Efficiency of Asynchronous Generators for Variable Speed Wind Turbines Kali McLaughlin, Flowtrack Pty Ltd P.O. Box 5 Nimbin 2480 N.S.W. Greg Clitheroe, Biomass Energy Services & Technology 1 Davids Close, Somersby Industrial Estate, 2250, N.S.W. Abstract A modern high efficiency induction motor was tested as an induction generator over a broad range of operating points far from the nameplate rating. Speed, voltage and power have been all varied independently and the resulting data has been presented as a convenient design tool for optimally matching a variable speed prime mover. 1 INTRODUCTION In order to have fixed blade pitch in a wind turbine it is desirable to have variable speed operation of the generator so as to maintain aerodynamic efficiency over a range of wind speeds. Our wind turbine design uses a standard induction motor as a generator. It was necessary to determine the best voltage at which to operate the machine as a function of varying speed and power. Previous design was based on extrapolation from motor data and loss analysis (ref 1). Recently the market for energy conservation has brought about a new product range of induction motors with substantially higher efficiency. Induction motors do not have a reputation for high efficiency, but that is because they have been optimised for price and operating parameters other than efficiency. Also the familiar single phase version of the squirrel cage machine is a considerable design compromise. A 4 kW motor was investigated which had specified efficiency above 90% for much of its load range and the question was whether this would be maintained when the machine was generating, and whether it would be maintained at very different conditions of speed and power. We took a double approach to the problem. Firstly the machine was measured on a test bed in order to establish its machine parameters in order to extrapolate from manufacturers data to predict generating performance. Then as a model check we took the fundamentalist approach of running the machine on a dynamometer so as to measure the input mechanical power and thus calculate the efficiency from the output power. 2 METHODOLOGY The simple calculation of generator efficiency by dividing electrical power out by mechanical power is inherently inaccurate because the energy lost in conversion is a small fraction of the total, and even small errors in the many variables measured can lead to wide confidence intervals in the difference between power in and power out. We felt however that an independent check of this sort was justified, and should give us more information about extreme operating points. The method of measuring power transferred to a load versus the method of analysing the losses in the generator was thought to be simpler for comparing one generator to another, and safer from errors in modelling and extrapolation. Also, our interest was not in finding the precise peak of efficiency. In a very variable application like wind power the system would rarely operate at the optimal point. We were more interested in finding the areas where efficiency fell sharply, and designing matching relations which avoided these. Efficiency of Asynchromous Generators for Variable Speed Wind Turbines 3 K. McLaughlin EXPERIMENTAL SETUP The mechanical drive was provided by a car with automatic transmission. The properties of this rig approximated a turbine and was free from control loop problems, being very heavily damped. The throttle of the car was set by hand to achieve various operating points. The generator was belt driven in a cradle with torque electronically sensed. The balance and friction of the cradle were both designed carefully to avoid offsets and nonlinearities over the wide operating range of torque. Excitation of the generator was by capacitors. This meant there were no high frequency currents in the machine, and the efficiency of a motor drive inverter did not have to be taken into account. Again we were making the test as realistic as possible. The load we used had to be active to stabilise some of the operating points, and injected some harmonic currents into the machine, but this also was considered to be a recreation of the practical application. Our initial load was a pseudo resistive load created by a 3 kW 20 kHz switchmode regulator. After destroying this several times we had to move to one which clamped the three phase rectified voltage. The problem we were having was that the sudden voltage surge at the moment of generator excitation was exceeding the 600 volt rating of the switchmode before its response time acted to protect it. Speed was measured by optical pickup on the generator shaft, electrical frequency voltage and current with good multimeters. Power factor was calculated from the capacitors connected to the machine at that operating point, and slip was calculated by recording a section of the output waveform which had synchronising pulses injected into it by the optical shaft pickup. (fig #1). This proved necessary because considerable jitter of rotor phase was noticed, probably originating from drive train irregularities. The generator cradle was constructed of wood, and much of the instrumentation was left at floating voltage to minimise interference and arguments about earth between the array of different test points. This proved a very bad idea as the belt drive to the generator transferred charge to the extent of several thousand volts which arced across the isolation of what used to be a very nice Tectronix oscilloscope. 