15-484/15-884, Fall 2013:
Computational Methods for the Smart Grid

Course Info

Number: 15-884 (PhD), 15-484 (Undergrad), also listed as 18-473
Time: TR 10:30-11:50
Classroom: GHC 4101
Instructor: J. Zico Kolter
Units: 12 (884), 9 (484)
Office hours: Mondays, 2-3pm, GHC 7115
Syllabus: Available here

Course Description

Building a sustainable energy ecosystem, and controlling a complex electrical grid to provide this energy, poses one of the largest challenges facing the world. Although energy topics span many disciplines, ranging from power systems to public policy, it is becoming apparent that computational techniques such as simulation, prediction, optimization, and control have the potential to drastically impact virtually all of these areas. This course provides an introduction to recent advances in computational methods applied to sustainable energy and the smart grid; the goal is to provide students with a broad background in state-of-the-art computational methods that repeatedly arise in these domains, such as machine learning, optimization, and control, and to provide hands-on experience applying these methods to real-world domains. In particular, much of the class will use real data from the Pennsylvania electrical grid as a running example, and address issues regarding the prediction, modeling, and control of electricity from existing and renewable energy sources. Although listed in Computer Science, the course is expected to be of interest to students in many departments, including ECE, MechE, CEE, and EPP.

This course is offered in undergraduate and PhD levels, under course numbers 15-484 (crosslisted as 18-473) and 15-884 respectively. Lectures are the same, but the PhD version will have additional problems and will require an original research final project.

Application areas include:
  • electricity demand and renewable resource prediction
  • modeling energy consumption in buildings
  • electrical power systems, power flow, and power markets
  • control of distributed storage
Computational techniques include:
  • regression and classification
  • time series prediction
  • Newton's method for non-linear equations
  • convex optimization
  • model predictive control
  • biweekly problem sets and a written final project
  • some programming experience (assignment use MATLAB)
  • basic background in linear algebra