Selection of Potential Sites for Sustainable Development of Solar Photovoltaic Plants in Northeastern Brazil Using GIS and Multi-Criteria Analysis

Brazil has one of the highest percentages of solar radiation in the world and which it is a favorable condition to generate electricity using solar photovoltaic systems. The construction of photovoltaic plants depends on the site selection that influences the electricity generation capacity and the socio-economic benefits that can be derived from it in the future. This article proposes to combine Spatial analysis that involves the acquisition and management of spatial data using a Geographic Information System (GIS), and the fuzzy multi-criteria AHP-TOPSIS model to identify potential locations for the installation of solar photovoltaic systems in the northeastern region of Brazil, State of Pernambuco. The combination of GIS and fuzzy AHP-TOPSIS offers the user the possibility of assigning merit categories to the mapping according to multiple assessment criteria. The GIS include factors of the restrictions and criteria. The restrictions are inserted into the GIS using layers defined from current legislation (urban areas, undeveloped land, community sites, infrastructure, etc.), which reduce the study area by eliminating zones in which PV installation is not permitted. The results show that 22 and 40% of the area of the state of Pernambuco has the very high and high potential, respectively, for the implantation of PV.


Introduction
A country's economic and social development is directly linked to its energy production. The more a country develops, the greater its need for energy, and these costs will directly impact its economy. Because of growing concern for the environment due to the use of fossil fuels since the industrial revolution, renewable sources have begun to play an important role in the diversification of the global energy matrix, contributing to sustainable development.
the highest levels of potential solar radiation in the country. Specifically, in the state of Pernambuco, the annual average of global radiation exceeds 5 kWh/m2 per day over most of its territory (Macedo et al., 2015). Because of these excellent climatic characteristics, this territory has become attractive for the implantation of photovoltaic solar plants (PV). Pernambuco also has Law No. 14,090, of 17 June 2010, which establishes goals for mitigating greenhouse gas emissions, encourages sustainable construction, and requires projects with a high environmental impact to inventory their greenhouse gases emitted. In order to obtain higher returns at their facilities, developers and investors need to use decision support models and methods that make it possible to maximize PV efficiency.
The analysis of the criteria that affect resource availability is an important part of the solar energy planning process. The criteria used in selecting the location for new plants may include: energy production, orography (slopes and orientation), environment (land use, land cover, erosion, flood risk, accessibility, and visual impact), distances (from roads, from energy substations, and from urban areas), financial, and climatic (solar radiation, temperature, dust, and wind). In fact, selecting a PV site is a complicated decision-making process, because the site is required to be climatically and geographically satisfactory, and must simultaneously have the greatest possible generation potential (Macedo et al., 2015). Therefore, it is essential that technological, economic, and environmental aspects be considered, as it is a complex process that needs complete information on a wide range of criteria that impact deciding-making regarding the available areas (Tahri, Hakdaoui, & Maanan, 2015). Multi-criteria decision analysis (MCDA) approaches are the most appropriate means to provide decision support.
However, according to the reviewed literature, there are some gaps in knowledge associated with photovoltaic solar plant site selection. The number of variables is limited in the literature and does not distinguish exclusionary aspects that make solar plant projects unfeasible like areas of risk or legal exclusion, from those aspects that limit or condition the activity like slope, orientation, temperature, radiation, etc.). In other words, the criteria considered may not identify the most suitable areas (Yushchenko et al., 2018).
In this context, Geographic Information Systems (GISs) are suitable tools for analyzing and visualizing spatial information, which have been used in energy applications to assess resources and plan infrastructure. GIS can be used to spatially solve problems related to data integration and resource management such as analytical and spatial modeling, spatial display, and reporting (Yushchenko et al., 2018). With the ability to manipulate data in digital models (raster and vector) GIS offers a collection of procedures, techniques, and algorithms to structure data to instantiate decision-making problems that deal with the design, evaluation, and prioritization of alternative decisions (Macedo et al., 2020).
