Direct Detection Lidar for Wind Measurement
Wind-sensing lidars are considered a promising technology for the high quality wind measurements required for different applications such as hub height wind resource assessment, power curve measurements and advanced, real-time, forward-looking turbine control. Until recently, the only available lidar technology was based on coherent Doppler shift detection, whose market acceptance has been slow primarily due to its high price. Direct detection lidar technology provides an alternative to remote sensing of wind by incorporating high precision measurement, a robust design and an affordable price tag.
By Nathan Sela PhD, co-founder and VP of R&D, Pentalum Technologies, Israel
The direct detection method is based on sampling the atmosphere with laser pulses whose reflected intensity alone is received and sampled in time. The sampled time-series represents the aerosol density in the line of sight of the transmitted pulse. By attaining aerosol map snapshots in different locations, at different times and tracking the aerosol structures, one is able to extract the wind velocity.
The Method of Cross-Correlation
A description of the principle of operation of wind measurement using the cross-correlation method is presented below. For the purpose of simplicity we will assume a two-dimensional (2D) wind is sampled using two single-beam transmitters/receivers (see Figure 1).
As the emitted light pulse travels upwards through the atmosphere it interacts with the aerosols in the air and is continuously reflected by them, with a portion of this reflected energy being detected by the receiver. Each signal received, for each transmitted pulse, as a function of time is proportional to the aerosol density as a function of the range along the line of sight of the pulse. Thus, by accumulating each received signal and projecting it onto a 2D plot of intensity as a function of range and time, one obtains a 2D map of the aerosol density that has passed through the beam (see Figure 2).
Wind Velocity Calculation Using Image Processing Cross-Correlation
Assuming, without loss of generality, that the average wind is blowing from left to right, and assuming that the aerosol density structure does not substantially change when it passes from one beam to the other, then the aerosol density maps which will be generated by the two transmitters/receivers will look very similar and include a time delay which corresponds to the time it takes the wind to flow from one beam to the other (see Figure 3, which is a typical aerosol density map generated by both transmitters/receivers).
Now, since the distance between the two beams is known, it is sufficient to calculate the time that it takes for the aerosol structures to flow from one beam to the other. The latter is performed by the cross-correlation method, of which a typical implementation is described below.
Following the work by Buttler (2001), this may be performed by defining a subregion of the left map, denoted by ‘Kernel’. The main idea then is to find the feature enclosed in the kernel in the map on the right. This is done by a 2D swiping of the kernel along the map on the right, looking for the point which shows maximum correlation with the Kernel. Once the point of maximum correlation is found, it is treated as if it represents the same feature in the two maps, and thus the time difference between them represents the time that it has taken the wind to pass from the left beam to the right beam. (This process is depicted in Figure 3.)
The process described above outlines, in a simplified manner, the full 3D cross-correlation method of wind measurement using an elastic lidar. Obviously, there are numerous mathematical methods for the implementation of the cross-correlation method, some of which are documented in the literature, but the basic principles are common to all.
As is now clear, the accuracy of the cross-correlation method relies on the assumption that the aerosol structures do not change significantly in the time it takes for the wind to pass from one beam to the other. The above condition is found to be valid in most atmospheric conditions by virtue of the well-known Taylor’s frozen-flow hypothesis, which states that under certain circumstances (not too high turbulence) the turbulence can be regarded as frozen (i.e. the change of eddies is negligible compared with the timescale of the mean wind flow).
The SpiDAR
The SpiDAR (see Figure 4) is a wind measurement pulse elastic lidar specifically designed for wind measurement within the wind energy industry. The SpiDAR is the first commercial lidar for wind measurements that is based on direct detection. The SpiDAR system generates eight conically scanning beams with a full cone angle of 10 degrees. The calculation of the wind velocity (speed and direction) at every height, up to 200 metres, is performed as a generalisation of the cross-correlation mechanism described above.
SpiDAR Field Validation Tests
Figures 5 and 6 summarise two field measurement campaigns performed with the SpiDAR in simple and complex terrain sites. The field experiments were performed in operating wind farms.
Figure 5 depicts a comparison between the SpiDAR measurements of wind speed and direction, as compared with cup anemometer and vane, at a height of 98 metres at a US wind farm (flat terrain).
