. Wind integration: Experience, issues, and challenges. Wiley Interdisciplinary Reviews: Energy and Environment, 1(3):243–255, 11 2012. ISSN 20418396. doi: 10.1002/wene.18.

Rob Hovsapian. Role of Electrolyzers in Grid Services. Technical report, US Department of Energy Fuel Cell Technology Office, Houston, 2017.

Eric Hsieh and Robert Anderson. Grid flexibility: The quiet revolution. Electricity Journal, 30(2):1–8, 3 2017. ISSN 10406190. doi: 10.1016/j.tej.2017.01.009.

Lawrence E Jones. Renewable Energy Integration Practical Management of Variability, Uncertainty, and Flexibility in Power Grids. Elsevier Inc., New York, London, San Francisco, San Diego, 2014. ISBN 1865843830.

K. A. Kavadias, D. Apostolou, and J. K. Kaldellis. Modelling and optimisation of a hydrogen-based energy storage system in an autonomous electrical network. Applied Energy, 227:574–586, 10 2018. ISSN 03062619. doi: 10.1016/j.apenergy.2017.08.050.

Jong Suk Kim, Richard D. Boardman, and Shannon M. Bragg-Sitton. Dynamic performance analysis of a high-temperature steam electrolysis plant integrated within nuclear-renewable hybrid energy systems. Applied Energy, 228:2090–2110, 10 2018. ISSN 03062619. doi: 10.1016/j.apenergy.2018.07.060.

(46)

Trevor M. Letcher. Storing Energy: With Special Reference to Renewable Energy Sources. Elsevier Inc., 4 2016. ISBN 9780128034408.

Peter Lund. Advances in Energy Systems: The Large-scale Renewable Energy Integration Challenge. John Wiley & Sons Ltd, Hoboken, 1th edition, 2019. ISBN 9781119508281.

B. Lyseng, T. Niet, J. English, V. Keller, K. Palmer-Wilson, B. Robertson, A. Rowe, and P. Wild. System-level power-to-gas energy storage for high penetrations of variable renewables. International Journal of Hydrogen Energy, 2018. ISSN 03603199. doi: 10.1016/j.ijhydene.2017.11.162.

Geoffrey J. May, Alistair Davidson, and Boris Monahov. Lead batteries for utility energy storage: A review, 2018. ISSN 2352152X.

Ministerie van Economische Zaken en Klimaat. Klimaatakkoord - 28 juni 2019. Technical report, Ministerie van Economische Zaken en Klimaat, 2019. URL https://www.rijksoverheid.nl/documenten/rapporten/2019/ 06/28/klimaatakkoord.

Rodrigo Moret´on, Eduardo Lorenzo, and Javier Mu˜noz. Comparison of decentralised and centralised grid- com-patible battery storage systems in distribution grids with high PV penetration. (January 2012):2–6, 2013. doi: 10.1002/pip.

P Ramamurthy. Operations research, volume 73. 1971. doi: 10.1016/S0076-5392(08)62705-8. REN21. Renewables 2019 Global Status Report. Technical report, REN21, Paris, 2019.

Matthias Resch, Jochen B¨uhler, Birgit Schachler, Rita Kunert, Andreas Meier, and Andreas Sumper. Technical and economic comparison of grid supportive vanadium redox flow batteries for primary control reserve and community electricity storage in Germany. International Journal of Energy Research, 43(1):337–357, 1 2019. ISSN 1099114X. doi: 10.1002/er.4269.

Hans Kristian Ringkjøb, Peter M. Haugan, and Ida Marie Solbrekke. A review of modelling tools for energy and electricity systems with large shares of variable renewables, 2018. ISSN 18790690.

O. Schmidt, A. Gambhir, I. Staffell, A. Hawkes, J. Nelson, and S. Few. Future cost and performance of water electrolysis: An expert elicitation study. International Journal of Hydrogen Energy, 2017. ISSN 03603199. doi: 10.1016/j.ijhydene.2017.10.045.

