How can pyomo & ipopt calculate different results each time

Question

I am using pyomo and ipopt to solve a nonlinear problem involving over 100 variables. I don't know too much math to know what the options in ipopt mean.

If my conditions are set relatively loosely, I find that every time my program is restarted, the results are far different from the last, and even if they are optimal, only some of them are satisfactory to me. I've tried to make the constraints a little bit tighter, but that also reduces the probability that I'm going to get a valid result.

So my approach is to loop solver.solve multiple times, storing the appropriate result into a file and then constrainting it. If I put model.cons = ConstraintList() into the loop, the program is running too slow; If I just put solver.solve(...) in the loop, as long as I don't run the program again, I'll get almost the same solution no matter how many times I loop through it.

I am not sure whether this problem is about pyomo or ipopt. I hope anyone can help me so that I can get different solutions in each cycle, thank you.

Here is my code:

# Create model is extremly slow, can't bear it in loop
model = ConcreteModel()

model.x = Var(range(n),range(n),[0,1],within=NonNegativeIntegers)
model.tolerance = Param(initialize=s_tolerance, mutable=True)

# Some constraints
model.cons = ConstraintList()
for j in range(n):
    model.cons.add(sum(model.x[i,j,0] for i in range(n)) == data[j][3])
    model.cons.add(sum(model.x[i,j,1] for i in range(n)) == data[j][4])

# Get almost the same solutions with these code
while True:
    # Not work for random solution
    seed = random.uniform(-0.5, 0.5)
    model.tolerance.set_value(seed + s_tolerance)

    solver = SolverFactory('ipopt', keepfiles=False)
    results = solver.solve(model, tee=False, symbolic_solver_labels=False, options={'max_iter':2000})
    
    if good_result(results):
        break