Distribution Costs and Distributed Generation
We need to get a handle on distribution costs as we get serious about electrification.
(This post is co-authored with Duncan Callaway)
Our household gets unusually excited about power system infrastructure. We reroute road trips to check out high voltage transmission towers. Our photo album includes pictures of our kids posing with transmission insulators.
Recently our dinner table conversations have been wandering into the topic of distribution infrastructure: conductors, transformers, and substations. You might think this stuff is a bit of a snooze (our kids share this sentiment). But this topic is increasingly important.
For one thing, the amount that utilities are spending every year on distribution system costs is escalating fast. And we have some big decisions to make about where to locate new renewable energy investments (think solar and storage) and new loads (think heat pumps and electric vehicles). How we account for distribution system costs when designing the grid of the future could significantly impact these choices.
So let’s dig into what we know — and what we don’t — about distribution system costs.
Distribution costs are increasing
Every year, major U.S. electric utilities file public expenditure reports. These data aren’t perfect. They only represent about 70% of total US electric load. But they do provide a window into utility spending trends. The chart below shows total annual distribution expenditures (inflation adjusted) by major spending category:
Source: FERC Form 1. These are annual expenditures reported by over 250 US utilities.
Distribution expenditures have been rising over this whole period. But something changed in the last few years. Between 2014 and 2019, capital investments in distribution infrastructure rose 65%! What the heck could be happening? Let’s unpack two possible explanations.
More customers? (No) Looking across utilities, a big driver of distribution system cost variation is the number of connected customers. Cost increases over time could reflect expansions and upgrades to serve an increasing number of new customers. But the total number of connected customers has only increased 6% from 2014 to 2019. So real distribution costs per customer are also increasing (particularly among the California utilities).
Notes: Real annual distribution spending per customer reported by over 250 US utilities. Half of all utilities fall in the range defined by the “boxes”. Approximately 99 percent fall within the “whiskers”. California utilities are highlighted for our California readers.
Increasing peak demand? (Nope) Utilities could be making more distribution system investments to keep up with peak demand. But peak demand for grid electricity (net of behind-the-meter solar) has been declining on average in recent years.
Notes: summarizes real annual distribution spending per kW of peak demand reported by over 250 US utilities. 50 percent of utilities fall in the range defined by the “boxes”. Roughly 99 percent fall within the “whiskers”.
You can see that utility distribution spending is increasing per kW of peak demand. This is particularly true for the California utilities. If rooftop PV can reduce distribution costs, it’s not showing up in these data.
The bottom line is that distribution spending is on the rise across the country, and we can’t explain it with growth in customers or demand alone.
What we don’t know about distribution costs could cost us.
In our own research and past blogs, we (and Severin and Lucas) have argued that distribution costs can’t matter much when making a case for distributed solar. But we want to take a minute to think about how distribution costs could matter in a different situation — one where we’re electrifying new loads — to highlight the importance of getting good estimates of future distribution costs.
Consider this thought experiment that adds electrification to the mix. Suppose we anticipate a growing number of electric cars (EVs). Let’s assume we’re going to build more solar — somewhere — to charge them. And assume we have the technology to charge in perfect unison with output from the solar (a strong assumption). This would likely mean workplace charging, at times when commercial distribution circuits are nearing peak demand. We’ll have to pay to expand the capacity of those circuits to make this scenario work if the solar is utility-scale, but not if it’s located where the EVs are charging.
In California, E3 estimates that avoided costs — coarsely at the utility zone level — range between $15 and $168 per additional kW per year in 2019. A year of production from 1 kW of PV, depending on location, would get you about 1,500 kWh per year. That works out to between 1 and 11 cents of added distribution costs per kWh of solar delivered to our cars in this simple scenario, if it’s coming from far away.
This 2019 Lazard report estimates the unsubsidized levelized cost for grid-scale solar at 4¢/kWh. Using this estimate, and the distribution cost estimates, our clean car scenario would cost between 5 and 15¢/kWh if we choose the utility scale solar path today.
Lazard also estimates residential rooftop solar costs at 20¢/kWh, and larger-scale distributed solar (commercial/industrial rooftop or community-scale) solar at 11¢/kWh. If these distributed options mean we can avoid the distribution upgrade costs in our simple thought experiment, residential rooftop solar still costs more than utility scale and its associated distribution costs. But larger scale distributed options could possibly pencil out.
