Because you want to make plots, I'm going to assume that you have numerical values for each parameter in your model. For example, suppose we have:
Kh = 2; Tl = 1; Ti = 0.5; wn = 1; zetan = 0.7; tau = 0.2;
At this point, there are several options to form the transfer function F(s). Perhaps the easiest is to start with a product of low order terms:
>> F = Kh*tf([Tl 1],[Ti 1])*tf(1,[1/wn^2 2*zetan/wn 1])
F =
2 s + 2
-----------------------------
0.5 s^3 + 1.7 s^2 + 1.9 s + 1
Continuous-time transfer function.
Now the only part that is missing is the exp(-tau*s). That can be included by either setting the 'InputDelay' or 'OutputDelay' property of F or by mutiplying F by exp(-tau*s) explicilty. For example of the latter:
>> F = F*exp(-tau*tf('s'))
F =
2 s + 2
exp(-0.2*s) * -----------------------------
0.5 s^3 + 1.7 s^2 + 1.9 s + 1
Continuous-time transfer function.
If you'd rather see F expressed in terms of the variable p, you can do:
>> F.Variable='p'
F =
2 p + 2
exp(-0.2*p) * -----------------------------
0.5 p^3 + 1.7 p^2 + 1.9 p + 1
Continuous-time transfer function.
which is just for display purposes and doesn't change any mathematical properties of F. It's not clear what is meant by "plot the ... transfer function." Once F is defined, all sorts of different plots can be developed, most typically:
doc bode
doc step
doc impulse