Comprehensive measurement and analysis for cost-effective backup battery monitoring solutions

The infrastructure that supports contemporary society must operate with very high reliability. In order to ensure nearly 100% "failure-free operation time" or system availability, Internet server clusters and communication exchange centers mostly rely on a very mature technology - lead-acid batteries, while data storage centers use high-tech. Typically, these key nodes and many other important departments are equipped with backup power supplies. The first layer of backup power is generally an inverter, which provides power conversion for battery packs assembled from valve-regulated lead-acid (VRLA) batteries or sealed gel batteries with similar performance.

This old technology is popular for many reasons, not least the economics of lead-acid batteries and their outstanding reliability. But while outstanding, it is not perfect. VRLA batteries have a limited lifespan (typically designed for 12 years), and are often used as backup power in critical systems, but are replaced periodically. Failures can and do occur. In a typical backup power system, the batteries do just what their name implies—they are kept fully charged in anticipation of a failure in the primary power source. The fully charged state is maintained by a continuous, low-current “float” charge. If the float current falls below a set limit, the gases produced by electrolysis within the cell recombine. In this case, a float voltage slightly above the standard 2.27 V per cell can damage the cell. A small overvoltage will cause the electrolyte to release more gas than it can handle, and this unresolved gas will escape through the safety valve. If the cell temperature is too high, even the charging voltage is appropriate, which can lead to electrolyte loss.

Other failure modes include early sulfation, poor connection between pole and grid, poor connection between plate and grid, electrolyte stratification and accelerated grid corrosion. There is also a rare but catastrophic failure mode - thermal runaway, which is a failure mode unique to VRLA and GEL batteries and can cause explosion and fire. The only way to prevent thermal runaway is to monitor the internal temperature of the battery.

It is generally acknowledged that monitoring cell voltage alone is of limited use in detecting capacity degradation in lead-acid batteries. When a battery is degrading, it will typically show nominal voltage until a large current is discharged, at which point its capacity has already been severely degraded and the terminal voltage has dropped prematurely. Determining the condition of a battery by measuring the exact specific gravity of the electrolyte is not applicable to sealed VRLA or GEL cells; conventionally, the only way to verify battery capacity is to discharge the entire battery pack to a controlled level, which requires the batteries to be removed from service. In addition, deep discharges reduce the life of lead-acid batteries; this test regimen is often used to determine battery life in systems where backup batteries are regularly discharged and where the primary power source has high reliability.

Recently, non-intrusive electronic methods that provide continuous monitoring can detect individual cells approaching failure, a cost-saving approach that maintains overall system availability. Predecessors of such systems typically measured cell or battery pack voltage (battery industry term for multiple cells enclosed in the same housing)—although its limitations are well known—plus charge/discharge current and ambient temperature. Some systems have attempted to measure or infer the battery's internal resistance, with varying success.

LEM's Sentinel system is a leading product based on the transformation of simple basic parameter analog measurements, and has now evolved to the third generation, Sentinel III. It integrates analog and digital technologies on a single custom designed SoC (System on Chip) integrated circuit. The device is configured in a module that measures terminal voltage, internal battery temperature and internal impedance, and is a key element in the design of a system that can provide accurate measurements at a cost that is within the budget range of most backup system configurations.

The data logging system monitors the trend of the data over time to identify potential impending failures, based on the cell temperature and/or exponentially increasing internal impedance values ​​(Figure 1). All Sentinel III modules are configured with an external temperature measurement probe or patch that can be attached directly to the casing of a single cell or battery pack to track the battery temperature as accurately as possible.

Figure 1. Battery internal impedance is not a good indicator of impending failure. The exponential curve means that early failure is difficult to detect, but later performance deteriorates very quickly.

A mature technique can be used to assess the internal impedance of a battery while it is being used or charged. Usually, a weak AC voltage is superimposed on the floating DC voltage, the AC voltage and current are measured, and then the internal impedance is calculated based on the measurement results. However, this method has certain limitations. It can only handle exponential curve shapes. A single battery that is about to fail may appear to be in good condition before the data logger recognizes its failure trend; conversely, by the time the failure problem occurs, the battery may fail completely in a short period of time.

LEM has developed a more sophisticated algorithm that can detect declining performance of individual cells at an early stage. The result is a very reliable test method that can completely penetrate the energy layer of individual cells to ensure maximum reliability. It is based on the so-called Randles equivalent circuit, which represents the electrochemical cell as a circuit network composed of electrical components, each of which is related to a physical factor that makes up the individual cell. (See figure)

Figure 2. Randles equivalent circuit of an electrochemical cell.

