There are approximately 200 million centrifugal pumps used in industrial processes, consuming annually about 3000 TWh of energy, or 10-20 % of the global electrical energy produced (Motiva, 2011). Roughly 75 % of these pumps are driven with direct-on-line (DOL) electric motors, with the output (flow) of the pump controlled via a throttling valve. While easy and cost-effective to set up, this approach is often hugely inefficient from the energy consumption and total cost of ownership point of view.
Indeed, significant monetary and energy savings can typically be achieved by different approaches. Nowadays, a common solution can include switching to a variable-frequency drive (VFD). However, other approaches, such as swapping a different pump or lathing the impeller of the existing pump, are also possible. Indeed, finding the optimal solution often requires specific expertise and reliable data both.
Naturally, being able to make an informed decision on a pumping process is generally desirable. Thus, quantified data on the process can be a crucial aid. This data can include the performance of the current setup, and estimates of the energy and cost savings achievable via alternative approaches.
This white paper describes an energy audit process on centrifugal pumps with DOL supply and throttling valves. While the paper is organized around the proprietary Viimatech Digital Energy Audit (VT) process, the presented theory itself is fairly general.
This paper is organized as follows. First, the overall energy auditing process is briefly described. Next, the VT-specific information pipeline is briefly discussed, starting from the current measurement and ending with the energy consumption estimates. Finally, the outline of analyzing alternate energy consumption scenarios is presented.
Authors:
PhD Antti Lehikoinen (Viimatech Digital Oy)
MSc Mikko Joronen
MSc Matti Viita (Viimatech Digital Oy)
Energy auditing process
Goals
The goal of an energy audit process, in general, in the scope of this paper, is two-fold:
- Obtaining reliable quantitative data on the current behavior of the system
- Providing estimates on alternative scenarios
For the first goal, the total energy consumption per year or per installation lifetime is perhaps the single most commonly requested quantity of interest. However, depending on the end-user and the type of the application, other aspects such as observed operating range (e.g. pump duration curve) or the distribution of pump starts and stops may be of crucial importance.
For the second goal, some very common outputs include the payback time, or e.g. the total savings over a 5-year period, after switching to a variable frequency drive.
Viimatech Digital Energy Audit
In addition to the general goals of energy auditing, the Viimatech Energy Audit process aims to offer a unique combination of low monetary and effort cost combined with satisfactory accuracy. Figure 1 below illustrates this, comparing the accuracy of different auditing methods with their requirement in person-hours, required expertise, and monetary cost. In addition to the VT process, shown are the ‘non-audit’ process of only utilizing typical characteristics of a particular type of a site with little to no actual measurements, and a detailed audit including installation of flow meters and a case-specific evaluation by experts.
Figure 1. Illustration of different audit approaches in the cost-benefit plane.
Process
The VT energy audit process can be divided into the following steps
- Installation of measurement devices
- Data collection period
- Reporting and recommendations
The first step includes setting up the VT field device, visualized in Figure 2 for reference. One of Viimatech’s primary goals is to make this step as simple as possible, exemplified by e.g. the use of a clamp-on current sensor facilitating very easy installation. Collecting the required data, such as motor name plate data and pump characteristics, is also included in this step.
Figure 2. Viimatech measurement device.
Next, a data collection period follows. This period should be long enough to be representative of the system’s behavior. For instance, any process including weekly variation should be monitored for one week at the very least – preferably several to rule out basing decisions on an anomalous period. In practice, the VT-standard 30-day measurement period is sufficient for many applications. It should be noted that this period is fully automatic, requiring input neither from the client organization nor from VT personnel.
Finally, the analysis and recommendations step concludes the audit. This step includes analysing and reporting the observed behaviour, and estimates on alternative scenarios. An illustration of the entire audit process can be found in Figure 3.
Figure 3. Flowchart of the audit process.
