How to answer questions in a conference talk that I should know the answer to, but don't?

It will depend on the research field but there are ways to get out these situations quite well, in most cases. I also don't see any problems in admitting you don't know the answer.

Usually you won't get questions exam like style. No one will ask you:

What's the formula for calculating XY?

Here some types of questions I've heard a lot before and some simple answers you can give without talking about the subject:

Have you tried to use this approach?

In that case you can answer with

No we have not but thank you for your input, we will look into this!

If the question is:

[Name] has used this approach, how does it differ from yours?

You can, again, simply answer with

I am not familiar with this work but I will look into that.

The following question is tricky:

I think there might be a problem with this part of your work, can you comment on this?

Here of course you somehow have to defend your own work and this is the part you should know. What I would do is ask the person who asked the question if I understood it right (a lot of the time that's a problem) and then if you still don't know the answer you can eitehr go with something like before

We haven't investigated this yet, we will have to look into this in detail, thank you for your comment.

or you shift the whole discussion to a more "private" setting:

This is an intersting point, maybe we could discuss this after this session? (in view of the time)

This is quite common to do and usually there's only time for a few short questions.

The same is true for a question like:

Why did you use this method?

where it's easy to shift the discussion to after the session.

Whatever you do, don't try to hide the fact that you don't know the answer with strange answers. At the last conference I attended a speaker answered every question with:

This question is spot on, but the project is still ongoing and there's a patent filed and I don't know if I can reveal that right now.

Which told anyone in the audience: that guy has no clue.

To add to an already good accepted answer, here's to maybe calm your nerves a little bit: The majority tend to go easy on young researchers giving their first talk. Few people attend conferences to ask gotcha' questions to Ph.D. students, and the few who do, are usually frowned upon by other researchers.

More directed to your question: Not knowing the answer up front is ok. I actually disagree with the answer above stating that no one asks questions exam-style, because the above mentioned few actually tend to do. I vividly remember having been asked such exam-style questions a few times, usually by older, Russian professors. If the situation should arise, know that both the chairperson and the audience will be on your side, if you simply say: 'I do not know the answer by heart, let us take this discussion afterwards.' No one wants to sit and hear a lengthy cross examination, they might have a question which is actually interesting, and there is usually some degree of time pressure.

Good luck with your first talk!