您可以使用dateutil库来执行此操作：

pip install dateutil

它不是矢量化的，所以你必须在一个循环中完成，这比内置的panda函数要慢得多：

from dateutil.relativedelta import relativedelta

def get_duration(row):
    delta = relativedelta(row["End_Date"], row["Start_Date"])
    return f"{delta.years} years {delta.months} months {delta.days} days"

df.apply(get_duration, axis=1)

例如，使用dateutil中的库relativedelta时，格式可能会有所不同

import pandas as pd
from dateutil.relativedelta import relativedelta

df = pd.DataFrame({'Start_Date':pd.to_datetime(['2018-01-27','2018-08-31']),'End_Date':pd.to_datetime(['2019-03-13','2019-12-13'])})
cols = ['End_Date', 'Start_Date']
df['relativedelta'] = df[cols].apply(lambda x: relativedelta(x[0], x[1]), axis=1)
print(df)

输出：

Start_Date   End_Date                                 relativedelta
0 2018-01-27 2019-03-13  relativedelta(years=+1, months=+1, days=+14)
1 2018-08-31 2019-12-13  relativedelta(years=+1, months=+3, days=+13)

并将其设置为所需格式（'% Y：：% M：：% D'），您可以

df['diff'] = df['relativedelta'].apply(lambda x: f'{x.years} :: {x.months} :: {x.days}')
print(df)

产出

Start_Date   End_Date                                 relativedelta           diff 
0 2018-01-27 2019-03-13  relativedelta(years=+1, months=+1, days=+14)   1 :: 1 :: 14 
1 2018-08-31 2019-12-13  relativedelta(years=+1, months=+3, days=+13)   1 :: 3 :: 13