Scientific Ethics Compliance

Full content available in the PDF download.

Guideline Compliance

Full content available in the PDF download.

Acceptance and Approval Minutes

Full content available in the PDF download.

Preface

Full content available in the PDF download.

Abstract

This dissertation investigates optimal pathways for Turkey’s electricity sector transition to renewable energy using reinforcement learning (RL) and long-horizon demand forecasting. Six RL algorithms are compared — PPO, A3C, DQN, SAC, TD3, and TRPO — in a custom OpenAI Gym environment representing Turkey’s electricity system over a 2023–2053 planning horizon. Key findings show that PPO achieves an 18.6% reduction in total system cost relative to a Business-As-Usual scenario, while LSTM-based demand forecasting achieves R²=0.9799 on a 5-year horizon. Sensitivity analysis identifies carbon pricing and battery storage costs as the critical policy levers for cost-optimal transition pathways.

Abstract (Turkish)

Full content available in the PDF download.

List of Symbols and Abbreviations

Key abbreviations used in this dissertation: ENSM (Enhanced Neural S-curve Model), MAPPO (Multi-Agent PPO), LCOE (Levelized Cost of Energy), PICP (Prediction Interval Coverage Probability), SHAP, CTDE, TEİAŞ, NDC, YEKA.

Full content available in the PDF download.

Extended Summary

The extended summary provides a comprehensive overview of the dissertation findings, methodology, and policy implications (PDF pp. 34–68).

Full content available in the PDF download.

Final Reflections

Full content available in the PDF download.

Downloads

Document Standards (Pandoc)

---
documentclass: report
fontsize: 12pt
mainfont: "Times New Roman"
papersize: a4
geometry:
  left: 3.5cm
  right: 2.5cm
  top: 3cm
  bottom: 2.5cm
linestretch: 1.5
parindent: 1.25cm
number-sections: true
toc: true
toc-depth: 3
lof: true
lot: true
figPrefix: "Figure"
tblPrefix: "Table"
eqnPrefix: "Equation"
chaptersDepth: 1
bibliography: references.bib
csl: apa.csl
link-citations: true
colorlinks: true
citecolor: "0369A1"
---

Ethics Committee Approval

Full content available in the PDF download.