Dittenheim Site

The developed model was applied to a real Ericsson Telecom Site, located in Dittenheim, Germany. It is a 4G LTE site that provides network service for an area of approximately 2 km radius. There are 6 Polycrystalline PV panels of 390 Wp, tilted 32 degrees and facing south with a lifetime of 25 years. It is a grid-connected site which

45 utilizes PV for energy savings. Information about the load profile over the month of April 2021, together with the measured PV generation, shown in Figure 40, were found among Ericsson Internal Documents. The load spans from a minimum and a maximum value of 2204 W to 3499 W with an average of 2499 W, whereas the solar PV output reached a maximum of 2044 W with an average 290 W. It appears clear that the solar contribution does not replace the grid but only helps reducing the amount of energy withdrawn from it.

Figure 40 - Dittenheim 4G LTE site hourly consumption and PV generation profiles for the month of April [43]

To integrate this informations in the tool, a hourly load was built starting from the graph in Figure 40. A 24 hours portion was copied on an excel sheet, trying to follow the same trend. Figure 41 illustrates how close the load profile built is to the actual trend. This was approximated to be the same for every day of the year.

Figure 41 - Hourly load profile comparison between the one found in the Internal Ericsson Documents (yellow) and the one built in Excel (blue)

The Capex of the system was partially documented and the rest was calculated based on current average costs found online or in the literature. For instance, for the lithium ion batteries, it is known that at the moment the

46 lowest cost has reached around 150€/kWh [87], however, it was found that an average price is 250€/kWh [88].

SolarPower sells 100 Ah 48 V lithium battery for 1584 €, which would make 330€/kWh [89]. They also included the information about their lifetime, which will be useful to calculate any cost deriving from the replacement of the battery pack. With a DoD of 80% their battery can perform more than 5000 cycles. The depth-of-discharge chosen in our model is 70%. Relion is another company that sells lithium batteries, it was found in their webiste a graph about the relation between DoD and number of cycles before reaching the end of life, it can be seen in Figure 42.

Figure 42 - Number of cycles depending on the depth-of-discharge for a 100 Ah 48 V lithium battery from Relion [90]

Whereas, in the solar power data sheet, it was reported that for a DoD of 100%, the lifetime is around 2500 cycles and for a DoD of 80%, the lifetime is about 5000 cycles. Therefore, the Relion battery has on average 1,4 number of cycles more than the SolarPower. Based on this, a graph of the two batteries lifetime was built and a linear trend was found, through the function LINEST, in Excel. The resulting number of cycles of the SolarPower for 70% of DoD is equal to 6500.

Figure 43 - Number of cycles as a function of the DoD for a lithium ion battery of 100 Ah and 48 V according to Relion (in blue) and SolarPower24 (in red)

A cycle of the batteries is calculated in the tool as a full charge and discharge of the 70% of capacity, following the equation:

𝑛 𝑜𝑓 𝑐𝑦𝑐𝑙𝑒𝑠 𝑖𝑛 𝑎 𝑦𝑒𝑎𝑟 =𝑒𝑛𝑒𝑟𝑔𝑦 𝑑𝑖𝑠𝑐ℎ𝑎𝑟𝑔𝑒𝑑 𝑖𝑛 𝑎 𝑦𝑒𝑎𝑟 70% ∗ 𝑇𝑜𝑡𝑎𝑙 𝐶𝑎𝑝𝑎𝑐𝑖𝑡𝑦

0 2000 4000 6000 8000 10000 12000 14000

40 50 60 70 80 90 100 110

number of cycles vs depth of discharge [%]

47 in practical terms, the amount of energy discharged by the battery is calculated for the whole year and divided by 70% of the total capacity, this gives the number of cycles performed in one year. As well as the number of cycles, another parameter to keep into account for the lifetime of batteries is the number of years that the supplier can guarantee for the battery to last. Beltran et al [91] retain that the average life expectancy of current lithium ion batteries are between 8 and 12 years. 10 years will be considered in the model. The actual lifetime will be given by the shortest timespan between the number of years and the 6500 number of cycles.

The Opex was not documented, therefore, an approximation of 1% of Capex as a yearly maintenance cost was adopted, as assumed by Perez [92].

