|
INSTALLATION AND OPERATION
The Fair Isle people have their own myths
about island life, which include the mysterious Trows who live in the caves
around the cliffs and go about the island, unseen, making mischief for
honest folk. We must have flushed out a whole tribe!
It started the moment the first shipment
arrived at the island harbour. The ship's crane refused to lift the blades
in their protective crate and the blades had to be taken ashore
individually. The process of moving the unprotected blades to site was nerve
wracking but uneventful, and the assembly of the turbine in the prone
position was accomplished without major snags. The lift was started and was
not going well. After raising the nacelle some three metres it was clear
that there was a problem with the wire rope
puller, although it was
working well inside its rated
capacity. The nacelle had to be lowered again onto a ground support. A quick
trip to Shetland produced a
secondhand wire rope puller
of the same capacity, which, after stopping off the
tackle, was placed on the wire in tandem with the new
puller.
This rig allowed us to erect the turbine safely, and avoided the need to
dismantle and store the sub assemblies.
Commissioning of the turbine against its
dump load; always exciting with stand alone machines, went smoothly, and the
temporary tie into the system was hooked up. Operation in light winds was
satisfactory, but, as the wind strengthened, more load automatically came on
on the south heating cable, which was being used to feed the turbine output
to the original control panel. This produced a volt drop in the cable, and
when wind power was also fed to the service network there was an irritating
flicker on incandescent lamps. It was then noticed that, at fairly high
power outputs, the turbine was "shivering". This effect was tracked down to
flexing of the generator feet, which caused a slight side to side rocking
motion. This in turn was varying the belt tension, and modulating the
rotational speed of the generator very slightly. The load control units are
sensitive to frequency (speed) and were responding to, and maintaining, this
disturbance vigorously. A temporary brace to resist the rocking motion
alleviated the problem whilst an engineered solution was prepared. This
modification reduced, but did not eliminate, the flicker effect.
The blade settings and speed of rotation
had been arrived at, in full discussion with the blade manufacturers, to
give good self-starting and stall at l00kW. After the initial stiffness wore
off we got good starting, but
found that stall was close to l40kW. This is where a belt drive link can
save the day, and we had a step
change wheel in hand which brought the speed - down to give stall
at 120kW
The old machine was lowered and
dismantled. The generator and main gearbox were removed and returned to the
factory for a full overhaul, including new bearings and shaft seals. The top
section of the tower went to Shetland for detail modification of the top
flange. We returned to the factory to progress the rebuilding of the nacelle
for the 60kW turbine and the remaining control panels.
The
compressed air supply in the turbine nacelle is only used to release the
mechanical brake on a `start' command. Thus, if the air system is leak
proof, the compressor motor is only called upon to recharge the receiver
after four `starts', and takes about 3 minutes to do so. In practice it is
difficult to make the system totally leak free, and the compressor will,
typically, run once or twice each day. At the end of a charging run, or when
the motor supply is interrupted, the delivery line from the compressor to
the receiver is vented, as the motor can not start against pressure. We had
considered the possibility that a changeover of supply would be long enough
for the motor to stop, though not long enough to vent the delivery line,
but, after all, the possibility of a compressor
run
coinciding with a
changeover was remote. We overlooked the fact that a compressor run is
likely to follow a start command, and did not anticipate that the turbine
could start, take load and initiate a changeover within the compressor run
time. After a number of thermal trips, following compressor stall, the
compressor failed. A timed disconnect circuit was added to automatically
hold off
the
compressor supply for 20 seconds after any supply interrupt, however brief.
On the 3rd of March 1997 at 05.48 the
turbine suffered a serious failure of the rotor hub flange on the wind
shaft, which resulted in the blade set being damaged beyond repair. All work
on the refurbishment of the original turbine was
put on hold to await the recovery
of the broken shaft and a detailed analysis of the fracture. The design of
the shaft and flange had been the subject of a finite element analysis of
working stresses to determine its fatigue life, and had been found
satisfactory. This analysis was redone with more recent software and higher
loadings. The fatigue life predicted exceeded 30 years. The author's view
is that the failure must have been due to a defect in either material
or process, but, as the hub had
spun on the fracture for at least several revolutions, the evidence of
cause, if any, had been destroyed. As a precaution, the shaft flange detail
design was revisited and all stress raisers eliminated or reduced. We
also added 100% inspection during
manufacture by our sub-contractor. As a further precaution, the wind shaft
brake disc and the rotor hub were returned to the factory to check
for fractures or distortion. These
were, rather surprisingly, undamaged. The tower, which had taken direct
strikes from all three blades, had some bent bracing bars. These were
successfully straightened with purpose-made tools on site. No weld fractures
or cracks were found on the tower. Up to the instant of failure the turbine
had generated power for 713 hours producing 53,966 kWh an average output of
75.7 kW.
The new
shaft and the hub and brake disc for the l00kW turbine were shipped to the
island and the nacelle rebuilt and assembled to the tower to free the
special trailer. The new nacelle assembly and blades sets for both turbines
were shipped with the control panels shortly afterwards. The 1OOkW turbine
was' bladed and erected first, with the wire rope
puller
now working properly. The erection tackle was moved to the
old site and assembly proceeded well until we found a problem pitching the
blades. A short delay followed, to get the hub bolt holes slotted two
degrees on Shetland, and assembly was completed without further difficulty.
