Best Practice SABP-A-002 Load Management for Energy Efficiency: Pumps and Compressors Document Responsibility: Energy Systems Unit, CSD 26 October 2005 Load Management for Energy Efficiency: Pumps and Compressors Developed by: Energy Systems Unit Consulting Services Department Issue Date: October 2005 Previous issue: None Next Planned Update: 1 November 2006 Primary Contact: jimmy.kuamana@aramco.com, phone +966 (3) 874-6157 Copyright©Saudi Aramco 2009. All rights reserved. Page 1 of 49 Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Table of Contents 1.0 Introduction 1.1 Purpose 1.2 Scope 1.3 Intended Users 1.4 References and Related Documents Page 5 5 5 5 2.0 General 2.1 Definitions 2.2 Principles and Concepts 2.3 Degrees of Freedom 2.4 Affinity Laws 2.5 Drivers 2.6 Data Quality 6 6 6 7 7 8 3.0 Pump Networks 3.1 Flow Profile 3.2 Number of Operating Trains 3.3 Recycle Minimization 3.4 Best Efficiency Point 3.5 Load Allocation by Efficiency 3.6 Composite Characteristic Curves 3.7 System Curve 3.8 Controls and Instrumentation 14 15 17 23 28 29 30 34 37 4.0 Compressor Networks 4.1 Thermodynamics of Gas Compression 4.2 Performance and System Curves 4.3 Control Strategies 4.4 Process Modifications 38 38 41 42 46 ATTACHMENTS none Page 2 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors List of Exhibits Exh. No 2.1 2.2 2.3 2.4 2.5 2.6 2.7 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 3.10 3.11 3.12 3.13 3.14 3.15 3.16 3.17 3.18 3.19 3.20 3.21 3.22 3.23 3.24 3.25 3.26 3.27 3.28 3.29 3.30 3.31 Title The Affinity Laws Simplified Schematic of CDU in an Oil Refinery Raw PI Data for Crude Oil Flow rates Calculated Yield Trends for Crude Oil Product Schematic Diagram of Butane Vapor Recovery System Measured PI Data for Butane Vapor Recovery System Schematic Diagram of Oil Storage and Loading Facility Typical Pump Network and Control System Fluid Flow Historical Data (sample) Fluid Flow Profile Histogram Determination of Ideal Trigger Points for Pump Switching Ideal Operating Policy for AM Shipper Pumps Indicative Relationship between Trigger Point and Reliability Basic Pump Data Pump Operating Status and Flow Data Estimating Power Savings from Minimizing No of Operating Trains Actual Pump Trains in Operation versus Minimum Required Power Cost Savings Potential vs Trigger Point Impact of Sampling Interval on Calculated Savings Typical Pump Control System Typical Variation of Power Consumption with Flow Rate Power Savings Potential from Minimizing Recycle Inferring Recycle Requirements from Flow Profile Estimating Power savings from Minimizing Recycle Operation at Best Efficiency Point vs Minimizing Pump Trains Operation at Best Efficiency Point vs Minimizing Recycle Load Allocation by Equipment Efficiency Generic Network of 3 Pumps in Series/Parallel Characteristic Curves for Individual Pumps Correlation of Pump Characteristic Curve Data Composite Characteristic Curve for Pumps in Series Composite Characteristic Curve Data for Pumps in Series/Parallel Composite Characteristic Curves for Entire Network Schematic Diagram of Simple Pumping System System Data at Design Conditions Static and Dynamic Heads at Design and Maximum Flow System and Characteristic Curves Recommended Control Philosophy for Parallel Pump Trains Page 7 9 10 10 11 11 13 14 15 16 17 18 19 20 20 21 22 22 23 24 24 25 26 27 28 29 30 31 31 32 32 33 33 34 35 36 36 37 Page 3 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 4.10 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Typical Data for Centrifugal Compressors Typical Centrifugal Compressor Operating Curves Compressor Performance with Variable-Speed Drive Suction Throttling Control of Fixed-speed Parallel Compressors Proportional Loading of Parallel Compressors with ASDs Control of Parallel Compressors (One ASD and Rest Fixed Speed) Power Consumption for Suction vs Discharge Throttling Process Modifications to Reduce Compressor Load Power Conservation by Minimizing Compressor Discharge Pressure Shedding Fan Load vs Minimizing Compression Ratio 40 41 42 43 44 45 46 47 48 49 Page 4 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 1.0 Introduction 1.1 Purpose SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Large industrial plants commonly use multiple parallel equipment trains for improved reliability. Very often, installed equipment capacity far exceeds normal production requirements. This excess capacity can be translated into energy cost savings through “optimum load management”. The purpose of this New Best Practice is to describe ways in which energy efficiency improvement can be achieved for different kinds of equipment. 1.2 Scope Many types of equipment commonly used in Saudi Aramco plants are significant energy consumers and amenable to operational optimization through Load Management, including: • • • • • • • Pumps Compressors Fired Heaters (furnaces) Boilers – fired and unfired Process Coolers – air, water, refrigerant Steam turbines Gas turbines This Best Practice manual focuses on methods to determine the optimum load management policies for pumps and compressors only. The rest are covered in other complementary Best Practice manuals. 1.3 Intended Users This Best Practice manual is intended for use by the engineers working in Saudi Aramco plants, who are responsible for efficient operation of their facility. 1.4 References and Related Documents SAES G-005: Centrifugal Pumps SAES K-402: Centrifugal Compressors SAEP-14: Project Proposal Requirements SAER-5968: Detailed Energy Assessment at Safaniya Onshore Plants, TIC library, Dhahran (January 2005) Electrical Power Savings in Pump and Compressor Networks Via Load Management, Proc of 27th National Industrial Energy Technology Conference, New Orleans, La, USA (May 2005) No conflict is expected between the optimum load management policy and other standard Saudi Aramco operating practices with respect to reliability, safety, etc. Page 5 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 2.0 General 2.1 Definitions SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Best Practice: A process or method that, when correctly executed, leads to enhanced system performance. Load Management: An operating policy that distributes the load among multiple machines or equipment installed as series-parallel networks in a way that minimizes their energy (fuel + power) consumption, without compromising safety or reliability. 2.2 Principles and Concepts The first priority in any energy conservation program should be to capture the “Easy Pickings”, that is, energy cost savings that can be achieved with little or no investment. Managing the load on various items of energy-consuming equipment falls into this category. The fundamental concept is to extract some operating cost savings in the form of reduced energy consumption from the capital that has already been invested in equipment assets, but is not being utilized for production capacity. The objective is to operate the equipment at the lowest total cost while still meeting the process objective. Several general principles and strategies apply in all cases: • • • • Minimize number of machines being operated in parallel Reduce the rate at which individual machines are being run, through minimizing recycle flows Operate equipment at conditions that will maximize the system efficiency of the network, even if it means that individual items are operating away from their maximum efficiency points Assign maximum duty to the most efficient equipment (in a parallel set), and use the least efficient equipment as the “swing” machine It must be recognized, though, that there is always a trade-off. The fewer the number of parallel machines that are running at any given time, the less redundancy there will be, with consequent loss of some operating flexibility. The analysis procedure outlined in this manual will help establish the quantitative relationship between operating flexibility and energy costs, thereby enabling the operating engineers and foremen to jointly make intelligent choices about what the optimum operating policy should be. 2.3 Degrees of Freedom Optimization implies that one has multiple choices to accomplish the desired objective, and the only problem remaining is to choose the best option. The range of options available is limited by constraints – which can be either “hard” or “soft”. A hard constraint is one which we cannot or are unprepared to violate at any cost – e.g. the laws of physics, market realities, or the directives of upper management. A soft constraint is one that we have imposed on ourselves, and which could be relaxed at our discretion upon penalty of incurring some additional costs Page 6 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors elsewhere. An example of a soft constraint is the requirement for redundancy in installed equipment in order to increase the level of operator comfort. It follows that the range of available options can be increased by relaxing soft constraints, and by finding some other way to alleviate the problem that the constraint was intended to prevent/mitigate. The range of options can be increased by introducing new Degrees of Freedom, which are parameters or design features over which one has some control. For example, in an existing pumping network, one can increase the range of options available for optimizing operating policies by adding inter-connective piping (eg. “headers”) between parallel trains, retrofitting fixed speed motors with variable frequency drives, etc. Basically, one must keep an open mind. Think “out-of-the-box”. Do not accept the existing plant configuration as inviolate; try to think of the ideal solution, and then systematically add features to the existing design that will help to reach that ideal solution. Learn to recognize the difference between hard and soft constraints. 