Mapping of Suitable Sites for Concentrated Solar Power Plants in the Philippines Using Geographic Information System and Analytic Hierarchy Process

Abstract

Solar energy is a renewable source of energy harnessed from the sun. Concentrated solar power (CSP) plants harness this energy by focusing sunlight on a limited area to heat a working fluid, which is used to generate steam and power a thermodynamic cycle that produces electricity. There are currently no CSP plants in the Philippines, and this study aimed to locate the most suitable sites for this type of power plant. The first step was to mask out areas totally unsuitable as plant sites; we identified five exclusion factors for this: protected areas, slope, direct normal irradiance (DNI), water bodies, and land cover type. A scoring gradient was then applied to the remaining suitable areas according to seven ranking factors: DNI, slope, typhoon frequency, capacity of the nearest grid line, distance to the nearest grid line, distance to the nearest road, and distance to the nearest water body. Next, to reflect the actual degrees of influence of the factors to site suitability, we determined their relative numeric weights using analytic hierarchy process, with the weights derived from inputs from interviews with academic and industry experts. Finally, using ArcGIS Pro, we used a weighted sum of the ranking scores to produce a suitability map covering the entire Philippines. Suitable sites in the following provinces were found: Ilocos Sur, Pampanga, Mindoro, Masbate, and Maguindanao. These areas have a total area of 27.9 km² and a projected total power output of 733 MW.

1. Introduction

Solar energy is the energy that comes from the sun in the form of radiation [1]. The sunlight that reaches the Earth’s surface is nearly 50% visible light, 45% infrared radiation, and small amounts of ultraviolet and other forms of electromagnetic radiation. It is the cleanest and most abundant renewable source of energy [2]. It can be harnessed to produce electricity without emitting carbon. In 2019, the world consumed 1793 TWh of solar energy, which is 5% of the total renewable energy mix and 1% of the total energy mix [3].

In 2008, Republic Act No. 9513, or the Renewable Energy Act of 2008, was passed with the goal of accelerating the exploration and development of renewable energy technologies in the Philippines [4]. This law encouraged companies to invest in renewable energy through incentives, such as the feed-in tariff (FIT) system. The FIT system entitles renewable energy producers to the benefit of a fixed payment for every kilowatt-hour of energy distributed to the grid within a minimum period of 12 years [5]. However, the FIT period increased up to 20 years as per the Energy Regulatory Commission Resolution 16 series of 2010 [4].

As of December 2021, the Philippines has 1317 MW total installed capacity in solar power connected to the grid, with 757 MW in Luzon, 476 MW in Visayas, and 84 MW in Mindanao [6]. These numbers are nameplate capacities, as calculated from electric generator nameplates based on the rated power factors. The dependable capacities, i.e., maximum capacities when modified for ambient limitations for a specified period of time, are lower: 586 MW for Luzon, 381 MW for Visayas, and 67 MW for Mindanao, for a total of 1034 MW. Both capacities exceed the National Renewable Energy Program’s (NREP) goal of 285.0 MW installed capacity in solar power by 2030 [4]. Despite this achievement, the Philippines is still lacking compared to its neighboring countries. Vietnam is currently leading the solar power industry in Southeast Asia with 16,504 MW in 2020. Thailand is second with a total installed capacity of 2983 MW in the same year. They are the only Southeast Asian countries with solar thermal technology [7].

Even though the country has surpassed the NREP goal for solar power, the Philippines still has a long way to go in terms of harnessing renewable energy sources overall. As of 31 May 2021, the total installed capacity for renewable energy is 5468.91 MW, which is still far from NREP’s goal of 15,304.3 MW by 2030 [8]. The country is still mostly reliant on fossil fuels for energy production, according to the 2020 energy supply mix with 66.5% from oil, coal, natural gas, and ethanol, while the remaining 33.5% is from renewable energy sources [8]. Installing more solar power plants will help achieve the NREP goal and improve the energy mix.

Concentrated solar power (CSP) is a relatively new energy technology. CSP plants use reflectors to focus sunlight on receivers, boiling working fluid. The thermal energy is converted to electricity using conventional thermodynamic cycles.

Researchers conduct site suitability studies to determine ideal locations for renewable energy plants, including CSP systems. Most studies use geographic information system (GIS) in creating, modifying, and processing geographic data. Methods of multi-criteria decision-making (MCDM) are used to weigh and compare the different factors that affect the suitability of a site; the most common MCDM method used in site suitability studies is the analytic hierarchy process (AHP).

In this study, exclusion criteria were established to mask out areas totally unsuitable as plant sites, such as areas with rough terrain and low solar irradiation; the GIS software ArcGIS Pro was used in processing map data. Then, with the use of AHP, ranking criteria were used to determine the most important factors in site suitability. Finally, ArcGIS processing yielded the desired site suitability map.

