Thermo-Cognitive Advanced Process Control (TC-APC) models, simulates and optimizes your process in real time to improve yield, conversion and throughput, and reduce energy use, with a quick payback.
Boilers and combustion
An oxygen reading tells you the excess air is right for the coal you assumed you were burning.
Your boiler is safe, and it is running in a wide comfort band rather than a tight window.
The width of that band is a fuel bill you pay every hour.
What oxygen trim cannot do.
Trim is not the problem and we are not proposing to replace it. It holds oxygen at a number, accurately, which is what it was built to do. There are three things it was never built to do, and all three cost fuel.
01
It does not choose the number
The setpoint comes from a fixed excess-air table drawn up once. It does not know what this load, this coal and today's ambient conditions actually want, because nothing in the loop is working that out.
02
It does not know what you are burning
Nothing in the loop estimates calorific value, moisture or ash. So the controller is trimming air for a fuel it has assumed, and the assumption is a monthly average of something that changes by the rake.
03
It is a steady-state device
Trim settles a boiler that is already settled. Every ramp is therefore uncontrolled, and ramps are where the fuel actually goes: overshoot, excess air spikes, and a burst of CO on the way back.
TC-APC knows the coal you burn. The O2 target moves with each coal lot.
Engineering proposal. Begins with a data audit on your own operating data.
The science under TC-APC
A boiler's result is decided in the flame and the fuel.
The air your fuel needs, how completely it burns and what leaves the stack are all set in the flame, by the fuel you are burning that day. That's the science TC-APC is built on, and AI makes it sharper.
This is the field of our Chief Scientific Advisor, Prof. Pratim Biswas, Member, US National Academy of Engineering and former Dean of Engineering, University of Miami. His work includes:
- Paint to Power, with Chrysler and Ameren: paint solids from Chrysler's St. Louis plants replaced about 570 tons of coal a year at Ameren's 855 MW Meramec power plant. He led the project to capture mercury with the titanium dioxide in those solids. 2007 Chrysler Environmental Leadership Award. Washington University, 2008
- Aerosol and gas emissions measured at a working coal-fired power plant under five operating conditions. Journal of the Air & Waste Management Association, 2019
- Mercury and fine particles from burning coal from different seams. Fuel, 2013
- Fine-particle formation and mercury under O2-CO2 coal combustion. Energy & Fuels, 2006
- Control of toxic metal emissions from combustors using sorbents: a review. Journal of the Air & Waste Management Association, 1998
- Patent: enhanced capture of heavy-metal emissions from combustion exhaust. US 6,248,217, 2001
The table was drawn for a fuel you had in a particular week, and it has been the setpoint ever since. The optimum moves with every rake of coal, with the moisture in it, with the ambient, and with the slow state of the boiler itself.
Nothing in the loop is watching it move, so the margin is set wide enough to cover the worst case and carried at every load.
What that looks like on your plant.
None of this is exotic. Every one of these is visible in a month of your own trend data, which is why the first thing we would ask for is a month of your own trend data.
Efficiency drifts and nobody sees it
Sensor aging, coal variation and air leaks move the plant slowly away from where it was tuned. There is no reading that says so, so it is found at the next overhaul.
Control is reactive
Fuel and air move after the pressure has already dropped or risen. The correction is therefore always chasing, and the chase itself costs fuel.
Operators nudge
Dampers and feeder speeds get adjusted by hand through the shift to hold steam pressure. It works, it depends on who is on, and it is nobody's actual job.
Zones are not alike
One burner zone runs rich while another runs lean. The average oxygen at the economizer looks correct while carbon monoxide, unburned carbon and fouling say otherwise.
Transients overshoot
After every load change the excess air spikes and the loops oscillate before settling. On a plant with frequent swings this is where a large part of the loss lives.
The margin is inflated on purpose
Plants run higher excess air than they need so a CO alarm never appears. That is a rational thing to do when you cannot see combustion health, and it is a steady fuel penalty for insurance against a signal you do not have.