2.4 Affinity Laws Sometimes it is necessary to determine the performance of an existing pump or compressor for a different impeller diameter or speed. The pump performance at off-design conditions can be estimated using what are known as the Affinity Laws, sometimes also called the Fan Laws, as summarized in Exhibit 2-1. Exhibit 2-1: The Affinity Laws 2.5 Constant Impeller Diameter Constant Impeller Speed Capacity Q1 N1 = Q2 N 2 Q1 D1 = Q2 D 2 Head H1 N1 = H 2 N 2 Horsepower BHP1 N1 = BHP2 N 2 2 H1 D1 = H 2 D2 3 2 BHP1 D1 = BHP2 D2 3 Drivers Pumps and compressors are usually driven by electric motors, but not always. Sometimes the motive power is provided by steam turbines (usually in the 500-10,000 HP range) or by gas turbines (usually >10,000 HP). Electric motors generally operate at fixed speeds. For 60 Hz power supply, these are usually around 1200, 1800, or 3600 rpm. For 50 Hz a/c power supply the corresponding speeds are 1000, 1500, or 3000 rpm. When speed variation is desired for either process reasons or for power savings, they have to be fitted with some sort of a speed control, such as a belt & pulley system (obsolete technology), a hydraulic clutch and gear box, or a variable frequency drive. Page 7 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Steam and gas turbines, on the other hand, are inherently variable speed devices, and elaborate controls are required to make them operate at constant speed. The correct choice of driver – whether motor or turbine – depends on whether speed control would be beneficial in that particular application, and on the size (power consumption) of the pump or compressor. The overall site steam and power balance also has a considerable influence on the economics of driver selection (especially for the larger sizes), and should not be ignored. A separate Saudi Aramco Engineering Procedure (still under development at the time of writing) addresses the issue of how and when to select adjustable speed drives (ASDs). 2.6 Data Quality Data quality refers to both the consistency and accuracy of measured values. Consistency is necessary; accuracy (within the specified limits) is sufficient. {The terms necessary and sufficient are used here in accordance with their strict mathematical definitions.} It should be common sense that “bad” data will lead to the wrong decisions no matter how brilliant the quality of the analysis. But how do we define “Bad” and “Good”? There is no such thing as perfect accuracy. An acceptable level of error in data accuracy is that which will not lead to the wrong process design or operating decision. As long as the correctness of the decision is not affected, the data quality can be considered to be “Good”. 2.6.1 Data Validation for Consistency Data Validation is the process of checking the various related values measured and recorded in the DCS/PI systems against independent sources and found to be in agreement. In general, we need to ensure the measured data are consistent of with the laws of mass and energy conservation, and with known physical and thermodynamic properties. The methodology is best explained using illustrative examples. Example 1: Material Balance check Consider the Crude Distillation Unit shown in Exhibit 2-2, with the following measured data. Stream Feed P1 P2 P3 P4 Measured flow (MBD) 100 10 25 35 30 Density Mass flow (Klb/h) 0.85 0.70 0.80 0.90 0.95 6198 510.4 1458.3 2296.9 2078.1 Sum of product flows on volume basis = 10 + 25 + 35 + 30 = 100 MBD, which appears to be in exact agreement with the feed rate of 100 MBD. Page 8 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Exhibit 2-2: Simplified Schematic of CDU in an Oil Refinery CRUDE OIL DISTILLATION UNIT P1 P2 Feed P3 P4 Would it be right to conclude that the data are consistent? No, because what is conserved is mass, not volume. This is a common mistake that should be avoided. Sum of product flows on mass basis = 510.4 + 1458.3 +2296.9 +2078.1 = 6343.7 Klb/h. It is not possible for the flow out to be more than the flow in. So strictly speaking the measured data should be considered to be inconsistent. However, if we look at the magnitude of the error, it is 146 Klb/h, or 2.4% of the feed rate, which is within the accuracy of the meters, and so we would accept the data as being acceptable despite being inconsistent; in effect we deem the data to have acceptable consistency. If, on the other hand, the error was found to be greater than the meter accuracy, then the data would be determined to be unacceptable, and some action would be required to reconcile the discrepancies before analysis can begin. Example 2: Material Balance check Let us say that we want to check the quality of flow data for the AM and AH product shipment pumps from one of the GOSPs. Sample raw data from the PI system are shown in Exhibit 2-3, columns 4 and 5. How can we check for consistency? One way is to calculate the “yield”, which we shall define as the ratio of product flow to feed flow. Because a certain (variable) amount of vapor flashes off in the wet crude receiving tank, the yield is expected to be less than 100%. The computed values shown in columns 6 and 7 reveal that the yield for AM crude is fairly steady throughout the year at around 94%, indicating the data are consistent. The computed values of yield for AH crude, on the other hand are often in excess of 100%, and occasionally in excess of even 200% (see circled areas in Exhibit 2-4). Page 9 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Exhibit 2-3: Raw PI Data for Crude Oil Flow Rates Exhibit 2-4: Calculated Yield Trends for Crude Oil Product Material Balance Check 250 AH Crude 200 Dry/Wet Yield, % AM Crude 150 100 50 0 1 31 61 91 121 151 181 211 241 271 301 331 361 Day of Year (2003) Page 10 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Since it is impossible for the yield to be higher than 100% (see numbers highlighted in brown in Exhibit 2-3), we conclude that the measured data for AH crude feed and product flow rates are inconsistent with each other, and there was obviously some problem with the metering system for the first 5 months that appears to have been fixed subsequently. Example 3: Properties check Consider the product recovery system in Exhibit 2-5, in which vapor from a liquid-butane storage tank is compressed and condensed against air before being returned to storage. Exhibit 2-5: Schematic Diagram of Butane Vapor Recovery System P2 T2 Air Cooler Compr Sat. Butane Vapor P1 P3 Receiver T3 T1 Exhibit 2-6: Measured PI Data for Butane Vapor recovery System Point # Measured Pressure (psia) Measured Temp (F) 1 2 3 13 110 107 30 169 160 Tsat at measured Pressure 25.5 155 153 Psat at measured Temp 14.3 130.1 116.7 Let us examine whether the measured pressures and temperatures make sense. The first thing to do is list the equilibrium temperatures and pressures for the measured pressures and temperatures. Page 11 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors For point 1, which is known to be a saturated vapor, the measured temperature of 30°F is 4.5° higher that the equilibrium (or saturation) temperature of 25.5°F . This is too great a discrepancy. So either the temperature or pressure measurement must be wrong. Of course, it is also possible that they are both wrong. Generally temperature measurements are more reliable, so we would probably choose T=30° and P=14.3 psia for point 1 as the reconciled values. For point 2, which is known to be a superheated vapor, the measured temperature would be expected to be higher than the saturation temperature, which is in fact the case. So we would accept the data for point 2 as being consistent. For point 3, which is known to be a saturated vapor, there is once again a discrepancy between the measured pressure and temperature. If we assume, as for point 1, that the temperature reading is more reliable, we get a pressure value in the receiver of 116.7 psia, which is more than the measured value of P2. Since that is not possible, we choose P3=107 and T3=153°F. This gives the following reconciled values: Point 1 2 3 Pressure, psia 14.3 110 107 Temperature, °F 30 169 153 An alternative possibility is that the measured pressure reading of P2 is also wrong. If we postulate that both the temperature readings are right and both readings are wrong, then we get a pressure drop in the air cooler of 130.1 – 116.7 = 13.4 psi. This is too high. Therefore we conclude that our first reconciliation decision is probably correct, and these are the values that should be used for design. 