Figure 6 depicts a summary of a field experiment performed in Spain at a complex terrain site.
Direct Detection Versus Coherent Detection (Doppler)
The main property of direct detection is that it is based only on the intensity of the back-reflected signal, regardless of the exact wavelength or phase of the light source. This renders a simple and robust transmission and reception design, which can be based on common off-the-shelf components such as the laser source, detectors and others. In contrast, coherent detection devices necessitate the use of carefully controlled wavelength and phase laser sources, acousto-optic modulators and unique detectors. Thus, coherent detection devices are substantially more expensive, have a less robust design and consume more power.
Another fundamental difference is related to ground-based, upward-looking wind measurement applications. Doppler devices must scan the atmosphere at a large angle in order to attain a reasonable wind radial component. This results in the need to average over large scales. This is also the reason why Doppler devices have inherent measurement errors when used in complex terrain, where the wind has spatial gradients over a short scale. Such a deficiency is not evident in direct detection.
Table 1 summarises the main differences between direct detection and coherent detection (Doppler) wind sensors.
Table 1. Comparison between direct detection and coherent, Doppler-based detection technologies
Biography of the Author
Nathan Sela is Pentalum’s co-founder and VP of R&D. Nathan is a long-time entrepreneur and executive, with 25 years of technology management and research experience. Past experience includes former founder and CEO of Neurosonix (ultrasound medical devices), VP R&D at X-technologies (medical X-ray), Project Manager at the IMOD and the IDF, leading large-scale multi-disciplinary R&D projects. Nathan is a graduate of the IDF Talpiot program (V) and holds a PhD in Physics from the Tel Aviv University.{/access}
Wind-sensing lidars are considered a promising technology for the high quality wind measurements required for different applications such as hub height wind resource assessment, power curve measurements and advanced, real-time, forward-looking turbine control. Until recently, the only available lidar technology was based on coherent Doppler shift detection, whose market acceptance has been slow primarily due to its high price. Direct detection lidar technology provides an alternative to remote sensing of wind by incorporating high precision measurement, a robust design and an affordable price tag.By Nathan Sela PhD, co-founder and VP of R&D, Pentalum Technologies, Israel
{access view=!registered}Only logged in users can view the full text of the article.{/access}{access view=registered}
The use of elastic lidars to measure wind velocity in the planetary boundary layer is well known. Generally, the methods of measuring wind are divided into two categories:
- (i) coherent detection;
- (ii) direct detection.
The direct detection method is based on sampling the atmosphere with laser pulses whose reflected intensity alone is received and sampled in time. The sampled time-series represents the aerosol density in the line of sight of the transmitted pulse. By attaining aerosol map snapshots in different locations, at different times and tracking the aerosol structures, one is able to extract the wind velocity.
The Method of Cross-Correlation
A description of the principle of operation of wind measurement using the cross-correlation method is presented below. For the purpose of simplicity we will assume a two-dimensional (2D) wind is sampled using two single-beam transmitters/receivers (see Figure 1).
As the emitted light pulse travels upwards through the atmosphere it interacts with the aerosols in the air and is continuously reflected by them, with a portion of this reflected energy being detected by the receiver. Each signal received, for each transmitted pulse, as a function of time is proportional to the aerosol density as a function of the range along the line of sight of the pulse. Thus, by accumulating each received signal and projecting it onto a 2D plot of intensity as a function of range and time, one obtains a 2D map of the aerosol density that has passed through the beam (see Figure 2).
Wind Velocity Calculation Using Image Processing Cross-Correlation
Assuming, without loss of generality, that the average wind is blowing from left to right, and assuming that the aerosol density structure does not substantially change when it passes from one beam to the other, then the aerosol density maps which will be generated by the two transmitters/receivers will look very similar and include a time delay which corresponds to the time it takes the wind to flow from one beam to the other (see Figure 3, which is a typical aerosol density map generated by both transmitters/receivers).
Now, since the distance between the two beams is known, it is sufficient to calculate the time that it takes for the aerosol structures to flow from one beam to the other. The latter is performed by the cross-correlation method, of which a typical implementation is described below.
