‘THE CONTROL OF VARIABLE SPEED PUMPS IN PARALLEL OPERATION by Anthony B. (Tony) Crease Adviser, Rotating Equipment Engineering Lagoven S.A. Maturin, Venezuela Anthony B. (Tony) Crease graduated in Mechanical Engineering from Imperial College, London (1966), and obtained a Master's degree in Tribology from the University of Leeds, England (1968). fer ‘wo years in the English Electric Company, he spent nine years with Michael Neale and Associates as a Consulting Engineer, ‘working on a wide variety of tribological and mechanical engineering problems for private and public industry. In 1979, he Joined the Rotaing Equipment Engineering group of Lagoven, one ‘of Venezuela's three major oll companies, becoming the company's Rotating Equipment Advisor in 1985. He is involved in all aspects ‘of roxating equipment ownership, from specification and selection, through commissioning, troubleshooting and maintenance, to per formance, reliability and economic assessments, relating t0 over ‘one milion horsepower of major machinery, including pumps and ‘centrifugal compressors and their gas turbine and electric motor drivers, ABSTRACT ‘The features aze discussed of a PLC based contol scheme that was developed for a multipamp pumping station 10 be used to supply ol field produced liguid (crude oil and associated wate) from tanks through a dehydration facility 0 dewatered crude storage. The hydraulic resistance characteristic of this system ‘exhibits substantial variation andthe required flow is variable and ‘covers wide range from inital statup conditions to the eventual peak capacity requirement. The avoidance of unnecessary energy Input to the product was of special importance, since it would promote the formation or stabilization of emulsions of the Produced fluids, thereby complicating the dehydration process. The contol scheme developed is applicable to any variable-speed. parallel pump station. ‘With the optimum pump number fixed in this way, pump speed is simply controlled 40 provide the desired overall How. As a minimum, values of Q and N (or excitation frequency) have to be ‘measured and input othe control scheme. The individual values of (QIN, besides providing the basis of pump number optimization, also’ permit protection against excessively high or low flow. ‘Additional practical aspects are discussed. INTRODUCTION ‘The potential benefits of using variable speed drives for cen- ‘vfugal pumps, including flexibility and the reduction of energy consumption, have been widely presented (for example, Murphy [1). Variable frequency drives and variable speed drives (hydraulic couplings, etc.) may be considered for a new applica- lion, and the benefits will have to be weighed against the higher initial cost and complexity ” ‘The author's company recently adopted the use of variable frequency drives for & station using multiple horizontal centrifugal ‘pumps to supply ol field produced liquids (erude oil and associat fed Water fom tanks, through a dehydration facility, to dewatered exude storage. The main factor inthe decision rouse variable speed in tis case vas the need to minimize unnecessary energy input tothe pumped fluid which is particularly susceptible to the formation of stable caude/Wwater emulsions, thereby rendering more difficult the down: Stream dehydration process in electrostatic precipitaiors. Energy cox savings and flexibility were important additional benefits of the variable speed drive, ‘The Wydraulic resistance characteristic of the system varies considerably due to tank level and fluid viscosity changes and the required flow varies from $0 to 300 mbpd (1460 to 8460 gpm) lover the project life, The range of system resistance is shown in Figure 1 HEAD [ff] 300 No.of pumps | 4 3 200 ly 7 if 100 7 i] - it Tax. eventual it 77 Fequirement 7 ‘mak initial requirement min, requirement 0 , ~—___+__| © — 2000 4000 6000 8000 10000 FLOW [gpm] Figure 1. System Resistance Characteristic and Pump Performance Envelopes for Speeds Between 800 and 1800 RPM ‘and QIN Berween 50 and 120 Percent ofthe BEP Value 1c was elected to use parallel pumps: 4 + 1 for the eventual ‘maximum requirement and three for the inital operation, The8 PROCEEDINGS OF THE THIRTEENTH INTERNATIONAL PUMP USERS SYMPOSIUM eventual maximum requirement determined the basic headifiow capability needed from the porps and the envelope of perfor- ‘mance a variable speed from 45 o 100 percent of nominal for one, ‘wo, three, and four pumps is shown in Figure |. The eletion of parallel pumping is in line with a published criterion (2) chat rec- fommends the use of a parallel rather than a series configuration when the fixed part of the system resistance (geodetic heady is ‘more than half the total (fixed + dynamic) resistance. Having selected the basic pumping scheme and pump type, attention was tured to the requirements for automatically control ling their operation, for inclusion in the project specification fora PLC based control