A compact in-blade five hole pressure probe for local inflow study on a horizontal axis wind turbine

Abstract

Understanding local inflow conditions on a wind turbine blade on an operating wind turbine can further understanding of aerodynamic variations and help predict loads on the turbine. Turbine blades are generally designed with a two-dimensional steady approach, however the real wind conditions are highly three-dimensional (3D) and unsteady. Detailed measurements are not common for validating aerodynamic models. The aim of this theoretical and experimental study is to build, calibrate, install and test a compact in-blade five hole pressure probe system to be used to retrieve these measurements. Wind tunnel calibration of the five hole pressure probe has been successfully completed using an automated traversing system over a ±45° range, with 5° increments. Error analysis showed that the multi-zone pressure coefficient data reduction approach is the most suitable for this application. This approach not only extends the measurable local inflow angles up to ±70°, but also it allows any reference pressure for differential pressure readings. A new data acquisition system (DAQ) internal to a rotating small wind turbine blade section was developed. Space limitations resulted in a custom built DAQ of very contained dimensions. This included, among others, five pressure transducers on a printed circuit board, a 16 bit analog to digital converter, an Arduino microcontroller, and a Bluetooth transceiver to transmit the data wirelessly to the main computer. The new blade section was designed and 3D-printed in such a way that the DAQ instrumentation could be easily accessed and, at the same time, had an acceptable structural solidity. A series of tests were conducted on a 3.4 m diameter wind turbine in a large scale wind tunnel in order to assess the correct functioning of the probe system. As expected, the inflow measurement obtained while the turbine was operating under yawed conditions showed a periodically oscillating inflow vector. The period of this variation was the same as the period of the rotor rotation.

Keywords

Five hole pressure probe wind turbine blade local inflow conditions

Introduction

Wind turbine instruments placed in the external environment face extremes of harsh weather conditions, but must provide reliable information that cannot be obtained by any other means. Typically, there is a significant knowledge gap between the aerodynamic design and analysis of wind turbines and their operation. Wind turbine blades are generally designed with a two-dimensional (2D) steady flow approach with modifications that are either empirical or model based, then the predictions are verified in scaled wind tunnel testing or fundamental numerical approaches. In reality, a large number of operational situations should be simulated. The wind around the turbine blades in field operation conditions is three-dimensional (3D) and unsteady. Sufficient field measurements to create and validate new aerodynamic models to predict the effects of the 3D flow field on the turbine parameters are currently not available.

In field operation conditions a wind turbine operates for the majority of the time under some yaw loads, where a yaw load is any situation when the blade experiences a non-uniform inflow velocity and direction during its rotation. This yawed condition can be the result of various situations, for example in the case of horizontal or vertical wind shear, or when the turbine is not perfectly aligned with the wind direction. Misalignment occurs frequently as it is impossible to continuously follow the changing wind direction. These conditions apply to all wind turbines but are very apparent in small wind turbines that have free yaw in comparison to larger turbines that have a controlled yaw mechanism. These yaw loads may be cyclic and have a frequency that is usually the same as the rotational speed of the blade. The load variations must be accounted for during the design phase, as the loads on the wind turbine structure can be very different from the loads estimated with 2D blade element momentum methods, and therefore adequate safety parameters must be adopted.

Here a compact in-blade system is developed to characterize the local inflow conditions on a wind turbine blade. The development, construction, calibration, installation and testing of the system is described to correlate these inflow conditions with the dynamically changing loads on the turbine’s structure. The instrument shown to be most suitable for this kind of experiment is a multi-hole pressure probe measuring dynamic pressure, allowing determination of both instantaneous velocity and angles of the incident flow.

Multi hole pressure probes (MHPP) use pressure tube concepts to determine flow in more than one dimension. A typical MHPP consists of three or more holes on the measuring tip, allowing the instrument to measure the direction of the flow in addition to its magnitude. Three linear holes determine the velocity vector in two dimensions. More holes, for example five in a cross formation, can measure the three-dimensional velocity vector.

A few cases of local inflow condition determination on medium and large scale wind turbines operating in the field have been described. For example, Maeda and Kawabuchi (2005) instrumented a 10 m diameter wind turbine with a five hole pressure probe to determine local inflow angle and dynamic pressure. Their aim was to investigate how the normal force acting on the turbine structure changes in yawed conditions compared to non-yawed conditions. Madsen et al. (2010) provided insights on aerodynamics and aero-acoustic issues. They instrumented a 3.6 MW wind turbine with five hole pressure probe and used this instrument to measure local inflow conditions. Petersen et al. (2015) described the use of pressure probes to rebuild the undisturbed inflow conditions to a turbine.

However, field measurement does not allow easy correlation of the inflow condition to loads on the turbine structure, due to the continuously changing meteorological conditions. Data obtained in large wind tunnel testing allow better control of the turbine’s working conditions.

The first and most notable attempt to perform this kind of study was the Unsteady Aerodynamic Experiment (UAE) conducted by the National Renewable Energy Lab (NREL) between 1987 and 1996 and is reported by Hand et al. (2001). The data collected on a 10 m diameter wind turbine are a major contribution to the community working on wind turbine aerodynamics, and are often still used as a reference to validate models and predict horizontal axis wind turbine (HAWT) behavior. Wind tunnel testing provided the opportunity to fully control the inflow conditions and therefore to separate the effects of inflow anomalies from the effect of strictly operating in a 3D environment. Differential pressure probes or multi hole pressure probes were found to possess all the necessary characteristics to characterize local inflow conditions: fast response, wide angle range and high sensitivity.

Here, an approach to reduce the complexity of the above approaches and develop a compact on-blade system to provide reliable and accurate measurements is described.

