Pre-Master’s Research Paper

(1)

Pre-Master’s Research Paper

Sizing battery storage for different groups of households that

utilize PV systems and peak shaving strategies

22

nd

June 2020

Technology and Operations Management

University of Groningen, Faculty of Economics and Business

Author: Rein Hiemstra

(2)

Abstract

Purpose: The purpose of this paper is to determine how shared household batteries should be

sized for different groups and group sizes of households using PV systems and peak shaving strategies. This intermitted nature of PV systems causes problems in the stability and reliability of the grid. With the help of batteries and peak shaving strategies, demand and feed-in peak loads can be reduced, which decreases stress on the grid.

Methodology: A simulation study is performed on an energy system involving different groups

and sizes of households, using different sizes of batteries per installed kWp and peak shaving strategies. With the help of 120 scenarios, in which parameters were changed, differences and similarities between groups were found.

Findings: The findings show that a 1kWh battery is too small for most groups and that a 3kWh

and 5kWh battery can reduce peak loads in demand and feed-in around 40%-50% for all groups. In groups ranging from 6 to 15 households, no significant additional benefits were found when increasing to a 5kWh battery, except for groups of 3 households that also performed better when batteries got significantly larger. It is also found that bigger groups have a smaller potential in a failure of the peak shaving strategy at any given battery size because the demand is smoother. However, smaller groups were able to reach higher reductions with a 5kWh battery because less energy needed to be reduced at any given moment due to the higher peaks in demand.

Implications: The findings are indicating that a 3kWh is most fitting for groups of 3 households

up to 15 when looking to decrease peak loads between 40% and 50% with peak shaving strategy that uses limitations to the demand and feed-in peak loads. Curtailment levels will occur between 2.96% and 9.65%. Groups of 3 households could scale up to a 5kWh battery to reduce peak loads by up to 60%, which comes with a higher curtailment level of 18,03%.

Originality: This study contributes to existing literature and problems as it addresses possible

(3)

3

Content

1. Introduction ... 4

2. Theoretical background ... 5

3. Methodology ... 7

3.1 Introduction to the energy systems... 7

3.2 The conceptual simulation model ... 7

3.2.1 Model inputs ... 8

3.2.2 Model outputs ... 11

3.2.3 Content... 12

3.2.4 Assumptions and simplification ... 14

3.3 Experimental setup ... 14

4. Findings... 15

4.1 Base case scenario ... 15

4.2 Storage size & Different groups ... 17

4.3 Sensitivity analysis ... 20

5. Discussion and conclusion ... 22

(4)

1. Introduction

Increasing activities in stationary storage research and development can be seen in many different countries around the world (Yang et al., 2011). The growing interests suggest a bright outlook for developing improved stationary energy storage technologies for future electric grids. With

lithium-ion battery prices dropping faster than expected over the last years (BloombergNEF, 2019), household-level battery storage is now emerging in different sizes (Agnew & Dargusch, 2017). With technology still improving and the need for even more cost reduction, battery storage utilization in households are still minor. Battery storage benefits like reducing peak loads during specific time intervals throughout the day are becoming of greater interest. (Rahimi, Zarghami, Vaziri, & Vadhva, 2013).

At the same time, increasing problems arise in both the energy supply to the households, as well as the feed-in energy by solar energy in the Netherlands (Liander, 2020). This problem is

occurring because of the mismatch in energy production by this fast-growing renewable energy source and household demand peak loads (Luo, Wang, Dooner, & Clarke, 2015). Solar energy is produced around midday and the highest peak load in demand is occurring in the evening, making it a challenging to maintain a stable and reliable power grid (Luthander, Widén, Munkhammar, & Lingfors, 2016; Klingler & Schuhmacher, 2018; Luo et al., 2015).

Previous studies have already shown that the use of battery storage could bridge some of the current capacity grid issues. With the help of batteries, both peaks in household demand and feed-in can be reduced with peak shavfeed-ing strategies that set limits on the energy flows (Luthander et al., 2016). An interesting new finding in the study of Luthander et al. (2016) is that shared batteries for groups of households show better results in peak shaving and self-consumption than a stand-alone household battery.

Future groups of households can apply shared batteries and peak shaving strategies to help reduce stress on the power grid based on these new findings. Still, information is missing on how these configurations need to look. Different battery sizes can be used to utilize peak shaving strategies within groups of households that can be configured in various group sizes and the size of the household based on the sum of their daily load profile (Braun, Büdenbende, Magnor, & Jossen, 2009; Klingler & Schuhmacher, 2018).