1.1. Statement of the Problem

There are currently no CSP plants in the Philippines. CSP technologies can store thermal energy, which means they can produce electricity at night and during low-sunshine periods. Thus, installing CSP plants can help improve the baseload grid capacity. However, installing a large-scale renewable energy plant, no matter what the type is, requires the determination of the most suitable plant locations, and there have been no papers that have done this.

1.2. Objectives

This study aimed to achieve the following:

• Construct a weighted ranking of factors that affect the suitability of a site as a CSP plant location;
• Determine the most suitable locations for CSP plant installation in the Philippines.

1.3. Significance of the Study

To combat climate change, greenhouse gas emissions must be cut down, and this requires the expansion of our renewable energy capacity. Determining suitable sites for CSP plants in the Philippines will help in accelerating the harnessing of renewable energy in the country, given that the country is still far from the goal of 15,304.3 MW by 2030 set by NREP. The results of this study will be a significant contribution to the first steps in deploying CSP in the country. Moreover, the addition of a new type of technology can encourage scientific research in the country.

1.4. Limitations of the Study

The results of this study were highly dependent on the availability of maps. The researchers had to make do with what were available from credible sources. Additionally, the factors for site suitability considered in this study were limited only to topographic, meteorological, and logistic factors. Economic factors, such as the installation costs of the plants that can be built on the suitable sites, and design factors, such as the sizing of the plants, are beyond the scope of this study.

2. Related Literature

This section reviews related papers on CSP technology, exclusion and ranking factors, and specific AHP methodologies, from which we drew the present paper’s methodology.

2.1. CSP

2.1.1. CSP Technologies

Current CSP designs are primarily divided into four categories, as shown in Table 1.

Table 1. Types of CSP technologies.

Columns: Type of Focus Rows: Type of Receiver	Line Focus (Uses a Linear Receiver)	Point Focus (Uses a Point Receiver)
Fixed	Linear Fresnel reflectors	Towers
Mobile	Parabolic troughs	Parabolic dishes

Linear Fresnel reflectors (LFR) utilize the Fresnel lens effect, where small thin lens fragments are used to produce the same reflecting effect of thicker lenses. Parabolic troughs (PT), on the other hand, use parabolic reflectors. PT designs are currently the most mature in the market today [9]. They have significantly lower costs and an operating range of (20–400 °C).

The tower system uses an array of reflectors called heliostats that reflect sunlight on a central tower. Parabolic dishes, on the other hand, have their own individual receivers attached to them. The tower system has been the dominant design for the past few years. Due to the point focus nature of this design, it can achieve temperatures up to 565 °C [10], with certain companies reporting temperatures even north of 1000 °C in their prototypes.

2.1.2. Cooling Requirements

In a typical power plant, water is usually needed for condensation purposes. The same is true for CSP, and as most ideal locations for the plants are in dry, arid regions, sourcing and distribution of the required water for operation is a major hurdle for CSP. Water requirements for each CSP technology family vary widely but are still relatively high: 3000 L/MWh for LFR and PT plants, while solar towers require around half of that. For comparison, coal plants typically only require around 2000 L/MWh [10]. Aside from figuring out the logistic requirements of transporting that amount of water to dry regions, one thing that CSP plants are currently implementing is a hybrid dry–wet cooling system. Dry cooling uses air as the cooling medium, and it has shown promising results in CSP plants in North Africa. There are certain risks however, as CSP plants that use dry cooling have shown lower efficiencies after some time.

The hybrid cooling system is designed to curb the water requirements and at the same time reduce the effects on efficiency. The way that it is configured is that CSP plants use dry cooling systems during winter, where cooling loads are lower, then switch back to wet and dry cooling when ambient temperatures rise.

2.2. Overview of MCDM Factors

2.2.1. Meteorological Factors

For CSP, one of the most important metrics to consider is the direct normal irradiation (DNI). It is the amount of radiation received by a surface that directly faces the sun. While there are universal repositories of measured DNI values online, most of them are usually only measured monthly. Since site selection is more focused on long-term planning, historical data should be consulted when it comes to creating the final recommendations.

Geographical location also plays a significant role, as this dictates the amount of daylight time, which depends on the time of the year and has a direct impact on investment and operational costs. Seasons are also dictated by geographical location, and this factors in the decision process as well.

Lastly, microclimate also has a significant impact on the DNI measurements. If a large area is being analyzed, there is a great risk of error within the analysis [11]. For location analysis, it is preferable to take on-site measurements then overlap them with public records of DNI across multiple sources to minimize the errors. This could then be used to create projections to supplement the arguments to be used for the site selection.

2.2.2. Land and Infrastructure

One of the biggest concerns with location-sensitive power plants is the issue with the transportation of electricity. Since CSP plants usually cover large swaths of land, they are ideally located in areas with low population density. On the other hand, many existing grids, especially in these areas, use small power lines. Unless certain infrastructural changes are made, the size and capacity of CSP plants to be built will be severely limited by these power lines. To scale up, additional investments must be made for larger transport lines.

