Puerto Rico Low-to-Moderate Income Rooftop PV and Solar Savings Potential

Puerto Rico Low-to-Moderate

Income Rooftop PV and Solar

Savings Potential

Meghan Mooney and Katy Waechter

The National Renewable Energy Laboratory

(NREL)

December 17, 2020

NREL | 2

Why Does Rooftop Solar Potential for LMI Matter?

LMI households represent 50% of PR population and spend

disproportionately more of their income on energy

=> Dollar savings from solar likely more impactful

Solar has been disproportionately adopted by higher-income

households

=> Growing backlash against rooftop solar policies that appear

to favor the more affluent

Increasing interest in policy interventions to include LMI

households and create more equitable rooftop solar access

=> However, poor understanding of how much rooftop solar

LMI communities can accommodate may lead to ineffective

policies

NREL | 3

Building off previous research

Sigrin, Ben, and Mooney, Meghan. 2018. Rooftop Solar Technical Potential for Low-to-Moderate Income Households in the United States.

Golden, CO: National Renewable Energy Laboratory. NREL/TP-6A20- 70901.

https://www.nrel.gov/docs/fy18osti/70901.pdf.

Percent of LMI electrical consumption offsettable by rooftop

solar generation in the U.S. (Sigrin and Mooney, 2018)

• Estimates for LMI PV Rooftop technical

potential and solar savings potential exist for

the U.S.

• Sigrin and Mooney (2018) found:

– 43% of U.S. is LMI

– Nearly 1000 TWh of residential potential,

with 42% belonging to LMI buildings

– 60% of LMI potential belonging to non-

traditional buildings (i.e., renter and multi-

family)

• This study extends this work for Puerto Rico

NREL | 4

Key Questions for Technical Potential Analysis

1. How is rooftop solar potential

distributed geographically, by

income group, building type, and

tenure of the building occupants?

2. How much electrical consumption

can be offset by rooftop solar?

NREL | 5

Summary of methodology

• This study processes 2015-2017 LiDAR scans

(<0.35 nominal resolution) of Puerto Rico’s

building stock (96% of Puerto Rico building

stock).

• LiDAR data intersected with Census

demographics tables of household counts

by income, tenure, and building type. LiDAR

data does not allow direct observation of

the tenant’s attributes, thus a bootstrapping

method is used.

• A statistical model, trained on LiDAR tracts,

is used to impute building stock

characteristics (area, orientation, shading,

etc.) for 4% of building stock without

sufficient LiDAR data.

• Finally, simulate solar generation for each

roof plane using NREL PVWATTS and

aggregate at the tract and county level

NREL | 6

Assumptions for Building Suitability

PV System Characteristics Value for Flat Roofs Value for Tilted Roofs

Tilt 15 degrees Tilt of plane

Ratio of module area to suitable roof area 0.70 0.98

Azimuth 180 degrees (south facing) Midpoint of azimuth class

Module Power Density 183 W/m

Total system losses Varies (SAM defaults + individual surface % shading)

Inverter efficiency 96%

DC-to-AC ratio 1.2

Roof Physical Characteristics Description

Shading Measured shading for four seasons and required an average of 80% unshaded surface

Azimuth All possible azimuths

Tilt Average surface tilt <= 60 degrees

Minimum Area >= 1.62 m

(area required for a single solar panel)

Rooftop Suitability Assumptions

PV Performance Assumptions

NREL | 7

Puerto Rico Residential PV Solar Rooftop Potential

Residential PV rooftop technical potential by income group

• Annual residential solar potential is

24.6 TWh

• Roughly 4x of residential

electricity consumption

• LMI opportunity is 11.87 TWh,

nearly half (48%) of total annual

residential solar potential

• Average household potential is

19,883 kWh

– Potential is slightly greater for

higher incomes but still

considerable for even very low

income group (17,924

kWh/household)

Income Group

Households

(thousands)

Suitable

Buildings

(thousands)

Suitable

Module Area

(millions of

m2)

Capacity

Potential

(GW

)

Annual

Generation

Potential

(TWh/year)

Very Low

(0-30% AMI) 267.8 203.6 21.9 4.0 4.8

Low

(30-50% AMI) 151.2 129.1 13.5 2.5 3.0

Moderate

(50-80% AMI) 203.3 177.4 18.6 3.4 4.1

Middle

(80-120% AMI) 297.8 267.7 28.2 5.1 6.2

High

(>120% AMI) 317.1 279.5 29.6 5.4 6.5

All LMI Buildings 622.3 510.1 54.0 9.8 11.9

All Residential

Buildings 1,237.2 1,057.3 111.8 20.4 24.6

NREL | 8

Total LMI Solar Potential By Tract

Census tract LMI technical potential (GWh)

• Highest potential of LMI in densely populated areas with more building stock, though lots of

potential throughout the territory in more rural communities.