(5)

5 In this study, various battery sizes and peak shaving strategies within different groups and sizes of households are investigated with the goal see if differences occur between groups. This study contributes to existing literature and problems as it addresses possible future battery connected household configurations so that every household can be supplied with renewable energy with minimum changes to the existing grid.

The remainder of this paper is organized as follows. In the theoretical background, existing studies around the subject’s energy storage, PV systems, and peak shaving will be described. Then in the methodology, a description of the simulation study is provided. The outcomes of the different scenarios, produced by the simulation, will be discussed in the findings. Finally, there will be a discussion and conclusion section in which the results are reflected on existing

knowledge, and concluding remarks on this study will be given.

2. Theoretical background

Previous studies have not yet answered what the effects of different groups and group sizes of households are on the battery size when using peak shaving strategies. Batteries can help smoothen the intermittency of PV production and bring more balance to the grid to improve reliability (Yang et al., 2011; Uddin et al., 2018). Using the right battery size is essential to reducing peak loads and can be established with the preferred approach called peak shaving (Uddin et al., 2018). Where previous research focused on reducing peak loads in demand or feed-in with optimal battery storage (Mulder, Ridder & Six, 2010; Moshövel et al., 2015), this study will focus on both demand and feed-in peak loads. Where previous research focused on single household configurations or differences between single households and grouped households (Braun et al., 2009; Luthander et al., 2016), this study will look at different groups and group sizes of households. The following paragraphs go deeper into previous conducted studies and addresses differences with this study that aims to add new insights.

Mulder et al. (2010) and Moshövel et al. (2015) both looked at optimizing the storage size for grid-connected residential PV systems and reducing the feed-in peak loads, but they did not consider reducing demand peak loads. This study will include demand peak loads as it is part of the current power grid problems addressed by Liander (2020). Mulder et al. (2010) showed a method to optimize the storage size for residential PV systems to cover the peak loads to the grid. The research also did not include limitations in battery storage, which are needed to extend the life cycle. State of Charge (SOC) is a limitation for the inner state of the battery to function properly (Yang et al., 2011). SOC levels will be used in this study to make the simulation more real. Moshövel et al. (2015) looked to lower the peak feed-in power by using self-consumption. A forecasting method was used that showed a high potential for PV systems with battery storage to relieve stress on the power grid on the feed-in side, instead of standalone PV systems.

(6)

self-consumption. Braun et al. (2009) used different battery sizes connected with PV systems in German households looking to increase self-consumption and Luthander et al. (2016) did research in the effects of grouping households together with PV systems and shared lead-acid battery storage. Both studies did not address different groups of households with shared batteries. Conclusions in both studies were affected by costs, which affected the results for the optimal size of the batteries. By not looking at the costs, this study will look for the optimal size of batteries for groups of households wanting to reduce peak loads by using peak shaving. More reasons for not involving costs are the expected reductions in costs for batteries in the upcoming years (BloombergNEF, 2019). Braun et al. (2009) found that with 2.3kWh batteries per household, it is possible to increase the self-consumed PV energy with 45% in comparison with systems without batteries. The same amount of storage can be found in the study of Luthander et al. (2016) but resulted in significantly better results because of the shared battery configuration. Batteries with 2kWh capacity per kWp installed PV increased the self-consumption with 14% when all

households shared batteries. The significantly better results with a shared battery make it interesting to look at the effects when different group sizes of households and sized households are used, which will be conducted in this study.

The research of Luthander et al. (2016) also provides a straightforward method for peak shaving with a battery. Batteries are charged when there is a surplus in PV power and discharged when there is a surplus in consumption. Curtailment can help reduce the feed-in peak to the grid when the battery is full, and PV energy is still being produced. The main drawback of this strategy is that it is uncertain if the peak shaving will last until the end of the demand peak and is therefore not fully sufficient for this study.

An addition to better-utilizing peak shaving strategies is provided in the studies of Barzkar & Hosseini (2018) and Leadbetter & Swan (2012), which used plateau values that set limits only on the demand peak loads for single households. Differently and new to these studies is that

households will get grouped together with shared batteries, and limitations are given for both demand and feed-in peak loads. Results in the study of Barzkar & Hosseini (2018) showed that setting the limit on 62% of the maximum peak load in combination with the ideal storage was the best result for single households. Leadbetter & Swan (2012) showed slightly fewer reductions in peak loads varying between 42% and 49%.

Slightly different use of the plateau value can be seen in the studies of Wang & Wang (2013) and Tan et al. (2017). Both studies looked for the optimal power load curve using load leveling and secondary storage in the form of EV’s with power supply directly from the grid. Differently to these studies is that a stationary shared battery will be supplied with renewable energy generated with the PV systems to reduce stress on the grid. This means that the battery only gets charged during the hours when there will be enough solar.