Various decision-making techniques have been developed in previous studies to select the location of solar plants and other energy projects based on geographic information systems (GIS) and spatial analyses (Yang et al., 2018). For example, Lee et al. (2014) applied the MCDA method based on the Analytic Hierarchy Process (AHP) to the selection of wind farm installation strategies. Janke (2010) studied MCDA for solar and wind farms using GIS. Charabi and Gastli (2011) studied the location of solar farms using GIS and fuzzy logic. Arán-Carrión et al. (2008) and Tahri, Hakdaoui and Maanan (2015) applied Geographic Information Systems (GIS) and AHP to assess solar park locations. Merrouni, Mezrhab and Mezrhab (2016) developed a goal programming model to select appropriate locations for different types of renewable energy installations. Sánchez-Lozano, García-Cascales and Lamata (2016) evaluated seven regions of hybrid wind/solar power stations via ELECTRE-and found the result to have better correction than results from related studies. Maleki, Hizam Ⅱ and Gomes (2017) conducted a multi-criteria assessment of photovoltaic technologies using the TOPSIS and AHP methods. In northeastern Brazil, Tiba et al. (2014), Azevêdo, Candeias and Tiba (2017) and da Ponte, Calili and Souza (2021) analyzed the development of a management and planning system on a GIS platform and multi-criteria analyses for renewable energy source administrators, planners, or consultants.
For the current study, a combined approach using GIS and the fuzzy AHP-TOPSIS method is described in order to classify the possible locations for photovoltaic parks in the northeastern Brazilian state of Pernambuco, into categories of merit according to the assessment criteria. First, certain areas are excluded, simplifying the subsequent analysis and allowing for more information to be included. After that, to carry out more detailed analyses, the aspects to be considered when selecting locations are classified into criteria and sub-criteria.
This two-stage approach combines criteria, which include solar radiation, local physical terrain, environment, climate, location, distance from roads and transmission lines, technology employed, and deployment costs. GIS data (solar radiation time series, digital elevation model (DEM), ground cover, and temperature) were used as additional input parameters. The objective of this study is to provide a methodology that identifies locations with potential for photovoltaic generation, and thereby supports the development of new photovoltaic plants. The methodology applied for Pernambuco, in the northeast of Brazil, considered the installation of 1 to 5 MW of a near infrared (NIR) band compared to a red band. This takes advantage of the high reflectance of vegetation in the NIR spectral range and the high pigment absorption of red light (Tahri, Hakdaoui, & Maanan, 2015). The Normalized Difference Vegetation Index (NDVI) is the best indicative factor for the state of plant growth and the spatial distribution of vegetation, which is linearly related to the vegetation distribution density.
Other areas with restrictions must be considered. These areas include bodies of water, ecologically sensitive areas, wildlife conservation areas, floodplains, towns and cities, roads, railways, and steep hillside areas. Photovoltaic modules contain some toxic and dangerous material components that must not be disposed of in the environment, for example, waste produced over the long term for which there is still no suitable destination (Zekai, 2014). Therefore, solar plants should be installed at a safe distance from sensitive areas. Large-scale installations of photovoltaic systems can also damage areas of land with potential for agriculture (Zekai, 2014). On the other hand, the protection of land with potential for agriculture is a principal environmental objective. Based on Aydin, Kentel and Duzgun (2013) agricultural areas should be excluded from being selected for PVP installation.
Slope is a topographic feature that can strongly affect project costs, and which therefore plays an important role in the selection of a suitable location for a solar power plant. On steep terrain, panels can create shade on neighboring panels, reducing energy conversion efficiency. It is also easier to set up the infrastructure on level ground, reducing overall construction costs. The ideal slope for the module that depends on latitude (for example, in Pernambuco, at latitude 9° south, this can be found between 1° and 5°, with 10° being recommended to avoid the formation of dust layers that can absorb part of the incident radiation). The orientation of the panel is another factor that affects the output of the PV modules. The ideal orientation of the module is always towards the north in the southern hemisphere (Oliveira & Gómez-Malagón, 2018).
Altitude and temperature are also factors that affect the normal functioning of solar panels and the production of solar energy. High altitudes increase the difficulty of building photovoltaic plants and can affect transmission facilities, however, the highest density areas of solar energy occur in high-altitude desert areas. Altitudes above 5800m are not recommended, although this depends on the study area (Pinto, Amaral, & Janissek, 2016). According to Skoplaki and Palyvos (2008) areas with an average temperature below 10°C or above 20°C should be excluded. The c-Si photovoltaic modules lose energy as temperature increases above the standard test conditions (25°C) at the rate of 0.5 and 0.6%/°C and therefore, exceeding the critical limit of 25°C is not recommended. With regard to moisture, the greater the amount of relative humidity in an area, the greater the absorption of short-wave solar radiation, which decreases the total amount of incident solar irradiation usable by the solar panel (Pinto, Amaral, & Janissek, 2016).

Step 2-Criteria Definition
The computer programs used were ArcGIS 10.3 (ESRI) and QGis (OSGeo 4W). Various cartographic sources were used, including maps from the Brazilian Institute of Geography and Statistics (IBGE), the National Department of Traffic Infrastructure (DNIT), the Brazilian Agricultural Research Corporation (Embrapa), the Pernambuco State Planning and Research Agency (Condepe/Fidem), and the National Institute for Colonization and Agrarian Reform (Incra). The solar radiation maps were obtained from The National Solar Radiation Database-NSRDB (Sengupta et al., 2018) and temperatures were obtained from Global climate and weather data (WorldClim).

