Hydro turbine wicket gate assembly

Hydro Turbine Control and Protection

Hydro turbine wicket gate assembly
Hydro turbine wicket gate assembly

A hydropower unit is a control problem with a water column attached. The governor can only move as fast as the penstock will tolerate, and every decision about speed of response is really a decision about pressure surge.

We have worked on two kinds of machine on hydroelectric plants: a Pelton unit, where the jet is cut by needles and a deflector, and vertical Francis units with a spiral case and wicket gates. They fail differently, are regulated differently, and need different things from a control system.

Pelton — an impulse machine

Horizontal Pelton unit rated 1.5 MW at 500 rpm, two nozzles with needle control and a deflector ahead of them. The control problem is nozzle staging — the second nozzle brought in at a defined opening of the first — plus deflector timing, which is what actually protects the penstock, because the deflector can cut the jet in seconds while the needles take twenty.

Measured on the machine: deflector servomotor closing in 2 seconds over its full 190 mm travel; needle travel of 60 mm taking 36 seconds to close and 9 to open. The load characteristic is strongly non-linear — 86 % of cylinder travel gave 130 kW, 95 % gave 950 kW — which is the part a controller tuned on paper gets wrong.

We delivered this plant complete: specification, design documentation to the national standard for automated system documentation, cabinet build, field instrumentation, commissioning and witnessed testing, and it runs unattended.

Francis — a reactive machine

Vertical Francis unit with a spiral case, wicket gates, turbine cover, thrust bearing and upper and lower brackets. The instrumentation set is different from an impulse machine in kind, not only in quantity: twelve stator winding points and twelve core points, twelve thrust bearing pads, four points on each guide bearing, shaft seal temperature, water level on the turbine cover, spiral case pressure, and absolute vibration and shaft runout on both brackets.

Settings from the unit we commissioned and maintain: stator winding and core trip at 120 °C with one-out-of-twelve voting and alarm at 110 °C; guide bearing and thrust pads trip at 70 °C, alarm at 65 °C; turbine guide bearing at 65 °C with its oil bath at 55 °C; shaft seal at 40 °C; spiral case pressure trip at 0.72 MPa; water level on the turbine cover at 191 mm with a 2-second delay. Overspeed trips at 160 % of rated. Below 15 Hz of shaft frequency for 60 seconds the gates are driven closed and the brakes applied.

The sensor-failure logic is as interesting as the trip settings. A single failed stator sensor raises an alarm; four failed sensors out of a group of four, or all twelve, trip the unit. A protection scheme that trusts a measurement it can no longer make is worse than one that admits it is blind.

Shear pin failure on the wicket gates, the shield release and the brake sequence are all in the same scheme, because on a reactive machine the gates are the only thing between the water and a runaway.

Governing systems, from oil to all-electric

A hydro governor can be built several different ways, and the differences matter far more than the brand of controller sitting above them.

Oil-hydraulic on Pelton nozzles. An accumulator unit feeding the needle and deflector cylinders, with the deflector on its own much faster closing path. The accumulator is protected at both ends — a shutdown on low pressure and another on high — and the drain tank level is watched against limits above and below normal. We supervise spool position against the controller's demand and trip on a mismatch of more than 2 % held for 2 seconds, which catches a sticking spool before it becomes a runaway. In the rejection tests the accumulator never fell below 22.5 MPa against a 23–25 MPa standing pressure, which is the margin that says the system is sized correctly.

All-electric wicket gate drive. Gates driven by electric actuators through frequency converters, with no oil pressure system at all. The converters are supervised at both 380 V and 364 V, and a stored-energy cabinet holds enough charge to close the gates when the supply is gone — alarmed below 300 V, because a stored-energy cabinet that is quietly flat is the same thing as no gate closure at all. On a plant like this, proving the stored-energy capacity is a maintenance task of the same weight as proving a UPS battery.

