Meeting Notes

Load Shift Working Group (LSWG)

In person meeting: CPUC Golden Gate Room, 505 Van Ness Avenue

May 23, 2018

These notes summarize the Load Shift Working Group workshop facilitated by Gridworks. For a stakeholder list and presentations for this meeting (and for previous meetings), go to or contact Laura Wang at mailto:or more information.

Agenda

  1. Introductions
  2. ESDER 2 baselines and Baseline Analysis Working Group Findings – Kathryn Smith, SDG&E
  3. Analyzing ESDER 2 baselines and load increase – SilaKiliccote, SLAC and John Hernandez, PG&E
  4. Lunch
  5. Performance evaluation methodology for PDR-LSR – Eric Kim, CAISO
  6. Putting this into practice: lessons learned from XSP Baselines – Jonathan Burrows, PG&E
  7. Conclusions and next steps
  1. Introductions

Matthew Tisdale (Gridworks) opened the workshop. The following participants participated in person:

Meeting Notes

Load Shift Working Group (LSWG)

In person meeting: CPUC Golden Gate Room, 505 Van Ness Avenue

May 23, 2018

-Anja Gilbert, PG&E

-Brian Kooiman, Ohmconnect

-Shelly Lyser, CPUC ORA

-Nathanael Gonzalez, SCE

-Ola Huem, SCE

-Robert Anderson, Olivine

-Peter Worley, NRDC

-Christian Miller, NRDC

-Nora Sheriff, CLECA

-Erica Keating, SCE

-Anna Brockway, NRDC

-Sergio DueñasMelendez, Strategen

-Matthew Tisdale, Gridworks

-Laura Wang, Gridworks

-Eric Kim, CAISO

-LeslieWiloughby – SDGE

-Kathryn Smith, SDG&E

-SilaKillicote, SLAC

-John Hernandez, PG&E

-Fabienne Arnoud, PG&E

-Jonathan Burrows, PG&E

-Gil Wong, PG&E

-Jennifer Chamberlin, CPower

-Monica Schwebs, AMS

-Cherish Balgos, SCE

Meeting Notes

Load Shift Working Group (LSWG)

In person meeting: CPUC Golden Gate Room, 505 Van Ness Avenue

May 23, 2018

The following participants participated over the phone:

Meeting Notes

Load Shift Working Group (LSWG)

In person meeting: CPUC Golden Gate Room, 505 Van Ness Avenue

May 23, 2018

-Mary Ann Piette, LBL

-Mlee – Smarter Grid Solutions

-Doug Karpa, Clean Coalition

-Craig Sherman, SMUD

-Helena Oh, CPUC ORA

-Eric Woychik, Strategy Integration

-Brian Im, EnergyHub

-Ben Hertz-Shargel, EnergyHub

-Pramod Kulkarni, Customized Energy Solutions

-Julie McNamara, Union of Concerned Scientists

-Peter Alstone, LBNL

-Henry Richardson, WattTime

-Michael Volpe, PG&E

-Peter Schwartz, LBNL

-Elizabeth Ingram

-Malcolm Ainspan, NRG Curtailment Solutions

-Michael Lee

-Paul Nelson, CLECA

Meeting Notes

Load Shift Working Group (LSWG)

In person meeting: CPUC Golden Gate Room, 505 Van Ness Avenue

May 23, 2018

-

  1. ESDER 2 baselines and Baseline Analysis Working Group Findings – Kathryn Smith, SDG&E

Kathryn Smith (SDG&E) presented on conclusions from the Baseline Analysis Working Group (BAWG).The BAWG was tasked with the development of additional baselines (to supplement the 10 in 10 used by CAISO) that may work for additional load and customer types. Three new baselines were approved by CAISO in 2017 and once they are approved by FERC they are estimated to be implemented in the Fall of 2018. In addition, a new change was to require Scheduling Coordinators to calculate baselines and submit those results to CAISO.

From BAWG Final Report

The presentation discussed the types of settlement methods used (day-matching, weather-matching, and use of control groups), use of adjustment factors, and differences between different customers.