4 RESULTS 4.1 Excitation The initial question about this new high efficiency generator was its self-excitation properties. The answer was interesting. Residual rotor flux was observed below 10 Gauss, but this did not prevent reliable self excitation. Notable also was the very slow rate of excitation. The combination of the low residual field and slow build up was that the machine could take ten seconds to come to life. This is a problem in self excited wind power systems as the aceleration a turbine can be faster than this, causing runaway. The low residual flux was expected because of the high quality of the lamination material, but the slow excitation rate seems to be a result of the dual cage rotor construction that this type of motor employs to keep slip, and thus rotor losses, low. A 50 kW machine of the same brand was found to have very similar properties, but was even slower in its excitation response time. This can be calculated as a function of the electrical response time of the deep rotor cage. 4.2 Operating Points The form of the data is that the variable of interest, generator efficiency, depends on three independent variables, usually: voltage, current and frequency. Instead of using these variables we used the three variables: voltage, mechanical power input and speed. We did this because we were interested in matching the generator to existing turbine data, stated of course in speed and mechanical power. These three variables were enough to specify any achievable operating point of the machine. Our data structure was able to yeild any convenient set of variables. Presentation of the data was as slices of constant power but varying voltage and speed. This allowed optimisation of the wind turbine at a particular windspeed. A series of slices at increasing power allowed an optimal turbine power point track to be created. 2 Proceedings for Solar ’97 – Australian and New Zealand Solar Energy Society Paper151 Efficiency of Asynchromous Generators for Variable Speed Wind Turbines K. McLaughlin We had intended to take data only on slices of constant power, thus reducing the three dimensional data to a series of sets of two dimensional data. In practice this proved impossible because the continuously variable load was not sufficient to traverse the operating space, and we found the need for continuously variable capacitors, which of course are not possible in the sizes necessary. Furthermore the car throttle was difficult to set accurately enough to fine tune the power level, and the automatic transmission had a tendency to kick down a gear at inconvenient times. The designers of the gearbox had allowed for people to select a low ratio, but obviously had not considered that anyone would want to hold the car in high range! Thus the data came out as 150 points in a three dimensional space. We got the computer to first project the data onto convenient slices of power, and then to draw contour lines of efficiency as presented . There is obviously a lot of noise in the data, but the trend is clear, the areas to avoid are quite well delineated, and the areas of high efficiency are gratifyingly broad. 4.3 Slip Our plan to measure slip was to measure the difference between the optical shaft sensor and the electrical frequency generated. This method worked for only part of the large range of slip we covered, from less than 1% up to 30% ! The slip data is not presented, as it is not germane to the matching problem, but it is worth noting that we had to take recordings of the generators waveform with the shaft sensor output injected into it. Slip was calculated for every operating point by counting the electrical cycles between phase concurrences. With enough data this was quite an accurate, if slow, method. Typical recordings were as reproduced below. This buried cage motor showed very low slip when running at good efficiency. This strong coupling to the stator field in the steady state was however at odds with transient response. The rotor was very compliant to high frequency variation in rotor position, and this is possibly a contributor to the phase jitter we encountered in the shaft position sensor, in spite of efforts we took with a hydraulic