Instruments

Five hole pressure probe

Five hole pressure probes come in different shapes and configurations. For this experiment, a probe was prepared composed of an array of 5 tightly packed stainless steel tubes lying on two perpendicular planes in the shape of a cross, with the intersection between the two planes passing through the center tube as shown in Figure 1. The five hole pressure probe was machined so that the inlet of the four outer tubes forms an angle with the axis of the probe. The five tubes have an exterior diameter of 1 mm and were bonded together with soft solder at about 2 tip diameters downwind of the probe tip. The extremity of each one of the four outer tubes was machined to form an angle of 45° between the tube’s inlet and the axis of the probe. The five tubes were then inserted into a 12.7 mm diameter brass tube through a collar, and five 3 mm stainless steel tubes were connected on the other side of the collar. The brass tube gives structural solidity to the instrument and a support for its installation on both the calibration device and the wind turbine blade.

Figure 1.

Details of the five hole pressure probe tip and angle conventions (image on the left seen from the front).

Calibration

Multi hole pressure probes can be used in two different configurations: either a nulling or non-nulling mode (Morrison et al., 1998). The nulling technique requires a complex traversing system and time to balance the pressures of all the outer tubes, although data analysis is straightforward. It would be impractical for characterizing highly dynamic or turbulent flows, as in the case of this experiment. The non-nulling technique allows the determination of the direction and magnitude of the flow with respect to the coordinates of the probe’s axis, as every particular combination of the five measured pressures corresponds to one pitch, yaw and velocity situation. Since the multi hole pressure probe doesn’t need to be moved during data acquisition, this approach is particularly utilized for applications where limited space is available and when a fast data acquisition rate is required (such as the case of a rotating blade of a wind turbine); therefore this is the approach that will be used here. This method requires the completion of an extensive calibration of the probe, since each combination of pressures must be associated with a flow direction and magnitude during the calibration stage. As mentioned above, the characteristics of the flow can be inferred from the correlations established between the pressures measured at the five holes. It is apparent that the calibration characteristics must include data that account for pressure change on both the pitch and yaw plane, and also the difference between the measured and true local total and static pressure (Treaster and Yocum, 1978). It is commonly accepted that the best way to express these relationships is to make them independent of velocity and a function of the flow angularity only. Dudzinski and Krause (1969), pioneers of this flow measurement technique, were the first to introduce pressure normalization. Without normalization, knowing the freestream total and static pressure was required in order to determine the flow direction, surely impossible in the case of the rotating blade of a wind turbine. They found that the best way to normalize the pressures ( $P_{norm}$ ) is to use the difference between the center hole pressure ( $P_{5}$ ) and the average value of the pressures measured by the four outer holes ( $P_{1}$ to $P_{4}$ )

P_{norm} = P_{5} - \frac{1}{4} (P_{1} + P_{2} + P_{3} + P_{4})

(1)

Three different sets of pressure coefficients (PCs) will be tested and analyzed for wind turbine inflow characterization during this experimentation. The first was used during the NREL unsteady aerodynamics experiment described above and in Fingersh and Robinson (1998). They used as a normalization parameter the free stream dynamic pressure of the flow incident to the probe, defining three dimensionless pressure coefficients.

The center hole pressure coefficient is described by

C_{p_{c}} = \frac{P_{5} - P_{\infty}}{q_{\infty}}

(2)

where $P_{\infty}$ is the free stream static pressure and $q_{\infty}$ is the free stream dynamic pressure

q_{\infty} = \frac{1}{2} ρ_{\infty} V_{\infty}^{2}

(3)

where $V_{\infty}$ is the free stream velocity.

The pitch and yaw pressure coefficients are defined as

C_{p_{p}} = \frac{P_{1} - P_{3}}{q_{\infty}}

(4)

and

C_{p_{y}} = \frac{P_{2} - P_{4}}{q_{\infty}}

(5)

The second set of pressure coefficients analyzed is the one introduced by Dudzinski and Krause (1969). As already explained, this is the first data analysis technique to include pressure normalization. This can be considered the basic set of pressure coefficients since all literature on multi hole pressure probes uses these coefficients or a slight variation of them. A single surface or mono-zone is defined where four different pressure coefficients are determined.

The pitch and yaw pressure coefficients, used to determine the flow direction, are described by

C_{p_{p}} = \frac{P_{1} - P_{3}}{P_{5} - \bar{P}}

(6)

and

C_{p_{y}} = \frac{P_{2} - P_{4}}{P_{5} - \bar{P}}

(7)

where $\bar{P} = \frac{1}{4} (P_{1} + P_{2} + P_{3} + P_{4})$ is the average value of the pressures measured by the four outer holes. The total and static pressure coefficients, used to determine the velocity of the flow are

C_{p_{total}} = \frac{P_{5} - P_{total}}{P_{5} - \bar{P}}

(8)

and

C_{p_{static}} = \frac{\bar{P} - P_{static}}{P_{5} - \bar{P}}

(9)

Lastly, a variation of the Dudzinski and Krause (1969) pressure coefficients has been analyzed. This variation employs the so-called sectoring scheme, a technique that allows extension of the functioning angular range of the multi hole pressure probe.

At high flow angles, separation will occur on the downstream hole (Zilliac, 1993). To overcome errors associated with this sectoring scheme, attempts to reduce these errors were first proposed by Gallington (1980). Based on the hole that is sensing the highest pressure, a combination of holes where the flow is not detached are used to define the pressure coefficients. For example, when the probe is almost parallel to the flow, hole number 5 is measuring the highest pressure and the set of pressure coefficients is the same as the one presented in Dudzinski and Krause (1969). When, for example, hole number 1, an outer hole, is measuring the highest pressure, the reading from hole number 3, opposite to number 1, is ignored in the creation of the pressure coefficient. Paul et al. (2011) have modified this approach to include the central hole pressure in all sectors.

The correlations between pressures at the five holes of the probe cannot be predetermined with an analytical procedure or with numerical simulation. This is mainly due to possible imperfections during the manufacturing stage, therefore, using a five hole pressure probe means performing a long experimental calibration procedure.

Calibration of a five hole pressure probe

Precise calibration is crucial to obtaining usable readings from a multi hole pressure probe. In a few words, the goal of calibration is to establish a correlation between the five measured pressures, the directional pressure coefficients obtained from them, the flow angles and total and static pressure coefficients. The calibration procedure consists of inserting the probe in a known axial flow field, typically an open-jet wind tunnel, and then varying its pitch and yaw over a matrix of angles which exceeds the angularity expected in the flow field to be measured. The five pressures from the probe, as well as the static and total pressure of the flow, are measured and stored at each position of this grid for subsequent data analysis and the creation of the pressure coefficients. The spacing of the matrix of angles to which the probe must be moved needs to be chosen carefully. A lower spacing means better accuracy, but, on the other hand, the increasing number of positions that the probe must be moved to requires much more time spent on calibration.