Moreover, the PV output is always lower than the load, therefore, no batteries would be considered at this stage if the assumption of a constant electricity price of 0,18 EUR/kWh and no need to perform any demand side management to reduce the costs was made. However, an online research about the hourly German electricity prices was performed, concluding that they are one of the highest in Europe and in the World according to [93] [94] [95], accounting for an average value of 0,30 EUR/kWh for the household sector and 0,18 EUR/kWh for the non-household sector. On EPEXSPOT the day-ahead wholesale prices for the geographical area of Germany-Luxemburg was found [96]. This information was combined together to the average electricity prices found above to deduce a rough 24 hours profile, reported in Figure 44, which follows the same trend proposed by EPEXSPOT with an average value of 0,18 EUR/kWh. It can be noticed how the electricity is more expensive during classic peak demand times, around 8 am and 8 pm and becomes cheaper during the night. As a result, demand side management would make sense in this case, charging the batteries when the grid prices are down and discharging it during high prices and high demand hours, for peak shaving.

Figure 44 - Germany hourly Electricity price for the non-household sector

Another benefit of integrating a renewable energy system in a fully reliable grid-connected Telecom Site, where there are no power outages issue, is the reduction of carbon emission related to the grid itself. It is well-known that after the Fukushima accident in 2011 Germany policymakers decided to phase out nuclear energy, leaving a significant gap in the electricity mix, that has been replaced by wind, solar but mostly coal production, which is the most polluting type of generation. According to [97], the emission factor of the German grid accounted for 0,485 tonCO2/MWh of electricity consumed. To find the exact value of emission factor of the German grid a simple calculation was performed, starting from the electricity generation mix by source found in the International Energy Agency webpage, for the country Germany [98]. German electricity is produced by a large variety of sources, such as fossil fuel, mainly coal and gas, renewables and a decreasing nuclear. Each source has its own emission factor, that can be found online, like for coal [99]. An average of all the emission factors of the sources, weighted on the yearly contribution of electricity production was made, concluding that, for every MWh of electricity consumed from the grid in Germany, 0,412 ton of CO2 was produced in 2019, which is close to the 0,485 estimated by the German Environment Agency [100]. The calculation is shown in Table 12.

0.00

Non household Electricity prices in Germany [€/kWh]

48 Table 12 - Calculation of the emission factor of the German grid, based on the electricity generation mix

German grid 2019 https://www.eia.gov/tools/faqs/faq.php?id=74&t=11

Source GWh % of mix Emission factor [ton/MWh] Mton CO2

Coal 185495 30,1% 1,00433 186,30

NG 94461 15,3% 0,41236 38,95

Oil 5083 0,8% 0,96956 4,93

Nuclear 75071 12,2% 0

Biofuels 44629 7,2% 0

Waste 12571 2,0% 0

Hydro 26201 4,2% 0

Solar PV 47517 7,7% 0

Wind 125975 20,4% 0

total

generation 617003 0,3731 ^tonCO2/Mwh generated

electricity

consumed 558900 0,4118 ^tonCO2/MWh consumed

Several scenarios will be examined with an increasing numbers of PV panels and Lithium-ion batteries, investigating the techno-economic performance of each of them.

When the number of PV panels are increased, there will be some part of the year in which the solar production exceeds the load, the logic of the battery will change and adjust to the new situation where we will give the priority to store the energy from the PV array, rather than storing the low cost energy from the grid during the night. A summary of all the possible cases and the different behaviour of the system is illustrated in the flowchart in Figure 45. The first check that is done is whether the peak PV production in the month can ever be higher than the minimum daily load. This control is actually applied on the radiation value rather than the production itself.

Depending on the number of PV panels and the array’s maximum output after the MPPTs it is roughly estimated what radiation value could acheive an output that overcome the load and in which month that radiation value is found, at least once. This method requires the knowledge of the radiation data of the whole month “a priori”, which is available from the Typical Meteorological Year data in the software. Other studies have investigated this

“a priori” logic to determine expected solar production from weather data [73]. In this way, the combination of pv, grid and batteries is utilized in a smart way, so to avoid discarding high amount of energy output from the PV, during the summer, that may derive from oversizing the system in order to meet the demand in the worst moment of the year, the winter.

