Professional Documents
Culture Documents
com
Received 1 September 2005; received in revised form 6 July 2006; accepted 30 January 2007
Available online 14 September 2007
Abstract
Based on the dynamic model of the gas engine heat pump (GEHP) system, an intelligent control simulation is presented to research
the dynamic characteristics of the system in the heating operation. The GEHP system simulation model consists of eight models for its
components including a natural gas engine, a compressor, a condenser, an expansion valve, an evaporator, a cylinder jacket heat exchan-
ger, an exhaust gas heat exchanger and an auxiliary heater. The intelligent control model is composed of the prediction controller model
and the combined controller model. The Runge–Kutta Fehlberg fourth–fifth order algorithms are used to solve the differential equations.
The results show that the model is very effective in analyzing the effects of the control system, and the steady state accuracy of the intel-
ligent control scheme is higher than that of the fuzzy controller.
2007 Elsevier Ltd. All rights reserved.
0196-8904/$ - see front matter 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.enconman.2007.01.033
Y. Zhao et al. / Energy Conversion and Management 48 (2007) 3146–3153 3147
Nomenclature
Ahf refrigerant side area of control volume (m2) r vapor latent heat (kJ/kg)
Ahw water side area of control volume (m2) t time (s)
B fuel consumption (kg/s) T1 inlet temperature of water cooling engine (C)
c clearance volume factor T2 outlet temperature of water cooling engine (C)
Cp specific heat of pipe wall at constant pressure T3 air temperature of condenser’s outlet (K)
(kJ/kg K) Te outdoor air temperature (K)
Cpa specific heat of air at constant volume (kJ/kg K) Tin temperature of refrigerant entering control vol-
Cpr specific heat of exhaust gas at constant volume ume (kg/s)
(kJ/kg K) T1 indoor air temperature (K)
Cw specific heat of water cooling engine (kJ/kg K) Tout temperature of refrigerant out of control volume
d pipe diameter (m) (kg/s)
Gw flux of water cooling engine (kg/s) Tpa air temperature of engine’s inlet (K)
Gw air consumption (kg/s) Tpr outlet temperature of exhaust gas (K)
hcp,out specific enthalpy in compressor outlet (kJ/kg) Tf average temperature of refrigerant (K)
hcp,in specific enthalpy in compressor inlet (kJ/kg) Tp average temperature of pipe wall (K)
hl liquid specific enthalpy of refrigerant in heat ex- Tw average temperature of water (K)
changer (kJ/kg) DT temperature difference (K)
hg gas specific enthalpy of refrigerant in heat ex- DTsh(s) evaporator outlet superheat (K)
changer (kJ/kg) DT 0sh ðsÞ receptive superheat (K)
K1, T1 characteristic factor Vcp compressor cylinder volume (m3)
K characteristic factor of expansion valve vcp,out specific volume in compressor outlet (m3/kg)
L length of control volume (m) x refrigerant quality
LHV low heat value of natural gas (kJ/N m3) apf coefficient of heat transfer between heat exchan-
mf,cp refrigerant mass flux (kg/s) ger and pipe wall (kJ/K m2)
min mass of refrigerant entering control volume apw coefficient of heat transfer between water and
(kg/s) pipe wall (kJ/K m2)
mout mass of refrigerant out of control volume (kg/s) x compressor speed (rps)
P pressure (Pa) k coefficient of discharge
Pc condenser pressure (Pa) kv volumetric factor
Pe evaporator pressure (Pa) kp pressure factor
DP difference between condenser pressure and evap- q refrigerant density (kg/m3)
orator pressure (Pa) qf density in expansion valve inlet (kg/m3)
DPs1 suction pressure loss (Pa) qg gas density of refrigerant in heat exchanger
Qb total heat from natural gas (kW) (kg/m3)
Qe effective shaft work (kW) ql liquid density of refrigerant in heat exchanger
Qr waste heat recovered from exhaust of gas engine (kg/m3)
(kW) c void factor
Qs heat loss (kW) c average void factor
Qw waste heat recovered from cooling water of gas
engine (kW)
control system in which a power law control action is intro- can be used to build better models and controllers. Fuzzy
duced to provide the necessary compressor speed variations controllers have been implemented by many researchers
according to ambient conditions [4]. With the development and good results were attained [5]. So, it is possible to com-
of artificial intelligence, there are also other new methods bine fuzzy set theory with the classical PID control
introduced into the control schemes, such as the fuzzy set problems.
theory, the function of the Smith predictor, the radial basis The function of the Smith predictor is to predict the pro-
function neural network and so on. cess variables well before it is possible to measure them.
