What are the upstream impacts of internal combustion and electric vehicles?

My previous post comparing the nitrogen oxide emissions of electric and fossil-powered vehicles generated a number of comments (via social media and the blog) to the effect of:

“Power plants may be dirty, but you’re leaving out the upstream impact of refining the fossil fuels to power an ICE vehicle!”

That I did. It was only an analysis of (mostly) local NOx emissions, either from the car’s tailpipe, or the power plant used to generate the electricity (in MInnesota). NOx emissions can have significant local health impacts (inhalation, formation of ozone and smog, etc), but they can also travel long distances. Lifecycle or upstream impacts are also important to consider since we shouldn’t only be concerned about the local air quality impact of our transportation choices.

In an attempt to respond to these comments, I went digging for information on the upstream NOx impacts of the different vehicle fuels discussed in the post (electricity, gasoline and diesel).

The best, most usable, resource I found was the National Energy Technology Laboratory’s Upstream Dashboard Tool (NETL is part of the US Department of Energy). The UDT is “a fast and easy to use tool to determine the environmental profile of various energy feedstocks” according to the documentation. It includes emissions impact information on various “upstream” portions of energy fuels acquisition like raw material acquisition (mining or drilling), raw material transport (the truck, train or pipeline trip fuels must make before they are used) and the “energy conversion facility” (in the case of gas and diesel, this means the refinery process). Basically, you tell it to look at a fuel, and it will give you outputs like air and water emissions, solid waste generation, and water and land use requirements. And it’s all in an easy-to-use excel spreadsheet, just like they advertised!

This tool allowed me to look at the “upstream” impact of the coal and natural gas used in Minnesota’s power plants (to “fuel” an EV), and gasoline and diesel which would be used in ICE vehicles. So now I can answer the question: how do the fuels of ICE and electric vehicles used in Minnesota compare to each other in terms of NOx emissions, accounting for both the “tailpipe” emissions and the upstream emissions?

The upstream emissions

Here are the results from the Dashboard, translated into grams of NOx pollution per mile of vehicle travel:


The Tesla is in fact cleaner in terms of upstream emissions (before the creation of any kWh sent to the electric grid) – responsible for one third as much upstream NOx as either an average gas vehicle or one select diesel vehicle (without a defeat device). This is assuming Minnesota’s electricity mix, which includes 50 percent coal and about 14 percent natural gas. The tesla has no “energy conversion facility” emissions equivalent to a refinery for the liquid fuels, since I’m counting the burning of the coal as part of the downstream emissions.

So, extracting and refining liquid fossil fuels is in fact dirtier (in terms of NOx pollutants) than the extracting and transporting the fuel for an electric vehicle (powered from roughly 75 percent fossil fuels).

The complete picture

Now let’s take a look at the whole picture: from extraction to turning wheels. Here are the results with the downstream emissions included:


Electricity production in Minnesota is responsible for 1.4 pounds of NOx pollution per MWh, meaning an electric vehicle like the Tesla produces 0.21 grams of NOx per mile driven. With Minnesota’s current electricity generation mix, an EV is responsible for emitting about 30 percent more NOx “well to wheels” than typical gasoline vehicle, and 50 percent more than a diesel vehicle.

Electric vehicles can be better for NOx, if we kick coal

NOx rates change significantly based on the amount of coal in the electricity mix. In Oregon, which only gets 6 percent of its electricity from coal (and 32 percent from natural gas), the lifecycle NOx emissions of an EV like the Tesla would be 20 percent lower than an average gasoline vehicle. In Washington, where 75 percent of electricity comes from hyrdopower or other renewables, emissions would be less than half of a gas vehicle. In Wyoming, where 89 percent of electricity generation comes from coal, an EV would emit 80 percent more NOx per mile than a gasoline vehicle on a lifecycle basis. The United States as a whole emits 1.2 lbs of NOx per MWh, translating into a lifecycle per mile emissions rate for an EV that is about 10 percent higher than a gasoline vehicle.

Some Minnesota utilities are on a trajectory to reduce their coal use. However, there is an active political discussion going on right now about the future of coal-fired electricity in Minnesota, and the outcome is uncertain. As I said in the previous post, I think we need to electrify the transport sector to reduce climate risks. However, the choice cannot be between deploying electric vehicles or cleaning up the grid, both must be done simultaneously.


  1. If you’d like to check my math, here is the spreadsheet.
  2. This is still not a full supply-chain analysis, which would account for things like the mining of materials used to build EV batteries and car parts. This is just an analysis of fuels.
  3. Liquid fossil fuels, like gasoline, are getting dirtier over time, measured by carbon impacts, with the addition of new sources like tar sands. The UDT is based on national average compositions for gasoline, so the impact on pollutants like NOx is unclear using this approach. Regardless, electricity still needs to get a lot cleaner.

