Sie sind auf Seite 1von 12

Time and Locational Value of Distributed Energy Resources

McGee Young, Recurve, Mill Valley, CA

ABSTRACT
The reality of grid decarbonization is shifting the tectonic plates of the energy industry. Within
the family of distributed energy resources (DERs), energy efficiency is the most established, but is poised
for the greatest disruption. Energy efficiency has long relied on deemed savings values to determine its
grid impact and to compensate implementers. Now that tools are available to quantify the actual time
and locational impacts of efficiency projects, a wholesale shift to a performance-based valuation approach
that is based on addressing grid requirements is within reach. This paper navigates between competing
approaches to DER valuation and illustrates how hourly usage data, the use of locational marginal pricing
of energy at the grid node, and the strong correlation of weather to energy use can be combined to value
energy efficiency at any given time of day and location. We introduce the concept of a “CarbonWatt” and
discuss how this concept of valuing DER portfolios can enable new technologies and business models that
will allow us to reach our decarbonization goals.

Introduction

The shift in system peak from afternoon to evening, precipitated by the widespread adoption of
rooftop solar, highlights the potential for energy efficiency (EE) and other distributed energy resources
(DERs) to offer clean, non-wires supply alternatives that align with patterns of grid demand. With new
policies steadily embracing concepts of grid decarbonization, we must begin to create mechanisms that
value time and locational savings. Moving from the current generalized valuation model of EE to one that
is tied to how EE alleviates time- and location-based grid constraints, as well as the value that EE can
generate by reducing our reliance on carbon-based fuels, will align EE’s valuation with its actual value.
This paper focuses on DERs as a broad category and EE as a specific subcategory of resources that
stand to benefit from an emphasis on time and locational valuation. We mix a combination of fresh
empirical research with an extensive discussion of the practicalities of moving towards a valuation schema
that meets the needs of decarbonization goals. Our research features meter-based hourly savings analysis
of a major residential energy efficiency program as well as a deep dive into the dynamics of wholesale
energy market pricing. Together, these inform the development of a framework for valuing the time and
locational value of distributed energy resources.
As we discuss EE in this paper, we are referring to downstream programs that target specific
buildings. We recognize that, for the most part, efficiency measures are not “dispatchable,” in the sense
that the effects of a measure (like insulation) cannot be switched on and off to meet grid needs. However,
there are patterns as to when efficiency gains are realized, and these patterns can be forecasted. In this
sense, efficiency can be viewed as an investment in grid capacity whereby buildings and measures are
targeted to produce efficiency savings at optimal times and in optimal locations.
Finally, our discussion focuses on policy developments and case studies from California. We
recognize that other states (New York, Arizona, and Florida for example) are experimenting with similar
policy initiatives, but California’s unique mix of policy experimentation, alongside its investment in data
infrastructure, creates a unique opportunity to imagine what a decarbonized grid of the future might
resemble.
California DER Regulatory Context

