The loadtxt function, by default, tries to convert everything to a float. It is getting confused by the text in the header and the datetime objects in the first column. You can tell it how to use the datetime objects, and you can also have it read the header. However, the simplest thing to do is to tell loadtxt to ignore the first row and the first column, like this:
data = np.loadtxt('data.csv',delimiter=',',usecols=range(1,7),skiprows=1)
It might also be convenient to unpack you data into separate variables, like this (I'm kinda guessing what some of the fields are):
day,hour,minute,Net_Shortwave_Avg,Net_Longwave_Avg,Net_Rad_Avg = np.loadtxt('data.csv',delimiter=',',usecols=range(1,7),skiprows=1,unpack=True)