def interpolate_series(self, series: pd.Series) -> pd.Series:
logger.debug(f"Interpolating series; series len = {len(series)}")
series = pd.to_numeric(series, downcast="float")
interpolated_series = series.interpolate(method="linear",
limit_area="outside",
limit_direction="both")
return interpolated_series
def read_temp_graph_from_binary_stream(
self, binary_stream: BinaryIO) -> pd.DataFrame:
logger.debug("Loading temp graph")
temp_graph_df = pd.read_csv(binary_stream,
encoding=self._encoding,
sep=self._separator)
return temp_graph_df
def __init__(self, filepath: str,
writer: AbstractSyncHeatingObjWriter) -> None:
self._filepath = filepath
self._writer = writer
logger.debug(f"Creating instance:"
f"filepath: {filepath}"
f"writer: {writer}")
def predict(self, weather_df: pd.DataFrame,
system_state_history_df: pd.DataFrame,
control_action_df: pd.DataFrame) -> pd.DataFrame:
logger.debug("Requested prediction")
boiler_temp, control_action_timestamp = self._unpack_control_action(
control_action_df)
heating_system_reaction = []
heating_obj_ids = self._timedelta_df[
column_names.HEATING_OBJ_ID].to_list()
timedelta_list = self._timedelta_df[
column_names.AVG_TIMEDELTA].to_list()
for obj_id, timedelta in zip(heating_obj_ids, timedelta_list):
coolant_temp = self._calc_coolant_temp_for_object(
obj_id, boiler_temp)
heating_system_reaction.append({
column_names.TIMESTAMP:
control_action_timestamp + timedelta,
column_names.HEATING_OBJ_ID:
obj_id,
column_names.HEATING_OBJ_TYPE:
self._objects_type,
column_names.CIRCUIT_TYPE:
self._circuit_type,
column_names.FORWARD_PIPE_COOLANT_TEMP:
coolant_temp,
column_names.BACKWARD_PIPE_COOLANT_TEMP:
None
})
return pd.DataFrame(heating_system_reaction)
def load_temp_graph(self) -> pd.DataFrame:
filepath = os.path.abspath(self._filepath)
logger.debug(f"Loading temp graph from {filepath}")
with open(filepath, mode="rb") as input_file:
temp_graph_df = self._reader.read_temp_graph_from_binary_stream(
input_file)
return temp_graph_df
def find_lag(self, x: np.ndarray, y: np.ndarray) -> int:
logger.debug("Lag calculation is requested")
rounded_x_column = "rounded_x"
y_column = "y"
min_std = float("inf")
lag = None
for test_lag in range(self._min_lag, self._max_lag):
x_with_lag = x[:-test_lag]
y_with_lag = y[test_lag:]
rounded_x = x_with_lag // self._x_round_step * self._x_round_step
correlation_df = pd.DataFrame({
rounded_x_column: rounded_x,
y_column: y_with_lag
})
mean_series = correlation_df.groupby(
rounded_x_column)[y_column].mean()
y_mean_group_value = (correlation_df[rounded_x_column].replace(
mean_series)).to_numpy()
y_delta = y_with_lag - y_mean_group_value
# noinspection PyUnresolvedReferences
std_var = np.std(y_delta)
if std_var < min_std:
min_std = std_var
lag = test_lag
return lag
def __init__(
self,
heating_system_model: AbstractHeatingSystemModel,
temp_requirements_constraint: SingleTypeHeatingObjOnWeatherConstraint,
controlled_circuit_type: str = circuit_types.HEATING,
min_boiler_temp: float = 30,
max_boiler_temp: float = 85,
min_regulation_step: float = 0.3,
) -> None:
self._heating_system_model = heating_system_model
self._temp_requirements_constraint = temp_requirements_constraint
self._min_boiler_temp = min_boiler_temp
self._max_boiler_temp = max_boiler_temp
self._min_regulation_step = min_regulation_step
self._controlled_circuit_type = controlled_circuit_type
logger.debug(
f"Created instance: "
f"min boiler temp: {self._max_boiler_temp}"
f"max boiler temp: {self._max_boiler_temp}"
f"min regulation step: {self._min_regulation_step}"
f"controlled circuit type: {self._controlled_circuit_type}"
