Independent Study Course Goals

I am interested in an independent study within the ENGR department under the guidance of Dr. Peter Alstone. The course content would revolve around a demand response study involving end-use load profiles, grid load time series data, and renewable curtailments time series data.

The primary research goal of this study would be to understand and predict curtailments in the California grid.

The primary learning goals of this study are as follows:

• To learn statistical methods required for describing and predicting large-scale timeseries data
• To understand the mechanistic or otherwise correlated dimensions that result in renewables curtailment

Secondary goals include:

• Developing a competence in physical data modeling of grid physics
• Implementing a full research stack through modern analytics tools
• Working with other researchers to further understanding of broader research questions (e.g. how does I something smaller that I do plug into a bigger picture?)
• Work toward producing or assisting the development of a white paper about the study's primary research goals

Course Design and Schedule

Learning goals and cadence are outlined below. Course schedules were roughly based on the complementary Introduction to Statistical Learning (IST) MOOC from Trevor Hastie and Rob Tibshirani. The course design was paced to roughly maintain 2-3 levels of cognitive load across different levels of Bloom's taxonomy.

Figure 1: Preliminary Course Schedule

• IST as a foundation to develop competence in advanced statistical methods
• A central research problem (predicting curtailments) as an application of statistical methods
• 5 major assessment components and deliverables:
  • IST MOOC Assessments
  • Descriptive report of data collection and exploration
  • Methods report on logistic regressions
  • Methods report on SVMs
  • Final report detailing and comparing results from both approaches (what can we say about desired bias and variance tradeoff? How well does our model perform on new data? Next steps, shortfalls)

Methods Briefs

Methods briefs should represent short, 2-3 pagers that concisely describe model results and approaches for predicting curtailments.

Final Report

The final report should survey different approaches to predicting curtailments. Structurally it should describe:

• A research question presented alongside appropriate context, background and motivation
• Data sets used in the analysis, and a general description of how the data were retrieved and transformed
• An overview of different methods used to approach the research question
• Results and discussion of each approach about potential applications, and drawbaks
• A conclusion for additional research areas, other promising methods and approaches and direct applications above

Alongside the report, all code and data should be published in a format that enables reproducible results.