Chiller selection and control strategies, particularly in California, with its volatile deregulated electricity marketplace require a more proactive approach when making choices among multiple chillers. Functionally integrated hybrid chiller plants with or without thermal storage can offer significant benefits for our clients when a real-time pricing (RTP) approach - as opposed to a more traditional life-cycle cost analysis - is employed as an operational control strategy.

One way of dealing with the pricing uncertainty in California is to use RTP as the integral driver in chiller sequencing and control. In this way, building operators can meet foreseeable needs for flexibility in the purchase of electricity on a real-time basis, and natural gas at released capacity prices.

Managing Uncertainty

Only by taking full advantage of a more responsive control system integration can HVAC designers and owners evaluate the availability - and profitability - of thermal energy at their sites. The solution lies in an integrated system design, which combines a variety of chiller types, each designated for a specific purpose. Taken together, they result in an overall chiller plant that takes full advantage of the unique characteristics of each separate chiller for the flexible exchange of chilled water among chillers on a real-time basis, thus reducing overall energy consumption and related emissions.

In the real world, the major factor in chiller energy consumption is efficiency at off-design conditions, where chillers typically spend 99% of their operating hours. Energy simulations of the conventional type often do not allow one accurately to determine a prescribed sequence of operation for staging chillers in hybrid plants if based on the stable or predictable utility rates. In recent years, California has experienced substantial fluctuation in both gas and electric prices.

In this article, we will explore two alternative methods of sequencing chillers in a gas-electric hybrid chiller plant using both projected (i.e., simplified bin method) and real-time utility rates (RTP algorithm method) to automatically control individual chiller sequencing for cost-effective operation.

Method 1 - Simplified Approach

One must first examine the relevant building loads, electric rates, and their relationship to the latest published weather data given by ASHRAE to understand the benefit of employing bin weather data vs. conventional computer simulation methods on sequencing individual chillers in multiple chiller plants.

Although the assumption of a straight-line relationship between building loads and weather data is not always accurate for some industrial and commercial cooling applications, it can be used to represent a simplified approach for purposes of the comparative chiller sequencing case studies that follow.

Fortunately, ASHRAE has already organized operating hours for both domestic and international microclimates into five-degree temperature bins. Chiller operating costs, therefore, can be analyzed by estimating the load in each of the ASHRAE temperature bins and assigning a cost to the power sold during each hour in that bin and estimated hourly gas/electric rate schedules.

The first step in this simplified analysis is to obtain bin weather data for the given city or area to determine the total number of (annual) hours spent in each temperature bin. The second step is to determine typical entering condenser water temperatures (ECWTs) available for each of the listed bins.

For purposes of overall air conditioning analysis in the three chiller case studies that follow, it is also assumed that airside economizers are used below 55 degrees F outdoor db temperature to satisfy the cooling load.

While most chiller plant operating hours fall within lower loads and lower electricity costs, there are significant hours of operation when electric prices are high. Thus, the conventional wisdom is that in order for facilities to benefit economically in a deregulated environment, electric chillers should operate primarily during low-load, low-cost periods of operation, while gas chillers run during high-load, high electric demand cost hours.

Typical Application

First, let's take a look at an electric-only multiple chiller plant in comparison with two comparable but different alternative fuel plants. Let us assume, for example, that the base case chiller plant consists of two nominal 500-ton electric centrifugal chillers and that the design day load is 800 tons. When applying the ASHRAE data methodology given above, one arrives at an annual operating cost of $95,799 and equipment cost of $252,000 as indicated in Table 1.

Next, compare the listed installed equipment and annual operating costs with those for the following all-gas or hybrid alternative-drive chiller plants, respectively, comprising Option 1: two nominal, two-stage direct-fired absorption chillers; or Option 2: same as Option 1 except that one of the two-stage absorption chiller is replaced by a nominal 500-ton, electric-driven centrifugal chiller.