NREL | 9

Solar Potential By Income, Building Type and Tenure

Generation potential by income group & building type

• In Puerto Rico, 50% of the LMI potential is on on renter-occupied and multi-family buildings, with the

highest percentages among the lowest income group (58%).

• Though less than in the U.S. (60%), non-traditional types of buildings (i.e. multi-family and renter), are an

important factor. Puerto Rico has 50% of potential in non-traditional types. Improving LMI solar access in

Puerto Rico will likely require novel deployment models (e.g. shared solar).

Generation potential by income group & building tenure

NREL | 10

We Can Offset All Residential Electricity

Consumption With Residential Rooftop Solar in PR

Comparison of fraction of consumption met by generation

potential for single-family owner-occupied buildings alone, by

income group

Technical feasibility of matching residential electric

consumption with rooftop solar, by income group

Solar generation can technically meet all Puerto Rico electrical

consumption for all income groups

NOTE: The residential electric consumption reported here are from PREPA in 2018. Due to Hurricane Maria, it is estimated that 5% of PREPA

residential customers were still without power in April, 2018.

NREL | 11

Offsetable Consumption is High No Matter How You

Slice It

Percent of LMI electrical consumption offsettable by rooftop solar generation

% of Offsetable Electric

Consumption Offsetable by

Rooftop PV – LMI Single-Family

Owner-Occupied Buildings

% of Offsetable Electric

Consumption Offsetable by

Rooftop PV – All LMI Buildings

NREL | 12

High Solar/ Low Load: The Case of Puerto Rico

Rooftop Solar Potential

• High Solar:

– The average annual GHI (5.89 kWh/m

/day) in Puerto Rico is 22% greater than the average

US GHI

– In total, PR has more residential buildings to households. This contributes to higher technical

potentials per household electric consumptions.

• Low Load:

– PR has a significantly lower per capita electric consumption compared to U.S. (4,665 kWh vs

12,900 kWh per household annually)

– Even if PR consumed electricity at the rate of the U.S., there would still be nearly 150% the

amount of rooftop potential than electric consumption for the entire residential electric

sector.

• The Result:

– On an aggregate, for all residential buildings, Puerto Rico has 425% more rooftop generation

potential than electric consumption. For LMI buildings only, Puerto Rico has 570% more

rooftop generation potential than electric consumption.

– Even if we assume that 50% of the LMI buildings in Puerto are structurally unsuitable for

rooftop PV, there would still be >2.5x the amount of rooftop potential than current

consumption.

NREL | 13

All data used in the study is publicly available

Puerto Rico LMI PV Rooftop Technical Potential and

Solar Savings Potential Dataset – Access tract and

county level data on residential rooftop potential and

solar savings potential, by income, building type, and

tenure.

https://data.nrel.gov/submissions/144

SolarForAll web application - Explore, download, and

intersect data in interactive web application; Updated

with PR data layers.

https://maps.nrel.gov/solar-for-all/

Screenshot of SolarForAll app

NREL | 14

All data used in the study is publicly available

PV Rooftop Database – Puerto Rico (PVRDB-PR) – Access the raw GIS

data of roof planes for all buildings in Puerto Rico. Available through

the DOE Open Energy Data Initiative (OEDI) and hosted on Amazon

Web Services

• OEDI:

https://data.openei.org/submissions/2862

• S3 Path: s3://oedi-data-lake/pv-rooftop-pr/developable-planes/

• Bucket Viewer:

https://data.openei.org/s3_viewer?bucket=oedi-

data-lake&prefix=pv-rooftop-pr%2F

• Documentation:

https://github.com/openEDI/documentation/blob/master/PVRO

OFTOPS_PR.md

Use the AWS CLI command to list the data:

Use the AWS CLI to download the data:

Screenshot of the PVRDB-PR GIS data

aws s3 sync s3://oedi-data-lake/pv-rooftop-pr/developable-planes/ pv-rooftop-pr --no-sign-request

aws s3 ls s3://oedi-data-lake/pv-rooftop-pr/developable-planes/ --no-sign-request

www.nrel.gov

Contact: Meghan.Mooney@nrel.gov | Katy.Waechter@nrel.gov

Puerto Rico PV Rooftop LMI Technical Potential and Solar Savings Potential Data

Set: https://data.nrel.gov/submissions/81

PVRDB-PR (GIS): s3://oedi-data-lake/pv-rooftop-pr/developable-planes/

Web App: https://maps.nrel.gov/solar-for-all

Thank you

This work was authored by the National Renewable Energy Laboratory, operated by Alliance for Sustainable

Energy, LLC, for the U.S. Department of Energy (DOE) under Contract No. DE-AC36-08GO28308. Funding

provided by the U.S. Department of Energy Office of Energy Efficiency and Renewable Energy Solar Energy

Technologies Office. The views expressed in the article do not necessarily represent the views of the DOE or

the U.S. Government. The U.S. Government retains and the publisher, by accepting the article for

publication, acknowledges that the U.S. Government retains a nonexclusive, paid-up, irrevocable, worldwide

license to publish or reproduce the published form of this work, or allow others to do so, for U.S.

