An EnergySaving Control Strategy on Central AirConditioning System.docx

1、An EnergySaving Control Strategy on Central AirConditioning SystemAn Energy-Saving Control Strategy on Central Air-Conditioning System By Sheng Xing1, Yin Juan2, Shang Jingfu3, Yu Yang4, Wu Zhangxian51. State Grid Information & Telecommunication Co., Ltd.2. China Huadian Finance Corporation Limited3

2、. State Grid AC Engineering Construction Company4. Northeast China Electric Power Dispatching & Communication Center5. North China Electric Power UniversityAbstract:An energy-saving control strategy based on predictive control for central air-conditioning systems is proposed in this paper. The cold

3、load model is developed to describe the dynamic characteristics of temperature control systems, and then parameters in the cold load model and in the central air-conditioning system model are estimated. Generalized predictive control (GPC) is used to establish an optimization model to minimize the c

4、onsumption of energy and the control error of temperature. The simulated annealing (SA) algorithm, combined with quadratic programming, is adopted to solve the optimal problem. Contrasted with the simulation of traditional PID control, the results prove the effectiveness of this proposed strategy.Ke

5、ywords:central air-conditioning; predictive control; energy saving; time-delay system; simulated annealingIntroductionElectric energy is the most important form of purchased energy in China1. For this reason, the efficiency of electric energy should be improved, since energy conservation is a necess

6、ary concern for developing countries which have experienced rapid development. In China, the boom in the construction, tourism and commerce industries has created an increasing demand for air conditioning2. Currently, electrical central air conditioning systems account for 20% of the total electrici

7、ty consumption in China3. Inevitably, for air conditioning systems,more efficient control strategies based on energy-saving targets need to be developed.According to recent studies, numerous researchers have focused their efforts on fuzzy control theories to be implemented on central air-conditionin

8、g systems4,7. Meanwhile, some researchers use predictive control strategy to compensate the long delay in temperature control5, 6. Besides, Wang and Burnett proposed anonline optimal control strategy, in which an adaptive and derivative method is used to determine the set-point for pressure8. As the

9、 main factor influencing temperature in a building, the cold load must be considered. Furthermore, the cold load could dynamically reflect the needed cold energy. In order to enhance the system operating efficiency and get a stable control performance, a generalized predictive control strategy based

10、 on cold load is proposed in this paper. An introduction of cold load and modeling the temperature control function are illustrated in section 2, while the optimal formulation and algorithm are proposed in section 3. In section 4, the simulation is presented and the conclusion is given in section 5.

11、Modeling of temperature control systemThe description of cold energy outputIn central air-conditioning systems, hot air from the building transfers heat to chilled water, causing the temperature of chilled water to increase. Chilled water pumps are used to recirculate the chilled water and transfer

12、heat to cooling water by refrigeration machines, which make chilled water cold again. Next, cooling water is recirculated by cooling pumps and finally heat is released into atmosphere in the cooling tower. In this paper, it is assumed that a set of circulating systems for air-conditioning could be d

13、ivided into one cooling water pump, one chilled water pump and one refrigeration machine, and there are totally Nre sets of such systems in the central air-conditioning system of a building. According to the laws of thermodynamics, the cold energy output in time k can be written as follows:cold_ener

14、gy (k) =sk Cwt1F1, k (1)skCwt2F2,k=skCwt1F1,k-Qrm (2)where, sk: the number of circulation system used in time k;Cw: the speci c heat capacity of water; F1, k , F2, k: the respective ows of chilled and cooling water; t1,t2: the respective temperature variation of the chilled water and cooling water;M

15、eanwhile, Qrm represents the heat generated by the refrigeration machine, which is proportional to the cube of ow and could be de ned as follows:(3)where is the ef ciency of the refrigeration machine and is a proportion factor.The model of cold loadThe cold load in the building can be determined by

16、the following four factors: the heat brought by the building surface, which is proportional to the buildings surface area (Pt), the heat brought by winds from central airconditioning system (Pw), the heat emitted by people in the building (Pf), the heat produced by electric equipment in the building

