cpuc’s distribution forecasting working group planning impacts from distr… · a short version...
TRANSCRIPT
CPUC’s Distribution Forecasting Working Group
on Distributed Energy Resource Allocation
Mark Quan
Itron Inc.
October 8, 2019
A SHORT VERSION OF HISTORY
“On August 14, 2014, the Commission opened Rulemaking (R.) 14-08-013 in order to establish
policies, procedures, and rules to guide California Investor-owned Utilities (IOUs) in developing their
Distribution Resource Plan (DRP) Proposals.”
This new code section requires the electrical corporations to file distribution resources plan proposals
by July 1, 2015. According to the Code, these plan proposals will “identify optimal locations for the
deployment of distributed resources.” It defines “distributed energy resources” as “distributed
renewable generation resources, energy efficiency, energy storage, electric vehicles, and demand
response technologies.”
The Code also requires the CPUC to “ review each distribution resources plan proposal submitted by
an electrical corporation and approve, or modify and approve, a distribution resources plan for the
corporation. The commission may modify any plan as appropriate to minimize overall system costs
and maximize ratepayer benefit from investments in distributed resources.”
(https://www.cpuc.ca.gov/General.aspx?id=5071)
AUTHORIZING DECISION
February 8, 2018:
D. 18-04-002
• “We direct Energy Division to develop a process and schedule for resolving the issues discussed
in this decision through the Distribution Forecasting Working Group.”
• The two main unresolved issues are the “DER forecast disaggregation methodologies, and using
alternative scenarios for resource planning purposes.”
ORDER …
• “The IOUs shall vet disaggregation methods through the Growth Scenario Working Group and
incorporate best practices in their planning process”
Mandate
DER DISAGGREGATION
» Disaggregation is the process by which a system
level forecast of DER for a given IOU service
territory is allocated at the circuit level.
» The goal of disaggregation methodologies is to
identify, to the extent possible, where new DERs
will be adopted.
» DERs studied are:
• Photovoltaic Generation
• Electric Vehicles
• Additional Achievable Energy Efficiency
• Energy Storage
• Load Modifying Demand Response
System
(IEPR)
Disaggregation
Methods
Circuit
THE ISSUE: DATA GRANULARITY VS . . .
System Region Substation
Data Stability
Model Accuracy
Power of Economic DataHigh Low
High Low
High Low
Circuit
Items to ForecastFew Many
Data Granularity Challenge
Customer
Issue
Load Shape ImpactsLow High
FORECASTING SPECIALIZATION
PNM Resources Weather Impact Analysis | ‹#›
FORECASTING SYSTEM LOAD CHANGES» Top-Down System Forecast
» Disaggregation to obtain load impacts
OFFICIAL REPORTS
https://drpwg.org/growth-scenarios/
Distribution Forecasting Working Group
(DFWG )April 18, 2018 to June 13, 2018
RANGE OF METHODS
Complexity
Adoption
Models
Proportional
Allocation
Propensity
Models
• Key adoption data
• Time series data
• Develop S-Curve
model
• Forward Looking
• Key adoption
characteristics
• Forward and/or backward
looking
• Develop cross-sectional
model
• Known utility data
(e.g. class sales,
customers)
• Backward looking
HighLow
METHOD: PROPORTIONAL ALLOCATION
System
(kWh)
Circuit 1
(kWh)
Circuit 2
(kWh)
Circuit 1 kWh = System kWh x Circuit 1 Ratio
Circuit N kWh = System kWh x Circuit N Ratio
System kWh = Circuit 1 kWh + … + Circuit N kWh
Circuit 1 Ratio = Circuit 1 Customers
System Customers
Circuit N Ratio = Circuit N Customers
System Customers
. . .Circuit N
(kWh)
. . .
. . .
» Disaggregates based on utility data for the circuit (load, energy, or number of
customers)
» Refinements using sector or rate class data
METHOD: PROPENSITY MODELS
Circuit 1 Score = f(Customers1, Income1, …)
Circuit N Score = f(CustomersN, IncomeN, …)
. . .
Circuit 1 Ratio = Circuit 1 Score
Total Score
Circuit N Ratio = Circuit N Score
. . .
Total Score = Circuit 1 Score + … + Circuit N Score
Total Score
» Disaggregates based on propensity score
» The scores (models) may be based on regression, machine learning or any
other method using cross section data with key identifying variables that
correlate with adoption.
METHOD: ADOPTION MODELS
Circuit 1 Adoption
Circuit 2 Adoption
Total Adoption
Time
Time
Time
Ad
op
tio
n U
nits
Ad
op
tio
n U
nits
Ad
op
tio
n U
nits
Circuit 1 Ratiot = Circuit 1 Adoptiont
Total Adoptiont
Circuit N Ratiot = Circuit 2 Adoptiont
. . .
