我找到了一种方法来做到这一点，但它明显慢了（我不在乎〈10000个时间戳，但对于数百万或更多的时间，我希望保持快速的时间）。
在Jupyter notebook中：

import datetime 
import pandas as pd

def iso_to_datetime64_1(iso):
    return np.datetime64(iso)

def iso_to_datetime64_2(iso):
    return np.datetime64(datetime.datetime.fromisoformat(iso)
                          .astimezone(datetime.timezone.utc)
                          .replace(tzinfo=None))

def iso_to_datetime64_3(iso):
    return pd.Timestamp(iso).to_datetime64()

test_time = '2022-05-01T00:00:00.123-07:00'
for f in [iso_to_datetime64_1, iso_to_datetime64_2, iso_to_datetime64_3]:
    print(f(test_time))
    %timeit f(test_time)

它打印出：

2022-05-01T07:00:00.123
574 ns ± 2.11 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)
2022-05-01T07:00:00.123000
3.32 µs ± 27.9 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)
2022-05-01T07:00:00.123000000
6.94 µs ± 44.1 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)