As already mentioned, CSP plants require huge amounts of water for cooling purposes. While certain measures can be taken to reduce the demand, it is still expected to have a connection with the main water supply to be fully operational. Around 90% of the water is used for cooling purposes, while the remaining 10% is for maintenance [12]. This latter water reserve is for cleaning and boiler blowdown. Bird droppings, dust, and dirt greatly affect the overall efficiency of the plants, and the reflectors must be routinely cleaned [13].

While the reflectors could be repositioned, it is preferable to have them laid out on a flat surface. Since CSP plants on medium capacities require large areas, this immediately puts a strain on site selection. In the worst-case scenario where the ground slope is not ideal, additional investment costs would go to land-clearing operations. This goes hand in hand with existing infrastructure, as serviceable roads are required for the heavy machinery, especially during plant construction.

2.3. Exclusion Factors

The first step in GIS analysis is to exclude areas where the CSP plants cannot be built. In a CSP site suitability study in the United Arab Emirates, Alqaderi et al. [14] used the following exclusion criteria: protected areas, land cover, slope, and solar irradiation. Land cover refers to the type of land use, whether it is urban, agricultural, or biodiversity. In their study, terrain slope that is greater than 4% was excluded; steep areas can increase installation and maintenance costs. CSP plants normally require 1800 kWh/m² of DNI to operate with profit, thus areas with values lower than that are excluded.

A study by Aly et al. [15] to determine solar PV and CSP hotspots in Tanzania additionally used urban expansion and distance to water body as exclusion criteria, aside from solar radiation, slope, and protected areas. Water bodies were excluded because plants cannot be constructed there. Considering urban expansion, areas within 8 km of cities with a population of 250,000 were excluded, and for cities with populations of 100,000 and 200,000, areas within a 6 km buffer zone were excluded. The slope threshold used, meanwhile, was 2.1%. A site assessment study of CSP plants in eastern Morocco by Merrouni et al. [16] also used buffer zones around urban areas; they excluded big cities with a buffer of 5 km and small ones with a buffer of 2 km.

In a study in Algeria, Haddad et al. [17] considered more exclusion criteria. They included distance to the electricity grid, to transport networks, and to airports. Less cost is needed to inject produced electricity into the grid if the grid and CSP plant are near each other [17]. Thus, they excluded areas more than 200 km away from the grid. Transport networks are needed to lessen additional costs in constructing roads to navigate to the site; areas located 40 km from the transport network were thus excluded, and areas less than 100 m away from a road or railroad were also excluded to avoid disturbing them. In terms of populated zones, Haddad et al., excluded areas that are located more than 200 km away to avoid energy losses and facilitate distribution and supply. Lastly, they also used a maximum slope of 2.1%.

Table 2 summarizes the exclusion factors used in related studies reviewed. As can be seen, protected areas, slope, water bodies, urban areas, and DNI were used in most studies, with some of them using specialized factors such as wind load, religious and tourist areas, and airports.

Table 2. Exclusion factors used in related site suitability studies.

Exclusion Factor	United Arab Emirates [14]	Tanzania [15]	Eastern Morocco [16]	Algeria [17]	Western Australia [18]	Mauritius [19]
Protected areas	✓	✓	✓	✓	✓	✓
Slope	✓	✓	✓	✓	✓	✓
Water bodies	✓	✓	✓		✓	✓
Cities		✓	✓	✓	✓	✓
Direct normal irradiance (DNI)	✓	✓	✓	✓	✓
Roads and railroads			✓
Wind load					✓
Other land use	✓			✓	✓	✓
Religious and tourist areas						✓
Airports						✓

2.4. AHP

Several factors decide the final suitability of a candidate power plant site with different degrees of influence, i.e., some factors are more important than others. MCDM methods are used to determine the weights of these factors. Across the literature on site assessment for renewable energy plants, the AHP, developed by Thomas Saaty in 1977, is the most widely used MCDM method [17,19,20,21]. In a 2015 review of MCDM papers [22], the field of energy, environment, and sustainability was found to use MCDM the most, with AHP being the most widely used method. AHP is not only used in site suitability studies but, more generally, in solving complex decisions with different criteria uses [23,24,25].

The AHP method works on a pairwise comparison matrix, which encodes the importance of each factor compared to the others [15,20,21]. Each element x_ij of the comparison matrix is a number that compares the importance of criteria i over criteria j in determining site suitability. Saaty prescribed using a scale of 1 to 9 (see Table 3), with 1 meaning equal weight. The element x_ji is the reciprocal of x_ij. The matrix is usually filled in after interviews with experts and stakeholders on CSP plants were conducted. Interviewees include government institutions, research institutions, electricity providers, private companies, local and international NGOs, development partners, and sustainable energy advocacy groups [15,18].

Table 3. Saaty’s scale for the entries in the comparison matrix. Reprinted from [26], Copyright 1977, with permission from Elsevier.