Discussion:

-Day matching: Participants discussed day-matching methodology (e.g., 10 in 10, 3 in 5, etc.) which takes an average of the most recent “similar” (e.g., weekend (including holidays) vs. weekday) days, removing event days, without explicit screening for outliers. In the current 10 in 10 methodology, there is a look-back of 45 days to find the 10 most recent similar days. If 10 days cannot be found, a minimum of 5 days are needed. If the minimum of 5 days cannot be found, then the methodology takes the average of 5 event days where load is highest during those event hours. In this last option, actual meter data is used. Participants noted that it becomes incrementally more difficult to demonstrate changes in load with each successor in methodology. Participants noted that this methodology may not be best suited when there are cases of frequent dispatch.

  1. Discussion on the study data: Data from all three utilities’ programs were used for this study (residential customers:AC cycling program customers from all 3 IOUS; agricultural customers: SCE’s agriculture program; BIP customers: participating customers from all 3 IOUs). The data set goes back at least 2 years.

-Adjustment factors: Kathryn discussed that an adjustment factor (a ratio based on recent energy use, applied on the event day itself), is applied to adjust for the difference between energy use on the averaged days and the current event day. It has been shown that using adjustment factors increase the accuracy of baselines. New baselines that will be in operation this November 2018 may use different caps (e.g., 40% cap). With caps higher than 40%, the study noted increased variability in results, affecting precision. Study results also showed that adjustments pre-and post- event, with a 2 hour buffer on each end, were the most accurate.

  1. Participants discussed retail vs. wholesale level alignment issues. It was noted that retail contracts allow up to a 40% day-of adjustment, which may cause “operational heartburn” when settling between the retail vs. wholesale level if a 20% cap is used at the wholesale level. In addition, retail baselines are calculated at the site level and aggregated up, while wholesale baselines are calculated at the aggregated resource level. Ohmconnect noted that, depending on the customer class, customers can still exceed the cap – for example, at the residential level, AC could be a main driver of load, and on hot days running the AC could lead to high spikes.
  2. It was noted that post-event adjustment factors are fairly new in CA, but have been used in other states. The Pacific Northwest states (accounting for cold weather days) and Ontario were suggested as examples.

-Weather matching baselines: in weather-matching, the baseline is calculated over a 90 day lookback using similar days with closest maximum temperature to the maximum temperature of the event day. Average energy use over 4 baseline days is calculated for each hour of the event.

-Control groups: Control groups compare customers enrolled in a program but are not participating in that particular event.

  1. Participants discussed that using a control group is more difficult for pay-for-performance events whereby customers are incentivized to participate, and may be more suited for residential customers. Also, the minimum size of a control group (150 participants minimum) is difficult to identify for larger C&I customers.
  2. DRPs are required to participate in control group validation, with periodic updates to account for seasonal changes or population changes, submitted to CAISO for approval. CAISO conducts two checks, one initial check to make sure the evidence is submitted properly, and a second after baselines are calculated. DRPs can use registered control groups across various resources after they pass the validation check.

-New baseline asymmetry/symmetry: It was noted that the new baseline calculation for 10 in 10is technically more asymmetric (upper bound 1.4, lower bound 1/1.4), though based on the Nexant analysis, this calculation leads to greater accuracy, which makes the calculation more “symmetric.”

-Kathryn highlighted the strengths of the new developed baselines and identified potential concerns with estimating load increases using these methods.

  1. If a day exhibits both load increase and load decrease, it can be difficult to calculate baselines and identify sufficient adjustment windows.
  2. SDG&E noted that these baseline analyses were tested on high-price days where load decrease is needed, and were not tested on low-price or negative-price days. Future studies may need to look at a larger number of days to make sure the baseline methodology still performs well on negative pricing days.
  3. The BAWG recommended that, in instances of frequent dispatch where it’s difficult to identify enough similar non-event days, day-matching baseline use is not recommended.
  1. Analyzing ESDER 2 baselines and load increase – SilaKiliccote, SLAC and John Hernandez, PG&E

John Hernandez (PG&E) introduced the purpose of the PG&E-commissioned SLAC study. The research was commissioned to help stakeholders better understand deficiencies (if any) with the existing and proposed methods under CAISO’s ESDER Phase 2 baselines in order to propose ways to improve these methodologies. It is examining ESDER 2 baselines in the context of load curtailment, load consumption, frequent dispatch, and issues related to the premise vs. devise participating in DR using the ESDER 2 baselines. SilaKiliccote (SLAC) presented on the study that started this year.

