Updating Californias Weather Files Introducing CALEE 2018 Joe

Updating California’s Weather Files: Introducing CALEE 2018 Joe Huang, White Box Technologies Rick Ridge, Ridge & Associates Brian Arthur Smith, Pacific Gas and Electric Company 12/2/2020 July 12, 2019 1

Agenda • • • What is Weather Normalization? Motivation: Why Update the Weather Files? Regulatory Engagement: Outlining the Process Long-Term Weather Normalization: CA’s Existing Protocols Background on CZ 2010, the Older Standard Key Research Objectives for the CPUC and the CEC New “Typical Year” Weather Files for Long-term Weather Normalization Differences between CZ 2010, CZ 2022 and CALEE 2018 Additional Historical Weather Data How to Access the Updated Weather Files Data Sources Questions? 2

What is Weather Normalization? • Traditional long-term weather normalization process includes two steps: 1. Develop a model based on actual observations of historical load and weather; 2. Apply this model to a year with typical weather conditions to estimate the load reductions that would have occurred under normal weather conditions. • The rationale for normalizing is that expected savings over the life of the measure or program should be a function of normal weather conditions, not of the temperatures observed during a particular performance period. • Reported savings should be based on temperatures that are expected—based on long-term historical weather. • Reporting normalized savings allows us to compare savings year-over-year absent the influences of atypical weather. 3

Motivation: Why Update the Weather Files? 1. The last ten years have been observed to be among the hottest on record, and temperatures are forecasted to increase. 2. Climatic data are a critical input to the models used by the California Energy Commission (CEC) to develop the Title 24 building standards. 3. The California Public Utilities Commission (CPUC) and the California investor-owned utilities (IOUs) rely on climatic data, especially temperatures, in estimating savings for their energy efficiency (EE) programs. 4

Regulatory Engagement: Outlining the Process 1. Commission and IOU evaluation staffs convened a Program Coordination Group (“PCG-2”) in early 2018 to identify key study areas. Updating California’s weather files used for normalization was identified as a critical research need. 2. The weather study is included in the 2017 -2019 Energy Division & Program Administrator Energy Efficiency Evaluation, Measurement and Verification Plan Version 8 (published by the CPUC on December 31, 2018) as the sole Program Administrator-managed study that focuses on Normalized Metered Energy Consumption (NMEC). 3. The study team will engage with the CPUC’s NMEC Best Practices Study Team (Group D contract, SWB Consulting, lead contractor) to facilitate the transition to the CALEE files for 2018 impact evaluations. 5

Long-Term Weather Normalization: California’s Existing Protocols • The requirements and guidelines for long-term weather normalization have a long history in California: – 2004: California Evaluation Framework, pp. 97 -98. – 2006: California Energy Efficiency Evaluation Protocols: Technical, Methodological and Reporting Requirements for Evaluation Professionals, pp. 27 and 29. – 2013: Chapter 8 of the UMP: Whole-Building Retrofit with Consumption Data Analysis Evaluation Protocol, p. 8 -18. – 2014: ASHRAE Guideline 14 -2 -14: Measurement of Energy, Demand, and Water Savings, p. 92 – 2017: Chapter 24 of the UMP: Strategic Energy Management (SEM) Evaluation Protocol, p. 38. – 2018: Rulebook for Custom Program and Projects Based on Normalized Metered Energy Consumption (NMEC), p. 14. 6

Background on CZ 2010, the Older Standard • The files were created for 86 locations based on 12 years of data from 1988 through 2009 and were updates to the original RV 2 weather files created in the early 1980 s. • The data sources are NOAA’s Integrated Surface Database (ISD) and satellite-derived solar radiation data from Clean Power Research. • Since Time Dependent Valuations (TDVs) are dependent on statewide weather conditions, CZ 2010 uses the same “typical months” for all 86 locations. 7

Key Research Objectives for the CPUC and the CEC • CPUC and IOUs. To support Energy Efficiency: – Create a set of “typical year” weather files (CALEE 2018) for the same locations for use in CPUC-led EE impact evaluations and in other EE savings calculations that are based on normalized meter-based energy consumption (NMEC) and similar methodologies. – Release last five years (2014 -2018) of historical weather data of the same locations for use in calibrating models. • CEC. To support Title 24 Energy Code: – Update the CZ 2010 (and CZ 2016 and CZ 2019) weather files to CZ 2022, suitable for assessing compliance with the 2022 Title 24 update. – The update should use a similar methodology to CZ 2010, but build in some improvements where possible. 8