Effects of Reynolds and Mach number

The effects of Reynolds number ( $R e = ρ U d / μ$ ) on the performance of a multi hole pressure probe have been widely investigated in the literature. Zilliac (1993) explains that the effects due to Re are certainly present but they are small compared to other effects, like errors in readings. Treaster and Yocum (1978) more specifically say that pitch, yaw and total pressure coefficient are basically unaffected by Re, however there is a measurable change in the static pressure coefficient. This means that the measurement of the flow angularity does not present considerable errors (due to the influence of Re), while the measurement of velocity presented a maximum error of 2.8% over the investigated Re range. Since this error is small, one possible solution to reduce this error source is to perform the calibration at the expected Re that will be encountered during experimentation, so the result can be considered accurate if Re remains close enough to the calibration value. Dudzinski and Krause (1969) confirm the goodness of this solution and add that the influence of Re is higher at higher angle of incidence and lower Re. In conclusion, Re certainly has some influence on the performance of the multi hole pressure probe, especially at a high angle of incidence, however this influence can be considered small enough to calibrate the probe at a single Re.

The Mach number (Ma) also has an influence on the response of a multi hole pressure probe, although there is no need to define compressibility coefficients for low speed flow (below 0.3Ma) as is the case in this HAWT study. Turbulence can also greatly affect the response of a multi hole pressure probe. This is particularly important because pressure is proportional to the square of the flow velocity, and highly turbulent flows can have an effect on the calibration data when they are time averaged. Rapid changes in the flow direction due to turbulence can result in a non-linear probe response Telionis et al., 2009).

Data acquisition apparatus

For accuracy reasons the data acquisition system (DAQ) that was developed to perform the instrument calibration is the same one that will be used during the experimental measurement campaign. This avoids problems that could come from using different materials (Fingersh and Robinson, 1998). Designing and adapting the DAQ for the five hole pressure probe in order to work in a small wind turbine blade presents numerous challenges. First, finding a way to fit all the devices necessary for data acquisition into a confined space such as the blade of a small horizontal axis wind turbine means not being able to use laboratory scale or conventional plug and play devices. The second problem comes from the fact that the instrumentation will be located on a rotating support, which means that data transmission between the data acquisition system and the computer that controls the test rig has to be performed wirelessly. This same problem affects the power supply to the DAQ. Lastly the DAQ must be able to correctly work in an environment that is subject to high centrifugal forces and high frequency vibrations. All the above factors, and a limited budget, had to be considered during the design stage.

Pressure transducers

Various models of pressure transducers were comparatively evaluated for the five hole pressure probe application. Keeping in mind the space limitations and required pressure range based on local inflow conditions, differential pressure transducers (Honeywell HSCDRRN005NDAA5) with a range of ±1245 Pa were chosen. Considering that the initial intention was to install the five hole pressure probe at a location of about 1 m from the hub, both the rotor maximum speed of 200 rpm and the maximum wind speed in the wind tunnel of around 13 m/s mean that the estimated maximum local inflow wind speed is 24.7 m/s, corresponding to a dynamic pressure slightly below 400 Pa.

Before the five hole pressure probe was calibrated, the as-received pressure transducers were calibrated. To simplify mounting the pressure transducer into the blade a simple circuit board has been printed in-house and the board mount pressure transducers have been soldered to it. The circuit board also houses a voltage regulator and potentiometer to convert the 9 V DC from a battery to the 5 V DC required by the pressure transducers. The signal from the pressure transducer is then transmitted from the circuit board to a differential analog to digital converter.

Analog to digital converter

The analog signal (voltage) from the pressure transducers was then converted to a digital signal by an analog to digital converter (ADC) chip, the AD7606 (Devices, 2015). The chip has a 16 bit resolution, and can read 8 simultaneous bipolar channels. It uses SPI (serial peripheral interface) to communicate the voltage readings. Simultaneous reading is crucial for multi hole pressure probe applications as the five pressures need to be measured at exactly the same time to get a correct reading from the probe. The ADC chip was pre-mounted on a precision voltage shield (Micro, 2015) that mounted on a microcontroller (Arduino Uno).

Arduino microcontroller

As this is a prototype development, the core of the whole data collection system is the low cost Arduino Uno microcontroller with open source hardware and software (Uno, 2015). The module features a USB for computer connection, 6 analog input pins and 14 digital I/O pins that simplify the task of connecting it to the above mentioned voltage shield. It is very compact, 68 mm long, 53 mm wide with a thickness of less than 1 cm. It is very easy to connect to other devices, and can be connected to a shield using its unique and standardized pin configuration, or to other devices through Serial, Parallel, SPI and I2C protocols.

Data storage and communication

As anticipated, one of the biggest challenges for the application of the multi hole pressure probe (and any instrument in general) to the rotating blade of a wind turbine is the difficulty of performing wired communication between the equipment on the blade and the computer used to collect the data. To overcome this problem, the device was designed to have the capability to internally store the pressure data (as voltages) and to communicate them wirelessly to a ground based computer. For the wireless transmission two possibilities were taken into account: Bluetooth and WiFi communication. After an in-depth evaluation, Bluetooth serial communication was selected as the most suitable technology, primarily because of the ease of use and high data transfer rate. For data storage purposes the data acquisition system was also equipped with a on board microSD flash memory card.

Power supply

Another challenge due to the impossibility of wiring the instrumentation to the stationary world was providing the correct power supply level to each piece of equipment. The choice was made to power the whole system through a 9 V battery, and this voltage level was adjusted in different ways to provide the correct voltage to each part of the instrumentation (such as the pressure transducer printed circuit board, the Arduino Uno microcontroller, the SD card reader, precision voltage shield and the Bluetooth). In order to increase the energy efficiency of the instrumentation, a three position toggle switch allowed selection between two working modes and an off position. In the first mode everything is turned off and in the second mode everything is turned on except the Bluetooth. This choice was made because the Bluetooth transceiver was found to be by far the most energy demanding component in the whole setup.