The monitoring phase
This section summarizes the theoretical background of the monitoring part of the audit process, beginning from the phase current measurement and ending up in the computation of total energy expenditure. Analysis of alternative scenarios is discussed in the next section. The following subsections each then discuss the important steps in the audit pipeline:
- Data collection
- Estimating the motor operating point
- Estimating the pump operating point
- Data analysis and cost calculations
Data collection
The data collection step of the audit process can be divided into two parts
- One-time compilation of background data
- Continuous monitoring
At the minimum, the background data includes the name plate data of the induction motor used to drive the pump in question – figures like the nominal power and voltage, power factor, and similar. These are practically always listed on the physical name plate of the motor, and thus available unless the plate itself has been damaged.
Additionally, some information about the pump and pumping system is needed. Ideally, the pump curves from the pump manufacturer would be used. However, as this data is often not available, reasonable accuracy can be obtained by using typical characteristic data. For this approach to work, the pump specifications such as the speed and nominal flow are still needed. Finally, the static head of the pumping setup is needed for optimal performance.
The continuous monitoring part of the audit is then fully automatic. Once the VT field device has been placed in a suitable position and the clamp-on current sensor has been engaged, it will automatically measure and transmit the phase current data to a secure cloud server.
Motor operating point
Next in the processing pipeline, the operating point of the electric motor is computed. This includes the torque and mechanical power and the speed of the motor, but also aspects like the input power and power factor.
While Viimatech is using a proprietary digital twin of an induction motor, the approach of determining the induction motor operating point from the phase current measurement alone is in itself well-known (Holttinen, 2018). Technically, also the input voltage would be needed, but in practice most sites only see relatively minor fluctuations from the assumed value. While this minor issue could be solved by a voltage measurement, this would present significant practical problems, such as safety requirements and having to take the system offline during the installation. Finally, it must be noted even though the well-known torque curve of an induction motor is quadratically dependent on the input voltage, the relationship from the current to the torque for instance can be shown to be roughly linearly proportional to it.
Another point that might not be immediately apparent is that it is indeed sufficient to measure the current from one phase only. Electric motors are indeed ideally symmetrical across their three phases; in practice the deviation is often on the sub-percentage level.
Now, the digital twin establishes a nonlinear mapping from the measured current, to the complete state of the motor. This is visualized in Figure 4, using the shaft power as an example. As can be seen, a measured phase current and the current-power trace trivially yields the estimated shaft power as a result.
Figure 4. The current-power characteristics of an induction motor, and shaft power estimation from the measured current.
Pump operating point
As the next step, the operating point of the pump is determined, based on the estimated shaft power. This point consists of the flow and head (proportional to the pressure difference across the pump) of the pump.
Several alternative approaches can be used to achieve this step (Pöyhönen, 2021). In the ideal case, the pump datasheets are available, in which case curves provided by the manufacturer already establish the relationship between the shaft power and the flow, and the flow and the head. In this case, the estimated shaft power from the motor model is used to estimate the flow, which is then in turn used to calculate the head. This has been illustrated below in Figure 5 and Figure 6.
In many cases, the pump datasheets might be unavailable. In that case, it is often possible to utilize one of the well-known approaches for estimating the pump behavior from the nominal pump speed, flow, and head, based on typical pump behavior (Stepanoff (1957), Gulich, (2014)).
Figure 5. The power-flow characteristics of a centrifugal pump, and flow estimation from the estimated shaft power.
Figure 6. The flow-head characteristics of a centrifugal pump, and head estimation from the estimated flow.
Energy consumption and duration curves
Finally, further analysis of the collected data concludes the monitoring phase of the audit. At the minimum, the total energy consumed during the measurement period is calculated. However, most of the time, more detailed analysis is desired.
Most commonly, analysis of the distribution of the pump behavior is desired. This can be presented in a histogram form, for instance. This has been done in Figure 7, for a hypothetical 1-month measurement period with near-constant pump operation. The pump in question had a best-efficiency point (BEP) of approximately 105 l/s. As can be seen from the histogram, the pump actually spent the huge majority of time at flow rates significantly smaller than that – a very typical observation with valve-throttled installations.
