The elements you give describe an idealised orbit that does not exist in reality. Those numbers are parameters to an approximate model. Earth's closest distance to the sun is different each and every year, by a lot (about 20,000 km in fact).
Are there any exact data about Earth's orbit?
There are certainly far better models than the 6-parameter elliptical one. Your best bet for very accurate positioning is an ephemeris like the JPL DE or VSOP. These models provide very long series with literally thousands of terms, which you must compute at a given point in time to get the value of a parameter. They supply such series for a wide range of orbital parameters in various coordinate systems and reference frames.
In particular, VSOP87 claims an accuracy of around ± 4 km for Earth – and this is one of the older ones. A basic elliptic approximation cannot come close to this, no matter what parameters you choose. I happen to have an implementation lying around, so I ran some numbers spanning the years 2001-2010, using the "Barycentric rectangular variables J2000" series. The code computes X, Y and Z for both the Sun and the Earth and then finds the distance between them. The numbers are:
Observe how variable these are. A perfectly elliptical orbit would have the same min and max every year, but the solar system is quite far from that.
Having said this, the average min/max distance agrees with NASA's values as provided by John Rennie, and not Wikipedia. If you wish to use the elliptical model, those are the better numbers to go by.