Analysis of calibration results

In this section the post-processing procedure will be explained, and the results from the calibration will be analyzed in order to determine the best practice for the application of this instrument to characterize the local inflow on a horizontal axis wind turbine blade.

General calibration decisions

As described, the calibration of a five hole pressure probe consists of placing the instrument into a known flow field while moving it through a matrix of known combinations of pitch and yaw angles. The calibration matrix size must exceed the expected maximum flow angle that will be encountered during the experiment. A total range of ±50° for both pitch and yaw angle was chosen in order to considerably exceed the expected flow angle. Steps of 5° were chosen as the resolution for varying the pitch and yaw angle. This is a common value that was encountered a few times in multi hole pressure probe calibration literature (Morrison et al., 1998; Zilliac, 1993). Using this resolution, calibration data have been collected on a matrix of 441 individual locations. As previously mentioned, carefully choosing the speed at which the calibration is performed reduces the small effects due to the influence of Re. At the time at which the calibration was performed, the actual experimental conditions were yet to be defined. However, the expected range of velocity was between 25 m/s and 33 m/s based on experimental parameters such as the location of the probe on the wind turbine blade; therefore a velocity of 29 m/s was chosen for the calibration. The calibration was performed by inserting the probe into the 0.3 m by 0.6 m outlet of a low turbulence open jet wind tunnel. Another issue that had to be defined was the number of individual measurements that had to be taken at every point of the calibration grid and then averaged to create the calibration data. This was very important in order to mitigate the random effects. In order to determine the number of readings to be taken at each position of the grid, a preliminary calibration was performed. A thousand data points have been taken with the probe in two different positions: low inflow angles (yaw 0°, pitch 0°) and high inflow angle (yaw 0°, pitch 30°). Results show that after 400 measurements the standard deviation tends to stabilize, therefore this is the minimum number of measurements obtained at each location.

In order to create the pressure coefficients, the total and static pressure of the flow had to be acquired. For this purpose a pitot static probe was used and the pressure reading was taken using a sixth pressure transducer (Honeywell HSCDRRN005NDAA5).

Traversing device

In order to accurately calibrate the multi hole pressure probe, a traversing device was designed and built to be able to cover the full range of calibration of ±50° in 5° increments, while keeping the tip in the same position during the whole process to calibrate it in constant flow conditions. Automation in the movement of the probe is required for reasons of precision and to reduce the time spent manually traversing the device. Lastly the structure of the device must be strong enough not to vibrate when inserted into the flow of air. The device was entirely designed, machined and assembled at in-house university facilities. The device was designed in such a way that the tip of the probe was positioned exactly in the intersection between the axis of the yaw and pitch rotational movement.

The probe was traversed through the pitch plane with a rotary platform positioned vertically on a stainless steel post. The platform was moved using a stepper motor, with a resolution of 200 steps per revolution, connected to an endless screw that moves a gear attached to the platform about its axis. The motor was controlled by the Arduino microcontroller through a microstep drive (GeckoDrive G201X) that increased the resolution of the motorized system to well below 0.1° (the sensitivity of the digital level used to measure the pitch angle of the probe). The pitching device was connected through a vertical stainless steel post to a rotating platform connected to a rotary table, forming the yawing device that was located manually. A precision protractor on the rotary platform allowed for a resolution of 1°.

Collection of calibration data

Voltage data from the pressure transducers were collected from the developed data acquisition system and stored using a desktop computer, in order to be converted to pressures during post-processing. Data were both transmitted to the computer to be displayed on serial monitor built-in to the Arduino IDE, and stored on the microSD present in the DAQ.

Standard deviation of the data

Typical standard deviation of the calibration pressure data is small, with some variation due to the measurement location. If the calibration grid is divided into two portions, a high angle zone (with one or both of the pitch and yaw angle above 35°) and a low angle zone (with both angles below or equal to 25°), the high angle zone presents a standard deviation of 3.4% while, for the low angle zone, this is 1.9%. This different level of uncertainty in the measurements confirms what has been stated previously, i.e. that at high angle of incidence the flow on the probe begins to separate, resulting in more uncertainty in the pressure readings.

Visual analysis of the data

The calibration data were also analyzed visually by creating a 3D surface representing the voltage readings versus the pitch and yaw angle at which the readings had been taken. Figure 2 shows an example of the transducer voltage measured at hole 5, a surface plot from the pitch versus voltage plane. It shows a smooth surface with no particular spikes, which indicate anomalies, in the data. Similar graphic representations were created for the measurement at each one of the five holes of the pressure probe.

Figure 2.

Voltage measured at hole 5, surface seen from the pitch versus voltage plane.

Pressure coefficient determination

Figure 3 shows a flow diagram that synthesizes the post-processing procedure of the calibration data. Raw calibration data are imported from the .csv file retrieved from the microSD installed in the DAQ during calibration. The complete set of data is then processed, the 400 data points for each location are averaged and the standard deviation is calculated. At this point some partial results are examined. Using a visual representation of the voltage measurements as an aid, like the one shown in Figure 2, anomalies in the data can be spotted, as explained in the previous section. The standard deviation of the population of data points at each calibration location is also checked for values that exceed 2% of the measurement average itself. If the calibration data is good enough to pass this validation, the voltage values are converted to pressures, using the calibration relation obtained for the pressure transducers. Pressures are then used to create the three different sets of pressure coefficients, which are then written on separate .csv files to be used during the experiment. The pressure coefficients are then visualized on surface plots in order to look for anomalies.

Figure 3.

Flowchart representing the procedure used to create the calibration surfaces.