In 2015, by order of the California Public Utilities Commission (CPUC), California’s investor-owned
utilities (IOUs) began developing Distribution Resource Plans (DRPs) that focused on identifying the
optimal locations for Distributed Energy Resources (DERs) in their service territories. The goal of this DRP
initiative was to figure out how to deploy a combination of solar, storage, efficiency, demand response,
and electric vehicle infrastructure to reduce grid congestion and to delay or to eliminate infrastructure
projects.
An initial attempt to quantify the grid value associated with DER deployment (conducted by the
consulting group E3) illuminated the complexity and the challenges associated with valuing DERs. Known
initially as DERAC (Distributed Energy Resources Avoided Cost Calculator), the tool attempted to quantify
seven distribution, transmission, and generation costs that DERs could offset. Post-mortems of this
initiative pointed to three factors that posed challenges to DER valuation. First, system models capable of
projecting more than a year or two into the future necessarily over-simplified, and in this case largely
failed, to capture the long term avoided costs of deferred capital projects. Second, even if a location could
be identified for DER valuation, the costs of developing DERs for specific locations would vary widely: a
simple metric for grid value would obscure the variable costs of DER deployment. Third, because of the
uncertainty around the fully-burdened avoided costs and the reality of differential DER development
costs, the tool was inadequate to compensate DER providers for the grid benefits they would be providing
(Hall 2018, Trabish, 2017).
PG&E’s assessment of its DRP pilots points to challenges to effective deployment of DERs as grid
resources. First, utilities have little control over where DERs will be deployed or to coordinate DER
deployment since such decisions are driven largely by customers. Second, deployment decisions are a
function of customer cost, not of utility benefit: there is little to no grid price signal sent by utilities that
might influence DER decisions. Third, consumers are known to be fickle: past patterns of DER adoption
may not be indicative of future trends (in both directions). Therefore, projecting the naturally-occurring
rate of adoption of DERs in a specific location is challenging. Fourth, what these choices reflect, currently,
are the result of past policies that might be revised (subsidies for solar, for example). Without being able
to forecast the changes in policy that might occur, it is impossible to know how consumers might act in
the future. Finally, many DERs are part of a class of “emerging technologies” with limited penetration into
consumer markets. Given the foregoing, it is difficult—if not impossible—to predict the grid impact of
these technologies (Aslin 2015).
In short, the DRP approach to DER valuation has left market participants struggling to develop
viable business models or to deploy technologies that meet the restrictive requirements and valuation
constructs developed to date. There are very few viable paths to DER deployment and massive uncertainty
that the assorted benefits of DERs would even be recognized by procurement or regulatory agencies.

Regulated Efficiency and Avoided Costs

By contrast, EE programs funded under the regulatory oversight of the CPUC are valued through
a cost effectiveness test1 designed to calculate the net present value of a program relative to the myriad
costs that would be borne by program participants, ratepayers and other stakeholders. The primary
determinant of the value of efficiency programs is an “avoided cost” calculation that is derived from a
combination of infrastructure and societal benefits, including the more obvious factors like reductions in
transmission, distribution, and capacity costs, as well as line items for greenhouse gasses and more
obscure benefits like reductions in ancillary services costs. This approach illuminates how the value of
efficiency is both enabled by a framework that prioritizes physical infrastructure and is constrained by the

1
The Total Resource Cost Test (TRC), see the California Public Utilities Commission, 2001
2019 International Energy Program Evaluation Conference, Denver, CO 2
very same set of mechanisms (CPUC 2001). As we show in Figure 1 below, the lion’s share of avoided cost
is realized from capacity – a grid constraint – while greenhouse gas reductions comprise a much smaller
portion of the total value of the savings.
If time and locational savings matter for the future, to the credit of regulated EE programs, the
avoided cost framework at least recognizes the value of savings during particular times of the year. For
most programs, benefits are concentrated in the summer months, where grid constraints are highest, as
demand related to cooling reaches its peak. On an hourly basis, the time of day in which efficiency is most
valued skews toward the late afternoon and early evening hours as the confluence of daily solar
downcycling and evening demand peaking result in maximum efficiency value.
For programs whose savings tend to be realized most during these critical times, the grid benefits
are quite striking. Consider Figure 1 that shows the average annual value of savings by hour for PG&E’s
Advanced Home Upgrade Program (AHUP).2

Figure 1. Average annual value of savings by hour for PG&E’s Advanced Home Upgrade Program. Values developed
by measuring hourly savings for program participants and using the Avoided Cost Calculator to determine the hourly
values. Source: Recurve 2018.