Number	Definition	Explanation
1	Equal importance	Two criteria contribute equally
3	Weak or moderate importance	Experience and judgment slightly favor one criterion over another
5	Essential or strong importance	Experience and judgment strongly favor one criterion over another
7	Very strong or demonstrated importance	A criterion is favored very strongly over another; its dominance is demonstrated in practice
9	Absolute importance	The evidence favoring one criterion over another is of the highest possible order of affirmation
2, 4, 6, 8	Intermediate values	Used to represent compromise between the priorities listed above

There are several methods for deriving the weights, properly called the priorities, from a given matrix. The two most commonly used are the eigenvalue method and the geometric mean method, both of which have strong mathematical foundations [27]. The eigenvector method is the one proposed by Saaty himself, and several authors have proposed axiomatic frameworks that only this method satisfies [27]. In this method, the priority vector is simply the eigenvector corresponding to the eigenvalue of the comparison matrix with the highest magnitude. The Perron–Frobenius theorem guarantees that, given a positive matrix, this largest eigenvalue is a real number and the corresponding eigenvector is positive.

The geometric mean method simply computes the geometric mean of the entries in the rows of the comparison matrix, i.e.,

$$w_i={\left({\displaystyle \prod }_{j=1}^n{a}_{ij}\right)}^{1/n}.$$

(1)

This vector is then normalized, so the entries will add up to 1. This method has equally strong proponents. Barzilai [27] proposed a framework that only the geometric mean method satisfies, while Bana e Costa and Vansnick [27] drew contrast with the eigenvector method by showing that the eigenvector method does not satisfy a consistency measure called the condition of order preservation.

A level of consistency is expected in a comparison matrix [15,20,21]. A matrix is consistent if

$$x_{ik}=x_{ij}\times x_{jk}.$$

(2)

To measure the level of consistency, the consistency index (CI) is computed using the formula

$$\mathrm{CI}=\frac{\lambda -n}{n-1},$$

(3)

where λ is the largest eigenvalue of the comparison matrix and n the number of criteria. The consistency ratio (CR) is then computed as

$$\mathrm{CR}=\frac{\mathrm{CI}}{\mathrm{RI}},$$

(4)

where RI is the consistency index of a matrix randomly filled with values using Saaty’s scale. There are tabulated values of this variable as a function of $n$ [15].

The maximum allowed value of CR is 10%. If a higher value is obtained, the comparison matrix needs to be reconstructed [20], or the consistency adjustment procedures devised by Saaty should be performed [21].

Consolidating multiple comparison matrices presents a layer of complexity to an AHP problem. When given multiple matrices, there are two ways to arrive at a single priority vector [27]. In the method of aggregation of individual priorities, the individual priority vectors of the comparison matrices are aggregated. In aggregating the priority vectors, two methods are most commonly used: the weighted geometric method and the arithmetic method, with the former being slightly preferred over the latter [27]. In the method of aggregation of individual judgments, the individual comparison matrices are aggregated into a single matrix, from which the priority vector is derived. Using the arithmetic mean does not result in a reciprocal matrix [27]; Aczel and Saaty, and Saaty and Alsina [27] proposed a set of properties that the aggregate matrix needs to satisfy and showed that the geometric mean is the only applicable function:

$$a_{ij}{}^G={\displaystyle \prod }_{h=1}^m{\left[a_{ij}{}^{\left(h\right)}\right]}^{{\lambda }_h},$$

(5)

where a^(h) is the hth individual comparison matrix and a^G is the aggregate matrix.

2.5. Ranking Factors

Alqaderi et al. [14] used AHP to determine the weights of the ranking factors they used in their United Arab Emirates study. DNI obtained the highest weight of 64.9% while slope and proximity to resources (roads, water bodies, and power grids) obtained 27.9% and 7.2%, respectively.

Aly et al. [15] used seven ranking criteria in their Tanzania study: solar resources; water availability; proximity to roads; proximity to the utility grid; proximity to cities with over 250,000 inhabitants; proximity to cities with 100,000 to 250,000 inhabitants; and proximity to mines. The weights they obtained were as follows: 61.8% for solar resources, 20.3% for water availability, 13.4% for accessibility to resources, and 4.5% for urban demand. They were able to determine four hotspots in Tanzania.

Haddad et al. [17] obtained a significantly lower weight for DNI (27.05%), but it remained the heaviest factor. The top factors after DNI were distance to high population density and distance to electricity grid, with weights of 23.0% and 22.95%, respectively. Their results showed that almost half of Algeria is incompatible for CSP implementation, due to factors including slope and proximity to the grid and water bodies.

In Merrouni et al.’s [16] study of eastern Morocco, climate also had the highest weight, followed by slope. Contrary to other studies, they considered two types of power plant cooling. They obtained a higher weight for distance to water in the wet cooling scenario (11.7%) than in the dry cooling scenario (5.3%).

Yushchenko et al.’s [28] GIS-based assessment of PV and CSP generation potential in western Africa used solar irradiance, distance to electricity grid lines, distance to roads, distance to settlements, and population density as the AHP factors. They developed two scenarios. Solar irradiance was chosen as the main criterion in the first scenario. The second scenario was based on an existing study conducted in Oman by Charabi and Gastli [29] that assigned more weight on the distance to electricity lines and major roads. Results showed that there is a significant difference between the two scenarios. Less areas were seen as suitable with distance to electricity lines and major roads as the main criteria.

