About
This web page is an interactive model of the New England
electricity grid. It projects power supply and demand
from 2020 through 2040 and lets you adjust plans to
build or retire various sources of electricity, to see
the cost and carbon impacts. The goal is to help people
concerned about climate change explore and communicate
our options.
The model is designed to be simple enough to play with, but
detailed enough to draw plausible conclusions. The top part
of the model display can show projected electricity demand
and supply for the New England grid (ISO-NE) an overview of
20 years with monthly averages, or 1 week with hourly values
(either the first week with missed demand, or the week with
highest demand if demand is satisfied), or 20 years of CO2
emissions along with a cost summary.
Below the output display, there are 8 tabs for the 8
modeled electricity generation sources: solar, wind,
hydroelectric, nuclear, natural gas, coal, batteries,
and hydrogen. Click on a tab to adjust the controls for
that source. The plan sliders are controls for
future decisions about changing the capacity of the
source. Below the plan sliders are
the assumptions sliders, which reflect my
understanding of the characteristics of the
source. These parameters determine the intitial
conditions of the model as well as the costs,
capabilties, time constraints, and CO2 impacts of the
source. You can adjust these assumptions to account for
better information, to model a different scenario than
the default, or just to explore. When assumption sliders
have been changed from the defaults, they are
highlighted so you know that the outcome might not be
comparable to plans under the default assumptions.
As you change the controls, the webpage URL captures
the slider positions. If you save or share the URL,
re-opening the URL will restore your settings and show
the resulting outcome. If you want to return to the
defaults, you can use the Default button at the bottom
of the page, or double-click on the individual sliders.
Caveats
- Numbers and code haven't been fully reviewed and
validated, expect revisions.
- This tool does not optimize, it just shows you the
results of the plan that you give it. Planning &
optimizing is up to you.
- There's no attempt to model locational value or
import/export.
- LCOE and CO2 are for the period 2020-2040. The
initial fleet is inherited at no cost, and there's no
value ascribed to the fleet at the end of 2039. Watch
for boundary effects.
- No spinning reserve or capacity reserve requirements
are modelled. Real grid operators would build more
capacity than what this model requires and so costs
would be a bit higher.
- The gas model is based on simple cycle turbines,
which are cheap and ramp fast but burn more gas per kWh
than combined cycle.
- Initial conditions are based on the ISO-NE grid mix but
the model does not capture in detail current construction
in progress as of 2020, or the age of the existing fleet
in 2020. This might penalize solar & wind scenarios
somewhat.
Details
The demand profile
is
observed hourly data from 6 years (2013-2018) of the New
England electricity grid (ISO-NE), along with
the
hourly solar and wind production for the same hours. The
profiles were preprocessed to normalize annual demand and
solar/wind capacity (see inputdata/prepare.py), and then the
initial grid mixes are set to approximate the
ISO-NE grid
mix for the first 11 months of 2019.
I chose ISO-NE partly because it is a challenging case
for wind and solar. If renewable energy can decarbonize
ISO-NE, it can work in many other places.
To get 20 years of data, the six year profiles are
repeated 3.33 times. Demand is scaled by the
(adjustable) demand growth rate, to model economic
growth and coming changes like EV charging, heat pumps
for heating, and electrification of industry.
CO2 outcomes reported by the tool
The CO2 budget for 2020-2040 is chosen as 2020 emissions
times 7.5. This is represents ramping emissions from
present rate down to zero over 15 years. I believe this
is roughly compatible with 1.5C warming scenarios
(assuming similar reductions across the world and across
other sectors). Please let me know if you have a better
rationale & target. [TODO footnote]
The CO2 reported is total lifecycle CO2 emissions from
electricity generation in ISO-NE over the 20-year
timespan of the simulation. It also includes (smaller)
effects of direct waste heat and water vapor
emissions. [MZJ 2019] gives a range of
values for various sources, broken out by lifecycle,
thermal and water vapor, and other categories.
Lifecycle analysis for a source includes the CO2
impacts of the processes used to construct the source,
and is notoriously tricky to pin down because there are
so many variables. One of the most important variables
is the CO2 intensity of the electricity grid used by the
factories making the materials in a source. For
fossil-fuel sources this is less tricky because the
direct fuel emissions dominate. But for solar, wind and
nuclear, the range of estimates is large. For the
purpose of this model, which reports system CO2 across
the whole grid, my solution for the solar, wind and
nuclear values is to use the low end of reported
lifecycle CO2 intensity, plus whatever (small) direct
emissions are involved. The justification is that the
CO2 from low-carbon sources does not significantly add
to system CO2 until the system becomes relatively
decarbonized. But as that happens, the CO2 impact of
industrial production should similarly decline. This
does not correctly handle effects of plant build and
lifetimes but it's simple and I think it leads to
reasonable results.