Figure 8 then shows the same information, but presented in the so-called duration curve format. The duration curve is essentially equivalent to a cumulative density function, visualizing the time spent pumping flows at or below the x-axis value. This presentation format is commonly preferred by pumping specialists.
Figure 7. Histogram of the flow vs hours distribution.
Figure 8. Flow duration curve
Energy savings potential
The final part of the energy audit report commonly includes estimates of alternative scenarios. The most common two cases are discussed shortly – installing a VFD, or swapping the pump itself.
In both cases, it is assumed that the measured data is representative of the typical behavior of the installation. Indeed, the analysis is most commonly performed by running the same flow-histogram scenario (aka ‘pumping this much flow for this length of time each month’) but either with a VFD or with a different pump, and estimating the amount of energy consumed.
VFD scenario
If a variable-frequency drive were to be installed, the control mode of the pump would change. Where before the flow would be controlled by throttling the valve, now the valve would be permanently fixed to the fully-open position, and the flow controlled via the speed of the pump.
From the hydraulic efficiency point of view, this has some important implications. With a DOL motor, the operating point of the motor will follow the pump-specific head-flow curve, shown in grey in Figure 9. Whenever a smaller flow is required, for instance, the valve is ‘tightened’, resulting in an increased head and decreased flow.
By contrast, in a speed controlled pump, the head-flow point is following the so-called system curve shown below in blue. The system curve represents the natural ‘resistance to the flow’ of the system, or more technically the pressure loss at any given flow. Now, changing the speed of the pump changes the characteristic curves of the pump. Thus, informally expressed, for any desired flow, the speed of the pump is adjusted until the system curve and the pump curve intersect.
Importantly, this often represents a significant saving in energy. Indeed, the area between the pump and the system curve, highlighted below in yellow, represents the pressure loss in the control valve. In a hydraulic system, the flow times the pressure loss equals power lost as heat. Thus, removing this loss at the valve often results in very significant energy savings over time – although a minor portion of it may be lost again in the losses at the VFD and perhaps changes in the pump efficiency.
Thus, the approach for handling the VFD scenario is normally as follows. First, the system curve is estimated. This requires knowing the static head of the system, and the system flow (or head, or both) with the valve fully open. Ideally, these values are made available by the client. Alternatively, they can be estimated from the data. With these two quantities, the system curve is then trivially obtained.
Figure 9. Visualization of the head loss associated with valve control.
With the system curve known, analyzing the VFD scenario is then relatively straightforward. For each measured flow during the monitoring period, an alternative operating point is determined. This consists of determining the speed of the pump, such that the flow-head point lies on the newly-determined flow-head curve. Now, as the afore-discussed power loss at the valve has been nigh-eliminated, this new operating point practically always has a shaft power smaller (or rarely equal) as with the DOL-operated case, often by a significant factor. As modern VFDs are highly-efficient, the electrical input power is almost always reduced, too.
The results are then finally presented in a desired format. Commonly, the total energy cost over one year is calculated. For a visualization, a histogram of the distribution of the energy cost across different flow rates with the two scenarios can be an invaluable aid, too. An example is shown below in Figure 10.
Figure 10. Energy consumption histogram with valve and VFD control.
Pump-swapping scenario
In other cases, a VFD may not be the optimal solution. Typically, these cases see relatively minor variation in the flow, but with the pump always operating too far from its best-efficiency point. In these cases, a VFD may be an unnecessary up-front investment, and also suboptimal from the energy consumption point of view, with its small but significant losses of its own. In fact, it is often possible to utilize spare pumps already on-site, resulting in very low initial capital costs.
Instead, either swapping the pump for a different one, or lathing the impeller of the existing pump may yield better results. The goal of this change is such that the typical flows pumped fall within the best-efficiency point of the new pump, with the minor flow adjustments handled by the existing throttle valve system.