A more general selection process was proposed by Schlecht and Meyer [11] in their site suitability study. Aside from the factors used by the studies previously discussed, they proposed considering other climactic factors such as temperature, humidity, and pressure, as they also have contributing effects to the overall performance of a CSP plant. They also proposed considering wind direction and velocity, as, aside from blocking the reflectors, certain soil particles could also scratch the surface of mirrors, depending on the material composition.

Table 4 summaries the top five ranking factors computed in related site suitability studies reviewed. DNI and sunshine hours were most commonly computed to have the largest weights [15,18,20,21,30,31,32,33,34]. Other factors that rank high are slope and proximity to water resources, roads, and grid lines. Interestingly, distance to shoreline ranked highest in Giamalaki and Tsoutsos’ 2019 study [21] on the Mediterranean, with DNI ranking the lowest.

Table 4. Top five ranking factors determined in related site suitability studies.

Rank	Tanzania [15]	Mediterranean [21]	China [30]	Serbia [31]	Iran [32]
1	DNI	Distance from shoreline	DNI	DNI	DNI
2	Water bodies	Water bodies; land cover	Temperature	Sunshine duration	Distance to grid
3	Distance to grid	Slope	Distance to roads	Slope	Distance to roads
4	Distance to roads	Elevation; visibility	Slope	Temperature	Elevation
5	Cities	Distance to roads; slope	Aspect; distance to grid	Aspect	Slope
Rank	Spain [33]	Isfahan-Iran [34]	Algeria [20]	Western Australia [18]
1	Distance to grid	DNI	DNI	Water bodies; distance to roads; wind
2	DNI	Sunshine duration	Distance to grid	Auxiliary fuel; cities
3	Temperature; aspect	Aspect	Sunshine duration	Sunshine duration
4	Distance to roads	Elevation	Roads	DNI
5		Distance to grid	Cities	Protected areas

3. Methodology

This study used GIS analysis, with AHP as the MCDM method, in mapping suitable locations for CSP plants in the Philippines; Figure 1 presents the methodology. The first step was the identification of exclusion criteria that mask out areas unsuitable as plant location. This study used exclusion factors collated from the related literature reviewed in the last section. These masks were mapped and overlaid using ArcGIS Pro. Next, a list of factors to be weighed using AHP were constructed, and tiers of scores were developed for each factor. Interviews with experts and stakeholders were conducted, and their inputs were used in constructing the pairwise comparison matrix. AHP computations yielded the required weights. Lastly, the suitability score of locations in unmasked regions were computed using a simple weighted sum of the scores and the AHP weights.

Figure 1. Methodology.

3.1. Exclusion Criteria

Table 5 shows the exclusion factors that were used to mask out totally unsuitable areas. For each factor, the exclusion factor and the map used, including its source, are shown.

Table 5. Exclusion factors and criteria used.

Ranking Factor	Exclusion Criterion	Map Used and Map Source
Protected area	Protected areas were excluded.	Map of protected areas from the Department of Environment and Natural Resources Biodiversity Management Bureau
Slope	Areas with slope more than 2.1% were excluded. 1	SRTM (Shuttle Radar Topography Mission) 1 Arc-Second Global digital elevation model
DNI	Areas with DNI less than 1600 kWh/m2/year were excluded. 2	Solar map from the Global Solar Atlas 3
Water bodies	Lakes were excluded. 4	Map of lakes from the United Nations Office for the Coordination of Humanitarian Affairs
Land cover	Urban and agricultural areas were excluded. 5	Land cover map from the National Mapping and Resource Information Authority

¹ This slope value is the one most used in the related literature [18]. ² This value was proposed in a 2022 study by the World Bank [35]. ³ The solar map is available for download in GeoTIFF format in the Global Solar Atlas website (https://globalsolaratlas.info/download/philippines; accessed 16 April 2022). ⁴ There are no available maps of rivers in the Philippines with data on the river widths. Using a single value will misrepresent a large fraction of the rivers. Thus, only lakes were considered in the exclusion map in this study. ⁵ The following land cover types were the ones considered suitable: brush/shrubs, grasslands, open forests, and open/barren lands.

3.2. AHP

Respondents were given questionnaires where they compared every pair of ranking factors and for each pair determined the more important one, if there is one. They were asked to use the numbers 1 to 9 in comparing the factors, as prescribed by Saaty in his comparison scheme (see Table 3). Their inputs were arranged into the comparison matrices.

In consolidating the multiple comparison matrices, the method of aggregation of individual judgments was chosen because it yielded a better set of weights or priorities, as will be shown. The geometric mean method was used to aggregate the comparison matrices, and the eigenvector method was used to derive the weights from the aggregate matrix.

