You find the known structure on www.rcsb.org. You can search by protein name, but it is more useful to use "Advanced search" to search with the sequence of the protein https://www.rcsb.org/search/advanced.
Of course, if you go to Uniprot https://www.uniprot.org and search for your protein (human Beclin 1 https://www.uniprot.org/uniprotkb/Q14457/entry , human BCL2 https://www.uniprot.org/uniprotkb/P10415/entry), you also get a list of links to the known experimental structure, as well as an alphaFold model of the full-length protein: https://alphafold.ebi.ac.uk/entry/Q14457 , https://alphafold.ebi.ac.uk/entry/P10415 .
You have to decide for yourself which of the available structures best suit your needs (sequence coverage, experimental technique, resolution, isolated protein or structure of a complex). It can be very useful to superimpose and compare the different structures to get an idea of the structural flexibility of the binding partners. You may have to pre-process these structures before you can start docking (e.g decide how to deal with missing residues in the experimental structure / low confidence or unfolded segments in the models). Available complex structures and mutational data may also inform your simulation (e.g. conformational changes between free and complex structures, pathogenicity data for missense mutations.
There is no single "best structure", but rather some structures are better suited for the specific questions you want to answer than others.
For example, if you go to https://www.uniprot.org/uniprotkb/Q14457/entry (see my previous answer) and click on the "structure" link on the left, you get to a list of available structures. Click on the "PDB" in the leftmost column of the table to display that structure. Next come the name of the structure in the structure databases, the experimental technique, the resolution, the chain labels of beclin in the asymmetric unit of the crystal (having multiple letters means you have multiple instances of your protein in the asymmetric unit), the fraction of the sequence covered by that structure (e.g. residues 107-135 of 450 residues in the full-length protein in structure 2P1L), the links to the relevant entries in four different databases where you can get the structure, a download link for the structure and a link to the foldseek search engine to search for other proteins with similar fold (https://www.nature.com/articles/s41587-023-01773-0).
Look at the entries for beclin1, download the structures, look at them in a molecular viewer, and tell me which structure you would use and why, and I will then critique your results and give you my assessment. Do the same for BCL2. While I will have to do the work in parallel, I will not do the work for you.
I usually use RCSB-PDB ( https://www.rcsb.org ) to get the structures and use PyMOL ( https://www.pymol.org ) for a quick look and comparison of the structures. I cannot give you step-by step instructions if I do not even know which specific programs you are working with.
Especially if you are new in the field, it is important that you do the work yourself and learn how and why it is done in a certain way. Someone experienced in the field and doing research rather than run of the mill technicians work will know how important it is to look at the different structures and the biology of the system rather than just feeding arbitrary structures into a program. You usually learn at least as much or even more about the system from this pre-run analysis of the input structures as from the results of the docking analysis. This is also why it is important to be aware of the questions you want to answer with the analysis before you even start with the project.