ubc-solar-physics 0.1.11__cp311-cp311-macosx_13_0_universal2.whl

physics/environment/weather_forecasts/OpenWeatherForecast.py

@@ -0,0 +1,307 @@
+"""
+A class to extract local and path weather predictions such as wind_speed, 
+    wind_direction, cloud_cover and weather type from OpenWeather.
+"""
+import numpy as np
+import datetime
+
+from physics.environment import OpenweatherEnvironment
+from physics.environment.weather_forecasts.base_weather_forecasts import BaseWeatherForecasts
+from physics.environment.race import Race
+
+import core
+
+
+class OpenWeatherForecast(BaseWeatherForecasts):
+    """
+    Class that gathers weather data and performs calculations on it to allow the implementation of weather phenomenon
+    such as changes in wind speeds and cloud cover in the simulation.
+
+    Attributes:
+        coords (NumPy array [N][lat, long]): a list of N coordinates for which to gather weather forecasts for
+        origin_coord (NumPy array [lat, long]): the starting coordinate
+        dest_coord (NumPy array [lat, long]): the ending coordinate
+        last_updated_time (int): value that tells us the starting time after which we have weather data available
+
+        weather_forecast (NumPy array [N][T][9]): array that stores the complete weather forecast data. N represents the
+            number of coordinates, T represents time length which differs depending on the `weather_data_frequency`
+            argument ("current" -> T = 1 ; "hourly" -> T = 24 ; "daily" -> T = 8). The last 9 represents the number of
+            weather forecast fields available. These are: (latitude, longitude, dt (UNIX time), timezone_offset
+            (in seconds), dt + timezone_offset (local time), wind_speed, wind_direction, cloud_cover, description_id)
+    """
+
+    def __init__(self, coords, race: Race, origin_coord=None, hash_key=None):
+
+        """
+
+        Initializes the instance of a WeatherForecast class
+
+        :param origin_coord: A NumPy array of a single [latitude, longitude]
+        :param str provider: string indicating weather provider
+        :param coords: A NumPy array of [latitude, longitude]
+        :param hash_key: key used to identify cached data as valid for a Simulation model
+
+        """
+        super().__init__(coords, race, "OPENWEATHER", origin_coord, hash_key)
+
+        self.last_updated_time = self.weather_forecast[0, 0, 2]
+
+        start_time_unix = self.weather_forecast[0, 0, 2]
+        end_time_unix = self.weather_forecast[0, -1, 2]
+        start_time = date_from_unix_timestamp(start_time_unix)
+        end_time = date_from_unix_timestamp(end_time_unix)
+
+        print("----- Weather save file information -----\n")
+        print(f"--- Data time range ---")
+        print(f"Start time (UTC)f: {start_time} [{start_time_unix:.0f}]\n"
+              f"End time (UTC): {end_time} [{end_time_unix:.0f}]\n")
+
+
+    def calculate_closest_weather_indices(self, cumulative_distances):
+        """
+
+        :param np.ndarray cumulative_distances: NumPy Array representing cumulative distances theoretically achievable for a given input speed array
+        :returns: array of the closest weather indices.
+        :rtype: np.ndarray
+
+        """
+
+        """
+        IMPORTANT: we only have weather coordinates for a discrete set of coordinates. However, the car could be at any
+        coordinate in between these available weather coordinates. We need to figure out what coordinate the car is at
+        at each timestep and then we can figure out the full weather forecast at each timestep.
+
+        For example, imagine the car is at some coordinate (10, 20). Further imagine that we have a week's worth of
+        weather forecasts for the following five coordinates: (5, 4), (11, 19), (20, 30), (40, 30), (0, 60). Which
+        set of weather forecasts should we choose? Well, we should choose the (11, 19) one since our coordinate
+        (10, 20) is closest to (11, 19). This is what the following code is accomplishing. However, it is not dealing
+        with the coordinates directly but rather is dealing with the distances between the coordinates. 