(7)

7 reducing both the peaks in demand and feed-in in a grouped household configuration. The power limits settings mentioned by the studies can be used as parameters for this study. By giving variations in these settings, different peak shaving strategies can be defined.

3. Methodology

3.1 Introduction to the energy systems

An overview of a grouped household configuration is shown in Figure 1. Renewable energy is being produced by the PV systems located at the roofs of the households. The energy can be directly used through self-consumption, stored in the shared battery, or feed into the grid. At the same time, energy from the grid can be used to supply the households of energy when the PV production cannot meet the demand of the households. The amount of energy that will be stored or used during any time (t) depends on the peak shaving strategy, battery capacity, energy demand, and PV energy produced. The peak shaving strategy consists of the power limitation (Plimit), at which point the battery needs to take over the grid and the feed-in limitation (PGlimit) when curtailment occurs.

3.2 The conceptual simulation model

The objective of this study is to gather insight into the battery sizes that different sizes of grouped households and sizes of households can use to reduce demand and feed-in peak loads with peak shaving strategies. These loads are the obtained peak power for the grid to meet the household's demand if there is not enough PV energy produced and the feed-in power into the grid when there is a surplus in PV energy. This objective will be reached by using a simulation so that it can become clear what different strategies and sized batteries to use in different sizes and combinations of households.

(8)

A simulation study is suitable for this research because it is possible to model complex and flexible situations (Robinson, 2004). It can be described as the process of designing a model of a real system and conducting experiments with the created model for the purpose of either

understanding the behavior of the system or evaluating strategies. One of the real strengths of a simulation is that it can simulate systems that already exist as well as systems that could be brought into existence (Shannon, 1988). Robinson (2004), addresses the importance of the conceptual model for simulations. The model design has an impact on the aspects of the

simulation study. Key components of the conceptual model, besides the already stated objective of this research, are the inputs (parameters), outputs (variables), content, assumptions, and simplification.

3.2.1 Model inputs

Inputs for the model are the energy demand per hour of combined households (Ht), the energy production of the PV systems from the households in hourly quantities (Pt), battery storage capacity (Cbatt), and the peak shaving strategy containing the grid limitation for the demand (Plimit) and feed-in grid power (PGlimit). The total amount of periods indicated with t consist of 8760 periods, which make up for one calendar year. All inputs will be further explained and can be seen in Table 4, which shows al the content of the simulation.

Household demand

In the simulation, three different household types are used to configure different groups of households in terms of group size and variability between household size based on their energy consumption. The household types are related to the household types mentioned by Braun et al. (2009). The different groups vary between 3 and 15 households. This is because 3 households are the smallest groups that can be configured to still be a group, and 15 households are the

maximum of complete household data from 2009 that could be provided by Liander (2020). The total amount of households can be seen in Table 1 and Appendix 1. The type of households does not play a role in the simulations but only indicate differences in size and demand.

Individual Ht profiles are the main input for the simulation, which can be seen in Figure 2 and were used to create 6 different groups (Appendix 2). Because multiple variations in the smaller groups could be made, two different household groups are made for the groups of 3 and 6. The “Extreme” groups have a bigger difference between the annual energy consumption of each household than the “Equal” groups. This way, there is a distinct difference between household groups with the same number of households. It is necessary to use individual profiles instead of the average profiles. Averaging would lead to a higher simultaneity of electricity demand than actually occurs on an individual level (Klingler & Schuhmacher, 2018). An example of the sum of all the 15 household demands can be seen in Figure 3, which show a peak in demand at 6 and 7 pm.

Household types Number of households Avg. demand

Single parent family 3 2,3kW

Childless family 6 3,5kW

Family with children 6 4,5kW

(9)

9

Figure 4 PV-production (PVGIS, 2016) PV production

For the PV production, solar radiation data from The Netherlands is used. This is obtained from PVGIS (2016). As can be seen in Figure 4, PV production is lower during the winter and gets more intents during spring and summer. This calendar year data will be used in combination with the amount of kWp of PV installed on each household.

(10)

Capacity battery

Li-ion batteries store electrical energy in electrodes made of Li-intercalation compounds. A number of battery pack units are integrated, depending on the size of the energy storage, ranging from few kW to MW level (Yang et al., 2011). Li-ion batteries have a better cycle life and degradation than lead-acid batteries Luthander et al. (2016). Storage must be sized to meet the required power demand during the highest peak of the year to prevent a system failure

(Leadbetter & Swan, 2012). The battery sizes are set at 1kWh, 3kWh, and 5kWh per installed kWp, based on the studies of Mulder et al. (2010) and Luthander et al. (2016). Larger batteries are also added in sizes of 10kWh and 15kWh, to see if the trends in the results of the smaller batteries continue as significant larger batteries are used. The SOC levels are set at 85% and 15%, which are shown for each battery size in Table 2.