Source: EVs charging while you work
This rough calculation is designed to show how distribution costs could steer decisions about where we build electrified load and distributed generation. Obviously, there are a lot of caveats and complications that our simple example ignores. For example, in the distributed scenario it might be necessary to upgrade part, but not all, of the distribution system in order to share solar between locations. This all goes to show why it’s important to build distribution cost considerations into the state-of-the-art planning models we are using to plan a greener grid.
Some recent studies are trying to tackle this challenge — for example this one from Vibrant Clean Energy and this one from researchers at Princeton. However if you dig into the details of these studies, you’ll find they’re using long-term historical trends in distribution spending (going back at least 25 years) to estimate average distribution capacity expansion costs per kW. You’ll also find that these studies are using total capital investments in distribution grids, rather than the marginal or incremental costs of adding a new kW of capacity.
The devil’s in the (distribution system) details
A host of factors (low interest rates + generous authorized rates of return?) could be driving distribution spending growth. At least some of this spending rise has been attributed to concerted efforts to overhaul the nation’s aging distribution grid. If that’s the major driver, perhaps spending will settle back down in the future.
Whatever’s driving changes in distribution spending, we need to get a handle on it. And figure out how to account for the going-forward distribution cost implications of big changes to electricity consumption. The distribution system may not be the most exciting piece of the greener grid planning puzzle, but it’s critically important to understand it better.
Keep up with Energy Institute blogs, research, and events on Twitter @energyathaas.
Suggested citation: Fowlie, Meredith and Callaway, Duncan. “Distribution Costs and Distributed Generation” Energy Institute Blog, UC Berkeley, February 8, 2021, https://energyathaas.wordpress.com/2021/02/08/distribution-costs-and-distributed-generation/
Categories
Energy Institute Blog 
One way to get at the extent of the Averch Johnson effect would be to compare distribution costs of IOU to those costs of POU and Coops, which do not have the incentive. At least in generation investment, we proved that reliability was a stronger driver than the AJ effect, this is, levels of investment where different not by IOU vs POU but by utilities that had balancing area obligations vs the ones that had not. I’ll dig into the data above later to understand if there’s marked differences in distribution cost growth between ownership structures.
AJ (or gold plating) is hard to measure if the economic costs are made up. Consider that “costs” for a public utility are typically defined in a two step process: (1) calculation of revenue requirements = sum of operating costs plus the allowed rate of return times the depreciated value of assets in the rate base; and (2) utterly arbitrary allocation of those revenue requirements to customer classes (residential, business, wholesale, etc) and/or services (e.g.. transmission, distribution, etc). Given this is how “costs” are calculated for electric utilities, I doubt that it would be economically meaningful to compare “costs” of a privately owned utility to the “costs” of a publicly owned utility or coop. Consider that coops and munis often lease facilities and buy power from private utilities at prices based on this cost calculation, so I doubt that coops and muni’s costs would reflect anything the resembles economic costs.
I think you may be conflating “costs” with “rates”. The logic behind class cost allocation is to mimic cost causation and making certain customer classes pay for costs that they accrue. This is not relevant for the first step you mention: the determination of revenue requirement. It’s true that POUs won’t have a allowed rate of return, but they will have operating expenses and capital expenditures that can be compared to those from IOUs. And to your second point, coops and munis will generally not own generation and transmission assets, but they will almost always own their distribution system (which are the costs we are interested on in this post). We would need to identify the distribution portion of the costs for vertically integrated utilities, or we could also look at distribution-only IOUs in restructured markets. I think the effort would be worthwhile.
Except if a managed utility’s stock is traded on an open market there is no clear method to figure it’s MTB proportion. The market and book costs of a utility’s parent are not of much worth.
Aside on Averch-Johnson. Recall that the AJ firm is a regulated monopolist, i.e. they choose inputs to achieve the profit-maximizing point on the demand curve given an allowable rate of return. AJ and Baumol-Klevorick show that the firm overinvests in capital, depressing its marginal product below the firm’s cost of capital. The limit to this apparent infinity machine is that more capital also depresses price. Now suppose there are multiple types of capital, some more productive than others. The firm has an incentive to choose the least productive capital! The “gold-plating” metaphor fits this extension even better than the original model.