Optimized operation of combined chilled ceiling displacement ventilation system using genetic algorithm.
The combined chilled ceiling and displacement ventilation (CC/DV) system is a system that has potential for energy savings in spaces where the cooling load does not exceed 100 W/[m.sup.2] [Novoselac and Srebric (2002)]. The CC/DV air conditioning system involves the simultaneous operation of two subsystems with multiple variables to be controlled under dynamic loading unlike conventional systems. The setting of several operational parameters need to be simultaneously considered and controlled to insure meeting thermal comfort and IAQ inside the space and preventing condensation on the ceiling. The operational variables include: the supply conditions (flow rate, temperature and humidity) of the DV system and the chilled ceiling temperature. Energy consumption is strongly dependent on the settings of these variables and the selected control strategy of the system operation. The purpose of a control strategy of the CC/DV system operation is to provide acceptable thermal comfort and best indoor air quality (IAQ). A system approach to optimal control of conventional HVAC and building systems operation has been proved in several studies to result in energy savings of up to 24% in comparison to conventional localized control in spaces [House and Smith (1995), Zaher-uddine and Patel (1993), Nasif et al. (2004), and Jiang and Reddy (2007)]. Chilled ceiling and displacement ventilation system has features that allow exploitation of optimized control strategies under dynamic loading to improve the CC/DV system transient response and reduce energy use of the system.
Typical control strategies derived specifically for DV/ CC systems reported by Novoselac and Sebric (2002) are not totally optimized since they concentrate on controlling the temperature of the cooling panel. Researchers have tackled independently displacement ventilation system control strategies and hydronic radiant cooling ceiling system control strategies. Lau and Chen (2006) reported that displacement ventilation systems use in general more fan and reheat energy but less chiller energy compared to conventional systems. Seppanen et al. (1989) summarized control strategies of the DV/CC system including constant capacity displacement strategy and variable flow rate displacement strategy with terminal reheat. The constant capacity displacement system control strategy is based on following a thermal comfort criterion of maximum permissible temperature difference of 5 [degrees]C between room air at the height of 0.8 m from the floor and supply air temperature. Room air temperature increases in response to load increase, so the supply air temperature is decreased to remove the additional load until the maximum temperature difference is reached. The variable air displacement system control strategy is characterized by varying both the supply airflow rate set point and temperature set point. The control of radiant ceiling panel temperature is normally done by either using variable water flow / fixed supply water temperature system or fixed flow/ variable water temperature system [Strand and Baumgartner (2004), Conory and Mumma (2001)]. Conory and Mumma (2001) cooling loop control strategy was based on modulating chilled water flow rate to the ceiling panel in response to space thermostat. They also used two approaches to control condensation at start-up or extreme off-design conditions by either adjusting panel set point to be always a few degrees above measured space dew-point or by simply allowing the radiant panel to only operate when humidity control has been achieved. Niu et al. (1995) used, while assessing a CC/DV system performance, a control strategy by which the supply airflow rate and temperature are kept constant at design values of ventilation requirements and temperature of 15 [degrees]C, and further cooling requirement is met by activating the ceiling panel.
There are many possibilities of operational control strategies that we can associate with a selected design of CC and DV subsystem to meet peak load. There is a need to develop optimal control strategies in CC/DV system operation that minimize energy cost but still sustain thermal comfort and best indoor air quality. The objective of the work is to explore optimal operational parameters for different control strategies of the combined chilled ceiling displacement ventilation (CC/ DV) system subject to transient load and to recommend an optimal control strategy for minimizing energy consumption. The control strategies under consideration are: 1) varying chilled ceiling temperature (base strategy), 2) varying displacement ventilation supply conditions, and 3) varying both chilled ceiling temperature and supply air conditions (multiple control variables). In the process of developing optimized control, suitable simulation models are used for the ceiling panel and the fan coil unit (air handling unit), performance prediction of water chiller, fans, and pumps [Lu et al (2005)]. The transient plume-multi-layer model of the space cooled by the CC/DV system will be used to predict indoor thermal comfort and indoor air quality as a function of the load and system operational variables and will be coupled with models of primary equipment and air handling unit. A genetic algorithm optimization technique [Mitchell (1997), Nassif et al. (2005)] will be used to solve for the minimization problem of the energy consumption associated with the proposed strategies to predict optimized settings and performance. The integrated simulation/optimization performance model of the CC/DV system operation and predicted energy consumption will be validated experimentally for the base control strategy and the optimization problem will be simulated to a case study of 5m x 5m office space to predict savings for optimized system operation in Beirut Climate.
Space and System Mathematical Models
A system modeling approach is adopted in this work. The CC/DV system is initially verified to have adequate size at peak load using Ghali et al. (2007) design procedure and Ghaddar et al. (2007) design charts that account for the interaction of the chilled ceiling and the displacement ventilation to ensure thermal comfort (room air vertical temperature gradient is less than 2.5 K/m) and ensure acceptable stratification height at level higher than 1.1 m (the level of a seated person). It is important to account for the level of the stratification height in the design stage of the DV/CC system since it has a high impact on internal air quality, thermal comfort and energy consumption.
The simplified plume-multilayer model of the CC/DV conditioned space is implemented to predict the room transient response to changes in load and settings of the supply conditions and chilled ceiling temperature. The plume-multilayer model divides the space into four horizontal and equal lumped air layers with their adjacent wall sections. Each layer interacts with its sidewall segments to account for external load. The chilled ceiling temperature [T.sub.c] is assumed uniform. The supply fresh air enters into the floor air layer at flow rate [M.sub.a] and temperature [T.sub.s] and the exhaust air leaves at the upper air layer adjacent to the ceiling [T.sub.c]. The internal load is represented by a number of heat sources present in the floor-adjacent layer. For a given geometry, wall material, heat sources, and external solar and environment conditions for external walls, the simultaneous transient mass and energy balances are solved in each layer and wall section while estimating the plumes from sources and non-uniform walls to predict the transient vertical temperature distribution in the space and the corresponding stratification height and Predicted Mean Vote in the occupied zone. The complete plume-multilayer model formulation can be found in the work of Ayoub et al. (2006) and Ghali et al. (2007).
