Stokastik Programlama Yaklaşımı ile Elektrik Üretim Endüstrisinin Modellenmesi

Abstract

This thesis focuses on the development of a framework and mathematical models to formulate generation expansion planning (GEP) models that include multiple objectives; by means of appropriately considering conventional and renewable electricity generation technologies faced with demand growth uncertainty. The growing need to simultaneously analyze a range of energy-economy-environment interactions, assess possible future impacts of energy and environmental policy decisions, consider uncertainties related with long-term electricity demand projections, and further optimize costs and emissions of main air pollutants; requires the development of flexible GEP models. To address emerging concerns, models that represent multi-period, multi-objective GEP problems to optimize different objective functions concurrently while satisfying a wide-range of technical and reliability constraints under demand growth uncertainty are proposed to serve as a flexible structure for decision-makers in evaluating the impacts of various scenarios. In this study, presented models include multi-period, multi-objective GEP and stochastic, multi-period, multi-stage, multi-objective GEP models to determine optimal expansion alternatives while minimizing total costs, CO2, NOx, (and SO2) emission objective functions concurrently under a set of economic, technical, operational and environmental constraints. The models are formulated as mixed-integer linear programming problems, and multi-objective mathematical programming approaches are implemented to obtain Pareto-optimal solutions. A fuzzy decision-making method is used to select the most preferable compromise solution for decision-makers. One of the key challenges of stochastic programming is to represent the continuous stochastic processes by generating a finite number of scenarios. To model demand growth uncertainty, the actual annual electricity demand was verified to ensure consistency with the correlated Geometric Brownian Motion processes. A finite number of scenarios with a multi-stage tree structure is constructed by using a matching method, based on moments of marginal distribution functions, and simulations carried out by using Monte Carlo simulation methods. Cost and technical performance estimates for conventional and renewable power generation technologies utilized by different energy models are compared and synthesized to satisfy the reliable data needs for such models. Finally, a real case study based on Turkish electricity supply industry-planning problem shows the relevance of the proposed models and performance of solution methods. That said, an overview of Turkish energy demand growth, supply and associated emissions for the year of 2030 are also examined for different scenarios. Using the energy-economic modeling platform LEAP; primary and final energy demands, electricity generation projections, and emissions of the main air pollutants (greenhouse gases, SO2, NOx, and CO2) are all studied and analyzed. In conclusion, considering the simulation outcomes, this thesis provides an appropriate tool for policy makers to analyze various scenarios and technology options considering energy-economy-environment interactions under demand uncertainty based on deterministic and stochastic models. The models can serve as an end-to-end decision-support tool to evaluate different generation expansion alternatives and to quantify the long-term implications of policy decisions on energy systems, energy economics, and the environment; while concurrently achieving a path for sustainable development.