GE Power uses data analytics on AWS to help power plant customers save millions of dollars, stream 500,000 data records per second, and scale to support the ingestion of 20 billion sensor-data tags. The company provides utilities and power companies throughout the world with solutions for power generation. GE Power runs its key data-analytics application on AWS, using Amazon Kinesis Data Streams and Amazon Elastic MapReduce.
GE Power equipment generates more than 30 percent of the world’s power. To make sure these plants are performing efficiently, the company relies on Predix, an application that collects and analyzes data from thousands of Internet of Things (IoT) sensors placed on power plant equipment. “We use this data to monitor the overall health of our customers’ equipment, so they can proactively take maintenance actions during scheduled downtime, instead of in response to failures, which is magnitudes more costly,” says Eric Pool, the director of Data and Analytics Infrastructure for GE Power.
For the past several years, GE Power has been moving more applications to the cloud in support of a company-wide digital transformation. To support this initiative, GE Power sought to migrate Predix to the cloud. “We hosted the application in our on-premises data center, which wasn’t scalable enough to meet data growth,” says Pool. “Data coming into the system had grown from 8 gigabytes to 200 terabytes, and that was a scale that was difficult for us to manage. We also needed to lower the latency of the application for monitoring purposes, because our customers expect us to alert them quickly if something is wrong with equipment, so they can take swift action to fix it.”
After considering various cloud solutions, GE Power selected Amazon Web Services (AWS) as its cloud provider. “AWS maturity made it the right choice for our data-analytics solution,” Pool says.
The new cloud version of Predix collects IoT data from equipment-based sensors at 900 GE Power customer sites across the globe. The data is ingested into Amazon Kinesis Data Streams and is then consumed by GE Power analysts or customers accessing Predix at power plants. The solution streams 500,000 data records to Predix each second.
GE Power uses Amazon Elastic MapReduce (Amazon EMR) as its base data-processing infrastructure. Amazon EMR manages the process of ingesting data into the Predix production database. The company also stores archival analytical data in Amazon Simple Storage Service (Amazon S3) buckets. Overall, the Predix application runs more than 1 million data executions every day. In addition, GE Power runs a performance-management application on top of Predix that sends automatic alerts to customers when equipment issues are found.
By running Predix on the AWS Cloud, GE Power has more than enough scalability to meet its fast-growing volume of customer data. “Using Amazon Kinesis Data Streams and Amazon EMR, we can ingest close to 20 billion machine-data records per day from sensors at power plants, a number that continues to grow,” Pool states. “Our existing on-premises infrastructure would have been very difficult and costly to scale to meet that kind of growth.”
GE Power is also helping its customers accelerate problem resolutions by using AWS to host its performance-management application. “As recently as one year ago, it would routinely take us weeks or even months to collect data, run analysis, generate results, and implement a change request for a customer,” says Pool. “Now, as soon as the application notifies us of an equipment problem, we can have the issue fixed within days.”
Using AWS, GE Power customers have better, faster visibility into their power plant operations. “Previously, when we received an alert about a potential equipment problem, we had to call or email a customer,” says Pool. “Now, our customers have full visibility into how their plants are operating because we can give them real-time insights into their equipment through the application. They can make better decisions on how to operate and manage that equipment.”
As an example, GE Power recently used Predix to promptly identify a problem with a customer’s power plant combustion system. Because the application revealed the issue early enough, the customer could schedule the equipment shutdown instead of having to perform a forced outage that would have shut the system down for too long. “The customer minimized downtime and loss of revenue,” says Pool. “Because of the real-time alerting we could provide, they repaired the problem at a fraction of what a catastrophic failure would have cost. They saved millions of dollars as a result.”
GE Power is also helping its customers improve plant operational efficiency. “Because we get deeper and faster insights by running our application on AWS, we can do more,” Pool states. “We can measure how long a given part will last in a plant, for example, because we can get better analysis of the conditions in that plant and the likelihood of impact on the equipment. Overall, this means we can certify that customer parts will run longer, minimizing downtime and additional costs.”
In the future, GE Power hopes to continue driving efficiencies for its customers by taking advantage of emerging technologies on AWS. Pool says, “We are exploring the use of machine learning and artificial intelligence applications, and we plan to use AWS technologies to help us build new solutions for our customers as we move forward.”