Assume a gas price of $0.35/therm for the above all-gas and hybrid chiller plants. Now by ratioing the installed equipment cost differential for each of the above referenced Option 1 and 2 chiller alternates to the estimated annual operating-cost differential above base-case electric centrifugal chiller plants (Table 1), one can establish a simple payback for each option, as will be shown below.

Additionally, at the preliminary design development/schematic phase of a project, equipment costs (not total installed costs) can be used as a first approximation. Although the latter alternative absorption units require exhaust systems, steam piping, or larger cooling towers in most early-stage analysis, these details, while needed subsequently for a more accurate comparative analysis, will often not be the determining factor in establishing the proper chiller operating mix for a minimum annual owning and operating cost estimate.

However, a maintenance premium should be included in the projected operating cost of the gas engine-driven centrifugal units where used, since it can have greater cost impact and requires an annual run-time estimate.

Referring next to Table 2, notice that the operating cost of the Option 1, dual, nominal, 500-ton, two-stage direct-fired absorption chillers are represented as $18,843/year less than the electric-only chiller plant. With the price premium for these chillers at $177,000, the result is a simple payback of 9-plus years.

For example, for the case of the hybrid chiller plant incorporating both a nominal 500-ton centrifugal and a nominal 500-ton, two-stage absorption chiller, the simple payback drops significantly to 2.48 years (Table 3). Therefore, the latter hybrid plant would appear to be a better choice than either the all-gas or all-electric chiller plant alternatives described above.

By definition, the traditional hybrid plants incorporate at least two chillers, comprising an electric reciprocating, centrifugal, or screw chiller, in combination with either an absorption or gas engine-driven chiller equipped with any one of the above referenced compressor types.

In summary, conventional industry practice has been to adopt an operating scheme that employs a base-loaded alternative gas or gas drive chiller during hours of high electric cost, with the electric chiller handling only the remaining load.

Method 2 - Use of RTP Chiller Control Algorithms

RTP-based chiller sequencing can replace the more traditional off-peak, mid-peak, and on-peak electric rate schedules. Alternative-drive chillers can operate during all hours when the hourly electricity-to-gas-cost portion of the total cost of cooling delivered (adjusted for associated parasitic loads such as tower fans, pumps, etc.) is high, and therefore, can be switched to operate only electric chillers when hourly electricity to gas cost is low.

A further benefit of RTP-based chiller sequencing is that building operating schedules can be made to change daily in response to real-time electric and gas rates.

For RTP-based operations, the chiller operating strategy now becomes baseload - that combination of gas and/or electric chillers that projects the lowest operating costs at any given hour. In this way, BASs that monitor the hourly electric and gas rates can determine the most cost-effective operating chiller sequence to help building owners operate plants at the lowest possible cost.

This case history of a large chiller plant, originally designed in 1986 to serve a major hotel project, will demonstrate how Method 2 works. The chiller plant was initially comprised of two nominal, 1,000-ton, electrical centrifugal chillers and located near Palm Springs, CA. It was subsequently modified years later to incorporate two additional nominal, 500-ton, two-stage absorption chillers following an extensive study of alternative energy conservation measures was undertaken to reduce then spiraling operating costs.

Adding the gas-fired, two-stage absorption chillers allowed the client to reduce peak (growth) demand and/or to permit transferring loads during on-peak and/or mid-peak periods to gas or vice versa back to electric during off-peak and/or mid-peak depending on crossover utility rate opportunities.

The two chillers also provided some redundancy benefits for the expanded chiller plant, which was then experiencing loads above 70% to 100% of peak for more than 50% of the summer months. The original chiller plant with two 1,000-ton, electric centrifugal chillers was configured as shown in Figure 1.

Not shown, however, are chiller isolation valves (one for each chiller) that were activated open whenever each respective chiller was operating. This required an automatic differential pressure CV to operate as needed to maintain flow balance through bypass (where shown) for the constant volume, primary chilled water pumps to serve all AHUs equipped with either two- or three-way valves at its respective cooling coils, etc.