The avoided costs generated by AHUP peak between 5PM and 7PM, driven in large part by the
grid savings, specifically the capacity savings of those two hourly blocks. Because of the way grid value is
recognized for programs like AHUP, PG&E is piloting a Pay-for-Performance EE program that incentivizes
savings that are concentrated during the key window between 4PM and 9PM weekdays by paying three
times the base rate for savings that materialize at these times. (Clearly, avoided costs due to the grid
benefits of efficiency are helping justify the procurement of time-based savings.)
At the same time, the fact that this avoided cost framework rewards efficiency based on the grid
value of demand reductions is also fundamentally limiting. The avoided cost structure of the CPUC’s cost-
benefit framework is only as granular as the state’s 16 climate zones, while grid constraints are typically
very localized. At best, this framework is providing “average” savings, while the types of resource
adequacy requirements envisioned in the DRP and Integrated Distributed Energy Resources (IDER)
proceedings are more explicit and targeted. Furthermore, neither the existing cost effectiveness
framework for EE programs nor the various DER initiatives provide space for explicit policy objectives like
decarbonization of the grid. So long as benefits like greenhouse gas reductions are viewed as a reflection

2
Chart values were developed empirically by measuring the hourly savings for the program’s participants and using
the Avoided Cost Calculator to determine the hourly values.
2019 International Energy Program Evaluation Conference, Denver, CO 3
of current carbon pricing rather than aspirational carbon reduction goals, EE will never meet cost
effectiveness thresholds beyond the most conservative and unambitious forecasts.

Moving Toward Time and Locational Savings: Finding a Middle Ground

There is likely a middle ground between the resource expectations of the DRP and the broad
valuation of efficiency that characterizes regulated programs. This middle ground would recognize the
value of EE at specific times and in specific locations but would do so with the recognition that there is a
societal dimension to EE (as well as other DERs) that must be recognized that is fundamentally different
than a power plant, substation, or transmission system. Whereas a power plant needs to be built and
commissioned (no small task, but the necessary inputs and outputs can be specified), DERs must be
procured through a process by which third parties give their consent to the project and, in most cases, co-
invest in the outcomes of the project.
Fortunately, this middle ground is closer than we may imagine. Commitment to Pay-for-
Performance programs will encourage DER developers to target customers with the highest savings
potential (Scheer et al 2017; Borgeson et al 2018). Extra payments for savings during critical evening
periods will encourage investment in the types of measures that combine to reduce or delay energy
consumption during that time.
Where policy needs to catch up is by valuing peak demand reductions independent of the existing
net present value of the existing avoided cost framework and toward adopting a broader policy objective
of reducing evening peaks, since these peaks are the largest impediment to grid decarbonization.
Similarly, procurements need to reflect the multidimensional nature of DER adoption. Rather than
stipulating explicit resource adequacy requirements, procurements can tap into DER adoption in a more
creative fashion. Ultimately, clear price signals to the market may do more to meet resource needs than
an approach by which the DER deployment becomes an extension of a utility-centric paradigm.

Lessons from the Field

Now we take a step back to examine how energy efficiency effects are detected at a time and
locational level when savings are measured using interval smart meter data. To demonstrate how
residential EE projects contribute to changes in the shape of energy consumption, we analyzed some
4,000 projects that were completed as part of PG&E’s AHUP Program.3 We clustered these projects into
portfolios based on the weather station that was used to provide temperature data for the regression
analysis. We applied the CalTRACK4 methodology to calculate a counterfactual consumption value for
each hour of the post-treatment year and subtracted the actual consumption to determine the hourly
savings for each project, or what we call the Resource Curve (RC). For ease of illustration, we show the
savings values as aggregated monthly RCs. The figures below show the pre- and post-intervention load

3
AHUP projects are “whole house” multi-measure interventions that typically include some combination of HVAC
replacement, insulation and/or ductwork. Its emphasis on building shell and mechanical system improvements
suggests that energy savings should largely match seasonal weather variations. Even the program has been plagued
by underperformance relative to predicted savings, its emphasis on deep shell and system measures suggests that
energy savings should largely match seasonal weather variations.
4
CalTRACK is an open-source savings calculation methodology that uses outdoor temperature to project a
counterfactual consumption value from baseline conditions using a building’s pattern of energy consumption
relative to weather conditions prior to the installation of EE measures. It can be used to predict what usage would
have been in future years by using the statistical relationship established in the pre-retrofit period and the weather
conditions of the post retrofit period. CalTRACK can be used for forecasting as well.