NREL pressure coefficient

The iteration sequence to determine the NREL pressure coefficients shown in Figure 4 is the same one described in Fingersh and Robinson (1998) for the application of a five hole pressure probe on the turbine used for the NREL experiment; the only change is that the initial assumption for dynamic pressure is equal to the highest pressure measured at the five holes, and not the reading from pressure 5. This expedient enhanced the capacity of this methodology to determine high flow angles. The iterative process ends after 100 iterations or when the new $C_{p_{c}}$ differs from the old one by less than 0.001. Convergence of the result with this method on average is obtained after less than ten iterations. More details of this approach can be found in Moscardi (2015).

Figure 4.

Flowchart representing the iterative procedure to determine the flow characteristics using the NREL set of PCs.

Mono zone pressure coefficient

This is the approach to determine velocity and angles of the flow using the mono-zone pressure coefficients. Schematics of this approach are shown in Figure 5.

Figure 5.

Flowchart representing the procedure to determine the flow characteristics using the mono-zone PCs.

Multi zone pressure coefficients

Figure 5 also represents a schematic of the approach used to determine velocity and angles of the flow when the multi-zone pressure coefficients are employed. It differs from the previous approach only in the fact that, based on the hole with the highest measured pressure, a different combination of pressures is used to obtain the coefficients and consequently the characteristics of the flow.

Surface interpolation method

For a precise determination of the characteristics of an unknown flow using a five hole pressure probe, it is fundamental to find the best way to interpolate the calibration data. As explained, calibration was performed over a grid of angles equally spaced by 5° step increments. As the flow during experimental measurement will not necessarily come from one of the angles on the calibration grid, the necessity for a valid interpolation method becomes clear.

Three different interpolation methods have been analyzed, the most conventional of which is based on the interpolation of calibration data with a fourth order polynomial, and two direct interpolation methods, i.e. linear and cubic interpolation. Silva et al. (2003) give a clear view on the different results that the application of a fourth order polynomial or linear interpolation has on the validity of the measurement. They found that the inaccuracy errors for both angles and velocity decrease from 1.5% with a fourth order polynomial to 0.8% with linear interpolation. This same result was also confirmed by Zilliac (1993).

Since the considered five hole pressure probe literature is strongly in favor of direct interpolation methods instead of polynomial interpolation, the choice was now between linear interpolation and cubic interpolation. In order to allow an assessment of the error resulting from the use of one interpolation method or the other, data at probe positions other than the ones from the calibration grid were acquired during the calibration procedure. For those points the characteristics were calculated using the pressure readings from the five hole pressure probe and compared with the data obtained from the calibration using interpolation. Since the flow characteristics were actually known from the beginning, it was possible to estimate the errors caused by different interpolation methods by comparing the calculated value to the known one.

Figure 6 represents the computed errors. In a general way it is possible to see that the two direct interpolation methods give almost equivalent results. Normally one would expect cubic interpolation to be more precise, and the reason for this substantial equivalence can be identified in the fact that the calibration grid has quite high resolution. Another notable fact is that errors on the yaw plane tend to be much higher than errors on the pitch plane (0.28° pitch compared to around 1° yaw). The reason for this trend can be attributed to the higher uncertainty generated by the manual yawing mechanism compared to the automated pitching mechanism. After an accuracy evaluation of the difference in the two interpolation methodologies, cubic interpolation was chosen as it gave slightly better results in the determination of the flow velocity.

Figure 6.

Computed five hole pressure probe errors due to the cubic and linear direct interpolation methods.

Results and discussion

Interpolated calibration surfaces

As previously explained, three different sets of pressure coefficients have been comparatively evaluated during data post-processing. This process has the scope to lead to the identification of the set of pressure coefficients that is most suitable for the application of a five hole pressure probe for local inflow studies on an horizontal axis wind turbine.

NREL PCs

As an approach to understand the goodness of the data collected during calibration, 3D surfaces obtained with pressure coefficient cubic interpolation have been compared to some examples that were found in literature. The plots in Figure 7 represent the surfaces obtained interpolating the pressure coefficients obtained using the same approach that was used during the NREL unsteady aerodynamics experiment (Fingersh and Robinson, 1998). In comparison to Figures 4 and 5 in Fingersh and Robinson (1998) it is clearly visible that the surface shape matches with a good degree of accuracy. The surface representing the yaw angle was omitted for brevity.

Figure 7.

Calibration surfaces obtained with the NREL set of PCs: center hole PC (left); pitch PC (right).

Mono-zone PCs

The second set of pressure coefficients analyzed was first introduced by Dudzinski and Krause (1969), and is considered the more usual way to normalize pressures. The calibration surfaces referring to the static pressure coefficients and the pitch pressure coefficient obtained using this approach are shown in Figure 8. These were not as smooth as the one obtained with the NREL pressure coefficients, and errors were present using this approach. The errors were caused by the fact that the flow on the lee side of the pressure probe begins to separate at high flow angles. As anticipated in the section ’Calibration’, Zilliac (1993) reports that flow incidence angles exceeding approximately 30° indicate separation and that pressures from holes with separated flow will not accurately indicate flow angle. This hypothesis finds additional confirmation when the same interpolated surface with a calibration range with a maximum 30° pitch and yaw angle is evaluated and presented in Figure 9. This restricted calibration range is not affected by flow separation on the lee side of the probe, and therefore the obtained calibration surface is smooth. Accurate results can be obtained using it as long as the flow angle remains in this limit. The particular shape of the static pressure coefficient surface is the same as was found in Treaster and Yocum (1978), a further confirmation of the validity of the calibration data collected.

Figure 8.

Calibration surfaces obtained for ±50° yaw and pitch angle range using the mono-zone PCs: static PC (left); pitch PC (right).

Figure 9.

Calibration surfaces obtained for ±30° yaw and pitch angle range using the mono-zone PCs: static PC (left); pitch PC (right).

Multi-zone PCs

The last set of pressure coefficients analyzed is based on a particular approach, the sectoring scheme (introduced in the section ’Calibration’), which is based on the hole that is sensing the highest pressure at each location during calibration.