f"heating system model: {self._heating_system_model}"
f"temp requirements constraint: {self._temp_requirements_constraint}"
)
def __init__(
self, temp_graph: pd.DataFrame,
weather_temp_round_algorithm: AbstractFloatRoundAlgorithm) -> None:
self._temp_graph = temp_graph
self._weather_temp_round_algo = weather_temp_round_algorithm
logger.debug("Creating instance")
def predict_on_weather(self, weather_df: pd.DataFrame) -> pd.DataFrame:
logger.debug(
f"Predicting on weather is requested; weather_df len = {len(weather_df)}"
)
weather_df = self._round_weather_temp(weather_df)
temp_requirements_df = self._calc_for_weather_df(weather_df)
return temp_requirements_df
def __init__(self, encoding: str = "utf-8", separator: str = ";") -> None:
self._encoding = encoding
self._separator = separator
logger.debug(f"Creating instance:"
f"encoding: {encoding}"
f"separator: {separator}")
def write_weather_to_binary_stream(self, binary_stream: BinaryIO,
weather_df: pd.DataFrame) -> None:
logger.debug("Storing weather")
weather_df.to_csv(binary_stream,
encoding=self._encoding,
sep=self._separator,
index=False)
def write_temp_graph_to_binary_stream(self, binary_stream: BinaryIO,
temp_graph_df: pd.DataFrame) -> None:
logger.debug("Storing temp graph")
temp_graph_df.to_csv(binary_stream,
encoding=self._encoding,
sep=self._separator,
index=False)
def _load_from_file(self):
filepath = os.path.abspath(self._filepath)
logger.debug(f"Loading weather from {filepath}")
with open(filepath, mode="rb") as input_file:
weather_df = self._reader.read_weather_from_binary_stream(
input_file)
return weather_df
def write_timedelta_to_binary_stream(self, binary_stream: BinaryIO,
timedelta_df: pd.DataFrame) -> None:
logger.debug("Storing timedelta")
timedelta_df = self._convert_timedelta_to_seconds(timedelta_df)
with io.TextIOWrapper(binary_stream,
encoding=self._encoding) as text_stream:
timedelta_df.to_csv(text_stream, index=False, sep=self._separator)
def write_heating_obj_to_binary_stream(
self, binary_stream: BinaryIO,
heating_obj_df: pd.DataFrame) -> None:
logger.debug(f"Storing heating object; len = {len(heating_obj_df)}")
heating_obj_df.to_csv(binary_stream,
index=False,
encoding=self._encoding,
sep=self._separator)
def __init__(self, filepath: str,
reader: AbstractSyncTempGraphReader) -> None:
self._filepath = filepath
self._reader = reader
logger.debug(f"Creating instance:"
f"filepath: {filepath}"
f"reader: {reader}")
def round_series(self, series: pd.Series) -> pd.Series:
logger.debug(f"Rounding series; len: {len(series)}")
multiplier = 10 ** self._decimals
rounded_series = series.copy()
rounded_series = rounded_series.abs() * multiplier + 0.5
rounded_series = np.floor(rounded_series) / multiplier
rounded_series = np.copysign(rounded_series, series)
return rounded_series
def read_timedelta_from_binary_stream(self,
binary_stream: BinaryIO
) -> pd.DataFrame:
logger.debug("Loading timedelta")
with io.TextIOWrapper(binary_stream, encoding=self._encoding) as text_stream:
timedelta_df = pd.read_csv(text_stream, sep=self._separator)
timedelta_df = self._convert_timedelta_from_seconds(timedelta_df)
return timedelta_df
def load_weather(self,
start_datetime: Optional[pd.Timestamp] = None,
end_datetime: Optional[pd.Timestamp] = None
) -> pd.DataFrame:
logger.debug("Loading weather")
weather_df = self._load_from_storage()
weather_df = self._filter_by_timestamp(end_datetime, start_datetime, weather_df)
return weather_df
def read_weather_from_binary_stream(self, binary_stream: BinaryIO) -> pd.DataFrame:
logger.debug("Loading weather")
