By CLT, the distribution can be approximated by Normal distribution as Normal(0.5*N*(a+b), 0.5*(b-a)*sqrt(N/3) ).
Your N>1000, that means the approximation should be very good.
Your only requested a 50 times simulation and "不需保留小数点后面的数", that means the simulation result will be very un-reliable.
What do you expect? Use such a meaningless simulation result to beat a beautiful theoretical approximation?