Can you clarify your question a bit more.  What do you mean by "lower bound"?  Are you talking about the objective value of the LP relaxation?  What do you mean by "impact"?  Providing information for these will help people likely answer this problem.

I have seen in some optimization problem(linear programming),that to strengthen linear programming relaxation and to have better lower bound ,add some valid inequalities to the model,without consideration redundant condition,is it neccessary to add some constraints like this?

It boils down to the problem of consideration and if your goal is to get a feasible solution to the IP after.  Sometimes yes, sometimes no.  It depends heavily on what the IP the relaxation is formed from is "up to".  Sometimes artificial constraints are included for very specific reasons in a LP relaxation of an IP.  Most times it is to ensure that an optimal solution to the LP relaxation can guarantee a feasible solution of the IP when rounded (may not be optimal).   If you give the LP relaxation a feasible or even optimal solution to the IP, it will be a feasible (may not be optimal) solution to the LP relaxation;  the other way around isn't true most of the time.  I can't just use a feasible solution of the LP relaxation to get a feasible solution to the IP.  Usually algorithms perform "rounding" techniques to get the feasible IP solution from the relaxation.  

Now back to the artificial constraints.  The artificial constraints when we talk about the IP mean nothing and do not "add" anything more to the model and its capacity to produce optimal solutions (unless it is proven to be the case and is non-trivial).  When we get to the LP relaxation, it can play a very important role.  For example, they could set certain variables to 0 that you absolutely do not want to have non-zero, because it may cause issues when you want to round to a feasible IP solution.  They also tend to give you properties to work with when you want to prove things about the LP relaxation when you want a feasible IP solution.

I am sure there will be somebody who can give you more information (as my answer is a bit vague as I think this concept can get very specific depending on the optimization problem in question).  For example, in scheduling a lot of the time, redundant constraints often are used to force the LP relaxation to not assign jobs that exceed durations that would normally not meet the given deadline for the IP (e.g., a feasible solution to the LP may assign some variable 0.3 to a job of duration of 10, when you want the deadline to be 2, which cannot be possible to do, but it may be possible in the LP relaxation to get a feasible solution that does this).

I hope this helps!