8.1
Outage Prediction and Response Optimization (OPRO)

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 5 February 2014: 4:00 PM
Room C114 (The Georgia World Congress Center )
Amith Singhee, IBM Thomas J. Watson Research Center, Yorktown Heights, New York; and A. P. Praino, A. Sabharwal, D. Melville, J. P. Cipriani, L. A. Treinish, S. Abbaspour, S. Siegel, Z. Li, G. Labut, H. Storey, J. Esser, P. Whitman, R. Foltman, R. Mueller, and W. Harlow

Severe weather events can have major damaging impacts on electric grid infrastructure and on the utility's operations. Utilities, consumers and the general society suffer huge losses because of such events. The ability to predict specific weather conditions and their impact on the grid with sufficient spatial and temporal precision, and lead time has the potential to enable proactive allocation and deployment of resources to minimize restoration time. We will present our work on rigorous prediction of weather-driven impact on the grid infrastructure and optimal planning of storm response based on such prediction. The goal is to help the utility to improve its readiness for weather-related unplanned outages, and business objectives such as CAIDI. The solution relies on three coupled models:

1. A highly precise and accurate, cloud-scale, physical numerical weather prediction model focused on the utility service territory

2. A statistical learning-based damage model that can predict the spatial and temporal distribution of relevant damage types over a horizon of one to three days

3. A pre-positioner that solves a declarative optimization model to generate optimal resource plans in preparation for the upcoming weather event

We are developing and applying these analytics in the context of a real utility environment at DTE Energy. The three models consume a large variety and volume of data including:

1. Weather observations from public and private sensor networks

2. Coarse-scale weather model data from NOAA

3. Remotely-sensed observations from NASA spacecraft

4. Land use and related surface data from USGS

5. Restoration crew dispatch ticket data from the utility

6. GIS data from the utility

7. Asset data from the utility

8. Resource data (crew, equipment) from the utility

9. Business rules and operating procedures

We will describe our outage prediction and response optimization solution, the underlying analytics and data, and how it integrates into the storm planning workflow at DTE Energy.