Power and heat plants: built for base load, now ramping daily

Boilers and engines face tighter nitrogen oxide and dust limits

What happens. Rules for medium-sized combustion plants have tightened for nitrogen oxides and dust. Older boilers and engine plants meet them with little room, and the combustion is set for the worst case so that no limit value is exceeded.

What it costs. Retrofit measures such as flue-gas cleaning, extra fuel for safety margins, and the risk of a shutdown order if the limit is exceeded.

Why the usual fix fails. A retrofit is expensive and takes a long time. Setting the combustion conservatively costs efficiency every hour. A measuring instrument tells the figure after the fact and does not steer.

How a fixed-price remote diagnosis checks it. From operating logs (load, air, fuel, flue-gas values) and emission records I look at how the emissions depend on the operating point. The fixed-price remote diagnosis tells you whether emissions computed online can let the combustion run just inside the limit.

Background: Computed measurements

Plants built for base load must ramp every day

What happens. Plants designed to run steadily day and night now follow solar and wind. They ramp up in the evening and down at midday, over and over, with loads they were not designed to follow.

What it costs. Material fatigue, more maintenance, fuel burnt in part-load operation and start-ups that follow one another.

Why the usual fix fails. Slow fixed ramp rates protect the plant and lose market income. Faster ramps by operator decision risk damage that shows up years later. A tighter control with the old loop structure does not know the stress limits.

How a fixed-price remote diagnosis checks it. From load logs and metal temperature data I look at how close the daily ramps come to the stress limits. The fixed-price remote diagnosis tells you whether predictive load control within the stress limits can ramp faster without extra wear.

Background: Industrial process predictive control

Thick-walled parts limit how fast the plant can ramp

What happens. Drums, headers and turbine casings have thick walls. If the temperature changes too fast, the inside and outside differ and stress builds up. The plant has ramp rates that protect the parts, set for the worst case.

What it costs. Slow starts and slow load changes, income lost to the delay, and still a worry that the limit is reached on a bad day.

Why the usual fix fails. Fixed ramp tables do not know the actual wall temperatures. Putting in more thermocouples measures the surface. Reducing the margin on the table risks life consumption nobody can see.

How a fixed-price remote diagnosis checks it. From temperature and pressure logs of past starts and ramps I estimate the stress in the walls and compare it with the allowed value. The fixed-price remote diagnosis tells you whether stress computed online can let the plant ramp as fast as the parts allow.

Background: Computing unmeasured states for control

Starts take long and burn fuel before any income

What happens. A start follows a procedure with holds and checks. It takes hours, and the plant burns fuel and uses auxiliary power long before it feeds into the grid.

What it costs. Fuel without income, missed market windows, and wear from every start.

Why the usual fix fails. Starting earlier costs more fuel. Cutting the holds by hand risks stress and trips. A start recipe made for the cold case is used for the warm case as well.

How a fixed-price remote diagnosis checks it. From logs of past starts (temperatures, pressures, loads, durations) I look at which holds are needed for the actual state and which only follow the recipe. The fixed-price remote diagnosis tells you whether an optimised start trajectory is realistic for your units.

Background: Trajectory optimisation and setpoint generation

Combustion runs with excess air as a margin

What happens. Boilers are operated with more air than the fuel needs to be sure that combustion is complete. The air value is set at commissioning and corrected when the carbon monoxide rises.

What it costs. Heat leaving with the extra flue gas, more nitrogen oxides at high excess in some burners, and more fan power.

Why the usual fix fails. An oxygen probe gives one point in a flue-gas duct with layers. Taking the margin out by hand lets the carbon monoxide rise on the day the fuel changes. A tighter setting needs a controller that knows the fuel.

How a fixed-price remote diagnosis checks it. From operating data (air, fuel, oxygen, carbon monoxide, load) I look at how much excess air the plant carries and what limits it. The fixed-price remote diagnosis tells you whether combustion optimisation can lower it safely.

Background: Economic predictive control

Steam temperature spray control oscillates

What happens. The steam temperature behind the superheater is controlled by spraying water. The reaction is slow and delayed, and load changes disturb it. The spray valves swing back and forth and the temperature oscillates around the setpoint.

What it costs. Thermal cycling that shortens tube life, a setpoint kept low for safety, and efficiency lost.

Why the usual fix fails. Standard controllers tuned for one load do not work at another. Detuning them makes them slow. Operators put the valves on manual in difficult phases.

How a fixed-price remote diagnosis checks it. From trend logs of spray valves, steam temperatures and load I look at the dead time and the load dependence. The fixed-price remote diagnosis tells you whether predictive steam temperature control can hold the temperature closer to the limit.

