Retrofitting BWR Recirculation Pumps with Adjustable-Speed Drives

Exelon Nuclear recently replaced the original motor-generator sets for its boiling water reactor (BWR) recirculation pumps at its Quad Cities Generating Station Unit 1 with adjustable-speed drives. We examine the actual energy savings, motor-starting characteristics, control accuracy and stability, and motor and cable thermal behavior of this retrofit project.

Exelon Nuclear’s Quad Cities (QC) Generating Station Unit 1 installed medium-voltage adjustable-speed drives (ASDs) to replace the original reactor recirculation pump (RRP) motor-generator sets in spring 2009. An earlier article (“Upgrade your BWR Recirc Pumps with Adjustable-Speed Drives,” November 2007) described the complex equipment selection process and potential operating and maintenance advantages promised by the retrofit. In this article, we compare post-retrofit operating experiences with our original design goals and share important lessons learned on this project that will be used to improve future projects.

Project Description

The project replaced the existing induction motor-generators (MG) that power the RRPs (Table 1) with a dual single-train ASD system that is expected to be more reliable than the original MG design. For example, to improve the reliability of the new ASD, a cell bypass feature maintains full voltage output even if one or two power cells were to fail, resulting in an expected system availability of between 99.99% and 99.999%. A more detailed description of the ASD system design was included in the earlier article.

Table 1. Original specifications of the Quad Cities Unit 1 reactor recirculating pump motors. Source: Exelon Nuclear

AREVA NP’s supplier, Siemens Large Drive Applications, designed and fabricated the ROBICON Perfect Harmony ASD installed at QC (Figure 1). This model is a water-cooled, third-generation, high-availability medium-voltage ASD. Table 2 presents key ASD design and performance specifications used for the retrofit project. We’ll use these specifications to guide the following discussion of the ASD system’s post-installation performance.

Table 2. Technical description of the Siemens Perfect Harmony adjustable-speed drives used at QC Unit 1. Source: Exelon Nuclear
1. Precise speed control. Exelon Nuclear recently completed an adjustable-speed drive (ASD) retrofit for the two reactor recirculation pumps at Quad Cities Unit 1. This photo shows the equipment after installation on the Unit 1 turbine deck. The ASD cooling water heat exchangers, located on the original motor-generator oil cooler pedestals located on the mezzanine below, are not visible. Courtesy: Exelon Nuclear

Reduced In-house Load.
Observed MG set efficiency was about 75%. When considering the elimination of two 40 horsepower (hp) lube oil pumps and one 60 hp ventilation fan and an expected ASD efficiency of 96.5%, estimated in-house load savings were 1.8 MW. Actual observed savings have proven to be 1.86 MW. The indicated drive efficiency found during tests is 97%, slightly improving the project’s net savings.

Soft Motor Starts. Figure 2 illustrates the motor inrush current reduction. The green line indicates the inrush current experienced by the original motor; the yellow line indicates the inrush current during a start on the ASD—a reduction of about 50%. This current reduction should reduce overall stress on the motor and slightly extend motor life.

2. Controlled start-up. This chart shows the start-up characteristics of the reactor recirculation pump before and after the ASD retrofit. Note the significant decrease in inrush current experienced with the ASD. Source: Exelon Nuclear

RRP flow regulation was sometimes unpredictable with the scoop tube positioning system used to adjust the MG fluid coupler and the generator voltage regulator. This old electromechanical system would not always respond to small speed demand changes and would occasionally move erratically. The ASD, in contrast, is electrostatic and has no moving parts associated with the drive output controlling the pump speed. As a result, the observed speed changes are crisp, accurate, and very stable.

As shown in Figure 2, it takes about 6 to 8 seconds after depressing the start pushbutton to achieve motor magnetization. Once breakaway torque is achieved, the motor speed ramps at a specified 5%/sec up to 20% speed, and then it changes to 2.5%/sec up to the minimum operating speed of 32%. At this speed, the operator will open the pump discharge valve and then demand speed increases as required.