+
+        Furthermore, once we have chosen a week's worth of weather forecasts for a specific coordinate, we must isolate
+        a single weather forecast depending on what time the car is at the coordinate (10, 20). That is the job of the
+        `get_weather_forecast_in_time()` method.
+        
+        """
+
+        # if racing FSGP, there is no need for distance calculations. We will return only the origin coordinate
+        # This characterizes the weather at every point along the FSGP tracks
+        # with the weather at a single coordinate on the track, which is great for reducing the API calls and is a
+        # reasonable assumption to make for FSGP only.
+        if self.race.race_type == Race.FSGP:
+            result = np.zeros_like(cumulative_distances, dtype=int)
+            return result
+
+        # a list of all the coordinates that we have weather data for
+        weather_coords = self.weather_forecast[:, 0, 0:2]
+
+        # distances between all the coordinates that we have weather data for
+        weather_path_distances = calculate_path_distances(weather_coords)
+        cumulative_weather_path_distances = np.cumsum(weather_path_distances)
+
+        # makes every even-index element negative, this allows the use of np.diff() to calculate the sum of consecutive
+        # elements
+        cumulative_weather_path_distances[::2] *= -1
+
+        # contains the average distance between two consecutive elements in the cumulative_weather_path_distances array
+        average_distances = np.abs(np.diff(cumulative_weather_path_distances) / 2)
+
+        return core.closest_weather_indices_loop(cumulative_distances, average_distances)
+
+    @staticmethod
+    def _python_calculate_closest_weather_indices(cumulative_distances, average_distances):
+        """
+
+        Python implementation of calculate_closest_weather_indices. See parent function for documentation details.
+
+        """
+
+        current_coordinate_index = 0
+        result = []
+
+        for distance in np.nditer(cumulative_distances):
+
+            # makes sure the current_coordinate_index does not exceed its maximum value
+            if current_coordinate_index > len(average_distances) - 1:
+                current_coordinate_index = len(average_distances) - 1
+
+            if distance > average_distances[current_coordinate_index]:
+                current_coordinate_index += 1
+                if current_coordinate_index > len(average_distances) - 1:
+                    current_coordinate_index = len(average_distances) - 1
+
+            result.append(current_coordinate_index)
+
+        return np.array(result)
+
+    @staticmethod
+    def _python_calculate_closest_timestamp_indices(unix_timestamps, dt_local_array):
+        """
+
+        Python implementation to find the indices of the closest timestamps in dt_local_array and package them into a NumPy Array
+
+        :param np.ndarray unix_timestamps: NumPy Array (float[N]) of unix timestamps of the vehicle's journey
+        :param np.ndarray dt_local_array: NumPy Array (float[N]) of local times, represented as unix timestamps
+        :returns: NumPy Array of (int[N]) containing closest timestamp indices used by get_weather_forecast_in_time
+        :rtype: np.ndarray
+
+        """
+        closest_time_stamp_indices = []
+        for unix_timestamp in unix_timestamps:
+            unix_timestamp_array = np.full_like(dt_local_array, fill_value=unix_timestamp)
+            differences = np.abs(unix_timestamp_array - dt_local_array)
+            minimum_index = np.argmin(differences)
+            closest_time_stamp_indices.append(minimum_index)
+
+        return np.asarray(closest_time_stamp_indices, dtype=np.int32)
+
+    def get_weather_forecast_in_time(self, indices, unix_timestamps, start_hour, tick):
+        """
+
+        Takes in an array of indices of the weather_forecast array, and an array of timestamps. Uses those to figure out
+        what the weather forecast is at each time step being simulated.
+
+        we only have weather at discrete timestamps. The car however can be in any timestamp in
+        between. Therefore, we must be able to choose the weather timestamp that is closest to the one that the car is in
+        so that we can more accurately determine the weather experienced by the car at that timestamp.