system to allow unattended operation if necessary ofthe pumping station The general criteria and requirements 10 provide optimum ‘operation of parallel, variable speed centifugal pumps, for appi- cation inthe above mentioned project, were evaluated. The results and conclusions of the evaluation are described. FLOW CONTROL IN VARIABLE SPEED PUMPS ‘The main objective of the contro scheme is to vary the total Flow of the pumping station to match system or process require ments, without the use of control valves. This is achieved by ‘varying the numberof units in service and thei speeds ‘An important aspect of the control scheme is to ensure that the pump units operate within a safe continuous duty range, as deter- ‘mined in conjunction with the manufscturer. The envelopes in Figure 1 indicate these safe ranges for one to four pumps. In addition, as illustrated by point A in the Figur, criteria are needed to choose the best combination of pumps for a given overall duty, between two, three or more alternatives that satisfy the basic requirement of having the pumps run in the acceptable range. DEFINITION OF THE SAFE CONTINUOUS DUTY RANGE, ‘The continuous duy range of a variable speed pump may be considered as the range of conditions that can accep without sulfering an unacceuable adverse effect on unit reliability. This ‘ange is normally welletablshed by the pump manufacture fr ‘opertion at nominal speed and is then simpy the usable pump flow range as determined by high and low flow cavitation or in some cass by vibration bearing loads, ete. Guidance on cavitation limits is now avalable from many sources (2) and. special, attention ta precise evaluation of minimum flow required for high specific speed tod high energy pumps, where the ow range may be very resrictea An accurate definition ofthe safe conin- Gus duty flow range at nominal speed should be obtained In onsultton withthe pump manufacturer es part ofthe evalaton of any pump fora vaable sped application Ate bs fcecy pe oft pnp (ep, Now ee he pump with an optimum ineence angle ono the impeller vanes, At Finer and lower flows the incidence angle deviates from ts ‘optimum valu, lading eventaly to flow separation turbulence, flow induced vitraton and cavtton. At the seme tne trust and radial bearing loads, which are designed tobe minimized tor neat the best efficiency point, increase progressively asthe percentage ‘deviation of the flow from bep increas 'At diferent speeds thn nomial, the ow incidence angle onto the vanes wll Connie tobe a fundameneal factor inthe occur fence of flow separation, turbulene and cavitation, ‘Sinee the incidence angle is directly related tothe normalized flow, QIN it ‘is reasonableto assume tat the useble flow range will continue to be determined by the maximum and minimum limits of QN {tablished for nominal speed operation, For reduced speeds, this Sssumption is conservave inthe sense tha, a given QM, all forees and pressures inthe pump vl end to reduce with speed ‘On the contrary, at increased speed, for a given QIN, pressures and loads on a pump wil increase. Thus speeds sign ‘cantly higher than nominal should be used with caution snd only afer consultation with the manufacturer. Avoidance of Critical Speeds Weis partially imorant, when applying variable spent pump, to eny with he manufac, he torsional nd lateral Creal speds ofthe equipment and to ensre that the chosen Speed cage ofthe pomp exclude all lateral creas wit an Aquat separation arg. Except for special pplicions where the cost of detailed eng ering evainton can be jose i desable to Ose, fr ‘rile spec epplietions, pimps, nd motors that rn below the fs lateral erieal sped. This permits he uns tobe a any Speed upto maximum witout cance Single tage pumps of ha iontal or vertical ‘nine type, using electic motor driver nominal frequency or below, will generally meet this sim, ‘cording to APL GIO, Mh edfon, paragraph 281 On the ter hand for molistage, veel tune and high sped pumps. 4 eed otordynamics analysis necessary to lest the “ne” trite speeds a funtion of clearances, an to determine ro response the cial, as 8 bast for determining the tkesble Speed range Daring the purchase of variable sped pumping vis, ead and transient torsional analysis should be spel pee parsraps 2RRA and 289.5 of APL 610 Pump set tha ute det lsc ‘oor dive ar uty to present sonal problems tht cannot te resolved bys torsional analysis and proper selection of oping sass, to move eal aay from exetation uence. (nthe other hand geared pumping sits with mslpe ors are particulary sensitive to torsional problems, due tothe many nical speeds and exction frequencies, and twill often be imposible to avoid “iterences” In such cates, it ill be necessary © limit che torsional sesseswhich result when ‘peratng righ on one or more resonances. rhe infomation is !ailaNe inthe paper by Pele a. (3, who mention he value of Alamping clement (Aesible eerey-absorbing couplings, damper’) insted