Minnesota residential solar installation prices in line with national trends

Depending on how you count, this is my sixth post in a row here about solar. I’ll do my best to discuss something non-solar next time.

CY 2014 was the first year the Minnesota Department of Commerce required electric utilities to report on the total installed cost of distributed generation sources (like solar!) in their annual reporting on DG. Prior to that, they’d just been reporting the amount of resources that were interconnected. LBNL recently put out Tracking the Sun VIII, a report on national cost trends, so I thought I would compare.


I was kind of surprised to find out that in Minnesota, we’re basically right on track with national price trends. It would be nice to look at older data, but it doesn’t exist in the reports.

Nationally, installation prices have fallen an average of over 12 percent each year since 2009. If those kinds of cost declines continue, residential solar could be sub-$2/watt by the early 2020’s. Even in wintry Minnesota, sub-$2/watt rooftop solar (the most expensive kind of solar) begins to look competitive with natural gas and coal on a levelized cost basis.

The next Minnesota DG reports come out on March 1, 2016.

Should more solar panels should face west? (with local data)

Here are two charts I made using data from the output of our grid-tied solar array and data from the Midcontinent Independent System Operator (MISO), the people who keep the regional grid (including Minnesota’s) running.


This shows the total load the grid is using in a given hour and the total production in that same hour from our solar array. Our array faces due south. The shape of MISO’s load curve is fairly consistent from day to day, and so is our production curve (on sunny days). The peaks are not coincident, and solar production drops off dramatically in the peak load time. Data from our home use is even uglier (on weekdays).


The second chart compares the same solar production to price. For this one, I averaged all the locational marginal prices at the Minnesota Hub for all the days in July (average prices can vary somewhat by day and hour). The blue line basically shows how expensive wholesale electricity is in Minnesota at a given hour based on supply and demand (side note: check out this great real-time map of marginal electricity prices in MISO at different locations). Solar production seems to match this curve better in the morning, but still misses the opportunity to offset some of the more costly electricity in the afternoon. In July, west-facing panels could frequently be producing until 8 pm and beyond, when the price peak seems to start dropping off.

Rooftop solar usually doesn’t point west, probably because the incentives and utility rates (at least in Minnesota) are designed to maximize production rather than meeting peak load or reducing electricity prices. The New York Times has covered this dilemma. At current levels of solar generation in Minnesota, the issue of west- vs. south-facing panels doesn’t matter much. As generation grows however, this issue is something for utilities and regulators to consider.

The grid is good

We’ve had an Egauge monitor that tracks our home’s solar production on for about a month now. It’s fabulous. For the last week, it’s also been tracking our whole-home electricity usage. Here’s a graph of hourly average power (in kW) from production and consumption.

weekofsolarconsumptionThe solar installers are still adjusting breakers, so output is down, but here are a few things I’ve learned mostly based on our usage.

  • Our house is never using less than 100 watts. We do a pretty good job with power strips, turning off lights, etc. But even at night, even when the refrigerator is cycled off, we’re using around 100 – 200 watts. There are a few things constantly plugged in (battery chargers, sleeping computers, clocks, etc), and I’ll be tracking them down. I’m curious how close to zero watts we can get during the night.
  • Our weekday usage peaks do not match solar production peaks. In the morning, the coffee pot and the hair dryer are running before the sun rises above the trees. The microwave, dishwasher and washing machine often run in the evening on weekdays. On sunny weekdays, we’re sending a lot more energy to the grid than the house is using (since we’re at work). Weekend usage and production peaks match better (see the last to days on the chart). For many reasons, the grid is good.
  • I’m now more obsessed with finding out how much energy things use, and switching them off. The little red line moving in real time on the screen is a great motivator.
  • Our peak demand is probably about 4 kW. The above graph doesn’t show it, because it’s an hourly average, but we hit about 3.5 kW on a very hot weekend day when a number of window air conditioners were running. This is above our peak solar output, which is probably 2.8 kW under optimal conditions. This is also important for a future post I hope to write exploring what it might cost to actually take our house fully off-grid (at least on the electric side).

Only Wikipedia is accurately tracking solar PV capacity in Minnesota

GreenTech Media recently showed that the main energy statistics agency for the United States, the Energy Information Agency (EIA) was missing information on a whopping 45 percent of installed solar PV. The problem is with their methodology – they don’t count customer-sited solar, like systems on rooftops. GreenTech found that actual solar production was 50 percent higher than official estimates. Three states (CA, AZ, HI) now get more than 5 percent of their electricity from solar, something you wouldn’t know if you only consulted EIA.

Data for Minnesota is even worse. Users of wikipedia will get a far more accurate assessment (which is based on the Interstate Renewable Energy Council’s annual solar market trends report).