Following the work by Buttler (2001), this may be performed by defining a subregion of the left map, denoted by ‘Kernel’. The main idea then is to find the feature enclosed in the kernel in the map on the right. This is done by a 2D swiping of the kernel along the map on the right, looking for the point which shows maximum correlation with the Kernel. Once the point of maximum correlation is found, it is treated as if it represents the same feature in the two maps, and thus the time difference between them represents the time that it has taken the wind to pass from the left beam to the right beam. (This process is depicted in Figure 3.)
The process described above outlines, in a simplified manner, the full 3D cross-correlation method of wind measurement using an elastic lidar. Obviously, there are numerous mathematical methods for the implementation of the cross-correlation method, some of which are documented in the literature, but the basic principles are common to all.
As is now clear, the accuracy of the cross-correlation method relies on the assumption that the aerosol structures do not change significantly in the time it takes for the wind to pass from one beam to the other. The above condition is found to be valid in most atmospheric conditions by virtue of the well-known Taylor’s frozen-flow hypothesis, which states that under certain circumstances (not too high turbulence) the turbulence can be regarded as frozen (i.e. the change of eddies is negligible compared with the timescale of the mean wind flow).
The SpiDAR
The SpiDAR (see Figure 4) is a wind measurement pulse elastic lidar specifically designed for wind measurement within the wind energy industry. The SpiDAR is the first commercial lidar for wind measurements that is based on direct detection. The SpiDAR system generates eight conically scanning beams with a full cone angle of 10 degrees. The calculation of the wind velocity (speed and direction) at every height, up to 200 metres, is performed as a generalisation of the cross-correlation mechanism described above.
SpiDAR Field Validation Tests
Figures 5 and 6 summarise two field measurement campaigns performed with the SpiDAR in simple and complex terrain sites. The field experiments were performed in operating wind farms.
Figure 5 depicts a comparison between the SpiDAR measurements of wind speed and direction, as compared with cup anemometer and vane, at a height of 98 metres at a US wind farm (flat terrain).
Figure 6 depicts a summary of a field experiment performed in Spain at a complex terrain site.
Direct Detection Versus Coherent Detection (Doppler)
The main property of direct detection is that it is based only on the intensity of the back-reflected signal, regardless of the exact wavelength or phase of the light source. This renders a simple and robust transmission and reception design, which can be based on common off-the-shelf components such as the laser source, detectors and others. In contrast, coherent detection devices necessitate the use of carefully controlled wavelength and phase laser sources, acousto-optic modulators and unique detectors. Thus, coherent detection devices are substantially more expensive, have a less robust design and consume more power.
Another fundamental difference is related to ground-based, upward-looking wind measurement applications. Doppler devices must scan the atmosphere at a large angle in order to attain a reasonable wind radial component. This results in the need to average over large scales. This is also the reason why Doppler devices have inherent measurement errors when used in complex terrain, where the wind has spatial gradients over a short scale. Such a deficiency is not evident in direct detection.
Table 1 summarises the main differences between direct detection and coherent detection (Doppler) wind sensors.
Table 1. Comparison between direct detection and coherent, Doppler-based detection technologies
| Direct detection | Coherent detection (Doppler) | |
| Hardware/costs | Simple design, low cost, non-coherent components | Complex design, high cost, sensitive temperature controlled, coherent components |
| Power | Low power consumption | High power consumption |
| Geometry | Narrow cone results in almost local wind measurement | Wide cone results in aggressive averaging over large distances |
| Range | Effective up to the convective boundary layer CBL (> 500 m) | Can measure beyond the CBL provided enough power is transmitted |
| Complex terrain | Accurate measurement in complex terrain | Inaccurate measurement in complex terrain |
| Robustness | Reliable telecom telecordia qualified components | Usually complex optics with thermal stabilisation and mechanical shock sensitivity |
Biography of the Author
Nathan Sela is Pentalum’s co-founder and VP of R&D. Nathan is a long-time entrepreneur and executive, with 25 years of technology management and research experience. Past experience includes former founder and CEO of Neurosonix (ultrasound medical devices), VP R&D at X-technologies (medical X-ray), Project Manager at the IMOD and the IDF, leading large-scale multi-disciplinary R&D projects. Nathan is a graduate of the IDF Talpiot program (V) and holds a PhD in Physics from the Tel Aviv University.{/access}