Government purposes.

NREL/PR-6A20-78756

Supplemental Material

NREL | 17

PR LMI Household Statistics

HH by Income

by Building Type

HH by Income

by Tenure

HH by Income

NREL | 18

PR Residential Rooftop Technical Potential by

Building Type and Tenure

NREL | 19

PR PV Rooftop Technical Potential

Results Compared to U.S.

Puerto Rico Mainland U.S.

All Residential LMI All * Residential ** LMI Buildings**

Suitable Buildings (millions)

1.34

1.06

78.68% of all

suitable

buildings

0.51

48.25% of all

residential

N/A 67 - 25.50

37.95% of all

residential

Households (millions)

1.24 0.62

50.30% of all

residential

- 117 - 49.80

42.60% of all

residential

27.5

20.35

73.97% of all

suitable

buildings

9.84

48.32% of all

residential

1,118.00 794

71.02% of all

buildings

329.40

41.49% of all

residential

TWh

33.18

24.55

74.00%of all

suitable

buildings

11.87

48.36% of all

residential

1,432.00 999.80

69.82% of all

buildings

415.90

41.60% of all

residential

* = Gagnon et al. 2016

** = Sigrin and Mooney, 2018

There are 1,343,874 suitable buildings in Puerto Rico of 1,500,106 total buildings in Puerto Rico.

Compared to the U.S. technical potential numbers, Puerto Rico is based off on higher precision LiDAR data (3-30cm) and updated solar

technology assumptions of solar technologies, such as solar panel power densities, more accurate (and stricter) percent shading assumptions

when estimating generation at the developable plane level, different assumptions related to minimum area requirements, different

assumptions related to panel orientation as candidates for rooftop PV suitability.

NREL | 20

Notes on PR Technical Potential Compared to the U.S.

Estimates

• Puerto is lower in latitude than the rest of the US; therefore, shading from adjacent trees is less impactful on the suitability of a rooftop.

– We assume an 80% of illumination criteria for the PR study, meaning that a plane must be illuminated for at least 80% of the year’s

daylight hours to be considered suitable for rooftop PV. In the Northern Hemisphere, there is a 2% increase in the solar elevation

angle for every 1-degree decrease in latitude. In PR’s lower latitude, its higher solar elevation angles creates shorter shadows and,

therefore, less shading, especially during the winter seasons, relative to its higher elevation states.

• Puerto Rico technical potential estimates are based on NREL’S PV ROOFTOP Model v2.0 updated assumptions, such as:

– Power Density is assumed to be 182 W/m

(compared to the 160 W/m

assumption in the 2016 U.S. study).

– North facing planes are not excluded for PR, however, they were excluded from the 2016 U.S. study. In PR, 13% of the technical

potential is a N, NE, NW facing plane.

– Minimum size requirement for a developable plane is set to 1.62 m

, which is the average size required for 1 250-Watt solar panel.

A building is considered suitable if it meets the other criteria and it has at least one plane large enough for a solar panel. In the 2016

U.S. Study, a suitable building needed a minimum 10m

. If we applied the same >= 10m

assumption to Puerto Rico, generation

would be reduced by ~4 TWh for the total building potential (all residential buildings).

– The shading assumption in this PR assessment was updated to apply percent shading directly at the developable plane level into the

System Advisor Model (SAM) when calculating generation potential. For the U.S. assessment, % shading was used to screen

potential planes, but it was not used directly at the plane level when processed in SAM to get the generation; instead, the SAM

default of 3% was applied. This new approach is more accurate than previous estimates, but it results in a lower kWh/kW estimate

for Puerto Rico compared to the U.S..

Taken together, the PR and the Gagnon et al. (2016) and Sigrin and Mooney (2018) U.S. estimates are not an apples-to-apples

comparison.

