Cool Rhythms

With greater industry interest in sustainable design and green buildings, there is a renewed attention to developing HVAC optimization strategies. These focus on specific operating characteristics of major chillers and related system parasitic loads—pumps, towers, fans—to improve estimation of predictable performance, and therefore, more accurately predict and determine multiple-chi...

By Milton Meckler, P.E., CPC, ASHRAE Fellow, President, Deisng Build Systems, Los Angeles January 1, 2004

With greater industry interest in sustainable design and green buildings, there is a renewed attention to developing HVAC optimization strategies. These focus on specific operating characteristics of major chillers and related system parasitic loads—pumps, towers, fans—to improve estimation of predictable performance, and therefore, more accurately predict and determine multiple-chiller plant annual energy usage and cost.

But to actually achieve more sustainable building operations may require some rethinking of current methods employed by HVAC system designers when planning the sequencing control of hybrid electric-gas multiple-chiller plants.

Based on stable utility rates in years past, HVAC designers have generally relied on commercially available computer energy simulations or temperature bin calculations to determine a prescribed sequence of operation for staging multiple chillers in central plants. The ASHRAE Handbook distinguishes forward models as those whose objective is “to predict the output variables of a specified model with known structure and known parameters when subject to specified input variables.” Inverse models, on the other hand, are defined in the scenario where “the input and output variables are known and measured and the objective is to determine a mathematical description of the system and to estimate the system parameters.”

Inverse modeling may better predict “whole building energy use.” But forward models provide a more accurate prediction of actual overall chiller plant energy use—or the desired output variable—for a specified configuration or mix of different or similar chillers of known structure and known parameters. This, of course, is subject to input variables, which are measured or specified within the same operating timeframe.

The facts

Chillers typically operate at design load for only a very short time, operating at less than capacity most of the year. As building load increases, however, if the chiller(s) operate in parallel and are set to deliver chilled water to the secondary circuit at a constant temperature, the entering return chilled-water temperature increases, then falls on a subsequent drop in building load by virtue of decoupling the primary and secondary chilled water circuits.

Development of chiller-sequencing control algorithms can be based on manufacturer-provided polynomial equations relating to the direct measurement of combined net effect of the following operative factors:

  • The actual cooling load seen each hour for each sequenced set of operating chiller(s).

  • The manufacturer’s rated kW/ton performance for electric chiller and/or the MBH/ton for both gas-engine driven and absorption chillers at above part-load condition adjusted for all associated system fan and pump loads.

  • The average entering condenser-water temperature (ECWT) entering the chiller(s) condenser over the same operating period.

  • Associated parasitic loads (e.g., pumps, fans, etc.) for all varying past load conditions as determined by performance interactive control strategies.

  • The average chilled-water return temperature over the same operating period put to a lesser degree than ECWT due to the decoupling effect of bypass.

  • Electrical cost ($/kWh).

  • Gas cost ($/Therm).

  • Recognition that utility rate schedules, if updated hourly with actual corresponding ECWT’s, could allow one to achieve more cost effective and predictable control of chiller sequencing using forward simulation models capable of bringing together all of the information collectively listed above.

Getting there

The first step requires the specification of a “real-time integrated” programmable logic controller (PLC) programmed with the capability of establishing all of the foreseeable chiller operational combinations as load both ramps up and down during normal diurnal and design-day periods.

To illustrate, let us assume a hypothetical multiple chiller plant comprising four chillers operating in parallel: two 1,000-ton electric centrifugal chillers and two 500-ton direct-fired double-effect gas absorption chillers. In doing the permutations, eight potential operating combinations can be programmed in the order of ascending building load (see table, p.59).

Since all chillers are configured in parallel, all operating chillers will load equally to the same percentage of chiller load if operative. For example, combination No.3, with a load of 1,000 tons, would load each chiller to 50% or 250 tons on each gas chiller.

