The advent of the smart meter has dramatically changed the landscape of the energy industry. Instead of utility companies relying on one readout per month for billing purposes, they are now provided with several readings per day on multiple measurements.
Smart meters represent a significant opportunity for utility companies, allowing them to predict and respond to energy spikes and equipment failure.
However, with that great opportunity comes great challenge, especially from the data perspective. SunGard held a webinar with several of their experts and CommodityPoint’s Patrick Reames to address these challenges through identification, problem-solving, and cost-cutting strategies, many of which point to Hadoop from a data analytics standpoint.
According to SunGard’s Melista Anderson, there will be a 500% growth in smart meter deployment from 2010 to 2015. This growth would represent a huge spike in the amount of data that utilities are collecting.
“For a mid-sized utility of a half-million customers,” Anderson said “the deployment of smart meters will result in an explosive growth of customer data, increasing the number of reads from six million per year to almost 18 billion per year. That’s a 3000-fold increase in the amount of data that then must be captured, ordered, stored and analyzed in near real-time.”
Real-time analysis is critical in the energy industry and it’s not difficult to see why. Utilities need to respond as quickly as possible to energy spikes to avoid costly shortages and even blackouts. But, of course,there exist other less dramatic examples, such as correlating weather changes to energy usages in certain areas.
One challenge not to be underestimates is the increased federal regulation that arose as a result of increased data access. According to SunGard’s Stephen Nimmo, utilities are almost required to process and analyze their data to show their intent to regulators while also keeping their raw data intact. “The solution to this problem, Nimmo remarked, “is to stop throwing data away.”
While conceptually simple, that response requires a significant data investment on the part of utilities. That’s before the utilities even start using the data to do work and deliver any sort of usable insight.
According to Reames, the smart meters are a big reason as to why traditional relational databases are being overrun. Instead, it is time for utilities to turn to Hadoop-type frameworks for storing and running their data.
“Big data sets go up to multiple petabytes in size and this requires new tech and approaches such as Hadoop,” said Reames, “a framework for processing these massive data sets.”
There are several reasons that Reames and SunGard’s Neil Palmer point to for why a Hadoop-based system could prove advantageous to the energy industry, including the ability to handle both structured and unstructured data, high levels of availability, andits open-source nature which, according to Reames, allows for “near-constant improvement.”
Most important, however, is its scalability, answering the fundamental question of utilities wondering how to store and process the petabytes of data they’re receiving from their smart meters. “To some extent,” Palmer said “Hadoop is replacing the initial pit stop for all data coming into an organization.” Palmer cautions that this may not always be wise and that existing BI infrastructure should not be discarded out of hand, but that Hadoop does offer the scalability necessary in the energy industry.
In this context, Hadoop has its limitations. For example, while physically accessible to those who want to invest in it, it is relatively inaccessiblefrom a functional standpoint in that few know how to work confidently in it. “Adoption by the business users is critical,” said Palmer. “Disruption to their day-to-day lives is not acceptable and they need new ways to be able to understand and mine the data that’s out there.” However, Reames feels that as the Hadoop market advances and matures, so too will the ability of vendors to cater to the business users.
For Palmer, Hadoop belongs to one of four big data analysis subsets which will eventually be crucial to the energy industry. Other platforms, such as graph databases like Neo4J, could help utilities eventually model their entire grid in what would be a huge step for predicting shortages. Per Palmer, “Graph databases such as Neo4J are excellent for mining relationships. If you want to understand the connections between e-traders and counterparties and accounts, a graph database is an excellent place to start.”
As meters get smarter, so too must utilities. Those hanging around Hadoop World in New York City right now may be able to help them with that.