This article is a re-edited version of an excerpt from "Research on the observation accuracy of vertical Doppler lidar in complex terrain (examination of differences due to averaging and correction methods)" presented at the 2021rd Wind Energy Utilization Symposium held in November 11.
1.First of all
In wind surveys, in addition to the traditional wind observation method using a triple-cup anemometer (CUP) and a vane-type anemometer (VANE) on a wind observation tower, the vertical Doppler lidar (VL) method has become common in recent years.1)(hereinafter referred to as the "NK Guidelines") also clearly state that when observations cannot be conducted using wind observation towers alone at a position 2/3 or higher above the planned wind turbine hub height, observations can also be conducted in combination with remote sensing equipment such as VLs (hereinafter referred to as "RSD").
Regarding VL observations in mountainous areas and other complex terrain, it is clear that the accuracy of wind speed measurements decreases when the surrounding terrain is uneven.2)3)4)In an attempt to reduce such VL measurement errors, devices equipped with a correction called FCR (Flow Complexity Recognition) have appeared, but the effectiveness of this correction needs to be thoroughly verified using actual measurement data from CUP and VANE. In addition, there are differences in the averaging method for the 10-minute average wind speed and direction recorded as observation values depending on the VL model, and the results may vary slightly depending on the model and data used.5)This is also something to pay attention to.
Therefore, in this paper, we conducted a verification using measurement data from CUP and VANE with the aim of quantifying the difference in wind condition measurement accuracy using FCR and averaging methods for VL observations in complex terrain.
2. Target area and wind observation
The observation site was located on a ridge in the Kyushu mountain range (see Figure 1), and analysis was conducted using data from a wind observation tower and VL.
Figure 1. Positional relationship between the wind observation tower and the VL
An overview of the wind observation towers is shown in Table 1. In this paper, the 59.6-minute average wind speed and direction values obtained from two CUPs at a height of 2 m and a VANE at a height of 57.0 m were used as true values for analysis.
Table 1. Overview of the wind observation tower
An overview of VL is shown in Table 2. For the verification, Vaisala's Windcube V2.1 was used, and three types of 57-minute average values of wind speed at a height of 60 m and wind direction at a height of 3 m were used as VL observation values, namely, scalar average, vector average, and vector average-FCR, in accordance with the maximum altitude of CUP and VANE on the wind observation tower.
Table 2 Overview of VL
The latest Windcube stores three types of horizontal wind speed data: scalar average, vector average, and hybrid average.7)The scalar average is calculated by taking a simple average of the east-west and north-south components of the instantaneous horizontal wind speed over the samples within the averaging time, while the vector average is calculated by averaging the east-west and north-south components separately. In addition, the hybrid average is a combination of the scalar average and the vector average, but is not included in the analysis in this paper.
The FCR function installed in Windcube corrects wind speeds in complex terrain based on the results of terrain analysis downscaled from a 100m resolution terrain data set using the latitude and longitude information of the observation point. For details, see the Vaisala report.8)Please refer to.
3. Results and Discussion
3.1 Evaluation criteria for observation data
The accuracy evaluation index for the correlation analysis of VL with CUP and VANE as the true value was the stricter standard 1) of the NK guidelines (see Table 3).
Table 3 Correlation between observation data from wind observation towers and RSD as specified in the NK guidelines1)
3.2 Comparison of VL and wind observation tower
Figure 2 shows a scatter plot of wind direction and wind speed with the CUP and VANE observation values on the horizontal axis and the VL observation values on the vertical axis, and Table 4 shows the accuracy verification results of the VL observation values based on the NK guidelines.2The results were 0.972[-] for all, with no significant difference being observed. Although there were slight differences in error in the VL wind direction, all averaging methods met the standards of the NK guidelines in Table 4, and the results can be said to have been measured with high accuracy. The wind speed bias was -4.86[%] for the scalar average, -7.65[%] for the vector average, and -0.85[%] for the vector average-FCR. As such, we can see that the scalar average and vector average measure lower wind speeds than CUP. For wind speed, R2For the slope of the regression line, only the vector average-FCR met the criterion of 0.98 to 0.98, while for the wind direction, all the averaging methods met the criterion.
From the above, it was found that at this site, the vector average-FCR averaging method is corrected to the high wind speed side for the vector average wind speed, and shows a better correlation than other averaging methods.