3.3. Ranking Factors

After reviewing related literature, we arrived with the following set of seven ranking factors:

• DNI;
• Typhoon frequency or the average number of typhoons that hit an area in a year;
• Slope;
• Voltage rating of the nearest grid line;
• Distance to the nearest grid line;
• Distance to the nearest road;
• Distance to the nearest water body.

An additional factor, condition of the nearest road, was initially considered but was then excluded due to lack of data source. While only lakes were considered in the exclusion criteria, both lakes and rivers were considered in the ranking factor of distance to water body.

3.4. Respondents

The researchers created an initial list of respondents of 30 people, who are all stakeholders in the solar energy industry. Out of the 30 invited, only six people accepted the invitation to participate. Four of these respondents were from the academe, one from the industry, and one from the government. Table 6 shows brief profiles of them.

Table 6. Profiles of respondents.

Respondent No.	Sector	Profile
1	Academe	College professor. Published a paper on a solar desalination system.
2	Academe	Associate professor. Published papers on CSP and PV systems.
3	Industry	Registered electrical engineer working in a local grid corporation. Founded an organization that provides solar power to rural communities.
4	Academe	University research staff and chairperson of a college department. Completed work on CSP thermal design.
5	Government	Research specialist in a government agency. Part of a team that spearheaded a solar–wind integration study in the Philippines.
6	Academe	Assistant professor. Published papers on system modeling for electrical networks and renewable energy systems.

3.5. Scoring System

A linear scoring system for each of the ranking factors was constructed (see Table 7); all the papers reviewed used a linear system as well. For a given factor, bins of equal width were specified to contain all the values of that factor across the entire Philippines. For example, the maximum slope value considered for ranking is 2.1% while the minimum value on the generated slope map is 0. A lower slope corresponds to a higher score, so the lower bound for the highest score of 9 was set to 0. Setting the bin width to 0.25% situates 2.1% within the bounds for the lowest score of 1, which are 2.00% and 2.25%. For the distances to grid, road, and water body, we did not use a maximum allowed value. Thus, the highest values that appeared in the map were assigned to the lowest score of 1; the bin sizes were adjusted accordingly to make sure the highest values fit inside the intervals for the score of 1. For example, the largest distance to grid detected was 389 km; the bin for the score of 1, which is actually 360–405 km (notice that the bin size is 45 km), contains this maximum value.

Table 7. Suitability scores for the ranking factors.

Criterion	Score
	1	2	3	4	5	6	7	8	9
DNI (kWh/m2/year)	1600–1635	1635–1670	1670–1705	1705–1740	1740–1775	1775–1810	1810–1845	1845–1880	>1880
Typhoon frequency	>3.2	2.8–3.2	2.4–2.8	2.0–2.4	1.6–2.0	1.2–1.6	0.8–1.2	0.4–0.8	<0.4
Distance to grid (km)	>360	315–360	270–315	225–270	180–225	135–180	90–135	45–90	<45
Grid capacity (kV)	<45	45–90	90–135	135–180	180–225	225–270	270–315	315–360	>360
Distance to road (km)	>200	175–200	150–175	125–150	100–125	75–100	50–75	25–50	<25
Distance to water body (km)	>160	140–160	120–140	100–120	80–100	60–80	40–60	20–40	<20
Slope (%)	2.00–2.10	1.75–2.00	1.50–1.75	1.25–1.50	1.00–1.25	0.75–1.00	0.50–0.75	0.25–0.50	<0.25

4. Results and Discussion

This section discusses the individual exclusion maps and the final exclusion map generated, the weights and consistency ratios obtained using AHP, the final suitability map generated, and projected power output from the suitable sites.

4.1. Exclusion Map

The application of the overall exclusion layer significantly reduced the total viable area to only about 0.3% of the total Philippine land area, with the slope and DNI being the biggest eliminating factors.

Considering the threshold value for DNI of 1600 kWh/m²/year, only 6% of the Philippine land area is suitable as plant location, contradicting the expectation that the country will have a lot of well-lit areas due to its proximity to the equator. Several factors, including microclimate, pollution, cloud cover, and abundance of terraneous areas, may have contributed to the low turnout. Most of these well-lit areas are in Region I (see Figure 2a).

Figure 2. Exclusion map for each of the ranking factors: (a) direct normal irradiance (DNI); (b) land cover; (c) DNI (central Luzon closeup); (d) land cover (central Luzon closeup); (e) water bodies; (f) protected areas; and (g) slope.

Figure 2b shows the exclusion map for land cover. As expected, agricultural centers and highly urbanized areas with a considerably strong industrial base were mostly excluded, especially areas in NCR and Region IV-A. Areas in Region I and CAR were deemed suitable, since they are less densely populated and slope was not taken into account for this exclusion layer. The huge presence of mangroves across the Palawan province and some portions of Mindanao also contributed to the elimination of the said areas.

An area is deemed suitable if it receives high DNI and has a suitable land cover type. Referring to Figure 2c,d, it can be noticed that areas with high DNI have unsuitable land cover type, and areas with suitable land cover type have low DNI. The mutual exclusivity of these two types of suitable areas was the biggest factor in the significant reduction in suitable area in the country.

