def setUp(self):
self.date_string = "2008-08-13"
self.date = pd.to_datetime(self.date_string)
self.start_df = format_nrel_dataframe("../data/2003_2016.csv")
self.lag_features = [
'Temperature', 'Clearsky DHI', 'Clearsky DNI', 'Clearsky GHI',
'Cloud Type', 'Dew Point', 'DHI', 'DNI', 'Fill Flag', 'GHI',
'Relative Humidity', 'Solar Zenith Angle', 'Surface Albedo',
'Pressure', 'Precipitable Water', 'Wind Direction', 'Wind Speed'
]
self.intra_df = create_lagged_features(self.start_df,
self.lag_features, 4, 30)
self.df = create_future_target(self.intra_df, 'DNI', 1, 30)
self.columns = [
'Year', 'DOY', 'Month', 'Hour', 'Minute', 'Date', 'final_date',
'DNI_T_minus_90', 'DNI_T_minus_60'
]
self.target = 'DNI_T_plus30'
self.num_units = 2
def predict_x_mins(shift, row_time_steps=30):
'''
Reads and formats NREL data, and creates a target column for modeling
that is the future values of DNI.
shift * row_time_steps = number of minutes target column will be in the
future.
Example:
[1]: df = predict_x_mins(shift=4, row_time_steps=30)
# 4 * 30 = 120. The target will be DNI 120 minutes into
# the future
Parameters:
----------
shift : (int)
The number of observations to shift backward
row_time_steps : (int)
The time difference (in minutes) from one observation to the next in
the DataFrame specified
Returns:
----------
df : (Pandas DataFrame)
DataFrame with new column added
'''
df = format_nrel_dataframe("../data/2003_2016.csv")
lag_features = [
'Temperature', 'Clearsky DHI', 'Clearsky DNI', 'Clearsky GHI',
'Cloud Type', 'Dew Point', 'DHI', 'DNI', 'Fill Flag', 'GHI',
'Relative Humidity', 'Solar Zenith Angle', 'Surface Albedo',
'Pressure', 'Precipitable Water', 'Wind Direction', 'Wind Speed'
]
df = create_lagged_features(df, lag_features, 4, 30)
df = create_future_target(df, 'DNI', shift, row_time_steps)
return df
import os
import sys
import pickle
parentPath = os.path.abspath("..")
if parentPath not in sys.path:
sys.path.insert(0, parentPath)
from modeling_base import *
from eda import format_nrel_dataframe
from mlp_base import *
if __name__ == "__main__":
df = format_nrel_dataframe("../../data/2003_2016.csv")
lag_features = [
'Temperature', 'Clearsky DHI', 'Clearsky DNI', 'Clearsky GHI',
'Cloud Type', 'Dew Point', 'DHI', 'DNI', 'Fill Flag', 'GHI',
'Relative Humidity', 'Solar Zenith Angle', 'Surface Albedo',
'Pressure', 'Precipitable Water', 'Wind Direction', 'Wind Speed'
]
df = create_lagged_features(df, lag_features, 4, 30)
df = create_future_target(df, 'DNI', 1, 30)
mlp = build_neural_network(len(df.columns) - 3, [10, 40])
same_week_results = {}
for i in range(1, 6):
key = f"{i}_week_same_mlp_results"
mlp_error_dict = iterative_nn_testing(mlp,
df,
'DNI_T_plus30',
