A METHOD TO REDUCE EUROPEAN CHILLER HOURLY LOAD CURVES TO A FEW POINTS

Transcription

1 A METHOD TO REDUCE EUROPEAN CHILLER HOURLY LOAD CURVES TO A FEW POINTS Philippe Rivière philippe.riviere@ensmp.fr Jérôme Adnot - jerome.adnot@ensmp.fr Dominique Marchio dominique.marchio@ensmp.fr CENERG -Ecole des Mines de Paris 6 Bd St Michel, F Paris cedex 6 Luis Pérez-Lombard - lpl@tmt.us.es José A. Ortiz - jaog@arrakis.es AICIA-Grupo de Termotecnia, Escuela Superior de Ingenieros, Universidad de Sevilla, Camino de los Descubrimientos, s/n. E-492 Sevilla SPAIN Abstract. EECCAC is an EU funded project, where twelve participants (National Energy Agencies, Manufacturers, University laboratories and Electric utilities) from eight countries gathered to identify the most suitable measures to achieve market transformation in the direction of energy efficiency of central air conditioning systems in Europe. Its basement was the Eurovent manufacturers wish to develop a seasonal performance index for chillers in Europe, so as to compare their performances not only on full load, which is not representative of chiller average performances, but also on part load. To that extent, a method has been developed that enables to reduce the chiller hourly load curve to N triplets (weighting coefficient, outside air or water temperature, load ratio). In order to minimise the number of testing points, as well as to adopt a format compatible with the existing American IPLV standard, 4 points have been kept. The reduction is applied to annual hourly load curves, results of the DOE software simulation of an existing office building for four different air conditioning systems and within three different climates (London, Seville, Milan). Finally, the bias introduced on the seasonal efficiency by the reduction is compared to the uncertainty of measurement of the chiller seasonal performance. Keywords: Chiller, Annual hourly load curve, Seasonal performances, E (Eurovent Seasonal Energy Efficiency Ratio), EECCAC. INTRODUCTION HVAC systems constitute a very important and growing energy use in the European Union. Within EECCAC (24) (Energy Efficiency and Certification of Central Air Conditioning), an EU funded project, twelve participants from eight countries (including the EU manufacturers' associations, Eurovent), have been engaged in identifying the most suitable measures to improve the energy efficiency of these systems. The project has considered all cooling systems over 2 kw (here referred as Central Air conditioning, CAC), and has been structured in four main directions: market study and energy

2 impact of CAC in Europe, technical and economical study of CAC energy performance, procedure to compute average performance of chillers, selection and preparation of actions. Within the study of CAC energy performances, hourly load curves were calculated for a typical office building. Different system types were simulated but we focus hereafter on CAC systems. Extensive experimentation was led on chiller part load performances. By coupling these pieces of information, the following analyses were performed: - Influence of climate and system on chiller seasonal performances - Hourly load curves reduction to N triplets (load, temperature, energy weighting coefficient) - Computation of average seasonal conditions for the E (Eurovent Seasonal Energy Efficiency Ratio). This article presents the methodology that has been developed to summarise the hourly thermal load information and weather data to a limited number of points. 2. HOURLY LOAD CURVES The thermal load has been extracted from simulations lead with DOE-2.2, on which more details were reported by Pérez-Lombard (24). We recall the main input characteristics below. - A representative office building with a total area of 828 m2 (4795 m2 conditioned) has been adapted to the different European envelope regulations. Three different locations have been selected to represent European weather: London, Milan and Seville. - Four chiller based systems have been studied, namely: Single zone constant air volume (CAV), Single zone constant air volume with free cooling (CAV-FC), Variable air volume (VAV), Fan-coil four pipes (FC4P). - Thermal comfort conditions have been considered to guarantee similar comfort levels for the different system types. The same ventilation ratios have been specified for all the systems; for the FC4P system, a primary AHU that provides neutral (22ºC) air directly to every building zone. - Chilled water loops provide 7ºC water to cooling coils. Water delta T for cooling is 5ºC. The thermal hourly load was extracted at the inlet and outlet of the chiller. - The cooling system is turned off at night. 3. CHILLERS PERFORMANCES Technologies of chillers of the European market are very few despite of the great number of models: air cooled scroll chillers until about 2 kw and screw chillers from 2 kw to 75 kw. We do not consider here centrifugal chillers whose cooling capacity, always higher than 75 kw put them out of the Eurovent directory, frame of this study. Five chillers are kept to perform the required analysis (Table ). Table. Description of the five chillers Type Compressor Capacity staging (%) % EER increase by OAT C decrease Scroll 3,5 2 Hermetic reciprocating 5, 2,7 3 Scroll 33, 66, 3,7 4 Scroll 25, 5, 75, 4, 5 Twin screw 5, 5, 7- (continuous) 2,3 Instantaneous energy efficiency ratio of chillers, EER, is defined as the ratio of the cooling capacity (kw), CC, to the electric power (kw), EP, required (water pump consumption on the evaporator loop is excluded).

