Predictive Analytics for Enhancing Customer Service In Electric Utility

Open Access
Goyal, Raghav
Graduate Program:
Industrial Engineering
Master of Science
Document Type:
Master Thesis
Date of Defense:
Committee Members:
  • Vittaldas V Prabhu, Thesis Advisor
  • Power outage
  • Electric utility
  • customer service
  • prediction model
  • logistic regression
Electric power outages in USA alone causes loss of billions of dollars per year and affects nearly all kind of industries. Even a short duration of outage can potentially cause huge loss depending on the area affected. While many factors could be correlated with this problem, one of the major problems identified is weather condition. Extreme weather conditions have been known to cause power outages for a long time on regular basis. Since there are many areas within US, which regularly see harsh weather conditions such as tornadoes, lightning, hurricanes, ice storms, etc; it is imperative that weather forecast be used to better prepare utility companies to handle outages. This thesis is aimed at using historical data for outages and their corresponding weather conditions to come up with a prediction model that can be used by utility companies to predict whether there will be a power outage or minutes interrupted (MI). The model used has a accuracy of ~63% while predicting hourly occurrence of power outages. Moreover the model used to predict occurrence of just weather related outages has been found to show around ~70% accuracy. Further this thesis explores the most common service codes, equipment codes and problem codes identified with large customer minutes interrupted for power outages. This information could be used by utility companies to improve their manpower planning.