Subscribe Bookmark RSS Feed

profiler for solar power plant simulation


Community Trekker


Jun 23, 2011

I am simulating a solar power plant. I have data on 4500 silicon PV modules that will go into the plant. The data is operating current and operating voltage.
14 modules are clubbed in series (a string) and let's say for simplicity that the lowest current of this group will be the current of this string. The voltage will be the sum of all the voltages

Now these strings are connected in parallel at a combiner box and for simplicity at this point, the operating voltage of the power plant is the lowest of the strings. The current will be sum of the individual string currents.

These numbers will be used to calculate operating power. I am interested in performing a monte carlo simulation in the profiler to see what happens when my distribution of operating current and temperature change.

I have seen simple examples in the profiler and if anyone can send me in the right direction in performing this simulation, I'll be grateful.
Let me see if I understand your situation. A string consists of 14 modules. A string's current is the minimum current of the 14 modules, and the voltage is the sum of the 14 voltages. Since there are 4500 modules, I assume there are approximately 4500/14 = 321 strings. Like I said, I could be totally off, I'm just guessing. The voltage of the plant is the minimum of all the strings, and the current of the plant is the sum of the string currents. To Profile a formula with 4500 components doesn't seem manageable. Let me recommend something else, but it still might be too much.

Let's say I have a formula with 5 component parts. To use the Simulator on that formula, I can give each of the 5 inputs a distribution by assigning it to be Random, and then choosing the distribution I want. That is a common thing to do. Another useful thing is to use the Expression setting for a variable. Instead of making the value of that variable equal to a random draw from a distribution, it makes the value of the variable equal to the results of an expression. So, back to my formula with 5 component parts. If one of the components needs to be equal to the sum of 3 random Normal variables each with mean 100 and stdev 10, I put the following in the Expression box:

temp1=j(3,1,random normal(100,10));

Then, when you use the Profiler, that variable will have the distribution defined by the expression, and not one of the regular distributions you can choose from if you use the Randon setting. You essentially make one variable out of 3. I can see this maybe being useful for you, since the strings are the minimum and sum of 14 component parts. You make 1 variable out of 14. But, then you still have 321 of those. That would be like having a Profiler with 321 components. That is much better than 4500, but still unmanageable.

I did a JMP Blog entry that used the Expression setting in the Simulator. Go to JMP's website, then go the Blog. Look down the right hand side under contributors, and click on Jonathan Gatlin. The title of the blog entry is "Happy Birthday...I'll Take That $1".

This would be significantly easier in JSL. You lose the interactivity of the Profiler (but we can get around that), but you gain so much more. I'm a geek, which means I use JSL for monte carlo simulations for fun. I'm sure there is a solution for you, and would be happy to talk with you about it. If you want, call SAS (919-677-8000) and ask for Jonathan Gatlin. Feel free to leave contact info if I'm not there.