2019 International Energy Program Evaluation Conference, Denver, CO 4


shapes and the RCs (in light blue) corresponding to the portfolios in each geographic cluster specific to
the month of August.

Figure 2. Hourly savings values for PG&E’s Advanced Home Upgrade Program for Hayward (upper), Vacaville
(middle), and Fresno (lower) areas for a typical weekday in August. Source: Recurve 2018.

There are significant seasonal and regional differences in load shapes and RCs. Consider the top
chart, which shows the aggregate savings for projects in the San Francisco Bay Area. These projects show
an almost identical consumption pattern after project completion as before for the month of August
(performance is slightly better in some other months, but generally project savings are minimal).
By contrast, projects occurring in the climate zones about 50 miles northeast of San Francisco
(middle chart) show a more distinctive RC, both in the aggregate of a full year as well as during peak
summer months like August. Likewise, projects in the Central Valley, near Fresno, also show a significant
increase in savings during the peak summer months (bottom chart). For projects in hotter climate zones
such as this, savings peak during the late afternoon and evening hours, precisely the time in which the
delta between demand and renewable energy supply is at its greatest.

The Importance of Weather

Where many building model simulations focus on square footage, type of roof and windows, and
other physical characteristics, weather-based regression methods illuminate a larger principle concerning
the importance of weather for understanding patterns of energy consumption. The correlation of a
building’s energy consumption and the outdoor temperature, measured prior to an EE project, is one of
the strongest predictors of how likely a building is to save energy following a retrofit that involves weather
sensitive measures. To a large extent, it is possible to forecast energy consumption over a period of years,

2019 International Energy Program Evaluation Conference, Denver, CO 5


if not decades, in portfolios of buildings by establishing the existing relationship between energy
consumption and weather and projecting it forward using various future temperature scenarios. This
projection is important because it gives us both a way to forecast demand as well as calculate energy
savings over a multi-year period.
EE programs like AHUP have the potential to increase savings by as much as 50% just by
eliminating subsidies for building retrofits where the baseline relationship between increases in energy
consumption and increases in outdoor temperature is negligible. Similarly, interventions that focus on
buildings with high temperature-to-load correlations tend to yield savings above the median and are
therefore highly aligned with the needs of the grid.
Because each building’s temperature sensitivity is different, statistical certainty regarding the
savings achieved in each building will vary. When projects are aggregated into a portfolio, however,
energy savings tend to be more consistent and predictable. The deeper the savings, the greater the
number of buildings in the portfolio, and the more stable the relationship between temperature and
energy consumption, the more confidence that we have in the longer-term delivery of portfolio energy
savings. In this sense, portfolios of EE projects are good candidates for the types of resources envisioned
by the earliest DRP and IDER working groups.

Dynamic Pricing: Leveraging Locational Market Pricing Values as a Proxy for the Value of EE