Quad Cities uses a digital reactor recirculation control system to control the drives. This system provides a continuous speed demand—a digital signal equivalent to the analog 4–20 ma—to the drive. This signal range corresponds to a 20% to 102.5% pump speed range. The system is programmed to allow slow and fast changes in the raise speed and lower speed directions, as described in Table 3. Figure 3 illustrates the ramp rates that were repeatedly observed during factory acceptance testing, post-modification testing, and plant operation. No noticeable overshoots, undershoots, or hunting for the speed setpoint have been observed.

Table 3. Speed demand ramp rates for Quad Cities Unit 1 reactor recirculation control system. Source: Exelon Nuclear
3. Observed ramp rates. The ramp rates illustrated were observed during factory acceptance testing, post-modification testing, and plant operation. No noticeable overshoots, undershoots, or hunting for the speed setpoint have been observed. The top chart shows the acceleration or ramp rate when starting the ASD; the lower chart shows shutdown. Source: Exelon Nuclear

Speed can be adjusted in 0.1% (1.67 rpm) increments. The operators have become confident enough in the speed control accuracy (and, therefore, reactor water flow) that they approach the reactor’s power to flow limits within 2 MW thermal. This is an improvement over the 5 MW thermal that was possible with the MG scoop tube.

Input Harmonics. Input harmonics are a major concern when installing electrostatic equipment with front-end rectifiers. Historically, this kind of equipment can adversely affect the supply system, causing losses and overheating of upstream equipment. The multi-level pulse wave modulated technology of the Perfect Harmony ASD does not require input harmonic filters and can still meet the requirements of IEEE 519-1992, IEEE Recommended Practices and Requirements for Harmonic Control in Electrical Power Systems. Siemens provided input harmonics analysis that showed IEEE 519 would be satisfied.

Harmonics data were collected during uncoupled motor testing, during post-modification testing, during the primary system hydrostatic tests, and during power ascension testing. All results corroborated the analysis and proved that IEEE 519 requirements were met and that the drives are performing as advertized with respect to input harmonics.

Cable and Motor Effects. Infrared thermographic images taken of the output cables from the MG and the ASD, before and after ASD installation, showed no relative change in temperatures. Motor winding temperatures are continuously monitored during operation and show similar performance with both the MG and the ASD. There have been no noticeable changes in monitored pump-motor vibration either. Motor performance appears to have been unaffected by the ASD.

Cell Bypass Time. As described in the earlier article, when a cell fault is detected, the drive control system will momentarily turn off the output, perform a neutral shift of the associated phase angles and voltage magnitudes to balance the circuit, re-start the drive, and then catch and maintain the spinning load. The neutral shift technology is unique to the Perfect Harmony drive design. The duration of this evolution was specified to be less than 250 msec. Timing of the cell bypass was completed during factory acceptance testing. The QC drive exhibited an average time of about 180 msec for the cell bypass evolution.

The time for the complete evolution must include not only the electronics but also the bypass contactor, given the much longer response time associated with the mechanical movement of the contactor. The combined overall process time is measured in seconds.

A cell bypass was performed with a coupled pump motor during post-modification testing, prior to start-up. The reactor was at relative cold conditions and the pump was running at about 50% speed. Figure 4 shows the observed speed profile for a cell bypass, which is well within acceptable limits.

4. Redundant cell protection. The system is self-protecting: It performs a cell bypass after a cell fails. This figure shows results of a cell bypass performed with a coupled pump motor during post-modification testing, prior to start-up. The reactor was at relative cold conditions and the pump was running at about 50% speed. Speeds observed during the cell bypass were well within acceptable limits. Source: Exelon Nuclear

At this writing, QC has not yet experienced a cell bypass during plant operation. Even though a cell bypass may be a reactivity event, it is much less severe than a pump trip, which is the alternative to a cell bypass. A trip may occur if there is a multiple failure such that the bypass contactor also fails when a cell fault is detected.

A failure modes and effects analysis (FMEA) for the drive was performed during the project design phase to estimate the drive’s reliability. Failure modes were limited to ASD components and related sensors—in other words, to hardware only. The FMEA reported that an uncommanded speed decrease, which includes a cell bypass, has a probability of 0.2% over a 24-hour mission, or once every 1.8 years. The probability of a drive trip is 8.3% over a two-year mission, or one trip in 23 years.

