import numpy as np
import pandas as pd
[docs]
def conversionlts(blocks_per_day, chronological_timeslices, timeslices, chronological_sequence, representative_days, output_file):
# 3 VECTOR CREATION #
# Tile: [1, 2,..., timeslice, 1, 2, ..., timeslice ... ] repeated (Chronological timeslice) times
timeslice_vector = np.tile(np.arange(1, timeslices + 1), chronological_timeslices)
# Repeat: [1, 1, 1, 1, ... (timeslice) times, 2, 2, 2, 2, ... (timeslice) times, ... (Chronological timeslices), (Chronological timeslices), ... (timeslice) times]
chronological_time_vector = np.repeat(np.arange(1, chronological_timeslices + 1), timeslices)
# Vector (days_in_year * blocks_per_day * timeslices) with zeros
values_vector = np.zeros(chronological_timeslices * timeslices, dtype=int)
# Loops #
# Mapeia cada dia em chronological_sequence para o seu correspondente bloco de timeslices
for day_idx, rep_day in enumerate(chronological_sequence):
# O índice em representative_days determina o bloco de timeslices
timeslice_block_start = representative_days.index(rep_day) * blocks_per_day
# Define '1' nos timeslices correspondentes para o dia específico em chronological_time
for block in range(blocks_per_day):
timeslice_number = timeslice_block_start + block
index = (day_idx * blocks_per_day + block) * timeslices + timeslice_number
values_vector[index] = 1 # Marca com '1' a posição correta
# Cria o DataFrame com as colunas corretas
df = pd.DataFrame({
'TIMESLICE': timeslice_vector,
'TIMESLICECRO': chronological_time_vector,
'VALUE': values_vector
})
# Ordena o DataFrame baseado na coluna TIMESLICE_NUMBER e reordena o índice
df = df.sort_values(by=['TIMESLICE', 'TIMESLICECRO']).reset_index(drop=True)
# Save to CSV
df.to_csv(output_file, index=False)
return output_file
[docs]
def conversionlts_sparse(blocks_per_day, chronological_sequence, representative_days, output_file):
"""Write only the non-zero entries (VALUE == 1) of the Conversionlts mapping.
Equivalent input for intraday_kotzur(), which keeps only VALUE == 1 rows, but avoids the
dense (365*b) x (k*b) table (about 77 million rows at k365h1). This also removes the
dependency of the Kotzur reconstruction on running Model_Cluster first.
"""
rows = []
for day_idx, rep_day in enumerate(chronological_sequence):
start = representative_days.index(rep_day) * blocks_per_day
for block in range(blocks_per_day):
rows.append((start + block + 1, day_idx * blocks_per_day + block + 1, 1))
df = pd.DataFrame(rows, columns=['TIMESLICE', 'TIMESLICECRO', 'VALUE'])
df.to_csv(output_file, index=False)
return output_file