Background: Industrial process predictive control

Engine plants run by heat demand only

What happens. Combined heat and power engines run when the heat is needed. Power prices are moving a lot during the day, but the engines do not care.

What it costs. Income from the power market that is not collected, because the engine could have run in the expensive hours and heat from a store could have covered the rest.

Why the usual fix fails. A fixed daily schedule misses the days that differ. Switching by price alone empties the heat store at the wrong time. A store without a plan for its use is capital that does not work.

How a fixed-price remote diagnosis checks it. From heat demand, engine data and price series I look at how much the plant would earn with heat- and price-led operation. The fixed-price remote diagnosis tells you whether a dispatch with store and prices pays for your plant.

Background: Economic predictive control

The state of the heat storage tank is unknown

What happens. A heat storage tank has a hot layer on top and a cold one below. A few temperature probes show the layers at their heights. How much usable heat is in the tank is a guess.

What it costs. Wrong dispatch decisions, engines started when the store was nearly full, heat supply shortfalls when it was less full than thought.

Why the usual fix fails. More probes cost money and the layers shift between them. A simple state of charge from the average temperature is wrong when the layers mix. Operators plan with a safety margin.

How a fixed-price remote diagnosis checks it. From probe data and charging and discharging flows I look at how well the layers and the usable heat can be computed. The fixed-price remote diagnosis tells you whether a computed storage state can improve the dispatch.

Background: Computed measurements

Compressor fouling in gas turbines is not seen

What happens. Dirt builds up on the compressor blades of a gas turbine. Efficiency and output drop slowly, in a way that looks like ordinary scatter and ageing. A wash is done on a calendar.

What it costs. Fuel and output lost for months, and washes done too early or too late.

Why the usual fix fails. Washing more often costs downtime. Comparing against the commissioning figures hides the effect of weather and load. Waiting for an alarm means waiting for a big loss.

How a fixed-price remote diagnosis checks it. From operating data (pressures, temperatures, speeds, output, ambient conditions) I look at how fouling can be separated from other effects. The fixed-price remote diagnosis tells you whether a fouling estimate can plan the washes.

Background: Computed measurements

Cooling tower fans and pumps run at fixed settings

What happens. Fans and pumps of the cooling tower run in fixed stages. Ambient temperature and plant load change, but the stages follow simple rules.

What it costs. Power for fans and pumps beyond what the cooling needs, and a condenser pressure that is not the best compromise between cooling power and plant output.

Why the usual fix fails. Variable-speed drives cut the cost of a stage but not the logic. A rule per season does not see the day-to-day weather. Operators err on the side of cooling.

How a fixed-price remote diagnosis checks it. From logs of condenser pressure, cooling water temperatures, ambient conditions and drive power I look at the best operating point against the current one. The fixed-price remote diagnosis tells you whether optimised cooling saves net power.

Background: Economic predictive control

Converting a unit from coal to gas or biomass changes its dynamics

What happens. A unit is converted from coal to gas or biomass. The flame, the heat transfer and the response to load changes are different from before, and the old control concept is built around the coal behaviour.

What it costs. Long commissioning, unexpected oscillations, limits on output, and a unit that cannot do what the conversion promised.

Why the usual fix fails. Keeping the old loops with new parameters hides the changed structure. Waiting for operating experience takes seasons. A copy from a unit that was converted elsewhere does not fit your boiler.

How a fixed-price remote diagnosis checks it. From the design data, trial results and the old control I look at which dynamics change and which control parts need a new concept. The fixed-price remote diagnosis tells you what a concept for the converted unit would contain.

Background: Industrial process predictive control

Hydrogen co-firing brings a fuel with unknown behaviour

What happens. Turbines and engines are meant to burn natural gas with an admixture of hydrogen. Hydrogen burns faster and hotter, with a different flame, so the combustion dynamics and the limits shift.

What it costs. Risk of flashback, oscillating combustion and higher nitrogen oxides, and trials that cost running time on units that earn money.

Why the usual fix fails. Trial and error on the real unit is risky. The manufacturer’s data cover test conditions, not your fuel mix. A fixed limit on the hydrogen share keeps the potential unused.

How a fixed-price remote diagnosis checks it. From the unit’s data and any trial results I look at how the combustion responds to composition changes and what a model for it needs. The fixed-price remote diagnosis tells you what a model and control trial would take.

