The mission of this group is to bring together utility professionals in the power industry who are in the thick of the digital utility transformation. 

27,043 Members

Question

Advanced Pattern Recognition Vendors

As GE reconsiders its position regarding APM, I would like to see a list of vendors offering a platform for advanced pattern recognition for the OT side of the house. Does anyone have anything like that?

Answers

Ben I suggest the DEX  " Demand Exchange " as a new vallue added service. Have a look at the below coped message.

I am open and willing to discuss "how to make it happen" with GE. Hoa about having a conf call. My mobile/WhatstApp is 55.11.99986 5563.

BRIEFING


US power distribution companies collect about USD 100 Billion a year in demand charges. I came up with a new concept. A kW capacity exchange (not kWh energy). It is synergetic with a list of other initiatives in the power arena, such as energy storage, distributed generation, time of use rates, energy efficiency, and demand response among others.

This exchange, to be sponsored by each power distribution company to its industrial, commercial and institutional energy users (who pay demand charges) is a place where they are able to trade their demand differences in a win-win situation for all involved. The power distribution company will remain with the same total contracted demand (and associated billings) BUT with increased load factor as its clients will be able to better adjust their contract demands vis-à-vis their recorded demands (lowering their demand cost by reducing their “idle” demands or excess surcharges). You may watch this 6-min-video that summarizes the most important aspects of this concept, at https://youtu.be/6vz-lSapZyM

BENEFITS (SHOULD BE CALCULATED ACCORDING TO EACH CASE – THE BELOW BULLETS IS JUST FOR ILLUSTRATION OF THE PROPOSED LINE OF THOUGJHT)

  • Just for the sake of initial calculations, let us consider that the power grid investments involve USD 500/installed kW
  • DEX might for example postpone new power grid investments for 3 years considering a load factor improvement of 5%
  • This is a feasible assumption: the current distribution company’s load factor may be for example 75% and it could climb to 80%
  • Load factor = (kWh consumption of demand charged energy users)/(kW contracted demand x Hours per month)
  • Contracted demand per energy user may be a simple fixed value or a ratchet (% on maximum recorded demand in a one year period)
  • Amortization costs of USD 500/installed kW using a 10% per year (cost of capital + country risk) and a 20-year-timeframe is USD 60/installed kW/year
  • Postponing for 3 years results in USD 180/installed kW
  • If the aggregated load of industrial, commercial and institutional energy users (charged by demand) is 1 GW the postponement benefit is therefore USD 180 Million
  • If a medium sized power distribution company does have a 3 GW load of energy users charged by demand it is a USD 540 Million benefit

HOW DID I COME UP WITH THIS CONCEPT?

  • I was consulting for two right next door neighbors at the same time: a 2 500 kW mall and a 3 000 kW industrial plant
  • The mall wanted to increase its contracted demand by 500 kW and coincidently the plant wanted to decrease its contracted demand by 500 kW
  • The power distribution company required the mall to wait a year because the grid was already operating at its nominal capacity. They would have to reinforce it.
  • The plant was told to ask for the reduction and accordingly pay for the 500 kW idle demand for 6 months (regulations)
  • If DEX was there, the mall and the plant could have traded their demand differences right away (no need to wait)
  • The power distribution company would increase its energy sales (for the same total contracted demand) with no extra investments
  • The two clients would be “happy”
  • But in reality: the mall had to rent diesel gen sets and run them for a year to cope with the new load until the power distribution company was able to increase the contracted demand
  • The mall paid a lot higher for the kWh because the variable cost of a gen set is about 3 times the regulated rates
  • The plant paid the “idle” 500 kW demand for 6 months
  • It was the opposite of a win-win situation and most importantly the power distribution company was a looser as well


 

Ben, Unfortunately, I'm not aware of any firms that offer a machine learning, pattern matching approach as part of an asset performance management program. The only item I can think of that uses pattern recognition for asset management is the work underway at Stanford, called Deep Solar

Hope you find this useful.

Gabe Prado's picture
Gabe Prado on July 17, 2019

Hi Richard! One to put on your radar is SparkCognition. Happy to provide a tech briefing and answer any questions. 

|I have seen some startups or growth companies working or focussing in this field.

You can check for instance the Free Electrons Accelerator. If I remember, there were some companies doing or at least talking about it. Let me check whether I find more info. 

Thorsten 

 

 

The key players in the market for pattern recognition are IBM Corporation, Imagga Technologies Ltd., Amazon Web Services, Qualcomm Incorporated, Google LLC, Microsoft Corporation, NEC Corporation, LTU technologies, Catchoom Technologies, Intel Corporation and etc.

S D's picture
S D on July 17, 2019

I agree.  These companies have AI/ML platforms that can be applied to any dataset and tease out trends.  I'd also look to EPRI.

Hi Ben! For advanced analytics for OT I recommend looking into SparkCognition, an AI growth company, based out of Austin, TX. You may have seen us at PGI or the Connected Plant conference but we have an offering of deployed machine learning-based solutions that improve operational reliability through predictive and prescriptive maintenance. A few relevant resources below:

Scalable enterprise AI solutions for predictive and prescriptive maintenance: 

SparkPredict: Predictive alerting to nuanced and previously unknown equipment failure modes based on asset-centric modeling (high-accuracy anomaly detection and clustering, normal behavior modeling, pattern matching). SparkPredict provides additional evidence and diagnostics for each predicted outcome and can be configured to provide insights for various user types including management, engineers, and SMEs. With SparkPredict, customers have been able to take significant steps towards realizing value from their digital transformation efforts in a sustainable way. 

DeepNLP: Natural Language Processing platform that offers prescriptive insights from previously untouched data in maintenance logs, user manuals, notes, emails, etc. DeepNLP brings meaning and context to these types of unstructured data and will make recommendations on the optimal action to take in order to fix an asset problem. This technology is helping to document and make reusable all the tribal knowledge that exists in operations teams and is improving the outcomes of work orders.

Please don't hesitate to reach out directly with additional questions! gprado@sparkcognition.com

 

Tap Into The Experience of the Network

One of the great things about our industry is our willingness to share knowledge and experience.

The Energy Central Q&A platform allows you to easily tap into the experience of thousands of your colleagues in utilities.

When you need advice, have a tough problem or just need other viewpoints, post a question. Your question will go out to our network of industry professionals and experts. If it is sensitive, you can post anonymously.