Well, the latter is just called act utilitarianism

Given that, under that system, you have a duty to do something, it is deontological.

Basically, you create a deontological ruleset which tries to predict in advance what maximises utility

It's not really predicting anything, not necessarily. Who or what evaluates the possible consequences of an action can be determined in a lot of ways that are more granular than just being on a per-rule basis.

It only approximates utilitarianism as I see it because once a deontological ruleset is laid out, you can't change it

The same applies to all the other codes of ethics, considering that they are just systems of logic.

If you then encounter an action which will have negative conquerors, but you should do according to your ruleset, you have to do it, or else you are just doing utilitarianism and calling it deontology

If the predictions regarding the consequences of a particular actions are reevaluated and it is no longer a good action under a given utilitarian code of ethics, it also becomes a non-good action under the equivalent deontological code of ethics/you no longer have a duty to perform it.

I think, you have a very narrow and naive view of deontology, which seems to often be instilled in students when Kantian deontology is taught to them using very primitive examples.

But also, I very much do posit that, as prescriptive systems, deontological ones, consequentialist ones, and deontological-consequentialist mixed ones can't really be distinguished in any significant manner. As such (provided that we are working with a utilitarian system that does not involve any elements of virtue-based codes of ethics), I can just say 'utilitarianism is just deontology in a trench coat'.

You can improve the approximation arbitrarily by making a richer and richer ruleset

I have provided a method for finding an equivalent deontological code of ethics that differs from the original one in terms of the cardinality of the set of rules by an addition of just one axiom.