inthe ie sha. a8 4 very eflen id reducing tomona stresses Teahoul be noted hat, nthe cae of variable frequency drives, toslonal excation of 5.00 30 perent of bate ore pending onthe VED switching tecnlogy is penrated a equeny of 60s ate frequency [1], with lower levels ofexcaton a higher tulle of €0 x base fequency. Is there pareualy Important to avoid torsional etal atthe 60 » frequency a8 & Mechanical Seals ‘The possible influence of reduced speed on mechanical seal ‘operation and reliability should be reviewed. There are two main ceancems. The first is thatthe ratio of sealing pressure to speed, ‘which is what determines seal lubrication, may increase above the ange for which the seal was designed in nominal speed operation ‘This can happen ifthe sealing pressure is maintained at reduced speeds, due to being balanced 0 close to suction pressure. “The second concer is thatthe Seal temperature may rise, due to the seal Mushing flow falling more rapidly than the seal heat gen- eration. This is of panicular concer when handling light hydrocarbons or other Muids with high vapor pressure ‘To ensure a satisfactory seal selection and design, these aspects should be evaluated, in consultation withthe pump and seal man- ufactuers. Drive Motor ‘The final component whose behavior has to be considered for variable speed operation is the driver motor. In the case of VEDS, a deraing factor must he applied when selecting the motor to allow‘THE CONTROL OF VARIABLE SPEED PUMPS IN PARALLEL OPERATION 9 for higher heat generation as a result ofthe iregular shape of the ‘waveform supplied by the VED. For reduced speed operation, there is no real concern since the pump torque at any given QIN reduces in proportion to N In the case of VFDs, the maximum output torque available from the ‘motor stays practically constant down to at least 20 percent torque [1], thus ensuring ample power at reduced speeds. In the ease of VSDs, even though the losses in the coupling increase in proportion tothe slip and are typically greater than the ‘pump power at 50 percent speed, the total power a the fixed speed ‘motor output shaft will ypicaly be down to 25 percent, when the coupling output speed is down to 50 percent, so that once again reduced speed operation presents no problem forthe drive. ‘With VEDs, motors can be driven at speeds above nomi subject to the mechanical limit of the motor. For such applications, the drive and motor manufacturers should be consulted to obtain 8 ‘proper Selection and the respective Speed and power limitations. With due attention othe various aspects mentioned, including 2 proper selection of seal and flushing arrangement, an ample con- tinuous operating range can be extblished. In particular, direct sven electric pumps can normally run safely up to atleast the ‘nominal speed, with QIN limits equal to those established for ‘operation at nominal speed. Limits based on speed and QIN can readily be incorporated ina PLC based control scheme to protect, the pumps during operation. In addition, criteria must be established for minimum accelera- tion rates while bringing the pumps on fine. A balance must be found berween disturbing the process ifthe pump is brought up very quickly and running a risk of damaging the pump by spending to long at very low flow if it s brought up too slowly. A stepped ‘ampup schedule that brings the pump quickly up to a least 10 percent flow to avoid any heating problem and then more slowly ‘upto the same operating point asthe pumps already in service, was ‘adopted in the application of this study, Soundly based guidelines for rampup schedules do not appear to be available at this time, and Will have tobe developed as experience is acquired. CRITERION FOR OPTIMIZATION OF PUMP OPERATION, ‘The useable range of QIN for pumps other than those of high | cnergy’of high specific speed, is typically $0 to 120 percent of the bbep value or even more. This means that given overall pumping condition of a parallel pump station can generally be satisfied by various different combinations of pumps, from a small number a high speed toa larger number at lower speed, as was illustrated in Figure I. In fact, the pumps do not even have to be identical nor run at the same speeds the only strict requirement is that all pumps shouldbe able to make, with Some margin of safety, the maximum regoiced head under any possible operating condition when running at maximum speed aad minimum allowable QIN. ‘This eases the question of how to choose the pomp number and speed for a given overall ity, and how to specify criteria for automatic increase and decrease ofthe number of running pumps. For the application considered in the present work, it was decided to use minimum total energy output (0 the Muid (that is minimum pump shaft power, not minimum electrical power inpot) 45 the fundamental exitrion for developing 2 contol strategy. For applications where the main concern is minimum energy