When a major broadcast network in New York needed an innovative solution to keep uninterruptible power supply (UPS) battery maintenance costs down and restore the integrity of its backup DC power supply systems, the stakes were high. Historically, the network relied on a distributed network of static UPS systems with integral valve-regulated lead-acid (VRLA) and flooded cell batteries to maintain the critical broadcast elements. Increasing battery failure rates combined with escalating maintenance compelled the network to consider an emerging alternative technology: the flywheel energy storage system (FESS).
Since the network was already familiar with “batteryless” UPS systems (motor-generator sets provided unmatched power quality during the 1980 Winter Olympics), it considered both static and kinetic UPS units on this project. Additionally, with the impending conversion to digital broadcasting and the intrinsic growth of critical load, the network realized its broadcast infrastructure would also need an upgrade. The search for a FESS/UPS combination began.
The first challenge was to research and evaluate FESS products and develop site-specific performance criteria. We looked at the obvious measuring sticks: rating, ride-through time, dimensions, weight, cost, nearest service center, efficiency, mean time between failure (MTBF), and service intervals. With such a new technology, however, we focused on three key aspects: installed base, confidence, and FESS/UPS compatibility.
At the start of this project, there were less than a dozen functional sites in the United States where FESSs were supporting the DC bus of UPS systems. Furthermore, the majority of them had not yet logged a year of continuous service. Today, there are more than 50 sites—most of which have a chemical battery paralleled with a FESS.
This led us to another point regarding the confidence aspect. Despite the marketing claims backing this new technology, you can only measure a good product through years of successful performance. The FESS technology (as applied to the critical power arena) just wasn't there. Although the MTBF of chemical batteries is relatively poor, there is a certain comfort level that doesn't exist with the FESS product. The ride-through of a FESS battery is typically less than a minute, while most chemical batteries are sized for 10 min to 15 min.
This led us to consider a chemical and kinetic battery combination for the project. We planned to configure the chemical battery as a redundant backup to the flywheel, which would also extend the UPS ride-through time.
In terms of compatibility, our next step was to determine the level of experience that the UPS vendors had with the FESS products. Although the power cable connections are the same for a chemical battery, the dynamics of the power exchange and the discharge cycle for a FESS are quite different. We later discovered the importance of several of the UPS EEPROM settings and the dramatic effect they had on the “usable” ride-through time of the FESS.
After narrowing the selection down to UPS and FESS suppliers who had proven success with combining these technologies, we ultimately learned all three static UPS bidders had about the same amount of experience with FESSs. After reviewing the bids, the network placed an order for two stand-alone 240kW static UPS systems with 240kW kinetic batteries.
Chemical Battery Integration
The network's engineering staff was keenly aware that frequent power quality events would subject the battery systems to short-cycle discharges. From previous experience, they knew the battery would experience a short discharge about every other week.
The chemical battery life could be extended if it was normally isolated from the UPS DC bus. The FESS would then be used to absorb power quality events and provide ride-through to the generators for utility outages.
The system would engage the chemical battery if the diesel engine did not start on the first crank or for a FESS failure. Although the traditional approach calls for paralleling an individual chemical battery with each FESS, the decision to use a single chemical battery to back up both FESSs was driven by space and cost limitations.
We chose to isolate the chemical battery through a normally open contactor/blocking diode scheme. This setup closes the contactor when a standard contact closure in the FESS signals that it has discharged to less than 15% of its capacity.
If you're wondering why we did not simply parallel the FESS and chemical battery to the UPS DC bus, read on. This is where it gets interesting.
We were told you can set the FESS to discharge before the chemical battery, thereby using the FESS as the “sink” for PQ events and protecting the lead-acid battery from cycling. However, that's only partially true. Battery cycle monitors do not differentiate a chemical discharge of the battery from a “surface or capacitive” discharge. A chemical discharge is the one we all know well; it takes place when the UPS bus voltage drops below the battery's nominal rating (which is 480V for this facility). A surface discharge will occur if, for example, you lower the UPS DC bus voltage from a 540V float to 490V. Either of these events will result in a discharge cycle count.
In our application, we set the FESS discharge target regulation voltage at 485VDC. The FESS model required approximately 50 ms to 100 ms to fully ramp up the generator field coils. This is the transition period that an electromechanical machine undergoes from a motoring to a generator state.
