HomeMy WebLinkAboutStaff Report 14677
City of Palo Alto (ID # 14677)
Utilities Advisory Commission Staff Report
Meeting Date: 12/7/2022 Report Type: VII. NEW BUSINESS
City of Palo Alto Page 1
Title: Discussion of City of Palo Alto Utilities' Long-Term Electric Load
Forecast Through 2045
From: Director of Utilities
Lead Department: Utilities
EXECUTIVE SUMMARY
Staff prepared this report on the development of long-term electricity demand forecasts for the
Utilities Advisory Commission’s review and discussion; no action is requested at this time. Staff
developed three long-term electricity demand forecasts for the City of Palo Alto Utilities (CPAU)
that build off long-term demand trends while also incorporating the projected impacts of
industrial businesses leaving the City, data centers growing, building efficiency improving,
customers adding solar, increasing numbers of electric vehicles (EVs), and customers switching
from natural gas to electric appliances.
The purpose of these three forecasts is to create a realistic projection of energy demand based
on known trends while acknowledging the substantial uncertainty that exists around the
impacts of these trends in the future. These low, mid, and high long-term forecasts are updated
at a minimum every five years, as part of CPAU’s Integrated Resource Plan (IRP) development
process, and will be used as the basis for electric supply procurement decision-making. Staff will
continue to adjust these forecasts periodically, as actual changes in loads are observed
BACKGROUND
Staff updated the long-term electric load forecast to aid electricity supply planning and for use
in CPAU’s 2023 electric IRP. To coordinate with the state’s 2045 goal of powering all retail
electricity sold in California with renewable and zero-carbon resources, the forecast extends to
2045. The purpose of the load forecast is to create a realistic projection of future electricity
demand, incorporating the impacts of known trends while acknowledging the substantial
uncertainty around these trends given the time horizon of the forecast and the unprecedented
technology transitions that are currently underway.
This set of forecasts is the first step in designing a supply portfolio with the right fit for CPAU’s
electricity needs, understanding CPAU’s overall exposure to changing market price profiles, and
appropriately valuing supply resources.
4
Packet Pg. 63
City of Palo Alto Page 2
The range of uncertainty that exists around the three forecasts also will be factored into the
City’s IRP modeling for designing a supply portfolio that has sufficient optionality and flexibility,
and for appropriately valuing that optionality.
Staff will adjust these forecasts periodically on the basis of observed changes in load and usage
patterns, and adjust the City’s electricity procurement plans to reflect these changes as well. A
detailed long-term electricity forecast update is undertaken every five years at a minimum.
DISCUSSION
Staff develops CPAU’s load forecasts by starting with a linear regression based on past
consumption history, and then exogenously incorporating the nonlinear impacts of new trends,
which in this case include: new data centers planned in the next three years, the nonlinear
growth component of EV load, and load growth from building electrification. These three
subsets are modeled separately because staff anticipates that their future growth will far
exceed their recent historical growth, and therefore the linear regression analysis will not
adequately represent their future impact on total load.
Linear Regression Forecast Component
Staff developed the linear regression econometric forecast using a trend variable to capture
load reductions caused by industrial businesses leaving Palo Alto, building efficiency
improvements, and solar on homes and businesses.
The recession caused by COVID-19 required an additional ‘COVID-19 recession dummy
variable’1 to be superimposed on top of this linear load reduction trend variable, given the
exceptional nature of the pandemic. As shown in Figure 1 below, Palo Alto’s electricity
consumption is rebounding as the country exits the COVID-19 recession, and the pandemic’s
impacts on the City’s electricity demand are expected to largely disappear by the end of 2023.
1 A dummy variable is a variable used in regression analysis that can take a value ranging from 0 to 1, where the
values indicate the presence or absence of something. The COVID-19 recession dummy variable values are
depicted by the yellow bars in Figure 1, and they represent the relative magnitude of the pandemic-induced
economic recession each year.