The space transient thermal model is coupled with models for air handling unit of the DV system, the hydronic chilled ceiling panel, primary equipment, and chilled water loop control. A schematic diagram of a typical CC/DV system components configuration is shown in Figure 1. The system consists of chiller unit, cooling coil, pump, supply and exhaust fans, chilled ceiling panel, reheat coil, chilled water mixers, three-way valves, piping, and duct work. The chilled water circuit followed the design used by Conroy and Mumma (2001). The chilled water supply temperature is maintained by using a three-way mixing valve and constant volume pump that serves the chiller. A fraction of the chilled water leaving the chiller enters the cooling-dehumidifying coil of outdoor air drawn into the system by supply fan while the rest of the chilled water flow is mixed with the water leaving the cooling coil to be delivered to the chilled ceiling circuit. The circuit design eliminates the need of heating the water entering the chilled ceiling. In situations when low humidity supply air is needed (high latent load), the circuit has to be modified to allow higher chilled water temperature to be supplied to the chilled ceiling different than that coming from the wet cooling coil. The analysis of the water circuit is general and can be applied to other water circuit configurations like Jeong et al. (2003) where the chilled water leaving the chiller circuit is split into the cooling coil loop and the ceiling panel loop.
[FIGURE 1 OMITTED]
The system model is performed by integrating the thermal models of the system components comprised the of hydronic system, the cooling coil and reheat system, and the chilled water network to desired air supply conditions and the chilled ceiling mean plate temperature to stratify space-system constraints. The chilled ceiling panel is modeled by implementing the model of Conroy and Mumma (2001). Their model uses fundamental heat transfer equations that govern the radiant cooling panel to express the mean panel temperature as a function of geometry, materials, flow rates, coolant temperature and space temperatures. The model of Conroy and Mumma (2001) assumes that the hydronic panels have a short time constant of less than three minutes and that the overall heat transfer coefficients are relatively constant in the ranges of temperatures at which the panel might operate (10-22[degrees]C).
The quasi-static model of Braun at al. (1989) is adopted for modeling the cooling and dehumidifying coil operation to predict outlet air temperature and humidity conditions for known inlet conditions. The Braun et al. (1989) model utilizes effectiveness relationships for heat and mass transfer using a lumped formulation approach. Their model prediction of air and water outlet conditions is known to work very well under steady state operation. The quasi-steady assumption for water and airflow means that the flows are boundary conditions for the other system component models that have transient changes. Zhou and Braun (2007) identified the conditions under which the quasi-static lumped model could simulate coil transients reasonably well. They have shown that the model developed by Braun et al. (1989) predicts well step changes in coil outlet parameters in response to step changes in water flow rate while the response of the quasistatic model is much faster than transient models for sinusoidal variations in water flow rate. In the present work, step changes in load and water flow rates will be considered permitting the use of the lumped and computationally effective Braun et al. (1989) quasistatics model. The cooling coil model is used to validate the coil selection in meeting the system peak cooling load, in addition to predicting the coil performance at partial load during transients.
The performance of the water chiller is quantified by its COP (coefficient of performance). The COP is determined at part load ratio PLR using Visual-DOE 4.0 Program Library (2005) for chillers of capacity less than 20 kW and 0.2 [less than or equal to] PLR [less than or equal to] 0.9 as
COP = 1/0.0101858 + 1.18131PLR-0.246748[PLR.sup.2] + 0.0555745[PLR.sup.3] (1)
The chiller power [J.sub.cuiier] is determined from the overall chiller energy balance on the chiller and is given by
[J.sub.chiller] = [m.sub.1] x [C.sub.p,water] x ([T.sub.10] - [T.sub.1])]/COP, (2)
where [m.sub.1] is the chilled water supply mass flow rate, [C.sub.p,water] is the water specific heat, and [T.sub.1] and [T.sub.10] are the temperatures of leaving and entering chilled water from and to the chiller, respectively. The reheat power rating is calculated for the electrical air heater resistance as follows:
[J.sub.reheat] = [[M.sub.a] x [C.sub.p,air] x ([T.sub.a,1] - [T.sub.a,coil exit])]/[eta] (3)
where [M.sub.a] is the supply air mass flow rate, [C.sub.p,air] is the air specific heat, [T.sub.a], coil exit] is the air temperature leaving the cooling coil, [T.sub.a1] is the supply air temperature entering the space and [eta] is the reheat coil efficiency. The fan power is assumed to vary as a cubic function of the volumetric airflow rate based on the reference energy consumption for a nominal airflow rate [Mehta and Thumann (1989)] as follows:
[J.sub.fan] = [J.sub.ref]([M.sub.a]/[M.sub.a,ref]).[sup.3] (4)
where [J.sub.fan] is the fan power consumption at the air volumetric flow rate [M.sub.a], [P.sub.ref] and [M.sub.a,ref] are the reference power and airflow rate, respectively. Since the pump in the system is a constant volume device, its energy consumption is easily calculated from manufacturer data and water circuit pressure drop.
CC/DV SYSTEM CONTROL STRATEGIES AND OPTIMIZATION OF ENERGY CONSUMPTION
The existence of two subsystems in the CC/DV system makes it harder to decide during transient change on the choice of parameters to control for an adopted control strategy. For the CC/DV system to meet the constraints of comfort while minimizing energy consumption, it is not obvious which parameters should be adjusted. Would controlling the chilled ceiling temperature alone be a sufficient strategy or changing the amount of supply air into the room, or induce both changes simultaneously? Any strategy could meet the system constraints of comfort and IAQ, but the best strategy would be the one that uses least energy. In this work, three control strategies are investigated for optimized operation and their implications on the operational cost over time are assessed in comparison to the cost arising from the use of a common base case control strategy. The control strategies under consideration are summarized below.