Bypass Two Cells. The cell bypass feature of this drive is referred to as N+2; that is, the drive can perform at 100% even if two cells are bypassed. To test this feature on an operating reactor is too intrusive and would require shutting down the pump and drive to restore the bypassed cells, so this test was not performed on site. However, the drive was tested at the factory and was proven to be able to produce 60 Hz and greater than or equal to 4,304 kW with two cells bypassed.

Both the factory acceptance test and the post-modification test challenged multiple cell bypass scenarios, but only for uncoupled or unloaded motors. In fact, the configuration for QC can permit up to six bypassed cells before a trip occurs. It is unlikely that operation would continue to that extent, because pump performance would be derated after two cells are bypassed, except in the rare situation of one cell per phase being bypassed. This would yield the same results as two cells being bypassed in one phase.

Because online maintenance is not permitted on the power cells, the N+2 feature enables the plant to allow multiple failures without shutting down a recirculation pump. An outage and orderly shutdown or single-loop operations can be planned to accommodate restoration of the inherent cell redundancy. The drive must be de-energized to replace a faulty cell.

50C Environment. This drive model is rated for an ambient air temperature of 50C (122F). The drives are on the turbine deck at QC, where observed temperatures have reached 43.3C in the summer months. Summer months were relatively mild during the first phase of operation. During factory testing, the drive was enclosed in a tent with heaters such that ambient air was maintained near 50C for 100 hours (Figure 5). Inlet cooling water to the drive was maintained at the design value of 40C. All cabinet temperatures were maintained below rated values for the duration of the test.

5. Hothouse testing. A tent was installed over the equipment to maintain 50C while performing factory tests of the drive system. Based on this robust test, it is believed that turbine building temperatures will not challenge the equipment temperature ratings. Courtesy: Exelon Nuclear

Key Lessons Learned

Although the ASDs meet the project specifications and passed all the necessary performance tests, the installation experience did provide an opportunity to learn more about how to best retrofit these systems. The following important lessons were learned.

Excessive Drive Loss Fault. During initial start-up of the ASDs, the medium-voltage input circuit breaker would not stay closed. The fault observed on the ASD indications was “Excessive Drive Loss Fault.” This protection logic was factory-installed and is typically used during normal loaded operation. The fault appeared because the difference between the supply bus and the capacitor precharged condition was much higher than the expected 10%, which allowed a high in-rush current. The high in-rush current compared to a zero output looked like an excessive drive loss, resulting in the faulted condition. This event was found to be caused by excessive power consumption within the ASD system.

The system senses a short-circuit condition during this event on one of the transformer secondary windings or another internal location within the ASD. However, it is not expected during initial energizing of the drive.

The first part of the energization process for this drive is a precharge process. Each power cell has two or three large capacitors. During precharging, the capacitors are in fact charged with 480 VAC via a series of contactors. At the end of precharging, the capacitors should be at about 90% of the rated DC bus voltage, or about 1,013 VDC. Upon completion of precharging, the input breaker closes. The intent of this process is to prolong the life of the capacitors, although the stress on the capacitors is believed to be minimal. The expected life of these capacitors is about 25 years. Earlier generations of this drive exhibited a capacitor lifetime of about 10 years.

The bus voltage on the supply bus of the drives was relatively high because the plant was shut down and many loads that were usually running were off. The bus voltage was near 4,450 VAC. The input transformer tap setting was at zero, which should support 4,160 VAC ± 416 VAC. The initial thought was to change the tap setting to +5%, but this may have resulted in reduced performance once the plant was back to normal operating conditions.

The resolution was to desensitize the excessive drive loss calculation during the energizing process. Once the drive is running a loaded motor, the original sensitivity is restored. The start-up team quickly resolved this fault with minimal impact on the start-up schedule, and there has been no recurrence of the problem.

Medium-Voltage Low-Speed Hold. If the plant medium-voltage supply is degraded, a power cell or motor toque limit may be approached. To prevent a trip, or damage to the drive or motor, a medium-voltage low-speed hold was programmed for the drive.

