1 of 22

Increasing Renewable Access in the Philippines through AI

Increasing Renewable Access in the Philippines through AI

July 17, 2021

DISCLAIMER: This document is strictly private, confidential and personal to its recipients and should not be copied, distributed or reproduced in whole or in part, nor passed to any third party.

Omdena Philippines Chapter

2 of 22

PH ranks average in sustainable energy adoption�53 points in planning; 60 in incentives & regulatory support

Increasing Renewable Access in the Philippines through AI

July 17, 2021

:

Planning for renewable energy expansion indicator

Rank 1

Rank 51

Rank 131

Rank 1

Rank 96

Rank 130

Incentives and regulatory support indicator

Source: https://rise.esmap.org/analytics

3 of 22

In 2030, PH gov’t solar energy capacity �installation target to increase by 28K %

Increasing Renewable Access in the Philippines through AI

July 17, 2021

Source: https://www.doe.gov.ph/sites/default/files/pdf/nrep/nrep_books_021-087_re_plans_programs.pdf

Sector

Installed Capacity, (MW) as of 2010

Target Capacity Addition by

Total Capacity Addition (MW) 2011-2030

Total Installed Capacity by 2030

2015

2020

2025

2030

Geothermal

1,966.0

220.0

1,100.0

95.0

80.0

1,495.0

3,461.0

Hydro

3,400.0

341.3

3,161.0

1,891.8

0.0

5,394.1

8,724.1

Biomass

39.0

276.7

0.0

0.0

0.0

276.7

315.7

Wind

33.0

1,048

855.0

442.0

0.0

2,345.0

2,378.0

Solar

1.0

269.0

5.0

5.0

5.0

284.0

285.0

Ocean

0.0

0.0

35.5

35.0

0.0

70.5

70.5

Total

5,438.0

2,155.0

5,156.5

2,468.8

85.0

9,865.3

15,304.3

4 of 22

�������

Project Goals

  • Web App containing Map of the sites
  • GitHub repo with open source code
  • Curated dataset hosted in GDrive for open access

Learning Outcomes

  • Collect satellite images and extract relevant features
  • Prototype an ML model to for site identification
  • Curated project-based resource for computer vision and image processing for impact application

Increasing Renewable Access

in the Philippines through AI

5 of 22

Project Pipeline

Increasing Renewable Access in the Philippines through AI

July 17, 2021

Input

Data Preparation

Modelling

Output

Energy Demand

Create light vs no light mask computed using Philippines 2020 dataset

Estimate solar energy availability and consider where are the optimal geo-locations

Partition PH domain as Laguerre-Voronoi regions (DHS data)

Clean, aggregate installed electric capacity (power plant data)

Convert vector data to raster; re-classify

Multi-criteria weighted overlay analysis

K-Means

Clustering

Energy Availability

Night-time�Light �Density

Annual VNL 2 from Earth Observation Group

2017 PH National Demographic and Health Survey (DHS) Access

to Electricity��2020 PH List of Existing Power Plants��FB High Resolution Settlement Layer (2018 PH population data)

Satellite image of Solar Irradiance(GHI, DHI, DNI) and Environmental Temperature;��Satellite Image of WorldBank

6 of 22

Task 2 - Night-time Light Density

���

Increasing Renewable Access in the Philippines through AI

July 17, 2021

The nightlight data was obtained from Earth Observation Group. The dataset is called “Annual VNL 2”, a postprocessed version of VIIRS nighttime lights.

The satellite imagery captures the light emitted at night. It measures the radiance as nW/cm2/sr (nanowatt per square centimeter per steradian).

Annual VNL 2 has the advantage of being preprocessed for noise filtering, zeroing of the background with no detectable lightning, and outlier removal.

For this project, we used Philippine data for the year 2020.

7 of 22

Task 2 - Energy Demand

Increasing Renewable Access in the Philippines through AI

July 17, 2021

We determined settlements from FB’s high resolution settlement layer (HRSL) of high-energy demand using multi-criteria weighted overlay analysis.

We first converted the following source vector data to appropriate rasters:

We then re-classified the response of each input raster (night-time light, DHS, installed electric capacity) and obtained a total score based on site suitability (low night light, access to electricity, and installed electric capacity).

8 of 22

Task 2 - Energy Demand

Increasing Renewable Access in the Philippines through AI

July 17, 2021

We determined settlements from FB’s high resolution settlement layer (HRSL) of high-energy demand using multi-criteria weighted overlay analysis.

We first converted the following source vector data to appropriate rasters:

We then re-classified the response of each input raster (night-time light, DHS, installed electric capacity) and obtained a total score based on site suitability (low night light, access to electricity, and installed electric capacity).

9 of 22

Task 4 - Estimating Energy Availability

Increasing Renewable Access in the Philippines through AI

July 17, 2021

Using single diode model circuit we are, using norton’s theorem we are finding the total resistance of the circuit since we know the Vmax and Imax we wil get the total resistance.

10 of 22

Task 4 - Estimating Energy Availability

Increasing Renewable Access in the Philippines through AI

July 17, 2021

TOP 10 LOWEST PVOUT(JAP6-72-320/4BB) MUNICIPALITY IN THE PHILIPPINES

TOP 10 HIGHEST PVOUT(JAP6-72-320/4BB) MUNICIPALITY IN THE PHILIPPINES

11 of 22

Task 4 - Estimating Energy Availability

Increasing Renewable Access in the Philippines through AI

July 17, 2021

Single Diode Formula

Single Diode Formula:

Most of the barangay belong to the clusters with moderate and high estimated energy availability. This demonstrates how viable solar energy is as renewable energy source in the Philippines.