The 60kW turbine was connected up to its
turbine control panel and dump load, and was commissioned without any
problems. However, it was not stalling until close to 1IOkW output. A change
of belt wheels reduced the speed
and brought the stall point down to 85kW. This was within the rating of
gearbox and generator, but uncomfortably close to the 90kW rating of the
dump load. Additional load was
added by putting a 3kW water heater in parallel with each of the last five
sections of the dump load. These five immersion heaters were mounted in a 30
gallon tank in the diesel house, complete with overheat thermostat and float
switch. These last sections of dump will not be powered except when the
turbine is running off the system.
The changes to the network, and cabling to
the network control panel, were planned to minimise outages on the service
distribution, and for three days this supply was provided directly from the
diesels via a jumper cable. After the network panel had been checked out,
including the connections driving the mimic diagram, the diesel and service
cables were completed to the network control panel, and we were now able to
commission the complete system.
The testing of the first three plant
options, diesels only, diesels plus 60kW turbine and diesels
plus 100kW turbine, went
reasonably well. It was noticed that there was still some light flicker
present running on the 100kW turbine as sole supply, even though the
generator vibration problem was fixed. A clue was given that this tended to
be more marked at specific frequencies, suggesting that the controlled
domestic load frequencies had become bunched. This was checked and found to
be the case. A new programme was worked out for each section of the heating
distribution network, to balance the phase loadings and provide a smoother
growth of load with frequency. Such programmes can not be perfect, but this
exercise did improve the running of the turbine and reduced the flicker
until it was barely noticeable.
We did succeed in soft coupling the
turbines one at a time to the diesel. The 100kW turbine very quickly pulled
the diesel up to overspeed,
as it is nearly four times more powerful. The 60kW
turbine was tested in a steady light wind with about 15kW of capture, and
ran coupled for some time. However, it too pulled up and overspeeded the
diesel after a slight increase in wind speed. Clearly this form of coupling
will only work if the stall power of the turbine is less than the load on
the diesel, or if the diesel
governor is such that it will permit a limited degree of
overspeed
without tripping the engine.
Operating with diesel and both turbines
was much more exciting to commission, and' as it was winter time, work went
on after dark. The most perplexing problems were those which only occurred
at certain wind speeds or wind directions, and in specific modes of
operation. For example, running uncoupled, out of guarantee, in marginal
winds, the lead wind turbine would overload, slow down and drop the service
load. The lights would go out, the control system would recognise the second
turbine was up and running to heating and promoted it to lead turbine. The
proving period ran and the service was reconnected and the lights came on.
The two minute wait seemed interminable! The logic was changed to allow each
turbine to prove it was ready beforehand and wait. This allowed a fast
changeover and service was restored in less than one second. This worked
well, but inside guaranteed hours of supply the diesel had been told to
start, and by the time it was up to speed its load had been taken and the
engine then timed out and stopped. To prevent this,
a different and longer set up time
was introduced when in guarantee time. Most of the problems were resolved by
adjusting timing and set levels to obtain a satisfactory service from the
system. Some interlocks were added, for example to inhibit coupling when
either turbine power went negative. This particular change reduced the
outage time when the lead turbine failed service on power loss.
The final stage of commissioning, with the
synchronisation of the turbines, was anticipated to be difficult. Both wind
turbines have brushless alternators as generating plant, and each is
equipped with quadrature droop compensated automatic voltage regulators to
aid parallel operation. The normal set up procedures, of running at constant
speed and switching load on and off, were out of the question with this
system. We resorted to setting voltage exactly as required, when each
turbine was running steadily between 60% and 75% rated power. The system
neutrals, as mentioned earlier, are bonded. When enabled, the synchronising
circuit monitors the voltage between the matching phases of the two
busbars.
The closing circuit is armed at voltage
opposition (maximum voltage) and subject to a timer, which sets the maximum
slip at 6 cycles per minute.
Provided this time has elapsed, the synchronising contactor is closed when
the voltage difference falls below
the set minimum (30 - 50 volts). We expected that this rather crude method
would result in a noticeable surge on synchronising. When it operated the
first time we saw nothing, so it
was tried again and again and at different settings. This resulted in little
or no effect until either the minimum voltage was raised or the time delay
shortened significantly.
The stability of the system is much
improved in synchronised operation, and a little more by closing the two
distribution networks into a parallel arrangement when synchronised: We had
one more modification to make after these tests. When one turbine is tripped
or given a stop command when synchronised, that turbine applies full dump
load to brake. Both turbines will feed into the dump load and shed some of
the heating load to compensate. Since both are producing power, the negative
power trip does not operate and the mechanical brake of the "stopped"
turbine has to retard both turbines to the under-frequency cut off. Both
induction coupling and synchronising were interlocked to be inhibited by the
de-energisation of the "run" relay on either turbine. It is difficult to
give a full picture of the system in the scope of this paper, but it is
hoped that what is here will be interesting. The system manual has fifty
pages of text on just the network control panel.
|
