What makes smart grid so smart? Can you ever hope to master it?
By Alyssa Farrell, SAS
As the Internet of Things (IoT) brings more connected devices onto the electric utility grid with all their associated data-generating capabilities, the answer will increasingly include machine learning. The human brain simply is not capable of managing and interpreting such data volume on its own.
As discussed in part one of this series, “Six Steps to Becoming an IoT Ninja,” which was published in POWERGRID International last month, the IoT brings connected, intelligent devices that contribute timely, granular data to grid operations. When used properly, the data can increase visibility into asset performance and decrease the time between event detection and correction to increase grid reliability.
The Path to IoT Enlightenment is Littered With Devices
As IoT devices proliferate, utilities will find it more difficult to keep up with the torrent of data. Gartner forecasts that 6.4 billion connected devices will be in use worldwide this year-up 30 percent from 2015. That figure is expected to reach 20.8 billion by 2020. In 2016 alone, 5.5 million new devices will be connected daily. The data volume means utility companies will likely adopt machine learning strategies to interpret what the data means and how they can benefit from it.
For those uninitiated in the terminology, computer networks equipped to use machine learning techniques excel at strategically analyzing, adapting and learning from data coming from complex, fast-changing data sources like those created by IoT sensors. Because machine learning analyzes an entire data set, including historical data, rather than drawing conclusions from a sample, those networks can make precise predictions and reveal insights derived from hidden patterns. As more data becomes available, the system learns, applies the results and adjusts insights over time.
Take machine learning one level higher with cognitive computing, and you add the ability to interact with humans more naturally because of technologies like natural language processing and facial recognition. Using cognitive computing techniques allows a computer system to understand complicated questions asked by a human voice and respond with an equally complex answer. For example, a homeowner might talk to her smartphone to adjust the thermostat without getting off the sofa. Within a few days, the thermostat recognizes a pattern of preferences and prompts the homeowner to confirm the change so the temperature adjustment happens automatically instead of waiting for a prompt.
Although utilities currently are further along in implementing IoT, most plan to use both IoT, machine learning and, eventually, cognitive computing as partner technologies within the next few years.
When machine learning is applied to IoT data, it allows utility companies to realize the next generation of the grid: a distributed system with power flows among millions of distributed energy resources (DERs), microgrids and in-home devices like smart thermostats and home energy portals that help utilities deliver more reliable energy and customer choice.
Real World Applications
Many utility companies are beginning to dip their toes into projects that rely on machine learning to achieve business objectives. Examples include:
“- Demand response events. One utility uses machine learning to better fit the included participants and forecasted results for demand response events. While the company previously compared the baseline plan to actual performance, machine learning helps them come up with a better baseline-achieving objectives on a more consistent basis.
“- Outage management. One utility customized its analytics software to help deterine the outage type, where it is, the required response and the time necessary to repair it. In the future, it plans to expand its use of machine learning to manage resource use more effectively, identify outages faster and even predict outages before they happen.
“- Cybersecurity. In the SAS 2016 survey, the top benefit that utilities cited for application of machine learning is increased cybersecurity. The volume of network data exceeds that which can be analyzed by the human eye, so self-learning algorithms will compliment business rules to strengthen critical infrastructure protection.
The Horizon for Utilities and IoT is Hazy
Utilities are grappling with threats to the accepted business-as-usual model. In the future, the bulk electric grid might be the provider of last resort once consumers have exhausted renewable resources and Tesla wall batteries. The problem is the cost of the grid is based on everyone using it, so prices rise as people defect.
So how do techniques like IoT and machine learning reduce grid operational costs while allowing for a better integration of renewables? How do those utilities prepare for this new future? IT architects and data scientists play a key role as utilities shift toward more IT to complement OT (operational technology). Utilities must work on hiring and retaining this talent, which is in demand in many industries.
Realistically, a completely autonomous grid is still beyond our grasp. But with the increasing complexity of the grid and the number of real-time decisions to be made about it, technologies such as IoT, machine learning and cognitive computing will continue to decrease the distance between today’s reality and the ideal of the future.
Editor’s note: This article in Part 2 of a two-part series on IoT. If you missed Part 1, you can find it in POWERGRID International’s August issue. For more information about machine learning in the utilities industry, download the whitepaper, “The Autonomous Grid: Machine and IoT for Utilities,” at sas.com/whitepapers.
Alyssa Farrell is the Global Energy-Utilities Industry marketing manager at SAS. Follow her on LinkedIn and @alyssa_farrell on Twitter.