drive and multiple large radius drive pulleys. fig#1 Generator output 4.4 POWER FACTOR As with efficiency and slip, power factor was calculated for each operating point, using the load current (resistive) and the capacitor current (90 degree leading). The voluminous data is not presented here, but it indicated the expected fall-off in the areas of poor efficiency. The only surprise was the poor power factor in areas of relatively high rotor current, such as the full load operating point on the nameplate. This was consistent with the high rotor reactance parameter as measured by Dr Peter Freere. The buried cage construction of the rotor could be expected to cause this, but examination of the data for the full range of motors in the range indicates that this feature might be an artifact of frame size quantisation. It is common practice to supply several different ratings of motor with the one frame size to minimise the stock of frame sizes needed. This variation of machine design on the one frame will cause a change in the proportions of the machine parameters from one motor rating to the next. This falling power factor at full load caused great problems in the use of this particular generator with our power point tracking transformer as it forced a higher than desired rate of speed change with power. For this reason we ended up using a different type of generator! Proceedings for Solar ’97 – Australian and New Zealand Solar Energy Society Paper151 3 Efficiency of Asynchromous Generators for Variable Speed Wind Turbines 4.5 K. McLaughlin EFFICIENCY The following graphs were computer generated from the 150 data points we recorded. In areas where scarcity of data points make the interpolations meaningless the contours have been removed. SPEED (Revolutions/sec) 0 50 100 rev/sec 150 200 Generator voltage, phase/phase EFFICIENCY at 300 WATTS INPUT Data Location 0 4 Data Location 100 200 300 volts ph/ph EFFICIENCY at 600 WATTS INPUT Proceedings for Solar ’97 – Australian and New Zealand Solar Energy Society Paper151 Efficiency of Asynchromous Generators for Variable Speed Wind Turbines K. McLaughlin rev/sec 0 data location 100 200 300 EFFICIENCY at 1 kW INPUT Revolutions / second 100 150 Volts, phase/ phase Data Location 200 250 300 volts (phase / phase) EFFICIENCY at 1.5 kW INPUT Proceedings for Solar ’97 – Australian and New Zealand Solar Energy Society Paper151 5 Efficiency of Asynchromous Generators for Variable Speed Wind Turbines K. McLaughlin revolutions/sec 100 200 Data Location 300 EFFICIENCY at revolutions / sec 100 5 400 500 volts, phase / phase 2 kW INPUT Data Location 200 300 400 500 volts, phase / phase EFFICIENCY at 3 kW INPUT INTERPRETATION At each power level there is a plateau of speed and voltage which yields maximum efficiency. The graphs indicate that this is gratifyingly broad, making the job of power point tracking relatively easy. The trajectory of maximum efficiency is seen to run from 75 volts and 12 rev/second in the 300 watt graph to 400 volts and 35 rev/sec in the three kilowatt graph. The trajectory seems approximately linear, as suggested by our earlier theoretical analyses. Below 300 watts the data became unreliable as relative errors rose, and the effects of friction dominated efficiency. The analysis was not continued to higher speed and voltage because these areas were of little practical importance, though it would have been interesting to see what happened! 6 Proceedings for Solar ’97 – Australian and New Zealand Solar Energy Society Paper151 Efficiency of Asynchromous Generators for Variable Speed Wind Turbines 6 K. McLaughlin REFERENCES Kali McLaughlin 1992 A Wind Turbine with Power Point Tracking Induction Generator ANZSES Proceedings, SOLAR 92 7 ACKNOWLEDGMENTS This measurement of generator efficiency was done as part of a wind turbine commercialisation project conducted by Biomass Energy Services and Technology (BEST) after an investment by Energy Research and Development Corporation (ERDC). Work was done by Monash university Dept of Electrical Engineering, both in measuring machine parameters at a range of frequencies, and in validating our experimental apparatus. A report was produced showing the unusual equivalent circuit of the high efficiency motor tested. While not directly relevant to calculation of generator efficiency, data supplied by Newcastle University Engineering Department on turbine characteristics gave us the curve which the generator had to optimally track.. 8 CONCLUSION It has been empirically verified that Induction motors work efficiently as generators over a considerable range of voltage, speed and power. The trend of this three dimensional region has been presented in a series of graphs which are useful as a design tool for power point tracking applications. Proceedings for Solar ’97 – Australian and New Zealand Solar Energy Society Paper151 7