Introduction

The Imperial-Tsinghua Electrical Engineering Research Roulette is a series of webinars jointly held by PhD students from Tsinghua University and Imperial College London. In each webinar, Four PhD students or postdoctoral researchers, two from each university, are invited to give online presentations on a specific topic. The material used in the presentation is preferably published or preprinted.

Topics

The related topics include but are not limited to:

  • Power and Energy Systems
  • Power Electronics
  • Smart Grids
  • Control Theory and Applications

Upcoming Event

Applications of data-driven approaches in power systems

Wednesday, 21/04/2021, 9:25-11:00 (London time), 16:25-18:00 (Beijing time), held on MS Teams (click here or contact Kaiwen Chen kaiwen.chen16@imperial.ac.uk or Olayinka Ayo o.ayo17@imperial.ac.uk for access)

Presentation information

Speaker 1: Firdous Ul Nazir, Imperial College London

Title: Approximate load models for conic OPF solvers

Abstract: The global optimum of the optimal power flow problem can be sought in various practical settings by adopting the conic relaxations, such as the second order cone programs and semidefinite programs. However, the ZIP (constant impedance, constant current, constant power) and exponential load models are not directly amenable with these conic solvers. Thus, these are mostly treated as constant power loads. In this presentation I will talk about two simple methods to approximate these static loads with good accuracy. To resonate with the theme of the webinar, I will focus more on one of the proposed methods which uses regression techniques, while briefly touching upon the other method. The proposed methods perform much better than the traditional constant power approximation.

Biography: Dr Firdous Ul Nazir received the B. Tech. degree in electrical engineering from the National Institute of Technology Srinagar, India, in 2012, and the M. Tech. degree in electrical power systems from Indian Institute of Technology Roorkee, India, in 2015. He received the Ph.D. degree in 2020 from Imperial College London, where he is currently working as a research associate under the Joint UK-India Clean Energy Centre (JUICE) project. His current research interests include distribution system modelling, operation and control, and optimization theory.

 

Speaker 2: Guangchun Ruan, Tsinghua University

Title: Learning-assisted optimization for demand response

Abstract: With dramatic breakthroughs in recent years, machine learning is showing great potential to upgrade the toolbox for power system optimization. This presentation will first summary the latest progress of learning-assisted optimization, and then introduce an application for distributed demand response. With a tailored Lagrange multiplier selection model, the convergence iterations could be prominently reduced by 45-74%, while no oscillation is found during the entire procedure. Several key challenges and opportunities are discussed as well. Deep integration between machine learning approaches and optimization models is expected to become a promising technical trend in the near future.

Biography: Guangchun Ruan receives the B.S. degree (with honors) from Tsinghua University, China, in 2016, where he is currently pursuing the Ph.D. degree. He is a visiting student with University of Washington, Seattle, WA, USA, in 2019, and with Texas A&M University, College Station, TX, USA, in 2020. His research interests include electricity market analysis, demand response, and data science.

 

Speaker 3: Dawei Qiu, Imperial College London

Title: Reinforcement learning in deregulated electricity markets

Abstract: Bi-level optimization constitutes the most common mathematical methodology for modelling the decision-making process of strategic players in deregulated electricity markets. However, previous models neglect the physical non-convex operating characteristics of generation and demand sides, due to their inherent inability to capture binary/discrete decision variables in their representation of the market clearing process or demand response problems, rendering them problematic in the context of markets with unit commitment clearing mechanisms and specific flexible demand technologies. Aiming at addressing this fundamental limitation, this presentation delves into deep reinforcement learning methods that enable explicit incorporation of these non-convexities into the bi-level optimization model.

Biography: Dr. Dawei Qiu received the B.Eng. degree (Hons.) in electrical and electronic engineering from Northumbria University, Newcastle upon Tyne, U.K., in 2014, and the M.Sc. degree in power system engineering from University College London, London, U.K., in 2015. He received the Ph.D. degree in electrical engineering research in 2020 from Imperial College London, London, U.K., where he is currently employed as a research associate since 2020. His current research interests include game-theoretic and agent-based modelling in wholesale as well as retail electricity markets, and local energy market design.

 

Speaker 4: Chenyu Liu, Tsinghua University

Title: Multiple features and data processing in wind power forecasting

Abstract: Aiming at a low-carbon power system, the growing penetration of wind energy has generated a large demand on accurate wind forecasting, especially in areas like uncertainty dispatch and spot trade. This presentation will firstly introduce the typical features involved in the wind power forecasting and their characteristics. Two crucial concepts of data processing techniques and their reflections in the speaker’s work will be briefly presented, the superior effectiveness of which will be evaluated in a real-world wind farm case. Several key method groups and combined challenges in this area will also be reviewed in details.

Biography: Chenyu Liu received the B.S. degree in electrical engineering from Tsinghua University, China, in 2018, where he is also pursuing the Ph.D. degree since then. His research interests include forecasting and uncertainty modeling of renewable energy, and big data analysis.