Strategy A (Base Strategy): Variable Chilled Ceiling Temperature Constant-Volume-Constant-Temperature DV System
The sensible cooling by the chilled ceiling system is enabled in response to space thermostat/PMV set point. The temperature or the flow rate of the panel supply water is modulated to meet the balance of the space sensible load constrained to be 1.5 [degrees]C higher than the space dew point [Novoselec and Srebric (2002), Mumma and Conory (2001)]. The cooling capacity of the constant-volume constant-temperature displacement ventilation system is set at its optimized value at the system peak cooling load. The reheat is used when subcooling of supply air is necessary for humidity control. The supply water temperature from the chiller is allowed to vary from 8 [degrees]C to 17[degrees]C.
Strategy B: Variable Displacement Ventilation System at Fixed Chilled Ceiling Temperature
Cooling is controlled by altering simultaneously the DV system supply flow rate and supply temperature while ensuring that stratification height is above 1.1 m and that thermal comfort is met. The chilled ceiling temperature remains constant during the operation and is set at its optimized ceiling temperature value at the system peak cooling load.
Strategy C: Variable DV and Variable Chilled Ceiling Temperature
The displacement ventilation system as well as the chilled ceiling system are simultaneously controlled and are optimized to change in response to the space load. The design variables are defined as the parameters that must be optimized simultaneously so that the energy consumption is kept at minimum while respecting the thermal comfort guidelines. Seppanen et al. (1989) summarized control strategies of the DV/CC system including variable flow rate displacement strategy with terminal reheat and control of chilled water flow rate in sequential manner Molina et al. (2000) described an approach to implement a load management system with direct load control for residential HVAC applications using constrained multivariable predictive control algorithms. It is feasible include in the control algorithm, with real-time optimization, the CC/DV system constraints of comfort and indoor air quality and the HVAC load characteristics. This would be possible when system and space simulation models are computationally to interact with the predictive controller that can modify set points of the chilled water flow rate and supply air conditions simultaneously based on signals from the space about space load and comfort and IAQ level.
The optimal operational set points for the proposed control strategies are determined by formulating and solving an optimization problem using the genetic algorithm [Mitchell (1997), Wang and Jin (2000)]. Genetic algorithms are growing more popular and have been used successfully in more complex problems for optimization of online process control [Wang & Jin (2000)].
The optimization problem is multi-objective and targets to simultaneously achieve thermal comfort and indoor air quality while saving energy over the time interval of interest. The objective function is subject to constraints associated with the system physical model and the bounds on system variables. The objective function E represents the total energy consumption of the system during the optimization period.
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (5)
where [J.sub.chiller] is the chiller power, [J.sub.pump] is the pump power, [J.sub.fans] is the power consumption of the fans operating, and [J.sub.reheat] is the reheat rate. The various operating powers are expressed in terms of the design parameters of the CC/DV system through CC/DV system mass and energy balances and using models of the chiller performance, fans, pump, and reheat, the space plume multi-layer model [Ayoub et al. (2006)] as explained in the previous section, in addition to simulation models of dynamic losses associated with the chilled water circuit and air distribution system. Note that the constant volume pump can be dropped from the minimization process objective function since its energy consumption does not change during operation. The design variables that must be optimized simultaneously to keep the system operational cost at minimum over the 24 hour operation of the system while meeting thermal comfort and IAQ requirements. The design variables of the CC/DV system are:
* The supply air temperature [T.sub.s] after the reheat.
* The supply airflow rate into the room [m.sub.a].
* The chilled water temperature leaving the chiller [T.sub.1].
* The water mass flow rate into the system [m.sub.2].
* The water flow rate into the cooling coil [m.sub.3].
* The water flow rate supplied to the chilled ceiling [m.sub.6].
The design variables are to be optimized over 24 hours period with optimal setting determined over each prediction period. The objective function is subject to the following constraints:
* The ceiling temperature must be larger 1.5 [degrees]C above the dew point of the adjacent air zone.
* The predicted mean vote in the occupied zone "PMV" is within [+or-]0.5 for thermal comfort.
* The minimum permissible stratification height H in the room is limited by the height of a sitting or standing person in the space to breath within that zone. The value of H is assumed to be larger than 1.1 m.
* The vertical temperature gradient must satisfy ASHRAE comfort condition given by dT/dZ [less than or equal to]2.5 K/m, where Z is the vertical coordinate [ASHRAE (2005)].
* The water flow rate into the cooling coil < Water flow rate into the system ([m.sub.3] [less than or equal to][m.sub.1]).
* The circulated water flow into the CC < Supply water flow into the CC ([m.sub.6] [less than or equal to] [m.sub.2]).
The Optimization Tool
The optimal control strategy problem is solved using genetic algorithm which is able to identify the pay-off characteristics between daily energy cost and zone thermal discomfort and indoor air quality [Wright et al. (2002)].
Genetic algorithm is most efficient when the optimization problem is not smooth and uni-modal or when the cost function is noisy [Mitchell (1997)]. The interaction between the model simulation and the optimization tool (Genetic Algorithm) is a continuous procedure until finding the optimum result.
Our objective is to find the optimum set point for each design parameter within an appropriate prediction period over the desired total simulation time. The simulation is performed with a time step of one minute dictated by the space thermal model desired accuracy and the optimization process is applied for a prediction period of one hour in accordance with the dynamic load which is assumed to change on an hourly basis. The value of design parameters are fixed during a prediction period and may vary from one prediction period to another. The procedure proceeds as follows:
* Genetic Algorithm Generation Step: The genetic algorithm starts with random values of the six design parameters. For strategy A, the supply air temperature and mass flow rate are constant value. In strategy B, the water flow rate into the chilled ceiling is varied to maintain the constant chilled ceiling temperature. The algorithm interacts with the various system models and the space thermal model in the following steps:
* Cooling Coil Model: The cooling coil model uses the chilled water temperature value [T.sub.1], the water flow rate into the cooling coil [m.sub.3] and the airflow rate [M.sub.s], as well as the ambient dry and wet bulb temperatures as an input and solves for the outlet air temperature and humidity ratio, and the exit chilled water temperature.
