In Chapter 1 we made the case for hourly emissions modeling, and Chapter 2 we saw how to slice the data. We’ll continue our data discussion by turning to a unique grid region that tests the limits of our current ways of looking at emissions data.

One great advantage of living in the Pacific Northwest is that in addition to good coffee and non-threatening college mascots (Go Beavs!), we are a uniquely bizarre electric utility region. Our grid is nestled between emissions extremes of hydro (BPA) and coal (Pacific Power). It has large multi-state IOUs and small municipal PUDs. Our region is in the process of joining larger energy imbalance markets (WEIM) which means the past doesn’t necessarily predict our future. Honestly, no one seems to entirely understand what is happening in our little corner of the universe, but the lights do stay on so obviously something about it works.

From a modeler’s perspective, working in the northwest provides insight into the limits of the current emissions data sets. Since we aren’t following the rules quite as nicely as the more studied regions like California, the uniqueness of our grids allows us to accidently break the models and reveal cautionary tales of the power for these data sets to be used for good or evil.

To highlight the diversity in possible emissions rates for the NW, we’ve created the table below summarizing all the possible outcomes for our region.

(Note: This data is from the Cambium 2021 model, not the just released – woohoo!! – 2023 model. Excited to update this table after cracking the hood on the new model in the coming weeks.)

Its intriguing to see the range in values depending on which lens you look through. And the point of highlighting the variety in emissions rates is that these results can have very real implications that alter the outcome of an analysis.

There are two key ideas this summary brings to light:

The first is that our large amount of hydro (and increasing amount of wind) can obscure the fact that anyone who isn’t BPA still has a decent mix of natural gas and even coal. This is rapidly changing, but the difference between Cambium averages and the specific utility-reported mixes highlights that we still have a ways to go. It becomes a philosophical debate on what this actually implies, since much of our generation is sold on the imbalance markets to California and other regions. But for my team, we’ve decided to still use utility-specific numbers for annual calculations of operational emissions in PSE, PGE, and PAC territories instead of AVG.

The second takeaway is perhaps the more critical one as it highlights the importance of caution when using SRMER and LRMER. Used incorrectly, SRMER as compared to a natural gas baseline can indicate electrification isn’t the best option in the NW, which is an incorrect conclusion. The NW IOUs are changing and it’s happening quickly. Making a decision on a system with a 20-year lifetime for a grid state disappearing quickly is not the right approach. LRMER lands on the correct conclusion (electrify all the things!); SRMER does not. (There is a side conversation to be had on whether this is even the correct value to use for natural gas – spoiler alert: its not! – but that’s for another day)

Yet the comparison becomes even more interesting when we layer on the “shape” of the emissions profile – i.e. not just the annual average, but what is happening when. In the interest of space, these are from one case example - Oregon Mid-Case, 2022- AVG, SRMER, LRMER from top to last (keep in mind that blue is, perhaps counterintuitively, the high emissions and gray the lower emissions times):

The plot now thickens because this data tells us the West Coast remains challenged to meet peak events with low emissions resources. In this case, SRMER likely does provide accurate representation of the impacts of DERs and other flexible load solutions installed in the near term. This can become important if a project is considering budget for a triple panel of glass or an onsite battery. SRMER helps us see that avoiding peak times with the battery might be more impactful than the slight energy reduction from windows in the temperate NW climate. SRMER helps highlight this more nuanced opportunity for operating emissions reductions, while LRMER can undervalue the importance of flexible loads.

The point is with great power comes great responsibility. As data analysts and engineers, we have a responsibility to make sure we’re using the correct model to answer the right question. In our efforts to simplify and standardize around a single data set approach, I wonder if we might inadvertently miss opportunities for a deeper understanding on system decisions.

It's worth noting that part of the challenge is that this is an emerging field of study. The industry isn’t yet able to pull from years of historical models back-checked against decades of historical data. The models are all fairly new, most emerging within the last half decade. And the real-world data is limited at best, so back-checking is challenging and often impossible.

All that is to say – as our understanding improves, the data will also improve. But until then, it’s worth taking a bit of time to really consider the grid scenarios underneath the data options and compare them to the question you are trying to answer. Pairing the wrong scenario to your model could point you towards an incorrect future outlook.

Next up – Dr Strange Data or: How I Learned to Stop Worrying and Love the Hourly Emissions: The existential question of hourly emissions modeling – does it really matter?

Also published on LinkedIn by Karina Hershberg

Chapter 1: Not All That Glitters is Zero Carbon
Chapter 2: The Emissions Multiverse
Chapter 3: With Great Power Comes Great Responsibility
Chapter 4: Dr Strange Data, or, How I Learned to Stop Worrying and Love the Hourly Emissions
Chapter 5: Wear Nice Glasses and Design Beautiful Spaces
Chapter 6: No One is Net Zero Until Everyone is Net Zero