Historically this has proven difficult to do, as there has either not been enough data to source when things are going wrong - and thus relying on customers to inform of problems - or the data available was so siloed it was hard to get a clear picture of what was happening and where.
I got the chance to sit down with Andres Gomez, IT manager for Smart Grid Solutions at Florida Power & Light, at MongoDB’s annual user conference in New York this week, where he explained that his team of developers are working harder to making the company’s network smarter through a better use of data. He said:
Our smart grid is basically leveraging everything that we get from the smart meters, with an emphasis on reliability, which at the end of the day is keeping the lights on. And then when we do have power outages, minimising those power outages and the duration of those power outages so that our customers are not affected by it. And if they are, keeping it to a minimum.
We collect that data, it’s basically thousands of data points on a daily basis. With that meter infrastructure we have been able to develop applications to monitor the health of the network. The reality is that we are bound to have power outages, but with the data we are likely to know when the outages occur, the specific location and we are able to deploy our resources in a more effective way.
Florida Power started a proof of concept last year with MongoDB, where it has been able to migrate its siloed time series data into a single source. A unified platform that enables the company to make sense of the data in an efficient way - which had traditionally been very time consuming an expensive with its relational systems. Gomez said:
The big idea that we are trying to solve is moving into a predictive algorithm perspective, in which we are able to have enough information in place to know if a condition (a problem) is occurring in the field and will merit somebody, a specialist, to take a look at those conditions before an actual power interruption occurs.
Identifying problems early on
Florida Power & Light has been gathering data from all of the substations it has, and is collecting data from the meters, the outage system and data on the lines. It is now putting that all into one MongoDB platform, one application, which allows it to make sense of the problems on the network without a highly skilled engineer having to crunch data from Excel and data warehouses. He said:
It eliminates the need to go to different applications. One of the things we are able to address with Mongo is that on the traditional set-up you would have one database for real-time data, another for historical analysis. We are combining the two. With Mongo we are eliminating the need to go to different applications and the time associated to that.
The company is now getting email alerts every 30 minutes, which essentially flag if there is a problem somewhere on the network. And if there is a problem occurring, that problem is ranked in terms of low, medium or high criticality - giving the company an idea of how soon an engineer needs to get there to fix it. Gomez said:
The accuracy we have on that is 83%. Just to give you an example, let’s pretend that the main line that serves this building and this section of New York will be identified as one of those alerts. Instead of having to patrol that two or three miles that that line is made up of, it will give us a specific section that we need to go to from a graphical user interface.
The benefitsObviously a shift to predictive maintenance will hold a number of benefits for Florida Power & Light, both internally and externally. Gomez said that the system will “absolutely” save money for the company, but it is also likely to benefit the customer. He said:
For every trip that we make with a specialist, there is time and cost associated with that persons. What we are going to be doing with this is not only reducing those trips, that could be unnecessary or unwanted, we are going to be accurately reflecting the location of where they need to go. And also the end goal of the big idea is being able to predict when power interruptions will happen. So if we are able to predict that, we can plan the service on a particular location, give notice to the customer, increase customer value and customer satisfaction
Florida Power & Light has just signed an enterprise wide user agreement with MongoDB and has been working with its consulting arm on the implementation. However, given Mongo’s announcement of its database-as-a-service offering Atlas, I was keen to find out whether or not Gomez sees a future in the cloud for his network’s management. But it doesn’t seem likely in the short to medium term. He said:
It’s interesting. I think that there is a lot value in Atlas. However, for our industry, we are very prone to cyber attacks. We are also a highly regulated industry. We have our own data centres and we have a dedicated group in our company that is just monitoring cyber attacks.
I see a value in Atlas and the ability to have hardware up and running in five to eight minutes is amazing, but similar to the finance industry, we will probably have our own data centres for a long time because they don’t want data being exposed to an attack.