2.6.2 Data Validation for Accuracy Accuracy means that the measured values are equal to the true values. Checking for accuracy is much more difficult than for consistency. Consider this example of an oil storage and loading station, depicted schematically in Exhibit 2-7. The flow is 10,000 gpm for 2 hours. The meter has been calibrated recently and certified as accurate by the maintenance department. The same pump is used for both loading (filling) and unloading a fuel storage tank, according to the following operating policy: Filling Unloading Valve A open closed Valve B closed open According to the meter, the amounts of oil that flow in and out are each 1,200,000 gallons (28,571 barrels). So the data are consistent. But are they accurate? For this we need independent verification. Page 12 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Exhibit 2-7: Schematic Diagram of Oil Storage and Loading Facility FI B Level 2 A A Level 1 Truck or Ship B Let us say that the tank is a vertical cylindrical type with a diameter D = 80ft. Let us say that the difference between the initial and final level in the tank (after filling) is 30 ft. Then, the volume of oil pumped in is V = (πD²/4) x ∆h = 150,797 ft³ = 1,130,976 gal. This is 6% less than what the meter reading shows, and so would be considered inaccurate, because the standard of accuracy for custody transfer meters is usually less than 0.5%. Page 13 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 3.0 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Pump Networks The methodology used for estimating savings potential will be described for a single representative pumping system only (see Exhibit 3-1), as all systems can be evaluated in an identical manner. Exhibit 3-1: Typical Pump Network and Control System There are three principal steps: 1. Develop and assess flow profile 2. Determine optimum operating policy for each of the four load management strategies noted in section 2.2 3. Estimate the power cost savings potential by comparing the costs of operation under the prevailing operating practices against the those from following the optimum operating policy Each of these will now be described in more detail. Page 14 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 3.1 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Flow Profile First and foremost, we have to develop a histogram of the load profile from raw PI data, as in Exhibits 3-2 and 3-3. PI data should be recorded as daily averages for a period of at least 12 months to capture seasonal variations. Exhibits 3-2a and b: Fluid Flow Historical Data (sample only). Note: AM CRUDE OIL PRODUCTION RATE 1400 1200 Design capacity 800 600 400 200 A M Crude 12/1 11/1 10/1 9/1 8/1 7/1 6/1 5/1 4/1 3/1 2/1 0 1/1 Flow , MBD 1000 DATE Page 15 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Exhibit 3-3: Fluid Flow Profile and Histogram AM Crude production profile 140 D ays per year 120 100 80 60 40 20 0 550 600 650 700 750 800 850 900 950 Flow, MBD HELP NOTE For those of you using Microsoft Excel®, there is a useful feature that enables you to generate histograms easily from tabular data. However, this feature is not part of the basic installation of MS Office®, and must be loaded manually. If you do not see Data Analysis on the Tools menu, you will need to load the Add-In as follows: Click on Tools > Add-Ins Check the box named Analysis ToolPak After a few seconds, you should see Data Analysis on the Tools menu. Click on that, select Histogram, and follow the instructions. Page 16 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 3.2 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Minimize Number of Operating Trains One of the biggest sources of energy savings is to minimize the number of pump trains being operated in parallel. This section describes the criteria for determining this number, and developing an operating policy that balances energy savings versus equipment integrity, operating flexibility, and reliability. Two important considerations must be kept in mind: (a) There is a certain minimum flow required through each pump below which cavitation could occur and damage the pump. This type of cavitation (as opposed to the type caused by inadequate NPSH) is due to eddy formation in the pump suction/discharge nozzles at low flows, and generally begins when the flow falls below 60% of flow at the best efficiency point. Short term episodes of low flow are not a problem; damage occurs only if low-flow operation is sustained for several weeks or months. However, if the pump flow falls below 30% of the “best-efficiency” flow, the fluid could overheat due to low pump efficiency, and reach its bubble point inside the pump casing. If this happens, the pump will seize, and stop working altogether due to internal mechanical damage. The recycle line is designed to prevent these types of problems. (b) In general, the flow achievable by using N pumps in parallel will be less than N times the flow through a single pump. This is because there is a non-linear relationship between flow and number of pumps, which is determined by the intersection between the system curve and the composite pump characteristic curve. As a matter of principle, we should never operate more than the minimum number of trains needed to satisfy the production target set by the corporate dispatching department. On the other hand, if the required throughput is bordering between the capacities of one pump and two, or between 2 pumps and 3, it is not good practice to frequently start and stop the “extra” pump. Exhibit 3-4: Determination of Ideal Trigger Points for Pump Switching AM Crude Booster/Shipper Pumps 2000 1800 1600 Head, ft of oil 1400 1200 1000 system hd, 3 p/l 800 system hd, 2 p/l 600 1 train 400 2 trains 200 3 trains 0 0 500 1000 1500 2000 2500 3000 Total Flow , MBD Page 17 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors The fractional load at which we make the switch between N and N+1 pumps has been named the “Trigger Point”. Numerically, the Trigger Point is expressed as the ratio (%) of actual flow at which the switch is made to the “ideal” flow at which the switch should be made. The flow rate at which we should ideally switch from N pumps to N+1, and vice versa, is when the composite characteristic curve for the pump network intersects the system curve (inclusive of required minimum control valve ∆P), as in Exhibit 3-4. This corresponds to a Trigger Point of 100%. The pump characteristic curve is obtained from the data sheets, and verified against operating data in the PI (plant data historian) system. For pumping networks consisting of multiple pumps connected in series and/or parallel, we have to construct a composite characteristic curve from the individual pump curves, according to the procedure explained in section 3.6. For pumps connected in series, we must add the individual heads at a given flow rate. For pumps connected in parallel, we must add the flows at a given head. The system curve can either be determined from the data sheets, or from PI data, as illustrated in section 3.7. One should keep in mind that the design manual and data sheets are usually based on new pipe, for which the pressure drop per linear foot is less than for old pipe, and make the necessary adjustments. Another potential complicating factor is that sometimes there could be more than one pipeline available for use. In the case of the AM and AH crudes, there are several pipelines that could be used interchangeably. For AM crude either 2 or 3 pipelines are normally used, depending on the flow rate. The “ideal” operating policy at a Trigger Point of 100% as derived from intersection of the system curves with the pump composite