如果不太像0，您可以将以前的行的差异与组的累计总和进行比较，然后在列表理解中添加年份：

g = df['old_value'].diff().lt('0 min').cumsum()

df['new_value'] = [a + pd.DateOffset(years=y) for a, y in zip(df['old_value'], g)]
print (df)
   old_value  new_value
0 2022-11-10 2022-11-10
1 2022-12-11 2022-12-11
2 2022-01-10 2023-01-10
3 2022-01-20 2023-01-20
4 2022-01-01 2024-01-01

另一个想法是在lambda函数中使用GroupBy.transform添加每组的年数：

g = df['old_value'].diff().lt('0 min').cumsum()

f = lambda x: x + pd.DateOffset(years=x.name)
df['new_value'] = df.groupby(g)['old_value'].transform(f)
print (df)
   old_value  new_value
0 2022-11-10 2022-11-10
1 2022-12-11 2022-12-11
2 2022-01-10 2023-01-10
3 2022-01-20 2023-01-20
4 2022-01-01 2024-01-01