con- sumption by the driver rather than the pump, the fundamental criterion would be slighty differen from the criterion adopted here, due to the variation of driver efficiency with power and speed. However, it could be argued that ignoring the increase of diver losses with reduction of power and speed, would be sub- stantially offset in an economic evaluation, by the lower ‘maintenance cost forthe driver and pump. This would make the criterion of minimum pump power applicable toa wide range of applications, In any case, the conclusions derived from the present study would often give an acceptable approximation to minimum driver paver consumption and the procedure developed inthis study for determining. when to change pump number, could readily be ‘modified to better reflect a criterion of minimum input power. ‘The enteron of minimum pump power is used, in the next section, 10 show how the operating points of individual pumps of different characteristics must be adjusted 0 satisfy the criterion. In the subsequent section it is indicated how the eptimum number of pumps of nominally identical type may be determined, CONTROL OF OPERATING POINT OF INDIVIDUAL PUMPS If all the pump units of a parallel group are identical, and by design they normally willbe, ican be shown that minimum total pump power is obtained by running all the pumps atthe same ‘operating point. This requirement is not always immediately obvious to those involved. It is sometimes argued for example that it might be preferable to have one or more pumps operating at best efficiency And one operating less efficiently at lower flow to achieve the total required flow. Iti, therefore, convenient to present the reasoning for running identical pumps at the same point ‘A:second subject of intrest is thst of where t0 operate pumps ‘that are nominally identical but have somewhat different perfor mance due to deterioration in service. Of various options for controlling the operation, such as ofthe basis ofthe same speed, (or the same flow or the same value of composite parameters such as QIN, it is interesting to try to identity a best solution ‘Thirdly, although not often of practical concern, itis also of inerest to consider the criterion that should be applied to the control of variable speed parallel pumps with completely diferent ‘operating characteristics; tnd sine ths evaluation should provide answers tothe two previous points it has been considered here. Optimum Operating Points of Parallel Pumps with Difierens Chractrisics For a specified total differential head, the performance of variable speed pumps can be described by their curves of power, P vs flow, Q, which can be readily generated from the multispeed performance map, as illustrated in Figure 2. ‘The resulting P-Q curves for a pump of the type used in the present project (pump B) and for a second pump (pump A) of Tower efficiency, lower capacity and a steeper curve, operating in parallel against particular head. (170 ft), are show in Figure 3, ‘with axes interchanged to plot flow against power ‘This figute illststes how the operating points of the two ifferest pumps mus be adjusted to minimize te total power con- sumption for'a given total low, Qt It can be shown that the minimum power condition, is found by operating the two pumps such that: + the slopes atthe operating poats on each Q-P curve are equal, that ss (AQIUP)q a Qq = (AQP) Qp. (Note that he vals of {4Q/€P at any power can be derived deel from the curves of Q vs Pand have been aded t Figure 3) + the average flow, (Qu + Qg)2s is equal to half he required total ow, 2. ‘This result is logical inthe sense tha, if certain total flow is required, a small change in Qn. the Sow of pump A, would have 10 be accompanied by an equal but opposte change in Qp. If the ‘change in power for pump A, determined by the sage of the Q-P ‘cUNe for A, were different from the change in power for B, deter- ‘mined by the slope of the Q-P curve for B, the teal power would change, implying that a lower total power could be found at some100) PROCEEDINGS OF THE THIRTEENTH INTERNATIONAL PUMP USERS SYMPOSIUM. cer combination of Qy end Qy capabe of giving Qt. It follows that the point of mitmum total power, the slopes at the two ‘operating points must be equal HEAD [ft] 300 250 200 = = 150 © 100 2 50 a 150 100 50 o ° 1000 2000 3000 FLOW [gpm] Figure 2. Pump Performance at Variable Speed and Constant Head FLOW [gpm x 10*3]_ dQ/dP x 100 35 @ === caer] © = Points oftangency = = 170) 0 so [Pal 400 POWER [hp] Figure 3. Performance of Diferent Types of Pump, A & B, in Parallel (Construction 10 Deiermine Optimum Operating Points. Equal Marginal Efficiency As the Criterion for Optimum Operation Another way of stating the above is to say thatthe optimure operating, points are found where the incremental or marginal ‘increase of low per unit of power isthe same for both pump or, in terms of calculus, the optimum operating points are Where 4Ql4P isthe