Peak shaving strategy

In this study, peak shaving strategies will be modeled to reduce the peak loads in demand and feed-in. Peak loads are the highest overall loads that occur in an energy system. It is a sensitive factor for the grids, as it occurs occasionally and takes place only for a small percentage of the time in a day (Uddin et al., 2018). The peak shaving strategies represent the two limits in the system, which act as the plateau values. Both parameters for demand and feed-in limit will be the same so that potential peak load reductions for demand and feed-in will be uniform. This means that for the peak shaving strategies, both the Plimit and PGlimit will be set at the same parameter in the simulations but differ, as can be seen in Table 3.

These parameters will be a percentage of the highest demand peak load of the grouped

households in the calendar year for Plimit (Leadbetter & Swan, 2012), and a percentage of the total amount of installed kWp that is installed in the grouped households for PGlimit (Litjens, Worrell & Sark, 2018). Increasing the percentage will mean that the limits will be stricter, which should result in lower peaks loads if possible. Figure 5 shows an example of a peak shaving strategy.

Battery capacity per kWp 1kWh 3kWh 5kWh 10kWh 15kWh

Maximal SOC (85%) 0,85kWh 2,55kWh 4,25kWh 8,5kWh 12,75kWh Minimal SOC (15%) 0,15kWh 0,45kWh 0,75kWh 1,5kWh 2,25kWh

Table 2 Battery SOC levels

Table 3 Plateau values

Plateau value Notation Based on:

(11)

11 In this example, the battery is fully loaded and thus all surplus energy is being feed-in into the grid until PGlimit is reached. Plimit can for instance be set at 40% of the highest peak in the calendar year (21,6kw) and PGlimit is then also set at 40% of the total amount of kW per installed kWp.

3.2.2 Model outputs

Based on the inputs, the model will generate different outputs for every scenario in each

simulation in hourly quantities. A scenario is a run of the simulation with a set of conditions set by the model parameters. By changing the parameters, the scenario is changed (Robinson, 2004). The outputs are:

- Storage inventory - Energy flows - Lost energy (Ploss).

- Success, non-failures, and failures

Storage inventory, energy flows, and Ploss outputs are a list if numbers than can be transformed in graphs. By calculating averages and percentages, insight can be delivered with tables. Success, non-failure, and failures are indicators for the working of a scenario. When demand and feed-in peak loads are reduced as intended by the limits, the scenario is a success. When the highest peak load is not reduced, this will lead to failure. A failure event would indicate that the given

parameters are undersized for the given scenario (Leadbetter & Swan, 2012). When the highest peak load is reduced, but not as much as intended, this will lead to a non-failure.

Demand peak load reductions will be calculated when the scenario is completed, by comparing the highest demand peak load with the highest peak load received from the grid in the calendar year. Feed-in peak load reductions will be calculated by comparing the highest feed-in peak load that could be delivered with the PV system of the scenario and what the highest feed-in peak load is after the scenario is run. This will result in a reduction percentage.

(12)

By combining this information, it is possible to see which battery size is the best fit for the different groups of households when using various peak shaving strategies. The decision for the right storage size will be based on how much the peak loads are reduced, the amount of inventory that is used, and energy is lost over time.

3.2.3 Content

The content of the simulation is as follows:

Figure 6 is an overview of the conceptual model representing the simulation model of this research. There are two main flows in the conceptual model in the form of a logic flow model.

No PV production

When the value of Pt is not greater than 0, Ht will be checked to determine if Plimit is exceeded or not. If Plimit is not exceeded, Ht will be meet by only using grid power (PG). When Plimit is exceeded, the next step is to check if the shared household battery SOC is above 15%. If the SOC is not above 15%, there will not be any energy to shave the peak load. All Ht will be meet by using PG. If the SOC is above 15%, the battery will be used to meet the Ht that is above Plimit.

PV production

When the value of Pt is greater than 0, Ht will be checked to determine if it exceeds Pt or not. When Pt does not exceed Ht, all Pt is used directly for Ht. The remaining energy needed to meet Ht is done with PG. If Pt does exceed the household demand, Ht will be meet by only using Pt. When there is energy left. produced by Pt, the SOC of the battery needs to be checked. When the SOC level is under 85%, the battery can be charged with solar energy until it reaches 85% SOC.