17、 (Pe). So the cold load can be described as follows: cold_load=Pf+Pw+Pt+Pe (4) We assume cold load per person to be a constant (the constant equals 30 W in summer). Meanwhile, Pt can be described as a function of surface area: Pt=KF (tL,k-tn) （5）in which K stands for thermal conductivity and F stand

18、s for the surface area, while tn and tL,k stand for the set temperature in the building and the temperature outside the building in time k, respectively. Since the roof and the side of a building have different thermal conductivities, (2) would be rewritten as follows:Pt=(K1Froof+K2Fwall)(tL,k-tn) (

19、6)For simplicity, we assume Pw is proportional to Pt and the coef cient is , so (1) can be rewritten as follows:cold_load(k)=30Pfk+(1+)(K1Froof+K1Fwall)(tL,k-tn)+Pc (7)Pfk represents the ow of people in the building in time k, and obeys a Poisson-process, which has been proven as the most approximat

20、e distribution of people ow in large buildings, and can be de ned as follows: (8) in which is the expectation of people ow. From (1) to (8), the unknown parameters are , PE and , which can be tted using experiment data of least square method. The predictive model of temperature control In a central

21、air-conditioning system, the cold energy produced by air-conditioning would satisfy the cold load firstly, and then the redundant cold energy left would maintain the temperature at a set value. If the cold energy can not satisfy the cold load, the temperature in the building will rise. Meanwhile, th

22、e real-time temperature transmitted by the transducer in the building would feedback to make the output of the cold energy more exactly, as shown in Fig.1.Since the temperature in the building is difficult to change, a first-order inertia element with a pure delay is used to simulate the dynamic pro

23、cess of the building when input variables change. So the transfer function can be written as follows, (9) That is, (10)where T is the inertia time constant and g equal to CairFair is the gain.The equation above can be discretized as follows: (11) where T0 is the sample time and N is an integer. The

24、delaying time =NT0 (12)where DTout represents the change of Tout while U(k-N) stands for the difference between cold energy and cold load, and can be de ned as follows:DTout(k)=Tout (k) - Tout (k -1) (13)U(k -N)=cold_energy(k -N)-cold_load(k -N) (14)Cold energy and cold load can be calculate by (1)

25、and (7), by using data in time k-N.GPC based control strategy Since D.W. Clarke presented the principle of the Generalized Predictive Control (GPC) in Automation in 1987, GPC has become the most popular Model Predictive Control method in both industry and academic applications, in which it has been

26、found very effective and robust. In GPC, which belongs to the group of rolling optimization, a set of future control signals (P signals in this paper) are generated in each sampling interval, but only the rst element of the control sequence is applied to the system input. Optimization formulationAcc

27、ording to (2) and (3), since t1 and t2 must be maintained in an acceptable range (which makes them impossible to be control variables), but two of the other three variable (sk, F1,k and F2,k) can be used as control variables. In this paper, sk and F1,k are selected as control variables. In the objec

28、t function, energy cost and temperature control error are taken into consideration. We define:where F1,max and F2,max stand for the maximum ow of the chilled water pump and cooling water pump, respectively.In addition, temperature output in the future P steps can be predicted by the equation:Tout (k

29、0+n)=Tout(k0)-DTout(k0+n) n=0,1.P (17) DTout (k0+n) can be obtained by (12), and Tout (k0) is the current temperature in the building which can be transuded by sensors, so the temperature in the next P steps can be predicted by (17) and the error of temperature output can be de ned: (18) According t

30、o (15) - (18), the optimization model can be established as follows:where F1, min and F2, min represent the minimum ow of the chilled water pump and cooling water pump, respectively, and P is used to designate the number of future signals generated in a samples time. In addition, W1* and W2* represe

31、nt the process weighing and nal state weighing, while W*1 and W*2 stand for the energy-saving weighing and temperature-accuracy.Solution algorithm based on SA For every sampling interval, an output signal, including F1,k, sk, can be obtained by solving the optimization model described above. However

32、, this optimization model is a nonlinear programming problem with integer constraints, which cannot be solved by the traditional Diophantine equation. In order to ensure the accuracy of the solution as well as the real time capability, an optimization algorithm based on simulated annealing (SA) is presented in this paper. Since the solution efficiency largely depends onthe initial value of sk, a particular treatment should be adopted. Firstly, we assume the num