Considerations in Anemometer Calibration Presented by: John Obermeier, President Otech Engineering, Inc. In collaboration with: Rachael Coquilla, Ph.D. candidate Dept. Mechanical & Aeronautical Engineering University of California at Davis When invited to speak at this meeting, I was asked to discuss some items of interest in regards to the use of anemometers in the wind energy industry. 1. Calibrate or not calibrate? 2. Use manufacturer’s recommended transfer function? 3. Use calibration transfer function? 4. Uncertainty applied to anemometer component of measurement system? 5. Contributions to uncertainty Upon reflection on the kinds of questions being asked, it became clear to me that it might be most helpful to re-frame some of the thinking process. I will attempt to address each item in the course of my presentation.

Outline Current Standards and Test Protocols Anemometer Calibration Quality Control Calibration Comparisons Calibration Standard Considerations Summary First is an introduction on CURRENT STANDARDS available and some cup anemometer calibration TEST PROTOCOLS being currently applied. Next, will be a discussion on the ANEMOMETER CALIBRATION process, particularly focusing on how calibration is conducted in Otech Engineering, Inc. I will then describe how we use the anemometer calibration in the QUALITY CONTROL of the manufacture of these anemometers. Then, I will present some CALIBRATION COMPARISONS. Based on these discussion items, I will provide some considerations in developing an anemometer CALIBRATION STANDARD for wind energy applications. And finally a summary of this presentation and recommendations.

Current Standards and Test Protocols starting threshold distance constant transfer function ** off-axis response Current Test Standards for cup and propeller anemometers ASTM D5096-02 ISO 17713 Non-standard Protocols Measnet Otech Engineering, Inc. …moving vehicle method …wind tunnel method There are currently two test standards for defining the methodology and details for characterizing the performance of an anemometer. Both standards describe distinct tests for four different anemometer characteristics. Starting threashold Distance constant Transfer function Off-axis response Most attention in the wind energy industry is focused on one portion of the transfer function test in a speed range applicable to the range of wind turbine energy production. ISO 17713 is about to be released as an update and includes some improvements over the ASTM standard, from my perspective, when applied to the wind industry. There are other test protocols applied to the wind industry that do not specifically conform to existing standards. These are Measnet, widely applied in the European Union, and Otech Engineering, Inc. aka OEI in the United States. One OEI test process has used a moving vehicle with multiple anemometers per test. During 2005, OEI added a conventional wind tunnel laboratory shown in the photo.

Anemometer Calibration Background 30 year history Most widely used instrument in wind industry (over 150,000+ units installed) Most tested anemometer in the world NRG #40 Cup Anemometer Performance characteristics Rugged and survives high wind environments Linear range matches wind turbine energy producing speed range Independent of temperature Maintains performance for extensive periods

Anemometer Calibration Generation of an anemometer transfer function Steady state wind speeds at set intervals Increasing and decreasing speeds Include specified speed range Produce slope and offset from linear transfer function Provide a measure of uncertainty Typical transfer function test is based on: Steady state conditions at preset intervals, usually 10 or 12 speeds Both increasing and decreasing speeds to address hysterics Cover specified speed ranges, typically 4 m/s and up A report giving a transfer function in the form of slope and intercept The right most graph is the residuals – showing the difference between the predicted and measured values at each wind speed. The scale is ±0.2 m/s. Notice that the differences positive or negative may depend on the specific speed range selected.

Anemometer Calibration Uncertainty in Velocity Measurement The next issue is the uncertainty in the wind speed measurement. Presented are the results of uncertainty analysis for both the OEI test processes using the wind tunnel (blue line smoothly decreasing with increasing wind speed) and a sample test from the OEI vehicle test (cyan line). Average values for each process are 1.79% for the wind tunnel and 1.59% for the vehicle test. I repeat for emphasis that this is the uncertainty in the wind speed measurement for each of two test processes. This does not include the uncertainty of the anemometer or of the transfer function. Also shown in the graph is the reported uncertainty from the NIST wind tunnel from the calibration test of the OEI reference pitot tube and pressure transducer system. Again, this is the uncertainty in measurement of the reference wind speed, which averages to 0.71%. NIST is our national standard, and a such, is our recognized final authority in the United States. I note that the current NIST measurement system is based on Laser Doppler Anemometer (LDA) measurements – generally considered about the best air speed measurement technology currently available. I specifically present the NIST uncertainty because I have reviewed a number of anemometer calibration reports claiming uncertainty values even lower than the NIST data. I think this is a case of buyer beware and would suggest the buyer request and review a detailed uncertainty report from the calibration facility.

Recall sample transfer function Quality Control Recall sample transfer function U = 0.7707 f + 0.318 Converted transfer function using forced intercept U = 0.7693 f + 0.35 “slope+k” value One of the initial questions, asked about the use of the supplier’s recommended transfer function for the NRG #40 anemometer. This transfer function comes from a measure of the manufacturing process control. To illustrate this, I am showing a CONTROL CHART -- one of the historically significant developments in manufacturing quality control. A control chart requires a single variable measure of the manufactured item. Blittersdorf suggested using the slope of the calibration test with a fixed offset for this purpose. Based on industry consensus during a 1997 investigation, it was agreed that the constant intercept value is k = 0.35 m/s for the NRG #40 anemometer. The single variable used in a control chart is the transfer function slope calculated with a fixed intercept of 0.35 m/s. This single variable is called the “slope+k” value. As shown in the example, the control chart tracks the difference between a mean value and the individual value for each manufactured item. Acceptable upper and lower limits are established. Using this method, the supplier maintains control of the manufacturing process. The supplier states this to the buyer by providing a transfer function using a slope+k value. The buyer is now assured that the individual anemometer has a tested performance rating that places it within specified limits (± 1%) of a recognizable outside authority like NIST. This is a statement of accuracy and directly addresses the largest contribution to uncertainty.