+
+        For example, imagine the car is at some coordinate (x,y) at timestamp 100. Imagine we know the weather forecast
+        at (x,y) for five different timestamps: 0, 30, 60, 90, and 120. Which weather forecast should we
+        choose? Clearly, we should choose the weather forecast at 90 since it is the closest to 100. That's what the
+        below code is accomplishing.
+
+        :param np.ndarray indices: (int[N]) coordinate indices of self.weather_forecast
+        :param np.ndarray unix_timestamps: (int[N]) unix timestamps of the vehicle's journey
+        :param int start_hour: the starting hour of simulation
+        :param int tick: length of a tick in seconds
+        :returns:
+            - A NumPy array of size [N][9]
+            - [9] (latitude, longitude, unix_time, timezone_offset, unix_time_corrected, wind_speed, wind_direction,
+                        cloud_cover, precipitation, description):
+        :rtype: np.ndarray
+
+        """
+        weather_data = core.weather_in_time(unix_timestamps.astype(np.int64), indices.astype(np.int64), self.weather_forecast, 4)
+        roll_by_tick = int(3600 / tick) * (24 + start_hour - hour_from_unix_timestamp(weather_data[0, 2]))
+        weather_data = np.roll(weather_data, -roll_by_tick, 0)
+
+        weather_object = OpenweatherEnvironment()
+        weather_object.latitude = weather_data[:, 0]
+        weather_object.longitude = weather_data[:, 1]
+        weather_object.unix_time = weather_data[:, 2]
+        weather_object.wind_speed = weather_data[:, 5]
+        weather_object.wind_direction = weather_data[:, 6]
+        weather_object.cloud_cover = weather_data[:, 7]
+
+        return weather_object
+
+    def _python_get_weather_in_time(self, unix_timestamps, indices):
+        full_weather_forecast_at_coords = self.weather_forecast[indices]
+        dt_local_array = full_weather_forecast_at_coords[0, :, 4]
+
+        temp_0 = np.arange(0, full_weather_forecast_at_coords.shape[0])
+        closest_timestamp_indices = self._python_calculate_closest_timestamp_indices(unix_timestamps, dt_local_array)
+
+        return full_weather_forecast_at_coords[temp_0, closest_timestamp_indices]
+
+    @staticmethod
+    def _get_array_directional_wind_speed(vehicle_bearings, wind_speeds, wind_directions):
+        """
+
+        Returns the array of wind speed in m/s, in the direction opposite to the 
+            bearing of the vehicle
+
+
+        :param np.ndarray vehicle_bearings: (float[N]) The azimuth angles that the vehicle in, in degrees
+        :param np.ndarray wind_speeds: (float[N]) The absolute speeds in m/s
+        :param np.ndarray wind_directions: (float[N]) The wind direction in the meteorlogical convention. To convert from meteorlogical convention to azimuth angle, use (x + 180) % 360
+        :returns: The wind speeds in the direction opposite to the bearing of the vehicle
+        :rtype: np.ndarray
+        
+        """
+
+        # wind direction is 90 degrees meteorological, so it is 270 degrees azimuthal. car is 90 degrees
+        #   cos(90 - 90) = cos(0) = 1. Wind speed is moving opposite to the car,
+        # car is 270 degrees, cos(90-270) = -1. Wind speed is in direction of the car.
+        return wind_speeds * (np.cos(np.radians(wind_directions - vehicle_bearings)))
+
+    @staticmethod
+    def _get_weather_advisory(weather_id):
+        """
+
+        Returns a string indicating the type of weather to expect, from the standardized
+            weather code passed as a parameter
+
+        https://openweathermap.org/weather-conditions#Weather-Condition-Codes-2
+        :param int weather_id: Weather ID
+        :return: type of weather advisory
+        :rtype: str
+
+        """
+
+        if 200 <= weather_id < 300:
+            return "Thunderstorm"
+        elif 300 <= weather_id < 500:
+            return "Drizzle"
+        elif 500 <= weather_id < 600:
+            return "Rain"
+        elif 600 <= weather_id < 700:
+            return "Snow"
+        elif weather_id == 800:
+            return "Clear"
+        else:
+            return "Unknown"
+
+
+def date_from_unix_timestamp(unix_timestamp):
+    """
+
+    Return a stringified UTC datetime from UNIX timestamped.
