Mike Horrocks gave a foretaste of OSIsoft’s well integrity management system, a proof of concept that is being developed with help from International Well Integrity Ltd. Current well integrity management systems tend not to run with live operational data and may react to well events when they have already happened. PI System’s exception-based surveillance, Asset Frame analytics and Notifications can easily be used in this context. The new solution will be presented in a webinar, ‘Well Integrity Management through Real-Time Data’ early in 2021.
Luca Cadei presented ENI’s e-DOF, a PI System-based real time digital oilfield and a ‘game changer’ in upstream asset management. eDOF is configured from state‐of‐the art off‐the‐shelf components. Design and assembly are performed by a team of ICT specialists and business users. The system acquires over 350 million values per day. Simulators and predictive tools embed the framework and enable production optimization and better decision making. Site data on OPC servers is captured and forwarded through a PI aggregator, and ‘historicized’ in ENI’s ‘Green Data Center’. The Center houses, inter alia, HPC 5, a 52 petaflop supercomputer that is N° 6 in the TOP 500. Graphical ‘operating windows’ show how equipment performs with respect to ‘correct and acceptable’ KPIs. ENI’s data science teams have complete access to all data including PI AF template attributes, calculations, KPI’s, summary information and event frames. (Only) relevant PI System data is ingested into the big data environment. ‘The is no need to perform utopic complete PI data ingestion into the data lake’. Approved data sets are sourced through PI AF to all AI models and data scientists work in the same environment as used for data discovery. Analytics results are fed back into operations dashboards via PI AF & PI Vision. The PI data infrastructure has enabled successful use of data science findings. Cadei ended with an image of a huge PI Vision display with ‘everything a control room operator needs’. The display, nicknames ‘the Monster’ shows some 230 PI variables on screen. ‘It is now possible to anticipate process upsets, asset integrity issues or deviation from plant optimized parameters’. eDOF is considered to be ‘the main enabler of ENI’s digital transformation’.
Glen Milne spoke of a ‘eureka moment’ when a visit from OSIsoft persuaded Spirit Energy of the potential of PI AF. This led to the creation of an asset dashboard that increased daily usage of the PI System. ‘People started talking about performance and started solving business challenges.’ the dashboard provided a live overview of well stock some of which had not been visible for years. The next step was to address high value, low effort problems. For instance, high temperature trips occurred in the process plant causing significant production loss. Spirit worked with Merkle Dentsu to develop a predictive model that worked with the production loss reporting system to identify equipment ‘bad actors’ and allow operators to take preventive actions. Data from over 400 sensors was ‘feature engineered’ to create a 30 billion points data space. Merkle worked with subject matter experts at Spirit Energy to create a target variable that identified unrecorded trip events. A random forest classification was used to create a model that now provides an indication of a trip event in under two minutes. Spirit and Merkle received the 2020 DataIQ award for most innovative use of AI*.
* DataIQ also awarded the Oil & Gas Technology Center an award for its missed pay project.
Peter Van den Heuvel reported that Shell’s PI system, with some 35,000 users, is the largest PI in the world. In the face of a digital landscape that is growing in complexity, Shell is trying to simplify things and speed connections to PI Server. Currently, there are about twelve million events/sec streaming into PI servers. Data management is growing in importance as Shell leadership now considers data as an asset and PI as one of the main databases. A lot of Shell’s PI data has very poor quality. Business decisions are made quickly. If a meter is flatlining, there is the risk of making a decision based on the wrong data. Shell has a whole team working on PI data quality/data management, an issue that was highlighted in the OSIsoft user conference last year by Pat Kennedy who put PI data quality at the top of his list. With support from OSIsoft Shell has kicked off a Golden PI tags data quality improvement project which will decide which tags are used in decision making.
More from OSIsoft.
© Oil IT Journal - all rights reserved.