* Chilled Water Circuit: Adiabatic mixing is applied to chilled water streams bypassing the cooling coil at [T.sub.1] and the stream leaving the cooling coil at [T.sub.4] to obtain the chilled water temperature entering the chilled ceiling [T.sub.5] and to obtain the chiller return temperature [T.sub.10].
* Chilled Ceiling Model. Using the chilled ceiling water inlet temperature [T.sub.6] and the mass flow rate of the water [m.sub.6] as well as the room air temperature in the zone adjacent to the ceiling, the mean plate temperature [T.sub.7] and the exit water temperature [T.sub.7] are predicted using an iterative procedure. The chilled ceiling model is solved as steady state model at each time step, but it is updated with data calculated transiently at each step such as the room air temperature adjacent to the ceiling. The piping circuit flow and energy equations are used to determine the water temperature returning to the chiller [T.sub.10]. For each simulation step (one minute) the cooling coil model and the chilled ceiling model as well as the chilled water circuit equations are solved as a steady state models to predict the main parameters needed for the plume multi-layer model ([T.sub.s], [M.sub.a], [w.sub.s], [T.sub.c]).
* Plume-Multilayer Model: The transient plume-multi-layer model receives input data on ([T.sub.s], [M.sub.a], [w.sub.s], [T.sub.c]) for each minute and uses an initial guess of the air and wall layers temperature distribution to simulate the transient thermal conditions inside the conditioned zone for the duration of the selected prediction period of the genetic algorithm The plume multi-layer model simulation is repeated several times (number of generations) until reaching the optimal design parameters for this prediction period.
* Closing step: At end of each prediction period, the energy consumption for the chiller, the reheat, and the fan, are integrated to obtain its value during that period. The genetic algorithm checks the total energy consumption for the system and the constraints from space response generated by the space model during this prediction period and a new seed of the design variables for the next generation is done until reaching the optimal design parameters for the total load period of 24 hours.
The plume multi-layer model revaluates the optimal design parameters for the current prediction period and the final conditions are taken as initial conditions for the next prediction period. The procedure is repeated for the next prediction period until the final simulation time is reached. The effect of the initial conditions used to start the iterations is eliminated by repeating the whole optimization process over several simulation periods to reach steady periodic solution in the space thermal model and a steady value of energy use for the simulation period. Periodic load over a period of 24 hours is common in office space and commercial buildings. Simulations were normally performed at prediction period of 60 minutes.
EXPERIMENTAL VALIDATION OF THE CC/DV INTEGRATED SYSTEM MODELS
Although each component model in the CC/DV system has followed a known and validated model in literature, but it is necessary to test the validity of the integrated CC/DV system model response of the combined models of the space, cooling coil, chilled ceiling panel, and chilled water circuit. We need to assess accuracy of predicted performance parameters (E, dt/ dZ, and H) considering the inherent assumptions of quasi-static response of the chilled ceiling and cooling coil models and make sure that comfort and IAQ constraints are met during optimized operation.
The objective of the experiment is to compare the experimentally measured and predicted energy consumption at set optimized CC/DV design parameters generated using our optimization tool for control strategy A and verify that thermal comfort and stratification height constraints are met at all times in the space during the system operation under transient space load. The CC/DV experimental setup facility at the American University of Beirut (AUB) will be used to conduct validation experiments.
Experimental Setup and Measurements
The experimental setup includes a climatic chamber conditioned by chilled ceiling and displacement ventilation system. Figure 2 shows a schematic of (a) the CC/DV system and chilled water circuit and (b) the chilled ceiling panel. The CC panel is composed of three copper tube and plate panels. The chilled water is supplied from a 14.65 KW chiller with a constant flow pump. The supply air fan is a three speed setting fan that permits a supply flow rate at 0.05, 0.12 [m.sub.3]/s to 0.23 [m.sub.3]/s (250 cfm to 500 cfm). The cooling coil has a capacity of handling 4.5 kW sensible and 3.75 kW latent loads with a reheat capacity up to 2 kW. Since we are using the same experimental facility described in the work of Ghaddar et al. (2007), the details of the room dimensions and complete specifications of the system components are not provided here. The transient internal load is generated using two cylindrical (0.3 m in diameter) electrical heat sources provided with a proportional accurate controller that permits programming the resistors power according to the preset dynamic schedule shown in Figure 3a. The peak load was not allowed to exceed 100 W/[m.sub.2]to be within the operable range of the combined CC/DV system. The response time for the heat source control was less than 5 seconds. The load variation cycle was set to eight hours repeated over 48 hours measurement period to insure that steady periodic conditions are attained. The climatic chamber is located inside the lab facility with controlled indoor environment, thus eliminating external load variation.
[FIGURE 2 OMITTED]
[FIGURE 3 OMITTED]
The experimental setup is equipped with automated measurement and data acquisition systems to be able to determine room performance parameters of comfort (dT/dZ), IAQ (H) and energy consumption (E). The data acquisition system is capable of recording data every second on the vertical temperature distribution of air column at six locations in the room using 10 thermocouples at each location and the power consumption of the chiller, reheat resistance, and fans over the period of the system operation. The measured temperature distribution is used to determine the vertical temperature gradient. The power consumption is measured using a power analyzer C.A 8334B-QUALISTAR manufactured by Chavin Arnoux which provided testing characteristics and complete three-phase analysis of voltage, current, and power/energy. In addition, a hotwire IFA-300 anemometry system composed of 3-component velocity sensors and a transverse xyz table for remote sensor position control is used to provide measurements on the local velocities of air columns spanning several positions within the middle region of the room in three locations at incremental heights from Z = 0.55 m from the floor to Z = 1.3 m to experimentally evaluate the stratification height in the room. The specification of the various sensors used in the experiment can be found in the work of Ghaddar et al. (2007). Control of experimental conditions is achievable using Lab-VIEW software to control cooling coil unit temperature, chilled ceiling temperature, and the air dew point in the room to prevent condensation from occurring at the cold ceiling.