+
+    :param int unix_timestamp: A unix timestamp
+
+    :returns: A string of the UTC representation of the UNIX timestamp in the format Y-m-d H:M:S
+    :rtype: str
+
+    """
+
+    return datetime.datetime.utcfromtimestamp(unix_timestamp).strftime('%Y-%m-%d %H:%M:%S')
+
+
+def calculate_path_distances(coords):
+    """
+
+    Obtain the distance between each coordinate by approximating the spline between them
+    as a straight line, and use the Haversine formula (https://en.wikipedia.org/wiki/Haversine_formula)
+    to calculate distance between coordinates on a sphere.
+
+    :param np.ndarray coords: A NumPy array [n][latitude, longitude]
+    :returns path_distances: a NumPy array [n-1][distances],
+    :rtype: np.ndarray
+
+    """
+
+    coords_offset = np.roll(coords, (1, 1))
+    path_distances = []
+    for u, v in zip(coords, coords_offset):
+        path_distances.append(haversine(u, v, unit=Unit.METERS))
+
+    return np.array(path_distances)
+
+
+def hour_from_unix_timestamp(unix_timestamp):
+    """
+
+    Return the hour of a UNIX timestamp.
+
+    :param float unix_timestamp: a UNIX timestamp
+    :return: hour of UTC datetime from unix timestamp
+    :rtype: int
+
+    """
+
+    val = datetime.datetime.utcfromtimestamp(unix_timestamp)
+    return val.hour
+
+
+if __name__ == "__main__":
+    pass

physics/environment/weather_forecasts/SolcastForecasts.py

@@ -0,0 +1,235 @@
+"""
+A class to extract local and path weather predictions such as wind_speed, 
+    wind_direction, cloud_cover and weather type using data from Solcast.
+"""
+import numpy as np
+from physics.environment.weather_forecasts import BaseWeatherForecasts
+from physics.environment.solcast_environment import SolcastEnvironment
+from physics.environment.race import Race
+from haversine import haversine, Unit
+
+import core
+
+
+class SolcastForecasts(BaseWeatherForecasts):
+    """
+    Class that gathers weather data and performs calculations on it to allow the implementation of weather phenomenon
+    such as changes in wind speeds and cloud cover in the simulation.
+
+    Attributes:
+        coords (NumPy array [N][lat, long]): a list of N coordinates for which to gather weather forecasts for
+        origin_coord (NumPy array [lat, long]): the starting coordinate
+        dest_coord (NumPy array [lat, long]): the ending coordinate
+        last_updated_time (int): value that tells us the starting time after which we have weather data available
+
+        weather_forecast (NumPy array [N][T][6]): array that stores the complete weather forecast data. N represents the
+            number of coordinates, T represents time length with temporal granularity defined in settings_*.json.
+            The last 6 represents the number of weather forecast fields available. These are: time_dt (UTC UNIX time),
+            latitude, longitude, wind_speed (m/s), wind_direction (meteorological convention), ghi (W/m^2)
+    """
+
+    def __init__(self, coords, race: Race, origin_coord=None, hash_key=None):
+
+        """
+
+        Initializes the instance of a WeatherForecast class
+
+        :param origin_coord: A NumPy array of a single [latitude, longitude]
+        :param str provider: string indicating weather provider
+        :param coords: A NumPy array of [latitude, longitude]
+        :param hash_key: key used to identify cached data as valid for a Simulation model
+
+        """
+        super().__init__(coords, race, "SOLCAST", origin_coord, hash_key)
+        self.race = race
+        self.last_updated_time = self.weather_forecast[0, 0, 0]
+
+    def calculate_closest_weather_indices(self, cumulative_distances):
+        """
+
+        :param np.ndarray cumulative_distances: NumPy Array representing cumulative distances theoretically achievable for a given input speed array
+        :returns: array of the closest weather indices.