3 Since we only consider hereafter air cooled chillers, the EER varies along the season according to the cooling load and the outside air temperature (OAT), inlet of the chiller condenser. The outlet water temperature at the evaporator and the chiller water flow rate remain constant. 3. Impact on the EER of the variation of the inlet air temperature at the condenser The cooling capacities and electric power (and thus the EER) are defined for a given inlet air temperature at the condenser (OAT). The variations of the efficiency with OAT as compared to efficiency at rating (r) condition (OAT r =35 C) were found to be accurately translated by the simple model of Roujol (23). As compared to the ASHRAE (996) chiller model, the form of this model enables to reduce the number of parameters required to describe the variation of efficiency with the OAT. The control of the condenser fan flow rate for low inlet air temperatures was considered (Rivière, 24). However, it did not impact the accuracy of the methodology to reduce hourly load curves and thus it is not developed in this article. Consequently, the five chillers exhibit nearly linear increase of efficiency with decreasing OAT. The specific slopes for each chiller are reported in Table. 3.2 Impact on the EER of the variation of the cooling load We accept here the hypothesis commonly adopted (ASHRAE, 996) to represent chiller performances: the relative variation of efficiency with source temperatures remains valid for different load rates. The continuous representation for part load performances of the ASHRAE (996) model is well suited to chillers whose capacity control is continuous: centrifugal, screw with slide valve or inverter driven chillers; it also enables to lower the parameter number for simulation engines and to ensure continuity. However, since chiller performances are posttreated, there is no constraint on parameter nor continuity. Thus, part load performances are calculated as follows: - if load lies between two capacity stages (stages and 2), capacities of each stage and of the required thermal load will determine the time (t) spent on each stage and thus the part load efficiency, Eq. (). The electric power is calculated on the basis of Eq. (2). t CC - CC 2 = & EP t.ep + ( t).ep2 LOAD CC = () & (2) - if load is inferior to the smallest available step, the chiller cycles on and off with subsequent efficiency degradation, Eq. (3): EERPL LR = (3) EER ( - a).lr + a FL LOAD (kw) Where the part load rate (LR) is defined as: LR = CCFL (kw ) Equation 3, first published by Anglesio (2), was checked experimentally for a 7 kw air to water chiller (Rivière & al, 24). For that chiller, the coefficient a of Eq. (4) was,96 ; it appeared to be linked only to the sleep power consumption that gives the shape of the curve found by Anglesio, which differs from typical ON-OFF degradation curves as used in American standards. Based on available data (chillers and heat pumps below 5 kw), and also on literature field campaign results (Bettanini, 23), a default coefficient of,9 was kept (for higher capacity chillers, evidence lacks both on ON-OFF classical degradation impact and on the weight of the sleep power consumption so that this default figure is also used for chiller n 5). The part load performance curves for the 5 chillers considered Table are reported Fig..

4 Figure. Part load performance curves for the 5 chillers of Table 2% Part Load Performance OAT = 35 C Part load relative efficiency % % 8% 6% 4% 2% % % % 2% 3% 4% 5% 6% 7% 8% 9% % Part Load Rate (LR ) % n n 2 n 3 n 4 n 5 The relative efficiency at part load is to be understood as the ratio of the efficiency for a given load rate to the efficiency at full load for the same source temperatures. 4 EXISTING LOAD CURVE REDUCTIONS The objectives of the ASHRAE 6 (we focus hereafter on the cooling mode only) and ARI IPLV standards are to compare vapour compression cycles on the basis of their seasonal performances by integrating typical load and temperature variations. Both use the bin method to that extent. Its name arises from the use of a discrete temperature axe whose intervals are called bins. (Nota: EER are expressed in W (cooling) / W (electric)) 4. The ASHRAE 6 standard (ANSI, 995) This standard defines a method to calculate the (Seasonal Energy Efficiency Ratio) of unitary (central air to air conditioners) equipments (CC between 2 and 9 kw). STEP ) An average climate is used for the US. The outside air temperature is binned by 5 F and hours of occurrence of temperatures in each bin are calculated and noted n i. (Fig. 2) STEP 2) An average load curve (load versus temperature draws a straight line) is given for the US. (Fig. 2: L (%) represents the ratio of the building load to the sizing capacity of the unit). Figure 2. ASHRAE 6, load curve ratio (L %) and temperature occurrence [h/h] vs OAT n i [h/h], plain OAT ( F) STEP 3) Specific assumption for sizing is taken into account: at 35 C, the air conditioner provides % of the needs L % (W/W) striped