The fuzzy set theory is a theory about vagueness and When the appropriate measurement actually becomes
uncertainty and enables us to use non-precise, ill-defined available, the synchronized error can be used to correct
concepts and yet work with these in a mathematically strict the current model predictions [6].
sense. When only imprecise or indefinite information about Because the building load is continuously changing with
the dynamic behavior of a process is available, fuzzy sets time due to changes in weather, equipment use and occupancy,
3148 Y. Zhao et al. / Energy Conversion and Management 48 (2007) 3146–3153
In recent years, there has been a great deal of results Fig. 1. Schematic diagram of the GEHP working in heating mode.
and investigations concerning models elaborated to simu-
late the behavior of heat pump systems. In this paper, a " n1 #
simplified model of the GEHP system is developed to sim- n Pc n
W cp ¼ V cp xkv kp P e 1 ð3Þ
ulate the heating process. The model has been proved to 1n Pe
reproduce the actual behavior of the heat pump well and " 1 #
provides the foundation for analysis of the control scheme Pc n
kv ¼ 1 c 1 ð4Þ
[8]. Pe
The system model is composed of a natural gas engine DP s1
model, a compressor model, a condenser model, an expan- kp ¼ 1 ð5Þ
Pe
sion valve model, an evaporator model, a cylinder jacket
heat exchanger model, an exhaust gas heat exchanger
model and an auxiliary heater model. The cylinder jacket 2.2. Thermostatic expansion valve [12,13]
heat exchanger and the exhaust gas heat exchanger are to
recover waste heat from the cooling water and the exhaust The thermostatic expansion valve model is composed of
heat of the tail gas of the gas engine, respectively. The aux- the control unit and the flow unit. The two units are used to
iliary heater will be switched on when the temperature of adjust the flow velocity according to the superheat. Because
the supply water is lower than needed. Fig. 1 shows the of its small volume, the expansion valve is considered as
schematic diagram of the GEHP heating system. The flow adiabatic without any refrigerant mass variation in it.
routines of the different flow medium are denoted with dif- The enthalpy is assumed to be constant. So, in the flow
ferent line types. unit, the refrigerant flow is given by
pffiffiffiffiffiffiffiffiffiffiffi
mf ¼ K DP qf ð6Þ
2.1. Compressor [9–11]
while in the control unit, the relation between the receptive
No mass variation in the compressor is considered here, superheat and the evaporator outlet superheat can be given
so the mass flows at the inlet and outlet are the same, and a by
polytropic compression process is assumed. Then, the K1
refrigerant mass flux, mf,cp, can be calculated by the follow- DT 0sh ðsÞ ¼ ess DT sh ðsÞ ð7Þ
1 þ T 1s
ing equations:
V cp xk
mf;cp ¼ ð1Þ 2.3. Heat exchangers (plate type)
vcp;out
" 1 #
Pc n In this paper, a two zone model is utilized. The evapora-
k ¼ 0:98 0:085 1 ð2Þ tor is divided into two zones (evaporation and superheating
Pe
zones), and the condenser is divided into three zones
The power of the compressor is given by (superheating, condensation and subcooling zones). Each
Y. Zhao et al. / Energy Conversion and Management 48 (2007) 3146–3153 3149
zone is considered as a control volume. In every control Similarly, the heat transfer coefficient in the superheated
volume, the equations of continuity, energy and momen- zone between the evaporator pipe wall and the refrigerant
tum are established. can be taken from the following correlation:
In the model, some assumptions are made: (1) the pres-
Nu ¼ 0:18Re0:18 Pr0:4 ð18Þ
sure drop along the pipes is not considered and the pressure