New “Typical Year” Weather Files for Long-term Weather Normalization 1. Files developed in collaboration with the CPUC and the IOUs for use in energy efficiency. • CALEE 2018 weather files for the same 117 stations using station-specific “typical months. ” • Using the CALEE 2018 data means that EE evaluators will obtain better estimates of net life-cycle savings that have been normalized for long-term weather that is unique to each station. 2. Files developed in collaboration with the CEC for Title 24 compliance. CZ 2022 weather files for 117 stations using statewide “typical months. ” CBECC-RES uses 16 stations (one for each climate zone) and CBECC-COM uses all 117 stations. 9

Differences between CZ 2010 , CZ 2022 and CALEE 2018 • The basic methodology for creating a “typical year” is the same as was used for CZ 2010, with a change in data source for the satellite-derived solar radiation (CPR in CZ 2010, NSRDB in CZ 2022 and CALEE 2018). • CZ 2010 and CZ 2022 both use statewide “typical months”, but CALEE 2018 uses station-specific “typical months”. • CZ 2010 and CALEE 2018 both use a 12 -year period of record ending a year or two before the time of creation (1998 -2009 and 2006 -2017 respectively), but CZ 2022 uses a 20 -year period of record from 1998 -2017. 10

Additional Historical Weather Data • The observed weather for all 117 stations will be updated on a monthly basis (within the first two weeks of the following month) by White Box Technologies starting in July 2019. These data will be used by analysts who are estimating savings based on actual weather for NMEC types of programs. • These models are based on actual weather and represent the first step in the two-step process of long-term weather normalization, described later. 11

How to Access the New Weather Files • CALEE 2018 —as well as the historical data and future monthly updates—will be posted on the www. calmac. org website (url: http: //www. calmac. org/weather. asp) • These data will be made available to the public for download at no cost. • Documentation of the methods used to create these data and other relevant information will also be posted on the www. calmac. org website in late July. 12

Today’s Focus • Our focus today is primarily on the development of the CALEE 2018 weather files for the 117 stations that used station-specific typical months, but mention will be made on their differences with the CZ 2022 weather files. • It is these data that can be used for the long-term weather normalization of energy savings 13

Data Source: Integrated Surface Database for weather station data Maintained by the National Center for Environmental Information (NCEI) (www. ncdc. noaa. gov/isd) 22, 747 stations, of which approx. 10, 000 are active. For 2016, there are 8, 140 stations with sufficient data to create usable Hourly weather files (2, 110 US, 275 Canadian, 5, 775 rest of the world) 14

Data Source: National Solar Radiation Database (NSRDB) for solar radiation data 30 -minute solar data derived from satellite observations on a 3 -km grid for any place in the Western Hemisphere from 60°N to 20° S from 1998 to 2017 15

Data Source: Climate Forecasting Models Climate reanalysis weather data • Using Climate Forecasting Models in a retrospective mode to derive climate conditions around the world going back several decades. E-P Annual mean 2002 -2008 MERRA • Generally operates on a ½ degree grid • Should be calibrated or “bias corrected” using observational weather data. ERA-1 • Can be used to fill in missing data or extend station weather data back through time. -5 -4 -3 -2 -1 0 1 mm/day 2 3 4 5 16

Method: Initial Weather Stations for CZ 2010 86 locations 17

Method: Expanded Weather Stations for Update CZ 2022/ CALEE 2018 117 locations 18