Figure 2e shows the exclusion map for water bodies. As previously discussed, only lakes were excluded.

Figure 2f shows the exclusion map for protected areas. Protected areas in the Philippines occupy 15.82% of the country’s total land area. Notice how Palawan is almost entirely a protected area.

A huge portion, or 83.85%, of the country’s total land area has slope greater than 2.1%, making it unsuitable for CSP installation (see Figure 2g). The relatively flat areas are mostly in Regions III and XII.

Figure 3a shows the resulting aggregate exclusion map, with closeup insets of northern Luzon (Figure 3b) and Mindoro (Figure 3c). Notice how almost the entire country is not suitable as plant location; the suitable areas are small, few, and scattered that they are not appreciably visible in a magnification that shows the entire country, as in Figure 3a. The resulting viable areas were concentrated on Region I, Pampanga, Mindoro, Masbate, and Maguindanao.

Figure 3. Resulting exclusion map (a) with closeups of (b) northern Luzon and (c) Mindoro.

4.2. Weights and Consistency Ratios

Table 8 shows the aggregate matrix derived from the individual comparison matrices using the geometric mean method, with values shown up to two decimal places.

Table 8. Aggregate pairwise comparison matrix.

	DNI	Typhoon Frequency	Grid Capacity	Distance to Grid	Distance to Road	Distance to Water	Slope
DNI	1	1.71	5.19	4.73	4.71	4.73	2.80
Typhoon Frequency	0.59	1	0.63	0.93	0.96	0.83	1.48
Grid Capacity	0.19	1.59	1	0.92	1.14	1.08	1.22
Distance to Grid	0.21	1.07	1.08	1	2.47	0.99	1.74
Distance to Road	0.21	1.05	0.88	0.41	1	0.49	0.97
Distance to Water	0.21	1.21	0.93	1.01	2.03	1	1.20
Slope	0.36	0.68	0.82	0.57	1.03	0.84	1

Table 9 shows the consistency ratios of the individual comparison matrices of the respondents. Note how only one respondent, Respondent 5, yielded a ratio less than the threshold value of 10%, or 0.10.

Table 9. Consistency ratios of the individual comparison matrices.

Respondent No.	Consistency Ratio
1	0.30
2	0.29
3	0.29
4	0.25
5	0.09
6	0.19

The consistency ratio of the aggregate comparison matrix is 0.0368, which is less than 0.1. Table 10 shows the priorities computed using the eigenvector method, along with the respective ranks in descending order. For comparison, the priorities computed from aggregating individual priorities, also using the eigenvector method for deriving the individual priorities and then the geometric mean method for deriving the final priorities, are also shown.

Table 10. Priorities computed using aggregation of individual judgments and aggregation of individual priorities.

Ranking Factor	Aggregation of Individual Judgments		Aggregation of Individual Priorities
	Priority	Rank	Priority	Rank
DNI	0.3877	1	0.3759	1
Typhoon frequency	0.1111	3	0.1204	2
Grid capacity	0.1071	5	0.1087	5
Distance to grid	0.1214	2	0.1175	3
Distance to road	0.0761	7	0.0753	7
Distance to water	0.1105	4	0.1105	4
Slope	0.0861	6	0.0917	6

The ranking of the factors in both methods is the same except for typhoon frequency and distance to grid. Since we have incomplete data on typhoon tracks, we chose the first set of priorities, which puts less weight on typhoon frequency.

4.3. Final Suitability Map

Applying the scoring system to the raw maps, with the AHP weights, produced the unmasked suitability map shown in Figure 4a; note that the exclusion map has not been applied here yet. The suitability scores ranged from 2.8 to 8.0, with a mean of 3.5. More than half of the country (57.36%) has an intermediate suitability score of 3–4, while those with a score between 4 and 5 comprise 38.96% of the total land area. Figure 5 shows the distribution of the suitability scores for the entire country; the horizontal axis is the score, and the vertical axis is the number of pixels.

Figure 4. Final suitability map with the exclusion layer (a) not applied and (b) applied.

Figure 5. Distribution of the suitability scores in the Philippines.

It can be seen that the most suitable areas in the Philippines are the ones with the highest DNI (compare Figure 2a and Figure 4a). This is because DNI has the highest computed priority.

Lastly, the exclusion map previously created was applied on the unmasked suitability map to produce the final, masked suitability map shown in Figure 4b. As previously discussed, only about 0.3% of the entire Philippines remained after applying the exclusion layer, so only a few areas are left in the final map. They are also scattered across the entire country, so they are almost not visible in the scale of Figure 4b.

The most suitable areas for CSP installation are scattered across the Ilocos provinces (see Figure 6a). A contiguous area was found near Caoayan, Ilocos Sur, with an area of 5.18 km². However, this area is on a river delta, making it unsuitable for plant construction. The next most suitable areas are in Pangasinan. The researchers were unable to calculate the largest contiguous area here because the patches of suitable areas were too scattered (see Figure 6b).