Damian Shaw-Williams, Connie Susilawati, Geoffrey Walker, and Jeremy Varendorff. Towards net-zero energy neighbourhoods utilising high rates of residential photovoltaics with battery storage: a techno-economic analysis. International Journal of Sustainable Energy, pages 1–17, 9 2019. ISSN 1478-6451. doi: 10.1080/14786451.2019. 1668394. URL https://www.tandfonline.com/doi/full/10.1080/14786451.2019.1668394.

Tennet. Imbalance pricing system. Technical report, Tennet, 2019. URL https://www.tennet.eu/fileadmin/ user_upload/SO_NL/ALG_imbalance_pricing_system.doc.pdf.

Javier Valdes, Axel Bastián Poque González, Luis Ramirez Camargo, Meyl´ı Valin Fenández, Yunesky Masip Macia, and Wolfgang Dorner. Industry, flexibility, and demand response: Applying German energy transition lessons in Chile. Energy Research and Social Science, 54:12–25, 8 2019. ISSN 22146296. doi: 10.1016/j.erss.2019.03.003. Chao Zhang, Yi Li Wei, Peng Fei Cao, and Meng Chang Lin. Energy storage system: Current studies on batteries

(47)

(48)

Appendix A

Grid congestion in the Northern

Netherlands

In the Northern Netherlands large portions of the local distribution grids are congested. This means that either no grid connection is possible for RES energy production, or that connection is limited by a strict upper limit. It is expected for these areas that the congestion problems will get worse in the short term, as increasing grid capacity can take years and over 1 GW of solar production alone is scheduled to be ready in 2021, see figure A.1 (Enexis, 2019) . For many solar plants this will mean that they cannot deliver their electricity to the grid, or only very limited amounts of their production, as the grid connection limits are set to values far below their peak production. In this case, all the excess electricity produced will need to be curtailed, making renewable energy more expensive and reducing the return on investment of these solar plants.

(49)

Appendix B

Overview of the Dutch Energy Sector

(50)

(51)

Appendix C

Overview of energy storage systems

A major way of providing flexibility to the power system is by using energy storage systems. Breeze describes energy storage systems as systems that can take up and store electrical energy, holding it securely and making it available for delivery at a later time (Breeze, 2018a) . Energy storage comes with an efficiency cost, as the process of storing and delivering the energy loses some energy along the way (Letcher, 2016); (May et al., 2018) . The efficiency of storing energy and delivering it later is usually called the round-trip efficiency. This efficiency is an important factor in comparing different energy storage systems. There are many different energy storage systems, and they vary widely in efficiency, the amount of power they can provide as well as the amount of time they can provide that power for. These last two characteristics are deemed important parameters for any source of flexibility on the grid by Ela et al. (2016).

(52)

Figure C.1: An overview of energy storage methods and their effective timespans (Group Technology and Research DNV GL, 2017)

Name Capital Costs Cycles Efficiency LCoS Sources VRFB (Flow) 600 $/kWh 20000 70-85% 0.035-0.043 $/kWh (May et al., 2018)

Li-Ion 400 $/kWh 2500-5000 90-95% 0.084-0.17 $/kWh (Letcher, 2016)

NaS (Flow) 360-800 $/kWh 2500-4500 75-90% 0.089-0.26 $/kWh (Dinesh et al., 2018) (May et al., 2018) Lead Acid 200 $/kWh 2000 85% 0.12 $/kWh (Akinyele et al., 2017)

HBr (Flow) 350 $/kWh 10000 65-75% 0.05 $/kWh (Elestor, 2019) Table C.1: An overview of battery storage

or bidaily basis these batteries would wear out around 6 to 10 years. Flow batteries can last around 10000 cycles or even more, and are still under development. A problem with them is that they are still generally more expensive than solid batteries (Breeze, 2018e) . Large battery systems are expensive, ranging from $200/kWh to $800/kWh, but battery costs have been dropping rapidly the past decade and are close to economic viability (Breeze, 2018c) (Letcher, 2016). An overview of most promising battery types can be found in table C.1. The capital costs are the cost per kWh of capacity and the cycles represent the typical number of cycles the battery can go through without wearing out. The efficiencies are the roundtrip efficiencies and the Levelized Cost of Storage (LCoS) are the capital costs per physical kWh that is supplied by the battery, according to the formula below.