varies with time; (2) the heat loss is negligible; (3) the heat asup d
Nu ¼ ð19Þ
conduction through it is negligible. k
For every control volume, the following set of differen- while in the two phase zone, for the annular flow, the heat
tial equations can be written [9,14]: transfer coefficient is represented by [15]
k
• Mass conservation equation a00 ¼ 0:023 Re0:8 Pr0:4
l / ð20Þ
d l
d
f½ql ð1 cÞ þ qgcAhf Lg where / is a modification coefficient. The average heat
dt transfer coefficient between the water and the condenser
dL pipe wall can be derived from the following correlation
¼ min mout þ ½ql ð1 cÞ þ qgcAhf ð8Þ
dt [17]:
• Energy conservation
Nu ¼ 0:023Re0:8 Pr0:4 ð21Þ
d dP
f½ql hl ð1 cÞ þ qg hgcAhf Lg Ahf L while for the evaporator, it is given by [17]
dt dt
¼ apf AðT p T f Þ þ min hin mout hout Nu ¼ 0:023Re0:8 Pr0:3 ð22Þ
dL It is important to determine the thermodynamic and trans-
þ ½ql hl ð1 cÞ þ qg hgcAhf ð9Þ
dt port properties of refrigerant R134a. The P–R state equa-
• Pipe wall energy conservation tion is used here [18]:
dT p RT aðT Þ
Cp ¼ apw Ahw ðT w T p Þ apf Ahf ðT p T f Þ ð10Þ p¼ ð23Þ
dt v b vðv þ bÞ þ bðv bÞ
• Water (secondary fluid) energy conservation
2.4. Natural gas fueled engine
dT w
Cw ¼ apw Ahw ðT w T p Þ ð11Þ
dt Supposing that the natural gas engine is one-cylinder,
four-stroke and the speed can be adjusted between
The correlations utilized in this work to calculate the 1000 rpm and 3000 rpm, its related model can be found
convective heat transfer coefficients and void fraction are in Ref. [8]. The total fuel consumption of the engine can
listed as follows. The void factor is given by [10] be calculated by the following equation [8]:
1
c¼ 1 q ð12Þ B ¼ Qb =LHV ð24Þ
1 þ x 1 qgl
where the thermal value of the fuel per unit time can be cal-
So, the density in the two phase zone can be expressed as culated by the formulation in Ref. [8]:
q ¼ cqg þ ð1 cÞql ð13Þ
Qb ¼ 0:292 þ 0:011x þ 0:000479x2 þ 1:337143W
The heat transfer coefficient in the single phase zone (super-
heat or subcooling) between the condenser pipe wall and the 0:01139W 2 0:00075xW þ 0:000108xW 2
refrigerant can be taken from the following correlation [15]: 7:2 107 x2 W 5:3 107 x2 W 2 ð25Þ
0:72 0:3
Nu ¼ 0:18Re Pr ð14Þ
while in the two phase zone, it can be calculated by [16] 2.5. Heat exchanger for recovering the waste heat [19]
0:8 !0:25
Rel 0:33 ql The model of heat exchangers for recovering the waste
Nu ¼ 0:0048 Pr ð15Þ
H qg heat can be described by the energy equilibrium of the nat-
where H is a dimensionless parameter for consideration of ural gas engine:
a liquid membrane and c 0 (J/kg) denotes the influence of li- Qb ¼ Qe þ Qw þ Qr þ Qs ð26Þ
quid membrane thickness and subcooling:
The heat recovered from the cooling water of the engine is
C p DT
H¼ ð16Þ Qw ¼ Gw C w ðT 2 T 1 Þ ð27Þ
c0
0 0:68C p DT and the exhaust gas heat is given by
c ¼c 1þ ð17Þ
r Qr ¼ ðB þ Ga ÞðC pr T pr C pa T pa Þ ð28Þ
3150 Y. Zhao et al. / Energy Conversion and Management 48 (2007) 3146–3153
Q(kW)
3. Solution method 29
Time
Outdoor temperature
Building load at
Indoor temperature
………
time (t)
………
Solar radiation
Humidity
Building load at
time (t-1)
Indoor
temperature
ke e E Fuzzy contr -
+ Fuzzy Fuzzy U
Non-fuzzy u
A/D process D olling rule process ko
d decision
d/dt kd
-
Gas engine heat pump Speed adjuster D/A
of gas engine
Fig. 4 depicts the diagram of the fuzzy control principle. The control action of the intelligent controller is summa-
Its input and output variables are listed as follows: rized below:
Input variables: deviation between the temperature of the • Real time tracking of the influence of weather changes
supply water and the desired setting value – e (1.0 to on the building load through the load prediction.
