Feasibility Study for Poverty Alleviation by Energy Production Through Photovoltaic Panels in Underprivileged Areas of Iran

The purpose of this research was to alleviate deprivation in underprivileged rural areas of Iran through energy production through photovoltaic panels. First, the poverty map of the country has been analyzed along with the sunlight gain potential map of Iran. 79 villages with extremely high and high deprivation conditions were identified. To analyze these areas and prioritize them for decision-making in energy sector policy, the areas were categorized into four clusters by the K-means machine learning algorithm. According to the clustering results, the priorities for deprivation elimination through energy production have been identified. Also, the income of each village from two sources was calculated according to the existing tariffs: First, a 5 kW photovoltaic panel for each household and second, a one-megawatt power plant located in the barren lands around the villages. The results indicate that Sistan & Baloochestan Province is the top priority in the application of photovoltaic panels for energy production. For instance, Bampasht village with a deprivation index of 0.540 and energy production of 9.37 MW per year from a solar power plant, has obtained the highest rank among the villages for deprivation alleviation by energy production through photovoltaic panels.


Introduction
Due to the increased accessibility, development and Iran's potential in the application of renewable energy sources, photovoltaic panels have gained interest among the stakeholders in the country. According to the literature, Iran receives 1050 hours of sunlight during summer and 500 hours during winter [1] and the annual solar radiation is at least 1800 kWh/m 2 [2]. On the other hand, the installation of photovoltaic panels(PV) is one of the main recommended procedures for poverty reduction in literature [3]. Where the geographical areas possess a high potential for sunlight gain, the application of PV would be beneficial for poverty alleviation and clean energy production. Accordingly, to build up a sustainable framework in this field, two main steps need to be taken. First, identification of the potential areas with high poverty issues; and second, prioritization of the areas for higher efficiency. The first step has been taken in previous studies and the results support the idea of Iran's high potential for the application of PVs in energy production. comprehensive studies have hardly been conducted for the second step making this subject underexploited. Accordingly, this research proposes a method for prioritizing the intended areas to be equipped with PV panels technology and energy production with the initial goal of poverty reduction.
There have been several attempts in literature for the development of renewable energy sources adaptation with the aim of poverty reduction. These projects can be categorized into three main groups: 1) Energy production for underprivileged areas. 2) establish businesses based on energy sales for deprived areas and 3) financial support for the households in deprived areas to be equipped with PV technology.
The projects of PV systems in South Korea, the Rural Network of Renewable Energy in Indonesia and UNDP biomass in Bosnia are examples of the first category [4,5]. SEPAP project in China has been developed for poverty reduction in the second category [6]. Romania has formed a framework based on governmental subsidies for buying PV panels, geothermal energy and wind energy systems for households [7]. Most projects in Iran can be categorized in the third group where the financial support of the government is given to the related businesses. The projects of Barkat Aftab and Jahad Roshanayee are examples of these projects which were supported and planned by the Imam Khomeini Committee and the Program and Budget Organization of Iran. According to the available reports, 1000 houses were equipped with 5 kW photovoltaic panels, and the income of each household in 2019 was reported to be 2.5 million Tomans per month in the Barkat Aftab project. More than 20000 power plants are planned to be built in the Jahad Rohanayee project [8,9].
In this study, two factors of deprivation index and the potential of studied areas have been considered for the initial prioritization of areas according to the collected data. The villages with the potential to produce more energy and of higher deprivation status are in higher priority.
By reviewing the projects carried out in Iran and their reports, there will be a hope for more effective implementation of approaches leading to poverty alleviation through the production and sale of energy using photovoltaic panels. It should be noted that in previous works a research gap is observed when it comes to the development of a framework for prioritizing potential areas.

Material and Method
The method used in this research is the use of one of the unsupervised machine learning algorithms called kmeans, which categorizes areas with similar characteristics, speeding up the decision-making process about their prioritization. Accordingly, collecting the required data in the first step and application of the algorithm in the second step are the two main steps to achieve the correct prioritization of deprived areas. Therefore, the methodology of this research includes 1) matching the deprivation distribution map and the map of potential for receiving sunlight to find the proper areas 2) determination of the amount of energy production with photovoltaics at any selected points 3) Description of the obtained data 4) Designing the unsupervised machine learning model.
Based on the deprivation distribution map, 79 zones with high and extremely high deprivations were selected with the combined metric of deprivation. The metric categorizes the areas based on a comprehensive rating system which includes economic, cultural and social indicators. To calculate the amount of produced energy through two power plants of 5 kW and 1 MW, some places near residential areas in barren lands were selected. The energy production amount was then calculated with the web-based software of Globalsolaratlas [10].

