The Prescient Data Center Manager

How to predict standby battery life.

Today’s modern UPS systems employ cutting edge technology, and it is tempting to believe that deploying that technology is enough to protect mission-critical applications from a power outage. But here’s a simple truth: the UPS draws its power from banks of lead acid batteries, and those batteries can go from good to failing in a matter of days. Even the latest and greatest UPS system provides no protection without fully functioning batteries.

There are a number of ways to manage and maintain the batteries. The available labor resources, cost, and the importance of the load the batteries support all influence an organization's choice of techniques.

Labor resources not only include the amount of labor available but the quality of that labor and other cultural concerns. For instance, in some nations labor is plentiful; battery maintenance and monitoring often is carried out continuously and by hand. In some of these regions, the staff is reluctant to bring bad news to a manager, especially if the messenger was supposed to maintain the prematurely failing batteries. In other societies, organizations cannot afford a staff to continuously monitor batteries.

Most professionals would like to be able to see into the future to know when their batteries would fail. Instead many companies replace 10-year life batteries every three to five years. They try to anticipate the failure of their batteries; they try to peer into the future. A review of battery monitoring and maintenance methods shows that professionals do have the ability to look at historical data right through to looking at predictive data.

Historical Failure Rates

Organizations that keep good records know the age of a battery that dropped a load, dumped the prime web site, lost customer dollars, lost a client, and cost someone a job. Battery replacement is based on the principal that new batteries should be installed just before the age they were when they failed last time. But what happens if the data is not available?

Checking batteries once a year tests to see if the battery failed during the preceding 12 months. This is looking into the past to try to avert future problems. A lead-acid battery can go from useful to useless in less time than it takes to order up a replacement and schedule the people to come out and fit it.

Valve-regulated, lead-acid (VRLA) batteries often cannot take the load at time of install. Some of these batteries recover over the first three months of life, sometimes they just don’t. Others fail in a matter of 10-15 days during the remainder of their life. Scheduling more frequent testing is little help, as they could start to fail the day after the test and be useless in time of need, two weeks later.

Also consider the nature of the test. The portable handheld tester has a number of shortcomings. The condition of the cables used, or more specifically, their state of repair can affect the result considerably. Many working and good batteries have been replaced by accident instead of replacing the test leads. How the cable connectors are put on the terminals also affects the results, either way. These days many VRLA batteries do not have connection terminals big enough to accommodate a test clip.

The quarterly visit can introduce other problems. Some refer to the routine maintenance schedule as planned destructive maintenance, because equipment that is working perfectly well can be affected by tampering during maintenance. Frequently (more often than most battery technicians would like to admit) breakers get left open, intercell straps are left off, plugs are left open and bolts left finger tight.

Figure 1. Modern battery monitoring systems can show at a glance if a battery parameter exceeds a threshold and requires further investigation. This screen shot shows a string of 40 jars with one in alarm mode (red).

The Alternative

A permanently connected monitoring system can replace the hordes of workers needed to manually monitor batteries in low-cost labor regions, with additional benefits. The first step is to install a battery monitoring system and understand how it works and how to use it. Knowing how to adjust it when necessary is critical.

But can the monitoring system tell an organization anything useful about the future? The answer requires thinking about the periodicity of the measurements and what parameters the monitoring system measures. A permanently wired battery-monitoring system would be better than a handheld device carrying out the same tests every three months because variability of connections and doubt concerning the state of the test leads would be removed from the equation. But why stop at quarterly tests? Surely the more data points the wired system can gather and compare the better the detail. More detail helps identify problems earlier before batteries fail.

Figure 2. A permanently wired battery monitoring system featuring data collection modules (DCM) attached to each battery cell/jar. Each DCM is connected by fiber optic cable to a control unit that transmits data to a dedicated PC (iBMU). Modern systems will also expand to include generator and switch gear/telecom batteries.

Ideally, a permanently wired battery-monitoring system enables readings of the float voltage of every component of the battery every day. That includes every cell, block, jar, or even smaller minimum changeable component of the battery. If over several days the float voltage of a cell trends downwards compared to others on the string, the system has identified a problem. Finding a problem means that the battery should be investigated without panicking. The battery may need replacing, but other outcomes are possible.

Figure 3. This installation shows a permanently connected (OV) battery monitoring system installed neatly onto the battery.

Is the float voltage related to the performance of the battery? It can be, meaning that it can also be unrelated. In general the float voltage of a bad cell will be lower than its peers. But sometimes the cell has to have become useless before voltage alone can be used to identify it. So what then? The answer is Ohmic Value.

Figure 4. Data collection modules (shown attached at the front side) collect information including Ohmic Value and voltage data from each battery cell/jar and transmit it to a dedicated PC.

Ohmic Value

Before Ohmic Value, users could measure resistance, impedance, and conductance. The proper use of these metrics confused users, leading to a number of white papers and perhaps even some bad blood between proponents of competing techniques.

IEEE developed the concept of Ohmic Value (OV) to clear up the confusion. Ohmic Value is a figure that indicates how the battery behaves under a small load. The figure is not an absolute, comparable figure, but a relative measurement. It is relative to the OV on the battery when new and the OV of all the other cells in the battery.

Typically OV will rise and be detectible as faulty three months before voltage falls and is visible as faulty. The owner of a wired-in system with Ohmic Value has at least three months better forward vision than a system that relies only on voltage. The key is to be aware of the periodicity of OV measurements.

Figure 5. This report shows historical OV data for three years of a battery’s life, including three instances of battery failures and replacements. Notice in particular how much more quickly the data center manager replaced the faulty jars as time went by.

If 12-volt VRLA blocks can go bad in 14 days, then OV must be measured, recorded, and compared (analyzed) every day. This frequency provides sufficient time to investigate, order, and replace parts before the power system integrity is put in jeopardy. ‘Every day’ measurement of OV didn’t become practical until recently. To measure OV, a current must either be injected into the battery or removed from the battery for a very specific and accurately measured amount of time. During this process voltage readings are taken and compared to voltage readings taken immediately before the test took place. The difference between these voltage readings indicates the Ohmic Value of the cell or battery. Unfortunately removing a large amount of current from the battery or removing it for an extended period of time can cause problems in a parallel/string charging arrangement that can result in damage to the batteries. To avoid this damage, early versions of wired systems with OV measurement capability took readings only monthly or every two weeks. Once again there is a period during which a battery could fail and the load could be dropped.

The amount of current used in the test has been the barrier to more frequent testing. The greater the current the less frequently the test can take place. So a very small current would be best. But changes in voltage from a very small current applied to UPS-sized batteries are almost impossible to measure, especially with a UPS charger and inverter running on the same power line. The electrical noise, ac ripple, and switching pulses swamp the tiny signal.

Electronic engineers eventually found a way to discriminate the tiny signals produced by smaller amounts of current for very short periods of time. This task is still difficult, especially if the measurement signals are backed onto the noise output from a UPS and requires a modern, state-of-the-art battery monitoring system.

A wired system removes the variability of test leads and the risks of human error, eliminating once and for all planned “destructive maintenance.” Ohmic scans provide voltage trending three months earlier than previously possible, and daily scans identify the day that a battery starts to trend away from its correct value compared to when it was new or compared to its peers.

The final detail is that the battery monitoring system must have the ability to turn itself into a voltage datalogger the moment the battery goes on-line. This datalogger then must take a battery reading every few seconds on each and every cell for the duration of the discharge. Then it can provide documented evidence of the value of an investment in a proper wired, predictive, battery monitoring system that can measure Ohmic Value.