Referring next to Figure 2, the subsequently revised chiller plant served an income-generating property located in the Southern California desert near Palm Springs, which would otherwise experience large losses in revenue should one of the chillers fail.

Our firm originally recommended adding two 500-ton, direct-fired absorption chillers since the natural gas chillers could produce each ton-hour at a lower cost during the electrical on-peak period at that point in time. The renovated chilled water distribution system configuration shown in Figure 2 now corresponds to a primary-secondary (with VFD) pump configuration. The differential pressure CV was also removed since it was no longer needed.

Energy simulations were performed using the utility rates when the plant was retrofitted. At that time, electric and gas rates were fairly stable. The latter energy simulation results were used to determine a prescribed sequence of operation for staging the chillers. These simulations indicated it was more cost-effective to operate the gas chillers as lead during the higher summer on-peak period only.

In the past year, for example, California has experienced large fluctuations in the price of electricity and gas. Accordingly, we were recently asked by our client to explore a means of using varying gas and electric rates as inputs for giving better real-time decision capability to the control system operating the chiller plant. This would allow the operator to provide more cost-effective utilization of the existing gas and electric chillers in order to maximize savings.

During the course of analyzing a better control strategy for the hybrid plant, we recognized that the following additional elements could be implemented into the control algorithm:

  • Electrical cost ($/kWh);
  • Gas cost ($/therm);
  • Measured plant load (tons);
  • Chiller parasitic loads (kW); and
  • Chiller efficiency data for part-load conditions (kW/ton or coefficient of performance [COP]).

We also recognized the utility rate schedules need to be updated in the control system as the rates changed. We then proceeded to develop a better control algorithm that uses this data.

The first step in developing the RTP algorithm referred to earlier, is to define the potential chiller operation combinations. Since the expanded chiller plant now had four chillers operating in parallel, (i.e., two 1,000-ton electric centrifugal chillers and two 500-ton direct-fired, double-effect gas absorption chillers) the range of foreseeable potential chiller combinations is shown in Table 4.

Once the various possible chiller operating combinations are defined, the chiller loading for each option can then be determined for a given plant load. Since the chillers in this plant are configured in parallel, all operating chillers will load to the same percentage of chiller load. For example, chiller combination #4 with a load of 1,200-tons would load each chiller to 80% (800 tons on the electric chiller and 400 tons on the gas chiller).

The electric centrifugal chiller efficiencies can be estimated from manufacturers' test results in the form of published ARI 550 NPLV values, which can then be easily curve fitted as shown in Figure 3.

Direct-fired gas absorption chiller efficiencies can also be estimated from manufacturers' tests yielding full-load and part-load COP values in the form of published ARI 560 IPLV values. For example, the electric centrifugal chiller manufacturers' data shown above in Figure 3 indicated a concave-shaped efficiency trend ranging from 0.534 kW/ton at 100% load, dropping to 0.416 kW/ton at 50% load at part load, and increasing to 0.565 kW/ton at 20% load.

Similarly, as can be seen from Figure 4, the two-stage absorption chillers are characterized by a convex-shaped trend ranging from a COP of 1.0 at 100% load, rising to a COP of 1.154 at 50% part load, and thereafter falling to a COP value of 1.136 at 25% part load as shown. Readily available software can be used to determine the polynomial equation and coefficients that provide the best curve-fit for the above-published ILPV or NLPL manufacturer's data.

Parasitic electrical loads can be accounted for as a fixed electrical load when the chiller is operating. The electric and gas chiller parasitic loads included associated primary chilled water pumps and condenser water pumps (not shown). Gas chiller parasitic loads also include the electrical load for the burner blower, solution pumps, and a purge pump. The comparative total parasitic loads are 79.3 kW for the 1,000-ton electric centrifugal and 62.5 kW for the 500-ton direct-fired absorption chiller.

The chiller plant control system program logic controller (PLC) can, however, be programmed to determine the optimum chiller combinations to use for a given load by estimating the operating cost/ton-hour for each feasible combination. The RTP algorithm, however, must have limits to recognize when a chiller combination is not feasible due to chiller loads being below the lower practical chiller load limits or when the load exceeds the capacity of the chiller combination.

