To begin with, don't use LMILAB even in the linear case
http://users.isy.liu.se/johanl/yalmip/pmwiki.php?n=Solvers.LMILABSecondly, SET is obsolete
http://users.isy.liu.se/johanl/yalmip/pmwiki.php?n=Commands.SetThird, strict inequalities is not really supported
http://users.isy.liu.se/johanl/yalmip/pmwiki.php?n=Blog.Prepare-your-codeFourth, There is no equality in your model.
And finally, you will not be able to solve this with both z and P as decision variables, since that no longer leads to a linear expression. Instead, you have a bilinear matrix inequality, for which there really is no solver (besides penbmi, but most people don't have accessto that solver, and even if you do, it is not sure it will manage to solve the problem).
Easiest approach is to grid over z, i.e, solve a large number of problems for variaous fixed z, and pick the best solution. Alternatively, pick an initial guess on z and P, and then linearize, use trust-regions, path-following, alternating directions etc etc.