Ray tracing vs. experimental data

[jlefevre76]

So, here's the deal: I'm trying to complete the publishing of a paper dealing with my thesis from almost two years ago, and I finally wrote a program that I believe will tell me how different geometries will influence the outcome.

Here's the setup, I am modeling the radiation that goes through the cover of a solar still (google WaterCone for an example of what a solar still is) and is absorbed by the basin. In order to verify that my ray tracing code is functioning properly, I have performed three experiments with small, polycarbonate "covers" placed over a solar cell. The solar cell is given a small load and the voltage over the load is measured over the course of the day. Assuming the voltage is proportional to the total radiation flux seen by the cell (which may not be the case, I'm not certain), the radiation distribution over the course of the day does not match what I would expect it should according to my ray tracing analysis.

Attached is one of the data files (data file 1, with the data used to approximate the flux given the maximum solar radiation that day). The behavior as you may note, drops down a bit and then comes up in a small hump at solar noon. While, for the ray tracing done to approximate the same conditions, there is no center hump, it is one single curve without the jump at midday. Similar midday behaviors can be seen in the other covers (all are wedge shaped), as the back plate reflects some light that would otherwise be lost onto the solar cell.

Data files 2 and 3 are also included.

Please note, there remains some fine tuning to be done to get them to match up a little better, but the general behavior of the ray tracing model will not change even after the fine tuning.

Any help anyone can give me would be greatly appreciated (even if that help is "go ask somebody else who might know"). I can upload ray tracing codes if anybody here is equipped to understand them and troubleshoot them, but I'm guessing nobody here would have any experience with them, so redirection to an appropriate site/forum would be acceptable as well.

 

[256bits]

The instrument used to measure radiation flux is called a pyranometer.
http://en.wikipedia.org/wiki/Pyranometer

 

[jlefevre76]

Right, but a solar cell with a low enough load can pass itself off as a pyranometer, except for it will still have directional absorptivity (whereas I assume a pyranometer somehow compensates for that). In the past I've had exceptionally good results using solar cells to compare fluxes (and that was with open circuit voltage, which is actually less accurate than what I was doing in this case by adding a small load to it).

In any case, that's why I took a control case for all the measurements done (one cell measured without a cover). Had the fluxes on each cell been the same, one would have expected them to have the same voltage over the load.

 

[256bits]

The solar cell can receive radiation in 3 ways - direct from the sun, indirect from the sky, and reflected from another surface. The curve should actually be symmetrical, with solar noon being the greatest, but smooth with no bump. Have you ruled out reflected light from anything in the vicinity for the bump.
The reflectivity of the solar cell surface could be also something to check as the angle of the sun changes.
Any chance of a lens effect happening. Or the solar cell being frequency selective.

Your data files have not shown, by the way.

 

[jlefevre76]

The solar cell can receive radiation in 3 ways - direct from the sun, indirect from the sky, and reflected from another surface.

Check, the ray tracing model accounts for both diffuse (sky) and direct (from the direction of the sun) solar radiation. The other surfaces part comes in where the surfaces of the cover can reflect light that would otherwise miss the cell back onto the cell, increasing the net radiation to the cell. Attached is an image showing that setup.

The curve should actually be symmetrical, with solar noon being the greatest, but smooth with no bump.

Correct, but the second half of the day some clouds moved in, causing it to drop a little bit after solar noon. Generally, there is some noise that is unavoidable due possibly different sources, like EM interference in the wires, clouds, shadows of animals, wind bouncing the wires around causing temporary disconnects, etc. If the "bump" was caused by something like that, we would expect to see something similar in the control case that was running simultaneously (the uncovered data).

Have you ruled out reflected light from anything in the vicinity for the bump.

Again, if there were something major affecting it, we would expect to see something happen to the uncovered case as well, at least I would think. They were literally only a few inches apart during the testing, so reflection off a house window or anything like that would not likely have affected only one of them.

The reflectivity of the solar cell surface could be also something to check as the angle of the sun changes.

Not sure about that. Since the sun is at different angles over the day, the reflectivity of the solar cell does change. The ray tracing model takes that into account (and I have tried treating it like a diffuse surface for comparison, it doesn't seem to affect the general behavior of the output).

Any chance of a lens effect happening. Or the solar cell being frequency selective.

That's a possibility. I was actually kind of hoping for the cover to increase the total radiation (is that what you mean by "lens effect?") to the basin the cell. Also, most solar cells, I think, are at least somewhat wavelength selective. How that could have given the behavior seen in the data, I have no idea, it doesn't make sense that it would have changed anything.

Your data files have not shown, by the way.

Oops, that certainly doesn't help. I'll look into getting them up. Otherwise, I don't think anybody would have any idea of what I'm referring to.

 

[boneh3ad]

I would assume that the polycarbonate shell does not perfectly transmit the light and instead reflects and/or absorbs some of the incident rays. That would account for the covered case being lower than the uncovered case. Might it be that as the sun passes directly overhead and has a less oblique angle, the reflectivity and/or absorptivity of the polycarbonate shell decreases, bringing the results more into line with the uncovered case? Actually, it looks like the two cases match fairly well except on the "shoulders" of the graph, so maybe there is a reflectivity/absorptivity issue at those specific angles and your ray-tracing model simply doesn't account for this.

 

[jlefevre76]

Actually, it looks like the two cases match fairly well except on the "shoulders" of the graph, so maybe there is a reflectivity/absorptivity issue at those specific angles and your ray-tracing model simply doesn't account for this.

That's kind of what I was thinking, but for the life of me I can't find the error in the code. I've even experimentally validated the directional reflectivity of the cover material (and I was surprised how well they fit the properties of a dielectric with n=1.49).

Yeah, so that's what I was thinking, something wrong in the ray tracing code. I'll have to track it down and figure it out. Anybody know of any good resources when it comes to freeware codes similar to what I would need?