4
Packet Pg. 64
City of Palo Alto Page 3
Figure 1. Graph of Long-term Linear Load Loss and COVID-19 Recession
In the linear regression, staff captured all of the above variables into a single trend variable,
which is assumed to continue until some point in the future. For the ‘mid’ forecast, staff has
assumed that this load loss trend will continue at the same pace until 2035, at which point
loads will stabilize. For the ‘high’ forecast, staff assumed substantial data center growth
combined with load loss stabilizing in 2030. For the ‘low’ forecast staff has modeled the current
load loss trend that began in 2004 continuing all the way through 2045, since this has been a
long-standing trend that includes behind-the-meter solar, customer and building code
efficiency improvements, and the departure of industrial businesses.
Table 1. Assumptions Concerning Long-term Linear Load Loss Trend
Linear Regression Assumptions Low Projection Mid Projection High Projection
Year when load decline from
deindustrialization, efficiency,
and solar is assumed to end
2045 2035 2030
Nonlinear Additional Load Growth Component
After defining the linear regression trends, staff exogenously added the nonlinear component
of load growth from additional data centers, electric vehicles, and building electrification. The
assumptions for the low, mid, and high cases are shown in Table 2 below. For both electric
vehicle adoption and building electrification rate, the high case assumes achieving a 71%
reduction in the City’s total CO2 emissions from 1990 levels by 2030. The low forecasts for both
electric vehicle adoption and building electrification rate assume a linear extrapolation of the
trends of the last three to five years. The mid forecasts for both electric vehicles and building
electrification are an aggressive but realistic projection. The additional data center load growth
shown in Table 2 is based on projections provided by customers for new data centers which
4
Packet Pg. 65
City of Palo Alto Page 4
are planned for the next three years, with the mid case assuming that 30% of these customer-
projected load increases are actually realized, while the high case assumes 100% realization of
the projected load increases.
Table 2. Additional Load Impacts, Including Linear and Nonlinear Components
2020 2045 projection
Additional Modeled Load Growth Actual Low Mid High
Additional Data Centers, GWh
-
0
(0%)
70
(8%)
230
(27%)
Electric Vehicles, GWh
10
(1%)
46
(5%)
129
(15%)
165
(19%)
Building Electrification, GWh
1
(0.1%)
16
(2%)
69
(8%)
91
(11%)
Total, GWh
11
(1%)
62
(7%)
268
(32%)
486
(57%)
Table 3. Assumed Growth Factors for Data Centers, Electric Vehicles, and Building
Electrification
Low Projection Mid Projection High Projection
Assumptions for Additional Loads 2020 2030 2045 2020 2030 2045 2020 2030 2045
Data Centers
% Additional Data Center Planned - - - - 30% 30% 100% 100% 100%
Electric Vehicles
% Residents with EVs 10% 21% 42% 10% 31% 61% 10% 44% 86%
% EVs of New Vehicles Purchased 30% 40% 62% 30% 50% 80% 30% 85% 100%
Building Electrification
% Single Family All-electric 1% 7% 26% 1% 10% 87% 1% 100% 100%
% Gas Packs Converted - - - - - 75% - 100% 100%
% School Sq ft Converted - - - - - 75% - 100% 100%
% Large Commercial Converted - - - - 5% 100% - 100% 100%
4
Packet Pg. 66
City of Palo Alto Page 5
Figure 2. Building Electrification Load Assumptions
Overall Forecasts Including Linear and Nonlinear Trends
The combined low, mid, and high load forecasts are shown in Error! Reference source not
found., including the total expected additional data center load, EV load, and building
electrification load. The EV load projections include both a linear component (capturing the
roughly linear growth in EV numbers over the past decade) as well as a nonlinear component
(projected future growth exceeding the trendline from the past decade). Neither the data
center load nor the building electrification load has components captured in the historical
linear regression, so these load projections are entirely nonlinear.
Figure 3. Low, Mid, and High Load Forecasts through 2045
4
Packet Pg. 67
City of Palo Alto Page 6
Changes in Seasonal and Hourly Usage Patterns
Staff is also exploring the changing patterns of electricity usage and have updated the hourly
models to encompass these hourly and seasonal trends (higher winter and nighttime loads),
behind-the-meter solar and batteries, and elevated temperatures from climate change (i.e.
more air conditioning and less heating load overall).
