Met Mast Analysis

[thclark]

Yanhua and Tom Meeting 05 June 2019

Agenda for the 1.5 days:

• [x] Read the papers together, to get a deep understanding of what was done

• [x] As we go, note down function stubs that might be useful as steps

• [x] Organise function stubs, write document strings describing what goes in, what comes out

• [x] Write some tests to run the function stubs, ensuring data can be read in Created datastores together to read data, cross-checked between our computers

• [x] Ensure everything runs on Yanhua's PC. Install plotting library and make sure she can create figures that are of the right type.

• [x] Create issue for each function stub and assign responsibility - add to project board to track progress. Yanhua has responsibility for all functions in met_mast.py. Use this issue for any communications.

Papers

There are two key papers, we are attempting to do a very similar analysis:

• "Canadillas" Comparison of turbulence spectra derived from LiDAR and sonic measurements at the offshore platform FINO1

• "Westerhellweg" ONE YEAR OF LIDAR MEASUREMENTS AT FINO1‐PLATFORM: COMPARISON AND VERIFICATION TO MET‐MAST DATA https://www.dewi.de/dewi/fileadmin/pdf/publications/Publikations/S01_3.pdf

Tasks

Create octue app to access the met mast processing datastore from #29 and run met mast analysis to determine outputs as consistent as possible with the LiDAR processing app from #8, with missing items or zero values where appropriate.

[thclark]

Notes from Westerhellweg paper

LAT is Lowest Astronomical Tide USA is UltraSonic Anemometer

Section 3

Figure 3 - need to reproduce similar for ours (see the diagram in the LiDAR Campaign report - do zoomed in view)

Section 4 Mast shadow

May not be a point recreating the Figure 4, which because of the VAD technique having been applied tells us not much. Better is to just do the geometry required to determine which bins of the LIDAR can't be used because they're suffering from mast shadow (see Filters below)

Section 5 Data Filtering

• 10-min ave wind speed use the levenmouth turbine cut-in speed of 3.5m/s.
• 10-min ave wind-speed use the turbine cut-out speed of 25m/s
• Wind direction need to determine which beams of the LiDAR, and which bins of those beams, are affected by mast shadow.
• Wind direction need to determine which side cup anemometer to use base on wind direction.
• Rain Filter - is possible, BUT we only have dailies from the Met Office https://www.metoffice.gov.uk/climate/uk/data/haduk-grid/haduk-grid. BUT, all we can say is "if there was rain within 1km of the turbine on that day, we can't use the data". For now, let's not worry about it, and say we'll use all data. Hmm. To be thought about more. Possibly excluded by intensity quality
• Doppler Intensity. Needs to be related to CNR - Tom to email Wood Group to get section 4.7.3 of the Galion Lidar User Manual, this relates power intensity to SNR. Then apply CNR of -22db ?? See how much data is lost!
• Atmospheric stability. Unstable temperature gradients vs stable vs neutral ones can to be determined from the two temperature sensors on the Met Mast, allowing us to filter on atmospheric stability.

Implement a set of functions to define and apply filters

Section 6 Wind Speed Accuracy

Makes sense to reproduce this analysis -its a good basic check of one against the other. Reproduce Table 3 and Figure 5

Figure 6 - we can't reproduce figure 6, because we use a different scan pattern so can't compare standard deviations. But, calculaitng the pink line (check IEC 61400-121 for how to do so) will give us a usefulerror assessment later. NB save a copy of the standard if you can find it in the library or as a PDF!

Section 7 Turbulence intensity "accuracy"

We can't determine standard deviation of Lidar so cannot produce Figure 7.

To produce figure 8:

• Yanhua can determine TI for met mast (blue trace)
• Tom to determine TI for LiDAR (red trace) analytically (using the es-flow library)
• Counts is the number of 10 minute intervals in the dataset, not filtered out, where the wind speed had that value. Determine this using the cup speed.

Section 8

Reproduce wind direction accuracy results. Unlike this paper, we only have one height.

Section 9

Not relevant to us - don't bother.

[thclark]

Notes from Canadillas paper

Section 1 Field Measurements Campaign

"The limitations of this approach [VAD] include the assumption of horizontal homogeneity of the wind field over the sensed height.". No they don't - this is why we de-smear velocity profiles measured with LiDARs

Note the difference between temporal resolution and sampling rate - 4.6s vs 1.15s because the data overlaps

Section 2 Spectral Analysis

NB the availability >90% filter for hour-long blocks

Welch's algorithm used to estimate PSDs... can use scipy to do that.

"Data segments of one hour were chosen so that larger- scale turbulent structures are captured." Finally someone gets it.

Figure 3 should be straightforwardly reproducible. However, we expect very weird behaviour at the 10-minute line for the lidar, as there should be a run mean filter in effect caused by our 10 minute scan cycle.

The cup anemometer is unable to resolve which turbulent fluctuations are transverse and which are streamwise. (both R11 and R21 and all cross terms are seen). So we have to do some weird magic to the LiDAR results, which are properly resolved into components, in order to mimic the PSD of the cups. This paper shows a way of compensating cups for the correct spectra. Although we may want to convert Lidar to represent cup, rather than vice-versa.

"From Figure 5 it can be shown that there is a good agreement between both radial wind spectra which suggests that the probe-length averaging seems not to have an influence on the shape of the horizontal wind lidar spectrum." Promising. But actually what it means is that the energy is all in scales larger than the probe length. And it's only valid up to 0.1Hz, or 10s, so basically refers only to fluctuations at the nyquist criterion. You can see the beginnings of a departure as expected above that frequency.

Filtering by wind direction. We'll include only times where the wind is coming from offshore; to help converge on the spectrum and be most representative of an offshore flow.

[thclark]

GLOSSARY

Window

A 10 minute block of time, starting every 10 round minutes, eg 24th Jan 2017 2:10:00 pm is the start of a window. Any samples from cup, lidar or other instrument falling between that time and 10 minutes later belong in that window.

High resolution

Data at the original sampling rate of teh instrument, not averaged to 10 minutes