The current experimental setup permits only the testing of Strategy A since the fan is not a variable speed fan. The model validation will be performed by applying the optimization tool developed in this work to the existing CC/DV system of the AUB facility under the known load profile shown in Figure 3a using strategy A of controlling chilled ceiling temperature to generate (predict) optimal set points of the system design parameters ([T.sub.c],[T.sub.s] and [M.sub.a]). The optimal setting is applied to the experimental system to compare the measured and predicted energy consumption at the optimized setting and verify the satisfaction of both comfort and IAQ by model and experiment. During the length of the experiments, the environmental psychrometric conditions outside the chamber were recorded at temperature of 24[degrees]C[+ or -]0.5[degrees]C and relative humidity at 45%[+ or -]5%. These values are used as outdoor conditions in the optimization tool.
The upper and lower bounds for the design variables of the experimental system are fed into the genetic algorithm optimization tool as given in Table 1. The genetic algorithm parameters were set at 60 individuals for each population with a 0.6 crossover value and 0.001 tolerance function as a stopping criterion. The prediction period for the optimization process was 60 minutes in accordance with the hourly change of the load profile. The resulting optimized hourly chilled ceiling temperature for the system was predicted. Note that the supply air temperature for strategy A was set to 23[degrees]C based on sensitivity analysis to the effect of selected supply air temperature on energy consumption at the peak load and these values remained constant for the duration of the experiments. The supply flow rate is kept constant at 200 cfm the minimum acceptable of peak value that maintains the minimum required stratification height of 1.1 m.
Table 1. The Upper and Lower Bounds for the Genetic Algorithm Parameter Lower Bound Upper Bound Water flow rate into the system [kg s] 0.06309 1.261 Water flow rate into the cooling coil [kg/s] 0.06309 0.820 Water flow rate into the chilled ceiling 0 0.1261 [kg s] Supply air temperature [[degrees]C] 17 23 Supply airflow rate [kg s] 0.1 0.25 Chilled water temperature [[degrees]C] 5 17
The schedule for the chilled ceiling temperature, supply chilled water temperature, supply air temperature, and dynamic load profile were manually fed to the Lab-VIEW controller based on the results of the optimization and system model. Since latent load was not present in the experiment, the optimized chilled water supply temperature was 17[degrees]C. The chilled ceiling temperature set point [T.sub.c] generated for the optimal setting in response to transient load is given in Figure 3b for control strategy A. This value is fed into the controller circuit of the CC system to adjust the circulated water flow rate into the chilled ceiling pipes to the imposed [T.sub.c]. Instantaneous measurements of the chilled ceiling temperature from the 20 thermocouple locations mounted on the copper ceiling were averaged over each hour and are also shown on Figure 3b. The maximum deviation of the chilled ceiling temperature in the experiment was [+ or -] 0.4[degrees]C in any given hour. This deviation is due to spatial non-uniformity in the chilled surface temperature.
Model Validation Results
Figure 4 shows the hourly variation of (a) the stratification height during the system operation under load profile (Figure 3a) and (b) the vertical temperature gradient as obtained from experimental measurements averaged every hour and as predicted by the simulation model at optimal setting of the chilled ceiling temperature under strategy A. At peak load, the stratification height was at its lowest value of 1.2 m (H > 1.1 m) while at zero load the stratification height was 1.4 m. The maximum error in the predicted and measured stratification height is less than 0.05 m while the maximum error between the experimentally determined and by simulation of the vertical temperature gradient response was [+ or -] 0.2[degrees]C.
[FIGURE 4 OMITTED]
The hourly-averaged power consumption values obtained from the instantaneous measured values logged every second and recorded at 1 minute intervals during the system operation were compared with the hourly power consumption values predicted by the model. Figure 5 shows the predicted power consumption of the system by the model and the measured power consumption. The integrated system model predictions of power consumption agreed well with experimentally measured values with error not exceeding 5%. The power consumption was mainly by the chiller as the cooling coil remained dry for the most part of the experiment.
[FIGURE 5 OMITTED]
CC/DV SYSTEM OPTIMIZATION MODEL APPLIED TO A TEST CASE IN BEIRUT CLIMATE
A test case is considered for the study of the implications of CC/DV system optimized control strategies on energy consumption. The test case is an office space of dimensions 5 m width x 5 length m x 3 m height. The floor and two side walls are considered internal partitions while two walls are assumed external having South and West orientations. It is assumed that all the walls have the same construction with an overall heat transfer coefficient U of 2.5 W /[m.sub.2]K that is typical for construction material in Lebanese buildings. The internal load dynamic component is based on the occupancy schedule shown in Figure 6 while the static component due to equipment and lights is assumed to be 600 W. The solar load on the south and west walls reach 280 W and 150 W respectively at the peak time when outdoor design temperature is 34 [degrees]C 14:00 hr. The hourly direct and diffuse solar radiation incident on the walls and the values of ambient temperature are derived directly from actual hourly measured weather data files of Beirut for a typical day of each month of the season when cooling is needed from June to September [Ghaddar and Bsat (1998)]. The peak load calculated for the test case caused by both internal and external loads was 93 W/[m.sub.2] which is within the operational range of the CC/DV system.
