Post on 31-May-2020
NRG ENERGY CASE STUDY
HEC Paris J17 Team Pascal Lim
Sanjay Munka Albert Priori
MARCH 8, 2017
HEC PARIS
FRANCE
Table of Contents
1. Introduction ......................................................................................................................... 2
2. Datacenter Requirements and Constraints .......................................................................... 2
3. Clean Technology mix selection .......................................................................................... 3
3.1. Fuel Cells CHP ................................................................................................................ 4
3.2. Solar PV and batteries ..................................................................................................... 6
3.3. Resiliency and reliability .................................................................................................10
4. Financing structure ............................................................................................................12
5. Powering the data center: final design and financial model ................................................14
5.1. Decision-making process ...............................................................................................15
5.2. Financial results .............................................................................................................16
6. Conclusion .........................................................................................................................19
1. Introduction
NRG is the leading integrated power company in the US with over 51 GW of generation
capacity1. In 2015, NRG’s power generation in the North American market stood at 123 TWh,
with US CO2e emissions1 at 87 million metric tonnes. NRG is a pioneer in the drive towards
cleaner energy with renewable sources representing about 9% of their total generation portfolio1.
It has aggressive sustainability goals of reducing its carbon emissions 50% by 2030 and 90% by
20501.
To achieve its ambitious emissions goals, NRG is increasingly focusing on cleaner energy.
However, unlike established technologies, project financing for renewable or distributed energy
systems remains a significant obstacle. This proposal helps overcome this obstacle by adopting a
creative financial model that enables the development of a resilient clean energy distributed
generation system for a 30 MW data center in Texas, helping NRG meet its sustainability
goals at the same time.
2. Datacenter Requirements and Constraints
Data centers require energy throughout the day, without specific requirements for a particular
time of the day; making their energy demand constant.
Critical load and HVAC load are the two most important figures crucial in designing the
electrical system that will power the data center. Studies2 have shown that typically data centers
have 52% of their overall loads represented by critical load, comprising mainly of server racks &
routers and 30% represented by HVAC load. HVAC extracts heat generated by all equipment,
lighting, personnel and weather. Further, 6% of the overall load represents UPS’s operating at
88% efficiency and 12% represents a combination of lighting and protection equipment loads
such as fire, security & monitoring systems. Fig 2.1 illustrates loading requirements for a data
center.
Figure 2.1: Typical electrical sizing for a data center
3. Clean Technology mix selection
We evaluated several clean technologies for this project, summarized below:
● Gas Turbines: Typically the most reliable and robust solution for this application.
However, inefficiency at part load and high emissions made it a far from ideal solution
for NRGs mid-term and long-term goals.
● Wind Turbines: Good and cost effective renewable solution. However, non-homogeneous
wind patterns in Texas cause high energy production volatility. For this reason, to ensure
resilience, a 100% backup would be necessary, increasing CAPEX significantly.
● Biomass: High CO2 emissions. Furthermore, the suitability of such technology is
strongly related to a short fuel supply chain that does not seem possible in Texas based on
a preliminary analysis.
● Battery Storage: Proven and reliable technology that facilitates the effective utilization of
intermittent renewable sources and reduces the need for peak generation capacity.
Technology is still expensive but costs are reducing1 year on year.
● Fuel Cells: High efficiency, can run continuously and are a good complement to
renewables, since they are able to provide base load power. Certain types are well suited
to CHP applications.
● Solar PV: Clean and sustainable technology requiring low maintenance and increasingly
lower investment costs. However, performances are weather dependent.
To increase reliability, resilience and reduce greenhouse gas emissions, a combination of fuel
cells in CHP configuration, PV and battery storage was chosen for the project. The Installed
capacity of each technology has been set to maximize project IRR and minimize carbon
emissions.
3.1. Fuel Cells CHP
Fuel cells can provide high efficiency and low emissions energy over a broad load profile.
Among the different available technologies, the Molten Carbonate Fuel Cells (MCFC) was
chosen for its fuel flexibility, efficiency and high operating temperatures that makes it ideal for a
CHP application.
1https://www.bloomberg.com/news/articles/2017-02-21/big-batteries-coming-of-age-prompt-bankers-to-
place-their-bets
Figure 3.1.1: MCFC operating process2
A large portion of the power consumed by a datacenter is essentially converted to heat, for this
reason its cooling requirements are high. Fuel Cells in combination with an absorption chiller
driven by exhaust heat will also provide cooling to the datacenter, reducing client’s air
electricity requirements and consequently costs.