5 STEP 4) Air conditioners are tested with constant indoor temperature. All capacity stages are characterized by two testing points at low and nominal outside air temperature and simple modelling of each stage performance is drawn from these two points. STEP 5) Efficiency between two stages is calculated according to Eq. () and (2). A linear efficiency degradation is applied when load is inferior to the smallest capacity stage available. These hypothesis enable to calculate the using Eq. (4) (T i represents the average temperature of the bin i ): ni i = 5 (4) n.ep(t ) %-hours i.load(t 6 4 i 3 i i ) This standard applied 2 to a typical 4 steps chiller would lead to 8 testing 4 points (STEP 4). One supplementary testing point could also be performed to prove that the 2 unit performs better when cycling on the smallest capacity stage than the disadvantageous default law proposed, testing points thus amounting to 9. Moreover, chillers capacity control are likely to be less standardized Ti ( F) that the ones of central air conditioners since there is no upper capacity limit for chillers: this could lead to extensive case listing of the different capacity control schemes in order to supply complete modelling. 4.2 The ARI IPLV standard (ARI, 998) D C B A The ARI IPLV standard defines a method for rating seasonal performances for capacity staged chillers with only 4 testing points. The testing results are not used to model the chiller performances for varied load temperature conditions, but directly to calculate a weighted integrated part load index or IPLV. Let study the methodology used: STEP and STEP 2 remain unchanged, except that data differ, both for climate and load curves (linear by interval). For STEP 3, at 35 C, the chiller provides % of the needs. Here comes the main change: before to proceed to the testing, the load curve is reduced as follows. - The reduced load curve (in W / W) is multiplied bin by bin to the number of hours ; this gives the repartition of the amount of the seasonal energy needed (Wh / Wh or % x h / h) to cool the building, by temperature bin of 5 F (Fig.3). Figure 3. (ARI, 998) load curve (L % W/W) and cooling needs distribution by 5 F bin L % [W/W] %-hours, plain D C B A T C T B T D T A L % [W/W] striped 2 Ti ( F) integration intervals are then defined with 4 temperature limits (above 35 C, the lower temperature born is set to the temperature below which no cooling need occurs, the two

6 intermediary temperatures can be chosen by an iterative procedure to find at last the 5 and 75 % researched testing points) (Fig.3). - The sums of energy needs of each bin within the 4 intervals give 4 weighting coefficients A, B, C and D (Fig.3). - For each of the 4 intervals, the energy weighted average temperature is calculated, which gives 4 temperatures T A (35 C by assumption), T B, T C, T D. - % load rate are determined by intersection on the load curve, (the crosses show these points, Fig. 3); the procedure can be iterated to find the required regularly spaced load rates. STEP 4) The chillers are tested in the 4 conditions calculated (load and temperature). STEP 5) Results of testing are weighted by the energy coefficients according to Eq. (5): IPLV = A.EER + B.EER + C.EER + D.EER (5) A B C Thus, the seasonal performance can be calculated with 4 points and not 8, as in ASHRAE 6. 5 METHODOLOGY DEVELOPMENT FOR EUROPEAN The methodology was created in order to reduce hourly load curves to N points and then to study the accuracy when reducing the number of testing points. It is hereafter illustrated for the constant air volume system and Milan climate hourly load curve, and N equal to 4. For an ordinary hourly load curve, because of the envelope thermal inertia and the occupation rate, a time delay between maximum solar gain and peak day temperatures appears; thus, load is normalised by the maximal load value, according to Eq. (6). LOAD L = (6) LOAD max Then, the T temperature and load L axes are binned. In each rectangle [T] x [L], load rates are summed and the resultant 2D matrix, MAP is normalised. If we note: - T(k), the array of bin temperatures, with k between and K, the temperature bin number, - L(j), the array of bin load rates, with j between and J, the load rate bin number, MAP(K,J) gives the distribution of the cooling energy needs; it is represented for a 5 % L bin and a 2 K OAT bin on Fig. 4. The linear trend between load rate and temperature clearly appears. Figure 4. MAP matrix, temperature bin 2 K, L bin 5 %, Milan, CAV system D