same for both pumps. The parameter dQIAP could be termed a marginal efficiency factor and is relate to, but different from, the norte! pump efficiency as indicated in te following Pump power, P= kH-Q'y w where: onstant forthe unit system employed n= efficiency Reerranging: QUP = 1ykHt a Thus, QPP at constant head is proportional tothe efficiency, as normally used, and may be termed the “normal efficiency factor.” Tris the flow delivered by @ pump per unit of poWer utilized, Similarly, o marginal efficiency factor, dQlAP, equal to the incre mental flow delivered per unit increment of power utilized, may be ‘obiained by rearranging Equation (1) as: Q=Pq/kH 8 and differentiating Qiai fn + PanyP) KHL a Thus, dividing Equation (4) by Equation (2), we get 6QI4P, the marginal - slope of the Q-P curve atthe operating point slope of the line from the ‘operating point to the origin, sficiencyfactor (QM, the normal efficiency actor + 40/0 = 1 + percentage increase of efficiency per percentag L 4/9 1 + percentage feficieney per percentage increase of power. Since the percentage change of efficiency per percentage increase of power is positive at points below bep, falls to zero at bep (where the efficiency is at its maximum value) and becomes negative for points beyond bep, it follows thatthe marginal effi- ciency is equal to the normal efficiency at bep, is higher than the normal efficiency at flows below bep and i lover than the normal efficiency a flows above bep, This is alto evident geometrically from the graph of the Q-P curve, when one observes thatthe slope of the tangent from the ‘origin tothe Q-P curve, which defines, atthe point of tangency, the ‘ep point, is both the normal efficiency factor (QP, the slope of the Tine from the origin to the operating point considered on the Q-P curve) and the marginal efficiency factor (4QI¢P, the slope ofthe ‘QPeurve atthe operating point considered; that isthe two facts ‘are equal. At operating points below bep, the slope of the curve is. {greater than the slope tothe origin and at opersting points beyond bep the slope ofthe curve isles than the slope tothe origin, This point is further illustrated in Figure 4, where the normal and ‘marginal efficiency factors are shown plotted ageist flow tis interesting to note thatthe optimum situation, fo a required total flow equal to the sum of the best efficiency flows of two clferent pumps, is not to have each pump running sits best efi- ciency point. Instead, the lower efficiency pump should be run below its bep and the higher efficiency pump above its bep, ‘In terms then of the marginal efficiency factor, the general criterion for minimized total pump power for pumps of any type ‘operating in parallel with vatiable speed, is that all the pumps should run with the same value ofthe marginal efficiency factor‘THE CONTROL OF VARIABLE SPEED PUMPS IN PARALLEL OPERATION to 40 = Marginal Efficiency e Factor, d@/dP 5 30 é 20 ; @ 40 | [Normal Efficiency 2 Factor, UP e o PUMP B o ° 4000 2000 3000 FLOW [gpm x 10*3] Figure 4. Normal and Marginal Efficiency Factors vs Flow. Determination Of Optimum Operating Points Of Pumps Wish Different Characteristics In general, the optimum combination for a given pumping requirement, defined by the values of Qt and head H, could be determined by developing for each pump and bead H, the cures of the marginal efficiency factor, dQIGP, as shown in Figure 3. Then, for each value of €QIGP, the coresponding total flow could be cbiained fom (Qq +-Qp), andthe speeds required of pumps A and B would be deduced from the values of Qq and Qa respectively, togetier with the kaown valve of head, H. Thus, by repeating the exercise for various values of dQUP and H, a matin of required Speeds is obuaned that could, at leat theoretically, be used to Provide optimum control ofthe two different paral! pumps. Operation of Identical Pumps in Parallel “Turning now to the common situation of pumps that are identical for practical purposes, it follows, from the general conclusion that the optimum combination of operating points for two pumps is ‘obtained where the values ofthe marginal efficiency factor, @QIAP, for each pump are equal, that the two pumps must be run atthe same point (same speed, low, etc.) since the Q us P and hence €QUdP vs P curves, for identical pumps must be identical. It follows that, contrary to what is often assumed, its not con- venient to operate a group of identical, parallel pumps with all bt ‘one at best efficiency and the remaining one controlled to match the total flow tothe level desire. Effect of Performance Deterioration on the Method of Operating Point Control ‘The question of the best way to control nominally identical ‘pumps, of which the performance of one has been affected by dete Fioration inservice, is complicated by the variety of deterioration mechanisms and the different effets these may have on the pump performance curves [4]. A comprehensive evaluation of the