Sea states and deployment site(s)

[ryancoe]

• What deployment location should we use? The RM3 was originally designed for a site off the coast of Eureka, CA, specifically NDBC46022. This seems like a great place to start!
• How should we select individual sea states (and weightings) to represent the wave climate? This is functionality will likely involve multiple iterations. I would suggest the following stages of development. Each update can supplant the previous more or less seamlessly, so there's no need to hold up other development in this respect.
  1. Create a coarser version of Table 5-1 from the Reference Model project report. @nearyvs - do you have the Excel file for this table handy?
  2. Use the guidelines defined in IEC TS 62600-101 for creating a grid in Hs-Te space. @nearyvs has as good understanding of the Technical Specification (TS).
  3. Use k-means clustering (@mankleh), perhaps determining number of sea states based on the number of bins suggested by IEC TS 62600-101 (@BryonyDuPont). Note that, as pointed out by @gbacelli, strictly speaking the k-means should consider the sea state data as filtered by WEC. This means the process to find an optimal set of sea states via k-means is actually nonlinear (selecting sea states effects the final design and the performance of design effects the selection of sea states). One imperfect workaround (linearization), would be to filter the sea states by impedance of some starting point design. Anyhow, this is way down the road, but I just thought I'd take a note for posterity!
• How should waves be defined? The evaluation block can consider arbitrary power spectral density distributions. These represent an individual sea state (e.g., 1 hr), not an entire resource (e.g., 25 yrs). These can be defined in a number of ways:
  • Empirical data: You can pull data from a wave buoy (e.g,. NDBC or CDIP), process it (generally this is already done for you), and define spectral density (e.g., units of [m^2 s/rad])
  • Idealized spectrum: Bretschneider, JONSWAP, Ochi-Hubble, etc. are parametric spectra that are meant to approximate the real wave spectra we see in the ocean. They are defined by some finite set of parameters, such as significant wave height, Hs, peak period, Tp, peakedness factor, gamma, and spreading, mu.
  • How is this implemented in terms of the tool? Currently (as of 00f2ae63e1bb7d2909df2a8ccc3d50af48e0db9e), RM3_getPow takes a single arbitrary sea state, which is a structure with the following fields:
  • S: Spectral density [m^2 s/rad]
  • w: Frequency [rad/s]
  • ph: Random phasing (0<ph<2*pi) When we're ready, we can add another layer/loop to have RM3_getPow look at multiple sea states and weight the results according to probability of occurrence.

Ok, so what are the next steps:

@Amerlon @mankleh - whenever its helpful for you, you can begin playing with the code as is. The only thing that will change going forward is the size of the variable S being passed into RM3_getPow.

• [x] @ryancoe - add functionality to consider an arbitrary number of sea states with weightings
• [x] @nearyvs - look back and see if you still have the Excel file for Table 5-1 from the Reference Model project report - if not, I think we can get this by OCRing the PDF...

[ryancoe]

added functionality to consider an arbitrary number of sea states with weightings in de1ea1b0b8a0669c78f9ca95d2fa2876743b56e3

[ryancoe]

@Amerlon - I added data and a plotting script for two different assessments of the RM3 deployment site with e16320cf3d7f7fac31abaceb29ad59951ba03394:

1. Original assessment from the Reference Model project report (54 months hindcase)

2. Resource assessment performed by Annie Dallman and Vince (10 years hindcast)

Considering 2, which uses IEC TS 62600-101 to set the bin size, there are 30 sea states with likelihood of occurrence greater than 1%. At least initially, it would be good to reduce this to no more than 5 sea states.

[nearyvs]

OK.

As we discussed, one way to reduce the sea states is to define the cut-in and cut-out sea states. I have seen cut-in and cut-out Hs in reviewing technologies for the wave prize. For Te, the Humboldt Bay site has most of its energy above Tp=10 seconds, which I can translate to Te so we can cutoff any values here as well; But this should show up in the scatter diagram (JPD) with energy distribution anyway.

FYI - here is a EWTEC 2019 paper from Yves Perignon (ECN) that I am going to go through. I will see if they constrained to a reduced set of sea states. Also, Aurelien did optimization for his PhD, but used all the spectra, just like our previous work in the 2017 EWTEC paper.

Vince

---

From: Ryan Coe notifications@github.com Sent: Wednesday, April 24, 2019 4:27 PM To: SNL-WaterPower/WecOptTool Cc: Neary, Vincent; Assign Subject: [EXTERNAL] Re: [SNL-WaterPower/WecOptTool] Sea states and deployment site(s) (#7)

@Amerlonhttps://github.com/Amerlon - I added data and a plotting script for two different assessments of the RM3 deployment site with e16320chttps://github.com/SNL-WaterPower/WecOptTool/commit/e16320cf3d7f7fac31abaceb29ad59951ba03394:

1. Original assessment from the Reference Model project reporthttps://energy.sandia.gov/download/23111/ (54 months hindcase)
2. Resource assessment performed by Annie Dallman and Vincehttps://openei.org/wiki/Characterization_of_U.S._Wave_Energy_Converter_(WEC)_Test_Sites (10 years hindcast)

Considering 2, which uses IEC TS 62600-101 to set the bin size, there are 30 sea states with likelihood of occurrence greater than 1%. At least initially, it would be good to reduce this to no more than 5 sea states.

- You are receiving this because you were assigned. Reply to this email directly, view it on GitHubhttps://github.com/SNL-WaterPower/WecOptTool/issues/7#issuecomment-486265583, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ACCEMHVLD3IHVRV7YIG4IKTPSBU65ANCNFSM4HFS2Y5A.

[gbacelli]

I have doubts about using cut-in and cut-out. WECs have more flexibility in these terms than wind turbines. For the moment we can consider the cut-in and cut-out, but we are working with developers to about them, so they can increase capacity factor. Basically it's all about the ability to start throttling (regulating) the power flow through control detuning and PTO design.

[ryancoe]

@gbacelli - agreed. let's do a first implementation with binary cut-in/cut-outs and then progress to some sort of modulation.

[ryancoe]

Here's the Yves Perignon paper that Vince mentioned - seems very relevant:

20190328_EWTEC-2019_article1643_v2a.pdf

[ryancoe]

@Amerlon - Can you give a short description of your plans to incorporate the k-means averaging code you've written into the RM3 study?

[ryancoe]

@Amerlon - Just to reiterate from our conversation today, the workflow should go something like the following:

1. Call Sorting_and_Clustering_v2.m to create .csv file with columns for significant wave height (Hs), peak period (Tp), and weighting factor (mu). This file should be created once for our study and then remain static.
2. Write a function to read the .csv file from 1., and create an array of sea states using the bretschneider function that comes with WAFO. The effect should be that you can do something like the following, with the definitions for the elements of S coming from your .csv file.

clc
clear

S(1) = bretschneider([],[8,4],0);
S(2) = bretschneider([],[4,4],0);

S(1).mu = 0.3;
S(2).mu = 0.7;

pow = RM3_getPow(S,'CC','scalar',1);
disp(pow)

3. The function from 2. will be called in @mankleh's optimization script.

[BryonyDuPont]

@Amerlon Can you read back through this thread and touch base with me on implementing multiple sea states?

[ryancoe]

This work is superseded by #40.