Wind turbine makers: longer blades, new load cases, control

Longer blades flex more and bring new load cases

What happens. A new turbine generation has longer, softer blades on a larger rotor. The inherited controller was made for stiffer machines, and the blades bend and twist in ways it was never tuned for.

What it costs. Loads on blades, hub and tower rise or have to be covered by heavier parts, and yield is lost where the controller backs off to be safe. Each new platform repeats the effort.

Why the usual fix fails. Carrying the old controller over and adjusting the gains is the quick way; it ends in limits that are conservative everywhere. Stiffening the blades adds weight and cost, which defeats the purpose of the larger rotor.

How a fixed-price remote diagnosis checks it. From the load simulations of your turbine, the model data and, where measured, prototype logs I look at where the controller is the limit. The fixed-price remote diagnosis tells you whether model-based control can take the new load cases within the loads that the design allows. I have worked on wind-turbine control at IAV (an automotive and energy engineering firm).

Background: Control that adapts to drift and wear

Pitch activity wears the bearings

What happens. The pitch system turns the blades constantly to hold speed and limit loads. A lot of small movements go into the bearings, and the gain that gives the best power or the best load reduction also gives the most wear.

What it costs. Pitch bearings and drives are expensive to replace, often with a crane, and failures show up as downtime. Fleets with a short pitch bearing life carry a maintenance cost that was not in the business plan.

Why the usual fix fails. Reducing the gain slows the pitch and reduces wear, but gives up load control. A filter on the pitch signal removes the movement without asking which movements were useful.

How a fixed-price remote diagnosis checks it. From turbine simulations and measured pitch logs I look at how much of the movement is useful and how much only chases noise. The fixed-price remote diagnosis tells you whether load-aware predictive pitch control can keep the load benefit with a fraction of the movement, and what the controller would have to know.

Background: Predictive control of process plants

Low damping of tower oscillations at certain speeds

What happens. At certain rotor speeds the tower swings and hardly damps itself. Tall towers and soft foundations make it worse. The turbine either passes quickly through those speeds or avoids them.

What it costs. Fatigue in the tower and the foundation, and a speed range that cannot be used. Avoiding a speed range costs output in the winds where it would be needed.

Why the usual fix fails. A stiffer tower is heavy and expensive. Skipping the speed range is simple and gives away yield every time the wind sits there.

How a fixed-price remote diagnosis checks it. From accelerometer or load signals of the tower, the rotor speed and the model of tower and foundation I look at how much damping an active control could add. The fixed-price remote diagnosis tells you whether active damping by the pitch or the generator is promising on your design, and what you would measure.

Background: Control that adapts to drift and wear

Gusts hit before the controller reacts

What happens. A gust reaches the rotor and the controller learns about it only from the response of the machine: speed, torque, blade loads. By then the load is already in the structure.

What it costs. Peak loads that set the design of blades and tower, and speed excursions that trip the turbine. Higher loads mean heavier parts or reduced output.

Why the usual fix fails. A faster controller reacts sooner, but cannot react before it has the information. Lidar (a laser wind measurement ahead of the rotor) adds hardware and cost and is not accepted everywhere.

How a fixed-price remote diagnosis checks it. From turbine and site data I look at what a short-term wind prediction, from lidar or from the farm’s other turbines, could contribute to the control. The fixed-price remote diagnosis tells you whether a preview in the control reduces the peak loads enough, and what data it would use.

Background: Predictive control of process plants

The effective wind speed is not measured

What happens. The anemometer sits behind the rotor, in the disturbed flow, and shows something other than the wind the rotor sees. The controller works with a measured wind that is wrong in the direction and size that matter.

What it costs. Controller quality suffers: the optimum rotor speed is missed, and the power curve is below what the machine could give. The sensor drifts and ages and shifts the whole picture.

Why the usual fix fails. A correction table from the anemometer to the true wind is made once and needs re-fitting at each site. A better anemometer cannot see what is happening across the whole rotor disc.

How a fixed-price remote diagnosis checks it. From the machine’s own signals (rotor speed, torque, pitch, power) I check whether the effective wind speed can be computed with an estimator. The fixed-price remote diagnosis tells you whether such an estimate is good enough to replace or check the anemometer, and what that would mean for the control.

Background: Computing unmeasured states for control

Strain gauges fail and the lifetime stays unknown

What happens. Strain gauges on blades and tower measure the loads that determine how long the machine will last. They fail in the field, in the weather, and the load record gets holes.

What it costs. Without a record nobody knows how much life is used. The maker cannot offer a lifetime extension, and the operator cannot prove one.

Why the usual fix fails. Replacing the gauges needs a technician and often a crane, and the new gauges fail again. Taking the design loads as the real loads is safe for the maker and wasteful for the operator.

How a fixed-price remote diagnosis checks it. From the turbine’s regular signals (accelerations, power, rotor speed, pitch) and from the periods in which the gauges worked, I check whether the loads can be estimated when the gauges are out. The fixed-price remote diagnosis tells you whether load estimation from signals is accurate enough to fill the gaps and to monitor the fleet.

Background: Computing unmeasured states for control

New grid requirements must be met by the control

What happens. Grid operators tighten the requirements: how a turbine behaves in a voltage dip, how it feeds reactive power, how fast it reacts to a frequency step. Each country and each operator has its own table.

What it costs. Each change needs a new round of control changes, simulations, tests and certification. A turbine that cannot meet a new rule cannot be sold in that market.

Why the usual fix fails. Adding parameters for each grid code into the existing controller makes it a patchwork that nobody wants to touch. Redoing the whole controller each time is too expensive.