The CHP configuration increases overall energy efficiency from 42.5% (electric generation only)
to 74.8% (thermal and electric generation altogether). The following simplified flowchart shows
the CHP working principle and main figures in terms of electricity and cooling production for
100 units of energy (natural gas) entering the fuel cell.
Figure 3.1.2: Fuel Cell CHP General Flow Chart
2 http://www.fuelcelltoday.com/technologies/mcfc
Natural Gas has been chosen instead of hydrogen, since it’s a cheaper and more available fuel:
for this reason, the system produces carbon dioxide. However, thanks to the CHP configuration,
CO2 releases are calculated with a thermal credit, since supplementary emissions to generate the
useful thermal power by an on-site boiler are avoided.
Taking into account thermal credits, CO2 emissions can be lowered from 939 to 559 lbs/MWh.
The performance, costs and emissions for MCFC that will be used in our financial model can be
found in the table 3.1.1.
MCFC FUEL CELLS
Electric Efficiency %, HHV 42.5%
Thermal Efficiency (%, HHV) %, HHV 32.3%
Overall Efficiency (%, HHV) %, HHV 74.8%
Installed Cost $/kW $4,600
O&M Cost, excluding fuel ¢/kWh 4.0
CO2 Emissions Electric Power Only lbs/MWh 939
CO2 Emissions in CHP with thermal credit lbs/MWh 559
Table 3.1.1: MCFC Key Figures3
3.2. Solar PV and batteries
The design of the photovoltaic and battery system has been optimized to match the daily and
yearly production profile with the Data center load profile. To smoothen the daily PV production
profile, a single-axis tracking system has been preferred to a static one. Furthermore, to reduce
the seasonal production gap, the tilt of the panels has been increased to 55°.
3 https://energy.gov/sites/prod/files/2016/09/f33/CHP-Fuel%20Cell.pdf
Figure 3.2.1: Production profile of one PV MWp installed
With tracking systems, the panels power density (kWp per square meter) can be optimized,
hence more efficient polycrystalline silicon technology have been chosen over thin film.
Conversion rate (DC Electricity over Solar Ration kW) for this type of technology is around
18%.
All the data related to photovoltaic power production are supported by a PVsyst study (appendix
3.1). The photovoltaic & battery storage association is designed as a complement of the fuel
cells power production. Therefore, the power output profile of the association should also be a
constant over time.
According to our simulations, a 1 MWP PV system in Texas will produce on average 2,89 MWh
per day during the worst month (November), as illustrated by figure 3.2.2.
Figure 3.2.2: November Load Profile for PV and batteries
The batteries are sized to store the energy that cannot be immediately consumed by the
datacenter (grey line). This energy is then restituted by the battery during night time and
inadequate weather conditions (orange line). In order to absorb the totality of PV excess
production of one particular day, the battery storage needs to be sized to 66%4 of the total energy
produced that day. Therefore, because of our choice to dimension the system for an average day
of November, 1.93 MWh of battery storage (66% of 2.89 MWh) is required for a 1 MWp. This
association will ensure constant power for a 0,125 MW production.
However, this combination does not guarantee perfect autonomy from the network and energy
will still be sold and bought on the grid, as illustrated by the following table.
4 This figure has been computing from the following inputs: battery storage losses, inverter losses
avoided and PV average production profile in November.
Table 3.2.1: Power balance of PV & battery system for 1 MWp PV
1 kWp of Solar PV
Daily average production in November kWh 2.89
Installed Cost $/kW 1400
O&M Cost, excluding fuel ¢/kWh 0.5
CO2 Emissions Electric Power Only lbs/MWh 0
Table 3.2.2: Performance, costs and emissions for PV
The Battery costs, life-time and efficiency are dependent on the technology selected5. The final
solution has been selected in order to optimize the financial results.
Batteries Lead Acid Li-ion
Cycle Efficiency % 80% 92%
Life time years 8.2 4.1
Installed Capacity Cost $/kWh 378 350
Cost of storage $/kWh 0.32 0.32
CO2 Emissions Electric Power Only lbs/MWh 0 0
Table 3.2.23 Performance, costs and emissions for batteries
5 https://www.solarchoice.net.au/blog/levelised-cost-of-storage-compare-battery-value
3.3. Resiliency and reliability
It is crucial for data centers to have a reliable energy source and be protected from network
disruptions. Energy resiliency and reliability in data centers are usually achieved by traditional
UPS configuration to ensure critical load feeding.