Step 3-Criteria Selection
Based on the literature review, all of the information was synthesized, criteria were chosen, and a preliminary list was defined. This list of criteria was sent to a group of experts in photovoltaic solar plants for approval. Based on their assessment, the list of criteria for the selection of photovoltaic plant sites was established. The criteria were separated into positive and negative indicators or restrictions. The restrictions to be inserted in the GIS are obtained from the regional public administrative bodies. These indicators constitute the technical and environmental restrictions of the study area (Table 1).
After excluding the restricted areas, positive indicators were adapted to assess the suitability of locations for a photovoltaic solar plant. This was based on five criteria (climate, topography, environment, location, and economic) and included ten factors (direct solar radiation, temperature, slope, orientation, land use, NDVI, distance to transmission lines, distance to water resources, distance to main roads, distance to urban areas, and project cost). Table 5 presents all of the evaluations. jms.ccsenet The random positive re the matrix v ij ,= w ij ⊗ n ij j=1, 2, …, n;i=1, 2,…, The ideal positive (PIS, A +) and negative (NIS, A-) fuzzy solutions are determined according to Equation 6 and Equation 7, respectively.
The distances to PIS (D+) and to NIS (D-) are calculated by Equation 8 and Equation 9, respectively.
Finally, the CC i approximation coefficients are calculated for each of the evaluated alternatives, according to Equation 10. The CC i value varies between 0 and 1. The closer to 1, the higher the priority of the alternative. From this, the final ranking of the alternatives is defined from the CC i values.

Results
This section presents the results of the potential sites for sustainable development of solar photovoltaic plants appropriate to the reality in the State of Pernambuco.

Classification of Input Criteria
The five-point Likert scale was used to classify the sub-criteria. The maturity levels considered were: 1-very low; 2-low; 3-moderate; 4-high; 5-very high (  The climatic criterion was considered one of the most important in the evaluation of the spatial aptitude for the development of solar projects, and for that reason it has the highest weighted coefficient in this study. Other criteria (topography, environment, and location) were considered to be of lesser importance, as they can be adapted by human intervention on the ground. For the economic criterion, the inverter technology was not considered and the cost of the project was not considered in the analysis, as the criterion presented a consistency ratio of 0.00% and can be adapted according to available resources. DB ( Figure 6).

Multi-criteria Analyses-Fuzzy AHP-TOPSIS
The results of the previous step gave rise to a list of criteria and sub-criteria to be hierarchized in order to apply the model. The five-point Likert scale was used to classify the sub-criteria.

Fuzzy-AHP
The order of importance of the criteria based on expert assessment is shown in Table 6.  Once the decision rule for each scenario was established, the pairwise comparison be-tween the criteria could be performed. The consistency ratio (CR) for each specialist, the highest eigenvalues (λmax) found, and the consistency indices of the judgments (CI) are shown in Table 7. With all of the criteria organized hierarchically, the process of obtaining the vectors from the priority and consistency assessment was carried out for the criteria and sub-criteria using the fuzzy AHP method. As the CR value is less than 0.10 for all specialists, the estimated values of the sub-criteria are also confirmed to be consistent. Final weights are shown in Tables 8 and 9, respectively. In order to unify the weightings for the obtained criteria, a homogeneous aggregation was performed, that is, all specialists were considered to be equally important in the decision. The arithmetic mean was used as an aggregation measure. The criteria weightings obtained from the homogeneous aggregation are shown in Table  10.

