Gold I-V dataset with benchmarking

.gitignore

@@ -44,11 +44,12 @@ pip-log.txt
 pip-delete-this-directory.txt
 
 # Unit test / coverage reports
-.tox/
-.coverage
 .cache
-nosetests.xml
+.coverage
+.pytest_cache/
+.tox/
 coverage.xml
+nosetests.xml
 
 # Translations
 *.mo
@@ -61,6 +62,7 @@ coverage.xml
 .idea/
 *.sublime-project
 *.sublime-workspace
+.vscode
 
 # Rope
 .ropeproject
@@ -72,9 +74,9 @@ coverage.xml
 # HDF tables
 *.h5
 
-
-#Datafiles
+# Datafiles
 *.csv
+*.json
 
 # vi
 *.swp

ci/requirements-py27.yml

@@ -18,3 +18,4 @@ dependencies:
     - pip:
         - coveralls
         - pytest-timeout
+        - pyinterval

ci/requirements-py34.yml

@@ -18,3 +18,4 @@ dependencies:
     - pip:
         - coveralls
         - pytest-timeout
+        - pyinterval

ci/requirements-py35.yml

@@ -18,3 +18,5 @@ dependencies:
     - pip:
         - coveralls
         - pytest-timeout
+        - pyinterval
+

ci/requirements-py36.yml

@@ -17,3 +17,4 @@ dependencies:
     - nose
     - pip:
         - coveralls
+        - pyinterval

docs/sphinx/source/whatsnew/v0.6.0.rst

@@ -13,7 +13,8 @@ Enhancements
 * Add sea surface albedo in irradiance.py (:issue:`458`)
 * Implement first_solar_spectral_loss in modelchain.py (:issue:`359`)
 * Clarify arguments Egref and dEgdT for calcparams_desoto (:issue:`462`)
-
+* Add gold I-V data set for the single-diode model, along with ``i_from_v()``
+and ``v_from_i()`` accuracy and performance benchmarking
 
 Bug fixes
 ~~~~~~~~~
@@ -35,4 +36,4 @@ Contributors
 * Will Holmgren
 * Yu Cao
 * Cliff Hansen
-
+* Mark Campanelli

pvlib/data/gold/README.md

@@ -0,0 +1,56 @@
+Gold Datasets
+============
+
+This directory contains gold datasets and python modules for gauging various
+algorithms, enabling verification and benchmarking.
+
+## Single-Diode Model (SDM)
+
+### Gold Dataset Creation
+
+The I-V curve gold dataset for the SDM is generated by runnning the following
+python script in this directory
+
+```console
+$ python sdm.py
+```
+
+which will overwrite the existing `sdm.json` and `sdm.csv` files. `sdm.json` is
+the lossless, machine-interoperable format, and `sdm.csv` is formatted to be
+human-readable and to be easily viewed in a spreadsheet program.
+
+NOTES:
+- You will need to `pip install pyinterval` in order to generate gold
+datasets, because interval arithmetic is used to ensure computational accuracy.
+- I-V curves are generated over a range of device types and operating
+conditions.
+- See the `sdm.py` and `test_sdm.py` source code for details/examples.
+
+### Gauging SDM Functions Against a Gold Dataset
+
+Using an existing `sdm.json` gold dataset, you can guage various `i_from_v()`
+and `v_from_i()` implementations for accuracy/performance using a python script
+such as
+
+```python
+import os
+import pprint
+
+from pvlib import pvsystem
+from pvlib.data.gold import sdm
+
+json_path = os.path.join(os.path.dirname(sdm.__file__), "sdm.json")
+gold_dataset = sdm.load_gold_dataset(json_path=json_path)
+pprint.pprint(sdm.gauge_functions(gold_dataset, {
+        "i_from_v": pvsystem.i_from_v,
+        "v_from_i": pvsystem.v_from_i,
+    }))
+```
+
+NOTES:
+- Any functions compatible with the `pvsystem.i_from_v` and `pvsystem.v_from_i`
+APIs can be guaged against the gold dataset.
+- For each function guaged, the worst computed I-V curve for each device is returned along with performance timing information and statistics for
+the ensemble of I-V curve computations for that device (without any
+repetitions).
+- See the `sdm.py` and `test_sdm.py` source code for details/examples.