مطالعه و تحلیل مکان گزینی نیروگاه های خورشیدی در استان آذربایجان شرقی

IntroductionEnergy is one of the primary human needs to pursue its goals. Traditional energy sources such as fossil fuels are exhausted and cannot be accessed everywhere. They also pollute the environment and cause irreparable damage. Decision-makers and planners should replace traditional energy sources with non-harmful environmental sources. Renewable energy sources such as solar energy, wind energy and Biomass can be a potential alternative to traditional energy sources. Solar energy is converted into electricity using solar panels. but one of the most important things to be considered in this case is finding an optimal location for the deployment of solar panels. Geographical information system (GIS) enables researchers to perform various spatial analyses in a graphical environment and allow appropriate visualization of the results. There are various important factors and criteria that affect the efficiency and cost-effectiveness of solar panels so it is necessary to compare these factors together. Multi-criteria decision-making methods are one of the ways that allow researchers to make a preferential comparison at the same time on different criteria affecting a decision. The best-worst approach is one of the multi-criteria decision-making techniques. B-W is a new and efficient method and can effectively determine the priority of criteria on each other. Also Artificial Neural Network (ANN) capabilities can be used to solve site selection problems because it has been proven that ANN is capable of solving many decision-making problems.Materials and methodsEast Azarbaijan province in northwest of Iran which is a mountainous region with no oil and gas reserves was selected as the study. The fuel needed for domestic and industrial uses of the province is supplied by pipelines of oil and gas extending from southern Iran. In this study eight climatic, orographic and economic criteria including GHI, distance from cities, distance from main roads, altitude, slope, slope direction, mean annual precipitation and mean annual sunshine hours were used. The GHI Criterion Layer for East Azerbaijan Province was extracted from the Iran Layer. Height, slope and aspect layers were also produced using the DEM product of ASTER sensor. The layers of sunshine hours and mean precipitation were also interpolated using the values of synoptic stations for the entire province using IDW algorithm. The values for the distance from the cities and the distance from main roads raster layers were also obtained using the Euclidean distance function. Since the defined criteria had different units and nature the criteria values were normalized between ۰ and ۱ using different functions. The best-worst method was used for weighting the criteria and determining the relative priority of the criteria. After calculating the weight of each criterion, the initial suitability values were obtained by multiplying the criteria by the weight obtained. Then, ۷۵۱ samples were taken from the whole study area for ANN training. The values of each sample in the criterion layer as “input” and the values obtained by multiplying the layers in their weight as “Target” entered the ANN. Also, according to the laws and regulations, some lands such as rivers and their margins, cities and their suburbs, protected areas and fault lines are places where solar power plants cannot be installed. Hence, in this study, these places were identified as restricted sites.Results and discussionThe BW results showed that the “GHI” criteria and the “distance from city” have the highest weight and the “average precipitation” criterion has the lowest weight. Levenberg–Marquardt algorithm was used to train the network. ۷۵% of the samples were allocated to network training and ۱۵% to testing and ۱۵% to validation. The number of hidden layers was set to ۱۵. After ۳۴۴ iterations mean squared error value for the network training data reached its optimum. Also, the regression value between the network inputs and the Targets was ۱, indicating that the network training was performed best. After the network training was completed, the output values were obtained for the entire study area using the “net(y)” function. The results of ANN showed that ۲,۶۲۲,۹۷۵ hectares of the province have a suitability of less than ۰.۵ for solar power plant deployment while only ۹۲۰۷۵ hectares of land in the province have a suitability rate of over ۰.۷۵. To make the results more accurate, the binary constraint layer was applied to the results obtained from the neural network. The results showed that this time the proportion of lands that have suitability of less than ۰.۵ reached ۳,۳۱۵,۰۹۳.۷۵ hectares and the proportion of land suitability rate of over ۰.۷۵ included only ۶۹۷۶۲ hectares of land area. This indicates that some of the initially high suitability locations for solar panels were located in areas where it is not reasonably possible to deploy solar power. So relying on the values obtained for "land suitability" without regard to the constraints in the areas may be misleading.ConclusionIn this study, suitable locations for installation of solar panels were identified using a combination of multi-criteria decision making methods and artificial neural network. The best-worst method that is a novel and efficient MCDM technique was used for weighting the criteria. The results showed that the areas that are located in the proper distance with the cities and the main roads, provided they have a high GHI value, are more suitable for other solar plant deployments. In order to improve the results, constraints including rivers, cities, protected area and fault lines were defined and applied to the land suitability layer. In general, the western regions of the province were more suitable than the eastern regions, due to the proximity of the cities to each other, the high density of the main roads, the suitability of elevation and slope and aspect, as well as the high amount of GHI and sunshine hours and low amount of mean precipitation. Due to the high accuracy of the results, we can say The procedure used in this study can also be used in other mountainous regions of Iran to produce reliable results.KeywordsArtificial Neural Network (ANN), Multi-Criteria Decision Making (MCDM), solar energy, Renewable Energy (RE), Geographical information system (GIS)

Keywords:

انرژی خورشیدی , انرژی های تجدیدپذیر , تصمیم گیری های چندمعیاره (MCDM) , سیستم اطلاعات جغرافیایی (GIS) , شبکه عصبی مصنوعی (ANN)