Employing the polynomial data from the electric centrifugal chiller manufacturer, one can more accurately determine part-load efficiency employing the following generalized polynomial equations, applicable to both nominal 1,000-ton centrifugal, and nominal 500-ton, two-stage direct-fired absorption chillers referenced in the table:

Equation 1

Efficiency Electric Chiller = [(Efficiency FullLoad)/X]*(a + b*x + c*x2 +d*x3 +e*x4) – 0.5137X2 + 0.3882X3)* (f+g (85-ECWT)+h*(85-ECWT)2)

Inserting applicable coefficients “a” through “e,” obtained from chiller manufacturers, we arrive at the definitive polynomial characterizing the referenced centrifugal chiller. Notice also that Equation 1 also includes an adjustment for ECWT, which can also materially affect chiller performance, as will be seen when employing the equations.

Equation 2

Efficiency Electric Chiller = [(Efficiency FullLoad)/X]*(0.1111 + 1.0144X – 0.5137X2 + 0.3882X3)* (1+0.0248(85-ECWT) – 0.0004*(85-ECWT)2)

Similarly as described above, we can also arrive at a definitive polynomial equation characterizing above-referenced gas chiller using

Equation 3:

Efficiency Gas Chiller = [(Efficiency FullLoad)/X]*(0.1356 + 0.3944X + 4.0933X2 -4.4598X3 + 1.6248X4)* (1-0.00949(85-ECWT)+0.00014*(85-ECWT)2)

Equation 4

Operating Cost Electric Chiller = (Chiller kw/ton * Chiller Tons + Parasitic KW Load) * $/kWh

Equation 5

Operating Cost Gas Chiller = (Parasitic KW Load * $/kwh) + [COP*0.12 Therms/Ton-Hr* Chiller Tons * $/Therm]

The value of Efficiency(X, ECWT), determined by direct solution of Equation 2, can readily be seen to vary significantly depending upon actual part load and ECWT conditions. Furthermore, the locus of foreseeable upper boundary electric chiller part load and ECWT conditions can be developed from solution of Equation 2 at design (at 85°F ECWT) conditions and the corresponding lower boundary electric chiller part load and ECWT conditions also determined by direct solution of Equation 2, employing the manufacturer’s recommended minimum ECWT of 60°F for satisfactory chiller operations.

Additionally, by employing a three-way mixing valve on condenser water piping serving both gas chillers (not shown), which must be set to maintain the required chiller manufacturer’s recommended 68°F minimum delivered ECWT to each gas chiller, one can avoid adversely affecting simultaneous electric chiller operations, which can still operate with an ECWT, and can also drop to the lower 60°F level for lowest cost combined operations.

With respect to predictive operation of the above gas chillers, their combined comparable upper and lower boundary operating conditions can also be extracted using Equation 3 to compute their respective values of Efficiency(X, ECWT). In parallel fashion, the locus of foreseeable gas chiller part-load and ECWT conditions, with the upper boundary condition also reflecting an 85°F design ECWT and the lower boundary, can also be determined by direct solution of Equation 3. However, a higher manufacturer’s minimum coefficient of 67.5°F should be employed for maintaining satisfactory operations as described above.

How far to reduce ECWT?

This decision requires one to make a careful trade-off when estimating condenser pump and tower fan parasitic loads. Cycling constant-speed condenser water pumps has been recommended by some practitioners, but certain minimum flows must be maintained to keep tower contact surfaces wet whether fans run or not. Otherwise, scale can build over time. Others report that throttling back on variable-frequency drive tower fan drives to achieve a lower ECWT may offer greater power savings at the chiller compressor due to reduced refrigerant lift. Furthermore, parasitic loads are not necessarily constant over a wide range of part-load conditions.

Variable-speed tower fans provide the minimum power consumption when all cooling tower cells operate under all conditions except below manufacturer recommended minimum flow and above manufacturer prescribed minimum entering ECT. Regardless of tower fan unloading, tower cells should remain operational since resulting tower pressure drops (at the spray nozzles) will automatically reduce parasitic condenser pumping energy use at part load.