Electric linear actuators for modulating duty. Self-locking linear actuators with stepper or brushless DC motors, up to 15 kN and 85 mm of stroke, EN ISO 5210 F05 and F07 valve attachment, electronic torque or travel seating, rated for continuous modulating duty at up to 1,200 starts per hour. These are what replaces a small hydraulic positioner when the machine is small enough and the oil system is not worth keeping. We installed one on the auxiliary gate at the Pelton plant.

Retrofits we have designed

Distributor block replaced by a proportional servo valve. A two-stage pilot-operated proportional directional valve with integrated electronics and electrical position feedback of the main spool, rated to 350 bar, mounted to ISO 4401, with a spring-centred main spool. The gain is not only resolution — it is that the spool position becomes a measurement the control system can supervise, instead of a mechanical state nobody can see.

Making loss of cabinet power a safe state. On the Pelton unit the start/stop mechanism used two solenoids, pulsed, with a mechanical latch holding the block in whichever state it had last been driven to. Losing the control cabinet's supply therefore left the machine exactly where it was. We specified the replacement: one spring-loaded solenoid, energised to arm and to admit oil to the main spool, so that removing power lets the spring drive the spool to its safe position and close the main cylinder and the nozzles. A protection function implemented in a spring is a protection function that does not need the controller to be alive.

Overspeed and vibration added to an existing scheme. On the vertical machine, shaft speed sensors wired to speed transducers with their relay setpoints at 160 % of rated speed, taken into the trip scheme and onto the shield release; an unused fourth speed sensor cabled in and brought into service; vibration transducers added and collected by a separate controller whose analogue outputs feed the unit control system. That work is described under Hydro Vibration and Shaft Monitoring.

Unit control

Nozzle control with staged cut-in — the second nozzle brought in at a defined opening of the first — hydraulic cylinder positioning through the governor block and actuator, oil pressure unit management, and the pre-turbine and bypass valve sequence that equalises pressure before the machine is opened.

Start-up is not instantaneous and should not pretend to be: the regulating oil system is warmed by heating the drain tank hours in advance and then cycling the hydraulic cylinder through full travel with holds at each end, because a governor working on cold oil is a governor with a different response than the one that was commissioned. On an unattended plant that warming sequence is an algorithm in the control system, not a line in an operating instruction, because there is nobody there to follow the instruction.

Protection design and voting

Protection is a set of decisions about what a single failed sensor should be allowed to do. On the Pelton unit:

ProtectionSettingVoting
Overspeed600 rpm2 sensors, 1 out of 2
Generator winding temperature80 °C6 sensors, 1 out of 6
Guide and thrust bearing temperature80 °C2 sensors, 1 out of 2
Oil pressure unit accumulator pressure, low16.3 kgf/cm²single
Oil pressure unit accumulator pressure, high29 kgf/cm²single
Drain tank oil level, low150 mm below normalsingle
Drain tank oil level, high115 mm above normalsingle
Vibration velocity, worst transducer11.2 mm/s RMShighest reading
Nozzle cylinder spool position mismatch2 % for 2 scomparison
Speed sensor mismatch100 rpmcomparison

Plus generator electrical protection, outgoing line protection, generator breaker failure, excitation system failure, loss of the DC and AC incomers, and failure of two of the three power supply units.

One-out-of-two on overspeed is not arbitrary. It is a judgement about which way to fail: a spurious trip is cheap, a missed overspeed is not.

Load rejection, measured

A governor is only as good as its worst rejection. On the Pelton unit we ran a staged load rejection programme on the running machine, with the generator breaker opened by hand at the switchgear at each step and six people stationed around the plant, each recording a different parameter.

Rejected loadPeak speedPenstock pressure, peakOil pressure unit, minimum
0.10 MW514 rpm21 MPa22.5 MPa
0.15 MW560 rpm20 MPa23 MPa
0.25 MW500 rpm—23.5 MPa
0.45 MW596 rpm21.5 MPa24 MPa
0.95 MW—21 MPa24 MPa

Standing pressure in the penstock was 19 MPa, and the rejections pushed it to 21.5 MPa and pulled it down to 16 MPa on the rebound. The deflector servomotor closed in 2 seconds; the nozzle needles took up to 20.

