Virtual Seminar Series | ISE Graduate Student Colloquiums

All dates for this event occur in the past.

Yuzhou Jiang

Advisor: Dr. Ramteen Sioshansi

Committee: Dr. Chen Chen, Dr. Huanxing Yang

Summary: Electricity market design is an important and long-lasting topic because a slight improvement can create a substantial amount of welfare to the society. The integration of energy storage introduces more complexity to it. This research focuses on three aspects related to electricity market design, from which we hope to aid market design that enhances market efficiency and avoid distortion. The first part of the research refutes an unfounded claim that giving the market operator the authority to make energy-storage operational decision undermines its independence. The second part compares two market participation models for energy storage on economic efficiency. The third part explores the effect of strategic bidding on market outcome under different market equilibriums.  

 

Mehdi Mashayekhi

Title: Efficient & Convergent Methods for Hyperparameter Design of Deep Learning Models Applied to Pipeline Inspections

Advisor: Dr. Theodore Allen

Committee: Dr. Guzin Bayraksan, Dr. Cathy Xia

Summary: This research is sponsored by the Rosen Group, which is the leading pipeline inspection company in the world. In the U.S. alone, there are 2.5 million miles of oil and gas transmission pipelines and there have been more than 1,400 pipeline incidents since 1986, including 364 incidents in 2020. The pipeline operators in the U.S. spend around $1B/year for cleanups, but the total societal costs are likely much higher. Globally, there have been over 1,000 oil spills per year Since 1986. To reduce leaks, the Rosen Group employs hundreds of inspectors to study pipelines using a largely manual approach.

In our research, we seek to use deep learning artificial neural net image classification methods to efficiently and accurately warn pipeline operators that maintenance is needed to avoid possible leaks. We also seek to develop methods that advance the accuracy of many machine learning modeling methods.