When the UPS rectifier turns “off,” the UPS DC bus voltage drops to the discharge trigger of the FESS. At this instant, the capacitors located in the UPS momentarily support the DC bus, and the FESS increases its field currents to support the DC voltage and the critical load.
During this 50 ms to 100 ms transition, the FESS arrives at full power generator state (with batteries in parallel), and the DC bus drops slightly to a point somewhere above the open-circuit voltage of the battery. The chemical battery also experiences a short “surface discharge” that the battery cycle monitor registers as a discharge event, reducing the battery's warranted service life. That is why we chose to preserve the life of the lead-acid battery by isolating it through a normally open contactor circuit.
During factory testing, each FESS endured a capacity test with 240kW of load on the UPS output. The input breaker to each UPS module was opened. Each FESS was proven to deliver 240kW at 12.5 sec. At first, we thought the capacity test was the more rigorous integrity test for the FESS. However, we later found the UPS system (module and FESS) literally could not endure a loss of AC input to the UPS module for 12.5 sec.
To appreciate why, it is important to review the dynamics of a UPS module and physics of the FESS. First of all, the DC source must support the load during a loss of UPS input power. Once the UPS input has been reenergized, the DC source must ride through the rectifier “turn-on” delays and walk-in period.
The rectifier turn-on delays consisted of two delays: one after the input returns and one to match the rectifier output voltage with the DC bus voltage now supported by the flywheel. Adjusting the firmware EEPROM settings reduced the turn-on delays from 5 sec to ½ sec. This gave us a big gain in ride-through time.
The second factor to consider when you reenergize the UPS input is the rectifier walk-in rate, which is normally set up at amps per second. In our case, we set it for approximately 150A/sec—an increment we felt the standby generator could safely accept.
Therefore, our walk-in period at a full module load of 240kW was approximately 2 sec. Subtract this and the ½ sec “turn-on” delay from the 12.5 sec FESS run time, and the result is approximately 10 sec. Consequently, the UPS system could now endure only a 10 sec loss of input power.
The electromechanical nature of the FESS challenged the ride-through time even more. The FESS was set to discharge at 510VDC. This setpoint is what triggers the machine to a generator state.
When you reestablish the UPS input and transition the load to the rectifier, the FESS will remain in a generator state until the UPS DC bus voltage climbs to 515V. Above this, the machine becomes a motor and recharges. This can take an additional 6 sec to 7 sec—during which time the flywheel continues to discharge. We optimized this delay to approximately 2.5 sec.
After considerable collaboration between the vendors, owner, and engineer, we achieved our goal. The FESS/UPS settings were optimized so that the kinetic battery alone supported both power quality events and utility power outages. The chemical battery acted only as a redundant backup.
The FESS's discharge voltage trigger is set approximately 110VDC+ above the UPS DC undervoltage setting. This permits a safe margin to allow the FESS to transition to generator operation, during which time the DC bus voltage begins to collapse.
The FESS caps provide the ride-through energy during this 50-ms to 100-ms period. It's important that the DC voltage not reach the undervoltage settings during this transition period. If it did, the inverter would shut down and drop the critical load.
Ideally, the FESS should transition to its motor state immediately following the rectifier walk-in period. However, this requires that the UPS module DC bus voltage increase rapidly without detrimentally affecting the AC output voltage. Obviously, you have to strike the right balance.
In addition to meeting the network's system requirements, we feel this relatively new application will prove useful to the UPS industry. It has long been theorized that ripple current from the rectifier switching has adversely affected battery life. It is also well known that the number of discharges affects battery life.
In theory, isolating the chemical battery from the UPS/FESS through a normally open contactor scheme should allow the network to greatly extend the useful life of the chemical battery. Whether this theory holds up in actual practice will be interesting information the broadcast network will share with other UPS users.
As you can see, without careful planning for integration of the FESS and UPS, the system could require more ride-through than the full load rating of 12.5 sec provided by one flywheel. By making the correct adjustments to the UPS, we were able to work with the minimal ride-through for full load without adding extra flywheel ride-through time.
System designers should consider using a FESS that is rated for 20 sec to 30 sec at full load. Although the cost will be greater, all of the integration/timing issues that we faced are much less critical.
In upgrading the reliability of its power system, this broadcast network chose to use a new technology to meet critical power and infrastructure requirements. Recognizing that the FESS was different from other time-tested solutions, the company identified and met stringent research, testing, and evaluation criteria. As a result, the FESS solution is spinning the network into high gear.