Table 4 Simulation content

Component Detail Included/Excluded Comment

PV PV production (Pt) Included Affects peak shaving potential Curtailment (Ploss) Included Linked to feed-in

limitation Demand Households (Ht) Included Hourly data of grouped

households Power limitation (Plimit) Included Constraint of network Storage Capacity (Cbatt) Included Based on household

demand

Costs Excluded Out of the scope

SOC (15%-85%) Included Affects battery utilization Inverter losses Excluded Out of the scope

(13)

13 When SOC reaches 85%, and there is still more energy left or it was already at 85%, energy can be feed-in to the grid. Energy can be feed-in to the grid until PGlimit is reached. When the PGlimit is reached, the left-over energy is being curtailed.

All flows end up at the decision if it is the end of the horizon. When the last hourly data of the whole year is reached, the scenario stops. The scenario will loop hour after hour until this last hour is reached.

(14)

3.2.4 Assumptions and simplification Assumptions

- The grid capacity is limited and is the reason why limitations need to be setup.

- Ht and Pt data is known in advance. This is necessary because Plimit will be based on the highest demand peak loads of the year.

Simplification

- The model will use the data for one calendar year. This way seasonal effects, like reduced solar production in winter months, are included.

- Costs, inverters, and cycle losses are not taken into considerations.

- The size of PV systems per household is based on the annual energy consumption. This will be 1 kWp of PV per annual MWh that is consumed and is common in the

Netherlands (Litjens et al., 2018). This gives a more realistic setting for each scenario. Because of this, each combination of households will have a different amount of combined kWp.

3.3 Experimental setup

The research consists of five simulations. Each simulation contains 24 scenarios that will be run in Excel. Simulation 1 starts with 1kWh per installed kWp. Mulder et al. (2010) and Luthander et al. (2016) found that the first kWh has the most impact on the energy flows and is already

capable of reducing peak loads in single households. The parameters for the peak shaving strategy are set at 40%, 50%, 60%, and 70% of the highest demand peak load and the total

amount of installed kWp. These values are based on the demand peak load reductions in the study of Barzkar & Hosseini (2018) and Leadbetter & Swan (2012), mentioned in the theoretical

background. A 40% feed-in limit should result in a curtailment percentage between 3% and 5% (Matthiss, Stellbogen, Eberspächer, & Binder, 2015). Because these are values that have proven to be working, differences and similarities among the outputs of the different groups in each simulation can be analyzed.

(15)

15 In the second analysis, the effects of different sized batteries will be compared by analyzing all five simulations. Differences in capacity and the range of the different batteries used in each simulation are shown in Table 6. In this analysis, the behavior of the various storage sizes within the different groups will be investigated by looking at differences and similarities in peak shaving results.

The last analysis is a sensitivity analysis that focuses on the relationship between the various outputs of the second analysis and a parameter that affects the results. This will be done by looking at the effects of peak shaving strategies on the results of all simulations. Including analysis on this parameter will help shape a strong, more in-depth conclusion.

In each scenario of every simulation, hourly data on solar production and demand will be modeled for the duration of one calendar year. At the start of every scenario, the battery inventory will be charged for 50%. This way, there is no need for a load-up period in the first days and scenario results are not influenced by an already full or empty battery. An example of the output of a scenario can be seen in Appendix 4

4. Findings

In this chapter, we discuss the influences of the different parameters on the outcomes, which can be found in Appendix 5. This will be done with the three analysis described in 3.3.

4.1 Base case scenario

• 1kWh battery is too small for most of the groups

• Bigger household groups use more energy from the battery at any given moment

Table 7 shows that both the Plimit and PGlimit are met throughout the calendar year with minor energy losses in scenario 1. A 40% reduction of the demand and feed-in peak load is established in this configuration, confirming that a 1kWh battery can reduce peak loads successfully.

Main simulations Cbatt Range (kWh)

Simulation 1 1kWh per kWp 11,1-54,8 Simulation 2 3kWh per kWp 33,3-164.4 Simulation 3 5kWh per kWp 55,5-274,0 Simulation 4 10kWh per kWp 111,0-548,0 Simulation 5 15kWh per kWp 166,5-822,0

Table 7 scenario 1 results

(16)

Throughout the year, 83,60% of the time, the shared battery inventory is at its maximum. If we visualize the inventory in Figure 7, it is clear to see that the inventory of scenario 1 never reaches the minimum SOC level of 1,7kWh (15%). At the end of week 1, the inventory level of the battery is at its lowest point (2,8kWh) in the calendar year. At this point, the highest peak load of the year needs to be reduced, which can be seen in Appendix 6.

