我克服了在每次合并时使用 fillna(0) 的问题.

I overcame by using fillna(0) at each merge.

def write_dips(writer):
    df_dips = pd.read_excel(file_path, sheet_name='DipsSummary')
    df_sales = pd.read_excel(file_path, sheet_name='SaleSummary')
    df_delivery = pd.read_excel(file_path, sheet_name='DeliverySummary')
    df_mapping = pd.read_csv(mappingfilepath, delimiter=',', skiprows=[1])

df_dips = df_dips.merge(df_mapping, left_on='Site', right_on='SHIP TO NAME',how='left').fillna(0)
df_dips = df_dips.merge(df_sales, left_on ='IDASS ID', right_on='IdassId',how='left').fillna(0)
df_dips = df_dips.merge(df_delivery, left_on='SHIP-TO PARTY', right_on='Ship-To Party', how='left').fillna(0)