Refinery engineers are already dealing with a lot of data.
There are measurements, control system values, energy records, equipment information, lab results, emissions measurements, alarms, maintenance, and historical operating data.
The real challenge isn't necessarily getting another measurement.
The challenge is ensuring the right measurements can be connected to answer an engineering question.
For energy-efficiency and emissions projects, a simple data pipeline can make that process easier:
Measure → Collect → Validate → Connect → Analyze → Improve
Start With the Measurements
A refinery may monitor a variety of parameters depending on its process units and environmental requirements.
Examples include:
NOx
SO₂
CO
O₂
Particulate levels
Stack flow
Stack temperature
Fuel consumption
Steam use
Process temperatures and pressures
Each measurement tells you something different.
Gas concentration can provide information about the exhaust composition, flow can provide information about the movement of that exhaust gas, and temperature can add another layer of operating context that helps connect information from both measurements.
Looking at these values together can be more useful than looking at one of them individually.
Why Context Matters
Imagine a refinery makes a combustion condition change and subsequently sees a CO concentration change.
The value in isolation doesn't explain what happened.
Did they change the production rate?
Did the fuel conditions change?
Did the exhaust flow change?
Did the stack temperature change?
Was the reading part of a temp startup condition?
Without that surrounding information, investigation can become guesswork.
A more useful data structure would look like:
Gas concentration + Flow + Temperature + Time + Operating condition
This doesn't provide an automatic causation, but gives engineers more information to investigate.
Build a Reliable Collection Layer
Once measurements are available, the next challenge is getting them into a consistent data structure.
A simplified architecture might look like the following:
Sensors & Analyzers
↓
Data Acquisition
↓
Industrial Network
↓
Central Data Store
↓
Visualization & Analytics
↓
Engineering Decisions
The data-acquisition layer will get you some information from multiple instruments.
Connectivity takes that and makes it usable.
So that data is connected, but it should have a purpose.
You're not trying to connect every possible device as an end goal, but instead to get relevant information out when it is needed.
Data Quality Comes Before Analytics
A dashboard can look pretty good, but may just be based off of bad data.
Industrial monitoring systems should consider things like:
Missing measurements
Sensor drift
Calibration periods
Maintenance activity
Communication issues
Wrong timestamp
Duplicate records
Out-of-bound values
If those conditions aren't identified, an analytics system might interpret a data-quality issue as a process issue.
Which is why validation is in the pipeline:
Measure → Collect → Validate → Store
Only after that point should people be relying on trend and/or automated analysis.
Historical Data Adds Another Dimension
A real-time data stream answers the question "What's happening now?"
Historical data can be used for answers like "When did it happen? Has it happened before? What else changed?"
That's important, particularly when you're evaluating possible energy-efficiency projects.
A refinery that makes modifications to their steam system or has a heat integration project will want to ensure that they can compare operating conditions before/after the change was made as a way of judging its impact.
Historical data can do that.
And the same can be true when considering combustion changes, equipment modifications or waste-heat recovery opportunities, etc.
Connecting Energy and Emissions Data
Energy and emissions data are often treated as separate datasets.
They don't have to be treated as such.
A more connected approach could include:
Energy Use
+
Process Conditions
+
Emissions
+
Flow & Temp
+
Production
↓
Operational Context
This gives the engineering teams more to evaluate when reviewing some change or deviation.
A reduction in energy use is much more interesting, compared to looking at just a production change, if that can be considered in the context of current CO, NOx SO₂ measurements, and the production level at the point of those emissions.
It still has to be engineered, but you're not relying exclusively on a correlation of variables to decide why something might be happening.
Dashboards Should Answer Questions
A dashboard isn't about showing every available data point.
For a monitoring dashboard, it should be asking "What are we wanting to get out of this data?"
For example, a refinery might ask:
What are our emissions levels?
How much did they change in the last 24 hours?
Did flow/change at the same time?
Did the event occur during a a certain operating period?
Were there any missing or questionable measurements?
How does the current data compare to other days?
It's about visualization as an interface, rather than a presentation of charts.
Where IoT Fits
Industrial IoT will help you to connect physical measurements with digital systems.
Sensors and analyzers, networks, data platforms, and dashboards each play a role in getting the right information organized. You don't have to build a complicated architecture because it's all about the following:
Physical Process → Sensor → Data → Context ← Analysis ← Human Decision
The human aspect to this is pretty important.
An automated alert can help notify a controller when a value moves outside of a range, but it can't explain what that out-of-range impact was on the process without additional context.
Applying the Approach to Emissions Monitoring
Industrial emissions monitoring could include things like gas emission analysis, particulate / dust measurement, and stack flow / temperature monitoring.
Emissions and Stack gives technologies related to gas emission analyzer, particulate and dust monitoring and stack flow and temperature measurement.
The useful questions about those sensors shouldn't just be about installation.
What decision is the measurement going to help me make?
That way, you can avoid going off of a well-intentioned approach that just wants to capture everything because that would be more data, but ends up with a disconnected bunch of measurements.
A Starting Place for People Wanting to Improve
For people looking to improve the existing monitoring architecture, this is a good process to consider:
Know what operational questions are important
Know what measurements are already available
Define where important data gaps are
Make sure your measurement and data quality is good
Connect related measurements
Build useful historical views
Add analytics where they solve a specific problem
Review with your engineers and operators
It doesn't require you to get everything going at once.
It builds a foundation that can grow with the facility's needs.
Final Thoughts
Better refinery energy and emissions management is about making sure that you're not just collecting data for its own sake, but rather capturing relevant information and using it to help inform better decisions.
A good monitoring architecture is all about the following at a very basic level:
Measure → Validate → Connect → Understand → Improve
While the technology underpins it, the real benefit is made when using reliable information to drive forward the right engineering and operational choices.
Top comments (0)