curves in Exhibit 3-4 is shown in Exhibit 3-5. Exhibit 3-5: Ideal Operating Policy for AM Shipper Pumps (at Trigger Point = 100%) Unfortunately following the “ideal” policy runs the risk of having to throttle back production during the time it takes to get the extra pump/train up and running. In practice, therefore, it is safer to start up the N+1th pump a little bit before it is needed, and to keep it running a bit longer after it is no longer needed. In effect, therefore, the optimum Trigger Point for fixed speed motors drives is somewhat less than 100% (see Exhibit 3-6). The approximate relationship between Trigger Point and reliability (measured in terms of lost production during the switchover period) is shown semi-quantitatively in Exhibit 3-6. The optimum operating zone is around the sharp bend in the curve, when reliability falls off rapidly for small increases in Trigger Point. For fixed speed motors the optimum range of Trigger Page 18 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Points centers around 95%, which is the number recommended, and is the basis for estimating the energy savings potential compared to existing operations. Significantly greater power savings can be realized if the Trigger Point is raised from 95% to 100% or 105%. This can be achieved if the pump driver has over-speed capacity, eg. if the motor is fitted with a variable frequency drive (VFD), or the driver is a steam- or gas turbine. Exhibit 3-6: Indicative Relationship between Trigger Point and Reliability Reliability vs Trigger Point 120 Reliability Index 100 80 Optimum Zone 60 40 f ixed spd motor var spd drive 20 0 0 20 40 60 80 100 120 Trigger Point, % To calculate the energy savings potential, one has to compare the cost of the current operating practice versus the cost of following the optimum policy. The energy consumption and cost of actual operation can be obtained either from the power meters (if the pumps have them), or by following the methodology described below. Step 1: Prepare a summary of the pumps data. Step 2: Determine pump on/off status over a period that represents typical operation. Step 3: Calculate minimum number of pump trains required for each operating interval, for a range of Trigger Points, say 85% to 105%. Step 4: Estimate power savings potential on the basis of shutting down the excess pumps during each operating interval, and sum these savings for all intervals within the selected period of interest. Step 5: Prepare a table and plot of power/cost savings potential vs Trigger Point. Page 19 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Exhibit 3-7: Basic Pump Data (Example) From the PI system, determine the on/off status (and flow rate if ON) of the pumps at the mid point of the selected period. For example, if the year is divided into 365 24-hour periods, and the periods are counted from midnight to midnight, then you would check the on/off status at noon every day. If the period is elected to be a shift, and the shift timings are 6 am – 2 pm, 2 pm – 10 pm, and 10 pm – 6 am, then the mid-points of the periods would be 10 am, 6 pm, and 2 am. The selection of sampling interval can be important, and is discussed in detail at the end of this section. Sample output from the PI system for the AM Booster/Shipper pumps at Safaniya is shown in Exhibit 3-8 for illustration. Exhibit 3-8: Pump Operating Status and Flow Data (from PI system) Page 20 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors From the pump flow data and the ideal operating policy table (Exhibit 3-5), calculate the number of pump trains required during each selected period for a range of Trigger Points (eg. 85% to 105%, in increments of 5%). The computational logic is as follows: Let number of parallel trains required = NP, and assumed Trigger Point (%) = TP. Let the minimum required flow through a pump to avoid cavitation or seizure be FM. Then, For FM < Flow < 825*TP, NP = 1 For 825*TP < Flow < 960*TP, NP = 2 For Flow > 960*TP, NP = 3 Exhibit 3-9: Estimating Power Savings from Minimizing No of Operating Trains The power savings are estimated assuming that each excess pump will be operating for exactly one full interval. While this is not strictly true, it is not a bad approximation, as there will be some intervals during which an excess pump may be operating part of the time but does not get recorded because it happened to be off at the sampling moment, and these discrepancies should cancel one another on average. Page 21 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors It is helpful to also plot the “fractional” number of pump trains required against actual number of trains in operation (as in exhibit 3-8) to get a visual feel for how much of the time excess trains are being operated. Exhibit 3-10: Actual Pump Trains in Operation versus Minimum Required AM crude shipping pumps 2.5 Number pump trains in operation Excess Pumps in use 2.0 1.5 1.0 0.5 No . trains needed No . Trains Running 0.0 12/10 1/29 3/20 5/9 6/28 8/17 10/6 11/25 1/14 3/4 Date, 2003 Exhibit 3-11: Power/Cost Savings Potential vs Trigger Point AM Booster/Shipper Pumps 1200 Savings, K$/yr 1000 800 600 400 200 0 80% 85% 90% 95% 100% 105% Trigger Point It can be seen that savings can be substantially higher for pumping systems with adjustable speed (variable frequency) drives on the motors. In the case of the AM Booster/Shipper pumps at Safaniya Onshore Plants, an additional $520-760 K/yr of savings could be realized by fitting the fixed speed motors with VFDs, and operating at a higher Trigger Point. The economics of installing VFDs are very attractive because it is possible to design a control system such that only one VFD is needed for any number of parallel trains. For other types of adjustable speed drives, eg. hydraulic gears, this is not the case. Page 22 of 49 Copyright©Saudi Aramco 2009. All rights reserved. 110% Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors When evaluating the existing load management practices of a set of parallel equipment it is important to ensure that the data sampling technique is an accurate representation of reality, because in order to have an accurate estimate of savings potential, we need to get the closest correlation possible between the average flow rate over the sampling period, and the on/off status of the pumps (which is an instantaneous measurement) and the average flow rate during the interval. This would argue for the shortest possible interval, say 15 minutes. However, since it normally takes at least 2 hours to get a pump fully operational from a cold start, there is unlikely to be a disconnect between average flow rate and pump status for sampling intervals less than 2 hours. In order to reduce the computational effort, we can limit the number of samples to 365 by making the assumption that average flow rates over the sampling period are representative of the average flow rates for the whole day, and calculate estimated savings accordingly. The relationship between sampling time interval and calculated savings is shown in Exhibit 3-12, which confirms that for savings from optimizing the number of running pumps, the sampling period does not have a statistically significant impact on the results. Exhibit 3-12: Impact of Sampling Interval on Calc’d Savings from Minimizing Excess Pumps Savings from Excess Pumps 129 128 Savings, K$/yr 127 126 125 124 123 122 0 4 8 12 16 20 24 Sampling Interval, hr 3.3 Minimize Recycle Flows Flow control can be achieved in many different ways – by throttling the main discharge line, by running the pump at full throttle and recirculating the excess flow, or by using an adjustable speed drive. Flow recirculation is also employed for protecting the pump against mechanical damage that could occur at low-flow conditions, as explained at the beginning of section 3.1.2. A typical pump installation showing the piping and control scheme is illustrated in Exhibit 3-13. Page 23 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Exhibit 3-13: Typical Pump Control System Pump power consumption is a function of flow (Exhibit 3-14). Recirculation through the bypass line increases flow rate and wastes energy; therefore it should only be employed when the net process flow falls below the minimum flow requirement of the pump. The opportunity for energy savings arises when some flow is being recirculated through the by-pass line even when it is not needed. This usually happens when the pumps are grossly oversized for the required service, a consequence of excessive conservatism during the project planning and design phase. Exhibit 3-14: Typical Variation of Power Consumption with Flow Rate Pum p Pow er Consum ption 900 Total 800 PV