possible effects and their implications for optimum pump contol Was outside the scope of the present study. Nevertheless, some guidance was required on a specific question, namely: what parameter of the running pumps should be controlled to be the same forall of them. With all pumps identical, the selection is immaterial, since any parameter would serve and the easiest to measure could be adopted. Inthe face of possible differences of performance due to deterioration, the question canbe reformulated as: what parameter, Of those that could be determined with reasonable precision by simple measurements, would, when made equal for both pumps, provide the closest approximation tothe desired situation of equal {4Q{4P for both pumps, The curves of 4QIEP, for a new and deteri- rated pump, are shown in Figure 5, ploted now against Q instead OF. To obtain the curve of the deteriorated pump, constant reduc tions by 10 percent of bep values are assumed for head and efficiency over the whole flow range. A similar exercise was also cared out forthe case of 10 percent reduction of head with seven percent reduction of efficiency, and gave very similar results Based on the range of deterioration mechanisms discussed by ‘Yedidiah [4], the above two cases give a reasonable representation of the range of possibilities Figure 5. Effect of Deterioration on the Marginal Efficiency Factor vs Flow Curve of a Pump, ‘The curves are marked up to show points of equal QIN and speed on the (wo curves, The results show that coatrol of the ‘pumps fo equal QIN isnot a good solution since it results in arel= atively large difference of dQ/aP between the two pumps. Control ‘o equal speed or flow would give closer and similar, though not complete agreement. Control to equal speed is generally preferable ue to the ease and relative accuracy with which speed can be estimated from the excitation frequency in a VED. Pumps with "Plat" Curves ‘The exception to the previous point would be with pumps hhaving curves which aze very flat at the lower end ofthe reliable ‘operating range, as was the case for the pumps selected for the application ofthis paper. In this case, control to equal speed could well result in unacceptably large differences of flow between the ‘new and deteriorated pumps, a5 may be clearly seen at low flow in Figure 5. For pumps of this type, control to equal flow is prefer able. As an arbitrary criterion of “Ala,” the following i suggested: Determine forthe pump, the slope, s, of the head ~ low curve at ‘he minimum continuous flow point from: 8= -(UHVSQV Hye / oe) © If sis less than 0.25, the curve should be considered flat and pumps should be controled on the basi of flow rather than speed,102 PROCEEDINGS OF THE THIRTEENTH INTERNATIONAL PUMP USERS SYMPOSIUM, In practical terms the criterion signifies that a given percent drop. in head ata speed would result in a percent change inflow four times higher, for example a 3.0 percent head deficiency would resultin a 12 percent drop inflow. ‘Control r0 equal values of flow requires that each individual pomp flow must be measured. As aside benefit, knowing both the individual flows and speeds allows the relative deterioration between the running pumps to be continuously monitored, and in the event the difference between the worst and best ofthe running jumps becomes greater than some specified criterion, an alarm can be incomporated in the control system to alert to this situation. Its recommenced that in cases of doubt for any specific appli cation, curves of marginal efficiency function, 4Q/P, be generated as a basis for determining which parameter (0 use for conte. ‘Theoretically, a beter approximation to maximum overall effi ciency could be obtained by controlling to equal valves of a parameter ofthe type QN? where x could be found from the curves ‘of dQMP as the value which results in a best approximation 12 ‘equal dQUEP, and appears co be around 2.0 or 3.0. In practice his Would seem to be an unnecessary refinement. DEFINITION OF OPTIMUM NUMBER OF IDENTICAL PUMPS FOR AGIVEN OVERALL DUTY Applying the criterion of minimum pump power, the optimum ‘umber of identical pumps running in parallel at the same Condition isthe number that gives lowest tea power Pr for a given ‘overall flow Qt and resulting head H. In other words, the optimum ‘number is that which gives highest overall efficiency, k-H-QUPt (where kis the appropriate constant forthe unit system adopted) Now individual pump efficiency isthe same as overall pumping efficiency since: Individual pump efficiency =k H-Q/P=k Hin Qin.P=kHQUP += overall pump efficiency. So the optimum pump number is also that which results in highest ndividusl pump efficiency. ‘The question then is how to determine the individual pump eff- ciency for any given overall duty (Qt and H) and for different ‘numbers of pumps. For constant speed service, the efficiency is simply the well ‘knowin funetion of flow. For variable