In table 8, All similar scenarios in aspects of battery size and peak shaving strategy are

summarized below. The table shows that all scenarios reduced the demand peak loads and have a very similar curtailed energy percentage of around 3%. Differences can be found in the amount of demand peak loads reductions and maximum inventory level percentages of the battery.

Only two scenarios can reduce the demand peak loads successfully to 40%. Visualization of the inventory flows of each scenario can be found in Appendix 7. The reduction percentages in the table and flows in the graphs indicate that the battery is too small because inventory levels drop significantly when high peak loads need to be reduced. These occur only a few times over the calendar year.

The maximum inventory percentages indicate a relation between the amount of inventory that is used throughout the year and the number of households grouped. The shared battery is used more efficiently in bigger groups than in small groups. This effect will be further explained in 4.2.

Figure 7 Battery inventory scenario 1

(17)

17

4.2 Storage size & Different groups

• Bigger batteries can lead to bigger demand peak load reductions

• Batteries of 5kWh and larger show overall minor improvements in demand peak load reduction

• Smoothing effect occurs in the energy demand when more households get grouped together

Analyzing the required battery capacity for each grouped household can be done by first looking at the average performance of each battery size. Figure 8 shows the average demand peak load reductions of all scenarios. From the figure, we can see that larger battery sizes lead to larger demand peak load reduction.

The figure also points out that the steps in increasing battery sizes are not equal to the changes in reductions. Between a 1kWh battery to a 3kWh battery, the average difference is 16,63%. Going from a battery capacity of 3kWh to 5kWh, the difference is 5,61%. Increasing to the even larger batteries result in even less increased reductions. When all groups are separated, visualized in Figure 9, differences can be found between groups.

Figure 8 Demand peak load reduction

(18)

An interesting finding is that the group of 15 households is performing between 6,66% and 14,12% better than the other groups when using a 1kWh battery per household, but groups of 3 and 6 households perform up to 16,03% better in reducing peak loads with a 5kWh battery then groups 9 and 15 households.

The reasons for this effect can be found in Figure 10. It shows the differences in the average Ht of the groups of 3 and 15 households in the first two weeks of the calendar year. These first 14 days have a big impact on the battery levels due to fewer daylight hours and higher demand in the winter months (Klingler & Schuhmacher, 2018). The peak loads are twice as big on some days, indicating that the demand for the group of 15 households is much smoother. This smoothing effect is caused by the aggregation of the multiple Ht that are all different from each other Luthander et al. (2016).

This smoothing effect in the group of 15 households is causing two side effects, which are visualized in figures 11 and 12. Firstly, peak loads are smaller over the year, which is beneficial for smaller battery sizes. Smaller battery sizes are vulnerable to these higher peak loads because they reduce over half of the inventory. This was already discussed in 4.1 and can be seen again in Figure 11, in which inventory levels are made equal. The drop in inventory on day 6 is higher for the 3 households group, increasing the chance of failure.

(19)

19 Secondly, when Plimit and PGlimit get stricter, more energy needs to be reduced for a group of 15 households. This is shown in Figure 12. When a 5kWh battery per household is used with Plimit and PGlimit at 60%, it is clear to see that the group with 15 households needs to use much more inventory between days then the group of 3. In these scenarios, the group of 3 households is ending up as a success and the group with 15 households as a non-failure.

Even when batteries get significantly larger, the same effects can be found. Figure 13 shows scenario 100 and 120, in which Plimit and PGlimit are set at 70%. The group of 3 households can successfully reduce demand peak loads with 70%. The group of 15 households only reduces the demand peak load with 2,96%, which was already established with the smaller battery capacities. This information shows that using a significantly larger battery does not influence the discussed smoothing effects.

Figure 11 Battery inventory 1&21

(20)

4.3 Sensitivity analysis

• Bigger groups outperform smaller groups at non-failures.

• Smaller groups outperform bigger groups at demand peak load reduction. • Stricter peak shaving strategies show increasingly higher curtailment levels.

4.3.1 Peak shaving strategies vs. Different groups

Figure 14 provides an overview of all peak shaving results for the batteries 1kWh, 3kWh, and 5kWh. Successful reductions established by the peak shaving strategy are in green, and failures are in red. Peak shaving strategies are much more successful with 3kWh and 5kWh batteries. This again indicates that a 1kWh battery is too small.

(21)

21 A counter-intuitive finding is that larger grouped households have fewer failures than smaller groups, but smaller groups reduce the peak loads more easily when peak shaving strategies get stricter. This is in line with the smoothing effects discussed in 4.2. Even when batteries get significantly larger, the results stay the same. This can be seen in Figure 15.

It clearly shows that smaller groups get more success at reducing peak loads to 60% and even up to 70% at 15kWh batteries per installed kWp.

