正如@MichaelSzczesny在一条评论中指出的那样，np.block最接近于matlab所做的事情。

import numpy as np

rows = 103
cols = 103

Q = np.ones((rows,cols)) * 1
S = np.ones((rows,cols)) * 2
P = np.ones((rows,cols)) * 3
H = np.ones((rows,cols)) * 4
G = np.ones((rows,cols)) * 5
E = np.ones((rows,cols)) * 6

U = np.block([
    [Q, S, P], 
    [S, H, G],
    [P, G, E]
    ])

print('U has dimensions', U.shape)
# U has dimensions (309, 309)

我认为您可以使用np.concatenate进行数组的多个串联。
示例：

import numpy as np

rows = 103
cols = 103

Q = np.array([[0]*cols]*rows)
S = np.array([[1]*cols]*rows)
P = np.array([[2]*cols]*rows)
H = np.array([[3]*cols]*rows)
G = np.array([[4]*cols]*rows)
E = np.array([[5]*cols]*rows)

U = np.concatenate((np.concatenate((Q,S,P)),np.concatenate((S,H,G)),np.concatenate((P,G,E))), 1)

print("Output dimensions: ", len(U[0]), " by ", len(U))