This study extends a 2009 study conducted of ISO baselines across the country. Two major outcomes of this study included: 1) the study showed that customers can be categorized by weather sensitivity and load variability, and the most appropriate baseline can be chosen for them; and 2) using an adjustment factor increases the accuracy of the baseline. Sila introduced the baselining methods tested in the study and suggested that additional baselines can be added to the test easily with modifications to the code. WG participants are welcome to suggest additional analyses that may be useful for the LSWG.

Sila highlighted future research questions, including: should clustering (pre- or post- processing) be tested? Clustered by load shape, % error, etc.? Are these weather dependent? Should the study pre-assign some baseline methods based on customer capabilities or characteristics?

Discussion:

-The WG discussed how this study may inform discussions on more frequent dispatch of short-term event periods. SLAC explained that it may be out of scope for this study given lack of available 15-minute meter data (mostly hourly data).

-The WG discussed potential issues with gaming for customers enrolled in 2 different programs (i.e., retail and wholesale) - whether it’s technically possible, financially beneficial, and what real-world potential examples may exist.

-The WG discussed that device-specific baselines (i.e., CAISO’s Meter Generating Output methodology) are important to understand, with regards to capability and availability. SLAC noted that an additional field study assessment specifically on device-specific baseline data may be useful.

-The WG asked whether these baselines may be biased in a particular direction (e.g., load reduction vs. load increase). SLAC noted that the means of measuring accuracy and bias (% error) may be biased towards load increase, as there is a lower limit (0) but not upper limit. SLAC noted that application of baselines in both directions is something the study can look into.

-SLAC noted the study is not looking at dual participation, only baseline representation.

Next steps: Some members of the WG would like to meet as a subgroup and discuss suggestions on behalf of the LSWG to SLAC with regards to the study scope and potential next steps. These WG members were identified as:

Meeting Notes

Load Shift Working Group (LSWG)

In person meeting: CPUC Golden Gate Room, 505 Van Ness Avenue

May 23, 2018

-Robert Anderson

-Brian Kooiman

-Shelley Lyser

-Eric Woychik

-Jennifer Chamberlain

-Sergio Melendez

-Peter Alstone

-Cherish Balgos

-John Hernandez

Meeting Notes

Load Shift Working Group (LSWG)

In person meeting: CPUC Golden Gate Room, 505 Van Ness Avenue

May 23, 2018

Gridworks will host a call Monday, July 2, 10am – 11am for WG members interested in discussing these recommendations further. For those who cannot make the call, written comments are asked by Tuesday, July 3 EOB.

  1. Lunch
  2. Performance evaluation methodology for PDR-LSR – Eric Kim, CAISO

Eric Kim (CAISO) presented on PDR-LSR characteristics and performance evaluation. Eric identified how CAISO currently determines the performance value and identifies “typical use” of load shift as its defined through MGO (metered generator output), which currently allows BTM storage devices discharging only to provide load curtailment. The current PDR-LSR proposal separately calculates curtailment and consumption using 2 resource IDs, as a mirrored use of the MGO concept. Similarly to calculating baselines in load shift, MGO currently uses 10 in 10 typical use calculations to determine the performance value of load shift. CAISO looks at 10 non-event-like days, specific to a 15 minute interval of the event (“event days” are considered as either a dispatch or outage in the ISO market), to apply an adjustment to the storage device. Typical use value is capped a below/above 0, depending on consumption/curtailment resource ID. Calculations of typical use are not necessary if load shift is applied to the premise level. CAISO noted that this set up is specific to battery storage technology, and after studying how this set up functions in the wholesale markets (expected Fall 2019 launch date), it may inform a more technology-neutral model.

Discussion

-Some WG members noted that consumption on the wholesale side in response to an event is affected by charge (retail side) – should and how is this incorporated into baseline calculations?

-WG members discussed whether this method is the right method to use for frequent dispatch events, or whether an alternative baseline method should be used. Without enough qualifying day-matching days (look-back of 45 days), the incremental value for frequent dispatch resources amounts to 0. WG members also asked whether, with enough frequent dispatch, if that becomes the new load profile of the customer.