Fig 3.3.1: typical UPS configuration in data center
However, our generation mix produces DC electricity. We can improve the traditional
configuration by an alternative configuration illustrated in Figure 3.3.2. Individual fuel cells and
PV panels supply DC power to the critical load independently, thereby mitigating the risk of
overall power failure.
The loss of any one power generation unit will always be < 1.4 MW (the size of a single fuel
cell). This loss will be immediately compensated by a combination of increased batteries output
and eventual transient load shedding of non-critical AC users. The impact on event for the Data
center will be a temporary partial change from Inverter to Grid supply for non-critical loads
(HVAC and other).
Figure 3.3.2: Alternative configuration in data center with DC power
Bloom energy estimates the availability of similar solutions to be 99.998%6.
The solution is not fully independent from the grid since external power support will be needed
30% of the days7 to meet the data center demand. We considered that protection against long
term grid outages (>2 hours), given its additional costs, should be conditioned by the client risk
aversion profile.
An event similar to the NorthEast BlackOut of 1965 would have less than 30% chances to result
in a power failure for the data center (depending on time and day of the event). However, if the
client decides to implement such protections, he will only need to connect his own batteries to
the DC line.
6
http://c0688662.cdn.cloudfiles.rackspacecloud.com/downloads_pdf_Bloomenergy_Mission_Critical.pdf 7 This figure has been computed on the basis of PVsyst daily production results for the year 1990 with the
weather data of Austin.
4. Financing structure
Financing structures for renewable energy projects range from the regular bank loan and parent
company equity scheme, to more industry-specific structures, taking advantage of the energy
Investment Tax Credit (ITC) offered by the US government for projects that involve
photovoltaic solar panels. This ITC is a tax credit to the power company of up to 30% of the total
capital required for building the solar panels. However, in most cases including ours, the project
company does not have enough tax liability to fully benefit from the tax credit.
Table 4.1: ITC rate per technology and installation year
Source: https://energy.gov/savings/business-energy-investment-tax-credit-itc
In order to fully take advantage of the ITC though, the project company (called “developer”) can
partner with another company (called the “Tax equity investor”) with enough tax liability to
benefit from this tax credit. The required capital to be deployed for funding the project is for the
most part largely provided by the tax equity investor (for example 80%), the developer providing
the remaining amount (20%). However, in order for this partnership to be even more attractive
for the tax equity investor, the cash flows originating from the project will be distributed in a
90%/10% split, skewed towards the tax equity investor who initially receives 90% of the cash
distribution until its rate of return reaches the pre-agreed hurdle rate (15% including ITC) while
the developer receives the remaining cash flow. Once the hurdle rate is achieved, the distribution
flips and 90% of the cash distribution now flows to the developer, who can decide to buy the tax
investor’s equity out. Using this structure, the capital provided by the developer is small but the
rate of return is rather large, making this structure very attractive for both parties.
The tax equity investor can then claim almost all of the ITC credited for the project in order to
reduce its overall tax liability as a company.
The financing structure can be further improved with loans for funding the initial required
capital. Such debt will be repayed through the cash flows generated by electricity sales. The
interest rates for such projects are not projected to be very high (around 5 to 7%) because the
project provides assets as collateral and the demand carries a very low level of risk, as prices will
be locked in through a Price Purchase Agreement (PPA) and as global demand for internet data
will be on the rise for the foreseeable future, ensuring the long term sustainability of the data
center. It is true that tax equity investors generally would rather steer away from project debt
because senior debt takes precedence over equity in a liquidation scenario. However, the project
relatively low level of uncertainty and the projected returns are the best incentive in this scenario.
The final result is displayed in figure 4.1, illustrating the interaction between the different actors
in the financial structure:
Figure 4.1: Finance and cash distribution structure
Risk can even further be reduced by locking in natural gas prices through the purchases of call
options with the Henry Hub natural gas as underlying. The purchase of such derivative products
will be reflected in the financial model.
5. Powering the data center: final design and financial model
In order to produce the financial model, some key assumptions need to be made (see appendix
5.1). The model created can however accommodate for a variety of scenario in order to select
the best parameters.
