Machine learning: Why energy data analytics is getting smarter

Published: Fri 11 Oct 2013
A blog entry by Smart Grid Watch

Contributed by:

Smart Grid Watch
Smart Grid Blog

Smart Grid Watch's Blog

The value of energy data goes far beyond simple meter-to-cash billing improvements. New machine learning technology can help utilities spot patterns in energy data that can enhance operations, identify problems, aid long-term planning for assets and networks, and much more.

Machine learning algorithms can help utilities address with a broad range of practical and strategic questions, such as:

  • Which customers are more likely to switch providers?
  • Which meters are more likely to have problems?
  • How to adapt grid operations to accommodate fluctuating levels of renewable resources?

This month, at European Utility Week, Krishan Gupta (Director of Product Management for eMeter, a Siemens business) will give a presentation exploring practical applications of recent advances in AMI analytics technology.

Gupta explained how the advanced algorithms of machine learning work in this context: "Essentially you feed into your system some known patterns, and ask it to find similar instances. For instance, if you've identified some instances of energy theft, then you can feed into your analytics system data from those cases and have it look for similar patterns in current customer data. The system will return possible hits and offer a confidence rating for each. That confidence rating can help you decide whether you need to roll a truck to check out a certain instance of possible theft. Then, when you investigate the situation and feed the results back into your system, you fine-tune the algorithm."

That's just the first step. Machine learning algorithms can also reveal new useful patterns in AMI data. That is, energy data can start to speak for itself, in ways that help utilities plan better.

For instance, previously unnoticed patterns of momentary outages or other grid issues might help a utility better predict maintenance needs for transmission and distribution assets. "If you want to figure out which transformers will fail, you can feed in data about which ones have failed already, and the parts of the network they function within -- and then let the system reveal correlations," said Gupta.

Utilities can start to capitalize on machine learning-enhanced analytics even if they haven't yet deployed smart meters or meter data management technology. MDM solution providers, including eMeter, are introducing cloud-based analytics services that can help you better understand, and use, your existing energy data to help prevent problems and spot opportunities.

Details: Krishan Gupta's presentation, Utility analytics: advanced algorithms for practical gains, will be held at Thursday Oct. 17, 10:30 am. See the European Utility Week program

Read more at Siemens Smart Grid Watch!