Tower airflow is directly proportional to fan speed, and therefore, its related power consumption varies approximately with the cube of the fan speed. As building load varies, however, experience suggests that when adding tower capacity the lowest speed fans should be the first of the parasitic loads to be increased and then maintained at the same speeds in all cells at any given time. While tower fan use affects overall energy cost, its effect on estimated hourly energy cost is not believed to significantly affect the outcomes of proposed multiple chiller combinations selected, and if operated as recommended in this paper, can more easily be accounted for in estimating associated parasitic effects. Part-load conditions employing the above-referenced proportional fan-speed energy use and other simplifying operating assumptions can also be used, assuming that the manufacturer’s minimum flow rates are always maintained.

Normally parasitic load-adjusted kW/ton performance is directly affected by the entering condenser water temperature (ECT) rising and falling, thereby creating a dependency. This effect is generally more pronounced with mechanically driven vapor-compression (VC) chillers than with comparable absorption chillers, which have a narrower ECT permissible range below 70°F. Therefore, when formulating cooling tower part-load control sequences, a similar adjustment to estimates of MBH/ton values should be made.

Pumping considerations

Primary pumping systems serving building loads directly also experience changes to return chilled-water flows and temperatures (assuming two position control values at all air-handling units), primarily as a result of variations in building load. Such variations also impact chiller kW/ton or MBH/ton values when operating to maintain a predetermined chilled-water supply temperature setpoint through activation of their respective chiller integral load/unload control sequences. Accordingly, primary chilled-water flow and return temperature variations can make predicting kW/ton and MBH/ton rates more problematic than modeling decoupled primary/secondary chilled water distribution systems.

Design conditions at which most chillers are selected usually require one to allow for maximum possible condenser water temperature. As a result, one usually provides more capacity than needed for most real-time operating periods. Additionally, whenever the entering ECWT falls below that maximum, actual chiller loading will decrease, further lowering capacity requirements.

One must, however, keep in mind that optimum energy efficiency is achieved when the major HVAC system energy consumer, i.e., the compressor—for vapor compression cycle machines—operates at its least overall $/ton-hour cost. For comparable absorption chillers, this also requires the high and low temperature solution heat exchangers (HX) to operate at their highest log mean temperature differential (LMTD) at design load.

Parasitic electrical loads associated with all respective gas and electric chiller pumps can be computed from shop drawings and then accounted for as a fixed electrical load in kW/ton to accrue when each chiller is operating. The electric and gas chiller parasitic loads include its associated primary and secondary chilled water pumps and condenser water pumps. Absorption chiller parasitic loads also include the unit burner blower, solution pumps and purge pump.

Based on the above, the hybrid, multiple-chiller plant PLC’s control algorithm could now be programmed to determine the optimum chiller combinations on an hourly basis for any given load condition, thereby more accurately determining the real-time operating cost/ton-hour for each feasible chiller combination. The provided algorithm limits are to recognize when a given chiller combination is appropriate for loads below manufacturer recommendations or when loads exceed recommended chiller combination capacity limits.

PLCs lead the way

By importing selected critical real-time and related operating parameters, one can better utilize the latest PLC chiller-sequencing control system technology for optimizing both hybrid and same-energy-source chillers in any given central chiller plant configuration. The approach or logic selected remains the sole responsibility of the design professional and is beyond the scope of our treatment here. Understanding what goes into modeling of interrelated variables is intended to better familiarize both HVAC designers and central plant operators with how those parameters interact through a better understanding of the polynomial equations that can be developed from data provided by chiller manufacturers. And by employing the latest PLC control features, one can achieve sustainable lower operating costs years after the design team has turned over the project. HVAC designers must communicate in their plans and specifications all of the necessary real-time information variables when establishing their desired PLC managed sequence of operation within both new and retrofitted multiple chiller plants, if we are to achieve truly sustainable buildings.

Listing of Available Chiller Operating Combinations

Chiller Combination Chillers Operating Capacity (tons)
1 (1) 500-ton gas 500
2 (1) 1,000-ton electric 1,000
3 (2) 500-ton gas 1,000
4 (1) 500-ton gas &(1) 1,000-ton electric 1,500
5 (2) 1,000-ton electric 2,000
6 (1) 1,000-ton electric & (2) 500-ton gas 2,000
7 (2) 1,000-ton electric & (1) 500-ton gas 2,500
8(2) 1,000-ton electric & (2) 500-ton gas 3,000