49 Figure 45 - Flowchart of the priority of resources in the different electricity price and solar production cases in the Dittenheim site 3.5.2 Mexico site

In the north-west of Mexico City an offgrid Telecom Site is fully reliant on a 10 KVA Diesel generator and an analysis of the decarbonisation opportunity of it are discussed in this report. For this site, no detailed information about the hourly load profile were found, it is only known than the average load is 4 kW. The model will be applied to this site and three different configurations will be studied. The base case with the genset only, a case with Diesel Generator and batteries and a hybrid configuration of PV panels, battery storage and genset. In the first case, the diesel generator will have to follow the load every hour, running at a lower load ratio, which does not give the best fuel performance. In the second case, a storage is used in combination with the diesel generator so as to run the genset at a constant output, coinciding with its maximum fuel efficiency point, to cover the load and charge the battery, once the battery reaches its limits, the genset is switched off and the load is covered by the bess. This process repeats in a cycle throughout all years.

4 Results

4.1 Spectral Effect

Based on qualitative information and quantitative data found in the literature, a hourly relation between the spectral factor and AM for CIGS and high-eff c-Si was developed. The AM is calculated from the solar angle, which is already present in the Model, then, the relative spectral factor is included in the total harvest equation as a percentage variation of the energy coming from the global horizontal radiation. The equation were validated by comparing the average monthly values of the spectral factor obtained with the one in the literature. A linear trend was found to well fit the behaviour of high-eff c-Si whereas a polynomial trend was chosen for the CIGS.

50

For the CIGS, the spectral effect stays very low at the beginning, then it grows quickly, reaching its maximum value of 8 % when the AM is 10. On the other hand, for the hif-eff c-Si, the spectral effect grows linearly, reaching the maximum value of 10 when the AM is 20. This was chosen because, in Figure 34, it can be noticed that the points when the highest value of MM are reached for high-eff c-Si, are all concentrated at the lowest irradiance level and do not tend to happen again for higher radiation power. On the other hand, for the CIGS case, high values of MM are present at several level of irradiance and happen more often. An AM of 10 is more likely to happen during the whole year than 20, as can be deduced from Table 8.

Despite some imperfections in the case of the CIGS, it could be of interest to implement the method discussed in the methodology chapter, Error! Reference source not found.in the Excel Tool and examine how it affects the results in terms of backup needed. The same seven cities of the angular effect study were taken under examination and for each of them, a comparison between the performance of the same energy system with and without the consideration of the spectral impact was made, both for the case of high-eff c-Si and CIGS. In

Figure 46 is shown the result for the former technology, the blue graph represents the variation in total annual production, a positive variation reflects an improvement. The red graph is the variation in hours of backup needed, therefore a negative variation means that the system is more self-sufficient. As expected, locations at low latitude shown a loss of yearly production and self-sufficiency, due to the high solar angle throughout the whole year, which lead to a blue-shifted spectral distribution. Whereas, in mid-high latitudes the production becomes slightly bigger and there are less hours of backup needed, because the AM in these locations has higher values and the shorter wavelengths are filtered, especially in the mornings and in the evenings. However, the overall

Variation in total yearly production [kWh] and hours of backup needed [h] vs latitude for high-eff c-Si

Yearly production variation

backup variation

51 needed was witnessed for the city of Frankfurt at 50 degrees North, with very similar weather conditions as Freiburg, and it accounted for 172 hours less than the 2900 hours needed in the case without considering the spectral impact, an yearly improvement of 6 %, followed by Milan, registering 109 hours less of backup needed.

Improvements which are not repeated in Nordic cities like Copenhagen and Stockholm, which show no more than 22 hours less than the 3500 hours of the base case, an yearly improvement less than 0,6% and that, according to the logic, should have shown the largest effect.

Figure 46 - Variation in the total annual production of solar energy and hours of backup needed when implementing the spectral impact for 7 locations at different latitude in the case of high-eff c-Si

A similar analysis was performed for the CIGS technology. The relation between the spectral impact and the AM illustrated in Figure 35 was added in the Excel sheet, multiplying those values times the incoming radiation. For the same seven cities, the output in terms of annual PV production and hour of backup needed were compared

and plotted in

Variation in total yearly production [kWh] and hours of backup needed [h] vs latitude for high-eff c-Si

Yearly production variation

backup variation

52 Figure 47. In this case, the city of New Delhi presents the best gains in terms of backup requirements, increasing the system self-sufficiency by 201 hours compared to the 2600 needed in the reference case, an yearly improvement of 7.7 %, followed by Milan (131 h and 7.5 %) and Frankfurt (162 h and 5.6 %).

