算出来的列类型是pd.Timedelta，实际上可以视为时间戳，可以通过其他方式转换为数值
import pandas as pd
import datetime
delay_time = datetime.datetime.strptime('2020-10-31', '%Y-%m-%d')
df =pd.read_excel("C:\\Users\\Administrator\\Downloads\\1604563522_368980.xlsx")
df['time'] = pd.to_datetime(df['time'], unit='d')
df.loc[df['标识']=='推迟','天数_py'] = delay_time + datetime.timedelta(days=1) - df.loc[df['标识']=='推迟','time']
df.loc[df['标识']!='推迟','天数_py'] = pd.Timedelta(seconds=0)
