Delete null values ("nan") in an array

Simply assign the value you want by using the Boolean mask output numpy.isnan :

>>> import numpy as np

>>> a = np.array([[1, np.NaN, 2],
                  [3, 4, np.NaN],
                  [np.NaN, 9, 8]])

>>> a
array([[  1.,  nan,   2.],
       [  3.,   4.,  nan],
       [ nan,   9.,   8.]])

>>> a[np.isnan(a)] = 0      #<<<<<<<<<<<<<<<<<<
>>> a
array([[ 1.,  0.,  2.],
       [ 3.,  4.,  0.],
       [ 0.,  9.,  8.]])

In the supposition that your data comes from a list or any structure in which the NaN values are actually text strings ("NaN", "Nan", "nan", etc), you can use the argument dtype when building the array to define the type explicitly:

>>> import numpy as np

>>> l = [[1, "Nan", 2],
         [3, 4, "Nan"],
         ["Nan", 9, 8]]
>>> a = np.array(l, dtype = np.float)
>>> a[np.isnan(a)] = 0