NREL | 21

Puerto Rico Residential PV rooftop technical potential

by Income Group, Building Type, and Tenure

Income Group Tenure Building Type

Buildings

(millions)

Suitable Area

(millions of m2)

Capacity (GW)

Annual

Generation (TWh)

Very Low

0-30% AMI

Owner-Occupied Single-Family

0.102 13.245 1.694 2.050

Owner-Occupied Multi-Family

0.009 2.338 0.316 0.380

Renter-Occupied Single-Family

0.070 8.987 1.154 1.396

Renter-Occupied Multi-Family

0.022 6.105 0.831 0.998

Low

30-50% AMI

Owner-Occupied Multi-Family

0.007 1.876 0.254 0.304

Owner-Occupied Single-Family

0.080 10.516 1.349 1.631

Renter-Occupied Multi-Family

0.010 2.463 0.335 0.402

Renter-Occupied Single-Family

0.032 4.099 0.525 0.634

Moderate

50-80% AMI

Owner-Occupied Multi-Family

0.010 2.661 0.360 0.432

Owner-Occupied Single-Family

0.116 15.195 1.953 2.363

Renter-Occupied Multi-Family

0.012 2.970 0.403 0.484

Renter-Occupied Single-Family

0.040 5.166 0.662 0.800

Middle

80-120% AMI

Owner-Occupied Multi-Family

0.016 4.044 0.549 0.658

Owner-Occupied Single-Family

0.203 27.364 3.538 4.280

Renter-Occupied Multi-Family

0.011 2.743 0.375 0.448

Renter-Occupied Single-Family

0.038 5.116 0.664 0.804

High

>120% AMI

Owner-Occupied Single-Family

0.223 30.786 3.962 4.775

Owner-O

ccupied Multi-Family

0.014 4.062 0.546 0.652

Renter-Occupied Single-Family

0.034 4.509 0.579 0.698

Renter-Occupied Multi-Family

0.008 2.254 0.306 0.365

NREL | 22

Validation of PV Rooftop Modeling

• Without ground truth data on existing rooftop systems in Puerto Rico, most of the model validation had to be centered

around assumptions and modeling standards, comparing results to previous study’s estimates.

– Results on a per building basis for a random set of buildings were evaluated along side Google Sunshot’s rooftop

potential estimates. Though variations exist between our set of underlying assumptions, our results were

consistently within a 5% of Google’s estimates.

– Potential estimates for PR were compared, to the degree possible, to previous U.S. estimates of potential (see slide

9).

– LMI estimates were compared to previous studies (e.g., LEAD).

• Azimuth and Slope Validation:

– A sample of developable planes stratified by azimuth and slope was randomly selected to validate roof plane

classification. Each plane was compared to raw aspect and slope models as well as manual verification of roof

orientation. Roof planes have an overall agreement of 92.06% with reference data. Among unique orientation

classes, omissions of east-oriented planes are correlated with commissions of south-oriented planes while there is

additional confusion between north-oriented planes and northwest-oriented planes, which may be the result of

signal scattering

NREL | 23

PR Electric Consumption

• 5.77 TWh of annual residential electricity consumption in PR

(PREPA, 2018 consumption)

– This estimate may not be complete of all consumption in

PR. As of April 2018, it is estimated that as much as 5% of

the PREPA customers still did not have electricity from

blackout caused by Hurricane Maria

(https://www.eia.gov/todayinenergy/detail.php?id=36832).

NREL | 24

Solar Savings Results

Comparison of fraction of consumption

met by generation potential for single-

family owner-occupied buildings alone,

by income group

Technical feasibility of matching

residential electric consumption with

rooftop solar, by income group

NREL | 25

Electric Consumption by Income, Building Type,

Tenure

Income Group

Building Type

Tenure

Annual Electric Consumption

(TWh)

Very Low

0-30% AMI

multi family

own 0.073

rent 0.282

single family

own 0.363

rent 0.113

Low

30-50% AMI

multi family

own 0.098

rent 0.134

single family

own 0.467

rent 0.084

Moderate

50-80% AMI

multi family

own 0.064

rent 0.056

single family

own 0.313

rent 0.045

Middle

80-120% AMI

multi family

own 0.064

rent 0.041

single family

own 0.295

rent 0.035

High

>120% AMI

multi family

own 0.626

rent 0.173

single family

own 2.307

rent 0.138

NREL | 26

Underlying Datasets

• NASA G-LiHT: 3cm LiDAR data collected in spring of 2017 (pre-hurricane)

• USGS 3DEP: <0.35 m nominal resolution LiDAR data collected in 2015 – used to fill in coverage

where G-LiHT is incomplete.

• HOTOSM building footprints – used to identify building footprints across island and overlaid with

LiDAR data.

• PREPA 2018 residential monthly electric consumption and billing, by municipality

• Low-income Energy Affordability Database (LEAD): Electric consumption by tenure, income, and

building class by municipality, used to disaggregate PREPA by income, building type and tenure.

• 2011-2015 American Community Survey (ACS) – Various ACS tables used to identify cross tabs of

number of households by income group, building type, and tenure at the Census Tract level

– B19001: Household Income in the Past 12 Months

– B25118: Tenure by Household Income in the Past 12 Months

– B25032: Tenure by Units in Structure

– B25124: Tenure by Household Size by Units in Structure

• 2013 RECS: Used for distributions of building size by type and income; used to generate

probabilities of residential buildings from building footprint information.