Method: Drill-Down on Files in the Update What’s contained in the hourly weather files? • All files come in three formats - *. epw, *. BINM, and *. FIN 4 text. Other formats such as *. TMY 2 and *TMY 3. csv are also available if requested. • *. FIN 4 is the archival format that’s best for looking at the underlying weather data. A free FIN 4 to. Excel. xlsx template file can be downloaded from the WBT web site to import FIN 4 files into Excel. • The *. FIN 4 format includes hourly records in SI of dry-bulb and dewpoint temperature, pressure, wind speed and direction, global horizontal and direct normal solar radiation, rainfall, sky cover, and solar angle; it also contains data flags for each parameter. • The *. epw format contains the same information, except for solar angle, although the data flags are not very accessible; it also contains calculated values for relative humidity and illuminances. • The *. BINM format contains hourly records in IP of all of the above except for solar angle and rainfall, and replaces dewpoint with wet-bulb temperature. it also contains calculated values for absolute humidity and enthalpy. 19

Method: How the Data are Processed How are the historical weather files processed ? • Producing the historical weather files represented 90% of the effort. • All the climatic elements except for the solar radiation are taken from the ISD. • Various methods are used to fill missing data depending on the climatic element and length of the data gap. • Extended data gaps are filled by interpolation from a nearby station if available, or using reanalysis data if not. • Solar radiation data are taken from the National Solar Radiation Database (NSRDB), which provides 30 -minute data for all of North America on a 3 -km grid from 1998 through 2017. • Time synchronization of the solar data is challenging because the satellite-derived solar are instantaneous values while the weather files require hourly cumulative amounts. 20

Method: Spatial Interpolation (1 of 3) Example of spatial interpolation for missing data 21

Method: Spatial Interpolation (2 of 3) Example of spatial interpolation for missing data 22

Method: Spatial Interpolation (3 of 3) Funny glitch found in the solar radiation along the coast during the summer months Average Monthly GHI and DNI for Arcata 2014, from NSRDB Average Monthly GHI and DNI for Arcata 2014, revised 23

Method: Production of Weather Files How are “typical year” weather files produced? • “Typical year” weather files are not synthetic data! • “Typical year” weather files are a concatenation of 12 real months taken from different years, each of which has been selected as the most representative of the long-term weather condition for that month. • Typical or most representative is defined as having the closest correlation (the Finkelstein-Shafer Statistic) to the long-term distribution of climate parameters for that month. • The choice of climate parameters and the weights given to each followed that used by NREL for the TMY 2 s, i. e. , max, min, and average daily dry-bulb and dewpoint temperatures, average daily wind speed, and total daily global horizontal and direct normal radiation. The weights are 40% for the 6 temperatures, 10% for wind speed, and 50% for the 2 solar values. 24

Method: The Finkelstein Shafer Statistic Brief explanation of the Finkelstein Shafer Statistic • The Finkelstein-Shafer (FS) statistic is the absolute area enclosed between the Cumulative Frequency Distribution (CFD) of a climate parameter for each month and the long-term CFD for the same month over all years. • The weighted sum of the FS statistic for all the parameter considered is the Cumulative FS (CFS) for that month. • The month with the smallest CFS is picked as the “typical month” for that month. January dry-bulb temperatures in Sacramento 1991 -2006 25

Method: Getting to the End Result Exploring various formulations of “typical year” weather • • • WBT originally proposed to add trends in the selection of “typical months”. California Energy Commission wanted a single statewide “typical year” for all locations to facilitate the calculation of Time-Dependent Valuations (TDVs). California Energy Commission also requested that trending not be used, and to look at different periods of record from as short as 7 years to as long as 20 years (all ending in 2017). California Energy Commission ultimately selected 20 years with statewide “typical months” (and no trending) for the CZ 2022 weather files. Since the intended use of the CALEE 2018 is for weather normalization, it is not necessary to use a single statewide “typical year” The suggested CALEE 2018 uses location-specific “typical months” and a 12 year period of record, same as what was used for the CZ 2010. 26

Method: Selection of Typical Months Identifying statewide “typical months” • A two-step process where the CFS for all locations are calculated and then added together to derive the cumulative CFS for all locations. • The year with the smallest cumulative CFS is selected as the statewide “typical month”. • The location CFS are populationweighted, so that the selected statewide “typical months” tend to reflect conditions in southern California. Colored coded location FSs for June, 15 year period with trending. 27

Method: Comparisons to Historical Record (1 of 8) How do the various formulations compare to each other and to the historical record? 28

Method: Comparison to the Historical Record (2 of 8) How do the various formulations compare to each other and the historical record? 29