Figure 10 is an example of the typical calibration surfaces obtained with the sectoring scheme approach. The top figures show the center zone, when hole 5 is measuring the highest pressure, while the lower figures show the calibration surfaces for zone 3. For brevity zones 1, 2, 4 and total and yaw pressure coefficients have been omitted. Compared to the previous set of pressure coefficients, the sectoring scheme omits some of the pressure readings that are subject to flow separation. Figure 10 shows how, even for angles of incidence up to 45°, the surface remains smooth and no singularity points are shown. Zilliac (1993) claims that this calibration technique gives valid results up to 70° of incidence, after which extensive flow separation causes errors.

Figure 10.

Calibration surfaces obtained using the multi-zone PCs: static PC zone 5 (top left); pitch PC zone 5 (top right); static PC zone 3 (bottom left); pitch PC zone 3 (bottom right).

Experimental error analysis

Figure 11 shows the experimentally obtained errors for each one of the three sets of pressure coefficients. These errors have been determined by placing the probe into a known flow field. For this purpose the same setup used for the calibration was employed, the five hole pressure probe was moved to a number of known angles using the traversing system and the calibration wind tunnel was set to various wind speeds allowing the reproduction of virtually any flow condition that could be encountered during the use of the probe on the wind turbine. The accuracy of the yaw angle measurement is greatly influenced by the low resolution of the manual yawing device (1°). This is by far the dominant source of error when it comes to yaw angle determination. In fact the only difference between the pitch and yaw angle determination procedure is the mechanism employed to traverse the five hole pressure probe. Figure 11 reveals that the set of pressure coefficients that was used for the NREL study gives the least accurate results in comparison to the other two sets employed, both for flow angle determination and velocity determination. The other two methodologies, mono-zone and multi-zone pressure coefficient sets, present almost the same degree of accuracy, with the results of the error analysis slightly in favor of the multi-zone pressure coefficient approach.

Figure 11.

Experimental errors for the three sets of pressure coefficients.

Another factor that favors the choice of the multi-zone set over the mono-zone one is that the latter allows extension of the range of incident angles where this measurement method is applicable by ignoring the pressure measurement from the hole where the flow is detached. This factor is not shown in the result of the error analysis because the angles that have been measured for this analysis are all below ±30°, which according to Zilliac (1993) is the maximum angle where the mono-zone pressure coefficient set is accurate. As suggested by Zilliac (1993), it is also advisable to subdivide the onset flow angle studied in two zones for error determination, one for higher angles and one for angles that are below ±30°. This is because of the different probe response in those two different flow conditions.

Preferred method

Subsequent to the error analysis, it was decided to employ the multi-zone pressure coefficient set for the application of the five hole pressure probe, in order to study the local inflow characteristics on a horizontal axis wind turbine. This approach presented considerable advantages over the two others analyzed.

The experimental error analysis showed that this methodology is consistently more precise than the NREL approach and slightly better than the mono-zone approach.

The multi-zone pressure coefficient set allows extension of the range of the acceptable inflow angles up to ±70° compared to the ±30° of the mono-zone set.

Compared to the NREL pressure coefficient set, the multi-zone set doesn’t need actual flow static pressure as a reference. This is a considerable advantage for the application of the five hole pressure probe on a rotating support, like the blade of a wind turbine, subjected to highly unstable flow conditions

Experimental turbine setup

Wind facility

The University of Waterloo wind generation facility, where all turbine experiments were conducted, is a large open-loop tunnel with 6 identical variable frequency drive fans located near its entrance. Wind speeds between 0–13 m/s could be indirectly controlled by varying the fan frequencies. Each fan generated an air flow of 78.7 m³/s at 413.5 Pa. The air flow was discharged into a 8.54 m long, 8.23 m wide, 5.9 m high plenum, where it was conditioned before entering a 19.5 m wide, 15.4 m long test area that was 7.8 m high at the sides and 13 m high at the peak of a pitched roof. Compared to other wind tunnels this facility presents an intentional relatively high turbulence intensity (between 5.9% to 6.2% (Gertz et al., 2012; Johnson et al., 2012)) and, due to the low flow blockage ratio, this is a good resemblance of realistic flow conditions for wind turbines during actual operations. Detailed wind velocity measurements at the location where the wind turbine is installed, as well as flow characterization, can be found in Gertz et al. (2012); Johnson et al. (2012) and Johnson and Gaunt (2012).

Air temperature and pressure were recorded between every set of experimental measurements.

Wind turbine

The wind turbine utilized in this study was a specially designed and purpose built wind turbine test rig, shown in Figure 12. Tests have been run varying a series of parameters, including rotor rotational speed, turbine yaw angle and wind speed. The rotor speed was controlled through the motor/generator variable frequency drive and continuously monitored through its built-in encoder. Yaw of the wind turbine can also be adjusted in a range of ±30°, in increments of ±5°.

Figure 12.

Photograph of turbine test rig with 3D printed blade attached.

The turbine rotor was designed and optimized based on the operating conditions of the above cited facility and existing wind turbine rig. The rotation speed chosen for the blade design was 200 rpm, for flow blockage reasons the rotor diameter was set to be 3.4 m, while the design conditions for inflow wind speed were set between 6.5 and 8.5 m/s.

Wind turbine blade

Although the standard for HAWTs is to have a three bladed rotor, a single bladed balanced rotor was installed on the test rig. This choice was made to study the aerodynamic behavior of the blade and correlate it to the loads on it. The blade on which the multi hole probe was installed is a custom built 3D printed modular blade, printed using polycarbonate-ABS material in a rapid prototyping machine (Fortus 360mc), with a main tubular spar inserted in the aerodynamic center of the blade. It is described in Abdelrahman and Johnson (2014) and shown in Figure 13. There are five separate identical modules forming the blade, and it was previously used to perform a study on the possible use of trailing edge flaps for active flow control applications. The blade utilized a constant chord airfoil (NREL S833 Somers, (2005)), with a chord of 178 mm and a fixed blade angle of 6°. The modular design of the wind turbine blade made it easier to fit the instruments on the test rig, since only one of the five sections had to be redesigned to install the probe and the data acquisition system in it. Since all the sections present the same aerodynamic profile, it is also possible to move the instrumented section to five different locations on the blade by switching the section order.

Figure 13.

Photograph of turbine blade with probe attached (left side) and photograph of the blade with the cover removed, showing complete DAQ components.