Quality Control Histogram of Slope+K Values from OEI Vehicle Tests Dec 2002 – Jul 2005 (11,834 new NRG #40 Cup Anemometers) Standard Deviation = 0.0031 m/s per Hz 95 % of the distribution are included within ± 0.8 % of the mean A further statement of variability in the test result can be made using the slope+k value. This summary is based on slightly less than 11,834 calibration tests of new NRG #40 anemometers. It includes all OEI vehicle tests since the introduction of a new vehicle in late 2002 up to July of 2005. The mean value of this distribution is 0.7647 m/s per Hz which differs from the consensus mean value by only 0.02%. This indicates a process that is in control and maintains its relationship to NIST and the Round Robin Experiment. One standard deviation is 0.0031 m/s per Hz. Ninety five percent of the test results fall within a range of ± 0.8 % of the mean value.

Calibration Comparisons Comparison of five NRG #40 anemometer calibrations (year 2002) between the OEI Vehicle and CRES Wind Tunnel test methods To address the question of what transfer function should be used in wind industry applications, I would like to shift your attention to two comparisons of different test facilities. Presented are direct comparisons of transfer function calibration tests performed on the same five NRG #40 anemometers by a Measnet wind tunnel facility and by the OEI vehicle test. Tests were performed during the year 2002. The graph on the left shows the comparison of both the slope and intercept. The red symbol near the center of the graph is the supplier’s recommended slope and intercept. The graph on the right is a control chart presentation based on the slope+k value for each anemometer. The correlation coefficient between both facilities is 0.92. The differences between both facilities varied from near zero to about 1% for the five comparisons.

Calibration Comparisons Comparison of 100 NRG #40 anemometer calibrations (year 2005) between the OEI Vehicle and OEI Wind Tunnel test methods A similar comparison was repeated with a sample of 100 NRG #40 anemometers. In this case, the OEI wind tunnel was used to compare results with the OEI vehicle test. The left graph shows the same type of comparison including both the slope and intercept. The red symbol is the supplier’s recommended slope with fixed intercept. The control chart comparison for this set of 100 shows an overall correlation coefficient of 0.62. Based on these two comparisons, I draw the conclusion that even when different test methods are used there appear to be underlying measurable differences from one anemometer to the next. Based on this observation, I recommend that the best practice in industry is to use the tested slope and intercept values. I present this conclusion with a caution – beware of the test conditions from which you get a reported calibration. Not all comparisons like this indicate a correlation.

Calibration Standard Considerations Must agree upon a consensus standard anemometer calibration procedure and report process (i.e. Measnet). Consensus must remain an open process. Maintain a relation to a standard instrument via round robin comparisons (i.e. RR-3 experiment). The wind energy industry needs a standardized test protocol addressing issues specific to industry needs. Measnet in the EU is a start in this direction. A number of issues need to be addressed including the measurement speed range. The standard should assist transfer of critical performance information across different sensor types. Relation to a recognized standard, like NIST, is critical as is intercomparability by round robin testing of different sensor types.

Calibration Standard Considerations Something to consider…..SPEED RANGE

Calibration Standard Considerations Considerations for advanced calibrations: Over-speeding Definition: a cup anemometer’s over-estimation of the mean wind due to its faster dynamic response at the initial exposure to the wind than at the decrease of the same wind. Off-axis ASTM D5096-02 recommended testing test range ± 30° off-axis increments of 5° angles Temperature dependence

Summary To add value in the manufacture of an anemometer, it is recommended to maintain quality control by using anemometer calibration and a statistical guideline (“Slope+K”). This assures buyers that the product clearly conforms to a defined performance criteria. When a reliable calibration transfer function is available, it is recommended that users apply the tested transfer function rather than a generic transfer function. Uncertainty in reference wind speeds for anemometer calibration are in the range of 2%. Uncertainty claims lower than NIST using LDA technology should be questioned. An anemometer calibration standard needs to be defined.

Related Measurement Standards and References ASTM D5096-02 Standard Test Method for Determining the Performance of a Cup Anemometer and Propeller Anemometer ASTM D4430-00e1 Standard Practice for Determining the Operational Comparability of Meteorological Measurements ASTM D6011-96(2003) Standard Test Method for Determining the Performance of a Sonic Anemometer/Thermometer ASTM D7145-05 Standard Guide for Measurement of Atmospheric Wind and Turbulence Profiles by Acoustic Means ASTM D5741-96(2002)e1 Standard Practice for Characterizing Surface Wind Using a Wind Vane and Rotating Anemometer ASTM D5527-00(2002)e1 Standard Practices for Measuring Surface Wind and Temperature by Acoustic Means “The Maximum Type 40 Anemometer Calibration Project” Thomas J. Lockhart, CCM. Cnet, April 1998 For your reference, I am listing a number of related standards and references.