Method: Comparison to the Historical Record (3 of 8) How do the various formulations compare to each other and the historical record? 30

Method: Comparison to the Historical Record (4 of 8) How do the various formulations compare to each other and the historical record? 31

Method: Comparison to the Historical Record (5 of 8) How do the various formulations compare to each other and the historical record? 32

Method: Comparison to the Historical Record (6 of 8) How do the various formulations compare to each other and the historical record? 33

Method: Comparison to the Historical Record (7 of 8) How do the various formulations compare to each other and the historical record? 34

Method: Comparison to the Historical Record (8 of 8) How do the various formulations compare to each other and the historical record? 35

General Observations (1 of 2) Observations about the CZ 2022 and CALEE 2018 files • California has been experiencing a generally warming trend over the past two decades. • The CZ 2010 weather files seem significantly out of step with current weather trends. • The CZ 2022, by using a 20 year period of record, seem to account for this trend in a mild way. • The CALEE 2018, by using a 12 year period of record, seems to be responding more quickly to this warming trend. • The use of a single “typical year” for the entire state causes some degradation of the data compared to the location-specific “typical year”, but does not produce major errors. • There is a slight brightening trend in the solar data, but the NSRDB shows significantly (10 -15%) more direct normal solar gain than did the CPR. 36

General Observations (2 of 2) General Comments about the new weather data • By combining weather data from three different sources, i. e. , the ISD, NSRDB, and MERRA, the project has been able to produce weather files of high resolution and unprecedented accuracy. • Both the development of the historical weather files and the methodology used to create “typical year” weather files were extensively vetted by consultants and staff of the CEC over a three month period. • Comments received from the CEC are that they are satisfied with the CZ 2022 weather files and intend to use them as the standard weather files for the 2022 version of Title 24. • The release of the last 5 years of historical data, and the planned maintenance of the data into the future will insure that all the weather data needs for utility energy efficiency programs will be met. • The 2018 and 2019 weather data will contain modeled solar until NREL releases their annual update of the NSRDB, at which time the modeled solar will be replaced by satellite-derived solar. 37

Adoption of CALEE 2018 • • • No formal adoption process NMEC PCG-2 NMEC Best Practices Team sponsored by CPUC Recommend adopting Joe Huang and NMEC Best Practices team available to assist in adoption 38

CZ 2019 • Currently, the 2019 Title 24 update will continue to use CZ 2019 files: https: //ww 2. energy. ca. gov/title 24/2016 standar ds/ACM_Supporting_Content/ 39

Questions, Comments and Suggestions? • Energy Savings Brian Arthur Smith Assistance Program Energy Efficiency Evaluation (ESA) Overview Pacific Gas and Electric Company B 2 SG@pge. com 415 -973 -1180 READ AND DELETE For best results with this template, use Power. Point 2003 40

Appendix 41

Future Weather § There has been increasing concern about Climate Change and whether to account for its effects in the design of new buildings. § There are several ways to get a handle on potential climate change, including trend analysis of the historical data, as well as using the results from GCMs to create future year weather files. § There are several methods to create future year weather files, including: § 1. 2. “morphing” a historical typical year” weather data. 3. Frequency analysis of GCM data combined with historical data to produce future year files without dynamic downscaling Downscaling GCM time series using a mesoscale simulation program. There is a growing list of resources and expertise that can be tapped to produce future year weather files and design conditions. 42

Future Weather A brief description of “morphing” Change in average daily temperatures by month. Change in average daily range (daily max – daily min) by month. 43

Future Weather Temperatures in January and July for Washington in 2020 by “morphing” an existing TMY 2 file 44

Future Weather A brief description of downscaling: Temperatures for Oakland 1995 -2100 by downscaling GCM results. 45

Future Weather General comments about Future Weather 1. Don’t get too fixated about what the models say. They are still just models. 2. Trend has been to combine the results from dozens of models to get a consensus of results. 3. Climate scientists have focused on overall trends and patterns; for building energy efficiency, it’s important to know what are the changes in the diurnal profile and/or the coincidence of temperature humidity, and solar gain? 4. Although large institutions with supercomputers can do dynamic downscaling of GCM results, is it worth doing or computational overkill? 5. Is there a need for future year weather files? 46