Using the assumed values, the full financial reporting statements were computed and projected
for the next 20 years: income statement, balance sheet and cash flow statement. The cash flows
from the cash flow statement were used in order to determine the rates of return that each
investor gains.
5.1. Decision-making process
The model was built to determine the financial impact of the technical design choices as
described in the first part of the document. The financial outputs are the ultimate drivers behind
the technology mix and selection.
The first question to be answered is the technology mix that will be required to provide enough
energy for the 30 MW data center. In order to attract tax equity investors, the ITC amount needs
to be significant enough and consequently, the CAPEX used for energy from photovoltaic has to
be significant too. Figure 5.1.1 summarizes the energy provided by the solar and batteries
combination based on the remaining energy requirement complementing the fuel cells. Because
the heat from fuel cells is used to produce cooling, the final demand from the data center
decreases.
Figure 5.1: ITC, PV and batteries requirements as a function of fuel cell output for data center
As shown by Figure 5.1, the ITC amount drastically reduces with increasing fuel cell output. In
order to attract the tax equity investor, it was decided that $10 million would be the minimum
amount of ITC achieved (equivalent to a $33.3 million CAPEX spend for solar PV). The $10
million mark for ITC corresponds to a fuel cell energy output of around 20-21 MW.
5.2. Financial results
A table showing different pricing and debt/equity scenarios can be found in appendix 5.2. The
table shows that because of the large CAPEX required for batteries, the project return is largest
when the proportion of fuel cells output is maximized. Therefore the optimal proportion knowing
the constraints ($10 million ITC and maximized return) is for a fuel cells output of around 21
MW. In this case, the PV solar equipment is designed to provide 21.7 MW at peak in order to
charge batteries with a capacity of 43.4 MWh, enough to continuously provide 2.7 MW, the
remaining demand not met by the fuel cells output.
The PPA price is then determined by the lowest number possible that generates an attractive
return. Indeed, the pricing structure should still reflect and capture the customer’s willingness to
pay.
Figure 5.2: Project internal rate of return compared to PPA price for different debt and equity
proportions
The minimum PPA price in order to reach the tax equity hurdle rate (15% including ITC)
given by the model is $0.115/kWh and corresponds to a debt and equity proportion of
respectively 80% and 20%. The tax equity investor reaches this rate for its invested capital of
$24.2 million as soon as the first quarter of the fifth year, at which point the distribution flips.
NRG’s internal rate of return based on this flip becomes 32%, for an initial invested capital
of $6.05 million.
These numbers can be compared with the project IRR of 13.9% (assuming all the equity was
provided by one investor only).
Figure 5.3: Representation of cashflows with time
Figure 5.3 illustrates the distribution of all the cash flows with time per quarter. The peaks
correspond to the summer time when solar PV is able to gather more sunlight.
Even though the required PPA price is high compared to the average price of electricity paid in
Texas, the customer has resiliency thanks to a dedicated distributed energy system that also emits
significantly less carbon than the national average.
6. Conclusion
The proposed project provides a reliable and resilient energy source to the datacenter. Value has
also been added through the optimization of the power originating from fuel cells, through the
use of cooling in order to reduce the data center demand, while solar and batteries provide power
with no carbon emissions. A creative financing structure was also set up in order to determine
what proportion of each technology was to be used in order to generate maximum returns for all
the different actors, with the full use of the ITC, leading to the following mix:
17% of the energy is produced by PV and 83% is produced by fuel cells. The carbon footprint of
the project can be calculated:
559 x 83% + 0 x 17% = 465 lbs/MWh
The project features a 70% reduction of CO2 emissions compared to NRG’s current portfolio,
with an average 1559 lbs/MWh.
These environmental performances will be sufficient to drive NRG’s effort to reduce by 50%
current emissions. Future significant price reductions are also expected in fuel cells and battery
storage technology; with an increase in hydrogen penetration in the energy market, allowing
NRG to meet its longer term emissions goals.
APPENDIX 3.1: PVSyst study
APPENDIX 5.1: Main assumptions made in the financial model
Table: Debt information for the proposed technical solution
Debt arrangement fees 2%
Construction financing interest rate 7%
Interest rate on cash 0.50%
Table: CAPEX and OPEX information for the proposed technical solution
Batteries CAPEX costs were calculated based on required capacity.
Table: Other financial assumptions made for the model
APPENDIX 5.2: IRR output tables for different debt/equity and PPA price scenario