Quantifying User Engagement in Residential Demand Response Programs
Maximilian Balandat

Citation
Maximilian Balandat. "Quantifying User Engagement in Residential Demand Response Programs". Talk or presentation, 4, November, 2015.

Abstract
We study user engagement in Residential Demand Response programs. Lacking data from a randomized experiment, we estimate a user's counterfactual consumption during Demand Response (DR) events using tools from Machine Learning. Our models incorporate past consumption measurements and other fixed effects (such as weather or time of day / day of week). We employ nonparametric statistical techniques to compare the forecast errors during non-DR times and DR times, and use the results to define a metric for the level of a user's participation. Our study is based on a dataset obtained from 500 customers of the company ohmconnect, the only third-party residential DR provider in California. Our broader goal is to combine modern techniques from Machine Learning and Econometrics in order to estimate causal treatment effects based on non-experimental high-frequency data.

Electronic downloads

Internal. This publication has been marked by the author for FORCES-only distribution, so electronic downloads are not available without logging in.

Citation formats

• HTML

  Maximilian Balandat. <a
  href="http://www.cps-forces.org/pubs/97.html"
  ><i>Quantifying User Engagement in Residential
  Demand Response Programs</i></a>, Talk or
  presentation,  4, November, 2015.

• Plain text

  Maximilian Balandat. "Quantifying User Engagement in
  Residential Demand Response Programs". Talk or
  presentation,  4, November, 2015.

• BibTeX

  @presentation{Balandat15_QuantifyingUserEngagementInResidentialDemandResponse,
      author = {Maximilian Balandat},
      title = {Quantifying User Engagement in Residential Demand
                Response Programs},
      day = {4},
      month = {November},
      year = {2015},
      abstract = {We study user engagement in Residential Demand
                Response programs. Lacking data from a randomized
                experiment, we estimate a user's counterfactual
                consumption during Demand Response (DR) events
                using tools from Machine Learning. Our models
                incorporate past consumption measurements and
                other fixed effects (such as weather or time of
                day / day of week). We employ nonparametric
                statistical techniques to compare the forecast
                errors during non-DR times and DR times, and use
                the results to define a metric for the level of a
                user's participation. Our study is based on a
                dataset obtained from 500 customers of the company
                ohmconnect, the only third-party residential DR
                provider in California. Our broader goal is to
                combine modern techniques from Machine Learning
                and Econometrics in order to estimate causal
                treatment effects based on non-experimental
                high-frequency data.},
      URL = {http://cps-forces.org/pubs/97.html}
  }
  

Posted by Carolyn Winter on 4 Nov 2015.
Groups: forces
For additional information, see the Publications FAQ or contact webmaster at cps-forces org.