Thesis Title: Stochastic Planning and Operation in Energy Systems 

Advisors:  

Dr. Valerie Thomas (advisor), School of Industrial and Systems Engineering, Georgia Tech 

Dr. Shijie Deng (co-advisor), School of Industrial and Systems Engineering, Georgia Tech 

Committee members:  

Dr. Andy Sun, School of Industrial and Systems Engineering, Georgia Tech 

Dr. David Goldsman, School of Industrial and Systems Engineering, Georgia Tech 

Dr. Beste Basciftci, Industrial Engineering Program, Sabanci University 

Date and Time: Wednesday, July 14, 2021, 10:00 am (EST) 

Meeting URL: https://bluejeans.com/294450929/7146

Meeting ID: 294 450 929 (BlueJeans)  ​

Abstract:  ​

Stochastic optimization models are widely used in energy system planning and operation problems. As both planning and operation processes involve different levels of uncertainty, an intelligent way to address the underlying uncertainty is critical. This thesis presents three stochastic optimization models for emerging energy systems. The first two are for planning problems with a specific focus on developing countries. The last one considers an operation problem in a financial option market for prosumers (proactive consumers), who are able to both produce and consume electricity at the same time.  

In Chapter 2, we study the electrification planning problem for Sub-Saharan African countries whose electricity access rate is below 100%. We develop a chance-constrained multi-stage stochastic energy system planning model for determining the optimal energy infrastructure investment and operations under demand uncertainty. We consider both centralized and decentralized (locally installed) generation technology. The model utilizes chance constraints to allow some demand unsatisfaction in certain years for cost savings by waiting for the construction of the inexpensive centralized infrastructure. Chance constraints are linearized by introducing new binary variables and the problem is modeled as a mixed-integer linear program. The model can be large in size; a progressive hedging algorithm is applied to decompose the model by scenarios, greatly reducing the computation times.  

In chapter 3, an adaptive two-stage stochastic optimization model is applied to the energy infrastructure expansion planning (EIEP) problem for developing countries. Traditionally, a two-stage stochastic model provides a static policy which is restrictive and conservative, whereas a multi-stage stochastic model is fully adaptive but requires too many adaptions during the planning horizon. The adaptive two-stage model is a partially adaptive approach that lies between the two-stage and multi-stage stochastic models. In particular, each investment decision in the energy system planning problem has an associated adaption time, a period prior to which the decision is predetermined and after which it is revised to adjust to the uncertainty realized thus far. The adaption time is a decision in the optimization model. The adaptive two-stage stochastic EIEP problem is formulated into a mixed-integer linear program. We use approximation algorithms to solve the proposed models by the bounds between two-stage, multi-stage, and adaptive two-stage models. We further investigate four different approaches in sharing adaption times for related investment decisions. A case study of the country of Rwanda shows the adaptability and effectiveness of the model and illustrates the benefits of implementing the adaptive two-stage EIEP for developing countries.  

Chapter 4 analyzes the optimal operations of distributed energy resource (DER) management problem for prosumers incorporating options contracts. As the DER penetration level grows higher on the consumer side, the load serving entities (LSE), which supply electricity for a collection of end-user consumers, shoulder enormous amount of quantity and price risks. As a result, an LSE has strong incentives in trading load reduction and buying electricity delivery directly from prosumers. We consider an option market in which the prosumer sells options of load reduction or electricity delivery to the LSE for certain designated time periods. During these periods, the LSE has the right, but not the obligation, to request load reduction or electricity delivery from the prosumer. We formulate a Markov decision process (MDP) for the optimal DER management problem for the prosumer equipped with a renewable energy source, an energy storage system, and demand response technology. The model helps prosumers determining the optimal option quantity to procure in the first stage and generating the optimal real-time operational policy involving storage assets and demand response in the second stage.