EIA data says that Minnesota produced no electricity from solar in 2012, and 2.7 gigawatt hours in 2013. Here is the chart from their Electricity Browser:


In reality, Minnesota produced something like 18.7 gigawatt hours from solar PV in 2013, or 592 percent more than EIA estimates.

I didn’t just use wikipedia data to make this estimate, I used data from the Minnesota Department of Commerce’s Annual Distributed Generation Interconnection Report, which utilities are required to submit each year showing existing and new DG facilities. The number above comes from only six utilities in Minnesota, which I think are some of the largest in terms of number of customers.


Here is the breakdown of installed capacity at the six utilities:


I couldn’t find a report on total installed PV capacity or production on Commerce’s website, but it could be hidden in dockets somewhere.

If you’re looking for accurate data on the growth of distributed generation, like solar, you can’t (yet) count on EIA. GreenTech outlines a bunch of reasons why this is important, including making the EPA’s Clean Power Plan to regulate existing power plants look harder to accomplish than it might be.

In Minnesota, solar is a small (0.04% of total generation in 2014), but growing part of the energy mix. Accurately tracking this growth is important for making good policy, especially in regards to distributed (customer-owned) generation, which is usually outside the control of utility planning processes.

Low-carbon energy, land use and community planning

At Energy Collective, Jesse Jenkins looks at the land use impacts of three low-carbon energy sources: solar, wind and nuclear. On solar:

According to the MIT authors, powering 100 percent of estimated U.S. electricity demand in 2050 with solar energy would require roughly 33,000 square kilometers (sq-km) of land. That’s if we spread solar panels evenly across the entire country. If we concentrate solar production in the sunniest regions, the total land footprint falls to 12,000 sq-km.

Those sound like big numbers. On the one hand they are. Massachusetts (where I reside) spans about 27,000 sq-km, for comparison.

On the other hand, the United States apparently devotes about 10,000 sq-km of land just to golf courses. And as the infographic illustrates, it’s agriculture and forestry that truly drives humanity’s footprint on the natural landscape.

In reality, no one is calling for 100 percent solar energy. Even the most bullish renewable energy advocates typically envision solar providing less than half and usually no more than a quarter of U.S. electricity. (See: “Is There An Upper Limit to Variable Renewables”)

If solar provided one-third of Americans’ electricity, it would require just 4,000-11,000 sq-km.

In other words: with an area no larger than the amount of land currently devoted to golf courses, we could power a third of the country with solar energy.


The above is only assuming greenfield development. And wind:

Powering one-third of the country with wind farms would thus truly impact only on the order of 1,800 sq-km, of which only roughly 600 sq-km would be permanently removed from production.

That’s an almost trivially small amount of land, equal to only 6 percent of the land area wasted, er, devoted to golf in this country.

If well sited and co-located on already disturbed and productive agricultural lands, wind farms could thus fuel a sizeable fraction of America’s energy demand without expanding the human footprint on the land in any meaningful way, except aesthetically.

So, on a national scale, the potential land use impacts of really high proportions of renewable energy doesn’t look like a barrier. But most energy projects will have neighbors, and they may not like the look of solar panels, or the loss of borrowed views.

Additionally, putting solar near the busiest parts of the electrical grid, where it can have the most benefit, may conflict with local communities’ plans for future development. A solar farm will probably pay far less in property taxes than residential, commercial or industrial development.

Both of these local conflicts may mean that a significant portion of future solar needs to be co-located with buildings (on the roof), parking lots and other developed areas to minimize land use conflicts and/or reduce transmission costs.

Wind and solar now provide 10% of growth in China’s energy consumption

I got into a twitter discussion about a blog post that was challenging the idea that China is seeming a “renewables revolution”. I do not claim to be a China expert, and I hope I do not qualify as an insta-expert/pundit. However, I can read a spreadsheet.

The blog post compared the growth in solar and wind in 2014 to the average growth in total primary energy consumption averaged over the last decade. I argued that one data point was not a good way to judge a “revolution”, much less discern a trend.

Trying to be proactive, rather than argumentative, I produced the data I thought should be in the post from the same data source.


This chart shows the percentage of the growth in China’s primary energy consumption that is being met by new wind and solar sources (according to BP’s Statistical Review of World Energy 2015)

Before 2005, effectively none of the growth in energy use in China was being met by new wind and solar generation. In 2013, about 12 percent of the growth was provided by new wind and solar resources. In 2014, that figure was about 11.5 percent. So the share of China’s growing energy consumption that is provided by new wind and solar is definitely increasing, and hence so is total wind and solar production as a share of total consumption (1.4 percent in 2014 according to BP).