Figure 2. Scatter plots of wind direction (top) and wind speed (bottom), with VANE and CUP observations on the horizontal axis and VL observations on the vertical axis ((a) scalar average, (b) vector average, (c) vector average-FCR) *The gray line in the figure is the regression line, and the statistics are shown in the upper left.
Table 4. Accuracy verification results of wind speed and direction
3.3 Verification of VL wind speed by direction
When observing wind conditions in complex terrain, the accuracy of the observed wind speed may be reduced due to the terrain located upstream of the wind. Therefore, in order to evaluate the accuracy of wind speed observations by VL for each wind direction, Fig. 3 shows a scatter plot with the VANE wind direction on the horizontal axis and the ratio of VL (scalar average, vector average, vector average-FCR) wind speed to CUP wind speed on the vertical axis. For the scalar average and vector average, the bin average values are lower than 1 for all wind directions. On the other hand, for the vector average-FCR, the wind direction bin average values are lower than 0 near 180° and 1° and higher than 90 near 270° and 1°, confirming a clear wind direction dependency. Further verification is required in the future to determine whether this is a phenomenon unique to this site or due to the averaging method.
Fig. 3. The horizontal axis shows the VANE wind direction, and the vertical axis shows the VL wind speed relative to the CUP wind speed ((a) scalar average, (b) vector average, (c) vector average-FCR).
Scatter plot of ratios
4. Summary
The results of accuracy verification of scalar average, vector average, and vector average-FCR data for VL observations in complex terrain are summarized below.
(1) In accordance with the NK guidelines, the accuracy of VL wind direction and wind speed observation values was evaluated using VANE and CUP observations as validation values. As a result, all averaging methods met the criteria for wind direction, while only the vector averaging-FCR method met the criteria for wind speed.
(2) For VL observations, the scalar and vector averaging methods underestimated wind speeds, and it was determined that observation errors occurred due to the effects of complex terrain. Based on these results, it is considered necessary to apply some kind of correction to the VL observations in order to meet the criteria of the NK guidelines.
(3) It was confirmed that the vector average-FCR method includes a wind direction-dependent correction to the vector average, which results in a correction to the high wind speed side, and at this site, it showed a better correlation than other averaging methods.
(Written by Toshinari Mito)
References
1) Nippon Kaiji Kyokai (ClassNK), 2021: Wind Farm Certification - Onshore Wind Power Edition (KRE-GL-WFC01, Edition:Oct2021)
2) Toshinari Mito, Mizuki Konagaya, Hideki Kato, Teruo Osawa, Takumi Tsuji, Susumu Shimada, 2018: Consideration of the accuracy of wind observation using a vertical-illumination Doppler LIDAR, Proceedings of the 40th Wind Energy Utilization Symposium, pp.191-194.
3) Toshinari Mito, Mizuki Konagaya, and Hideki Kato, 2019: Introduction of a case study on the accuracy of wind observations using a vertical-illumination Doppler LIDAR on flat terrain, Journal of the Japan Wind Energy Society, Vol. 43, No. 2, pp. 193-196
4) Toshinari Mito, Mizuki Konagaya, Hideki Kato, Susumu Shimada, and Hisanori Tanaka, 2019: Achievements and challenges of wind observation using Doppler LIDAR in areas without electricity and complex terrain, Proceedings of the 41st Wind Energy Utilization Symposium, pp.199-202.
5) WIND GUARD, 2018: Observed Reduction of Sensitivities of Windcube Measurements by Vector Averaging, Workshop on Vector Averaging Versus Scalar Averaging
6) IEC 61400-1(ed-4.0): 11.2 Assessment of the topographical complexity of the site and its effect on turbulence
7) Andrew Hastings-Black, Principles of Hybrid Wind Reconstruction, Lidar without limits: Innovative WindCube® enhancements for wind energy, Vaisala Webinars
8) Leosphere、 Windcube FCR measurements、〔https://windweb.leosphere.com/windweb/assets/supportDoc/Windcube%20FCR%20measurements%20-%20detailed%20presentation%20(rev1.2).pdf〕（最終検索日 : 2022年 9 月21日）
9) Annette Westerhellweg, Philippe Beaucage, Nick Robinson, RSD correction for complex terrain effects with the linear wind flow model WindMap, 2021-09-08 Wind Resource Workshop WindEurope