+        :rtype: np.ndarray
+
+        """
+
+        """
+        IMPORTANT: we only have weather coordinates for a discrete set of coordinates. However, the car could be at any
+        coordinate in between these available weather coordinates. We need to figure out what coordinate the car is at
+        at each timestep and then we can figure out the full weather forecast at each timestep.
+
+        For example, imagine the car is at some coordinate (10, 20). Further imagine that we have a week's worth of
+        weather forecasts for the following five coordinates: (5, 4), (11, 19), (20, 30), (40, 30), (0, 60). Which
+        set of weather forecasts should we choose? Well, we should choose the (11, 19) one since our coordinate
+        (10, 20) is closest to (11, 19). This is what the following code is accomplishing. However, it is not dealing
+        with the coordinates directly but rather is dealing with the distances between the coordinates. 
+
+        Furthermore, once we have chosen a week's worth of weather forecasts for a specific coordinate, we must isolate
+        a single weather forecast depending on what time the car is at the coordinate (10, 20). That is the job of the
+        `get_weather_forecast_in_time()` method.
+        
+        """
+
+        # if racing FSGP, there is no need for distance calculations. We will return only the origin coordinate
+        # This characterizes the weather at every point along the FSGP tracks
+        # with the weather at a single coordinate on the track, which is great for reducing the API calls and is a
+        # reasonable assumption to make for FSGP only.
+        if self.race.race_type == Race.FSGP:
+            result = np.zeros_like(cumulative_distances, dtype=int)
+            return result
+
+        # a list of all the coordinates that we have weather data for
+        weather_coords = self.weather_forecast[:, 0, 1:3]
+
+        # distances between all the coordinates that we have weather data for
+        weather_path_distances = calculate_path_distances(weather_coords)
+        cumulative_weather_path_distances = np.cumsum(weather_path_distances)
+
+        # makes every even-index element negative, this allows the use of np.diff() to calculate the sum of consecutive
+        # elements
+        cumulative_weather_path_distances[::2] *= -1
+
+        # contains the average distance between two consecutive elements in the cumulative_weather_path_distances array
+        average_distances = np.abs(np.diff(cumulative_weather_path_distances) / 2)
+
+        return core.closest_weather_indices_loop(cumulative_distances, average_distances)
+
+    @staticmethod
+    def _python_calculate_closest_weather_indices(cumulative_distances, average_distances):
+        """
+
+        Python implementation of calculate_closest_weather_indices. See parent function for documentation details.
+
+        """
+
+        current_coordinate_index = 0
+        result = []
+
+        for distance in np.nditer(cumulative_distances):
+
+            # makes sure the current_coordinate_index does not exceed its maximum value
+            if current_coordinate_index > len(average_distances) - 1:
+                current_coordinate_index = len(average_distances) - 1
+
+            if distance > average_distances[current_coordinate_index]:
+                current_coordinate_index += 1
+                if current_coordinate_index > len(average_distances) - 1:
+                    current_coordinate_index = len(average_distances) - 1
+
+            result.append(current_coordinate_index)
+
+        return np.array(result)
+
+    @staticmethod
+    def _python_calculate_closest_timestamp_indices(unix_timestamps, dt_local_array):
+        """
+
+        Python implementation to find the indices of the closest timestamps in dt_local_array and package them into a NumPy Array
+
+        :param np.ndarray unix_timestamps: NumPy Array (float[N]) of unix timestamps of the vehicle's journey
+        :param np.ndarray dt_local_array: NumPy Array (float[N]) of local times, represented as unix timestamps
+        :returns: NumPy Array of (int[N]) containing closest timestamp indices used by get_weather_forecast_in_time
+        :rtype: np.ndarray
+
+        """
+        closest_time_stamp_indices = []
+        for unix_timestamp in unix_timestamps:
+            unix_timestamp_array = np.full_like(dt_local_array, fill_value=unix_timestamp)
+            differences = np.abs(unix_timestamp_array - dt_local_array)
+            minimum_index = np.argmin(differences)
+            closest_time_stamp_indices.append(minimum_index)
+
+        return np.asarray(closest_time_stamp_indices, dtype=np.int32)
+
+    def get_weather_forecast_in_time(self, indices, unix_timestamps, start_time, tick) -> SolcastEnvironment:
+        """
+
+        Takes in an array of indices of the weather_forecast array, and an array of timestamps. Uses those to figure out
+        what the weather forecast is at each time step being simulated.