To contrast both the conventional and alternative approaches, refer to Table 5.

Table 6 shows the results of the control algorithm using three different electrical rates with a constant $0.35/therm gas rate. The algorithm determines the chiller combination with the lowest predicted operating cost to automatically determine the best chiller sequencing strategy based on the utility cost data.

Note that at $0.05/kWh, it is more cost effective to use only the electric chillers when using a $0.35/therm gas rate. The electric parasitic loads on the gas absorbers accounted for 26% to 30% of the operating costs (cost/ton-hr), showing why parasitic loads should be considered in the control logic. The results also show a dramatic increase in operating costs for the plant at recent on-peak electric costs of $0.18/kWh when the plant has more than a 1,000-ton load due to the 1,000-ton capacity limit of the installed gas absorption chillers.

As can be seen from Table 6 above, one can better use current PLC control system technology to implement more complex calculation logic for optimizing hybrid plant operation.

This optimization allows plant operators to minimize operating costs many years after the design team has left the project. Chiller plant system designers must, however, be willing to detail and describe the necessary control logic in the sequence of operations in order for control system programmers to effectively implement them.

RTP electric costs reflect many hours of operation with very low electric rates, particularly when low ECWTs also are available as can be seen in Figure 5.

Effects of ECWT

One of the factors that can be used to hold down equipment cost is ECWT. For example, the nominal 500-ton electric centrifugal chiller can be selected for the off-design ECWT of 74 degree maximum, rather than the 85 degree maximum required in the all-electric or traditional hybrid plants. Thus, the same size unit can handle more tons with less energy, making nontraditional hybrids less expensive to operate while providing more redundancy in capacity.

Chiller operating costs at various load points and at corresponding ECWTs can also be used where real-time values are available to more accurately achieve proper chiller sequencing (vs. using NLPV or ILPV values).

For example, consider the hybrid plant illustrated in Table 1, an 850-ton, single-stage steam absorption chiller could handle the entire load above 79 degrees ECWT, when electric costs are highest, while a 650-ton electric centrifugal chiller could take over below a 79 degree ECWT. The operating cost would be $27,019 less than the equivalent all-electric plant, and the payback would be a very attractive 1.5 years.

Figure 5 was developed from curve-fitting actual data adjusted for the effects on COP with percent part load for varying EWCT values. If we now compare the latter values with earlier presented COP values with percent part load developed from published IPLV values for the same gas-fired absorption chiller, one will note some significant deviations in projected COP values, not only at ECWT design conditions as reflected in Figure 4, but at other part load conditions as well.

Summary and Conclusions

In summary, if we refer again to Figure 4, at a 50% part-load condition, the result would be an estimated COP of approximately 1.24 at an 85 degree ECWT. If we then compare the result extrapolated from chiller manufacturer-published ILPV values to the result obtained directly by curve fitting actual chiller manufacturers' test data already adjusted for the effect of EWCT, we will see a rather large variation in the result.

For example, at a 50% part-load condition, a design day ECWT of 85 degree, the COP is significantly less at 1.01 (by approximately 19%) than that obtained from earlier referenced ILPV values illustrated in Figure 4.

Furthermore, such ILPV values are published only at limited number of part-load conditions corresponding to 25%, 50%, 75%, and 100%, for example. If we allow the ECWT to drop to a permissible 67.5 degrees, the estimated COP rises to 1.31 or an increase of 30% above what is available at an 85 degree ECWT. In the absence of actual chiller manufacturers' performance data of the type illustrated in Figure 5, however, one can still use IPLV or NPLV data, but its use is advised only as a first approximation, recognizing that some underestimation of operating performance will result if EWCT is allowed to drop to levels approved by the respective gas or electric chiller manufacturer(s). ES

EDITOR'S NOTE: The images and tables associated with this article do not transfer to the Internet. To review the figures, please refer to the print version of this issue.