7 In order to extract that information, we define equivalent load rates and energy weighting coefficients for each temperature of the T vector, as follows (Eqs. (7) & (8)): with: J MAP(k, j).l(j) L eq(k) = in [W / W], relative to T(k) (7) E(k) J j= j= E(k) = MAP(k, j) relative to T(k) (8) At this point, hourly information has been transformed under a form comparable to the one supplied in the American standards (Fig. 5): - A load curve (L eq as a function of T) - An energy weighting curve (product of hours of operation by the load rate energy as a function of T). Figure 5. L eq and E as functions of T, temperature bin 2 K, Milan, CAV system Leq (%) [W / W].9 E(k) [W / W] OAT ( C) The repartition of energy weights is clearly gaussian shaped, while the equivalent load curve is nearly a straight line, as guessed in American standards. However, this is not the case for other central systems such as VAV, or CAV with free-cooling. Since 4 regularly spaced steps are searched, the IPLV methodology would require the iterative process described above. Moreover, another problem arises: the ARI load curve suits well the 35 C OAT rating point at full load. But how to choose this temperature limit for different European climates? In order to get around both problems, we try to determine an average set of temperatures corresponding to the required load rate, 25, 5, 75 and %. To that extent, we define 3 supplementary arrays, solutions of the problem: - L N = [25 %, 5 %, 75 %, %], the final load rate array, - W N, in Wh / Wh, the corresponding weighting coefficients, - T N, in C, the corresponding testing temperatures. Then, we assume that a virtual chiller with N stages, here N=4, is cooling the building: for each of the couples (L eq (k), E(k)), the chiller operates its stages to supply the required cooling load. Equation enables to find the time the virtual chiller has to operate on each of its stages. Thus,

8 the energy weighting coefficients E(k) can be allocated among stages. This splitting is done directly, to calculate W N according to Eq. (9a) and (9b), and by temperature value T(k). If L eq (k) lies between L N (n-) and L N (n), Eq. (9a, 9b, a and b) are used. If L eq (k) lies below the smallest step value L N (), then E(k) is just summed to W N (). W N (n)=w N (n)-e (k)*(l N (n-)-l eq (k))/(l N (n)-l N (n-)) W N (n-)= W N (n-)+e (k)*(l N (n)-l eq (k))/(l N (n)-l N (n-)) (9a) (9b) To determine corresponding temperatures, an intermediary matrix W T (n,k) ( n N) is filled with Eq. (a) and (b). If L eq (k) lies below the smallest step value L N (), then E(k) is just summed to W N (,k). No correction for cycling is taken into account. The splitting of weighting among stages is represented Fig. 5a, for a 4 step chiller and temperatures between 9. and 3. C. The disc surface corresponds to the weighting of each stage at T(k). W T (n,k)=w T (n,k)-e (k)*(l N (n-)-l eq (k))/(l N (n)-l N (n-)) W T (n-,k)=w T (n-,k)+e(k)*(l N (n)-l eq (k))/(l N (h)-l N (n-)) (a) (b) The representation of PT is given Fig. 5b. The resultant temperatures T N (n) are calculated as the energy weighted barycentre of the temperatures T(k), for each L N (n) stage, Eq.. K W (n,k)* T(k) T N (n) = in [ C] () W (n,k) T K k= k= T Figure 5a & b. (a) virtual chiller method of weighting stages; (b) W T matrix, Milan, CAV system.2 9. PT, per stage and OAT bin.8 WT, energy weight L % % 5 % 75 % % OAT ( C) STEP 4) Finally, the chillers will be tested in the N sets of conditions [T N (n),l N (n)]. STEP 5) is computed according to Eq. (2), N = W ( n) EER( n) (2) n= 6 METHODOLOGY VALIDATION N

