为什么Pandas对行数不同的两个CSV计算时输出全表NAN值

问题根源 #

NaN是pandas索引对齐规则导致的：仅1行的static DataFrame只有索引0，和24行的dynamic做运算时，仅索引0的行能匹配到static的取值，其余索引无对应值就返回NaN。你推测的行数不一致是完全正确的。

解决方案 #

你想到的把static的取值存为常量变量的方案完全可行，也是最简洁易读的解决方案。只需在读取static后，将所有字段的取值提取为标量即可，后续计算用这些标量和dynamic的列运算，不会触发索引对齐逻辑，也就不会出现NaN。

修改后的代码参考：

#Step 1 - Reading the input file
static = pd.read_csv("Static_Data.csv", sep=';', na_values='(missing)')
dynamic = pd.read_csv("Dynamic_data.csv", sep=';', na_values='(missing)')

# 提取static的所有静态参数为标量
height = static['Height(m)'].iloc[0]
depth = static['Depth(m)'].iloc[0]
length = static['Length(m)'].iloc[0]
window_wall_ratio = static['Awindow/Awall'].iloc[0]
air_change_rate = static['Air Changes/hour (h^-1)'].iloc[0]
u_walls = static['Uwalls(W/m2K)'].iloc[0]
q_people = static['Qpeople(W)'].iloc[0]
solar_gain = static['Window Solar Gain'].iloc[0]

#Physical properties calculations
A_win = height * (depth + length)*2 * window_wall_ratio #Area of walls
A_wall = height * (depth + length)*2 * (1 - window_wall_ratio) * 1.225 #Area of windows
t_ext = 20
m_dot = air_change_rate * height * (depth * length)
c_p = 1.0021
u_wall = u_walls                                                                     
#Calculation of energy flows inside the house
Energy_flows = pd.DataFrame(columns=[])
Energy_flows['Q_exterior'] = u_walls * A_wall * (dynamic['Temp'] - t_ext)/1000
Energy_flows['Q_interior'] = Energy_flows['Q_exterior'] 
Energy_flows['Q_windows'] = u_walls * A_wall * ( dynamic['Temp'] - t_ext)/1000
Energy_flows['Q_ventil'] = c_p * m_dot * (t_ext - dynamic['Temp'])/3600
Energy_flows['Q_intern'] = q_people * dynamic['Occupation']/1000
Energy_flows['Q_solar'] = solar_gain * height * (depth + length)* window_wall_ratio * dynamic['Rad (W/m^2)']/1000
Energy_flows['Q_envelope'] = Energy_flows['Q_exterior'] + Energy_flows['Q_interior'] + Energy_flows['Q_windows']
Energy_flows['Q_heating/cool'] = Energy_flows['Q_envelope'] + Energy_flows['Q_intern']  + Energy_flows['Q_ventil'] + Energy_flows['Q_solar'] 

内容的提问来源于stack exchange，提问作者Ilaha