(22)

4.3.1 Peak shaving strategies vs. Curtailed energy

Figure 16 summarizes Ploss for each peak shaving strategy. In 4.1, it was already shown that different groups do not affect Ploss.

It shows that stricter peak shaving strategies result in increasingly higher curtailment levels. Going from 40% to 50% will increase Ploss with 5,99%. Increasing to 60% will add an additional 8,11-8,29%, and increasing to 70% will add an additional 8,94 -9,74%, which is increasingly more. This means that there is a relation between the used limits in the peak shaving strategies and the amount of curtailed energy.

5. Discussion and conclusion

This last chapter consists of a discussion about the findings and a conclusion. Finally, a few remarks are provided about the research. These limitations address potential improvements to the modelled approach.

5.1 Implications of the findings

This study provides relevant results for future configurations of PV connected households that utilize shared batteries. The scenarios in simulation 1 showed that a 1kWh battery per grouped household is vulnerable for high peak loads. It also showed that this battery could not decrease more than 40% of its demand peak loads in any household group, but curtailment levels where kept below 4%. This means that peak shaving strategies within groups of households aiming to decrease peak loads with 40% or more need extra battery inventory.

Further findings show that a battery size of 3kWh can reduce demand peak loads in almost all group household’s configuration up to 50%. When increasing to a 5kWh battery, reduction in demand peak load reductions is overall minor. This indicated that the extra 2kWh has less impact on the reductions than going from 1kWh to 3kWh. This finding is also found by Braun et al. (2009), which found going from 2,3kWh to 4.6kWh of battery per household, the benefits are less significant.

(23)

23 However, both groups with 3 households showed demand peak load reduction up to 60%

utilizing a 5kWh battery. This is controversial to the study of Braun et al. (2009). Another

surprising finding is that a bigger group of households increases the potential of not failing a peak shaving strategy. Bigger groups where more able to reduce peak loads when limits got increased. This is both caused by the smoothing effect that occurs when household demands are grouped. The smoothing effect was already found in the study of Luthander et al. (2016) but did not address these additional findings.

The findings of the curtailment levels are very similar to the study of Matthiss et al. (2015). The 40%-70% feed-in limit range should result in curtailment percentages around up to 24%. All scenarios came up with curtailment percentages between 2.96% and 28.23%, which are quite similar. Differently to the study of Matthiss et al. (2015) is that curtailment levels did not drop as much when the battery size got increased because less focus was put on self-consumption within the groups of households. One-third of the curtailment could be decreased when considering more self-consumption (Matthiss et al., 2015).

5.2 Conclusion

In this research, battery sizes for different household groups are modelled when using various peak shaving strategies. The groups of households consist of six different groups varying in sizes and combinations. 120 scenarios were run to simulate different configurations. Results are concluding that a 3kWh battery is most fitting for groups up to 15 households when looking to reduce peak loads by 40%-50% with peak shaving strategies that use limitations for the energy flows. This gives curtailment losses between 2.96% and 9.65%. Groups of 3 households can use a 5kWh battery to reduce peak loads to 60%, but this doubles the curtailment losses up to 18,03%. The larger battery sizes of 10kWh and 15kWh showed that increased demand reductions in the peak loads are even less often achieved. At the 15kWh battery, only one of the groups with 3 households could successfully reach a 70% reduction in demand peak loads.

Based on the results, it can also be concluded that a 1kWh battery is not preferable for future household configurations, because of the vulnerability for high peak loads. The study shows in general, that future household configurations should focus on using 3kWh batteries per grouped household to reduce stress on the grid.

5.3 Limitations

Some aspects of a real situation were not modelled in the simulation like inverters and cycle losses. To be able to determine the exact battery storage and peak shaving strategy with more accuracy, future research should be added with inverters and cycle losses. Secondly, the

(24)

References

Agnew, S., & Dargusch, P. 2017. Consumer preferences for household-level battery energy storage. Renewable and Sustainable Energy Reviews, 75: 609–617.

Barzkar, A., & Hosseini, S. M. H. 2018. A novel peak load shaving algorithm via real-time battery scheduling for residential distributed energy storage systems. International Journal of

Energy Research, 42(7): 2400–2416.

Battery Pack Prices Fall As Market Ramps Up With Market Average At $156/kWh In 2019. 2019, December 3. BloombergNEF. https://about.bnef.com/blog/battery-pack-prices-fall-as-market-ramps-up-with-market-average-at-156-kwh-in-2019/.

Beschikbaarheid capaciteit per gebied. 2020, February 20. Liander. https://www.liander.nl/transportschaarste/beschikbaarheid-capaciteit.

