Genetic Algorithm Drifting to Worse Solutions With Each Generation

Hello all, So I am using a genetic algorithm for parameter estimation, and in the past 10 or so attempts a situation such as in the attached picture has been occurring. In that picture, the following was my code:

options = optimoptions('ga','PlotFcn',@gaplotbestf,'PopulationSize',600,'UseParallel',true,'Elitecount', 20,'Generations',50,'SelectionFcn',{@selectiontournament,2},'Display','iter'...
 ,'InitialPopulationMatrix',Sturt);

[x,fval,exitflag] = ga(@RED_vecfit1,Nvars,[],[],[],[],LB,UB,@RED_nonlinconstraint,options)

My number of variables is 692, and I have no linear constraints, and only an upper bound and lower bound (the first 690 variables are between 0 and 10, and the final two are just -Inf to Inf). My suspicion is that the feasible starting point I fed the algorithm was just located in a bad spot, so the algorithm drifted to a worse solution, because my nonlinear constraint is returning the the results I expect. However even when I do not start with an initial population, the algorithm always jumps up quickly in the beginning and levels off quickly, like an inverse of what it should be doing.

If anyone can tell me if I just chose the worst possible set of parameters for such a high-dimensional problem that would be great.

Also bonus thanks if someone can tell me why that same call (minus the ga options) would cause patternsearch to just shut my computer down? I've been wanting to try patternsearch but it either will run until forever or immediately kick back a barely changed solution set from my initial.

Much thanks! Jacob