pow er Power, BHP 700 Pw r to Heat 600 500 400 300 200 100 0 0 200 400 600 800 1000 1200 1400 1600 Flow , gpm Page 24 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors “PV” power is the useful energy absorbed into the process for increasing pressure or driving the fluid. However, a certain amount of input power is lost to heat due to friction. Observe that the pump efficiency (useful energy divided by input power) is not constant but in fact goes through a maximum over the pump’s operating range, falling off to near zero at extremely low flow rates. In general, there are two situations that we could encounter: (a) (b) Required Process Flow > Minimum Pump Flow Required Process Flow < Minimum Pump Flow In case (a), there should be no recycle; in case (b) some recycle is unavoidable, but should be kept to the minimum. Exhibit 3-15: Power Savings Potential from Minimizing Recycle Pump Power Consumption 1000 MINIMUM Power, BHP 800 REQUIRED ACTUAL 600 SAVINGS 400 200 RECYCLE 0 0 200 400 600 800 1000 1200 1400 1600 Flow , gpm Pump Power Consumption 1000 REQUIRED MINIMUM Power, BHP 800 ACTUAL UNAVOIDABLE RECYCLE 600 SAVINGS 400 200 ∆ RECYCLE 0 0 200 400 600 800 1000 1200 1400 1600 Flow , gpm Page 25 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Consider Exhibit 3-15, which shows the power-flow curve for a typical pumping system. Let us adopt the following nomenclature: Parameters: F =Flow, HP= Power Subscripts: A=Actual, R=Required, M=Minimum Then the potential power savings for a given time interval are: ∆HP = HPA – max ( HPR, HPM ) where BHP = Flow (gpm) x Head (psi) Flow (gpm) x Head (feet) x sp gr = 1714 x pump eff 3964 x pump eff To calculate the power consumption for each case (actual, required, minimum), use the average flow rate and head for that time interval. Pump efficiencies at the relevant flow rates should be obtained either from the pump manufacturer’s data sheet/curve or from the efficiency data generated during the most recent pump performance test. The power savings for each time interval must be added up for all intervals during the year to get the total annual savings. It is recommended to use either 365 intervals of 1-day each, or 730 intervals of 12 hours each. The pump flow profile histogram is a very good indicator of whether there is significant cost saving potential from elimination or minimization of recycle. Exhibit 3-16: Inferring Recycle Requirements from Flow Profile Flow Distribution Histogram 140 100 Days per year Design Capacity Minimum Flow 120 80 60 40 20 0 550 600 650 700 750 800 850 900 950 1000 1050 Flow , gpm Page 26 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Specific steps to be taken are listed below and illustrated in Exhibit 3-17. (a) Develop correlation for pump characteristic curve (from factory test or data sheet) (b) Develop correlation for pump efficiency curve (from factory test or data sheet) (c) Establish minimum flow requirement per pump (if not specified on pump data sheet, assume 35% of flow at “best efficiency” point) (d) Gather PI data for net process flow and actual flow through pump (= process flow + recycle flow) (e) Calculate power consumption and potential savings from recycle flow elimination or minimization using the formulas given above Pump Curve and Efficiency correlations: Head-flow correlation (pump characteristic curve) h (ft) = a + bQ - cQ^2, where Q = gpm/100 a b c 1702 15 2.5 Pump Efficiency correlation eff (%) = a + bQ - cQ^2, where Q = gpm/100 a b c 14.48 14 0.8 Operating data: liquid sp gr Cost of power Interval duration 0.86 26.7 24 $/MWH hours Minimum flow (surge point) Head (from char curve) Efficiency, % Power consumption 571 1706 68.3 309 gpm feet % HP Exhibit 3-17: Estimation of Power Cost Savings from Minimizing Recycle Page 27 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 3.4 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Operation at Best Efficiency Point The efficiency of pumps is a function of flow rate. Sometimes, the efficiency can be significantly lower at flow rates beyond the “best efficiency” point, and one has to check to make sure that minimizing the number of operating pumps will in fact minimize power consumption. The procedure for doing so is illustrated in Exhibit 3-18. If this is not the case, then the operating policy developed for minimizing pump trains in operation (as recommended in section 3.2) must be revisited and revised as necessary. Real efficiency curves are seldom as extreme as the one shown in Exhibit 3-18, but it makes the point. Exhibit 3-18: Operation at Best Efficiency Point vs Minimizing Pump Trains 100 1600 80 1200 60 800 40 Head Efficiency 400 200 400 600 Process Flow, gpm Number // pumps Flow per pump, gpm Head, feet Efficiency, % BHP per pump Total power, HP Case 1 2700 2 1350 1449 57.7 735 1471 normal Case 2 2700 3 900 1635 75.7 421 1264 best eff 20 0 0 Efficiency, % Head, ft of liquid Pump Head and Efficiency 2000 800 1000 1200 1400 1600 1800 0 2000 Flow, gpm In the case or recycle elimination/minimization (as recommended in section 3.3), however, there is never a case to be made for operating at higher flow than the minimum, because the increased power consumption due to higher flow always exceeds the savings from efficiency improvement, as illustrated in Exhibit 3-19. Page 28 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Exhibit 3-19: Operation at Best Efficiency Point vs Minimizing Recycle 100 1600 80 1200 60 800 40 Efficiency, % Head (ft), or Power (HP) Pump Head and Efficiency 2000 Head Power 400 20 Efficiency 0 0 200 400 600 800 1000 1200 1400 1600 1800 0 2000 Flow , gpm 3.5 Load Allocation by Equipment Efficiency So far we have assumed that all pumps in parallel are identical, and have identical efficiencies. In fact, this can never be strictly true; at best, it can only be approximately true. In some cases, it may not even be approximately true, eg. if one machine suffers mechanical deterioration at a faster rate than another. The appropriate operating policy, when we have parallel machines of unequal efficiency, is both simple and obvious: Use the most efficient machines for base load, and the least efficient machines for swing loads. The calculation procedure is straight-forward. Consider the case of three equal-sized pumps of varying efficiency of which only two are normally operated in parallel (see Exhibit 3-20). The best combination is pumps 1+3, while the worst is pumps 2+3. The savings potential between best and worst combinations is $1039 - $996 = $42K per year. While this may not be very great compared to some of the other savings, it can be achieved easily with zero capital investment. Page 29 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Exhibit 3-20: Load Allocation by Equipment Efficiency Process Flow, gpm Head, feet sp gr Pump efficiency, % Pump power, HP Pump #1 Pump #2 Pump #3 10000 10000 10000 800 800 800 0.9 0.9 0.9 75 69 72 2421 2631 2521 Alternative pump combination options: Option 1 2 3 Pumps ID 1+2 1+3 2+3 Σ HP 5052 4942 5152 K$/yr 1018 996 1039 n In general, the number of combinations to be evaluated is Cm, calculated using factorials as follows: n Cm = n! n(n − 1)(n − 2)...x2x1 = m! x (n − m)! [m(m − 1)...x2x1] x [(n - m)(n - m - 1)(n - m - 2)...x2x1] where n = total number of installed parallel pumps, and m = number required to be in operation simultaneously. 3.6 Composite Characteristic Curves for Pump Networks The head-flow relationship of pumping networks consisting of multiple pumps connected in series and/or parallel is described by the composite characteristic curve, which must be constructed from the individual pump curves. For pumps connected in series, we must add the individual heads at a given flow rate. For pumps connected in parallel, we must add the flows at a given head. Consider the series/parallel network shown in Exhibit 3-21, with individual pump characteristic curves as shown in Exhibit 3-22. Page 30 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Exhibit 3-21: Generic Network of 3 Pumps in Series/Parallel to Process Wet Crude Storage Tank Storage Tank Pump 2 Pump 3 Pump 1 Exhibit 3-22: Characteristic Curves for Individual Pumps Individual Pump Characterictic Curves 2000 Pump 1 1600 Pump 2 Head, feet Pump 3 1200 800 400 0 0 500 1000 1500 2000 2500 3000 3500 4000 Flow, gpm The construction procedure for the composite characteristic curve is illustrated in Exhibits 3-23 to 3-26. Step 1: Develop quadratic correlation (use curve fitting utility within Excel) for each pump curve in the form h (ft) = a + bQ - cQ^2, where Q = gpm/100. It should be noted that the quadratic formulation is a good fit only for heads less than 95% of the shut-off value. In the