The main challenge for the full integration of EE within a DRP framework is one of valuation and
procurement. In this section, we discuss how Locational Marginal Pricing (LMP) values from the California
Independent System Operator (CAISO) can serve as a proxy for valuing the precise time and locational
value of EE. The CAISO wholesale power market prices electricity based on LMP, the cost of generating
and delivering it from some 9,700 specific grid locations called market pricing nodes (PNodes). One
California energy market runs the day before the energy is needed (day ahead market), while another one
runs in real-time to balance last minute demand needs. Daily auctions set a day-ahead price while
automated 5-minute interval auctions allow fine tuning of energy procurement to meet instantaneous
demand requirements. Insights from this exercise provide some visibility into how procurements might
operate within a time and locational valuation framework.5
From a decarbonization standpoint, EE is most valuable when it reduces consumption in the peak
evening hours. At a basic level, what matters is the difference between the available supply of renewables
and aggregate demand for any given time of day. In reality, the rate of change in that relationship will
factor into the amount of non-renewable idle capacity that will be required to stay online to meet
expected demand. But for our purposes here, we will stick with a more simplistic value formulation.
We can understand this concept intuitively from the familiar Duck Curve. Solar production, which
peaks in the late afternoon, is replaced by generation from peaker plants that operate to meet spiking
demand during the early evening hours. Of course, there is no singular Duck Curve. Each local node on
the grid faces its own unique pattern of distributed generation and evening peak, a pattern which, over
the course of the year, will exhibit wide variation (a veritable zoo of animals might be projected onto the
load shape). Whereas the Duck Curve illustrates the relationship between renewable supply and demand,
a valuation framework for time and locational EE needs a richer concept of the costs associated with
delivering energy under specific scenarios of over and under supply. Fortunately, the concept of LMP
precisely captures this relationship.
LMP is not a perfect corollary to the type of dynamic valuation that might ultimately inform EE
procurements, but it provides us with a starting point to think about how the costs of delivering clean
energy might be captured by a single pricing signal. What we want to look for in LMP is the difference
between the cost of providing energy when renewables are plentiful—late mornings and early afternoons

5
See http://www.caiso.com/PriceMap/Pages/default.aspx
2019 International Energy Program Evaluation Conference, Denver, CO 6
and in the spring and early summer on an annual basis—and the cost of providing energy when
renewables are scarce and demand is high. Because LMP pricing differences are mainly a function of
congestion (local demand is high relative to local supply) and line loss (importing energy from outside load
aggregation points6), they provide a reasonable approximation of the valuation that we would be seeking.
Consider, for example, the LMP-shape of the node in Mill Valley on April 1, 2019 (see Figure 3).
The dual-peak LMP-shape on this day shows a morning peak, followed by an afternoon trough, followed
by a steep ramp into the evening hours, before dissipating overnight. This is a classic Duck Curve pattern
reflecting the prevalence of local renewable energy and likely the effects of intercity migration during
working hours.

Figure 3. Spring daily load shape for Mill Valley LMP. Source: Recurve 2018. Based on published data found at
http://oasis.caiso.com/mrioasis/logon.do

As the spring sunlight starts to generate solar production and while days are mild, daytime LMP
prices will often approach zero (or even dip into negative values). Negative LMP pricing represents a
situation in which the evening ramp requirements require spinning reserves of baseload to remain active
even though, in combination with solar, supply will exceed demand. In this case, excess electricity will
need to be offloaded to neighboring nodes or even neighboring markets. For example, consider the LMP-
shape from Oakland a little later in the month of April (see Figure 4). Morning demand gives way to zero
prices for the middle of the day followed by a steep ramp in the early evening hours.
By contrast, the same Node in Oakland in November maintains consistent pricing throughout the
day, followed by a smaller ramp in the evening hours. However, by November, hydro-electric power has
dwindled, resulting in higher pricing throughout the day than what is found in April. Even though the
afternoon ramp is much steeper in April, from trough to peak the dollar difference is greater in November
than it is in April.
These seasonal trends pale in comparison to what happens when summer heat waves cause the
widespread use of air conditioning. One such event can be seen in the LMP curve for August 8 in the
upper-right of Figure 4, which shows LMP exploding in the afternoon from $100/MW to nearly $500/MW
at peak. This sort of ramp is difficult for production systems to handle and often results from severe grid
imbalances.
When the heatwave passes, however, prices return to normal. Consider that, just a week later in
the same node in Oakland, LMP leveled out at around $80/MW on a much more gradual afternoon peak
(see lower-left chart). This pattern is common during summer months when higher outdoor temperatures
result in spikes of energy consumption.

