There are a myriad of possibilities with the situation you present. Quark answered the fundamental question about control. How you select the number of pumps is a question of optimization, and the answer depends on your preferences with the specific application.
Taken in context, cme's statement about recessed sumps also mentions a control valve on the sump outlet, and an oversized equalization line. The equalization line assumes multiple towers or cells, which is not clear to me. You mention 3 pumps, but you never stated whether there is more than one tower or cell.
At any rate, your original premise includes a potential discrepancy that cme was trying to address:
Design Flow - 300m3/h, Max Pump Flow (8 users) - 400m3/h.
So, if your max case occurs, the tower(s) may empty. This is because a cooling tower presents an open circuit to the flow. All head is lost at the point the water is distributed through the fill. Suction head to the pump is only atmospheric pressure plus the physical height of the sump. Accordingly, the fill and distribution basin will only support a maximum flow. A larger sump (recessed), a valve to keep additional back pressure on the sump, and an oversized equalization line (more volume), may all be attempts to ensure that the tower sump does not empty - in the isolated case where 8 users may exceed design flow. However, these controls may only address the transition. Your pump may start to cavitate from decreased suction pressure even if the tower(s) does not empty.
In any event, the first thing I would address is the discrepancy in your design flow (300 m3/hr) vs. the expected max flow rate (400 m3/hr). A better solution may be to pick a maximum design flow, and then limit the system to that flow. Otherwise, the deliterious effects mentioned above will over-complicate the controls (if they work at all).
To answer your question about sizing the pumps - it's an optimization preference:
1. If you have a load that varies most often between 1/3 to 2/3 max, then 2+1 pumps may be a good solution.
2. If you have a load that runs most often around 50% max, then a 1+1 arrangement may be best.
3. If you have a load that varies all the time from 0% to 100% max, then a single pump with a VFD may be the best solution.
4. You may still want a "+1" pump just for simple redundancy.
These questions can only be answered with your knowledge of how the system may run, suggest some solution scenarios, then compare them with cost. Converting 300m3/hr to gpm = 1321 gpm. (I don't live and breathe metric, so "gpm" gives me a better idea of size.) That is not an insignificant size. Fewer pumps - but with VFD's - may be the best at that size. Discriminating among the load with several towers or cells will depend on your operational judgment. You have such an obviously large diversity, I would be concerned with whether you should increase the size of the system to 16 users or more, then incorporate optimum sizing and VFD's to limit the energy usage of reduced flow.
Actually, the size of the system may warrant decoupling the tower from the distribution circuit. In that case, you may achieve a closer scenario to that described in your first description: Size the tower for 300 m3/hr (primary), then size a separate, decoupled circulation circuit (secondary) at the higher flow of 400 m3/hr. The tower will not be able to maintain temperature at the 400 m3/hr flow rate, but that may be acceptable for short periods. Since the tower will be decoupled, it will not empty or overflow. This scenario could also be fairly inexpensive by placing rudimentary, limited controls on the tower circuit and pump, then applying the VFD to the distribution circuit.
Forgive me if all of this sounds complicated, but that may really be the case. It sounds as if you have a system of decent size that could be expensive with first cost and long-term cost from energy and maintenance. Yet, it also has a significant load variance and diversity. That is when Engineers earn their salaries.