very low flow region, when head is between 95 and 100% of the shut-off value, the relationship is more accurately correlated as a linear function: h = a - dQ. (The shut-off head is the value at zero flow) Page 31 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Step 2: Add the a, b, c and d parameters of the pumps connected in series (#2 and 3 in the example) to get the composite values for the two together. Exhibit 3-23: Correlation (Curve-Fit) of Pump Characteristic Curve Data Exhibit 3-24: Composite Characteristic Curve for Pumps 2 & 3 in Series, and Pump 1 by itself Composite Characteristic Curves for Pumps in series 2000 Head, feet 1600 1200 800 Pump #1 400 Pumps 2/3 0 0 500 1000 1500 2000 2500 3000 3500 4000 Flow , gpm Step 3: Construct a new table showing the flow for pump 1 and the 2/3 combination at the same head. Then add the two flows together at each value of head, which gives the composite characteristic curve for the whole network. Page 32 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Exhibit 3-25: Characteristic Curve Data for Pumps 1 and 2 & 3 (in Series) in Parallel Flow, gpm Pumps 2/3 Head, ft 1550 1545 1540 1520 1500 1495 1485 1450 1400 1300 1100 800 Pump 1 0 0 0 0 0 250 750 1212 1477 1870 2440 3076 in series 0 147 294 882 1338 1360 1402 1536 1701 1975 2407 2911 1 + 2/3 0 147 294 882 1338 1610 2152 2749 3178 3845 4846 5987 Exhibit 3-26: Composite Pump Characteristic Curves for Entire Network Composite Characteristic Curves 1800 1600 1400 Head, feet 1200 1000 800 600 P ump 1 400 P umps 2/3 Co mpo site 200 0 0 1000 2000 3000 4000 5000 6000 7000 Flow, gpm Observe that if system head is greater than the shut-off head for Pump 1 (= 1500 ft), it will not be able to contribute any flow, and the combined network flow will be equal to the flow from Pumps 2/3 (in series) only. Both parallel lines can contribute flow only when the system head falls below the shut-off head for the lower one of the two. Page 33 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 3.7 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors System Head Curve for Pump Networks System head is the total head that the pump must overcome at any given flow. It has two principal components – static and dynamic. The static head consists primarily of the potential energy difference between the suction and discharge points. The dynamic head consists primarily of kinetic energy (fluid momentum) differences and frictional losses in the piping network. Exhibit 3-27: Schematic Diagram of Simple Pumping System P2 P1 Static Head (feet of liquid) = (P2 - P1)/ρ + (h2 - h1) It should be recognized that the “static” head is not necessarily constant. It will fluctuate somewhat as a consequence of variations in vessel pressure at the suction and discharge ends, as well as due to fluctuations in liquid level. If frictional losses dominate the system, then the static head may be considered to be approximately constant, but if not, then variations in static head would have to be taken into account in the analysis. Dynamic head (feet) = (α 2 V22 − α1V12 ) Σ(∆Pf ) + 2g ρ In normal industrial piping systems, the kinetic energy (V2/2g) term is generally small, and can be safely neglected. Strictly speaking, the frictional term in the Bernoulli equation includes pressure losses in the piping, equipment, instruments and the pump itself (bearings, seals, etc). It is common practice, however, to separate pump losses from piping/equipment losses. Internal losses within the pump are accounted for as pump efficiency, and only the piping, equipment and instrument losses are included in the dynamic head component of system ∆P. Page 34 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Frictional pressure drop in turbulent flow (Reynolds numbers > 10,000) can be very closely estimated by the equations ∆P = 2ρ f LV2 / gD and f = 0.0029 (ρDV/µ)-0.2 Because the “Moody” friction factor f is itself a function of velocity, the net proportionality between frictional pressure drop and pump flow works out to be approximately ∆P α Q1.8 With a proper understanding of these basic principles, it becomes easy to develop the system curve from available data. If the engineering design contractor and the procurement group have done their jobs right, the static and dynamic heads at the design condition will be recorded on the pump documentation supplied by the manufacturer. Only four items of information are needed – design flow rate, liquid density, static head, and piping/equipment frictional drop at design flow – to calculate the system curve over its entire range of operation: (∆Pf ) d H = HS + ρ Q Qd 1.8 , where subscript “d” refers to Design conditions. A more accurate method is to obtain this same information from PI data over a suitably wide flow range. Unfortunately, there is seldom sufficient instrumentation installed to enable disaggregation of the control valve drop and the frictional drop. The appropriate procedure in such instances is to estimate the piping and equipment frictional drops (using the equations and methods described in most engineering handbooks and college-level textbooks on fluid mechanics). The first step is to list the pump data (from drawings and design manuals or plant data), as in Exhibit 3-28. The next step is to calculate the static and dynamic heads according to the equations provided above, as in Exhibit 3-29. Exhibit 3-28: System Data at Design Conditions Liquid level (elevation), ft Source/destination pr, psig Equipment ∆P psi Piping ∆P, psi Instrument (meters) ∆P, psi suction discharge 24 86 12.0 31.1 3.9 25.7 2.0 293 0.0 4.3 Page 35 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Exhibit 3-29: Calculation of Static and Dynamic Head at Design & Maximum Flow Flow, gpm Liquid density (specific gravity) Static head, ft Dynamic head, ft System head, ft Delivered head, ft Control valve DP, ft , psi , % of TDH. Design Maximum 825 1070 0.8605 0.8605 100 100 884 1419 984 1519 1656 1576 672 58 250 21 41% 4% The control valve drop is the difference between the TDH of the pump and the system head. For good control, this should generally be about 1/3 of the total pump delivered head. Even in the fully open position, the control valve incurs some pressure drop, equal to 21 psi (58 ft) in the illustrative example, which defines the maximum flow possible from the pump and piping system. The maximum flow must be found by trial and error until the system head + control valve drop (in fully open position) equal the delivered head. Exhibit 3-30 shows the system curve in relation to the pump characteristic curve. Exhibit 3-30: System and Characteristic Curves 2000 Pump Curve System Curve Head, feet of oil 1600 Control Valve ∆P 1200 800 Dynamic head (frictional ∆ P) 400 Static head 0 0 500 1000 Flow, gpm 1500 2000 Page 36 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 3.8 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Controls and Instrumentation For effective operation of pumping networks, it is important that they be controlled properly. The control scheme shown in Exhibit 3-1 (section 3.0) is typical of Saudi Aramco facilities, but is not optimal for effective load management. A superior control scheme, which works equally well for both identical and non-identical pumps, is shown in Exhibit 3-31. [ Ref. Bela G. Liptak, Optimization of Industrial Unit Processes, 2nd ed, CRC Press, Boca Raton, Florida, USA (1999), pp 394-401.] There is no sacrifice in operating reliability; in fact the illustrated scheme features improved flexibility. Exhibit 3-31: Recommended Control Philosophy for Parallel Pump Trains The combined total flow may be set on either flow control (shown) or level control (not shown), depending on process requirements. If the stream is a process feed, we would normally prefer flow control, as this makes for better operating stability. If on the other hand, it is a product stream going to a pipeline or bulk storage facility, we may prefer to use level control. The flow controller output signal passes through a hand switch, controlled by the operator, which is routed to one of the three control valves in the individual pump discharge lines. Only one of the valves should be controlled at any given time; the other two would be either fully open or fully closed, depending on whether the pump is running or not. The valves should be set to the “fail-open” mode. Check valves, block valves, bleed valves, pressure gages, and other details of standard piping and instrumentation are not shown. Page 37 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 4.0 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Compressor Networks The methodology for estimating savings potential from load management of compressors is similar to that for pumps, except that several important differences must be taken into account: (a) Gases are compressible while liquids, for all practical purposes, are incompressible. Physical properties such as specific heat and compressibility can vary significantly at high compression ratios, affecting power consumption. (b) Density variations (due to composition and suction pressure drift) are more prevalent. (c) Compressors are generally more difficult to startup and shut down (normal startup period is on the order of 4 hours vs less than 1 hour for even very large pumps) partly because they usually operate between their first and second critical speeds, and partly because they have to be properly purged every time when compressing flammable hydrocarbon gases. (d) The system curve is generally dominated by static head, as opposed to dynamic head for most pump applications. (e) The surge limit generally occurs at 50% of the design flow at the design speed. 