LCoS = CapitalCosts

(53)

Name Capital Costs Efficiency Lifetime Sources AEC (P2G) 1000-1200$/kW 62-82% 60000-90000 h (Schmidt et al., 2017) PEMEC (P2G) 1860-2320$/kW 67-82% 20000-60000 h (Schmidt et al., 2017)

Table C.2: Overview of P2G processes

(54)

Appendix D

Mathematical model

Parameters

In this mathematical model the following parameters are used:

Let Pi be the day-ahead, or APX-price, at time i. This will supplied to our model as data, obtained from Tennet

(2019), and is known beforehand in real life situations, as well as in our simulation.

Let Si be the production of the solar farm at time i. This production will be supplied as data to our model, and

this data is generated in PV*SOL.

Let Ai be consumption directly connected to the solar farm at time i.

Let Li be the net production of the facility at time i.

Let Yi be the amount of energy delivered to the grid at time i and Ymax the maximum amount allowed to be

delivered to the grid in a time period i Let N be the timeframe of the simulation.

The storage capacity, efficiency and the charge and discharge rates are all dependent on the specifics of the storage system chosen. These specifics are constants within each specific scenario, and therefore they are parameters in the mathematical model.

Let Cmax be the maximum storage capacity for the energy storage system.

Let η be the charging efficiency of the energy storage system, meaning that of an amount of energy delivered to the energy storage system at time i, Ii, only η · Ii actually is stored in the energy storage system.

Let θ be the discharging efficiency of the energy storage system, meaning that of an amount of energy delivered from the energy storage system at time i, Oi, only θ · Oi is actually delivered to the grid.

Let α be the maximum charge rate of the energy storage system in energy per time period. Let β be the maximum discharge rate of the energy storage system in energy per time period.

Choosing to charge of discharge follows from the storage policy, which is under investigation in this research. The storage policy is a decision variable, so the following binary variables are decision variables, representing the storage policy.

Decision variables:

The storage policy will determine how much is the battery is charged and discharged at what time. This means these factors are decision variables in the mathematical model. Let Ii be the amount of energy delivered to the

storage system at time i.

Let Oi be the amount of energy delivered from the energy storage system to the grid at time i.

Cithe amount of energy stored in the storage system at time i. This value follows from the storage decisions taken

(55)

Wi is the amount of energy curtailed at time i.

Assumptions:

The net production of energy, available for storage or grid delivery, is assumed to be the difference between total production and consumption. Since direct consumption is always more efficient in terms of costs than storing the energy or delivering it to the grid we will work with net production as the input to the battery policy model.

Li= Si− Ai ∀ i (D.1)

The storage level of the energy storage system equals the level at time i-1 corrected for energy delivered to and from the storage system at at time i and the storage level at time 0 is 0.

Ci= Ci−1+ η · Ii− Oi ∀ i (D.2)

C0= 0 (D.3)

The net amount of energy delivered to the grid at time i equals the net production at time i minus the energy delivered to the energy storage system, minus the amount of energy curtailed and plus the amount delivered from the energy storage system at time i.

Yi= Li− Ii− Wi+ θ · Oi∀ i (D.4)

Constraints:

The amount of energy delivered to the grid can never exceed the maximum allowed by the grid operator, Ymax

0 ≤ Yi ≤ Ymax∀ i (D.5)

The energy stored should never exceed the maximum storage level.

0 ≤ Ci ≤ Cmax∀ i (D.6)

Since curtailment cannot be negative, the following constraint keeps curtailment bounded:

0 ≤ Wi∀ i (D.7)

The amount of energy supplied to the energy storage system should never exceed: 1. The charging rate α

2. The space left in the energy storage system Cmax− Ci

3. The net amount of energy produced Li

Ii≤ min(α,

Cmax− Ci−1

η , Li) ∀ i (D.8)

However, Ci is also bounded by its own constraint in equation D.6 and following equation D.2 this bounds Ii, so

this constraint can be reduced to:

Ii≤ min(α, Li) ∀ i (D.9)

(56)

2. The amount of energy left in the energy storage system Ci

3. The amount of energy still allowed to deliver to the grid Ymax− Yi

Oi≤ min(β, Ci−1,

Ymax− Yi

θ ) ∀ i (D.10)

The energy delivered to the grid is bounded by its own constraint in equation D.5, bounding Oi via equation D.4.