[FIGURE 6 OMITTED]
A design procedure based on Ghaddar et al. (2008) design charts was followed to size each of the two subsystems (chilled ceiling and displacement ventilation) and determine the chilled ceiling load removal ratio R to the total room load such that thermal comfort and stratification height are attained at the peak load [ASHRAE Project: 1438-RP]. The selected design parameters at peak load were at chilled ceiling temperature of 18 [degrees]C, supply air temperature of 18-19 [degrees]C, supply flow rate of 0.232 kg/s leading to stratification height of 1.26 m and CC load removal ratio of 40% of the total load. Hence the load removed by the displacement ventilation is 1.4 KW and the sensible load removed by the chilled ceiling is 0.93 KW. The next step is to select appropriate design parameters of the CC/ DV system primary components, the cooling coil, the chilled ceiling, and the chiller size.
The cooling coil can be selected using typical standard manufacturer data procedure to determine appropriate parameters for the cooling coil simulation model to predict the coil performance at partial load. The chilled ceiling is designed at peak load using the formulation of Conroy and Mumma (2001). The selected cooling coil and the chilled ceiling specifications and nominal design parameters are summarized in Table 2.
Table 2. The Cooling Coil and thhe Chilled Ceiling Specifications System Specifications Value Number of rows of tubes 4 Number of tubes per row 8 Number of feeds 3 Outside surface area 197 [m.sup.2] height 25.4 cm Cooling coil Size width 35.6 cm material copper Tube outside 12.7 mm diameter inside 11.7 mm diameter Panel material and 3 mm steel panel thickness Tubes material 12.7 mm copper tubing and size Chilled ceiling Panel area 4.5 m x 4.5 m system Number of passes 23 Spacing between 20 cm passes Total tube length 99 m
RESULTS AND DISCUSSION
The CC/DV system was simulated and the developed genetic algorithm optimization tool is used for comparing performance and energy consumption subject to the three proposed control strategies A, B, and C for a representative day of each of the Beirut summer months of June, July, August and September using the occupancy profile of Figure 6. The genetic algorithm searches for the optimum values of the design parameters within the appropriate ranges of design variables (see Table 1) and genetic algorithm parameters at maximum population size of 210 with a 0.6 crossover value, [10.sub.-6] tolerance function, and stall time limit of [10.sup.4]. A sensitivity analysis was performed to changes in the Genetic algorithm cross over and population size to ensure accuracy of obtained results. The prediction period for the optimization process was 60 minutes in accordance with the hourly change of the load profile. In this section, we will present the predicted system optimized operational parameters and space comfort condition and stratification height at various control strategies for the representative day of the month of August for comparison purposes. This will be followed by an assessment of energy consumption of the system under various control strategies for the summer months.
The optimized hourly optimal settings of the supply air temperature, supply airflow rate, and chilled ceiling temperature of the CC/DV test case system are given in Table 3 for strategies A, B, and C for the month of August. Figure 7 shows the hourly variation of the space predicted (a) stratification height and (b) Predicted Mean Vote (PMV) at optimized design conditions of strategies A, B, and C for the month of August. Note that the occupied zone is a high convective region with operative temperature close to the air temperature which means that the PMV is usually satisfied. The minimum cold ceiling is rarely below 16[degrees]C which does not cause uncomfortable plane temperature asymmetry. Figure 8 shows the total hourly power consumption (chiller, fan, and reheat power) in the month of August for the three optimized control strategies A, B, and C. The comfort and stratification height constraints are satisfied at all times. Strategy B of DV variables control and fixed chilled ceiling temperature resulted in higher stratification height than the optimized strategy A or C indicating more use of fan power (higher supply flow rate) to satisfy comfort condition. The hourly power consumption of optimized strategy C is the lowest (see Figure 8). Note that the supply air temperature for strategy A was set to 19[degrees]C based on sensitivity analysis to the effect of selected supply air temperature on energy consumption at the peak day and these values remained constant for the rest of the summer months. The energy consumption at other supply temperatures of 18[degrees]C, 20[degrees]C and 21[degrees]C is performed for the month of August. A lower supply temperature reduces the reheat energy but may affect comfort level. The total energy consumption increased from 557 MJ at [T.sub.s] = 19[degrees]C to 574 MJ at [T.sub.s] = 20[degrees]C showing less than 3% increase per degree C of supply air temperature. The supply flow rate of strategy A is kept constant at 0.2 kg/s representing the minimum acceptable of peak value that maintains the minimum required stratification height. In strategy B, the chilled ceiling temperature was contrained to fixed optimized value at 20[degrees]C that was determined from a sensitivity analysis done by incrementally changing [T.sub.c] values between 18[degrees]C and 22 [degrees]C in the peak load day. No temperature below 18 [degrees]C was considered for the chilled ceiling setting for strategy B to reduce added dehumidification load to prevent condensation on the ceiling. The total energy consumption during the peak day was 551.6 MJ, and 535 MJ, and 498 MJ for control strategies A, B, and C, respectively at optimized set conditions.
