www.un.org/Depts/DGACM/index_spanish.htm 拟议解决办法

grouped = (df.groupby(np.arange(len(df)) // 2)
             .agg({ timestamp :  first ,  current :  sum ,  apparent_power :  sum }))

grouped[ timestamp2 ] = grouped.groupby(level=0)[ timestamp ].transform( last )

grouped = (grouped[[ timestamp ,  timestamp2 ,  current ,  apparent_power ]]
             .rename(columns={ current :  total_current ,  apparent_power :  total_apparent_power })
             )

http://www.ohchr.org。

>>> grouped.columns
Index([ timestamp ,  timestamp2 ,  total_current ,  total_apparent_power ], dtype= object )

>>> grouped
                      timestamp  ... total_apparent_power
0 2024-02-01 00:00:00.944369920  ...                121.9
1 2024-02-01 00:00:01.945563136  ...                121.8
2 2024-02-01 00:00:02.944880896  ...                122.2
3 2024-02-01 00:00:03.944605184  ...                122.7
4 2024-02-01 00:00:04.944967936  ...                121.4
5 2024-02-01 00:00:05.944571904  ...                120.9
6 2024-02-01 00:00:06.945287936  ...                121.6
7 2024-02-01 00:00:07.944841984  ...                122.5
8 2024-02-01 00:00:08.945434112  ...                123.2
9 2024-02-01 00:00:09.944396032  ...                124.9

引自“无时能力”主张,另一种利用<条码>/在一行中减少该守则的方法:

grouped = (df.groupby(np.arange(len(df)) // 2)
             .agg(**{
            "timestamp_0": ("timestamp", "first"),
            "timestamp_1": ("timestamp", "last"),
            "total_current": ("current", "sum"),
            "total_apparent_power": ("apparent_power", "sum")
        })
             )

http://pandas.pydata.org/docs/ reference/api/pandas.Series.dt.hour.html 您需要的行动:

ts = pd.to_datetime(df["timestamp"])
grp = {"hour": ts.dt.hour, "second": ts.dt.second} # +more ?
# grp = {"time": ts.dt.floor("s").dt.time} # 3-components of time

#1 reduce the groups with more than two identical censors
uniq = (df.assign(**grp).sort_values([*grp, "sensor_id"]).loc[
        lambda x: ~x.duplicated(subset=[*grp, "sensor_id"], keep="last")])

#2 look for the previous cursors to complete the groups having a single one
miss = df.groupby(list(grp.values()))["sensor_id"].transform("nunique").eq(1)
exte = (uniq.mask(miss | uniq["sensor_id"].eq(uniq["sensor_id"].shift(-1)))
           .ffill().loc[miss].convert_dtypes())#.assign(timestamp=None))

#3 swap the censors in the groups with a leading censor-1
cln = (pd.concat([uniq, exte if not exte.empty else None])
         .sort_values(["sensor_id", *grp]))
out = (
        cln.astype({"line no": "str"})
        .groupby(list(grp.values()))
            .agg(**{
                # feel free to comment (#) any agg
            "lines_joined": ("line no", " & ".join),
            "sensor_0": ("sensor_id", "first"),
            "sensor_1": ("sensor_id", "last"),
            "timestamp_0": ("timestamp", "first"),
            "timestamp_1": ("timestamp", "last"),
            "total_current": ("current", "sum"),
            "total_apparent_power": ("apparent_power", "sum")
        })
     .reset_index(names=[*grp]) # to show up the groups
    #.reset_index(drop=True) # or just drop them if optional
)