Democritus University of Thrace (DUTH)
Πολυκριτήρια ανάλυση καταλληλότητας περιοχών με τη συνέργεια μοντέλων χωρικής παρεμβολής και μηχανικής μάθησης
Abstract
dc:descriptionLand use suitability analysis is a complex spatial multicriteria problem. It requires the collaboration of Geographic Information Systems (GISs) with Multicriteria Decision Analysis (MCDA) methods for the evaluation and the ranking of locations, as well as the provision of suitability index mappings. The current PhD thesis focuses on the PROMETHEE II multicriteria method and investigates, proposes, and implements methodological approaches to overcome PROMETHEE’s computational limitations in raster driven GIS-assisted suitability analysis. Spatial multicriteria decision-making frameworks are developed to conduct suitability analysis and to determine the most efficient locations for wind energy investments. The novelty of the proposed frameworks is the implementation of the PROMETHEE method to a sample of locations to produce suitability indexes and afterward the generalization of suitability indexes to the area of interest by using the predictive capabilities of spatial interpolation and machine learning models. A case study in the Thrace region for onshore wind farms siting is selected for modeling suitability mapping. The combined suitability results show that a significant number of locations are suitable for wind farm installation. The proposed spatial frameworks take advantage of the PROMETHEE method’s ability to indicate areas with positive and negative outranking flows and assist energy planners and wind developers in identifying efficient locations for wind farm development. The dissertation is organized into six (6) chapters with the following content: In Chapter 1, land use suitability analysis definitions are introduced and the fundamental role of GIS to support such analyses is highlighted. Types of spatial data and raster overlay operations are described. The synergy of MCDA methods with GIS tools is highlighted and a literature review regarding the implementation of PROMETHEE method in a GIS environment is presented. In Chapter 2, the main features of PROMETHEE method and “the alternatives profile” variation are described. The synergy with the multicriteria Analytic Hierarchy Process (AHP) is proposed and the method is presented with an educational example for determining the criterion weights. Chapter 3 focuses on spatial prediction models for suitability index estimations. The Inverse Distance Weighting (IDW) spatial interpolation method is presented and illustrated through an educational example. Machine learning predictive modeling techniques are introduced. Moreover, the general classification and regression framework is analyzed along with performance evaluation measures. Furthermore, the k nearest neighbor (kNN) and the Distance weighted k Nearest Neighbor (DWKNN) are presented for classification and regression problems. SVMs are discussed for linear and non linear classification problems as well as for multiclass problems. Support Vector Regression (SVR) is presented for numerical values prediction. In Chapter 4, methodological approaches are proposed to overcome the computational limitations of PROMETHEE II in raster driven suitability analysis. The first decision-making framework combines GIS technology with PROMETHEE II and the IDW spatial interpolation technique to conduct onshore wind farms suitability analysis of the Thrace region of north-eastern Greece. The second one proposes a novel decision-making framework for GIS-based land-use suitability analysis that combines PROMETHEE II with IDW and AHP, along with a discussion for a variety of net preference flows estimation options using both total and per criterion net preference flows. The spatial framework is illustrated through an indicative hypothetical example in a specific part of the Thrace region. The result of the research shows that IDW prediction is sensitive to the loss of information phenomenon as the generalization is formed exclusively by the coordinates of the pixels. The third spatial decision-making framework integrates GIS technology with PROMETHEE II and k nearest neighbor (kNN) machine learning (ML) prediction tasks. The framework offers two GIS, PROMETHEE, kNN variations according to whether land use suitability analysis is implemented as a classification or a regression problem. Chapter 5 focuses on wind energy and the main features of onshore and offshore wind turbines are presented along with the advances in their design development. The licensing process for wind farms in Greece is analyzed. Wind farm siting decision criteria and exclusion zones are described according to the national spatial framework. A suitability analysis case study is being implemented for wind farm siting in Thrace region. The results of the first methodological framework are presented and discussed (PROMETHEE IDW). The fourth decision-making framework is proposed, that combines AHP, PROMETHEE II and machine learning regression models. Models are developed and evaluated using the kNN and SVR to predict the most efficient onshore wind farm locations generating suitability index mappings. The results of the fourth framework are presented and discussed. Chapter 6 provides PhD thesis conclusions and highlights issues for further research.
Degree
thesis:*- Grantor dc:publisher
- Democritus University of Thrace (DUTH)
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Sotiropoulou, Kalliopi
- Σωτηροπούλου, Καλλιόπη
Subjects
dc:subject × 36- Ανάλυση καταλληλότητας περιοχών
- Γεωγραφικά Συστήματα Πληροφοριών (ΓΣΠ)
- Πολυκριτήρια ανάλυση αποφάσεων
- Μέθοδος σχέσεων υπεροχής PROMETHEE II
- Στάθμιση της αντίστροφης απόστασης
- Χωρική παρεμβολή
- Μοντέλα μηχανικής μάθησης
- k Πλησιέστερος Γείτονας
- Παλινδρόμηση διανυσμάτων υποστήριξης
- Χωροθέτηση χερσαίου αιολικού πάρκου
- Suitability analysis
- Geographic information systems (GIS)
- Multicriteria decision analysis
- PROMETHEE II outranking method
- Inverse distance weighting
- Spatial interpolation
- Machine learning models
- k Nearest Neighbor
- Support vector regression
- Onshore wind farm siting
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Φυσικές Επιστήμες
- Διεπιστημονικές εφαρμογές επιστήμης ηλεκτρονικών υπολογιστών
- Πληροφοριακά συστήματα
- Οικονομικά και Επιχειρήσεις
- Κοινωνικές Επιστήμες
- Επιστήμη λήψης αποφάσεων
- Διοικητική Επιστήμη και Επιχειρησιακή Έρευνα
- Computer and Information Sciences
- Natural Sciences
- Computer Science Interdisciplinary Applications
- Information Systems
- Economics and Business
- Social Sciences
- Decision Sciences
- Management Science and Operations Research
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/54790