Table 3. Optimized Set Point Operational Parameters of the CC/DV Test Case For Strategies A, B, and C Optimized Strategy A Parameters Hour [T.sub.c] ([degrees]C) [T.sub.s] ([degrees]C) [M.sub.a] (kg/s) 0-1 20.58 19 0.2 1-2 20.58 19 0.2 2-3 20.58 19 0.2 3-4 20.58 19 0.2 4-5 20.58 19 0.2 5-6 20.58 19 0.2 6-7 20.58 19 0.2 7-8 20.58 19 0.2 8-9 20.32 19 0.2 9-10 20.25 19 0.2 10-11 20.15 19 0.2 11-12 20.10 19 0.2 12-13 20.00 19 0.2 13-14 19.92 19 0.2 14-15 19.86 19 0.2 15-16 19.81 19 0.2 16-17 19.50 19 0.2 17-18 19.32 19 0.2 18-19 19.67 19 0.2 19-20 19.52 19 0.2 20-21 20.10 19 0.2 21-22 20.23 19 0.2 22-23 20.39 19 0.2 23-24 20.58 19 0.2 Optimized Strategy B Parameters Hour [T.sub.c] ([degrees]C) [T.sub.s] ([degrees]C) [M.sub.a] (kg/s) 0-1 20.00 22.16 0.2101 1-2 20.00 22.16 0.2101 2-3 20.00 22.16 0.2101 3-4 20.00 22.16 0.2101 4-5 20.00 22.16 0.2101 5-6 20.00 22.16 0.2101 6-7 20.00 22.16 0.2101 7-8 20.00 22.16 0.2101 8-9 20.00 21.65 0.2151 9-10 20.00 21.42 0.2133 10-11 20.00 21.24 0.2142 11-12 20.00 21.35 0.2164 12-13 20.00 21.15 0.2181 13-14 20.00 21.04 0.2183 14-15 20.00 20-98 0.2191 15-16 20.00 20.75 0.2202 16-17 20.00 20.84 0.2194 17-18 20.00 20.92 0.2173 18-19 20.00 21.21 0.2164 19-20 20.00 21.36 0.2145 20-21 20.00 21.44 0.2131 21-22 20.00 22.02 0.2128 22-23 20.00 22.20 0.2102 23-24 20.00 22.16 0.2101 Optimized Strategy A Parameters Hour [T.sub.c] ([degrees]C) [T.sub.s] ([degrees]C) [M.sub.a] (kg/s) 0-1 21.5 20 0.19 1-2 21.5 20 0.19 2-3 21.5 20 0.19 3-4 21.5 20 0.19 4-5 21.5 20 0.19 5-6 21.5 20 0.19 6-7 21.5 20 0.19 7-8 21.5 20 0.19 8-9 21.1 19.7 0.195 9-10 20.8 19.5 0.192 10-11 20 19.2 0.196 11-12 19.9 19.1 0.205 12-13 19.8 19 0.205 13-14 19.65 18.85 0.206 14-15 19.5 18.9 0.21 15-16 19.2 18.5 0.22 16-17 19.45 18.4 0.215 17-18 19.2 18.7 0.212 18-19 19.5 19.2 0.213 19-20 20.2 19.5 0.202 20-21 20.5 19.8 0.208 21-22 20.7 19.9 0.2 22-23 21.2 19.9 0.195 23-24 21.5 20 0.19
[FIGURE 7 OMITTED]
[FIGURE 8 OMITTED]
The total energy consumption are calculated at predicted optimized values of control strategies A, B, and C for the months of June, July, August, and September. The simulations were performed for one representative day for each month taken at the mid of the month and the mean power consumption averaged over the 24 hours is calculated per unit floor area in W/[m.sup.2]. The total monthly energy consumption can be calculated by multiplying the daily energy use by the number of days of the month. Figure 9 shows the monthly mean power consumption per unit area floor area and as distributed by component (chiller, fans, and reheat) at the optimal operation of the three control strategies of the CC/DV system. Optimized operation of the CC/DV system on strategy C resulted in the lowest electrical mean power consumption. The calculated mean power consumption of strategy C was 192 W/[m.sup.2] for all the four months when compared to 222 W/[m.sup.2] for strategy A and 210 W/[m.sup.2] for strategy B, respectively. The % reduction in energy use is 13.4% with strategy C and 8% with strategy B when compared with base strategy A. The optimum control strategy C resulted in energy savings of 13%, 11.9%, 13.8% and 15.13% for the months of June, July, August and September respectively when compared to the Strategy A and 10.2%, 4%,8.7%, 10.3%, 7% and 7.66% when compared to the strategy B for the months of June, July, August and September.
[FIGURE 9 OMITTED]
When considering energy consumption distribution by component, it is clear that under the three control strategies, the reheat energy consumption is greater than the fan energy consumption in all the months except for the month of August. The largest difference between the reheat and fan energy consumption is noticed for the month of September with a fan energy consumption almost half of the reheat energy consumption. The chiller energy consumption is maximum in August. For strategy C, the chiller consumes 64% of the total energy consumption during the month of August while it has a mean value of 50% during the summer season while the reheat resistance consumes 17% of the total energy consumption during the month of August and the fan consumption is 19%. The mean reheat energy consumption over the summer compared to the total energy consumption is 27% while that of the fan is 23% for control strategy C.
The operation of the combined chilled ceiling and displacement ventilation system involves several variables to be controlled unlike conventional systems. Energy consumption is dependent on the settings of these variables that include supply airflow rate, temperature and humidity, and chilled ceiling temperature and depends on the selected control strategy for the system operation while maintaining thermal comfort and indoor air quality dictated by the stratification height in the space.
In this work, three control strategies based on control of single parameter (chilled ceiling temperature) or simultaneous control of multiple parameters (DV system flow rate and temperature or all the DV and CC systems parameters) were optimized using a multi-objective genetic algorithm for minimal energy consumption during operation of the same CC/DV system under the various strategies with 100% fresh air brought into the space by the DV system. Simulation models of various system components and integrated system operation and the optimization tool were validated by experimentation showing less than 5% error in predicting energy consumption.
The optimized control strategy C that involved varying all system design variables has been shown to consume the lowest energy when compared with optimized strategies when one or two variables are optimized during operation. Savings of up to 15% are realized with optimal strategy C as compared to the base strategy when only the chilled ceiling temperature is varied.
Future work need to explore building online supervisory controller for the CC/DV system operation and to consider multi-zone system with the use of heat recovery option between exhaust and incoming air and the use of desiccant dehumidification rather than electrical reheat to reduce energy consumption.
The help of Mr. Amer Keblawi and Mr. Ralph Saadeh in setting up and conducting the experimental part is greatly acknowledged. The financial support of the ASHRAE (Project RP-1483), the Swedish Research Council - MENA, the Lebanese National Council for Scientific Research, and of the Qatar Chair in Energy Studies Endowment Fund are highly acknowledged.
