Sustainable Planet: Site Suitability Analysis for Solar Photovoltaic Power Plants Using Fuzzy MCDM and Network Science Based Approaches
Restricted (Penn State Only)
- Author:
- Almasad, Abdullah
- Graduate Program:
- Industrial Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- December 14, 2021
- Committee Members:
- Soundar Kumara, Chair & Dissertation Advisor
Saurabh Basu, Major Field Member
Hui Yang, Major Field Member
Greg Pavlak, Outside Unit & Field Member
Thamer Alquthami, Special Member
Steven Landry, Program Head/Chair - Keywords:
- Geographical Information System
Multi-Criteria Decision Making
Fuzzy Analytical Hierarchy Process
Site Suitability Analysis
Graph Theory - Abstract:
- The implementation of renewable energy sources has significantly increased in the past years due to the increase in the global energy demand and the awareness of the fossil fuel risks on the environment. Solar energy has many advantages over other renewable energy sources with inexpensive installation costs and location flexibility. Due to its ability to absorb both direct normal irradiance (DNI) and diffuse horizontal irradiance (DHI) and having low production cost, Photovoltaic has become the most common solar energy technology used in the world. The purpose of this research is to identify the suitable locations for implementing utility-scale photovoltaic (PV) farms. However, the evaluation process is greatly influenced by technical and economic factors. To address this problem, important questions that needed to be asked are: What are the factors that influence the performance of the solar PV panels? How can we identify the degree of the internal impact of each factor while minimizing the uncertainty in the decision? What is the appropriate method to evaluate and identify the suitable locations for implementing utility-scale solar PV farms? To answer these questions, we have divided the research into three problems. For the first problem, we first address the gap in the literature regarding the factors affecting the performance of the solar PV panels. There are multiple climatic factors that have not received attention in previous solar site suitability studies, despite their proven effect on the performance of PV panels. In addition to the common climatic factors incorporated in the previous studies (solar irradiance and temperature), we further investigate the influence of precipitation, air pressure, surface albedo and relative humidity toward the final PV site suitability map. Furthermore, economic factors are incorporated to ensure minimization of the construction costs of solar PV farms. The economic factors consist of slope, aspect, major cities, highways, transmission grids and power lines. In addition, we address the weighting process for these evaluation factors. Generally, not all factors have the same effect toward the suitability results. Therefore, it is important to assign the appropriate weights that reflect their true impact. The process of assigning weights to the factors is subjective as it is hugely influenced by experts’ knowledge. To mitigate the uncertainty and increase the flexibility concerning the weighting process, we utilize the extent analysis method on fuzzy analytical hierarchy process (AHP). For the process of location decision-making for solar PV farms, Saudi Arabia is used as a case study. We utilize a geographical information system (GIS) and multi criteria decision making (MCDM) method to evaluate the sites with respect to the selected factors. Preference Ranking Organization Method for Enrichment Evaluations (PROMETHEE II), a well-known MCDM method, is used to carry out the evaluation process. The resulting suitability map shows that Saudi Arabia has huge potential for implementing solar PV farms with approximately 376,623 Km2 (65.1%) of the total area considered, “most to highly suitable”. In addition, a validation of the model’s prediction accuracy was conducted through an evaluation of the suitability map with respect to the future solar PV projects that Saudi Arabia is developing. The results showed that 90.6% of the future projects fell within, “most and highly suitable” areas provided by PROMETHEE II suitability map. Furthermore, a sensitivity analysis was carried out by using different preference functions and higher weights for the economic criteria to examine the effect of economic factors toward the suitability results. The second research problem addresses the issue of the uncertainty in the suitability results when using MCDM methods as they do not have an appropriate way to group the sites into the suitability categories. Thus, a second approach is used to overcome the partitioning problem. A graph-based framework is developed where the sites are represented as the nodes and the edges in the graph represent the similarity among the sites. We propose a new similarity measure that has been used to construct the graph. Two graph-clustering algorithms, spectral and Markov clustering, are used to compare the results of the different similarity measures. The clustering results then ranked using Ideal point method to produce the optimal suitability layer. An-in depth validation of the clustering results shows the significance of the approach in identifying the suitable locations. The third problem aims to utilize the given suitability information from the existing solar PV projects to develop an adaptive site suitability analysis that would predict the suitability information for the rest of the areas. The challenging part is that there are only limited numbers of existing solar PV projects; thus, a semi-supervised graph neural network approach is utilized to carry out the classification process. The graph constructed in the second problem is used and only small number of the labels are available to train the model. The model has 77% accuracy and presented significant results in the suitability map. Solar Energy is an important component of sustainability and in this thesis an attempt is made to derive alternate site suitability selection models. Among the improvements, the third approach (Graph Neural Networks) can lead to better results if we can have more labeled data. This work is generalizable for any PV location problem and the current results using Saudi Arabia as a case study are readily usable by Saudi Arabia’s governmental policy makers.