Issue/52 sun azimuth elevation

notebooks/2021-09/2021-09-28/get_raw_sun_data.py

@@ -0,0 +1,120 @@
+############
+# Look into the differences from year to year in elevation and azimuthal direction
+
+# Looked from 2018-2020, for January, April, July and October,
+# Found the different from year to year was less than 1 degree
+
+############
+
+import logging
+
+logging.basicConfig()
+logging.getLogger().setLevel(logging.DEBUG)
+logging.getLogger("urllib3").setLevel(logging.WARNING)
+
+from datetime import datetime
+from pathlib import Path
+import pandas as pd
+import numpy as np
+import os
+import nowcasting_dataset
+from nowcasting_dataset.data_sources.gsp.eso import get_gsp_metadata_from_eso
+from nowcasting_dataset.data_sources.sun.raw_data_load_save import (
+    save_to_zarr,
+    get_azimuth_and_elevation,
+)
+
+# set up
+BUCKET = Path("solar-pv-nowcasting-data")
+PV_PATH = BUCKET / "PV/PVOutput.org"
+PV_METADATA_FILENAME = PV_PATH / "UK_PV_metadata.csv"
+
+# set up variables
+local_path = os.path.dirname(nowcasting_dataset.__file__) + "/.."
+metadata_filename = f"gs://{PV_METADATA_FILENAME}"
+
+# PV metadata
+pv_metadata = pd.read_csv(metadata_filename, index_col="system_id")
+pv_metadata = pv_metadata.dropna(subset=["longitude", "latitude"])
+pv_longitudes = pv_metadata["longitude"]
+pv_latitudes = pv_metadata["latitude"]
+
+# GSP Metadata
+gsp_metadata = get_gsp_metadata_from_eso()
+gsp_metadata = gsp_metadata.dropna(subset=["centroid_lon", "centroid_lat"])
+# probably need to change this to centroid
+gsp_lon = gsp_metadata["centroid_lon"]
+gsp_lat = gsp_metadata["centroid_lat"]
+
+# join all sites together
+longitudes = list(pv_longitudes.values) + list(gsp_lon.values)
+latitudes = list(pv_latitudes.values) + list(gsp_lat.values)
+
+# make d
+start_dt = datetime.fromisoformat("2019-01-01 00:00:00.000+00:00")
+end_dt = datetime.fromisoformat("2019-01-02 00:00:00.000+00:00")
+
+azimuths = {}
+azimuths_sin = {}
+azimuths_cos = {}
+elevations = {}
+months = [1, 4, 7, 10]
+years = [2018, 2019, 2020]
+for month in months:
+    for year in years:
+        print(year)
+        print(month)
+        start_dt = start_dt.replace(year=year, month=month)
+        end_dt = end_dt.replace(year=year, month=month)
+        datestamps = pd.date_range(start=start_dt, end=end_dt, freq="5T")
+
+        azimuth, elevation = get_azimuth_and_elevation(
+            longitudes=longitudes, latitudes=latitudes, datestamps=datestamps
+        )
+
+        azimuths[f"{year}_{month}"] = azimuth
+        azimuths_sin[f"{year}_{month}"] = np.sin(np.deg2rad(azimuth))
+        azimuths_cos[f"{year}_{month}"] = np.cos(np.deg2rad(azimuth))
+        elevations[f"{year}_{month}"] = elevation
+
+m_azimuths = []
+m_azimuths_sin = []
+m_azimuths_cos = []
+m_elevations = []
+for month in months:
+    for year in years[1:]:
+        print(year)
+        print(month)
+
+        m_azimuths.append(
+            (np.abs(azimuths[f"{year}_{month}"].values - azimuths[f"2018_{month}"].values)).max()
+        )
+        m_azimuths_sin.append(
+            (
+                np.abs(
+                    azimuths_sin[f"{year}_{month}"].values - azimuths_sin[f"2018_{month}"].values
+                )
+            ).max()
+        )
+        m_azimuths_cos.append(
+            (
+                np.abs(
+                    azimuths_cos[f"{year}_{month}"].values - azimuths_cos[f"2018_{month}"].values
+                )
+            ).max()
+        )
+        m_elevations.append(
+            (
+                np.abs(elevations[f"{year}_{month}"].values - elevations[f"2018_{month}"].values)
+            ).max()
+        )
+
+
+# for small radians, sin(x) ~ x, so sin(x)*180/pi ~ degrees
+m_azimuths = np.array(m_azimuths_sin) * 180 / np.pi
+# m_azimuths = np.array(m_azimuths_cos) * 180 / np.pi
+
+print(f"Maximum azimuth difference is {max(m_azimuths)} degree")
+print(f"Maximum elevation difference is {max(m_elevations)} degree")
+
+# largest different in both azimuth and elevation < 1 degree --> Happy to use one yea data

nowcasting_dataset/config/model.py

@@ -64,6 +64,11 @@ class InputData(BaseModel):
         description="Path to the GeoTIFF Topographic data source",
     )
 