A = area, [m.sup.2]
CC = chilled ceiling
COP = coefficient of performance
[C.sub.p] = specific heat, kJ/kg.K
DV = displacement ventilation
E = energy consumption, GJ
H = stratification height
IAQ = indoor air quality
J = electric power
[M.sub.a] = air mass flow rate, kg/s
PLR = part load ratio
PMV = predicted mean vote
R = chilled ceiling load ratio to space load
T = temperature, [degrees]C
w = humidity ratio, kg of [H.sub.2]O/kg of dry air
Z = vertical coordinate in the room from the floor, m
c = ceiling
s = supply air
American Society for Heating Refrigeration and Air conditioning Engineers ASHRAE, Book of Fundamentals, Ch. 12, 2005.
Ayoub M., N. Ghaddar, and K. Ghali. 2006. Simplified thermal model of spaces cooled with combined chilled ceiling and positive displacement ventilation system. ASHRAE International Journal of HVAC & R Research, 12 (4):1005-1030.
Braun, J.E., S.A. Klein, and J. W. Mitchell. 1989. Effectiveness models for cooling towers and cooling coils. ASHRAE Transactions 95(2): 164-174.
Conroy, C., and S. Mumma. 2001. Ceiling radiant cooling panels as viable distributed parallel sensible cooling outdoor systems. ASHRAE Transactions 107(1):1-8.
Ghaddar N., K. Ghali, R. Saadeh, and A. Keblawi. 2008. Design Charts for Combined Chilled Ceiling Displacement Ventilation System (1438-RP). Submitted to ASHRAE Transactions, TRNS-00263-2007.
Ghaddar N. and A. Bsat. 1998. Energy conservation of residential buildings in Beirut. International Journal of Energy Research 32 (2): 523-546.
Ghali K., N. Ghaddar, and M. Ayoub. 2007. Chilled Ceiling and Displacement Ventilation System: An Opportunity for Energy Saving in Beirut. International Journal of Energy Research, 31: 743-759.
House J.M. and Smith T. A system approach to optimal control for HVAC and building systems. ASHRAE Transactions: Symposia 1995; 101 (2): 647-660.
Jeong, J., S. Mumma, and W. Bahnfleth. 2003. Energy conservation benefits of a dedicated outdoor air system with parallel sensible cooling by ceiling radiant panels. ASHRAE Transactions, KC-03-7-1.
Jiang W. and T.A. Reddy. 2007. General methodology combining engineering optimization of primary HVAC&R plants with decision analysis methods- Part 1: deterministic analysis. HVAC&R Research, 13(1): 93-118.
Lau, J. and Q. Chen. 2006. Energy analysis for workshops with floor-supply displacement ventilation under U.S. climates. Energy and Buildings: 38(10): 1212-1219.
Lu L., W. Cai, L. Xie, S. Li and Y. C. Soh. 2005. HVAC system optimization in building section. Energy and Buildings, 37(1): 11-22.
Mehta D.P., and A. Thumann. 1989. Handbook of energy engineering. Lilburn, GA: Fairmont Press.
Mitchell M. An Introduction to Genetic Algorithm. Cambridge, The MIT Press; 1997.
Molina, A., A. Gabaldon, J.A. Fuentes, F.J. Canovas. 2000. Approach to multivariable predictive control applications in residential HVAC direct load control. IEEE Power Engineering Society Summer Meeting, (3): 1811-1816.
Nassif N., S. Kajl, and R. Sabourin. 2005. Optimization of HVAC control system strategy using two-objective genetic algorithm. HVAC&R Research, 11(3): 459-486.
Novoselac A. and J. Serbric. 2002. A critical review on the performance and design of combined cooled ceiling and displacement ventilation systems. Energy and Buildings, 34: 497-509.
Niu, J., L. Zhang, and H. Zuo. 2002. Energy savings potential of chilled-ceiling combined with a desiccant cooling in hot and humid climates. Energy and Buildings 34:487-95.
Seppanen, O., W. Fisk, J. Eto, and D. Grimsrud. 1989. Comparison of Conventional Mixing and Displacement Air-Conditioning and Ventilation system in U.S. Commercial buildings. ASHRAE Transactions, VA-89-19-3.
Strand R.K. and K.T. Baumgartner 2004. Modelling radiant heating and cooling systems: integration with a whole building simulation program. Energy and Buildings, 36(1): 1-9.
Wang S, X. Jin. 2000. Model-based optimal control of VAV air-conditioning system using genetic algorithm. Building and Environment, 35:471-487.
Wright J.A., H.A. Loosemore, and R. Farmani. 2002. Optimization of building thermal design and control by multi-criterion genetic algorithm. Energy and Buildings, 34: 959-972.
Visual DOE 4.0 software. Architectural Energy Cooperation, www.archenergy.com, San Francisco, USA, 2005.
Zaheer-uddin M. and R. Patel. The design and simulation of a sub optimal controller for space heating. ASHRAE Transactions 1993; 98:554-564
Zhou X. and J.E. Braun. 2007. Simplified Dynamic Model for Chilled-Water Cooling and Dehumidifying Coils-Part 1: Development (RP-1194) HVAC&R Research 13(5): 785-824.
Student Member ASHRAE
Kamel Ghali, PhD
Nesreen Ghaddar, PhD
Lars Jensen, PhD
This paper is based on findings resulting from ASHRAE Research Project RP-1438.
Mounir Mossolly is a graduate student and Nesreen Ghaddar is Endowed Qatar Chair and an energy studies professor in the Department of Mechanical Engineering, American University of Beirut, Lebanon. Kamel Ghali is an associate professor of mechanical engineering at Beirut Arab University, Lebanon. Lars Jensen is a professor of building services engineering at Lund University, Sweden.
|Printer friendly Cite/link Email Feedback|
|Author:||Mossolly, Mounir; Ghali, Kamel; Ghaddar, Nesreen; Jensen, Lars|
|Date:||Jul 1, 2008|
|Previous Article:||Granular filtration for airborne nanoparticles.|
|Next Article:||Development of adaptive algorithms for the operation of windows, fans, and doors to predict thermal comfort and energy use in Pakistani buildings.|