In Part 1, we looked at how electricity supply, demand, and emissions vary minute to minute and day to day. In Part 2, we’ll estimate the emissions impact of a specific residential solar installation: mine. In other words, if I had NOT installed panels on my roof, how much additional carbon would have been emitted in providing electricity to my house?
The Carbon Accounting Answer
We consumed 17,063 kWh over the first 365 full days after the solar panels were installed. This electricity, consumed from the grid, has emissions associated with its production (our “Scope 2” emissions, as covered in the Terra.do Carbon Accounting course).
The simplest carbon accounting estimate for those emissions would be to simply use the average carbon intensity of our particular region of the grid over that timeframe. In the case of the PJM interconnection in 2022, that was about 811 lbs CO2/MWh, or about 0.386 kg CO2/kWh. (Note: today we’re just looking at carbon dioxide, excluding other greenhouse gases with impacts on climate change.) This equates to 13,808 lbs CO2 or about 6.6 metric tons of CO2. If you have trouble picturing that, imagine 3,300 fire extinguishers full of CO2, just from producing the electricity to power one house.
So, we reduced our Scope 2 emissions by 6.6 metric tons—not bad! The average American’s carbon footprint is 16 tons per year, so we’ve made a good dent. To look at it another way, 6.6 metric tons is how much CO2 that 330 trees would absorb over the course of a full year.
If this were a business reporting its Scope 2 emissions, that would be enough math to call it a day. However, as we saw last time, the carbon intensity of electricity varies by minute and day, so the grid’s average carbon intensity over a period of time is a limited representation of the real world impact.
To get a better estimate of how much less carbon is actually in the atmosphere, we will need to perform additional calculations. But first, we have to make some decisions.
Hourly Emissions
In Part 1, we saw that sites such as Electricity Maps and WattTime track hourly emissions for each region of the grid. By looking at the sources of electricity generation hour by hour, they estimate the carbon emissions of the electricity on the grid. By adding up the emissions (in kg or tons of CO2) from the power generators contributing to the grid and dividing by the total energy produced (in kWh), they arrive at the carbon intensity of the average kWh of electricity for each hour of each day. This is a step forward from using annual averages.
Hourly emissions data also provides the ability to account for the separate timing of our generation and consumption of electricity. Given that our production and consumption don’t align hourly and that grid emissions vary by time of day, a strictly average accounting won’t give us the best estimate of our impact.
Marginal Emissions
However, an hourly reflection of average grid emissions is not necessarily the best reference point to evaluate real world decisions either.
As we are prone to do on this blog, let’s make a baseball analogy. Let’s say your starting shortstop gets hurt in May, and you need to find a new one. What kind of production should you expect? Well, the average shortstop hit .249 last year with a .309 wOBA. Is that what we should expect from a fill-in? In short, no. You probably won’t be able to find an average shortstop to sign with your team in May. You’ll be left choosing between the bottom of the free agent scrap heap and whatever minor league players you might be able to cheaply acquire from another team. Baseball has documented the concept of “replacement level”, which is the below-average level of production that you should expect from this type of fill-in.
Electricity markets have a similar feature due to the way they operate. For a given timeframe, the cheapest energy is bought on the grid first, with costs of additional load rising until electricity demand is met. Renewable energy happens to be the cheapeast energy much of the time, with dirtier, costlier nuclear, coal, gas, and oil filling remaining capacity. As a result, the “marginal” unit of electricity is typically dirtier than the average, which is weighted down by low-carbon renewables. (Electricity Maps has a great primer on the subject.)
So if I’m adding a marginal amount of electricity to the grid’s load by charging an EV, or if I’m removing a marginal amount with solar panels, that impact is felt by the generator of that marginal electricity, which is often dirtier than the average electricity on the grid at that moment.
So, while using an average emissions factor is a fine first order estimate of Scope 2 emissions for carbon accounting purposes, the true impact of these decisions requires an analysis of the marginal emissions on a hour-by-hour basis.
Marginal, Hourly Emissions
Fortunately, WattTime provides Marginal Operating Emissions Rate (MOER) data for different regions on the grid. I’ll pair this with my own electricity data to better compare the emissions saved by our solar production against the emissions released by our electricity consumption.