This is the same problem as every other application on this site. The variable that determines the money is not measured, so the plant runs a fixed recipe with a margin built into it.
What TC-APC adds
A layer above the controls you already have.
It does not replace your DCS, PLC, SLC or burner management, and it does not touch your safety logic. It sits above them, works out what the boiler should be doing, and writes setpoints down.
It chooses the oxygen number
The fixed excess-air table is replaced by a setpoint that is re-derived continuously for the load, the fuel and the ambient conditions in front of it, and that keeps re-deriving as the boiler ages.
It estimates what you are burning
Calorific value, moisture and ash are inferred online from the boiler's own response, so fuel and air are biased for the coal actually on the grate rather than the coal in last month's analysis.
It moves before the pressure does
Load is predicted ahead rather than answered afterward, and fuel and air begin adjusting before the header moves, so the transient is shaped instead of corrected.
It coordinates the loops together
Fuel, forced and induced draft fans and draft control are solved as one problem with the actuator limits and the delays between them respected, rather than as separate loops pulling against each other.
It notices drift
Air ingress, fouling and instrument bias show up as a disagreement between the model and the plant, which is an early warning rather than a finding at overhaul.
It shows its working
Live efficiency, combustion health and estimated coal quality on one screen, with every setpoint move logged alongside the reason and the model state that produced it.
What arrives on site
A system, wired in beside the one you have.
- ReadsLoad, steam pressure and flow, oxygen, draft and differential pressures, fuel and air flows, flue gas temperatures. Over Modbus or OPC‑UA from your existing DCS, PLC or SLC.
- WritesOxygen setpoint, fuel and air demand and draft setpoints, back to the same system. Your burner management and safety interlocks are untouched.
- Runs separatelyOn its own machine alongside your controls, so nothing in the existing logic is modified. Minimal wiring and no shutdown to install.
- GuardrailsHeartbeat watchdog and rate‑of‑change limits held in your own system. On any timeout or implausible estimate, control reverts with bumpless transfer to your existing setpoints.
- Before it writesRun in shadow mode on your live plant, writing nothing, until its predictions track what the boiler actually did, for as long as you want to watch it.
- For boiler makersThe same layer can be packaged with a new boiler, so the efficiency is part of what is sold rather than something the buyer has to chase afterward.
What this cannot do.
It cannot seal your air leaks
It will find them and say where, from the disagreement between the model and the plant. Closing them is still a maintenance job and the saving is yours either way.
It cannot fix bad instrumentation
If the oxygen probe is drifting or the flow elements are wrong, the estimate inherits that. It will usually tell you the instrument disagrees with the physics, which is the useful part.
It cannot beat the boiler's design
Heat transfer surface, burner arrangement and turndown are what they are. Running closer to the real optimum is not the same as raising the ceiling.
It does not touch safety
Burner management, flame supervision and the interlocks stay exactly as they are, under their own logic, and the supervisory layer writes nothing into them.
The mark
Find the optimum. Hold it while it moves.
The curve is a yield or selectivity response. The amber point is the optimum, and the dashed drop is where it sits on the operating axis.
The second half of the line is the part that matters. Finding an optimum is a study, and studies end up in drawers. Holding one while the feed, the fouling, the weather and the wear keep moving it is a control problem, and that is the work.
Start with your trend data
Send a month of operating data from one boiler
We will tell you how wide the comfort band actually is, how much of the loss sits in transients rather than in steady running, and what that is worth, with the uncertainty stated openly. No charge, and no commitment either way.
projects@yielddynamics.in Talk to IIT Bombay engineers Call +91 94230 07004Monday to Saturday, 9:00 to 19:00 IST
Or look at what we have actually run
Fourteen carbonation towers, validated over six months
This page is a proposal and says so. The one deployment we own is named, with the signed letter and the General Manager who still takes reference calls.