+
+        we only have weather at discrete timestamps. The car however can be in any timestamp in
+        between. Therefore, we must be able to choose the weather timestamp that is closest to the one that the car is in
+        so that we can more accurately determine the weather experienced by the car at that timestamp.
+
+        For example, imagine the car is at some coordinate (x,y) at timestamp 100. Imagine we know the weather forecast
+        at (x,y) for five different timestamps: 0, 30, 60, 90, and 120. Which weather forecast should we
+        choose? Clearly, we should choose the weather forecast at 90 since it is the closest to 100. That's what the
+        below code is accomplishing.
+
+        :param np.ndarray indices: (int[N]) coordinate indices of self.weather_forecast
+        :param np.ndarray unix_timestamps: (int[N]) unix timestamps of the vehicle's journey
+        :param int start_time: time since the start of the race that simulation is beginning
+        :param int tick: length of a tick in seconds
+        :returns: a SolcastEnvironment object with time_dt, latitude, longitude, wind_speed, wind_direction, and ghi.
+        :rtype: SolcastEnvironment
+        """
+        forecasts_array = core.weather_in_time(unix_timestamps.astype(np.int64), indices.astype(np.int64), self.weather_forecast, 0)
+
+        # roll_by_tick = int(3600 / tick) * helpers.hour_from_unix_timestamp(forecasts_array[0, 0])
+        # forecasts_array = np.roll(forecasts_array, -roll_by_tick, 0)
+
+        weather_object = SolcastEnvironment()
+
+        weather_object.time_dt = forecasts_array[:, 0]
+        weather_object.latitude = forecasts_array[:, 1]
+        weather_object.longitude = forecasts_array[:, 2]
+        weather_object.wind_speed = forecasts_array[:, 3]
+        weather_object.wind_direction = forecasts_array[:, 4]
+        weather_object.ghi = forecasts_array[:, 5]
+
+        return weather_object
+
+    def _python_get_weather_in_time(self, unix_timestamps, indices):
+        full_weather_forecast_at_coords = self.weather_forecast[indices]
+        dt_local_array = full_weather_forecast_at_coords[0, :, 0]
+
+        temp_0 = np.arange(0, full_weather_forecast_at_coords.shape[0])
+        closest_timestamp_indices = self._python_calculate_closest_timestamp_indices(unix_timestamps, dt_local_array)
+
+        return full_weather_forecast_at_coords[temp_0, closest_timestamp_indices]
+
+    @staticmethod
+    def _get_array_directional_wind_speed(vehicle_bearings, wind_speeds, wind_directions):
+        """
+
+        Returns the array of wind speed in m/s, in the direction opposite to the 
+            bearing of the vehicle
+
+
+        :param np.ndarray vehicle_bearings: (float[N]) The azimuth angles that the vehicle in, in degrees
+        :param np.ndarray wind_speeds: (float[N]) The absolute speeds in m/s
+        :param np.ndarray wind_directions: (float[N]) The wind direction in the meteorlogical convention. To convert from meteorlogical convention to azimuth angle, use (x + 180) % 360
+        :returns: The wind speeds in the direction opposite to the bearing of the vehicle
+        :rtype: np.ndarray
+        
+        """
+
+        # wind direction is 90 degrees meteorological, so it is 270 degrees azimuthal. car is 90 degrees
+        #   cos(90 - 90) = cos(0) = 1. Wind speed is moving opposite to the car,
+        # car is 270 degrees, cos(90-270) = -1. Wind speed is in direction of the car.