To minimize flow interference the structural probe support and all the instruments used for data acquisition are located inside the blade section. For obvious reasons only the five hole pressure probe can be outside of the airfoil. The tip of the pressure probe, where the five pressures are measured, must be far enough from the airfoil leading edge that the measured characteristics of the flow are not influenced by the presence of the airfoil itself. After a computational fluid dynamics (CFD) analysis of the flow around the airfoil, the tip was placed at one chord upwind of the leading edge, a similar value to that used in the NREL experiment (Hand et al., 2001). The tip of the five hole pressure probe was attached to the airfoil at 15° from the airfoil chord line, as this is the nominal angle of attack at the location the five hole pressure probe was initially installed (0.55 r/R). A fundamental requirement for the new section was structural solidity. This has been achieved with an internal web of thin reinforcements and by providing an instrumentation access lid with three internal supports. The supports were also used to tightly join the access lid to the section main body accessible from the blade pressure side, see Figure 13. Such a complex design could be possible only with 3D printing.

Finding a place for the complete DAQ inside the turbine blade volume was a significant challenge, and size was an important parameter in the choice of every single piece of equipment that composes the DAQ. The existing blade section design was modified in order to gain as much space as possible inside the airfoil boundaries. This resulted in a custom built data acquisition system of contained dimensions: 1.5 cm high, 7 cm wide and around 20 cm long.

Data acquisition system

The data acquisition system used for the experiment can be conceptually divided into two sub-systems. The first is installed in the turbine blade that is directly connected to the five hole pressure probe. Its only functions are to read the analog signal from the pressure transducers, then transform it to a binary digital signal that is converted into a string and sent to the main ground based computer wirelessly through Bluetooth. This is exactly the same setup used during the calibration of the probe described above. The second is conceptually more complicated, but easier to assemble and run, since it is composed of standard computer based components (National Instruments) and controlled using NI LabView. It retrieves every measurement except for the pressure probe, synchronizes the entire system and stores the data. This system will be described briefly and further details can be found in Moscardi (2015).

First the routine initializes the subroutine that reads the encoder signal (built-in in LabView), after the user has set a series of parameters (channels to read, encoder characteristics, etc). Similarly, the digital output signal used to trigger the Arduino measurement is initialized. Until the stop variable becomes true, the code keeps reading the blade’s position. If the blade position is found in an array containing all the blade positions where the pressure must be read, a digital pulse is sent to the Arduino and the blade position and revolution number are sent to the main code.

Subsequently serial communication is used by the Arduino via Bluetooth. The transmission includes the five voltages measured by the pressure transducers plus a time stamp and position. This code is a routine called by the main code that controls most of the wind turbine instrumentation. Bluetooth proved to be an adequate and relatively easy solution for wireless serial communication between the DAQ on the blade and the ground based computer.

Pressure system on the turbine

Pressure transducers used for the experiment measure differential pressure. As such, a reference pressure line had to be developed that was as constant as possible for the five pressure transducers used. This is to minimize acoustic distortion in amplitude and phase due to the compressibility of the medium and viscous flow inside the tubing. Distortion is a function of many factors, including tubing length and diameter. Tubing lengths were kept to a minimum (less than 0.3 m), an advantage of the on-blade system. No corrections for dynamic effects on pressure have been made following Hand et al. (2001). If the pressure varies at high frequencies, attenuation has been observed (Telionis et al., 2009). Installing the reference pressure inlet close to the rotational center of the blade was considered the best way to minimize fluctuation in the reference pressure level, and therefore to minimize its distortion. This is also convenient to minimize the errors due to pressure transducer hysteresis. Not cited above, another considerable advantage of the multi-zone pressure coefficients over the NREL ones (Hand et al., 2001) is that static pressure is not required as the reference pressure for the five hole probe pressure readings retrieved with differential pressure transducers (as long as the scope is to measure velocity and the five pressure readings are referred to the same pressure level, any reference pressure is acceptable). This is a large improvement compared to the setup used in the NREL unsteady aerodynamic experiment (Hand et al., 2001), where a complex tubing system had to be developed to bring the tunnel static pressure level all the way to the rotor to be used as a reference.

The influence of the centrifugal force and vibration on the pressure transducers has been assessed through testing. It was found that, at 200 rpm, the centrifugal force affects the pressure measurement by less than 1/1000 of the measurement itself, while the test rig vibration didn’t show any clear effect on the standard deviation of the population of collected data.

Lastly, the centrifugal force acting on the reference pressure column of air had to be considered. The differential pressures between the probe pressure and the hub reference pressure were reduced by the centrifugal force acting on the column of air in the pressure tubing caused by rotation of the blade (Hand et al., 2001). The correction for this effect can be carried out as follows

P_{cor} = P_{meas} + P_{cent}

(10)

P_{cent} = \frac{1}{2} ρ {(r ω)}^{2}

(11)

where $r$ is the radial location, $ρ$ is the air density and $ω$ is the angular velocity.

The reference pressure is in the same order of magnitude as the pressure measurement itself. It is therefore crucial to apply this correction to the data collected.

Another parameter that was varied during this first phase of data collection was the configuration of the reference pressure line. Four different reference pressure line configurations have been tested with the scope to provide the steadiest possible reference pressure level. Even if, as discussed above, reference pressure theoretically does not influence the determination of inflow vector characteristics, having a steady reference pressure level positively influences pressure transducers hysteresis errors and acoustic distortion due to the tubing system.

Turbine probe measurements

A series of tests have been conducted with the scope of measuring local inflow conditions with the turbine operating under yaw loads. Here the term yaw describes the turbine misalignment with the incoming flow. This is the ultimate reason why the whole instrumentation has been designed and built. Results in yawed load conditions have been compared to non-yawed conditions. As expected, the inflow measurement performed while the turbine was yawed showed a periodically oscillating inflow condition. The period of this oscillation was the same as the period of the rotor rotation, see Figure 14. These oscillating measurements have been compared to the measurements reported in Maeda and Kawabuchi (2005), and similar results have been obtained. These results are only presented briefly as the focus here is on the developed in-blade inflow measurement system.