A speed hold is initiated whenever supply bus voltage drops below 90%, which also correlated with the first alarm for medium-voltage low. During start-up of the recirculation pumps at minimum speed, a speed hold was initiated when a reactor feedwater pump was started, as both the reactor feedwater pumps and the RRPs are on the same bus. The bus voltage dipped below 90% of 4,160 VAC for about 6 seconds. The operators were not used to seeing this behavior before and identified it as a nuisance.

The team determined that the appropriate way to implement this feature was to relate it to a motor torque limit rather than to the raw bus voltage. Therefore, a linear relationship was developed relating input voltage to motor output current. This change was added to the system software, and its effectiveness was successfully tested during the Unit 1 commissioning in spring 2010.

Cooling Water Pump Suction Pressure Loss. The closed cooling system for the drive includes two pumps, heat exchangers, a surge or expansion tank, and a deionizer. The pump suction pressure should run between 5 and 10 psi. The QC drives have exhibited a loss of suction pressure over a few weeks’ time and require repressurization with air.

The surge tank is an assembly of two tanks paralleled together with metal to composite piping fittings. Unfortunately, the fitting were found to leak the air volume required to maintain the desired suction pressure, as the tank walls are relatively thin and don’t provide significant thread engagement. Sealants have been used to repair some of the leaks, including sealing a threaded cap at the top of the tanks.

Pump flow, noise, and vibration are routinely monitored, and no abnormalities have been discovered during the low–suction pressure conditions. The plant is closely monitoring and repressurizing until a new tank design can be provided by Siemens. A final improved design is expected to be available by December 2010.

Surge Tank Level Anomaly. The surge tanks are translucent, so water level can be observed through the tank walls, although a level gauge has been installed. The gauge is connected to the tank via flexible rubber hose. The indicated level can be affected if a loop seal is created at the upper connection of the gauge. The routing and length of the upper hose on one of the surge tanks has created a loop seal after the top hose was used to fill the tank, which prevented the water from draining. The hose was rerouted to prevent the possibility of a loop seal. Indicated tank water level now correlates with actual level.

One-Shot Redundancy. As described in the earlier article, the Perfect Harmony drive uses a redundant control system. The 480 VAC sources, 120 VAC sources, supervisory programmable logic controllers (PLCs), speed control microprocessors, and communication network are all redundant. However, many of the redundant components are not accessible during operation due to their proximity to active medium-voltage components. If an A-side component fails or power is lost, the drive will continue to run on the B-side component or source. However, if the component is not accessible, it cannot be restored with the drive online. Other components are accessible and are hot-swappable while the unit is online.

Even though the speed control microprocessors are accessible and can be replaced, they cannot be rebooted and resynchronized to the master processor during operation. These processors are controlling the firing of all the transistors in the power cells, and a hot resynchronization of that process is not recommended by Siemens. Exelon has requested that Siemens explore that process in hope of Siemens providing a future enhancement for online restoration.

QC has experienced several backup PLC failures. No PLC transfers occurred, because only the backup was affected. Alarms properly notified the operators of the failures.

In one case, Siemens suggested that the PLC be replaced, but doing so would have required a shutdown. This assumption was challenged, and an online resynchronization process was developed. Additional online software changes were required to ensure that successful online replacement processes were proven. The PLC was replaced and resynchronized with the drive online. Much testing was completed at the factory and on a site stand-alone drive simulator prior to performing this test in the plant. PLC redundancy was successfully restored. At this writing the root cause of the failure is being investigated.

The good news is that these single PLC failures did not cause trips, and the redundancy feature of the control system was successfully demonstrated.

Next, Exelon wants to make the inaccessible components online-accessible so that redundancy restoration can be maximized. However, even with the current design, a single failure of any control system component will not cause a drive/pump trip. Repair and restoration of redundancy can be planned and then implemented during a forced or a planned outage or during a planned single-loop operation.

James W. Morgan ( is a principal engineer for instrumentation and control with ILD Inc. ( On assignment to Exelon Nuclear’s corporate engineering department, he is the lead engineer responsible for Exelon’s fleetwide upgrade of reactor recirculation pump flow control systems. Timothy Gode ( is the reactor recirculating system manager for Exelon Nuclear’s Quad Cities Generating Station.