Approximately half of the projected building electrification load is expected to come from
heating homes and businesses, which will result in more usage in the winter and at night.
Electric vehicles used for commuting are also assumed to charge more in the evenings and at
night. Staff also expects solar installations at customer premises to continue at approximately
the current pace. Staff has integrated all of these changes in usage patterns into their hourly
forecast models and will continue to monitor them and adjust the models as needed.
RESOURCE IMPACT
The updated forecasts will be incorporated into long-term budgeting and will be used to design
the most cost-effective portfolio through the electric IRP.
POLICY IMPLICATIONS
This updated range of forecasts is consistent with the Utilities Strategic Plan, the Utilities
Electric Integrated Resource Plan, Sustainability Implementation Plans, and the City’s
Sustainability and Climate Action Plan (S/CAP).
ENVIRONMENTAL REVIEW
The UAC’s discussion of these electric load forecasts does not require California Environmental
Quality Act review, because it does not meet the definition of a project under Public Resources
Code Section 21065 and CEQA Guidelines Section 15378(b)(5), as an administrative
governmental activity which will not cause a direct or indirect physical change in the
environment.
NEXT STEPS
Staff will continue to refine these load projections as well as the hourly usage models and
incorporate them into the IRP modeling effort and electricity supply planning.
Attachments:
• Attachment A: Presentation
4
Packet Pg. 68
Electric Load Forecast
Discussion
Lena Perkins, PhD Senior Resource Planner
December 7, 2022
4.a
Packet Pg. 69
At
t
a
c
h
m
e
n
t
:
A
t
t
a
c
h
m
e
n
t
A
:
P
r
e
s
e
n
t
a
t
i
o
n
(
1
4
6
7
7
:
E
l
e
c
t
r
i
c
l
o
a
d
f
o
r
e
c
a
s
t
)
Figure 1. Graph of Long-term Linear Load Loss and COVID-19 Recession
4.a
Packet Pg. 70
At
t
a
c
h
m
e
n
t
:
A
t
t
a
c
h
m
e
n
t
A
:
P
r
e
s
e
n
t
a
t
i
o
n
(
1
4
6
7
7
:
E
l
e
c
t
r
i
c
l
o
a
d
f
o
r
e
c
a
s
t
)
Table 2. Additional Load Impacts, Including Linear and Nonlinear Components
2020 2045 projection
Additional Modeled Load Growth Actual Low Mid High
Additional Data Centers, GWh
-
0
(0%)
70
(8%)
230
(27%)
Electric Vehicles, GWh
10
(1%)
46
(5%)
129
(15%)
165
(19%)
Building Electrification, GWh
1
(0.1%)
16
(2%)
69
(8%)
91
(11%)
Total, GWh
11
(1%)
62
(7%)
268
(32%)
486
(57%)
4.a
Packet Pg. 71
At
t
a
c
h
m
e
n
t
:
A
t
t
a
c
h
m
e
n
t
A
:
P
r
e
s
e
n
t
a
t
i
o
n
(
1
4
6
7
7
:
E
l
e
c
t
r
i
c
l
o
a
d
f
o
r
e
c
a
s
t
)
Figure 2. Building Electrification Load Assumptions
4.a
Packet Pg. 72
At
t
a
c
h
m
e
n
t
:
A
t
t
a
c
h
m
e
n
t
A
:
P
r
e
s
e
n
t
a
t
i
o
n
(
1
4
6
7
7
:
E
l
e
c
t
r
i
c
l
o
a
d
f
o
r
e
c
a
s
t
)
Figure 3. Low, Mid, and High Load Forecasts through 2045
4.a
Packet Pg. 73
At
t
a
c
h
m
e
n
t
:
A
t
t
a
c
h
m
e
n
t
A
:
P
r
e
s
e
n
t
a
t
i
o
n
(
1
4
6
7
7
:
E
l
e
c
t
r
i
c
l
o
a
d
f
o
r
e
c
a
s
t
)