Figure 6. Closeups of the final masked suitability map: (a) Ilocos provinces; (b) Pangasinan; (c) Pampanga; (d) southern Mindoro; (e) southern Masbate; and (f) Maguindanao.

The largest contiguous area in Pampanga is 3.79 km², an agricultural land with bare areas (see Figure 6c). In southern Mindoro, the largest contiguous area is 6.03 km² (see Figure 6d). Similar to Pampanga, the land is used for agriculture and contains bare areas as well.

Southern Masbate contains the largest contiguous area suitable for CSP installation in the Philippines, specifically 8.35 km² (see Figure 6e). The last suitable area is in Cotabato City, Maguindanao, with an area of 4.56 km² (see Figure 6f). Both of these areas are bare land.

4.4. Projected Capacity

Column 2 of Table 11 shows the lowest annual DNI value in the largest contiguous area in the five locations discussed in the previous section; the corresponding available solar power values in W/m² are in the next column. The areas of those suitable contiguous sites are shown in Column 4. The lower bound value was utilized to provide a more grounded evaluation of the said areas.

Table 11. Lower bound energy generation of the largest contiguous area at suitable areas.

Location of Contiguous Suitable Area	Lowest Annual DNI (kWh/m2)	Lowest Available Solar Power (W/m2)	Size of Largest Contiguous Area (km2)	Projected Power Output (MW)
Ilocos	1740	199	5.18	144
Pampanga	1641	187	3.79	99
Mindoro	1600	183	6.03	154
Masbate	1641	187	8.35	219
Maguindanao	1600	183	4.56	117
Total			27.91	733

The power output, shown in the last column, was estimated using the formula

Projected Power Output = Available Solar Power × Area × Area Collection
Factor × Efficiency.

(6)

The area factor is the fraction of the total site’s area covered by heliostats. The value of 0.7 was used based on the maximum land occupancy of solar collectors with minimum shading effect [25]. Solar power plants have average efficiency values of 7–25% [35]; a fairly conservative value of 20% [18] was used. The computed projected power outputs are shown in the last column. The Ilocos site receives the highest DNI, but the Masbate site has the highest projected output because of its larger land area. Pangasinan was not included because the researchers were unable to determine a contiguous area.

5. Summary and Recommendations

Just like most renewable energy systems, the performance of a CSP plant is site-sensitive and highly dependent on the existing main energy source in the area. The following factors were chosen to determine the suitability of a location as a CSP plant: DNI, typhoon frequency, slope, voltage rating of nearest grid line, and distances to the nearest grid line, road, and water body. Their respective weights were determined using AHP; DNI, distance to grid, and typhoon frequency obtained the highest weights. This satisfies the first objective set in Chapter 1.

To satisfy the second objective, a map of the suitability scores of areas across the entire Philippines was generated. Despite the proximity of the country to the equator, the country receives relatively low DNI, with respect to CSP requirements, and this substantially restricted the total suitable area. The total suitable area in the country is about 905 km²; much of this area, however, is non-contiguous. Relatively large, contiguous suitable areas were found in the following provinces: Ilocos, Pampanga, Mindoro, Masbate, and Maguindanao. The Ilocos site, which covers 5.18 km² has the highest average DNI of 199 W/m². The Masbate site, however, yielded the highest projected power output of 219 MW due to its large land area (8.35 km²). The total area and total projected output of the five sites are 27.9 km² and 733 MW, respectively, or 26.3 MW/km².

Suitability studies such as the present paper rely heavily on the availability of maps on the factors to be used. In this study, only lakes were excluded as water bodies. There are data on the location of rivers, but there are no data on their widths. Using a single value for the width will misrepresent a large number of rivers, so it was decided to disregard them in creating the exclusion map. Hence, river deltas, such as the one in Ilocos Sur, were included in the final suitability map when they should not have been. There are also incomplete data on the typhoon tracks; the typhoon frequencies computed were underestimations. Thus, it is recommended, before conducting a suitability study such as this, to make sure there are complete data on the exclusion and ranking factors to be used.

While multiple factors were already incorporated in this study, further research can still be done to include other parameters that were not covered in this paper. To add more depth to the technical aspect of site selection, the wind velocity and soil composition could also be included in further studies to factor in possible loss of efficiency due to abrasion on the reflector surface. Moreover, since there are four standard CSP systems being used in the industry, a more specific study can be conducted that will consider the efficiency, cooling requirements, capital costs, and base load capacity of the said systems to determine the best fit in this country’s geographic and climatic conditions. Furthermore, since renewable energy is also subjected to market forces, an economic evaluation combined with sensitivity analysis on potential electricity prices can be performed to provide a more holistic view of the potential for a CSP plant in the country. The distance to the nearest existing solar power plant could also be added as a factor, since the existence of the said plants usually indicates that the necessary infrastructural support was already available within the vicinity.