6
Load Aggregation Points (LAPs) are divisions of electricity service that follow the divisions of the IOUs and are
further divided into sub-LAPs.
2019 International Energy Program Evaluation Conference, Denver, CO 7
Figure 4. Assorted LMPs. Source: Recurve 2018. Based on published data found at
http://oasis.caiso.com/mrioasis/logon.do

Such “peaky events” are not equally distributed. In some cases, especially where the weather
conditions are more consistent (even if that means just “hot” all the time), the afternoon ramp will be
more predictable. For example, during the August 8 heat wave in Fresno (see left-hand chart in Figure 5),
the outdoor temperature was 102 degrees (30 degrees warmer than Oakland). But while the afternoon
ramp bore a striking resemblance to the LMP-shape in Oakland, pricing maxed out at $150/MW—less
than what was found in Oakland. By the time temperatures cooled a week later, the LMP-shape in Fresno
on the same day (see right-hand chart in Figure 5) assumed an almost identical form as in Oakland (see
lower-left chart in Figure 4). These examples of LMP-shapes suggest that energy savings could be fitted to
a more dynamic valuation model than is currently provided. Savings in a certain location at a certain time
provides massively different value to the grid.

Figure 5. Assorted Fresno LMPs. Source: Recurve 2018. Based on published data found at
http://oasis.caiso.com/mrioasis/logon.do

2019 International Energy Program Evaluation Conference, Denver, CO 8


Time Value of Efficiency

Given these diverse LMP patterns, we can now assess the actual value of energy savings if they
were valued according to the LMP–that is, by their time and locational value to the grid. Recall that earlier,
we showed a few typical months of hourly savings associated with AHUP. We now create a valuation
framework by multiplying the LMP prices for particular nodes against the savings that were typical of
projects in those areas in those months.
Figure 6 shows the value of AHUP program savings for the average project within the nodal region
(roughly construed) for which a given price was identified. To obtain the values, we multiplied the hourly
kWh savings values from the average AHUP project by the hourly day-ahead LMP price per MW. The best
way to make sense of the monetary value is to construe the value as if each represented 1,000 treated
homes.
What is striking is the magnitude of difference between the EE value that would be delivered on
a particularly hot day in Fresno, versus a typical summer day in Fresno, versus any other location. The
combination of high LMP and high marginal savings from AHUP suggests that residential efficiency
investment in this part of California (the Central Valley) delivers massive value. By contrast, on the same
August day, Oakland homes increased their consumption relative to their baseline, thereby delivering
negative savings to the grid. The retrofits in these homes added load to the grid during the times that it
was most beneficial to have been saving. For most other locations, modest savings and modest LMP values
resulted in very low valuation relative to the projects highlighted by these examples.

Figure 6. Locational Value of selected AHUP savings results as a function of LMP. Source: Recurve 2018.

There are two initial takeaways from these examples. First, there are obviously great disparities
in the grid value of EE projects. However, we should not confuse this value with the other values that are
obtained from EE projects (health, safety, comfort, other policy goals). The point here is not to say that
we should only invest in EE in locations such as Fresno. However, we should recognize that time and
location will matter in determining the grid value of EE.
The second takeaway is that we now have a way to capture the dynamic value of EE as it pertains
to decarbonization. To the extent that demand for energy exceeds available carbon-free energy supply,
efficiency can be valued for the portion of the gap that it closes. This basic construct, that we can value
EE based on both when and where it occurs, relative to demand and to the availability of renewable
supply, provides a basis for new policy directions that will help scale the deployment of DERs.