speed, assuming the affinity laws apply, efficiency can be expressed as an approximately ‘unique (tha is, speed-independent) relationship against QIN For the particular numberof pumps in service at any moment, it would be an easy matter to determine the QIN by direct measure- ment of Q and N. For more or ess pumps in service atthe same ‘overall duty (Qt and H), the individual pump flow could also be ready determined from the (measurable) total flow and the pump number. However, to obtain the requited value of QIN, the expected speed with more of less pumps would sll be required and it snot immediately obvious how this may be estimate, Instead, the efficiency and corresponding values of QIN for ferent pump numbers and overall duty, may be deduced directly from cures of efficiency vs H/QE, for each pump number. The reasoning isa follows: For an individual pump, assuming the affinity laws apply, the efficiency and normalized head, h (HIN), are functions of QIN, as shown in Figute 6, Now: H=bNe o and Q=an o From (7): Sag © From (6Y(8): HUQ? = hig? o x F 9 (he HIN'2) EP 3 = x se | EFFICIENCY i} e #3 » Ad r= high? ae = ¢ o Os 1 15 = QIN [gpmirpm] Figure 6. Pump Performance Expressed in Terms of Normalized (Speed: Independent) Parameters. Using the symbol for We? note that since is 9 unique function of, then rnst also bea unigu function of gn since the efficiency is function of g, it can also be presented as function of, with he advantage that ony depends on tbe ove dit an the number of pumps. That is r= HQ? = HQ?) = 2 WIQE. This signifies that it must be posible to present tke ef. cienty os afaneion of FUQ? andthe numberof pumps n, wich ‘will provide a convenient way of selecting the OpGmum pump number for a given duty “The various functions of q of interest canbe deveined inthe following way, using the subject pumps diseused herein as an example From a table of several values of bea and efficiency vs ‘fora given speed, the corresponding values canbe obtained of h and efiieney vg” These values ean then e fied by polynomi- als to provide smooth continuous funcions of q. Using these finetions, a table such as Table I can te developed using a simple spreadsheet program ‘Table 1. Data Development for he Curves of Eficiency and QIN ws HOR, ‘The values of efficiency and g can then be plotted against H/QE for diferent numbers of pumps using the relevant columns of the ‘able, 25 shown in Figure 7. The intersection ofthe curves of effi ciency vs H/QE indicate the points where a change of pump ‘number is required to maximize pumping efficiency. Tn practice itis desirable to maintain a pump combination in sefvice until the addition or removal of one pump would result in an iinprovement of the efficiency by a small amount (chosen to‘THE CONTROL OF VARIABLE SPEED PUMPS IN PARALLEL OPERATION 3 EFFICIENCY % 80 10 43. 4 NO.OF 60 \ eT a wi 40 10 ws Zz 20 g z g ° on 1 10 100 1000 HEADI(TOTAL FLOW)*2 [{ftigpm*2) x 104.6] Figure 7. Determination of the Optimum Number of Pumps for ‘Any Operating Condition (H & Qt) as a Function of the Terms F708 and QIN (Showing Change Points Between 1 and 2 Pumps) avoid unnecessary cycling of the pump number). A zoom of the region of changeover between | and 2 pumps i shown in Figure 8. ‘The soli line shows the situation during flow increase, wen the pump number has to be increased, and the dated line represents the situation during flow decrease, when the purnp number has to be reduced. Using a suitable value forthe separation, 2, between the change points, © obtain the desired hysteresis effect, the specific values of HUQ und g (=Q/N) eam be chosen and incorpo: rated into 8 PLC based contol algorithm, The control values which were determined for the application of this paper are shown in Table 2 Table 2. Parameter Values atthe Optimum Change Points. Disrat erent —| Ir should be noted that when the conucl system has to adjust ow to contol another process variable such a evel, it should be esigned and tuned to minimize the amplitude of flow Tisctuations associated with control of the process variable. In this way the sep aration, 2, between the pump number change points during flow increase and flow decrease, can be reduced, andthe potential effi- ciency loss nea the change point (areas w, for wast, in Figure 8) can be minimized Tn addition all signols that go fo the PLC for control purposes should have a damping function applied to them either a¢ the signal source or electronically in the PLC, to avoid undesired fluctuations of speed and starstop of pumps. ©= pump number change point 76 : 5 a 1.6 g vd & 68 7 . i & a ’ g ae + 0s 16 20 25 30 35 HEADI(TOTAL FLOW)*2, Figure 8. Zoom of the Optimum Change Point Between } and 2 Pumps. Alternative Pecameters for Defining the Pump Number Change Points Definition ofthe change pots in the control algorithm can be ‘on the bass of H/Qe, H/Q?, QIN, or (QUO), depending on what prameters are to be measured and fed to the PLC. The firs two can