**4. Experiments to Verify the Proposed Algorithm**

A 3-kW ESS was implemented to verify the BMS algorithm of the ESS considering the battery efficiency.

The BMS algorithm proposed in this paper was applied to the ESS and the battery efficiency was tested during the charge-discharge process by charging several battery modules.

The internal resistance calculated from the battery efficiency was applied to the SoC. Then, the OCV, CCM, and proposed algorithm were compared and the SoC was confirmed in the case of a battery fault. The charge-discharge cycle was performed by converting the SoC calculated from the internal resistance of the battery into the charging-discharging time. Furthermore, the termination of the charge-discharge cycle was confirmed through the connection between the ESS and BMS in the case of a fault. In the additional part of the algorithm, the total efficiency of the ESS was further confirmed to verify its validity.

Figure 10 illustrates the ESS experiment hardware used in this paper, while Table 1 lists the experiment parameters. The PCS of the ESS consists of a two-level inverter, a full bridge converter, and a master controller. The output side comprised three lithium-ion battery modules (1 module: 24 cells × 4.2 V) and a BMS. The experiment was conducted using an oscilloscope and a laptop computer to confirm the operation.

**Figure 10.** ESS hardware configuration for the application of the proposed algorithm.

**Table 1.** ESS experimental parameters for the application of the proposed algorithm.



For the battery efficiency experiment, battery efficiency was confirmed by the chargedischarge of a faulty battery module and a normal battery module.

The profiling of the battery was carried out in four steps. The data were confirmed in the order of (1) securing the charge-discharge data, (2) deriving an equation through curve fitting, (3) performing the charge-discharge cycle, and (4) extracting the target data from the implemented correlation equation.

An experiment battery was proposed to verify the battery efficiency by configuring the battery with three modules and assigning modules 1 and 2 as the normal batteries and module 3 as the battery subjected to repeated charge-discharge cycles.

Figure 11 illustrates the efficiency graph of the battery module.

During battery charging, the difference in the final internal resistance values of the battery was confirmed, as depicted in Figure 11. If a specific range was set during the charge-discharge cycle for testing, the change in the state of the battery caused by aging was detected.

The battery efficiency test revealed a significant change in the efficiency of the battery after investigating the changes in the efficiency of the faulty or abnormal batteries that occurred during the charge-discharge cycle of the ESS and those of the normal battery. The difference between the efficiencies of the faulty (aged) and normal batteries was 38.4%. The results suggest that the battery efficiency of the proposed algorithm could be applied for predicting the SoC and SoH, which requires improved accuracy, while the change in the internal resistance (which has the greatest impact on the battery state) could also be applied to increase the accuracy of the battery state prediction.

**Figure 11.** Battery efficiency difference profile graph according to battery power.

Figure 12 and Table 2 illustrate the SoC profile of the battery to which the proposed battery efficiency equation was applied.

**Figure 12.** SoC comparison profile graph.



All three normal battery modules were discharged up to 20% and charged up to 80% of the maximum SoC.

By applying the battery efficiency, the OCV, CCM, and proposed SoC algorithm could be compared.

The SoC profile was confirmed using the proposed algorithm.

To confirm the SoC calculation, the OCV and CCM were compared with the proposed SoC calculation algorithm. The CCM was charged after accurately determining the initial value. The OCV could not accurately determine the SoC during charging. The CCM and proposed SoC operation seemed to accurately calculate the SoC; however, when using the actual CCM, the user could not directly and accurately set the initial value. As such, using the algorithm proposed in this paper, the SoC can be determined more accurately.

Figure 13 illustrates the SoH profile to which the proposed algorithm was applied, while Table 3 presents the CC termination time based on the battery state.

**Figure 13.** SoH profile with CC charging time, to which the battery efficiency was applied.


**Table 3.** SoH table for three battery modules to which the battery efficiency was applied.

50 cycles were charged and discharged at 0.3 C-rate, and CC charging time was compared in the 51st cycle.

The battery was charged by applying the internal resistance to which the battery efficiency was applied. The results demonstrated that the CC charging time of the module decreased when the battery failed or had other problems.

Equation (10), which compared the *SoH* profiles obtained using the three methods investigated, confirmed that the *SoH* prediction was possible based on the CC termination time of the battery. The ∆*SoH* is the amount of change between *SoHbefore* and *SoHafter*, while *SoHbefore* is the *SOH* before *SoHafter*.

$$
\Delta SoH = \left(1 - \frac{SoH\_{after}}{SoH\_{before}}\right) \times 100\tag{10}
$$

It is difficult to accurately diagnose faulty batteries based on environmental changes, such as battery aging. Because the characteristics of the battery vary when a cell comprises modules, the internal resistance and capacity deviation occurs, causing overdischarge; thus, because the safety and energy efficiency of the battery system is significantly reduced, in this paper we diagnosed the battery state using two methods, whereby faulty batteries were diagnosed based on when the (1) battery efficiency and (2) SoH battery efficiency were reached.

Figure 14a shows the charging voltage and current waveform at the time of a fault signal, while Figure 14b is the discharge voltage and current waveform at the time of a fault signal. Figure 14 illustrates the fault signals when the battery efficiency is reduced and the SoH is 40% or less. If a fault is detected, the charge–discharge cycle of the battery is terminated with the general purpose input output (GPIO) signal, which cuts off the battery MC through the BMS. The main controller then terminates the pulse width modulation signal, causing the ESS to enter into a stop state.

**Figure 14.** Waveforms for battery fault diagnosis: (**a**) charging voltage and charging current waveform; (**b**) discharging voltage and discharging current waveform.

Figure 15 illustrates the efficiency waveform of the ESS when the system was implemented by applying the proposed algorithms.

**Figure 15.** ESS efficiency waveform when the proposed algorithm was implemented.

The efficiency of ESS is caused by the decrease in the difference between the power consumed by charging and the power generated by discharging; therefore, the operating cost for using the battery increases. Efficiency was measured when applying the proposed EMS and BMS algorithms. When the algorithm proposed in this paper was applied, the maximum efficiency was 97.57%.

This paper proposes a BMS algorithm for an ESS. To apply the BMS algorithm to the ESS, the experiment was conducted by deriving the internal resistance of the battery from its efficiency. Moreover, the increase in battery state accuracy was verified through experiments by applying the battery efficiency to the SoC with the OCV and CCM and the SoH considering the charging time. Furthermore, increased safety through the diagnosis of faulty batteries was verified through experiments.