2019 International Energy Program Evaluation Conference, Denver, CO 9


Towards a New Model of Procurement

We began this paper by recounting some of the challenges that faced early efforts to create a
valuation function for distributed energy resources. We noted that some of the shortcomings of DERAC
were related to ambiguity associated with the value of infrastructure deferral. We also noted that a
higher-order focus on decarbonization would provide an alternative to asset-based valuation. We made
the case that time and locational value, as represented by LMP, could serve as a proxy for DER
decarbonization valuation, even if inadequate in its pure form. Still, we have not discussed the other main
shortcoming of DERAC, which was the difficulty establishing a framework for procurement. As was noted,
it is difficult to forecast the naturally-occurring rate of DER penetration in the grid and more difficult still
to determine a price function that matches the avoided costs of capital investments.
Decarbonization offers a more straightforward valuation approach. Each unit of carbon-based
energy produced to meet demand can be thought of as what we term a “CarbonWatt.” For any given hour
of the day, on any node on the grid, there will be some amount of CarbonWatts required to meet demand
(except hours in which renewable production outpaces demand). For example, if demand totaled 100GW
and renewable supply provided 60GW, there would be 40GW of CarbonWatts on the grid at that time. If
renewable supply fully met demand needs, there would be no CarbonWatts on the grid. If our goal is
decarbonization, the logic is simple: quantify the difference between demand and renewable supply (i.e.,
calculate the CarbonWatts needed to be reduced) and issue procurements based on the reduction in
CarbonWatt for each hour of a day.
Under this type of procurement scenario, a CarbonWatt would be technology agnostic. The same
carbon reduction goals could be achieved by adding more renewables, increasing energy efficiency, or
deploying energy storage systems capable of discharging during peak CarbonWatt times of day. All that
would need to be known by the procuring organization is the marginal hourly carbon intensity of the grid
and a single clearing price could be established for any viable combination of DERs.
For DER suppliers, participating in this type of market would represent a giant step forward
relative to existing EE programs or nascent IDER procurements. For starters, using a CarbonWatt or similar
type of concept would create the flexibility needed to deploy combinations of DERs to match customer
preferences and to support innovative technologies and business models. With the right locational value
granularity, differential costs of DERs would be reflected in procurement bids, rather than the one-price-
fits-all approach that currently prevails. We would likely see more investment in EE in Fresno and more
batteries in Oakland, each capable of providing the same CarbonWatt offset, but at prices reflecting the
challenges and opportunities of each region.
Most importantly, if LMP is any guide, DER providers would be able to forecast the long run value
of their investments. The uncertainty that grips current DER procurements, primarily related to T&D
deferral value, would be replaced by the quantification of risk not dissimilar to other futures markets. To
show why this is the case, we tested one additional hypothesis to see if we could predict the LMP value
over time.
Recall that the CalTRACK savings calculation could be used to calculate savings as well as to
forecast demand. If the LMP price of energy is related to both weather and patterns of demand, it might
be possible to predict LMP based on prevailing weather conditions and the time of the day. To take a stab
at this question, we modeled the day-ahead LMP using the same CalTRACK hourly model that we used to
calculate savings. Instead of energy consumption as the dependent variable, we modeled the price of
delivering energy, with the time of week and temperature as the independent variables.
We found that between 40% and 50% of the variation in price across nodes is predicted by the
combination of time of week and temperature. The model performed especially well in places like Fresno,
where temperatures are more varied, than in than in locations like Oakland, where exogenous factors
(such as grid congestion elsewhere) appear to account for a greater portion of the change in price.

2019 International Energy Program Evaluation Conference, Denver, CO 10


Figure 7 below, taken from the model fit in Fresno over the month of August, illustrates this point.
To a striking degree, we can anticipate when and by how much the price will rise given certain weather
conditions. We can even see that the heat wave that occurred during the early part of the month resulted
in higher LMP prices than the model predicted (a consistent trend). We are also able to forecast savings
from EE projects with the same approach, thus making it possible to forecast the future CarbonWatt value
of EE investments with confidence.

Figure 7. LMP in Fresno versus CalTRACK model prediction. Source: Recurve 2018. Data from
http://oasis.caiso.com/mrioasis/logon.do

This type of novel approach to DER valuation is sure to raise more questions than it is ready to
provide answers. For example, we still need a more precise auction function similar to the one that allows
LMP to reach a clearing price, but that considers the long-run value of a particular DER (perhaps a concept
of a Savings Purchase Agreement). We need to better understand the time and locational value that EE
can deliver relative to storage or solar contracts. There needs to be a methodology that allows demand
response to participate on equal terms as well. However, these challenges seem surmountable—
especially as they compare to the potentially catastrophic effects of climate change if we do not succeed
in decarbonizing our grid. With a functional market capable of delivering on the decarbonization promise
ready to stand up, we can hardly afford not to try.