4.1 Thermodynamics of Gas Compression The pressure-volume-temperature behavior of real gases is described by the equation: Pv = ZRT where P = pressure, psia v = specific volume, ft3/mole T = temperature, °R R = universal gas constant = 1545 ft-lb/mole°R = 1.987 Btu/mole°R = 10.729 for the units of measure indicated above and Z = compressibility factor (must be obtained from data charts for each particular gas) For diatomic gases at low pressures, Z is approximately 1. The power consumption of a compressor, in horsepower, is given by ZW(T1 + 460) 1 k P2 BHP = 1281.55 MW η ad k - 1 P1 where T1 P1 & P2 k Z W MW ηad = = = = = = = k −1 k − 1 suction temperature, °F suction and discharge pressures, psia or any other units average specific heat ratio Cp/Cv average compressibility factor gas mass flow, lb/hr molecular weight, lb/mole adiabatic efficiency, usually in the range 0.75-0.85 The constant 1281.55 = 60 (min/h) x 33000 (ft-lb/min-HP) /1545 (ft-lb/mole°R) Page 38 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors It is useful to know that Cp is usually about 7 Btu/lb-mole°R for most diatomic gases, and 5 Btu/lb-mole°R for ideal monatomic gases. For a gas mixture, Cp is evaluated as the weighted mole fraction average. It is also useful to know that for ideal gases, Cv = Cp – R, so that k can be approximated as k = Cp/(Cp-R) Overall Energy Efficiency of a compressor is easy to define: ηo = absorbed energy into process gas isentropic HP = delivered energy to the driver brake HP Unfortunately, it is very difficult to measure. The best we can hope for is to calculate efficiency based on measurements of other process parameters such as flow, temperature, pressure, and estimation of gas physical properties. For compressors, two types of efficiencies are commonly used: adiabatic and polytropic. As a practical matter, the single-stage adiabatic efficiency can be calculated as: ηa = T2 T2’ = = T2' − T1 T2 − T1 actual discharge temperature before any cooling, °F isentropic (adiabatic) discharge temperature, °F, calculated as: P T2 ' = (T1 + 460) 2 P1 − 1 − 460 k −1 k The polytropic efficiency is then calculated as: n k − 1 n − 1 k ηp = where n = polytropic constant, which is a function of gas properties only, and determined experimentally from the equation PVn=constant, unique to each machine. Generally n and ηp are provided by the compressor manufacturer, and can be found in the data sheet supplied with the equipment at time of purchase. The adiabatic efficiency can be back-calculated from this information: η ad P = 2 P1 k −1 k − 1 P2 P1 n −1 n − 1 1 Page 39 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Adiabatic efficiency varies with compression ratio, whereas polytropic efficiency is independent of the thermodynamic state of the gas, being a function of mechanical design only. The reason is that gas properties are implicitly included in the polytropic constant itself. Polytropic efficiency is therefore a better indicator of compressor mechanical condition and performance, and so load allocation decisions should be made on the basis of polytropic, not adiabatic efficiency. The overall efficiencies are given by: η oa = η a .η m η op = η p .η m and ηm = mechanical efficiency of the compressor = fraction of power delivered by the driver (motor) that is actually transmitted to the gas, usually 97-98% Exhibits 4-1a and b: Typical Data for Centrifugal Compressors Centrigugal Compressors 79.0 12000 78.5 8000 77.5 Efficiency 6000 Speed 77.0 4000 76.5 Nominal Speed, rpm 10000 78.0 Polytropic Eff, % where 2000 76.0 75.5 0 0 10 20 30 40 50 60 Impeller diam, in Page 40 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors For multi-stage compressors, the total power requirement is simply the sum of the power for each individual stage. From a thermodynamic viewpoint, the defining characteristic of a compression stage is that there should be no inter-cooling between successive impellers. Thus a compressor casing containing multiple impellers without intermediate coolers would be considered a single stage. Confusion often arises because some manufacturers and authors of technical articles refer to each impeller as a “stage”. These are not thermodynamic stages unless an intercooler is provided between each impeller. In order to keep the temperature rise within reasonable limits, the single-stage compression ratio is normally limited to about 3.0. When calculating the power consumption of the 2nd and later stages, one should take into account the pressure drop in the interstage cooler and piping, the new suction inlet temperature, and differences in gas properties at the new suction conditions. 4.2 Performance and System Curves In general, the head vs capacity curve (also called the “performance” curve) for a centrifugal compressor operating at a fixed speed is quite flat, with the total head at the minimum throughout (the surge point) typically being only 105-115% of the head at design throughput. Similarly, the system curve is also relatively flat, because the static head usually dominates frictional (dynamic) head. The operating point occurs at the intersection of the compressor performance curve and the system curve. Exhibit 4-2: Typical Centrifugal Compressor Operating Curves Centrifugal Compressor Curves 32 Head, 1000 ft 30 28 Surge limit 26 24 Performance Curve System Curve 22 20 0 1000 2000 3000 4000 5000 6000 7000 8000 Suction Flow, acfm These characteristics are important considerations in selecting a control scheme that will result in both stable and energy-efficient operation. Page 41 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 4.3 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Control Strategies For almost all compressor applications, some form of flow regulation is required, whether to maintain a constant discharge pressure or a constant flow. Furthermore, the flow set point could be in terms of volume or mass. The type of control scheme also depends on the type of driver – whether fixed speed or variable speed. The optimal control scheme therefore varies from case to case. Speed control is considerably more efficient than throttling the flow with a valve (or even worse, by employing flow recycle) at constant compressor speed, since the valve resistance creates an unrecoverable power loss. Steam and gas turbines are inherently variable speed machines, with speed control being easily achieved by regulating either the steam flow or fuel/air flow. Compared to fixed speed drivers, variable speed drivers permit a much wider range of control in a highly efficient manner. Speed variation can be used to alter the position of the H-Q performance curve such that it exactly intersects the system curve, as illustrated in Exhibit 4-3, with power consumption rising and falling roughly in proportion to the process load. The performance curves at different speeds are developed using the affinity laws. Exhibit 4-3: Compressor Performance with Variable-Speed Drive R92-K151 33 31 29 System Curve 1780 rpm Polytropic Head, 1000 ft 27 1767 rpm 1630 rpm 25 1484 rpm 1362 rpm 23 Current Compr Flow Rate, fixed speed 21 19 17 15 0 1000 2000 3000 4000 5000 6000 7000 Process Flow (vapor from storage tank), acfm Page 42 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Controlling two or more compressors operating in parallel and having identical characteristics would be similar to the case of a single compressor. However, because of age, wear, or design differences, no two compressors have identical performance characteristics. Slight variations in flow can cause one compressor to be fully loaded and the other to fall below its surge point, forcing needless recycle. The control scheme