Braun, M., Büdenbende, K., Magnor, D., & Jossen, A. 2009. Photovoltaic self-consumption in Germany: using lithium-ion storage to increase self-consumed photovoltaic energy. Frauhofer

IWES, 1–7.

Klingler, A.-L., & Schuhmacher, F. 2018. Residential photovoltaic self-consumption: Identifying representative household groups based on a cluster analysis of hourly smart-meter data. Energy

Efficiency, 11(7): 1689–1701.

Leadbetter, J., & Swan, L. 2012. Battery storage system for residential electricity peak demand shaving. Energy and Buildings, 55: 685–692.

Litjens, G., Worrell, E., & Sark, W. V. 2018. Assessment of forecasting methods on performance of photovoltaic-battery systems. Applied Energy, 221: 358–373.

Luo, X., Wang, J., Dooner, M., & Clarke, J. 2015. Overview of current development in electrical energy storage technologies and the application potential in power system operation. Applied

Energy, 137: 511–536.

Luthander, R., Widén, J., Munkhammar, J., & Lingfors, D. 2016. Self-consumption enhancement and peak shaving of residential photovoltaics using storage and curtailment. Energy, 112: 221– 231.

Matthiss, B., Stellbogen, D., Eberspächer, M., & Binder, J. 2015. Curtailed Energy of PV Systems – Dependency on Grid Loading Limit, Orientation and Local Energy Demand. 30th

European Photovoltaic Solar Energy Conference, 1: 1-4.

(25)

25 Mulder, G., Ridder, F. D., & Six, D. 2010. Electricity storage for grid-connected household dwellings with PV panels. Solar Energy, 84(7): 1284–1293.

PVGIS. 2016, January 11. PVGIS radiation tool. PHOTOVOLTAIC GEOGRAPHICAL INFORMATION SYSTEM. https://re.jrc.ec.europa.eu/pvg_tools/en/tools.html#MR.

Rahimi, A., Zarghami, M., Vaziri, M., & Vadhva, S. 2013. A simple and effective approach for peak load shaving using Battery Storage Systems. 2013 North American Power Symposium

(NAPS).

Shannon, R. 1998. Introduction to the art and science of simulation. 1998 Winter Simulation

Conference. Proceedings. 1: 7-14.

Tan, K. M., Ramachandaramurthy, V. K., Yong, J. Y., Padmanaban, S., & Mihet-Popa, L., et al. 2017. Minimization of Load Variance in Power Grids—Investigation on Optimal Vehicle-to-Grid Scheduling. Energies, 10(11): 1880.

Uddin, M., Romlie, M. F., Abdullah, M. F., Halim, S. A., Bakar, A. H. A., et al. 2018. A review on peak load shaving strategies. Renewable and Sustainable Energy Reviews, 82: 3323–3332. Wang, Z., & Wang, S. 2013. Grid Power Peak Shaving and Valley Filling Using Vehicle-to-Grid Systems. IEEE Transactions on Power Delivery, 28(3): 1822–1829.

(26)

Appendix 1

Household Category Family Situation Annual kW kWp PV

1 Single parent family 1057,0 1,1

3 Childless family 2076,9 2,1

4 Family with children 2153,1 2,2

Low

Mid

(27)

27

Appendix 2

Household Category Family Situation Annual kW kWp PV

3 Low Childless family 2076,9 2,1

8 Mid Family with children 3543,4 3,5

13 High Childless family 5500,8 5,5

4 Low Family with children 2153,1 2,2 7 Mid Family with children 2922,0 2,9 8 Mid Family with children 3543,4 3,5

14 High Family with children 6081,0 6,1 2 Low Single parent family 1901,5 1,9

4 Low Family with children 2153,1 2,2 7 Mid Family with children 2922,0 2,9 8 Mid Family with children 3543,4 3,5

9 Mid Childless family 3598,7 3,6

14 High Family with children 6081,0 6,1 1 Low Single parent family 1057,0 1,1 8 Low Family with children 3543,4 3,5 15 High Family with children 8206,4 8,2 1 Low Single parent family 1057,0 1,1 2 Low Single parent family 1901,5 1,9 8 Mid Family with children 3543,4 3,5

9 Mid Childless family 3598,7 3,6

14 High Family with children 6081,0 6,1 15 High Family with children 8206,4 8,2

(28)

(29)

29

Appendix 4

Time (hour) Hours per _day Solar energy produced (kWh)

Battery inventory

(kWh) Max SOC Min SOC

Household demand (kWh) Solar to energy storage Solar to

demand Max storage for demand Storage to demand Feed-in Power from the grid Curtailment

(30)

(31)

(32)

(33)

33

(34)

(35)