An example scenario is again visualized in the figures below. Figure 11 shows the histogram of the flow. Worth noting is the comparatively smaller variation of the flow rates, and them all being significantly below the best-efficiency point of roughly 100 l/s.
Figure 12 again visualizes the head loss at the throttling valve, both for the original pump and a significantly smaller pump. As can be seen, the smaller pump requires significantly less throttling within the observed flow range. However, it must be noted that the maximum flow that the pump can produce at the fully-open valve position is also reduced. Thus, care must be taken that the new pump is selected appropriately, with a suitable margin included for the head and flow rates.
Finally, Figure 13 again visualizes the energy savings potential, over a period of one year.
Figure 11. Flow histogram for the pump-swapping example.
Figure 12. Valve head loss illustration for the pump-swapping scenario.
Figure 13. Energy consumption histogram with old and re-sized pumps.
Conclusion
This white paper serves as a brief overview of energy auditing pumping systems, heavily focusing on centrifugal pumps driven by direct-on-line motors and controlled with a throttling valve. The paper follows the Viimatech Digital Energy Audit process, but most of the information presented is fairly general and applies more generally. As discussed, the VT audit aims to be an optimal compromise between a ‘rule of thumb audit’ and a traditional ‘full’ audit with specialist equipment and know-how, requiring a fraction of the man-hours and investment of the latter while still producing a significantly better-informed estimate compared to the former.
In the wider scope, energy auditing can be expected to explode in popularity in the near future. The moral and regulatory push for energy efficiency is continuously gaining momentum, and the savings potentials in the process industry – and any pumping installation for that matter – cannot realistically be ignored for much longer. At the same time, making informed decisions is becoming increasingly important. In the scope of pumps, auditing can indeed yield crucial information for choosing the best path forward.
References
Gulich, J. (2014). Centrifugal pumps, 3rd edn. Springer.
Holttinen, V. (2018). Oikosulkumoottorin laskentamallin toimintaperiaate ja tarkkuuden verifiointi [Bachelor’s thesis, Lappeenranta-Lahti University of Technology]. https://urn.fi/URN:NBN:fi-fe2018073033146.
Motiva (2011), Energiatehokkaat pumput. Online publication, funded by the Ministry of Employment and the Economy of Finland. Referenced 9.10.2023. https://www.motiva.fi/files/5343/Energiatehokkaat_pumput.pdf
Motiva (2020), Pumppujen hankintaopas, Online publication. Referenced 9.10.2023. https://www.motiva.fi/files/18150/Pumppujen_hankintaopas.pdf.
Pöyhönen, S. (2021). Variable-speed-drive-based monitoring and diagnostic methods for pump, compressor, and fan systems [Doctoral dissertation, Lappeenranta-Lahti University of Technology LUT]. https://urn.fi/URN:ISBN:978-952-335-624-5
Stepanoff, A. (1957). Centrifugal and axial flow pumps, 2nd edn. Malabar, FL, USA: Krieger Publishing.
Appendix A. Additional visualizations for the VFD Scenario
The following two pictures provide some additional visualizations for the VDF scenario. Figure 14 shows the behaviour in the flow-head plane. As the control valve is now permanently fixed to the fully open position the system curve (blue) does not change. Instead, varying the speed of the electric motor and thus the pump, changes the flow-head curve of the pump. The new curves have been visualized for a few rotational speeds. In each case, the steady-state operating point (dot) lies at the intersection of the pump curve and the system curve.
As can be seen, speed control would result in the pump operating at significantly reduced heads compared to the valve-controlled one. This follows from the fact that the head loss at the valve is now missing. The head loss is again visualized with the yellow shading.
Figure 15 then visualizes the dependency between the shaft power of the pump, and the flow rate. Again, the flow-power curves have been visualized for a few different speeds, along with the system-curve-imposed operating point for each speed. Finally, the yellow shading illustrates the power difference compared to the valve-controlled case.
Figure 14. Difference between valve- and VFD-control in the flow-head plane.
Figure 15. Difference between valve- and VFD-control in the flow-power plane.