shown in Exhibit 4-4 overcomes this problem. Typically, the suction valve that receives the lower flow is kept 100% open. Exhibit 4-5 illustrates how two compressors can be proportionally loaded and unloaded, while keeping their operating points at equal distance from the surge line. The lead compressor (31) is selected either as the larger unit or the one that is closer to the surge line when the load rises or further from it when the load falls. Improper load distribution is prevented by measuring the total load, and assigning a variable percentage to each compressor adjusting the set points of the flow ratio controllers. Each compressor must be provided with its own independent surge protection system. Exhibit 4-4: Suction Throttling Control of Fixed-speed Parallel Compressors [Ref. Liptak, op cit] Page 43 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Exhibit 4-5: Proportional Loading of Parallel Compressors with ASDs [Ref. Liptak, op cit] Various alternative control strategies for different conditions and scenarios are described in the following excellent reference texts: [1] “Optimization of Industrial Unit Processes”, 2nd ed, Bela G Liptak, CRC Press, Boca Raton, Florida (1999), Chapter 4. [2] “Compressor Handbook for the Hydrocarbon Processing Industries”, Gulf Publishing Co, Houston, Texas (1979), pp 103-124. In most Saudi Aramco plants, when the compressor is not turbine driven, the electric motor is normally operated at fixed speed (although there are a few rare cases where variable speed capability is provided either using a hydraulic gear box or a variable frequency drive). Since variable speed operation is one of the ways to improve energy efficiency, it is worth noting that this is a relatively easy retrofit that can have excellent economics when dealing with networks of parallel compressors. The reason is that only one of the compressors needs to be fitted with a VFD; the rest can be left on fixed speed, as illustrated in Exhibit 4-6 (surge protection controls not shown). It is not necessary, as some mistakenly believe, to install a VFD on each and every motor in the network. Page 44 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Exhibit 4-6: Control of Parallel Compressors (One ASD and Rest Fixed Speed) TI TI FC FI PI PI M M TI TI SC PC FI FI Whether the process objective is to maintain constant header pressure or constant flow, the fixed speed compressors must be on flow control, which can be accomplished by either suction throttling or discharge throttling. Butterfly valves are preferred because of their lower pressure drop. The appropriate valve location depends on whether the goal is to deliver mass flow (eg for most process applications, including sales gas compression) or volume flow (eg. most utility applications such as plant or instrument air). For equal mass flow rates, discharge throttling consumes less power, and therefore would be preferred. However, for equal volume flows, the situation is reversed. Consider the illustrative example in Exhibit 4-7, with suction and header pressures of 80 psia and 230 psia respectively, a k value (= CP/CV) of 1.32, and a control valve drop of 10 psi at the design flow. A comparison of the relative power consumption for the two cases clearly demonstrates that appropriate placement of the control valve can save a significant amount of energy. Page 45 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Exhibit 4-7: Power Consumption for Suction vs Discharge Throttling (a) Constant Mass Flow control Inlet pressure, psia Discharge pr, psia Compression ratio X = (k-1)/k (P2/P1)^X - 1 Relative power Suction throttling 80 230 2.875 0.1113 0.1247 1.00 Discharge throttling 90 240 2.667 0.1113 0.1154 0.9248 Suction throttling 80 230 2.875 0.1113 0.1247 1.00 1.00 1.00 Discharge throttling 90 240 2.667 0.1113 0.1154 1.2112 1.2112 1.1201 (b) Constant Volume Flow control Inlet pressure, psia Discharge pr, psia Compression ratio X = (k-1)/k (P2/P1)^X - 1 Relative inlet density Relative mass flow Relative power The final point to keep in mind is that the interaction of compressor operations with the rest of the plant must be given due consideration in the design of the control system. For example, if the process upstream of the compressors is under constant pressure control, then the compressor control system must be designed (or modified) in such a way that starting and stopping a compressor will not disturb the upstream process. In short, an integrated control philosophy is needed. 4.4 Process Modifications The principal process parameters that affect compressor power consumption are mass flow rate, suction (inlet) temperature, and the compression ratio, so anything we can do to reduce these three parameters through process modifications will help to reduce power consumption. Flow requirements are generally set by process conditions, but one should examine the overall process flowsheet to look for opportunities to change the material balance in such a way that the flow through the compressor is minimized. Inlet temperature can obviously be reduced by installing a heat exchanger in the process stream entering the compressor, but this seldom pays out, because the new cooler creates additional pressure drop in the system that will increase the required compression ratio. The increased power requirement from a higher compression ratio will almost always be more than the reduction from lower inlet temperature. The solution, once again, is to examine the overall process flowsheet, and look for places where the suction stream may be being heated. Page 46 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Exhibit 4-8: Process Modifications to Reduce Compressor Load From Exhibit 4-8a, it can be seen that the suction stream to the Sales Gas compressor is being used to cool the feed gas to the Gas Treating process. Effectively, we have a “heater” in the compressor suction line; by bypassing it, as shown in Exhibit 4-8b, we can reduce not only the suction temperature but also the compression ratio by eliminating the heater’s ∆P. The process stream which was being cooled against compressor suction must now be cooled against some other stream, with consequent net energy savings. Page 47 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Power consumption can also be reduced by minimizing the required discharge pressure. For example most Aramco compressors have a fin-fan cooler in their discharge line, whose cooling capacity varies with ambient temperature. One of the power conservation strategies used by the operators is to shed power load on the fans during cooler weather (a laudable attempt at thermal load management) once the temperature set point downstream of the cooler is being met. However, maintaining a constant condenser temperature is the wrong control objective if the compressor discharge stream is going to a condenser, because the required pressure for condensation is not constant but varies with ambient temperature. In such cases, even greater power savings could be obtained by following a different operating policy – viz. to maximize the fin-fan cooling capacity but save even more power by minimizing the discharge pressure (and therefore the compression ratio). A suggested control scheme is shown in Exhibit 4-19, with the supporting calculations presented in Exhibit 4-10. Exhibit 4-9: Power Conservation by Minimizing Compressor Discharge Pressure The tricky part is being able to determine when exactly we have achieved total condensation, something very difficult to do. The proposed solution is to have two condensers in series. The main condenser would condense only about 90-95% of the vapor, and the vent condenser would condense the balance. The control system would be set up to maintain a fixed 10:1 or 20:1 flow ratio between the main flow and the vent flow. Page 48 of 49 Copyright©Saudi Aramco 2009. All rights reserved. Document Responsibility: Energy Systems Unit, CSD Issue Date: 26 October 2005 Next Planned Update: 1 November 2006 SABP-A-002 Load Management for Energy Efficiency: Pumps & Compressors Exhibit 4-10: Shedding Fan Load vs Minimizing Compression Ratio The required discharge pressure in Case 2 is found by successive iteration until the calculated condenser surface area for cases 1 and 2 are identical. Although process modifications cannot strictly be classified as Load Management, the subject has been presented here because it is a way to introduce new degrees of freedom that enable optimal load allocation between the different energy consumers in the overall system. 26 October 2005 Revision Summary New Saudi Aramco Best Practice. Page 49 of 49 Copyright©Saudi Aramco 2009. All rights reserved.