You are allowed to do whatever you like if you are not working to a particular style, the issue is whether your readers will be happy following you: the key is that looking up the citation in the reference list should be unproblematic.
The two citation keys you give: "Percival & Walden 1993" and "PW1993" are from two different citation styles and I've not seen them be used together: I think there is a risk of confusion. The short form alone works well enough, and you could consider using that, although it is less flexible than the longer form, that allows discussion of the author at the point of citation, e.g., In the study of Percival & Walden (1993), 140 subjects were ... A point about author-date: do ensure that the author and date occur together in the reference list; it makes life needlessly difficult for the reader if the date is buried somewhere in the heart of the reference entry.
If you really care about space, consider numerical referencing, where citations are just numbers in order of appearance. It's used in journals like Nature and Science that are very economical on space. It's not very friendly to readers when the reference list gets long, though, and the trend generally seems to be towards longer citation keys.
In the humanities, note-based referencing is the rule, where the full citation is given —perhaps as a footnote, perhaps with additional material about the citation such as a short quote— and then later citations will be in short form, consister of the author(s)' surname and no more than three words from the title identifying the work. So this, often called author-title referencing, is an example of a citation style that does shorten subsequent citations, but it is still less economical than the author-date system.