Farming seems easy from 1,000 miles away. Today, noise regarding artificial intelligence (AI), IoT, and edge devices is growing astronomically. Many startups talk about edge devices and edge computing, but they may be surprised by the challenges.
Having deployed over 300 edge devices at the edge (on land rigs, plunger lifts, and offshore drilling and production platforms) and performing edge computing ourselves in the US, we have been able to learn a few things and gain priceless experience.
During one of our projects, we successfully piloted and scaled our plan, but we soon realized the nightmare of managing tasks with edge computing and the millions of dollars of investment it would require.
On land, we set up wireless networks at the rig location with our edge devices which could consume any format of data, buffer it, run smarts, share on an iPad and send the data via satellites and wifi to our cloud. Everyone wanted to replace NOV/Pason as an EDR provider and take data directly from the Omron System (PLC). The business value on paper suggested not needing and not paying for instrumentation providers such as NOV/Pason.
Offshore, we took data from HAL/SLB (LWD/MWD) to help geologist seamlessly consume different data sets to create real-time pore pressure prediction curves on the rig itself and to auto-sync the data to office applications like Petrel. We also helped send remote controls back from the master production platform to the satellite platform to shut valves to ensure we could continue production without risking human life in the case of an emerging storm.
One may argue, just six years back, the edge was not ready — but what many don’t know is — it’s a decade old concept, and only the name changed. It was called “smart service,” and edge devices were mesh devices often called “motes.” Today’s form factor processing power has increased, but the implementation challenges, especially at scale, have not. Let’s review some of them.
Access To Data: Many top rig providers will not even let operators access their Omron PLC, let alone startup/software vendors, citing proprietary and confidential nature of work. This led to legal discussions which quickly consumed six months with no concrete outcomes.
Data Quality: The Omron System PLC gathers better high-resolution data than the EDR system with a better sampling rate, but when you collect this data every second, every minute, every hour for 365 days a year, you see many issues with the quality. This low quality can throw off even the smartest algorithms like the ones we ran on the rigs. How do you train and maintain those algorithms for 365 days that can be fraught with challenges, especially at scale? I’ll give you an example.
Since these algorithms are hungry for many distinct parameters to work, if one of them lags behind another, the algorithms won’t give expected results. Also, the Omron PLC can be new or old depending
on the rig and have different software versions which result in data coming at different intervals. The upgrades can take months.
Wireless LAN/WWAN Challenges: Creating a wireless network is easy, but cost to install it at a variety of locations is not cheap. Ultimately, even if you can get a good WLAN set up, you need to account for the cost to move data over WAN (satellites) which can cost money and increase TCO dramatically. Don’t forget if anything goes wrong at the edge in WAN/WWAN to the office; operators expect you to resolve the issue.
Supply Chain: Hardware devices, by definition, are a piece of equipment. When they go down during after-hours, it can be a nightmare to manage and replace. You could install a smart mesh network, but the cost to replace failed devices is not cheap. At the edge device itself, you can install a secondary failover, but that means additional cost and additional real estate when there is already so much real estate in the field trailers. Overall the support network required to maintain them operational all year in different basins is a daunting task.
24/7 Support: At scale, cloud software must be brilliant to intelligently flag data quality issues from edge to office. That requires much human talent and smart training. Typically you need on staff drilling, completion, production, and IT staff at the minimum.
In our experience, the road from pilots to scaling and commercial deployments is a huge undertaking which requires hundreds of millions of dollars in investments. We have seen some companies raise close to 100 million dollars and bite the dust. Only time will tell if the IoT bubble is waiting to burst or not.