Figure 3 shows the asymptotic curves of various parameters during the life of a single cell. The same behavior is also confirmed during discharge or capacity degradation. All impedance factors of the equivalent circuit follow similar curves; there are no major changes during the early failure or capacity degradation phase. If impedance is used as the main indicator of the health of a single cell, it will not give any meaningful indication unless the capacity has dropped by more than 25-30%. Since the industry standard is to replace cells that have dropped to less than 80% of the specified performance, it is obvious that possible failures must be identified as early as possible.

Figure 3 Randles parameters evolve with battery life or discharge. Different resistance parameters show the same curve shape, while the double layer capacitance shows early changes that can be detected.

However, in the Randles equivalent circuit, there is one parameter that changes early in the failure of a single cell (except for simple metal corrosion, which will be manifested by an increase in the Rm parameter), and this is Cdl, the double layer capacitance. The bottom curve in Figure 3 shows its characteristics; again, the shape of the Cdl curve is similar for a normal performing battery in the normal discharge stage and a failing battery that is assumed to be fully charged.

Monitoring Technology

This monitoring technology is not described in detail in this article, but is briefly introduced below.

Feeding the test signal to individual cells one by one eliminates the need to inject large currents into the entire battery pack and eliminates the need to interfere with the DC connection of external systems. The original algorithm was improved using a bipolar test signal, but the results showed that a unipolar signal was more reliable. However, when testing with a unipolar signal, a DC drift occurs. Simply eliminating this drift does not maintain the characteristics of the data set, which is necessary for accurate parameter determination. Rearranging signal pulses of different frequencies (including the test signal) using a frequency sweep can make the battery voltage response match the predetermined curve.

Once the potential drift profile becomes regular, the firmware algorithm can be set to model this drift and eliminate it, resulting in an average zero voltage data set suitable for direct input to the Sentinel algorithm. This method can reduce the drift error to less than 0.1% without causing significant distortion of the data set. Therefore, this algorithm can also be used in waveform measurements, resulting in higher accuracy of equivalent circuit parameters.

Many measurement functions and algorithmic processing are integrated into a monolithic integrated circuit. The Sentinel module can measure both individual cells and entire 12V batteries). Up to 250 measurement points are measured in modular form and the results can be submitted to the battery data logger, the S-Box, via a dedicated data bus. In large-scale battery systems, several such data streams can be combined and made available to local or remote upstream management systems using the web server integrated in the S-Box via a standard bus or Internet connection.

By using the measurement SoC to determine the true state of each cell, it is possible to provide more than just the detection of impending failures that a mature monitoring architecture would provide; other functions and services can also be set up.

For example, the internal resistance of individual cells within a battery pack often varies. This condition can cause problems over time. The SoC intelligent control system can quickly detect these individual cells, and the terminal voltage optimization system can divert the floating current around the individual cells that can no longer be charged...

Real-time charge management can extend battery life: At the same terminal voltage, the float charge current in VRLA batteries is higher than that in flooded batteries. This may accelerate anode plate corrosion and reduce the effective service life of the battery by up to 30%. Eliminating float charge for a certain proportion of the service life can reduce this adverse effect. However, this side effect on cycle life also has a benefit, which is to reduce the incidence of thermal runaway.

A battery mounting module can also provide terminal voltage and temperature records throughout its life cycle for use by manufacturers and users.

Over discharge protection: This device is common in charger/UPS systems, especially battery monitors, which terminate discharge based on the average single cell voltage to protect the battery. However, the terminal voltage of a poor performance battery may be much lower than the average battery voltage, and it has been discharged well before it reaches the termination voltage. Therefore, a high-precision dynamic 'Time To Run' algorithm has been developed to give a warning when any single battery is about to be exhausted.

Backup battery parameter monitoring must be as detailed as possible in order to generate the most accurate results for the battery status. This is not only a technical issue, but also an economic one. Avoiding the failure of in-use batteries is essential, but it is extremely uneconomical to replace batteries that are not yet near the end of their life. In addition to measuring the voltage, impedance and discharge performance of each battery, LEM also sets the monitoring of the internal temperature of the battery as a standard function; this is a world leader. LEM is currently developing a floating charge sensor using fluxgate technology with a resolution better than 10mA, no or almost no temperature drift, almost no residual magnetism after high current discharge, and higher measurement repeatability. Integrating these advanced features, battery monitors are no longer expensive additional systems, but extremely cost-effective overall life management systems.