How a fixed-price remote diagnosis checks it. From the existing controller, the converter model and the grid requirements I check which requirements the control can meet by parameter changes and where a structural change is needed. The fixed-price remote diagnosis tells you what an adaptation would take, and how far a model can reduce the number of test runs.

Background: Replacing hand-tuned parameter tables by a model

Floating platforms couple with the control

What happens. On a floating turbine the platform moves with the waves and the rotor thrust, and the controller acts on the same motion. A controller made for a fixed tower can make the platform swing harder.

What it costs. Instability or heavy loads, and conservative settings that give up yield. Failures at sea are hard and costly to reach and repair.

Why the usual fix fails. Detuning the controller until it is stable gives a safe machine with lower yield. Model tests of the platform are expensive and cover only a few sea states.

How a fixed-price remote diagnosis checks it. From the platform and turbine model, the wave data and, where they exist, test-basin results I look at how the control interacts with the platform motion. The fixed-price remote diagnosis tells you whether a coupled model with a new controller keeps the platform calm, and what the model needs from your design.

Background: Coupled machines with saturating actuators

Field tests are expensive and depend on the weather

What happens. A new control software has to be tested on a real turbine, in the right wind, with the right crew and a free slot. A campaign can stand still for weeks waiting for wind.

What it costs. Delay and cost for a prototype that is standing there, plus the risk of a test that goes wrong at full scale. Every iteration is slow.

Why the usual fix fails. More prototypes or more test sites cost money and do not fix the dependence on the weather. Doing the same test over and over to cover the cases is too slow.

How a fixed-price remote diagnosis checks it. From your test plans, the simulation model of the turbine and the results of earlier campaigns I look at which tests can be done in simulation first and which need the field. The fixed-price remote diagnosis tells you how a better simulation reduces the number of field days, and what has to be proven there.

Background: Replacing hand-tuned parameter tables by a model

Control specifications are incomplete

What happens. The control specification for a new turbine is written while the design is still changing. Interfaces, load limits, safety functions and operating modes are not all defined, and gaps are found when the code is already written.

What it costs. Rework in software and tests, and discussions between the control, load, electrical and safety teams about who was meant to do what. Time to market suffers.

Why the usual fix fails. Writing more detailed requirements by hand is slow and is outdated by the next design change. Hiring more control engineers helps only if the experienced ones are available to guide them.

How a fixed-price remote diagnosis checks it. From your existing specifications, interface lists and change history I check where the gaps and the contradictions sit. The fixed-price remote diagnosis tells you what an outside review of the requirements and the control architecture can clear up before the code is written.

Each site is tuned by hand

What happens. Parameters for each site or turbine variant are found by engineers: filter settings, speed limits, load limits, shutdown thresholds. The engineers work through a list, starting from a similar site.

What it costs. Engineering time per site, which grows with the fleet, and a risk that two sites are tuned differently for no reason. Errors show up late in operation.

Why the usual fix fails. Larger parameter tables only multiply the work. Copying from a similar site is quick and copies its mistakes and its mismatch with the new conditions.

How a fixed-price remote diagnosis checks it. From your parameter sets, site data and commissioning logs I look at which parameters really depend on the site, and how they could be computed from a model instead of tuned. The fixed-price remote diagnosis tells you whether automated parameterisation saves enough engineering time for your fleet, and what it would need as input.

Background: Replacing hand-tuned parameter tables by a model

Small turbines lack good control

What happens. Small wind turbines often run with simple control that cannot be afforded to be better. They are built in small series, with a low price, and the controller is a fixed table.

What it costs. Lower yield than the rotor could give, loads and noise that could be smaller, and a poor reputation for the whole class of machines.

Why the usual fix fails. A professional controller made as for a large turbine costs too much. Software that is cheap but fixed does not adapt to the site.

How a fixed-price remote diagnosis checks it. From the machine’s data sheet, a few measured runs and the cheap hardware available I look at how much a model-based controller on a small processor could add. The fixed-price remote diagnosis tells you what yield and load improvements are plausible, and whether the effort is worth it for your series.

Background: Predictive control on embedded hardware

Wind farm and hybrid plant control needs decision support

What happens. Teams at turbine makers have to build control for whole wind farms and hybrid plants with storage and solar. The skills sit in small teams, and the control engineers who also know optimisation are hard to find.

What it costs. Projects wait for the same few people, and the plant-level functions arrive late or as a patchwork of rules. Customers want a plant that follows the market, not only the wind.

Why the usual fix fails. Hiring is slow in a small field. Turbine control and plant optimisation are different disciplines, and combining them is where projects get stuck.

How a fixed-price remote diagnosis checks it. From your plant model, the controller structure and sample operating data I check which plant-level tasks a predictive optimisation can take. The fixed-price remote diagnosis tells you what such a plant controller would look like, what decision support the operators need and how much of it an outside engineer can set up with you.

Background: Economic predictive control

Converter control is hard to verify and faults hard to trace in the field

What happens. Converter control software has to be verified against many grid and fault cases, and when something in the field behaves oddly, it is hard to tell whether the cause is the converter, the controller or the grid.

What it costs. Long verification cycles, field returns with unclear causes and engineers who spend weeks on one event. The cost is time, and a worse reputation with customers who wait.

Why the usual fix fails. More test benches test what was thought of in advance. A field fault happens in a mix of conditions no bench reproduces.

How a fixed-price remote diagnosis checks it. From field recordings of the faults, the converter model and the test results I check whether the events can be reproduced in simulation and where the cause lies. The fixed-price remote diagnosis tells you whether diagnosis from field data finds the cause faster, and what recordings would be needed.

Background: Replacing hand-tuned parameter tables by a model