I think that for material balance that takes into account water injection you should consider some waterflood calculation method. The best is to start with BUckley-Leverett and then to dig into craigs waterflood method (I think all is very well described in craigs SPE monograph).
Average pressure in a formation for material balance you can extrapolate to the average formation depth, by using hydrostatic gradient. However, if you perform simple material balance calculation (for example, Muskat), and then use the same basic input data for reservoir model with for example 10x10x5 grid, you'll get different solutions, because numerical reservoir model uses different pressures at each layer