Implications and Next Steps

In this paper we have shown that a gap exists between competing visions of EE as a distributed
energy resource. The DRP framework would position efficiency as a procurable resource, like existing
physical assets, but has struggled to effectively value and procure it. Efficiency programs are embracing a
meter-based savings and valuation framework and transitioning to Pay-for-Performance, where delivered
savings are valued based on total quantity and time of day. However, existing cost effectiveness tests
constrain our ability to target efficiency where it could matter the most: decarbonizing the grid by
synchronizing demand to the availability of renewable energy.
Our field research shows that residential home retrofits that involve building shell and HVAC
replacements tend to save the most energy during the late afternoon and early evening hours, especially
during the warmer months of the year. Likewise, Locational Marginal Price—the cost of delivering energy
at the nodal level of the grid—also tends to increase in the late afternoon and evening and is prone to
spikes during the warmer months of the year. Using a standard set of methods to calculate savings, based
on constructing a counterfactual consumption value using weather as a predictor of demand, allows us to
measure the impact of EE on an hourly basis in a consistent fashion. These same methods also allow us to
forecast demand to provide EE developers with targeting tools to maximize potential savings.
We calculated the extent to which LMP could be forecasted based on the time of week and
temperature. For the most part, the cost of delivering energy is highly correlated with the time of week
2019 International Energy Program Evaluation Conference, Denver, CO 11
and the outdoor temperature. While these factors did not entirely predict the spikes in price that
accompanied certain events (such as a really hot day), the residuals of these hours—the quantity of the
price increase unexplained by the base model—create an opportunity for synergy between EE and
demand response. Where significant curtailments of energy consumption are required and where price
spikes are unrelated to weather conditions, demand response can complement EE in providing temporary
relief to the grid. We suggest that a valuation framework based on the concept of a CarbonWatt can
include all DERs within the same procurement. Where we are single-mindedly focused on reducing the
difference between demand and renewable supply, we can provide the kind of focused policy
environment that allows new technologies and business models to transform the grid.

References

Aslin, Richard. “PG&E’s Distribution Resource Plan: Opportunities and Challenges for Energy Efficiency”.
ACEEE National Conference on Energy Efficiency as a Resource. September 21, 2015.

Borgeson, Sam, A.M. Scheer, R. Kasman. “Energy Efficiency Program Targeting: Using AMI Data Analysis
to Improve At-the-Meter Savings for Small and Medium Businesses.” 2018

California Public Utilities Commission. “Standard Practice Manual”. 2001.


https://www.cpuc.ca.gov/uploadedFiles/CPUC_Public_Website/Content/Utilities_and_Industrie
s/Energy_-_Electricity_and_Natural_Gas/CPUC_STANDARD_PRACTICE_MANUAL.pdf

Hall, Jamie, et. al. “Locational and Temporal Values of Energy Efficiency and other DERs to Transmission
and Distribution Systems.” ACEEE Summer Study on Energy Efficiency in Buildings. 2018.

Scheer, Adam, S. Borgeson, K. Rosendo, et. al. “Customer Targeting for Residential Energy Efficiency
Programs: Enhancing Electricity Savings at the Meter.” 2017. S. Borgeson, A.M. Scheer, R. Kasman.
“Energy Efficiency Program Targeting: Using AMI Data Analysis to Improve At-the-Meter Savings
for Small and Medium Businesses.” 2018

Trabish, Herman. “Have California's efforts to value distributed resources hit a roadblock?” Utility Dive.
March 21, 2017.

2019 International Energy Program Evaluation Conference, Denver, CO 12

Das könnte Ihnen auch gefallen