1.0 C), deviation variance rate – e (0.01 to 0.01 C/s); • Based on the feed forward prediction of load, accurate
Output variable: fuel consumption of the gas engine – B control of the system by the feedback controller is
(1.0–2.5 N m3/h). accomplished.
The decision making rules for the fuzzy control are sta- Once the GEHP is on, the compressor will run with
ted as follows: accelerating speed. The combination controller acts
according to the deviation between the setting value and
• If e is bigger and e is increasing, B should be greatly the actual value of supply water temperature, while the
increased. RBF predicts the building load for the next moment and
• If e is smaller and e is decreasing, B should be slightly reflect in real time the change to the prediction controller
reduced. at the same time. Then, a control action will be done. If
• If e goes to zero and e does not vary, B should be kept the temperature deviation lies between 0.02 and
invariable. 0.02 C, the fuzzy controller will be cut off and the PID
controller will act. The smooth transition between the
So, the two parts of the intelligent controller can be two controllers avoids oscillation. On the contrary, if the
illustrated as follows: The feed forward part predicts the deviation lies beyond the range of 0.02 to 0.02 C, adjust-
load of the building through the radial basis function ment will be made according to the fuzzy rules. The above
(RBF) neural network (prediction controller). The feed- process will continue until the system reaches steady state
back part is the combination control of fuzzy control with operation. The diagram of the intelligent control principle
Smith control and PID control (combined controller). of the GEHP system is shown in Fig. 5.
Meteorological Meteorological
Other parameters variation
parameters
Initialized temperature +
GEHP unit Temperature of
+ Combined +
controller supply water
Intelligent controller
B(Nm 3 /h)
door air temperature and (2) 8000 s of operation with
abrupt change of outdoor temperature. Figs. 6 and 7 refer 0.3
to case (1) and Figs. 8 and 9 refer to case (2). In both cases,
the desired setting temperatures are 21 C, and the original
outdoor and indoor air temperature are both supposed to 0.25
be 7 C.
Figs. 6 and 7 show the dynamic simulation results of the 0.2
compressor speed and indoor temperature with constant 0 2000 4000 6000 8000
55 6
0 2000 4000 6000 8000
t(s)
50 Fig. 9. Dynamic variation of the controlled room temperature under
falling outdoor temperature conditions.
n(r/s)
45
change of outdoor temperature. The initial operation is
the same as in case (1), but at 4000 s, the outdoor air tem-
40
perature falls by 5 C. It can be seen from these figures that
when the outdoor temperature changes, the indoor temper-
35
ature will fluctuate and then get to its steady value quickly,
0 1000 2000 3000 4000 5000 but vibration does not appear. Although the steady state
t(s) error increases at the same time, it is still acceptable. It also
Fig. 6. Dynamic variation of compressor speed. shows that when the difference between the indoor and out-
door temperature increases, the fuel consumption of the
gas engine will go up fast and then tend to a constant value
22 in a short time. Therefore, the control scheme has good
20 adaptability for disturbances caused by shifts of outdoor
temperature.
18
16 6. Conclusions
T(˚C)
14
12
(1) A simplified model of the GEHP system has been
constructed to simulate the heating process. It has
10
been proved to reproduce the actual behavior of the
8 heat pump well, and it provides the foundation for
6 the analysis of the control scheme. It is effective in
0 1000 2000 3000 4000 5000
analyzing the effects of the control system and the
t(s)
dynamic variation of performances in the operation
Fig. 7. Dynamic variation of the controlled room’s temperature. process.
