Systematic data gathering with statistical symptom recognition allows operators to anticipate and prevent problems before they occur.

Recent advancements in the capture of real-time production data have provided new possibilities for the understanding of production. With these developments, large quantities of real-time data are now starting to hit corporate databases. Transforming these data into practical information is crucial. We need to ensure that the 'right information' is getting to the 'right person' in the 'right timeframe.' Advanced analysis of real-time well data, coupled with predictive analysis and automated alerting, is the key to bringing this dream to fruition.

Syndrome recognition

Advanced production systems are able to identify problems as they occur, primarily through continuous monitoring activities together with the power of advanced analytics. The goal for these systems is to identify potential situations in production before they develop into actual problems. This prognostic capability is what will drive the value of advanced monitoring systems and is the core enabler of production optimization. The impact this will have on producers from both a cost savings and asset utilization standpoint has significant potential.

Syndrome Recognition Application (SRA) from Telegnomic delivers real-time optimization solutions for a wide range of production problems by providing a software environment for better managing emerging production problems - 'syndromes' - such as premature pump failure or gas breakthrough. Using real-time data gathered from a wellsite based intelligent gateway (or RTU), the SRA applies quantitative techniques to identify the statistical signature of these syndromes. As trends in the real time data readings develop in a way consistent with the event signature, warnings of both imminent and long-term problems can be given.

Example: Problems with pumps

Heavy oil production is tough on rotating equipment. The combination of sand, gas, water and oil is a constantly varying abrasive mixture whose consistency can change very rapidly. As a result, wear on moving parts is unpredictable and can increase rapidly and apparently without warning.

The pumps are inevitably remote, and get visited once daily at most. When weather conditions are bad, this interval can increase. It is hardly surprising, therefore, that catastrophic failure of pumps and compressors is a real concern and can cost huge amounts in lost production, repairs and supervision. In heavy oil production, interruptions in flow may have additional deleterious effects, which take many weeks to disappear.

One approach to resolving this dilemma is to provide a mathematical model of instrument failure that is then matched against incoming data. This statistical approach can measure the risk of instrument failure, which, if the model is constructed correctly, quickly climbs towards 100%. This enables better decision making, gives oil companies the opportunity to pool knowledge across the whole oilfield and promote best practice.

The model describes a 'syndrome' and it is incorporated in a 'syndrome detector,' which can create e-mails, faxes or pager messages and send them to the appropriate persons. Telegnomic uses its pattern recognition software to identify 'syndromes' or characteristic changes in the incoming data stream. This reduces the problem to a set of traffic lights, which indicate the risk of a particular event. Graphical analytics allow users around the world to further analyze and discuss these problems and to detect them if they arise elsewhere. Using thin-client technology (giving users access to information through an Internet browser) makes it possible to give a wide variety of people access to the information they need, irrespective of the type of computer and operating system they may have.

Having established that the necessary instruments are working properly, statistical models can then be used to detect changes in pump performance. For example, classic short-term changes in pump performance, such as 'slugging' due to sand or water, can be detected. Research shows that repeated slugging predisposes pump failure. The characteristics of pre-failure can be modeled and, as a result, give many hours of warning before actual failure occurs. Even when telemetry brings this data to a 24 by 7 control room, hard-pressed operators can miss these events unless there is software available to automatically and reliably recognize them and classify them in engineering terms.

Conclusions

In a world in which streaming real-time data threatens to overwhelm us, we need intelligent systems to detect significant patterns in data. The solution needs to be highly scalable in that it can search thousands of data streams from thousands of wells, looking for hundreds of syndromes. Not only does this put a continuous watch on crucial processes and thus reduces downtime through early detection, but the data warehouse can be mined to reveal further useful syndromes and further improvements leading to improved well and field management. To do this effectively requires specialist hardware and software able to manage tens of thousands of transactions per second, while searching for multiple syndromes, within terabytes of time-series data at the same time. The results of such advanced systems will enable producers to better manage and truly optimize the productivity and performance of their oil assets.

Comments