Adding an outlier check for dc capacity/power

docs/examples/outliers/pvwatts-outlier-detection.py

@@ -0,0 +1,98 @@
+"""
+PVWatts Outlier Detection
+=========================
+
+Identifying outliers in time series using
+PVWatts outlier detection.
+"""
+
+# %%
+# Identifying and removing outliers from PV sensor time series
+# data allows for more accurate data analysis.
+# In this example, we demonstrate how to use
+# :py:func:`pvanalytics.quality.outliers.pvwatts_vs_actual_abs_percent_diff`
+# along with
+# :py:func:`pvanalytics.quality.outliers.flag_irregular_power_days`
+# to identify abnormal daily behavior in a time series by measuring
+# a data stream's daily performance against a PVWatts model.
+
+import pvanalytics
+import pvanalytics.quality.outliers as outliers
+import matplotlib.pyplot as plt
+import pandas as pd
+import pathlib
+
+# %%
+# First, we import an AC power data stream from the SERF East site located at
+# NREL. This data set is publicly available via the PVDAQ database in the DOE
+# Open Energy Data Initiative (OEDI)
+# (https://data.openei.org/submissions/4568), under system ID 50.
+# This data is timezone-localized.
+
+pvanalytics_dir = pathlib.Path(pvanalytics.__file__).parent
+ac_power_file_1 = pvanalytics_dir / 'data' / 'serf_east_15min_ac_power.csv'
+ac_power_data = pd.read_csv(ac_power_file_1, index_col=0, parse_dates=True)
+ac_power_data.index = pd.to_datetime(ac_power_data.index).tz_convert("UTC")
+ac_power_series = ac_power_data["ac_power"]
+print(ac_power_series.head(10))
+
+# %%
+# Second, we import weather data from NSRDB PSM3 api.
+# We specifically pulled ambient temperature, dhi, dni, and wind speed values
+# at the SERF East site and for the entire duration of the AC power
+# measurement. By default, the pulled weather data is in UTC
+# tz-aware datetime. Users can sign up for an api key at NREL Developer
+# Network https://developer.nrel.gov/signup/ and use pvlib.iotools.get_psm3
+# to query the api. See pvlib documentaton for the full
+# list of fields in NSRDB
+# https://pvlib-python.readthedocs.io/en/v0.9.0/generated/pvlib.iotools.get_psm3.html
+pvanalytics_dir = pathlib.Path(pvanalytics.__file__).parent
+weather_file_1 = pvanalytics_dir / 'data' / 'serf_east_nsrdb_weather_data.csv'
+weather_data = pd.read_csv(weather_file_1, index_col=0, parse_dates=True)
+print(weather_data.head(10))
+
+# %%
+# We then get the metadata for SERF East site and
+# use:py:func:`pvanalytics.quality.outliers.pvwatts_vs_actual_abs_percent_diff`
+# to get the absolute percent difference between the actual and predicted
+# time series. The predicted AC power time series
+# is modeled from PVWatts with NSRDB data and site metadata.
+
+lat = 39.742
+long = -105.1727
+azimuth = 158
+tilt = 45
+tracking = False
+capacity = ac_power_series.max()
+# Get absolute percent difference
+abs_percent_diff_series = outliers.pvwatts_vs_actual_abs_percent_diff(
+    power_time_series=ac_power_series,
+    lat=lat,
+    long=long,
+    tilt=tilt,
+    azimuth=azimuth,
+    tracking=tracking,
+    nsrdb_weather_df=weather_data,
+    dc_capacity=capacity)
+print(abs_percent_diff_series.tail(10))
+
+# %%
+# We then flag the days where the absolute percent difference exceeds
+# the set precent threshold of 50%. We can use
+# :py:func:`pvanalytics.quality.outliers.flag_irregular_power_days`
+# for this flagging the irregular days.
+# The actual and predicted daily time series with detected anomalous days
+# are then plotted.
+
+irregular_day_series = outliers.flag_irregular_power_days(
+    abs_percent_diff_series, pct_threshold=50)
+print(irregular_day_series.tail(10))
+
+# Plot results
+abs_percent_diff_series.plot()
+abs_percent_diff_series.loc[irregular_day_series].plot(ls='', marker='o')
+plt.legend(labels=["Absolute Percent Difference", "Detected Outlier"])
+plt.xlabel("Date")
+plt.ylabel("Absolute Percent Difference")
+plt.tight_layout()
+plt.show()