Y. Zhao et al. / Energy Conversion and Management 48 (2007) 3146–3153 3153
(2) An intelligent controller is developed in this paper. It [5] James Li C, Tzou JC. A new learning fuzzy controller based on the P-
is composed of a prediction controller and a combi- integrator concept [J]. Fuzzy Sets Syst 1992;48:297–303.
[6] Mulholland Michael, Fernandes Luis AR. Combined Smith predic-
nation controller. The prediction controller based tion and recursive filtering for observation of a polyethylene reactor
on the RBF neural network can be used for predict- [J]. Comput Chem Eng 1997;21(11):1283–9.
ing the building load at the next moment. The com- [7] Bishop CM. Neural networks for pattern recognition. Clarendon
bined controller consists of a PID controller, a Press; 1995.
fuzzy controller and a Smith controller. The simula- [8] Rusk RP, Van Gerpen JH. Development and use of a mathematical
model of an engine-driven heat pump [J]. ASHRAE Trans 1990;Part
tion results show that its dynamic and steady charac- 2:282–90.
teristics are satisfactory. The intelligent controller can [9] Koury RNN, Machado L. Numerical simulation of a variable speed
eliminate the disadvantages of long time delay, large refrigeration system [J]. Int J Refrig 2001;24(2):92–200.
inertia and nonlinearity of the heating system and [10] Ding Guoliang, Zhang Chunlu. Simulation and optimization of
achieve good steady state accuracy and dynamic refrigeration and air-conditioning equipments. Beijing: Science
Press; 2001.
characteristics. [11] Li Shulin. Refrigeration in air conditioning engineering. Bei-
(3) It is feasible to use the intelligent controller in the jing: Mechanical Industry Press; 1995.
heating system to realize energy saving and comfort. [12] Chen Zhijiu, Que Xiongcai, Ding Guoliang. Refrigeration system
thermodynamics. Beijing: Mechanical Industry Press; 1998.
Acknowledgement [13] Chen Youhua, Chen Zhijiu. Comparison research of expansion valve
in refrigeration machines [J]. Refrig J 2003;3:10–5.
[14] Ge Yunting, Yan Qisen, Peng Xiongbing. Theory and experimental
This project was supported by NSFC, the National Edu- study on simulation mathematics of refrigeration and air conditioning
cation Department for Doctor Center Foundation, the system [J]. Refrig J 1995;4:9–18.
Tianjin Science and Technology Development Project [15] Ni Xiaohua, Xia Qing, Xiao Yuan. Calculation of heat transfer and
and National 863 Program. pressure drop in plate heat exchangers [J]. Fluid Mach 2002;30(3): 22–6.
[16] Wang Lieke, Yang Qiangsheng. Steam condensation heat transfer
characteristics in plate heat exchanger [J]. J Shanghai Jiaotong Univ
References 1998;32(4):18–22.
[17] Zhang Ximin. Heat transfer. Chinese Architecture Industry Press;
[1] Zhao Yang, Zhao Yi, Li Lixin, et al. Evaluation of five kinds of 1998.
heating and cooling system and some suggestions [J]. J Refrig [18] Su Changsun. Advanced engineering thermodynamics. Higher Edu-
2000;4:43–9. cation Press; 1987.
[2] Parnitzki D. Digital control of heat pumps with minimized power [19] Xu Ying, Qin Chaokui. Natural gas engine driven chiller systems and
consumption [J]. Int J Energy Res 1989;13(2):167–78. the waste heat usage [J]. Energy Technol 2002;23(4):162–7.
[3] House John M, Smith Theodore F. System approach to optimal [20] Matlab 6.5. The Mathworks.
control for HVAC and building systems [J]. ASHRAE Trans [21] Xiong Hongyun, Zeng Shaobiao, Mao Yunying. Applied mathemat-
1995;101:647–60. ics foundation. Tianjin University Press; 1994.
[4] Vargas JVC, Parise JAR. Simulation in transient regime of a heat [22] Zhao Yi. Research on intelligent control simulation of natural gas
pump with closed-loop and on-off control [J]. Int J Refrig engine driven water-to-water heat pump unit. Tianjin University
1995;18(4):235–43. Master Thesis, 2002.