Conference abstracts are usually due many months before a conference. Hence, it is not imperative that you follow the abstract exactly when presenting your final project, since many people find that by the time they prepare, related but new material is of interest. An abstract should be your best guess of what you will do for your final project in the allowed space. It should also include some big picture ideas that discuss the importance and relevance of your final project content and place it into the bigger context of related fields.
Abstracts are usually a few paragraphs long, and sometimes have space limitations on the number of characters or words. Your abstract will be presented to the class orally, in 1-3 minutes. While the final project may well be different, the abstract should be at least loosely adhered to since otherwise attendees will be confused and sometimes angry that the final product was unrelated to what you said it would be. After all, attendees are giving their valuable time to you, and might instead have chosen a different poster instead.
The first step in writing an abstract is to choose a title. The title should be chosen carefully to summarize the content of your project. At a conference, some people may not read the abstract (it is sometimes published in a separate book from the schedule), and may use only the title in order to decide whether they will attend.
Here are examples of some of my abstracts.
Here is an example of an abstract by Robert Ghrist on Barcodes: The Persistent Topology of Data.