For many years, chiller manufacturers encouraged HVAC engineers and cooling-plant operators to keep a constant water flow trough the chiller evaporator. The overriding concern was one of protection since reducing water flow faster than the chiller capacity control can accept could result in knocking it off line, requiring a manual reset to restart it. The most severe upset that a chiller could suffer operating at constant flow is load suddenly dropped in half from a fully loaded condition. Starting a second chiller in a two-chiller plant, where identical chillers operate in parallel, typically leads to being half the load to the active chiller.
In a constant-flow primary loop designed to produce CHWS at 45 F from 55 F of CHWR, a 50 % drop in load would cause that roughly half the primary flow of 45 F circulate through the common pipe to mix with the 55 F CHWR. The outcome is 50 F of CHWR at chillers inlet. Such CHWR entering the formerly fully loaded active chiller would initially be subjected to the full cooling capacity of the chiller until its controls could react to decrease capacity. The chiller would thus tend to drive the entering CHWR at 50 F down toward 40 F.
On the other hand, in a variable flow configuration the upset becomes far more severe: starting a second equal CHWS pump could cut the flow through the active chiller roughly in half. Then the chiller will try to apply its full output capacity to half the flow having its Delta T being doubled. It means to drive 55 F CHWR down to 35 F. That is pretty close to freezing. A simple low evaporator temperature sensor, would likely cause the chiller to trip off line to protect it from freezing. On the other hand, the constant-flow chiller whose living CHWS temperature dips only half as far, would probably remain on line.
However, modern chiller controls have incorporated sophisticated microprocessor-based control that allows the chiller to deal with such flow dips without tripping off line. Another reason stated by chillers manufacturers to keep the flow constant is to avoid the possibility of laminar flow through the evaporator. The VPF schemes proposed avoid this condition by two ways: with a bypass from CHWS to CHWR, as in primary-secondary schemes, opened via a signal from a flow meter or a pressure differential sensor across the evaporator. The other way is letting some terminals with three way control valves in order to ensure the flow will not drop below the minimum at any load condition. The second alternative becomes especially attractive in retrofitting projects because it allows some budget reductions although avoid reaching the entire energy saving potential of the VPF systems.
Designing a VPF chilled water plant that performs reliably at all load conditions requires careful attention to chiller selection. It is highly recommended a minimum evaporator-flow limit that is <= 60 % of the chiller's design flow rate and the greatest tolerance to large changes in flow rates.

The newest generation of unit controllers can reliably maintain the desired chilled water temperature with a flow-rate reduction of 67 %. It allows simplifying of system control by minimizing the need for “additional” demand limiting or isolation valve control as chillers come online. Chillers that are well-suited for variable primary flow can tolerate and respond to rapid flow-rate changes. Selecting chillers with these characteristics improves the likelihood of stable, uninterrupted operation. Another, less robust chiller controller permits flow-rate changes of less than 2 percent per minute and would need 30 minutes to adapt to a flow-rate reduction of 50 %. Fluctuations of 2 percent or more are typical, even during normal system operation. Attempting to limit flow-rate changes to this extent while starting or stopping a chiller is impractical, if not impossible.

However, with tandem disposition, flow through operating chillers will always change abruptly even though the pump has a VSD because no flow will go through the starting pump or chiller until the pressure at the pump discharge exceeds the backpressure on its check valve caused by pumps serving the operating chillers. At that point, the check valve will suddenly open and flow will abruptly change through the starting pump and chiller, causing an abrupt change in flow through the operating chillers.

Fig. Nº22. VPF scheme with separated pump control and chiller sequencing. Click on image to enlarge

The preferred scheme used in VPF systems separates pump control from chiller control (Fig. 22). It means the water distribution and circulation functions could be treated apart. A first appealing shape of such scheme is the prospective of implementing a flow supply control based on only one VSD, which allows some initial budget reductions. However, an isolation valve must be installed on each chiller inlet to avoid the flow circulation through an off-line chiller and also to modulate the flow through both during sequencing.

The election of a determined scheme should be based on a careful analysis of such factors in order to reach the best balance between equipment cost and system performance.
In the context of HVAC design, decisions made to save money often involve a tradeoff between acquisition expense and operating cost. If you can realize savings on both fronts, so much the better. Perhaps this explains the increased interest in chilled water systems with a VPF scheme. A VPF design uses fewer pumps and fewer piping connections than primary–secondary systems, which means fewer electrical lines and a smaller footprint for the plant. These factors reduce the initial cost of the chilled water system. However, the savings may be partially offset by additional costs for flow-monitoring and bypass flow (bypass line and control valve). VPF designs may also require more programming for system control than other designs.
As for operating costs, how much a VPF design saves depends on the pressure drops and efficiency of the pumps. A VPF design displaces the small, inefficient, low-head primary pumps used in primary–secondary systems allowing selection of larger and far more efficient pumps that satisfied the pressure drops previously satisfied by the primary pumps.
The Air-Conditioning and Refrigeration Technology Institute (ARTI) has released a final report on a research project entitled "Variable Primary Flow Chilled Water Systems: Potential Benefits and Application Issues." The final report can be downloaded from the ARTI Web site at www.arti-21cr.org
To measure the energy use and economic benefits of variable primary flow chilled water systems, ARTI conducted an extensive study that compared variable primary flow chilled water system energy use with that of other common system types including: constant flow/primary-only chilled water systems; constant primary flow/variable secondary flow chilled water systems; and primary/secondary chilled water systems with a check valve installed in the decoupler.
According to the ARTI study results, primary-only chilled water systems reduced the total annual plant energy by 3 to 8 percent, first cost by 4 to 8 percent, and life cycle cost by 3 to 5 percent relative to conventional constant primary flow/variable secondary flow chilled water systems. Several factors significantly influenced energy savings and economic benefits of the variable primary flow system relative to other system alternatives. These included the number of chillers, climate, and chilled water temperature differential.
“In view of both the state-of-the-art review and parametric study results obtained in this project, it can be concluded that variable primary flow is a feasible and potentially beneficial approach to chilled water pumping system design,” according to the study. “However, the magnitude of energy and economic benefits varies considerably with the application and is obtained at the cost of more complex and possibly less stable system control. The literature on effective application of variable primary flow is growing and should promote its appropriate and effective use in the future.”