Data Description
The obtained data from the previous analyses were drawn in the form of scatter plots with an x-axis of deprivation index and the y-axis of energy generation amount from PV panels in kilowatts to get a better view of the data. Besides, before clustering and in the preprocessing stage, the scales of deprivation index and energy production were standardized by the following equation:

Clustering Process
The MLA was designed concerning both main factors that were considered in the analyses: poverty factor and energy generation by PV panels. The purpose was the determination of similar areas and prioritise them to implement poverty alleviation policies by application of photovoltaic panels. Clustering is one of the unsupervised machine learning approaches for classifying multi-dimensional data which can be used to identify hidden patterns in the dataset [11,12]. A cluster is a dataset having a common internal pattern which is different from the others.
K-means is one of the most famous and well-known clustering techniques [13]. The distance of each data from the primary batch centres is found and the data are placed in the primary clusters. This operation is then repeated until there is no change in the members of the clusters created. The formula of the k-means algorithm (eq.2) is calculated by considering the Euclidean distance as follows: Where K is equal to the data in the cluster and μ is the average of the related cluster. One of the ways to find the optimal number of clusters is to use an Elbow diagram. This diagram indicates the sum of the squares of the points' distances from the centre of their categories. To evaluate the clusters made by the k-means algorithm and to confirm their differences statistically, the ANOVA test and its nonparametric equivalent Kruskal-Wallis test can be used. Also, to examine the differences of clusters in groups more accurately, TUKEY POST HOC and DUNNS tests have been applied [14]. For this reason, the normality of the data was first measured by the Chisquare test. The energy production index is normal, but the deprivation index does not follow a normal distribution. Therefore, ANOVA and TUKEY POST HOC tests were used to determine the differences between the clusters in the energy production index, and Croxall-Wallis and Dunn`s tests were used to determine the significant differences in the poverty index.

Matching the Maps
The potential of each deprived area in renewable energy production through PV panels along with the poverty index is presented in Figure 1.
Some areas are in complete deprivation, but in terms of energy production potential from photovoltaic panels are among the best areas in the world. Figure 2 is implemented by the yellow-brick library in Python, and the optimal number of 4 clusters is selected for it [15]. After finding the acceptable amount of clusters using the k-means algorithm, the data are divided into 4 categories based on the amount of deprivation and the amount of energy production from PVs ( Figure 2).

Evaluation of clusters by ANOVA and Kruskal-Wallis test
Cluster evaluation is necessary to ensure the independence of clusters from each other to implement specific policies for each cluster and prioritization. According to the Kruskal-Wallis test, the 4 clusters have significant differences in poverty indices. The villages located in at least one of the clusters have different deprivation rates from the villages located in the other clusters (can be more or less) and the difference is statistically significant. The same results are observed in the ANOVA test where the energy production indices are analyzed. By examining the clusters in pairs using the TUKEY POST HOC test, it is observed that there is a significant difference in all clusters and the PV variable except for two. Therefore, in prioritizing the villages of those two clusters, the amount of energy production can be the same and prioritized based on the deprivation index, which means that any cluster with a higher deprivation rate has a higher priority for installing photovoltaic panels to eliminate deprivation. A summary of the studied areas and their prioritization results are presented in table 1.
The results indicate that Sistan & Baloochestan Province is the top priority in the application of photovoltaic panels for energy production. For instance, Bampasht village with a deprivation index of 0.540 and energy production of 9.37 MW per year from a solar power plant, has obtained the highest rank among the villages for deprivation alleviation by energy production through photovoltaic panels.
Two successful projects of Barakat Aftab and Jihad Roshanai, which have reportedly planned to build 3,000 home solar power plants in deprived areas in the first phase, have been mentioned as successful projects to eliminate deprivation by producing energy from photovoltaic panels in Iran. According to the project reports [8] Qazvin, Ardabil, Zanjan and Khorasan Razavi provinces (with a share of 600 houses in the first phase and 2000 houses in the 2 nd ) have been considered. While the most deprived province of Iran, Sistan and Baluchestan, has a share of 100 of this sector. According to the findings of this study, the Bampasht section, with a deprivation index of 0.540 and 9.37 MW per year of energy production from a solar power plant, has the highest rank among the villages in terms of deprivation along with energy production potential. The priorities are allocated to almost all areas of Sistan and Baluchestan province.
It can be concluded that it is better to pay special attention to the priorities of each region in the forthcoming projects; because the deprivation alleviation process can be completed faster with full knowledge of the characteristics of each region.

Conclusions
Investment challenges in general (not just for disadvantaged areas) and the willingness of investors and business owners should be analyzed for such subjects of study. Since reliance on government plans can be completely overshadowed by budget cuts and decisions by top stakeholders, the need to study the feasibility of mechanisms such as cooperatives and participatory investments should be considered. Issues such as the low growth rate of the renewable energy industry in Iran, economic problems, the lack of governmental facilitating mechanisms, the high initial cost of operating solar power plants, lack of familiarity with their mechanism and low education and knowledge about the use of renewable energy, and the lack of sufficient experience in this field due to the cheapness and prevalence of fossil fuels are of the obstacles in the general use of PV panels [8]. In addition to the above, failure to improve the existing infrastructures for businesses operating in this field (due to unstable economic conditions) to facilitate the process of using renewable energy leads to long-term returns of investments (over 6 years), which is not satisfying.
For further research, indicators that can be effective in prioritizing deprived areas may include sociocultural ones (acceptance of the use of renewable systems in indigenous peoples, population and unemployment rate), political issues (government operational policies), Security (theft statistics in different areas and its impacts), geographies (how to access deprived areas and impassability of routes), infrastructure (existence of the electricity network to sell electricity from renewable sources to the electricity network).