Delivering reserve power to the grid brings penalties for missing it

What happens. A plant that sells reserve power must follow a signal from the grid operator quickly and exactly. The plant control was built for steady operation, and its response is slow or oscillating.

What it costs. Penalties and lost qualification when delivery is late or incomplete, and income given up because the plant offers less than it could.

Why the usual fix fails. Offering a smaller band is safe and unprofitable. Faster loops without a view of the plant limits cause trips. Testing on the real market gets paid in penalties.

How a fixed-price remote diagnosis checks it. From step response tests and operating logs I look at how fast and how accurately the plant follows set values and where it falls short. The fixed-price remote diagnosis tells you whether reserve-capable control is possible with your plant.

Background: Industrial process predictive control

Fans and pumps in the plant are oversized and use own power

What happens. Fans, pumps and compressors of the plant were sized with margin and run at throttled or fixed settings. The plant’s own power use stays high at part load.

What it costs. Power used within the plant, which reduces what it can sell, and wear from throttling.

Why the usual fix fails. Exchanging machines is an investment. A variable-speed drive saves only if the control uses it well. Each auxiliary is optimised alone, without the effect on others.

How a fixed-price remote diagnosis checks it. From power and flow logs of the auxiliaries across the load range I look at where most energy goes beyond need. The fixed-price remote diagnosis tells you which optimisation of the auxiliaries pays.

Background: Economic predictive control

Faults are found late and end in outages

What happens. A bearing, a pump, a sensor drifts slowly. The alarm limits trigger only when the fault is advanced, and the first sign is a trip.

What it costs. Unplanned outages, repair under time pressure, and lost income during the outage.

Why the usual fix fails. Fixed alarm limits are wide enough to avoid false alarms and so late. More alarms are ignored. A calendar of inspections misses faults that develop in between.

How a fixed-price remote diagnosis checks it. From operating data and records of past faults I look at whether early signs exist and how a model of normal behaviour could flag them. The fixed-price remote diagnosis tells you whether model-based fault detection works for your plant.

Unit controls from the nineties cannot be changed

What happens. The control of a unit dates from the nineties. Spare parts are hard to find, only a few know the system, and every change is a risk.

What it costs. No improvements possible, long outages when something fails, and a modernisation that is postponed from year to year.

Why the usual fix fails. A one-to-one replacement keeps the old concept and its weaknesses. A full replacement needs a long outage. Without a clear picture of what the old control does, no one dares to start.

How a fixed-price remote diagnosis checks it. From the documentation and logs I describe what the control does and which functions matter. The fixed-price remote diagnosis tells you what a new concept would need to contain, and in what steps it could be introduced.

Fewer operators per shift run the same plant

What happens. Shifts are staffed with fewer operators, while the plant needs the same attention. Routine tasks such as starts, load changes and checks take their time.

What it costs. Risk of errors in busy phases, overtime, and difficulty keeping skilled operators.

Why the usual fix fails. More alarms add workload. Checklists on paper do not shorten tasks. Hiring is hard where the shift work is unattractive.

How a fixed-price remote diagnosis checks it. From operator actions in logs I look at which manual tasks are repeated, rule-based and safe to automate. The fixed-price remote diagnosis tells you which tasks could be taken over by the control and how.

Several units sell into one market without a joint plan

What happens. A utility runs several units and sells into the same market. Each unit is dispatched on its own, and nobody weighs heat demand, fuel and prices across all of them.

What it costs. Lost margin from units that run at the wrong time, or that run against each other.

Why the usual fix fails. Local rules do not know the whole portfolio. A spreadsheet plan the night before cannot follow the day. Merging the plans needs a tool that can handle the technical limits of each unit.

How a fixed-price remote diagnosis checks it. From unit data, heat demand, prices and past schedules I look at how much a joint plan would earn against the separate ones. The fixed-price remote diagnosis tells you whether portfolio dispatch is worth building.

Background: Economic predictive control

Control and protection design for gas-turbine auxiliaries lacks specialists

What happens. Equipment makers for power plants need control and protection design for the auxiliaries of gas turbines: starting systems, lube oil, fuel supply. The work needs people who know both the process and the requirements, and few are available.

What it costs. A design that waits for people, requirements that change late, and a delivery date at risk.

Why the usual fix fails. Re-using the design of the last project carries over its errors. New staff need time to learn the requirements. Handing the design to an outside party without clear requirements returns rework.

How a fixed-price remote diagnosis checks it. From the requirement documents, process diagrams and existing control descriptions I look at what the control and protection functions must do and where requirements are open. The fixed-price remote diagnosis tells you how the design could be split and what is needed to complete it.