Fuzzy-TOPSIS
Once the weightings for the criteria that influence PVP location were defined, the alternatives were evaluated using the fuzzy TOPSIS method. The evaluation matrix was normalized, to arrive at the normalized and weighted matrix , using Equation 4 and Equation 5. With the standardized and calibrated fuzzy values, the distances between these data and the ideal positive and negative fuzzy solutions were calculated, which are the maximum and minimum values for each criterion and sub-criterion. Then, the distance matrices A+ and A− were generated using Equations 6 and 7, respectively, and the CCi of the criteria and sub-criteria was calculated. The results are presented in Table 11 and Table 12. The best alternatives are shown in decreasing order according to CCi, The results defined with this coefficient will be counted in ArcGis 10.3. le to combine ereby obtainin se different cr d ( Figure 15). Four areas were selected for evaluation. Area 1 is located in the Sertão (Figure 15). Assessing the influence of the weighting of the thematic layers in the definition of the suitability map, it was noticed that, the solar radiation layer presented values above 5.3 kWh/m2·day throughout its territory. Solar radiation had a weighting of 42%. The land slope layer shows that the slope of the municipality's land is less than 5% in the central and northern portions. The slope layer was the second in order of importance, with a weighting of 26%. Regarding land use, most of Salgueiro's territory is composed of land that is unfavorable for agriculture, which is the class that most favors the installation of PVPs. This parameter has a weighted value of 18%, third in order of importance. Finally, in relation to the infrastructural aspects for interconnection to the grid, the area is close to 69 kV transmission lines and the BR-232 highway.
Areas 2 and 4 are located in the Sertão and São Francisco mesoregions ( Figure 15). These areas also had solar radiation values above 5.3 kWh/m2·day (annual average) throughout their territory. This terrain also has slopes of less than 5% and was close to 69 kV transmission lines and state highways. The big difference with the first area is due to the land use, with most of the territory being composed of land having favorable agricultural potential, disadvantaging the installation of PVPs. Mainly in the São Francisco mesoregion, there are several farms with mango and grape plantations, responsible for the region's economic development and for supplying the international market.
The lowest suitability class, area 3 ( Figure 15), was found in the Recife Metropolitan Area, where the solar radiation is between 3.5 kWh/m2·day and 4.8 kWh/m2·day and the slope of the terrain is greater than 5%. These regions are closer to urban areas and electrical and road infrastructure. They also are for the installation of photovoltaic plant projects, considering the distances to urban areas, but when it comes to land use, there are restrictions due to the presence of remnants of the Atlantic Rain Forest.
To better refine the research methodology, a minimum area limit, based on the generation capacity of the plant, can be applied to systems connected to the grid on a large scale. The land occupation factor can have separate values for rural areas (or built areas) and other available surfaces. However, when working on a regional scale, it is difficult to determine the exact value of the land occupation factor for built-up areas. In addition, the population density may be different for systems connected to the network on a large scale. In other words, all inhabited pixels (population density > 0 inhabitants/km2) can be classified as the most suitable (score 5), and non-inhabited pixels can be classified as the least suitable (score 1).
A sensitivity analysis can be carried out with regard to the technical characteristics of the chosen technologies. In this study, it is proposed that large-scale systems connected to the network be located away from cities (that is, preference should be given to greater distances from urban areas). The objective is to avoid restrictions on urban development and to choose places with lower land value. However, a maximum distance limit can be applied in order to reduce losses during electricity transmission.
The uncertainty discussed in the methodological choices demonstrates a need for more research and dialogue (including academia, legislators, and other stakeholders actively involved in the deployment of photovoltaic plants) with respect to approaches that estimate potentials for large-scale solar energy generation (i.e., at a national or regional level).
Estimates should not be limited to only geographic and technical potentials, but should include an economic evaluation, as this information is essential for investment planning. For this reason, a more detailed analysis of this criterion is suggested to analyze any possible flaws in the methodology.

Discussion
For the sustainable development of a region, it is extremely beneficial to identify areas suitable for the deployment of solar photovoltaic plants to optimize the planning of transmission lines, strengthen the solar energy market, and develop master plans for the production of solar energy, among others. The evaluation of different alternative locations in planning is a preliminary and decisive step in creating maps with suitable and economically-viable locations for the use of specific technologies.
In this study, spatial analysis began from the definition of the criteria with which the areas for solar photovoltaic plant installation would be selected. The choice of criteria was based on the available literature on this subject: essential criteria imposed by government legislation and expert opinions on the performance of solar plants. The chosen criteria were then classified into three categories: technical, environmental, and social. Site selection involves screening a large geographic area to select a limited number of alternatives. The locations identified by the screening must be evaluated later, which will lead to finding the most suitable location among all available alternatives.
The principal contribution of this article was its methodology that combines GIS with a multi-criteria decision analysis method (fuzzy AHP-TOPSIS). The main advantage offered by this integrated approach is to be able to use a GIS to collect and organize the information that will be provided to the AHP. The application of the fuzzy AHP-TOPSIS for the locations of PVPs proved to be an accurate and adequate tool for hierarchizing the criteria and sub-criteria. The results showed that Pernambuco can be considered a highly suitable state for PVP. The areas having very high potential represent 12.61% of the entire state. The percentage of high and medium potential is 22.75% and 18.48% whereas inadequate sites with low and very low potential represent 0.12% and 0.31%, respectively.
GIS-based methods can be applied differently according to the scale of the study area, as well as the type of solar energy conversion technology. The results show that solar radiation maps can be used as banks of spatial data useful in spatial and temporal analysis of solar resources. In particular, site assessment using GIS is useful to support decision-making on a regional scale, and it is necessary to consider the economic, environ-mental, technical, social and risk factors, in addition to solar radiation. These factors can be used to exclude inappropriate regions using map algebra. The functions of GISs can extend beyond being a data and visualization inventory to sophisticated modeling, evaluation, and interdisciplinary studies of solar energy.