Some gory details: WattTime’s MOER is provided in five-minute intervals. I’ll downsample and pair this up with 15-minute electricity generation data (from Enphase) and 15-minute consumption data (from PPL’s net consumption data when subtracting out the Enphase production data) to look at both the emissions associated with my home’s consumption and production of electricity. This doesn’t work perfectly—there are some rough edges due to rounding or slight time sync mismatches where PPL says they received more kWh in a 15-minute period than Enphase says we produced in that same period—but I am generally confident this evens out, close enough for my non-professional needs here!
WattTime provides data from the northern California region (CAISO North) for free. Even though I’m in PA, I’ll use this data as a proof of concept. In general, I would expect PA’s grid (in the PJM interconnection) to be dirtier and to have less fluctuation than CAISO North’s, which probably means that my solar array will have a larger impact than demonstrated with CAISO North data.
Results
At long last, here are the results. I have a raw data set that looks like this:
For each 15-minute interval, I have the electricity (in kWh) produced, consumed, and net; I also have the MOER number from WattTime (originally in lbs/MWh but also converted to kg/kWh), which I multiply by the net, produced, and consumed electricity to arrive at the emissions associated with each. I have this data from the installation of solar panels on April 15, 2022 through November 2023.
Let’s look at the first 365 days so we’re not biased by seasonality. From the first full day of solar production, April 16, 2022, through April 15, 2023, our home produced 18,060 kWh of electricity and consumed 17,062—the solar panels covered 106% of our electricity needs for the year, with almost exactly 1 MWh of excess electricity added back to the grid in net. If we use an annual average emissions factor, the accounting is the same—a 106% reduction of our electricity-associate emissions.
However, when we move to a marginal, hourly emissions calculation, our emission impact takes a slight hit. The 18,060 kWh that our panels produced meant that the grid didn’t have to buy that same amount of energy from generators on the margin on those days and times, and those other sources would have released an estimated 5,623 kg of CO2 emissions in producing those kWh. Meanwhile, our 17,062 kWh hours at the days and times of consumption were associate with 6,831 kg of CO2. Even though we offset 106% of our electricity usage, this hourly, marginal accounting shows that we offset only 82% of the emissions from our electricity usage.
Why is that? As noted in Part 1, the grid is cleaner during the day, when solar is going strong. The marginal unit of electricity is cleaner (and cheaper, by the way) when renewable energy is plentiful. Unfortunately, our energy usage peaks right as the sun goes down (though it varies seasonally), when utilities are bringing non-renewable generators online and electricity is dirtier and more expensive. Thus, the energy we consume is dirtier per kWh than the energy we offset with our own generation. The mismatched timing of electricity supply and demand creates challenges.
There are solutions to these, of course. You could start by rearranging the timing of electricity supply and demand to better align. Consumers can shift electricity demand to the daytime and summer hours, or energy generators can shift supply to better match peak demand windows. There are many mechanisms for this, including electricity time-of-use rates, managed charging (such as EV charging), virtual power plants (VPPs), electric vehicle-to-home or vehicle-to-grid (V2X) technologies, and energy storage (e.g. residential or grid-scale batteries). At some point, as renewable energy continues to explode and electrical appliances approach 100% market share, these technologies will become fundamental parts of a reliable energy grid.
Conclusions
Overall, we’re still tremendously happy with our solar panels. The long-term economic return on investment is ahead of schedule, and an 82% reduction in emissions is no small feat. The panels also serve as a great conversation piece with neighbors (one family has followed us with solar already) and with our kids. (On sunny summer days, our four-year-old wakes up and proudly declares, “It’s a great day for solar panels!”)
This reduction in emissions is valuable in and of itself. Many homes and companies can’t totally eliminate their Scope 2 and 3 carbon emissions, at least not immediately, so they look for ways to contribute towards reducing emissions outside of their own scopes through offsets or other vehicles. WattCarbon recently closed its first renewable Energy Action Credits (EACs), helping incentivize the installation of more renewable energy. (Note that WattCarbon projected the impact of their EACs at 316 g CO2/kWh, almost exactly the same as the 5,623/18,060*1000 = 311 g CO2/kWh that my data show, even though we were looking at different grid regions and presumably different data sources.)
In the future, I’d like to get the equivalent WattTime data for my area and other regions of the grid to get a more accurate answer for my home’s true carbon impact. I’m curious how much it would vary across the country or the world.
At some point, our utility might introduce time-of-use rates, or our state might change its net metering policies as others have, at which point we might see a financial benefit to adding battery storage to complement our solar panels.