Figure 47 - Variation in the total annual production of solar energy and hours of backup needed when implementing the spectral impact for 7 locations at different latitude in the case of CIGS technology

4.2 Optimal Tilt Angle

In the Excel Tool the optimal tilt angles, both to maximize PV production and self-sufficiency, were searched for several locations and different cases. Moreover, the number of panels and batteries were varied to see if that affected the output. The results are shown in Table 13. For each location, the latitude, the best tilt angle in terms of production and self-sufficiency are listed in the first three columns highlighted in red. The column “comments”

explains what variation was made to the system, such as an increase in number of PV panels, an addition in storage capacity or a variation in the load. Then, during this process, it was noticed that these variations were not affecting the tilt angle for the PV production optimization, but only the one related to the maximum self-sufficiency.

Variation in total yearly production [kWh] and hours of backup needed vs latitude for CIGS

Variation in total yearly production [kWh] and hours of backup needed vs latitude for CIGS

Variation in production

backup variation

53 Furthermore, it became clear that an addition in number of panels and/or batteries can be well described by one parameter, the hours of backup needed. As the PV array power rating and the storage capacity rise, the hours of backup needed decreases. Therefore, the hours of backup needed was considered when searching for the best tilt angles for the following cases.

Table 13 - Optimal Tilg Angles to maximize the PV production and to maximize the self-sufficiency over a year, for several lcoations and different system set-ups

It was found that many cases present an optimal tilt angle that maximizes the PV production whose value is similar to the latitude angle. Whereas, the optimal tilt angle that maximizes the self-sufficiency often increases with the number of panels and batteries that are available, or the self-sufficiency itself, reaching values equal to 20 degrees more than the latitude angle. In the columns “latitude”, “max PV production” and “max self-sufficiency”, the cases that followed this logic are highlighted in yellow. All the cities whose results differ from the main trend are located below the 30^th parallel, in the tropical region. In those cases, the optimal tilt angle that maximizes the production is from 10 to 20 degrees more than the latitude angle, while the one that maximizes the self-sufficiency does not follow any consistent trend. The results are shown graphically in Figure 48. Two linear trends were drawn through the Excel function “LINEST”, to find an equation that could describes them.

54 Figure 48 - Optimal Tilg Angles to maximize the PV production and to maximize the self-sufficiency over a year, at different

latitudes

From the results above, it was decided to include in the Excel tool a suggestion for the tilt angle with a margin of error of about 10 degrees, which the customer may or may not choose to observe, as follow:

If the system is grid-connected, meaning that the power in excess can be sold:

𝑂𝑝𝑡𝑖𝑚𝑎𝑙 𝑇𝑖𝑙𝑡 𝐴𝑛𝑔𝑙𝑒 = 0,39 ∗ 𝐿𝑎𝑡𝑖𝑡𝑢𝑑𝑒 + 31° ± 5°

If the system is off-grid, the excess power is lost and the self-sufficiency is the main objective:

𝑂𝑝𝑡𝑖𝑚𝑎𝑙 𝑇𝑖𝑙𝑡 𝐴𝑛𝑔𝑙𝑒 = 0,76 ∗ 𝐿𝑎𝑡𝑖𝑡𝑢𝑑𝑒 + 24° ± 10°

4.3 Dittenheim site

The results were taken from the model for the case study in Dittenheim, Germany and the KPIs analysed are presented in this chapter. Several cases were taken under examination and are illustrated in more details in the Appendix.

Overall, the KPIs for 4 main scenarios will be analysed. The scenarios are:

 1: 6 PV panels and no batteries

 2: 12 PV panels and 3 batteries

 3: 18 PV panels and 3 batteries

 4: 24 PV panels and 4 batteries

In Figure 49 the fraction of solar energy of the total energy consumed by the site in the four different scenarios is shown. The higher the number of PV panels, the larger is the share of the renewable source.

0

55 Figure 49 - Fraction of Solar Energy in the total energy consumption for 4 scenarios

Then, Figure 50 illustrates the Discounted Cumulative cash flow value achieved in the 25th year of operation, in

Then, Figure 50 illustrates the Discounted Cumulative cash flow value achieved in the 25th year of operation, in