conveniently be transformed toa ratio of differential pressures that not only incteases the accuracy oftheir determination via wea surement and division in the PLC, but also makes the parameter independent of the specific gravity ofthe fluid. This isa distinct advantage in applications where the pravty may vary significant |y, as isthe casein the application of this paper due to substantial ‘aritions in the water content of the erude/water mixture. ‘The transformation ofthe HUQ? terms into a ratio of diterential pressures is achieved inthe following way: H= (ps Pos a an, using the equation forthe orifice employed for the flow @=CLA® 2g.49/0 ay where ifice Naw coefficient P= density of fd and Ap= differential stati pressure across the orifice, “Thus, dividing Equation (5) by Equation (6), we obtain: Hig =K (pap) AP a where K is constant If, as discussed in the previous section, individual pump flows are to be measured to enable controlling of individual poraps on08 the bass of flow rather than sped, (pg ~ p/p oF QIN could be ‘red a the parameter that determines changes of pump namber in the conta lgothm. The former woud be pefemed for applica: thos with vail gravity. ifthe pumps have a steep crv, the control of individual pump operating pots cold be done on the bss of sped equalization Sd change points could be derived frm the otal No, dsing (py Bava o GUNN Oe asin the former weal be pee ible gravity situa ‘CONCLUSIONS + For best over efficiency, identical pumps running in parle ‘hoald al be operated at the same contin; nok convenient 0 run some pumps at best efficiency with thers removed from the ep +The contol of aominaly identical, variable speed, parle amps tou include te folowing festurs: + (Protection of a reliable operating range using maximum (endif nesesary minimum) allowable speeds and maxim and minimum values of QIN. Con lam and shutdown ‘ales shouldbe incorporated foreach iit + Ifthe pumps have a flat curve (6 < 0.25), control all running fpumps to equal individual flow. Ifthe pumps have 8 step curve (5 > 0.25), control to equal speed. + Control the start and stop of pumps on the bass of prede- termined values ofthe parameters q = QIN. or (py ~ Pap, and include some measure of “hysteresis inthe valves (0 avoid cycling on and off of pumps when the required duty is, close to a change point. Use tte procedure discussed herein for determining the change values, Use (Pq~ pg(p asthe con- trong parameter. ifthe fiid gravity is subject o appreciable Variation. If not, either parameter mas be use. + Iitis chosen to measure individual pump flows (oe p's) as ‘well as speeds, monitor the differences between the individual Yalues of QIN (or VAIN) forthe running pumps and incorpo- rate an alarm to signal unusually 1 QIN relative tothe pump With highest QIN, in order to detect and warn of excessive pump deterioration + If ts required to operate pumps with different charactcrs- ties in parallel, a suitable control algorithm for thie purpose could be developed using the methodology indicated herein, Which is based on minimization of the energy input to the pumped fluid and translates into @ criterion of running the pumps at equal values of a “marginal” efficiency factor, the Slope at the operating point of the constant head, variable speed, Q-P curve for each pump. An additional basic require- ‘ment to allow nonideatical pumps to be run satisfactorily in PROCEEDINGS OF THE THIRTEENTH INTERNATIONAL PUMP USERS SYMPOSIUM. parallel is that each, tits maximum speed and minimum con- ‘invous QIN value, should generate head atleast equal tothe ‘maximum possible head which could arise in any possible ‘operating condition ofthe system, NOMENCLATURE QP Flow, power for an individual pump QPL Total flow and power for a number of pumps in parallel N Pump speed 6Q/dP Slope ofthe curve of Q vs P fora pump at constant head and variable speed H Differential head of a pump or a group of pumps in parallel 1 Efficiency of a pump Number of running pumps in parallel ho WN 3 N r hig? oP, Discharge, suction pressures of @ pump © Fluid density 8 Acceleration due to gravity 4p, 4p, Differential static pressure across an orifice for ‘measuring individual, total flow s slope of head-flow curve st minimum flow = @WGQ/M pers) REFERENCES |. Murphy, S.P, “Application of Variable Speed Electric Motors Tor Pumps,” Proceedings of the Tenth Inernational Pump Users Symposium, Turbomachinery Laboratory, Texas A&M. University, College Station, Texas (1993), 2. Sulzer Centrifugal Pump Handbook, 3rd. Ealtion, Sulzer Brothers Limited, Winterthur, Switzerland (1987), 3, Frei An Gre, A. Hell, W. and Luzi, A. "Design of Pump Shait’ Trains “Having ‘Variable-Speed Electric Motor Proceedings of the Third International Pump Symposium, Turbomachinery Laboratory, Texas A&M University, College Station, Texas (1986). 4, Yedidih, S., Centrifugal Pump Problems-Causes and Cures, pp. 4-12, Tulsa, Oklahoma: Petroleum Publishing Company (1980),