+    sun_zarr_path: str = Field(
+        "gs://solar-pv-nowcasting-data/Sun/v0/sun.zarr/",
+        description="Path to the Sun data source i.e Azimuth and Elevation",
+    )
+
 
 class OutputData(BaseModel):
     filepath: str = Field(

nowcasting_dataset/consts.py

@@ -27,8 +27,10 @@
 PV_SYSTEM_ROW_NUMBER = "pv_system_row_number"
 PV_SYSTEM_X_COORDS = "pv_system_x_coords"
 PV_SYSTEM_Y_COORDS = "pv_system_y_coords"
-PV_AZIMUTH_ANGLE = "pv_azimuth_angle"
-PV_ELEVATION_ANGLE = "pv_elevation_angle"
+
+SUN_AZIMUTH_ANGLE = "sun_azimuth_angle"
+SUN_ELEVATION_ANGLE = "sun_elevation_angle"
+
 PV_YIELD = "pv_yield"
 PV_DATETIME_INDEX = "pv_datetime_index"
 DEFAULT_N_PV_SYSTEMS_PER_EXAMPLE = 128

nowcasting_dataset/data_sources/gsp/gsp_data_source.py

@@ -26,6 +26,7 @@
     DEFAULT_N_GSP_PER_EXAMPLE,
     OBJECT_AT_CENTER,
 )
+import time
 
 logger = logging.getLogger(__name__)
 

nowcasting_dataset/data_sources/pv_data_source.py

@@ -3,8 +3,6 @@
     PV_SYSTEM_ROW_NUMBER,
     PV_SYSTEM_X_COORDS,
     PV_SYSTEM_Y_COORDS,
-    PV_AZIMUTH_ANGLE,
-    PV_ELEVATION_ANGLE,
     PV_YIELD,
     DEFAULT_N_PV_SYSTEMS_PER_EXAMPLE,
     OBJECT_AT_CENTER,
@@ -17,7 +15,6 @@
 import pandas as pd
 import numpy as np
 import torch
-from tqdm import tqdm
 from numbers import Number
 from typing import List, Tuple, Union, Optional
 import datetime
@@ -27,8 +24,6 @@
 import xarray as xr
 import functools
 import logging
-import time
-from concurrent import futures
 
 logger = logging.getLogger(__name__)
 
@@ -56,8 +51,6 @@ def __post_init__(self, image_size_pixels: int, meters_per_pixel: int):
     def load(self):
         self._load_metadata()
         self._load_pv_power()
-        if self.load_azimuth_and_elevation:
-            self._calculate_azimuth_and_elevation()
         self.pv_metadata, self.pv_power = align_pv_system_ids(self.pv_metadata, self.pv_power)
 
     def _load_metadata(self):
@@ -123,18 +116,10 @@ def _get_time_slice(self, t0_dt: pd.Timestamp) -> [pd.DataFrame]:
         del t0_dt  # t0 is not used in the rest of this method!
         selected_pv_power = self.pv_power.loc[start_dt:end_dt].dropna(axis="columns", how="any")
 
-        if self.load_azimuth_and_elevation:
-            selected_pv_azimuth_angle = self.pv_azimuth.loc[start_dt:end_dt].dropna(
-                axis="columns", how="any"
-            )
-            selected_pv_elevation_angle = self.pv_elevation.loc[start_dt:end_dt].dropna(
-                axis="columns", how="any"
-            )
-        else:
-            selected_pv_azimuth_angle = None
-            selected_pv_elevation_angle = None
+        selected_pv_azimuth_angle = None
+        selected_pv_elevation_angle = None
 
-        return selected_pv_power, selected_pv_azimuth_angle, selected_pv_elevation_angle
+        return selected_pv_power
 
     def _get_central_pv_system_id(
         self,
@@ -208,11 +193,7 @@ def get_example(
 
         logger.debug("Getting PV example data")
 
