It’s confusing out there in the oil and gas industry. Analytics has become a loaded and unclear term. Many companies rightly think they do analytics already, in that they are using data to make decisions and deploy resources to save costs and increase production. They wouldn’t be wrong.
Recently, however, analytics has assumed a newer definition in the industry that is called a range of buzzy sounding names: big data analytics oil and gas, advanced analytics oil and gas, predictive analytics, prescriptive analytics, machine learning algorithms, AI, and so on. This is in addition to some of the longer established terms such as SCADA analytics, SCADA big data analytics, and even the more plain SCADA data analytics. The first list is what a number of companies are starting to charge hard towards, thinking there is an opportunity to save significant time and money by doing these things.
Analytics in this context has become a mixed bag of high-tech, computer and data science, and different types of statistical analysis and model building. All of this implies next level capability, of course, which means finding a better way for people to wade through the onslaught and glut of daily data.
It fits in with the trend of oil and gas companies wanting to do more with less by working as smartly as possible. There’s a feeling behind all this that despite having more data than ever before, decision making is not faster and the clarity of what to do that is best for the business is not all it could be.
The question we want to answer is what makes eLynx’s approach to analytics fundamentally different from other approaches and why it solves the biggest challenges of doing it. Our ultimate goal is to help entire companies transform by using real-time data to create insights so they can immediately know and prioritize the actions that they should take to deploy resources to save costs and increase production.
What you will learn is that the hardest part of doing this isn’t necessarily the models or math, but building a production system capable of feeding models with quality and timely data while empowering decision makers with the information to make measurably faster and better decisions throughout an entire company.
The concept car is the flashy one-off that attracts attention at an auto show. The press writes about it and enthusiasts comment on it in forums and social media. And then it doesn’t get built. It’s often completely impractical for a number of reasons including that is not designed for the rigors and realities of daily driving while the technology and economic feasibility is not there to produce one every minute.
The concept car in the oil and gas analytics world are the models built from a large static historical data set. It is developed and used by a small team of data scientists to find some new insights looking across a static quantity of past -- not real-time -- data. Typically, there are many holes in the data in the form of missing values, inaccuracies because of faulty sensor readings, delayed data acquisition problems, and unlabeled data that has no identified data sources associated with it.
The team works to get the data in order for analysis in various ways, and one of the key steps is to clean it. Cleaning data is actually akin to patching it, or finding ways to fill holes or correct errors (at least the ones they can identify) in the data so that statistics can be calculated. Think of a tire that is riddled with patches, it’s not something you could or would want to drive very far on.
With so much cleaning, the number of assumptions and estimations for any given model compounds, making inferences from the data much less reliable. While it feels like higher-level analytics and computation is going on, in just one 24 hour period an oil field is generating thousands more data points. This data is not fit to feed these models, while the models themselves are not on a solid foundation.
The equivalent of the analytics concept car approach isn’t able to process this information because the data inputs are not configured to continuously generate clean data that can be immediately used to spot problems and opportunities. In other words, they are not built to generate real-time, real-world analytic insights that people can do something with at the right moment.
The hardest part of analytics is not the math, it’s creating the entire ecosystem to ingest, analyze, and make instantly useful the huge amount of new data generated in the present moment. In order to solve these kinds of problems, for instance, IT and OT have to get on the same page and figure out how missing and faulty data is occurring in the system and devise a plan to fix it.
The true engine of oil and gas real-time analytics is the ability to have both high quality and frequency of data that can be immediately analyzed and deliver the insights to teams who can quickly understand and prioritize and plan their actions with confidence. This is the ultimate key to what enables people in a company to make faster and better decisions.
The analytics concept car has no real engine under the hood. Most companies can not get out of the starting blocks for this very reason. The concept is only a one-off version that is not built to continuously produce time series data that analytics models can use to make all the daily decisions in the oil field.
The graphic story below summarizes the attempt of a big five oil company to do advanced analytics. Their approach is the concept car version where model building is based on a large, static data set. They realize after some time and investment that the inputs they have are not sufficient to the task.
The cleaning or patching can’t solve the deeper structural issues related to how they are generating, collecting, and preparing the data. And they can not find a way to turn what they were working on into a real-time analysis tool without having to start from the beginning and reconfigure their entire ecosystem for analytics. The roadblocks are daunting.
In order to build the real-time data engine, everything in the field from the sensors to the communications to the data warehousing -- just to name a few of the things -- has to be done with the goal of generating clean time series data with the proper frequency and granularity for analytics. The quality of the data has to be continuously monitored to find any problems and quickly correct them as they occur. These are essential to build models that are accurate and useful.
Because eLynx started as a SCADA company nearly 20 years ago, quality and timely data was always the focus because it is the key to making automations and alarms, for instance, work as they should. No matter if the type of analyses are trend visualizations or setting alarm thresholds, we knew that these will only be as effective as the data used to do these things.
eLynx was the first company to offer the turnkey SCADA solution for companies which means we had to figure out how to work and integrate with all lift and equipment types. We’ve been innovating and perfecting ways to produce clean, usable real-time data since the day we started. What we have today is a true full-stack analytics solution.
This is also why our product is cloud-based so all be integrated into one place and is instantly retrievable by anyone, anywhere. Making the data useful by more people in a company is just as important as getting the data. We have created the software and the user experiences to deliver information and insights in real time to anyone, anywhere on any type of device. All the daily decisions add up to save costs and increase production.
It is only through this kind of rigorous approach to generating real-time analytics quality data that enables data scientists and production engineers to do more than just establish correlations. Right now, most have to essentially guess which of the many correlations they should react to. Now they can discover causes and build predictive models that cut down analysis time and direct their resources to the things that can make the biggest impact.
By creating a reliable library of causes it becomes possible to spot more problems farther in advance, accurately classify them, and, because of this, be able to know exactly what to do and when to fix the problem with greater speed and certainty than ever before.
Doing production analytics requires the foundation of doing good SCADA. A big reason the concept car approach to analytics is used is that a company’s SCADA data is not currently up to snuff. The workaround eventually makes apparent the shortcomings from the beginning to the end of a company’s data ecosystem. The high ambition to do such things as predictive analytics is thwarted, but it also clearly revealed that the analysis done with SCADA tools will not be anywhere near what it could be.
This is good news for companies who want to improve their bottom line. The best way to do this is to simply improve the data quality and timeliness of their SCADA data and make sure the tools to deliver insights are user-friendly, work on any sort of device, load quickly, and work anywhere, including the remotest area of an oil field. What companies can do with good data and the SCADA toolset can be very advanced, particularly with the eLynx product. This is especially the case if people can understand and take action on this information.
This capability is also within close reach of any company with the help of eLynx’s product. If you work with us, you’ll be setting yourself up to add the super advanced analytics capability later with little effort. What a lot of companies are starting to forget in all this data frenzy is the broader definition of analytics: which is the systematic analysis of data to make better decisions. Doing SCADA well enables this to a greater degree and many companies are realizing that they are not happy with their data quality. They are experiencing many false positives and end up making gut decisions anyway.
Production analytics is about being able to do meaningful analysis of any sort in as close to real-time as possible with tremendous confidence. Only a production approach to doing this makes this possible and eLynx has proven it can do this with a wide range of customers.