For me, a project is about telling a story. It has a beginning, a middle, and an end and the images work well together in my eyes. I do projects for me. Projects for other people (RPS Submissions, formal qualifications) tend to have set requirements. Make a decision as to why you're doing it, and who you're doing it for. For you is easier than for a competition or qualifying body. I have done projects for formal qualifications and some I hated, and some I really enjoyed. All the ones I did for me though, I enjoyed.
Find a theme. I find that specific themes work well to really focus the mind. "Landscapes" for example is so wide I'd get lost doing it. "Ordinary Landscapes Taken Around My Local Village/Town" is more specific and forces me to actually look at my surroundings. It also sets a 'brief' that I can follow. Just 'Landscapes' would have me continuing to wander aimlessly with the camera. A statement of intent before you get cracking is always a good thing.
Minimum images depends on how you're displaying it. Are you creating a book? A collage print? A web site, or just a page? I've found that when I get started, I get a feel for how big it's going to be. I also get a feel for the sort of images I want in it, and keep shooting until I get them. My "Playing With Knives" project required 12 square images to make a simple print on A3. My Chester Street Photography is still ongoing and has been for a few years (due to ill health/inability to travel/life). Phantoms is a real labour of love that I'm not willing to finish! So make your own rules.
I think image style should be down to the creator. If colour & b&w images works for you - do it. If you want to keep them all the same, that's fine too. I've seen successful RPS panels that were mixed, so clearly *they* think it's ok.
Overdone themes. I don't think themes are overdone. It's all about the *why* and the *what story are you telling*. If you have a reason for doing it, the theme then becomes yours and no two projects are the same. If you look at projects that aren't well defined though they kinda blur into insignificance - because they literally have no meaning.
My process is shoot - evaluate - re-define the project if neccessary - shoot more - evaluate re-define etc etc etc. After two or three sessions you'll begin to see a pattern develop and can focus down on what interests you.
Hope this helps. I'm still mulling over a doable statement of intent for my ARPS and struggling, so please take the above as guidance rather than firm advice.
Edit to add that there is a thread that kind of talks about this
here and has some useful guidance there too.