World Data Bank:

Values from the global atlas show that most of the barangays can be classed as having low and moderate estimated energy availability��Recommendations:

For future implementation of the similar concept or a continuation of this task, deep clustering algorithms along with tuning of the hyper parameters can be added to the methodology. Along with the validation of the computed estimated energy available for areas where similar data has already been collected.

World Data Bank

Cluster

Value (kWh)

No. of Barangays

Percentage

Low

117 - 1134.83

4333

10.33%

Moderate

1134.83 -�1313.50

16373

39.05%

High

1313.50 – 1487.80

14236

33.95%

Very High

1487.80 – 1752.10

6991

16.67%

Model

Silhouette Scores

Number of Clusters

K Means

0.54013

4

Gaussian Mixture

0.5396

4

DBScan

0.2747

19

Model

Silhouette Scores

Number of Clusters

K Means

0.5444

5

Gaussian Mixture

0.5419

5

DBScan

0.1494

20

Cluster

Value

No. of Barangays

Percentage

Very Low

117 – 1312.96

2577

6.15%

Low

1312.96 – 1395.98

12234

29.18%

Moderate

1395.98 – 1464.33

13012

31.03%

High

1464.33 – 1552.86

9691

23.11%

Very High

1552.86 – 1695.31

4419

10.53%

12 of 22

Task 4 - Estimating Energy Availability

Increasing Renewable Access in the Philippines through AI

July 17, 2021

Single Diode Formula

Municipality PVOut Clustering:

And based on the Clustering, it is also the same as the which is composed of 5 cluster, based on the value of PVOut of each municipality of the philippines.

Recommendation:

For the Lowest PVOut Area, we are suggesting if they can explore any type of renewable energy especially Wind Energy because mostly on the lowest PVOut area the weather there are cloudy, and low temperature which is possible have a high wind strength.

World Data Bank

13 of 22

Task 1&6 - Visualization & Integration

Increasing Renewable Access in the Philippines through AI

July 17, 2021

Omdena Philippines Chapter

14 of 22

Task 1&6 - Visualization & Integration

Increasing Renewable Access in the Philippines through AI

July 17, 2021

https://omdenaph-solar.vercel.app/ ��Built with React, Next.js, Vercel, and Mapbox.

15 of 22

Task 1&6 - Visualization & Integration

Increasing Renewable Access in the Philippines through AI

July 17, 2021

We started with plotting the different power plants and energy sources in the Philippines. This way we identified various municipalities that are still off-grid based on the 2020 Department of Energy data. These are marked with Diamonds

16 of 22

Task 1&6 - Visualization & Integration

Increasing Renewable Access in the Philippines through AI

July 17, 2021

Our Web App has 4 map modes in order to highlight different information in our App. Below the Map Mode Menu, you can see the municipalities ranked by their need for solar energy. They are also highlighted in the map by Dots (see 'Marker Labels' legend).

17 of 22

Task 1&6 - Visualization & Integration

Increasing Renewable Access in the Philippines through AI

July 17, 2021

Our suggested places and settlements (estimated by Task 2 & 4) are marked in Green Dots for Area and Blue Dots for more granular Settlements data. When you click on the Green Dot for Area, you'll be shown more information about the area in the sidebar.

18 of 22

Task 1&6 - Visualization & Integration

Increasing Renewable Access in the Philippines through AI

July 17, 2021

Likewise, there is also a list of Areas/Municipalities below the Map Mode option. When you click on the Area, like 'Tapaz, Capiz', you'll be directed to its location, with the municipality's information at the sidebar.

19 of 22

Task 1&6 - Visualization & Integration

Increasing Renewable Access in the Philippines through AI

July 17, 2021

We matched the demand data with an estimated output that solar energy will produce for that area. We start with the population and settlements data from 2017 DHS data and settlements estimate from the our geospatial analysis.

20 of 22

Task 1&6 - Visualization & Integration

Increasing Renewable Access in the Philippines through AI

July 17, 2021

Using the Solar Global Atlas and World Bank datasets, we estimate the mean PVOut that installing solar energy could produce for that Area/Municipality. The granular settlements (in Blue Dots) are also marked for installing solar micro grids all over the municipalities.

21 of 22

Recommendations

Determine clusters of settlements that can be serviced by ground mount solar panels

  • Cluster analysis of suitable settlements by proximity with other settlements
  • Estimate energy demand of each cluster using another dataset, e.g. 2011 Household Energy Consumption Survey by the Philippine Statistics Authority)
  • Estimate energy availability of each cluster via raster sampling, e.g. PV electricity output from Global Solar Atlas)
  • Get clusters of settlements where PV electricity output > energy demand

Increasing Renewable Access in the Philippines through AI

July 17, 2021

22 of 22

Active Collaborators

Increasing Renewable Access in the Philippines through AI

July 17, 2021

Bryce Mercines

Michael Brian Manuel

Arjay Ronald Hosmillo

Andrea Marie Maderazo

Albert Yumol

Jose Marie Antonio Minoza

Neil Ruaro

Dana Redeña

Cyndi Ignacio

Raphael Bihis

Joshua Cortez

Carl Terence Valdellon

Bryan Pajarito

Abegail Lagar

Jofer Santiago