如果您将n_iter增加到1000，它将工作！
输出为

"global minimum: x = 0.0000, f(x0) = 0.0000"

这是一个随机算法，在这种情况下，它需要一些更多的尝试，实际上使用

import numpy as np
from scipy.optimize import basinhopping

while True:
    n_iter = 850 

    @np.vectorize
    def f(x):
        return ( x * np.sin(x) + 2*x)**2

    x0 = -6
    minimizer_kwargs = {"method": "BFGS"}
    ret = basinhopping(f, x0, minimizer_kwargs=minimizer_kwargs, niter=n_iter)

    print("global minimum: x = %.4f, f(x0) = %.4f" % (ret.x, ret.fun))

印刷品

"""
global minimum: x = -7.7230, f(x0) = 60.6709
global minimum: x = -0.0000, f(x0) = 0.0000
global minimum: x = -0.0000, f(x0) = 0.0000
global minimum: x = 0.0000, f(x0) = 0.0000
global minimum: x = -0.0000, f(x0) = 0.0000
global minimum: x = 0.0000, f(x0) = 0.0000
global minimum: x = -0.0000, f(x0) = 0.0000
global minimum: x = -0.0000, f(x0) = 0.0000
global minimum: x = -0.0000, f(x0) = 0.0000
global minimum: x = -7.7230, f(x0) = 60.6709
...
"""

具有n_iter=850的算法并不总是找到全局最小值。