Here’s another way to look at it: in 2000, wind and solar production in China was basically zero. In 2014, production from wind and solar sources in China was more than the total annual energy consumption of twenty six countries listed in BP’s data book (including developed countries like Kuwait, Austria, Switzerland, New Zealand and Denmark). So the solar and wind generators in China can now provide the equivalent of all the energy (including equivalent of fuels for transportation and heating, not just electricity) required to power a small, modern western European country.

Is this a “revolution”? I’ll leave that to others. I’d say the growth of solar and wind production in China is very, very strong. Of course, the growth in total energy consumption in China is very, very strong also.

Home solar update – to the roof!

Home solar installMy small experiment with solar started with one 100 watt, battery-connected panel resting on the ground in the backyard. I soon added a second and eventually a third panel. I learned a lot about every component, as well as the seasonal and weather-related variability of production.

Over the last few weeks, the hobby was turned over to the professionals, panels went up on the roof, and most of the system is now “grid-tied” instead of going to a battery. We now have 3,840 watts on the roof! Thanks to a very attractive financing option from Innovative Power Systems, our total costs should be similar or slightly lower than normal pre-solar electricity bills. The system should produce very close to the amount of electricity we use in an average year. We’re still waiting for the Xcel Energy engineer to sign-off on the install and switch on the grid-tied portion, but are told that will happen in days/weeks.

Of course, copious data will flow, as solar production AND total home usage will be monitored in real time. I look forward to trying to match usage to production curves.

I kept a piece of the system battery-tied for my hobbyist tendencies. The two smaller panels on the right connect to the battery charge controller and small battery bank in the basement. Both the panels and the batteries have been upgraded in size from the backyard setup, and of course the panels will no longer be shaded by the neighbor’s walnut tree. The usability of the battery system should be way up, and this should allow an interesting comparison between battery- and grid-tied systems.

Thoughts on Xcel’s 2030 Resource Plan

Xcel Energy, the state’s largest electric utility, has filed their 2016-2030 Resource Plan with the Public Utilities Commission. This begins a long process of commenting and modification until their plan is approved by that body (which can take years). The Resource Plan details what trends in usage Xcel expects, and what resources (like new power plants, etc) are needed to meet that demand. The plan is important because it identifies the infrastructure investments the utility will need to make, and also the resulting environmental performance, among many other details.

I’m slowly making my way through it, both for professional and personal interest, and hope to highlight some thoughts for you, my dozens of readers.

There are a lot of things to like in the plan, the first being that Xcel is planning to meet State greenhouse gas emissions reduction goals within their own system. This is unlike the previous plan, which showed emissions increasing between 2015 and 2030. The chart below, from Appendix D, compares the two plans. (State goals include a reduction of 15 percent by 2015, 30 percent by 2025 and 80 percent by 2050)

2030 CO2 Emissions Xcel

Most of the planned reductions in carbon pollution come from the addition of renewable energy resources to their system, as the chart below shows. By 2030, Xcel plans for 35 percent of their energy portfolio to be renewables.

Sources of CO2 reductions

However, I think the plan’s assumptions about the future cost of the solar portion of those renewables is probably too high.

Xcel plans to add over 1,800 MW of utility-scale solar to their system by 2030 (up from basically zero in 2015). This is a significant increase from the “reference case”, a ten-fold increase in fact. However, this slide was presented at a public meeting at the Public Utilities Commission:

Renewable Price ForecastXcel says this in Appendix J about their assumption:

As solar technology is still not fully mature, and costs are expected to decline and conversion efficiency to improve, it was assumed that the $95/MWh price holds throughout the study period. In effect, the assumption is that fundamental cost driver improvements will offset inflation.

So the rate of decrease in solar prices will match the inflation rate? Many sources have documented the dramatic decline in solar PV prices over recent years. Lazard seems to be an oft-cited source, and their 2014 Levelized Cost of Energy Analysis shows the price of energy from solar has dropped 78% since 2009. According to usinflationcalculator.com, the cumulative rate of inflation between 2009 and 2014 was about 10%. So, at least looking historically, this seems way off.

Of course, current precipitous declines probably won’t continue forever (most of the cost is now not modules). NREL says costs have been dropping on average 6 to 8 percent per year since 1998. If we assume just half of that decline per year (4 percent), solar energy would be around $51 per MWh in 2030. Using some very back-of-envelope calculations, a price difference of $46 per MWh in 2030 means costs for new solar energy shown in the Plan’s “Preferred Plan” scenario could be over-estimated by $97 million.

This is significant not just because the price estimates of the Preferred Plan may be too high. In preparing the plan, Xcel also ran seemingly dozens of other scenarios, some including CO2 reductions of over 50% in 2030 (compared with 2005). The price difference, according to Xcel, between the Preferred Plan scenario and the scenario with the largest CO2 benefit is $172 million (from Appendix J). These other scenarios which seem too costly may actually be more in line with what Xcel is currently asking to spend once dropping technology costs are factored in.