-The WG noted additional considerations necessary to develop a technology-neutral product, including 1) duration of the technology-neutral response to a market signal (is a longer duration, e.g., 1 hour, needed for non-storage products?), 2) how this product may be designed to solve for grid needs; and 3) metering granularity (15-minute meter data may not exist everywhere). One participant also noted that the product design should allow it to take advantage of low prices as well as negative prices. designing a product, also looking at low prices Another participant questioned how the WG would define similarity for a “take” product, given that the end technology “acts” in a certain way, which may make it more difficult in matching “non-event days”.

-One participant noted that design limitations in calculating frequent dispatch baselines should not be a limitation of use. Frequent dispatch issues may also be addressed with program limitations, similar to other retail programs (e.g., bidding 6 hours per event, 5 days per month, etc.).

  1. Putting this into practice: lessons learned from XSP Baselines – Jonathan Burrows, PG&E

Jonathan Burrows (PG&E) presented on how the excess supply pilot (XSP) conducted performance measurement. Performance calculations compare event load to typical usage, using whole premise meter data. The same approach is used for both load increase and load reduction At the time of the pilot, ISO type 1 (10 in 10) was the approved baseline methodology for wholesale DR participation, so that methodology was used. Actual settlement is determined using metered data. Overall, XSP has had success using in the 10 in 10 baseline, without encountering any issues “unique” to load increase or customer type. The pilot may look at the other baseline methodologies that were approved in ESDER Phase 2.

Discussion:

-PG&E conducts settlement separately from the retail bill for program participants. For the XSP pilot, it was noted that if participants’ TOU demand charge increased due to participation in XSP, PG&E would calculate what the demand charge would have been without the event, and paid for the incremental increase in demand charge – this method only compensates for coincident peak demand charge. Olivine used sub-meter data to ensure the assets controlled are the ones causing the coincident peak demand charge.

-PG&E’s XSP team currently works with the distribution planning team to make sure that participation of a party doesn’t overload the circuit. Each customer’s parameters of participation and load profile are sent to the interconnection team and compared to existing Rule 21 studies of that circuit, or a study may be conducted.

  1. Themes of Discussion from Baseline Discussion: More discussion will be needed on the following topics. These issues do not just affect baselines for load TAKE but also load SHED.
  1. Frequent vs. Infrequent Dispatch
  2. Background: With new technologies such as batteries providing load response, DR is moving from a product that is infrequently dispatched to one that is dispatched more frequently.
  3. Questions:
  4. What constitutes frequent use?
  5. Can baselines still capture incrementality with frequent use?
  1. Devise vs. Premise Participation
  2. Background: For some load it may be beneficial to only have a device participate, for other load it may be beneficial to have the premise participate.
  3. Question:
  4. How do we design baselines so that either the devise’s participation in a DR event does not impact the settlement associated with the premise participating in DR or visa versa?
  1. Retail and Wholesale Participation
  2. Background: There are both retail and wholesale DR settlement methods.
  3. WHOLESALE: The baseline is settled in aggregate for the performance of energy.
  4. RETAIL: Baseline is typically settled at the individual resource level for the performance of energy. However, in some cases like the Capacity Bidding Program, the retail capacity is tied to an energy baseline (CBP).
  5. Questions:
  6. How do we design programs that do not provide payment for the same capacity between retail and wholesale?
  7. How do we design programs so that consumption in the wholesale side in response to an event on the retail side does not affect the baseline?
  1. Participation in both TAKE and SHED services
  2. Background: With new models of demand response you could have a resource that provides both TAKE and SHED services.
  3. Question:
  4. How do we develop a baseline that does not introduce more bias in estimating the typical use when the resource provides TAKE and SHED services?
  5. How can CAISO integrate a DR resource that provides TAKE and SHED that does not result in conflicting dispatch signals?
  1. Technology Impacts on a Baseline
  2. Background: Different technologies may have different abilities or characteristics that influence the development of a baseline.
  3. Questions:
  4. How can we develop a baseline for storage so that its actions taken in previous intervals do not introduce bias into the baseline?
  5. How can we develop a baseline for a device that can move (i.e. EVs can move from the EVSE charging station to the premise)?
  6. Should each device have its own baseline?
  1. Conclusions and next steps

-The WG will adjust the schedule to accommodate WG members participating in multiple RA and DR proceedings.