Interested? Do you feel you will benefit?

All Rights Reserved. marcus evans ® 2019

Outage Prediction and Modeling

Harnessing and Harmonizing Methods to Forecast Outages, Prevent Damage and Expedite Restoration Decisions

19th-20th June 2019
Boston, MA, USA

marcus evans Noth America

What our delegates think of us:

An interview with Edward Kieser, Senior Meteorologist at American Electric Power

Ahead of the Outage Prediction and Modeling conference, we spoke with Edward Kieser, Senior Meteorologist at American Electric Power about monitoring weather aspects to proactively address pertinent information and remain updated.

One of the sessions you are speaking on is a joint discussion regarding weather aspects involved in the outage prediction model, from a meteorological perspective, what are the most important aspects to consider? What are the most unpredictable?
 
The most important aspects to consider from a meteorological perspective are the types of weather that are going to cause a large number of outage cases. Any type of storm can cause some outages, but we’re really trying to identify the ones that are going to require extra help, so it’s important that we find a better way to do that. We’d like an outage prediction model that can identify the types of weather that are going to cause widespread power outages. In terms of the most unpredictable, I’d say the risk with any outage prediction model is that it’s relying on a highly detailed accurate weather forecast. When I say highly detailed, getting down to small grid sizes for where winds will occur and now, some of that isn’t well done. There are different types of weather that are easier to forecast than others, so thunderstorms are probably the most difficult because they tend to pop up. We have less lead time on thunderstorms than some other types of weather because sometimes they develop under similar atmospheric condition and sometimes they don’t.

How do historical weather patterns contribute to forecast process? 

It’s very important. In the non-outage prediction model realm, we rely heavily on past, similar cases of weather to try to identify the potential impact an incoming storm will have on our assets. Every weather system is different, but it comes down to experience and pattern recognition. In terms of models, the model that is being developed for us is a learning model. As we acquire new data from storm occurrences, that data will go into the model and help train the model to improve down the road. The more data we have, the better the model will become. For some of the types of weather that we’re trying to forecast, there are not very many cases in the past to go on, so the sample size is rather small on some impactful events. 

You mentioned AEP took a hybrid approach on deciding what path to take in terms of creating an outage prediction model, how did AEP come to conclusion on that decision? What are the pros and cons that swayed the decision in that direction? 

This goes more into my second session. We have taken a hybrid approach between hiring a vendor versus building a model in-house.  We have hired a group from nearby The Ohio State University and the University of Michigan to build our model. Technically, this is a vendor, but not in the traditional sense. They are  working very closely with us. Thus, there is probably more in-house input than compared to an off-the-shelf model that the vendor will tweak for your company. The model is being built with our company’s needs specifically included, so we blended those two approaches. They also have more of a scientific component. They are learning and part of this is their scientific research and applying what they’ve already learned. They are also learning from what they get out of the models they’re developing. It’s a good approach because it’s more cost effective. We’re also getting more input into the model since it is easier to interact with the team. We have monthly, larger group meetings to go over the progress.  There is not a one-size fits all, a large vendor might work well for some and others may find building one on their own may work.

Have you come across any unexpected challenges? 

Not yet, but it is not operational yet, as it is still in development. We will show a wind storm case (at the conference) that they ran the model on, which performed fairly well. We can’t say one thing is doing great or not because we have not used it in an operational setting.

Tell us an interesting fact about yourself unrelated to work. 

I have been to 49 states, and want to get to Hawaii sometime to make it 50. I’ve enjoyed visiting all the states, but Alaska was awesome. While in Alaska, I submerged myself in the Arctic Ocean, so I am a member of the Arctic Ocean Polar Bear Club.

Edward Kieser is a facilitator and a panelist in interactive discussions during the conference

Why you should attend this marcus evans conference?

Practical insights from active practitioners in your sector

• Alex Rojas
  Director, Distributed Technologies
  Ameren
  
  

• Derek Roles
  Director, Emergency Preparedness & Restoration
  Hydro One Networks Inc.
  
  

• Michael Berlinger
  Project Specialist (Meteorologist) Emergency Preparedness
  Con Edison
  
  

• Joseph Muglia, P. Eng. 
  Director, Distribution Operations 
  Hydro Ottawa

• Emmanouil Anagnostou, Ph.D
  Eversource Energy Endowed Chair
  in Environmental Engineering
  Director, Eversource Energy Center
  University of Connecticut

Interactive round-table discussion: 
Monitoring Weather Aspects to Proactively Address Pertinent Information and Remain Updated (19th of June at 9:15 am)

• Reviewing historical weather patterns to provide an anticipated forecast
• Maximizing meteorological data inputs by understanding the expected level of ice accumulation, wind speed and other specific variables
• Acknowledging the weather is extrapolated for areas lacking recorded weather to consider alternate strategies

For registration pricing and multiple attendee discounts, please contact:

Melini Hadjitheori 
melinih@marcusevanscy.com