-        (
-            selected_pv_power,
-            selected_pv_azimuth_angle,
-            selected_pv_elevation_angle,
-        ) = self._get_time_slice(t0_dt)
+        selected_pv_power = self._get_time_slice(t0_dt)
         all_pv_system_ids = self._get_all_pv_system_ids_in_roi(
             x_meters_center, y_meters_center, selected_pv_power.columns
         )
@@ -229,9 +210,6 @@ def get_example(
         all_pv_system_ids = all_pv_system_ids[: self.n_pv_systems_per_example]
 
         selected_pv_power = selected_pv_power[all_pv_system_ids]
-        if self.load_azimuth_and_elevation:
-            selected_pv_azimuth_angle = selected_pv_azimuth_angle[all_pv_system_ids]
-            selected_pv_elevation_angle = selected_pv_elevation_angle[all_pv_system_ids]
 
         pv_system_row_number = np.flatnonzero(self.pv_metadata.index.isin(all_pv_system_ids))
         pv_system_x_coords = self.pv_metadata.location_x[all_pv_system_ids]
@@ -249,10 +227,6 @@ def get_example(
             pv_datetime_index=selected_pv_power.index,
         )
 
-        if self.load_azimuth_and_elevation:
-            example[PV_AZIMUTH_ANGLE] = selected_pv_azimuth_angle
-            example[PV_ELEVATION_ANGLE] = selected_pv_elevation_angle
-
         if self.get_center:
             example[OBJECT_AT_CENTER] = "pv"
 
@@ -267,9 +241,6 @@ def get_example(
         ]
 
         pad_nans_variables = [PV_YIELD]
-        if self.load_azimuth_and_elevation:
-            pad_nans_variables.append(PV_AZIMUTH_ANGLE)
-            pad_nans_variables.append(PV_ELEVATION_ANGLE)
 
         example = utils.pad_data(
             data=example,
@@ -319,74 +290,6 @@ def datetime_index(self) -> pd.DatetimeIndex:
         """Returns a complete list of all available datetimes."""
         return self.pv_power.index
 
-    def _calculate_azimuth_and_elevation(self):
-        """
-        Calculate the azimuth and elevation angles for each datestamp, for each pv system.
-        """
-
-        logger.debug("Calculating azimuth and elevation angles")
-
-        self.pv_azimuth, self.pv_elevation = calculate_azimuth_and_elevation_all_pv_systems(
-            self.datetime_index().to_pydatetime(), self.pv_metadata
-        )
-
-
-def calculate_azimuth_and_elevation_all_pv_systems(
-    datestamps: List[datetime.datetime], pv_metadata: pd.DataFrame
-) -> (pd.Series, pd.Series):
-    """
-    Calculate the azimuth and elevation angles for each datestamp, for each pv system.
-    """
-
-    logger.debug(
-        f"Will be calculating for {len(datestamps)} datestamps and {len(pv_metadata)} pv systems"
-    )
-
-    # create array of index datetime, columns of system_id for both azimuth and elevation
-    pv_azimuth = []
-    pv_elevation = []
-
-    t = time.time()
-    # loop over all metadata and fine azimuth and elevation angles,
-    # not sure this is the best method to use, as currently this step takes ~2 minute for 745 pv systems,
-    # and 235 datestamps (~100,000 point). But this only needs to be done once.
-    with futures.ThreadPoolExecutor(max_workers=len(pv_metadata)) as executor:
-
-        logger.debug("Setting up jobs")
-
-        # Submit tasks to the executor.
-        future_azimuth_and_elevation_per_pv_system = []
-        for i in tqdm(range(len(pv_metadata))):
-            future_azimuth_and_elevation = executor.submit(
-                geospatial.calculate_azimuth_and_elevation_angle,
-                latitude=pv_metadata.iloc[i].latitude,
-                longitude=pv_metadata.iloc[i].longitude,
-                datestamps=datestamps,
-            )
-            future_azimuth_and_elevation_per_pv_system.append(
-                [future_azimuth_and_elevation, pv_metadata.iloc[i].name]
-            )
-
-        logger.debug(f"Getting results")
-
-        # Collect results from each thread.
-        for i in tqdm(range(len(future_azimuth_and_elevation_per_pv_system))):
-            future_azimuth_and_elevation, name = future_azimuth_and_elevation_per_pv_system[i]
-            azimuth_and_elevation = future_azimuth_and_elevation.result()
-
-            azimuth = azimuth_and_elevation.loc[:, "azimuth"].rename(name)
-            elevation = azimuth_and_elevation.loc[:, "elevation"].rename(name)
-
-            pv_azimuth.append(azimuth)
-            pv_elevation.append(elevation)
-
-    pv_azimuth = pd.concat(pv_azimuth, axis=1)
-    pv_elevation = pd.concat(pv_elevation, axis=1)
-
-    logger.debug(f"Calculated Azimuth and Elevation angles in {time.time() - t} seconds")
-
-    return pv_azimuth, pv_elevation
-
 
 def load_solar_pv_data_from_gcs(
     filename: Union[str, Path],