Values by source, in gCO2e/kWh:
- solar: 8
- wind: 5
- hydro: 31
- nuclear: 13
- gas: 447 (adjusted in validation process below)
- coal: 927
- battery: 0
- h2: 4 (direct emissions from turbines)
Cost outcomes reported by the tool
The costs are derived by computing the timing and size
of expenditures for plant construction, operation, and
fuel, discounted by the specified discount rate
according to the year in which they occur. To compute
the
Levelized Cost of Electricity
(LCOE), the discounted expenditure is divided by the
discounted energy delivered to the grid. This does not
include the construction cost for the initial fleet of
sources, and does not credit the value and costs of the
fleet or energy generated after 2040.
Because the tool is computing LCOE of generation, it
does not give retail rates or do any market
analysis. Retail rates are usually much higher and have
to cover the cost of operating and maintaining the grid,
among other things. However, cost of generation is a key
driver of prices charged to customers.
Energy generation cost assumptions
Power Sources
For all power sources, we model costs arising from the following sources:
-
Capital: Cost to build the plant ($ per Watt of capacity)
-
Fixed: Cost to operate and maintain the plant ($ per Kilowatt of capacity per year)
-
Variable: Cost of fuel, operations and maintenance to
generate energy ($ per Megawatt hour).
The default capital, fixed and variable cost values for the sources are:
- Solar: $1.0/W, $10.5/KW/y, $0/MWh, 30y life.
- Wind: $1.3/W, $32/KW/y, $0/MWh, 20y life.
- Hydro: $2.95/W, $40.85/KW/y, $1.36/MWh, 50y life.
- Nuclear: $9.55/W, $120/KW/y, $12.0/MWh, 40y life.
- Gas: $0.825/W, $13.1/KW/y, $36.2/MWh, 20y life.
- Coal: $4.62/W, $61.1/KW/y, $18.9/MWh, 40y life.
- Battery (4hr): $1.26/W, $26.8/KW/y, $0/MWh, 20y life.
- H2 (720hr): $2.08/W, $20/W, $5.5/MWh, 20y life.
The solar (PV crystalline), wind (onshore), gas
(peaking), nuclear, coal and battery (lithium) values are
derived
from [Lazard LCOE v13
2019] and [Lazard LCOS v5]
by taking the average value of the cost & time
ranges. Hydro values are derived
from [EIA
Outlook 2019].
The H2 values are based
on [energy.gov]
for PEM
electrolyzers, [Sandia]
for salt cavern storage, and assumes turbine costs are
similar to gas turbines.
Validation
This model has not been validated by others; it
certainly has errors and inaccuracies. I would
appreciate corrections.
I did a very basic validation based on the 2017 ISO
emissions reported
at [iso-ne.com].
The report gives a grid mix of 48% gas, 1% oil, 2% coal,
31% nuclear, 7% hydro, 7% other renewables (trash,
biomass, etc), 3% wind, and ~1% solar. Production was
102,500 GWh, with 31.7 Mt CO2 produced, and average of
309 g/kWh. I checked and tweaked the initial values in
this model to produce a similar result. (I added the oil
and other renewables production to gas.)
Carbon capture and sequestration (CCS)
I'm a bit skeptical of CCS so I didn't include it
explicitly in the tool. However, you are encouraged to
adjust the assumptions for Natural Gas and Coal to
explore the effects of CCS.
Transmission
For simplicity I omitted transmittion and location from
the model. This makes the grid balancing easier because
it's assumed that generation is always close to use. But
it makes grid balancing harder because there is no
provision for import/export of electricity to
neighboring grids. ISO-NE actually imports a significant
percentage of electrity from Ontario, New Brunswick, and
New York, but that complicates the modeling and makes it
less useful as a proxy for other grids. The intention is
to make it slightly harder to decarbonize the model than
to decarbonize a real grid with ability to
import/export.
Hydropower
I modeled hydropower as a constant flat source. This is
not realistic -- depending on the facility, hydro can
have a lot of dispatchability and storage features, and
at other times can be constrained by weather and
environmental requirements. But, those complexities are
very particular to local conditions and work against
applying the results of this tool to other grids around
the world, so I chose to keep the hydro model very
basic.
Acknowledgements
Grid Transition is by Thatcher Ulrich, based heavily on
the wonderful
Energy
Strategies tool released in 2016.
Energy Strategies: concept, design and development by
Amber Robson, Corrie Scalisi, Doug Fritz, Drew Bryant,
John Platt, Jonny Mack, Josh Freed, Kate Brandt, Matt
Jones, Michael Terrell, Orion Pritchard, Philippe
Larochelle, Ross Koningstein, Ryan Fitzpatrick and
Saleem Van Groenou.
Similar tools
In addition
to energystrategies,
see also:
Further Reading
Code
See https://github.com/tulrich/gridx2020
Contact
email: tu@tulrich.com | Twitter: @ThatcherUlrich