weather_df = pd.read_csv(
binary_stream,
encoding=self._encoding,
sep=self._separator,
parse_dates=[column_names.TIMESTAMP]
)
return weather_df
def get_weather_start_end_timestamps(
self, control_action_timestamp: pd.Timestamp
) -> Tuple[pd.Timestamp, pd.Timestamp]:
logger.debug(
f"Get weather start end timestamps for control action timestamp {control_action_timestamp}"
)
min_timedelta = self._timedelta_df[column_names.AVG_TIMEDELTA].min()
max_timedelta = self._timedelta_df[column_names.AVG_TIMEDELTA].max()
return control_action_timestamp + min_timedelta, control_action_timestamp + max_timedelta
def filter_df_by_min_max_timestamp(
self, df: pd.DataFrame, min_timestamp: Union[pd.Timestamp, None],
max_timestamp: Union[pd.Timestamp, None]) -> pd.DataFrame:
logger.debug(f"Filter range: [{min_timestamp}, {max_timestamp})")
if min_timestamp is not None:
df = df[df[self._timestamp_column_name] >= min_timestamp]
if max_timestamp is not None:
df = df[df[self._timestamp_column_name] < max_timestamp]
df = df.copy()
return df
def __init__(
self,
filter_algorithm:
AbstractTimestampFilterAlgorithm = LeftClosedTimestampFilterAlgorithm(
)
) -> None:
self._storage = dataset_prototypes.HEATING_OBJ.copy()
self._filter_algorithm = filter_algorithm
logger.debug(f"Creating instance:"
f"filter algorithm: {filter_algorithm}")
def load_heating_obj(
self,
start_datetime: Optional[pd.Timestamp] = None,
end_datetime: Optional[pd.Timestamp] = None) -> pd.DataFrame:
logger.debug(f"Loading for {start_datetime},{end_datetime}")
heating_object_df = self._load_from_file()
heating_object_df = self._filter_by_timestamp(end_datetime,
heating_object_df,
start_datetime)
return heating_object_df
def read_heating_obj_from_binary_stream(self,
binary_stream: BinaryIO
) -> pd.DataFrame:
logger.debug("Loading heating object")
heating_obj_df = pd.read_csv(
binary_stream,
parse_dates=[column_names.TIMESTAMP],
sep=self._separator,
encoding=self._encoding
)
return heating_obj_df
def __init__(
self,
timestamp_round_algo: AbstractTimestampRoundAlgorithm,
interpolation_step: pd.Timedelta,
) -> None:
self._interpolation_step = interpolation_step
self._timestamp_round_algo = timestamp_round_algo
self._timestamp_column_name = column_names.TIMESTAMP
logger.debug(f"Creating instance:"
f"timestamp round algo: {timestamp_round_algo}"
f"interpolation step: {interpolation_step}")
def __init__(self,
x_round_step: float = 0.1,
min_lag: int = 1,
max_lag: int = 20) -> None:
self._x_round_step = x_round_step
self._min_lag = min_lag
self._max_lag = max_lag
logger.debug(f"Creating instance:"
f"x_round_step: {self._x_round_step}"
f"min_lag: {self._min_lag}"
f"max_lag: {self._max_lag}")
def __init__(self,
min_value: Optional[float] = 0,
max_value: Optional[float] = 100
) -> None:
self._min_value = min_value
self._max_value = max_value
logger.debug(
f"Creating instance"
f"min value is {min_value}"
f"max value is {max_value}"
)
def find_lag(self, x: np.ndarray, y: np.ndarray) -> int:
logger.debug("Lag calculation is requested")
x -= x.mean()
x /= x.std()
y -= y.mean()
y /= y.std()
corr = np.correlate(y, x)
lag = corr.argmax() + 1 - len(y)
return lag
def __init__(self,
temp_correlation_df: pd.DataFrame,
timedelta_df: pd.DataFrame,
objects_type: str = heating_object_types.APARTMENT_HOUSE,
circuit_type: str = circuit_types.HEATING) -> None:
self._temp_correlation_df = temp_correlation_df
self._timedelta_df = timedelta_df
self._objects_type = objects_type
self._circuit_type = circuit_type
logger.debug(f"Creating instance:"
f"objects type: {self._objects_type}"
f"circuit type: {self._circuit_type}")