The largest rejection the water available allowed was 0.95 MW against a rating of 1.5 MW, so the machine was never asked for its worst case. Even so, the number that matters is 596 rpm against an overspeed trip set at 600 — 119 % of rated speed against a trip at 120 %. Four revolutions per minute of margin, measured rather than calculated, on a machine that had to stay in service. After every step the penstock expansion joints were walked and inspected, and a man watched the vacuum breaker valve at anchor block 31 and logged every time it lifted.

We took the load characteristic the same day, so that the relationship between cylinder position and output was a measurement and not an assumption. The curve is steep in the middle, which is exactly where a load controller tuned on paper gets into trouble.

Governor structure

What a hydro unit should be controlling depends on what it is connected to and what the water is doing. The control system has to hold all of these and move between them cleanly:

  • Speed control, no-load — from first movement to synchronising, holding rated speed with no electrical load to stabilise it. The least stable condition the governor ever sees.
  • Frequency and active power control, synchronised — load setpoint with droop, so the machine shares load correctly with the rest of the system.
  • Isolated operation — the machine sets the frequency rather than following it, which is the mode the penstock constrains hardest.
  • Daily load schedule — a load curve loaded into the controller and followed without an operator, which is how an unattended plant participates in a dispatch programme at all.
  • Headwater level control — the unit's load moved to hold the level in the settling basin, so the machine follows the inflow. Level is measured by radar in the basin, brought back to the unit cabinet over fibre, and the loop writes to the load setpoint rather than directly to the nozzles.
  • Opening control, manual — nozzle or gate position held directly, with no outer loop, for testing and for conditions the outer loops cannot handle.

Moving between them without a bump

A plant spends its life changing which of these is in charge: synchronising, losing the grid, inflow rising, dispatch taking over, a test starting. Each transfer is a place where the machine can be kicked.

Transfer has to be bumpless in both directions and has to happen on two kinds of trigger — an operator command, and an automatic safety criterion when the plant's state no longer suits the active mode. The standing controllers track the active one, so that whichever takes over starts from the current output and not from its own stale integral. Done badly, this is the single most common source of load swings on a small plant, and it is invisible in a factory test because the condition that triggers it only happens on the river.

Why the characteristic has to be measured, not assumed

The relationship between opening and output is not linear, and it moves with head. On a run-of-river or derivation plant the head changes with the season and with what the intake is passing, so a controller tuned once is right on one day of the year and approximately right on the others.

That is why we take load characteristics and rejection records on the machine at commissioning: so that the gains, the limits and the nozzle staging point rest on a measurement rather than on a manufacturer's curve. Where the head range is wide enough to matter, the controller is parameterised by measured head rather than fixed.

Failure behaviour of the governing subsystem

Failure of the electronics has to stop the machine, not freeze its output. A short interruption of the supply must not produce a load step, which means the controller has to ride through it rather than restart and the actuator has to hold position without a command. And re-entry has to be bumpless in any operating state, after a loss of supply or a repaired fault, whatever the unit is doing at the time.

To that we add the requirement that survives all of the above: a manual path to the nozzles or gates from a local station, not passing through the controller at all. A governor that cannot be operated when its PLC is dead is a governor that defines its own worst case.

Failure modes we test for

A protection scheme is only as good as the failures it has actually been exercised against. On a unit shutdown condition we test, one at a time:

  • failure of the governor block, including loss of its action on the actuator
  • failure of the actuator itself
  • sharp loss of oil pressure in the oil pressure unit
  • loss of the 0.4 kV auxiliary supply downstream of its automatic transfer
  • failure of the DC supply unit
  • failure of the unit controller or of its I/O subsystem

Each is applied deliberately, with personnel stationed at the governor stop coil and at the pre-turbine valve to close them by hand if the auxiliary supply goes with it.

Test programmes and acceptance are covered under Factory and Site Acceptance Testing, and the telecontrol testing that goes with an unattended plant under Remote Operation of Unattended Hydro Plants.