+        return wind_speeds * (np.cos(np.radians(wind_directions - vehicle_bearings)))
+
+
+def calculate_path_distances(coords):
+    """
+
+    Obtain the distance between each coordinate by approximating the spline between them
+    as a straight line, and use the Haversine formula (https://en.wikipedia.org/wiki/Haversine_formula)
+    to calculate distance between coordinates on a sphere.
+
+    :param np.ndarray coords: A NumPy array [n][latitude, longitude]
+    :returns path_distances: a NumPy array [n-1][distances],
+    :rtype: np.ndarray
+
+    """
+
+    coords_offset = np.roll(coords, (1, 1))
+    path_distances = []
+    for u, v in zip(coords, coords_offset):
+        path_distances.append(haversine(u, v, unit=Unit.METERS))
+
+    return np.array(path_distances)
+
+
+if __name__ == "__main__":
+    pass

physics/environment/weather_forecasts/__init__.py

@@ -0,0 +1,9 @@
+from .base_weather_forecasts import BaseWeatherForecasts
+from .OpenWeatherForecast import OpenWeatherForecast
+from .SolcastForecasts import SolcastForecasts
+
+__all__ = [
+    "BaseWeatherForecasts",
+    "OpenWeatherForecast",
+    "SolcastForecasts"
+]

physics/environment/weather_forecasts/base_weather_forecasts.py

@@ -0,0 +1,56 @@
+import pathlib
+from abc import ABC, abstractmethod
+
+import dill
+import numpy as np
+from physics.environment import Race
+import logging
+import os
+
+
+class BaseWeatherForecasts(ABC):
+    def __init__(self, coords, race: Race, provider: str, weather_directory: pathlib.Path, origin_coord=None, hash_key=None, reduction_factor=625):
+        self.race = race
+
+        if origin_coord is not None:
+            self.origin_coord = np.array(origin_coord)
+        else:
+            self.origin_coord = coords[0]
+        self.dest_coord = coords[-1]
+
+        if self.race.race_type == Race.ASC:
+            self.coords = coords[::reduction_factor]
+            weather_file = weather_directory / f"weather_data_{provider}.npz"
+        elif self.race.race_type == Race.FSGP:
+            self.coords = np.array([coords[0], coords[-1]])
+            weather_file = weather_directory / f"weather_data_FSGP_{provider}.npz"
+        else:
+            raise ValueError(f"base_weather_forecasts has not implemented retrieving race {repr(self.race.race_type)}")
+
+        # if the file exists, load path from file
+        if os.path.isfile(weather_file):
+            if provider == "SOLCAST":
+                with open(weather_file, 'rb') as file:
+                    weather_data = dill.load(file)
+            elif provider == "OPENWEATHER":
+                weather_data = np.load(weather_file)
+            else:
+                raise ValueError(f"base_weather_forecasts has not implemented retrieving provider {provider}")
+
+            if weather_data['hash'] == hash_key:
+
+                print("Previous weather save file is being used...\n")
+
+                self.weather_forecast = weather_data['weather_forecast']
+
+        else:
+            logging.error("Get or update cached weather data by invoking cache_data.py\nExiting simulation...")
+            exit()
+
+    @abstractmethod
+    def calculate_closest_weather_indices(self, cumulative_distances) -> np.ndarray:
+        raise NotImplementedError
+
+    @abstractmethod
+    def get_weather_forecast_in_time(self, indices, unix_timestamps, start_hour, tick) -> np.ndarray:
+        raise NotImplementedError