Also the storage level Ciis bounded, in equation D.6 and following equation D.2 this bounds Oi, so this constraint

can be reduced to:

Oi≤ β ∀ i (D.11)

The aim of the model is to maximize profits in each scenario, in comparing outcomes of different scenario’s capital costs also play a role so, Life Cycle Analysis (LCA) will be the deciding factor when comparing scenario’s. Objective function: M aximize : N X i=1 Pi· Yi (D.12)

In the research of of battery storage systems, see Appendix C, the roundtrip efficiency is used to denote the efficiency of the energy storage and the efficiencies of the charging and discharging are not known separately. We will correct the energy stored by multiplying I with the full roundtrip efficiency, instead of only the charging efficiency, i.e. η = roundtrip and θ = 1 when considering battery systems. Although the roundtrip efficiency combines the charging and discharging efficiency, we apply it directly to only the charging process. This means this model underestimates the battery capacity necessary to store the energy with a factor of the discharge efficiency. This simplification is made because the separate charge and discharge efficiencies are not known for all energy storage systems reviewed.

For some energy storage systems self-discharge over time is a problem, but in this model self discharge is assumed to be negligible, see appendix 3.

In our first scenario we will consider a standalone solar farm, meaning Ai= 0 ∀ i

We consider the production Si as input data in the model. The data used is generated in simulation program

(57)

Appendix E

Scenario 1 - Linear Programming

solutions

To illustrate the patterns and the ROI values found in the linear programming solutions, the battery system solutions at 90% efficiency and the P2G solutions at 55% efficiency at the different grid connections are listed in the tables below.

E.1 Tables with results of normal data

1 MWh 2 MWh 3 MWh System cost e 300.000 e 600.000 e 900.000 Curtailment % 50% 47.8% 45.3% Battery Revenue e 18.500 e 35.100 e 50.600

Battery ROI 6.2% 5.9% 5.6%

(58)

1 MWh 2 MWh 3 MWh System cost e 300.000 e 600.000 e 900.000 Curtailment % 25% 23% 21% Battery Revenue e 14.800 e 28.000 e 41.000

Battery ROI 4.9% 4.7% 4.6%

Table E.2: Results from linear program: battery storage with different system configurations at 3MW grid restric-tion, with 90% efficiency

1 MWh 2 MWh 3 MWh System cost e 300.000 e 600.000 e 900.000 Curtailment % 3% 2.5% 1.8% Battery Revenue e 9.800 e 18.400 e 26.000

Battery ROI 3.3% 3.1% 2.9%

Table E.3: Results from linear program: battery storage with different system configurations at 6MW grid restric-tion, with 90% efficiency

0.5 MW 1 MW 1.5 MW System cost e 500.000 e 1.000.000 e 1.500.000 Curtailment % 43% 34.3% 27%

P2G Revenue e 43.800 e 81.100 e 112.800

P2G ROI 8.8% 8.1% 7.5%

Table E.4: Results from linear program: P2G storage with different system configurations at 1.5MW grid restriction at 55% efficiency 0.5 MW 1 MW 1.5 MW System cost e 500.000 e 1.000.000 e 1.500.000 Curtailment % 20.8% 15.7% 11.6% P2G Revenue e 28.900 e 53.400 e 73.800 P2G ROI 5.8% 5.3% 4.9%

Table E.5: Results from linear program: P2G storage with different system configurations at 3MW grid restriction at 55% efficiency 0.5 MW 1 MW 1.5 MW System cost e 500.000 e 1.000.000 e 1.500.000 Curtailment % 2.3% 1.2% 0.05 % P2G Revenue e 11.000 e 19.800 e 26.600 P2G ROI 2.2% 2.0% 1.8%

(59)

Figure E.1: ROI and curtailment ratio of a battery system at 90% efficiency at a 3000kW grid restriction

(60)

(61)

(62)

Investigating solutions for solar farms connected to congested grids A master Thesis TOM