9 In order to test the accuracy of the methodology, we compare the relative bias between the seasonal performance calculated on an hourly basis, Eq. (2), and based on the reduced representation (), and the relative uncertainty of measurement. 876 CC( i) i= = 876 (2) EP( i) i= The compounded uncertainty resulting from the testing of the 4 points and their weighted sum is majored by its square norm, according to Eq. (3) EER A EERB EERC EER D A. + B. + C. + D. (3) The EER experimental uncertainty for a single testing point is majored by its square norm with tolerances on individual measurement as specified in EN45 (24). 6.2 Comparison between hourly and 4 points The chiller is sized to match the maximum capacity for the coincident outside air temperature. An iterative process is used, in case a lower load but higher air temperature point could occur; in that case, the capacity of the chiller is consequently increased. Once sized, efficiency can be calculated from modelling, for varied OAT and load rates. The reduction methodology has been applied to the 2 available load curves (4 systems and 3 climates). Seville CAV-FC and London CAV load curves are both extremes as far as weighting coefficient distributions and temperatures are concerned. Results of seasonal efficiency comparisons are summarised in Table 2. Hourly seasonal performance, H, and 4 points, are compared with and without cycling. The 4 points uncertainty of measurement appears for comparison. Moreover, from the point of view of certification, the reduction methodology should ensure that H values, which translate field seasonal performances give the same efficiency merit order than 4 points. Thus, efficiency classification of chillers are also reported in Table 2. No cycling Cycling Exp. Unc., EN45 H sequence E sequence EER sequence Table 2. Validation of the methodology, chiller types according to table Conditions CHILLER Type 3 Type 2 SEVILLE CAV-FC LONDON CAV Type 4 Type Type n 5 EER H Bias.9% %.7% % 2.% %.% %.8% 4% H Bias 2.3% 5% 2.% 6% 2.8% 4% 3.% %.4% 6% 2.7% 6.% 2.6% 5.8% 2.9% 6.3% 2.6% 5.6% 2.5% 5.3% Whether cycling is not considered in the hourly calculation, and H fit for both climates except for the chiller n 5. The bias introduced is, for both load curves; inferior or

10 approximately equal to the uncertainty of measurement. When cycling is taken into account in the hourly calculation, the bias increases a bit for Seville and more for London. Why? For the Seville CAV-FC load curve, weighting coefficients are mainly on 5 and 75 % load points. On the contrary, for London CAV curve, 25 % load point is important. The 4 points method does not take into account the efficiency behaviour below 25 % and is thus penalized for the coupled - important weighting of very low loads- and - chillers with efficiency drops at low loads (below 25 %, see chiller n 5, Tab. ). On that point, the methodology can be improved by adding a point at lower loads (5 or % for instance), taking care of not considering too low load rates since it can make the final value highly uncertain. 7 CONCLUSION An easily reproducible methodology to reduce chiller hourly load curves has been established. It has been applied with a specific number of four points. However, the number of points can easily be adapted, for instance, to better take into account important degradation of performances at low loads. This method was used to compute the average seasonal performance index, E (E as Eurovent) for Eurovent (EECCAC, 24). It is also to be noted two important points about seasonal performances (Tab. 2): - EER alone cannot translate seasonal performance: the observation of efficiency indexes sequences shows that EER is really a poor indicator of chiller average performance and that part load should be considered, - Chiller specificity: for different load curves, depending on nominal EER, temperature impact on efficiency and part load behaviour, one chiller can be more adapted than another one, - Despite this fact, chiller n 4 ranks always higher than the four others (Tab. 2); it highlights the interest of ranking chillers on a that would translate average conditions. REFERENCES Anglesio P., Caon S., Caruso S., 2, Determinazione delle prestazioni energetiche di condizionatori elettrici a due unità in aria invertible, CDA, febbraio. ANSI, 995, ASHRAE Standard 6 - Methods for Rating Seasonal Efficiency of Unitary Air Conditioners and Heat Pumps. ARI, 998, Standard 55/59, Water Chilling Packages using the vapor compression cycle. ASHRAE, 996, HVAC Systems and Equipment Fundamental Handbook. Bettanini, E., Gasdatello, A. and Schibuola, L., 23, Simplified models to simulate part load performances of air conditioning equipments, Eighth International IBPSA Conference, Eindhoven, Netherlands, August -4. EECCAC, 24, Energy efficiency and certification of central air conditioners. Final Report. ARMINES for DG-TREN, the European Commission. EN 45, 24, Air conditioners, liquid chilling packages and heat pumps with electrically driven compressors for space heating and cooling. Pérez-Lombard, L., Adnot, J., Ortiz, J. A. and Rivière, P., 24, HVAC systems energy comparisons for an office building, Proceedings of the Climamed conference, Lisbon. Rivière, P., 24, Seasonal performances of chillers, Ph.D. diss., Ecole des Mines de Paris, Paris, France. Rivière, P., Flach-Malaspina, N. and Lebreton, J., 24, A new installation for part load testing of air to water single stage chillers and heat pumps, International Refrigeration and Air Conditioning Conference, July 2-5, Purdue University, West Lafayette, USA, paper R86. Roujol, S., 23, Methods of prediction of the energy consumption of air conditioned buildings uncertainty and validation, Ph.D. diss., Ecole des Mines de Paris, Paris, France.