Figure 14.

Measured parameters during one blade rotation of the turbine at 30° turbine yaw, turbine rotation 200 rpm, probe location r/R=0.55, facility wind speed 11.5 m/s: average variation of inflow velocity (top); average variation of angle of attack (middle); average variation of slip (spanwise) angle (bottom).

Conclusions

The development of an in-blade compact five hole pressure probe for wind turbine blade applications has been described. Calibration of the five hole pressure probe has been successfully carried out in all its aspects, including examining three approaches to pressure coefficient development. In order to successfully perform this task, a two axis traversing system has been designed and built, allowing a precise traverse of the probe on both pitch and yaw planes while keeping the probe tip in the same location throughout the whole calibration.

A large number of samples had to be collected at each calibration position. This number was identified by studying the cumulative standard deviation of 1000 pressure measurement samples, and 400 samples was found to be the limit over which the standard deviation of the population of measurements tends to stabilize.

The search for the most suitable method to interpolate the calibration data between each calibration point led to cubic interpolation as the most precise and reliable method over direct linear interpolation and 4th order polynomials.

Experimental error analysis found that the multi-zone pressure coefficient approach gave the most accurate results over mono-zone pressure coefficients and NREL pressure coefficients. This error has been quantified in 0.3° on the pitch plane, 1° on the yaw plane (higher error due to manual traversing system) and 0.3% in velocity determination.

Application of a five hole pressure probe to a small scale wind turbine required the development of a customized data acquisition system. The main challenges to this task were the limited space on the wind turbine blade and the fact that no wired connection was available between the rotor and the main computer located in the wind tunnel control room. Both these challenges have been successfully overcome. Initial results obtained on the operating turbine indicate a periodically oscillating inflow condition, indicating that the developed compact pressure probe data acquisition system is performing as designed.

Footnotes

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors would like to acknowledge the support of the Natural Sciences and Engineering Research Council of Canada and the Ontario Centres of Excellence.

References

Abdelrahman

Johnson

(2014) Experimental testing of axial load alleviation using trailing edge flaps: Static flap angles case. Journal of Physics: Conference Series 524(1): 012059. DOI:10.1088/1742-6596/524/1/012059.

Devices

(2015) AD7606. Available at: www.analog.com/en/products/analog-to-digital-converters/ad-converters/ad7606.html (accessed 1 October 2015).

Dudzinski

Krause

(1969) Flow direction measurement with fixed-position probes. Report NASA-TM-X-1904, NASA Lewis Research Center, USA, October 1969.

Fingersh

Robinson

(1998) Wind tunnel calibration of 5-hole pressure probes for application to wind turbines. Report NREL/CP-500-22134, National Renewable Energy Laboratory, USA, May 1998.

Gallington

(1980) Measurement of very large flow angles with non-nulling seven-hole probes. Report USAFA-TR-80-17, USAFA, USA, October 1980.

Gertz

Johnson

Swytink-Binnema

(2012) An evaluation testbed for wind turbine blade tip designs - winglet results. Wind Engineering 36(4): 389–410.

Hand

Simms

Fingersh

. (2001) Unsteady aerodynamics experiment phase VI: Wind tunnel test configurations and available data campaigns. Report NREL/TP-500-29955, National Renewable Energy Laboratory, USA, December 2001.

Johnson

Gaunt

(2012) Wind turbine performance in controlled conditions: Experimental results. International Journal of Green Energy. DOI:10.1080/15435075.2012.675799

Johnson

Abdelrahman

Gertz

(2012) Experimental indirect determination of wind turbine performance and blade element theory parameters in controlled conditions. Wind Engineering 36(6): 717–738.

10.

Madsen

Bak

Schmidt

. (2010) The DAN-AERO MW experiments final report. Report Risø-R-1726(EN), Risø DTU, Denmark, September 2010.

11.

Maeda

Kawabuchi

(2005) Surface pressure measurement on a rotating blade of field horizontal axis wind turbine in yawed condition. JSME International Journal Series B 48(1): 156–163.

12.

Micro

(2015) Precision voltage shield. Available at: www.rascalmicro.com/docs-precision-voltage-shield/ (accessed 1 October 2015).

13.

Morrison

Schobeiri

Pappu

(1998) Five hole pressure probe analysis technique. Flow Measurement and Instrumentation 9: 153–158.

14.

Moscardi

(2015) Five hole pressure probe for local inflow study on a horizontal axis wind turbine. Master’s thesis, Politecnico Di Milano, Italy and University of Waterloo, Canada.

15.

Paul

Upadhyay

Jain

(2011) A novel calibration algorithm for five-hole pressure probe. International Journal of Engineering, Science and Technology 3(2): 89–95.

16.

Petersen

Larsen

Madsen

. (2015) Turbulent wind field characterization and re-generation based on pitot tube measurements mounted on a wind turbine. In: 33rd wind energy symposium, AIAA SciTech, (AIAA 2015-1467), Kissimmee, Florida, USA, 5–9 January 2015. Reston, VA: AIAA.

17.

Silva

Pereira

Cruz

(2003) On the use of a linear interpolation method in the measurement procedure of a seven-hole pressure probe. Experimental Thermal and Fluid Science 28: 1–8.

18.

Somers

(2005) The S833, S834 and S835 airfoils. Report NREL/SR-500-36340, NREL, USA, August 2005.

19.

Telionis

Yang

Rediniotis

(2009) Recent developments in multi-hole probe (MHP) technology. In: COBEM 2009 20th international congress of mechanical engineering, Gramado, Brazil, 15–20 November 2009.

20.

Treaster

Yocum

(1978) The calibration and application Of five-hole probes. Report TM 78-10, Pennsylvania State University Institute for Science and Engineering, USA, 18 January 1978.

21.

Uno

Arduino

(2015) Microcontroller board. Available at: www.arduino.cc/en/Main/ArduinoBoardUno (accessed 1 October 2015).

22.

Zilliac

(1993) Modelling, calibration, and error analysis of seven-hole pressure probes. Experiments in Fluids 14: 104–120.