Total Adoptiont
» Bottom-up adoption forecast is used
to develop ratios.
» S-Curve models are used to capture
adoption through time.
» S-Curve models need saturation and
adoption rate parameters.
RANGE OF METHODS
Complexity
Adoption
Models
Proportional
Allocation
Propensity
Models
• Photovoltaics
• Electric Vehicles• Electric Vehicles
• Load Modifying
Demand Resource
• Additionally
Achievable Energy
Efficiency
• Energy Storage
• Load Modifying
Demand Resource
HighLow
QUALITATIVE RISK
» Uncertainty: Relative uncertainty is defined as the range of possible
outcomes within DER technologies. A High rating indicates a relatively wide
range of possible outcomes. A Low rating indicates a relatively narrow range
of possible outcomes.
» Impact: Expected Impact is defined as the relative size of one DER against
another for planning purposes.
» Risk: Risk combines uncertainty and expected impact and should guide the
relative level of attention that different technologies receive in the planning
process.
Risk = f(Impact , Uncertainty)
UNCERTAINTY AND RISK
CONCLUSION
» Disaggregation Methods. “the DFWG vetted each
method and found that they are reasonable for
disaggregating the IEPR DER forecasts considering
the state of each of the DER technologies and the
available data”
» Uncertainty: “the DFWG recommends that these
qualifications be used to guide, not dictate, future
analysis and modeling efforts”
» Final Report submitted July 2, 2018.
Final Note: The Commission directed the IOUs to annually hold a one-day workshop to
present and discuss updates to their disaggregation methods. The Commission did not
determine that a longer working group process would be necessary unless there was a
major policy change.
TECHNOLOGY LEVEL SUMMARIES
PHOTOVOLTAICS
PNM Resources Weather Impact Analysis | ‹#›
Allocation Method
• Adoption Model (BASS Model or S-Curve Model) at the zip code level.
• Proportional Allocation method to Circuit level.
Key Issues/Uncertainty:
Method: Good quality geographic data
Shapes: Good quality load shapes from utility and 3rd party sources
Lumpiness: Timing of large project beyond interconnection queue is
difficult.
Impact: Size of market may be large.
ELECTRIC VEHICLES
PNM Resources Weather Impact Analysis | ‹#›
Allocation Method
• S-Curve Adoption Model (1 IOU) driven by a propensity model.
• Propensity Model (2 IOUs) at the zip code level.
• Proportional Allocation method to Circuit level.
Key Issues/Uncertainty:
Method: Early adoption makes identifying customer choice
characteristics difficult for the mass market.
Shapes: Charging patterns are evolving. Home charging versus station
charging will look different.
Lumpiness: Location and timing of commercial charging stations is difficult.
Impact: Size of market is expected to be large in the future.
ADDITIONAL ACHIEVABLE ENERGY EFFICIENCY
PNM Resources Weather Impact Analysis | ‹#›
Allocation Method
• Proportional Allocation using sector (class) load at the circuit level.
Key Issues/Uncertainty:
Method: Location is of EE is known for more than 50% of past EE
programs.
Shapes: Lack of recent EE shape data studies.
Lumpiness: Timing and location of large EE project is impossible to
forecast accurately.
Impact: Large amount of AAEE forecast.
ENERGY STORAGE
PNM Resources Weather Impact Analysis | ‹#›
Allocation Method
• Proportional Allocation method
• Allocate residential ES based on PV adoption (2 IOUs).
• Allocate non-residential ES based on load (2 IOUs).
• Allocate all ES based on load (1 IOU).
Key Issues/Uncertainty:
Method: Very early adoption phase. Lack of adoption data makes
modeling difficult.
Shapes: Operating profiles vary among customers based on customer
objectives.
Lumpiness: Timing and location of large EE project is impossible to
forecast accurately.
Impact: Very small amounts of ES in the forecast.
LOAD MODIFYING DEMAND RESPONSE
PNM Resources Weather Impact Analysis | ‹#›
Allocation Method
• Proportional Allocation based on eligible customers (1 IOU)
• Propensity based on regression trend model (1 IOU)
• Customer level propensity model for nonparticipants (1 IOU)
Key Issues/Uncertainty:
Method: Opt-out TOU rates create large uncertainty in customer
behavior.
Shapes: Impact profiles from existing studies appear reasonable.
Lumpiness: Clusters of adoption may introduce uncertainty.
Impact: Very small amounts of LMDR in the forecast.