I’ll try to give you a long answer because I don’t have time for a short one.
Unless your “large system” is very unusual a stochastic analysis is unlikely to be very useful. Here are a few of the reasons:
1. In most public water systems, demands are NOT random but are somewhat predictable. Domestic demands are most often diurnal with peak flows occurring early in the morning and later around supper time. At night, these flows often drop to near zero.
2. Industrial demand will also follow a pattern closely related to each industry’s workshifts or similar scheduled operations.
3. Fireflows are usually the largest system demands. While fires may be random events, the critical locations of fire demands may be predicted based on Code required flow and such factors as whether the demand can occur at the top of a hill or alongside an undersized main where pressure and capacity are low.
4. Water quality analyses can be done using these same demand patterns, rather than assuming a completely random series of events.
5. If you don’t have demand patterns for your system you can either develop them from a pressure survey or you can infer them from similar nearby systems which do have such records.
6. Your time and effort are probably better spent calibrating your system model rather than attempting a random sampling approach which may be of academic interest....unless this IS homework, of course.
