Description

This PR fixes an overflow that led to incorrect results when using the detailed PV Heat Transfer Method when using array dimensions.

The root cause of the error was that the Length and Width parameters of the mcsp_celltemp_t class were being scaled by Nrows, Ncols on each iteration. The intended behavior was to scale them a single time.

A secondary issue was that the result of the temperature model was not checked. The temperature model takes in a temperature variable by reference and modifies it upon return. In cases where the temperature model fails (including the overflow condition above), the function fails and returns false. Because this return is not checked, and because the temperature variable is initialized to ambient temp, the error state is not detected. This results in the behavior described in nrel/SAM #2081.

This PR should solve these issues.

1. It removes all modifications of class variables, and instead modifies local copies. This removes the overflow.
2. It adds additional error checking in cmod_pvsamv1.cpp.
3. It adds additional error messages in the mcsp_celltemp_t operator() function.

One additional change I made is const-qualifying the pvcelltemp_t and pvmodule_t operator() methods, and cosnt-qualifying additional methods that are downstream of these operator() methods and therefore must be const as well. This required no changes in code besides those I already made in mcsp_celltemp_t to remove the overflow. Const-qualifying these methods should prevent errors of this class from arising in the future.

To test, follow @cpaulgilman's steps to reproduce the bug from the initial issue:

Steps to reproduce the behavior:

1. Create a default Detailed PV / No Financial case.
2. On the Module page, choose Heat transfer method for the temperature correction option.
3. Choose Array Dimensions for the heat transfer dimensions option.
4. Run a simulation.
5. On the Data Tables tab of the Results page, display "Subarray 1 Cell temperature (C)" and "Weather file ambient temperature (C)". Note that starting the last daylight time step of January 16, the cell temperature and ambient temperature are the same.

Observe that now the cell temperature does not revert to the ambient temperature on January 16.

Fixes # (issue(s)): nrel/SAM #2081

Corresponding branches and PRs:

This PR is an SSC only PR. There are no specific branches of other repositories that this PR depends upon.

Unit Test Impact:

No tests were changed. If tests exist which assumed the correctness of the detailed PV heat transfer method and used array dimensions, the output of those tests may now change.

Checklist

• requires help revision and I added that label
• adds, removes, modifies, or deletes variables in existing compute modules
• adds a new compute module
• changes defaults
• I've tagged this PR to a milestone