*4.5. Energy Arbitrage*

Energy arbitrage is the process of simultaneously purchasing and offering energy supplies in the marketplace. It was only initiated by commercial users because the power sectors of most countries do not have any form of regulation. The application of BESS pairs with DG or load, in which storage units are utilized to redirect energy production or generation, is aimed at maximizing profit irrespective of the fluctuations in market prices [43,52]

#### **5. Battery Energy Storage Technologies**

LA, Li-Ion, NaS, and RF are grid applications' most common battery technologies. These are classified according to their energy density, efficiency, lifespan, and cost when coupled to a storage network, as shown in Tables 6–8. The LA battery has high efficiency between 80 and 90% and low costs within the range of 50 to 600 \$/kWh [52,53] However, when compared to other technologies, it has a significant disadvantage in terms of lifespan (approximately 2500 cycles) [54] and low energy density (within the range of 20 and 30 Wh/kg). A high discharge depth shortens an LA battery's life [52,55].

The characteristics of Li-Ion batteries are based on the chemical composition of both the cathode and anode, which typically consists of graphite and lithium metal oxide. Interestingly, the cathode and anode give the battery its name and power, respectively. This technique is highly efficient, with a maximum efficiency of approximately 90%. On the other hand, some commercial devices boast reported round trip efficiencies of more than 95% with energy density within the range of 90 to 190 Wh/kg [56] and extended service life of relatively 10,000 cycles [54]. Cell temperature, an essential element in the deterioration process, significantly affects the battery life [30]. Li-Ion batteries are commonly found in electronic devices and recently emerged as the industry standard for EV. This technology is suitable for grid-connected network applications, even though it is still somewhat expensive. Presently, there are several Li-Ion technologies, for example, lithium manganese oxide (LiMn2O4), lithium cobalt oxide (LiCoO2), lithium nickel cobalt aluminum oxide (LiNiCoAlO2), lithium iron phosphate (LiFePO4), and cobalt-based Lithium nickel manganese oxide (LiNiMnCoO2) [57]. Tables 7 and 8 show details of the Li-Ion and nickel-based battery specifications, respectively.

NaS batteries have a high working temperature (approximately 300 ◦C), efficiency (>80%), energy density within the range of 150 to 240 Wh/kg, and a long lifespan of relatively 4500 cycles [58,59] As a result, this technique has been utilized to lessen the effect of renewable energy-based generators as an in-grid [58,60]. Vanadium redox flow batteries (VRB) batteries comprise two containers, one containing two chemical reagents and the other two electrodes partitioned by a membrane. Incidentally, when the two components combine, it results in an oxidation reaction. One of the containers holds the chemical reagents, while the other contains the electrodes. The amount of stored chemicals contributes to the flow cell's total energy capacity. Meanwhile, the electrodes and membrane filtering system are responsible for individual energy capacity flow cell. The power and energy ratings are separated, resulting in the increased design and operational flexibility. The energy density of VRB is relatively low, ranging from 15 to 30 Wh/kg, and its efficiency is approximately 75% in some cases [61]. On the other hand, they are not constrained by reactant life cycles or discharge depth [62]. Due to the low costs involved in their maintenance and operation, VRB have been suggested as viable options for large-scale grid-based energy storage [63]. The reactants have been investigated, and several chemical compositions have been proposed. The most utilized ones are vanadium and Zn-Br [64].



**Table 7.** Specification of technology lithium-ion batteries [70,71].


**Table 8.** Specification of technology nickel batteries [69].


#### **6. Battery Degradation**

Battery degradation leads to a reduction in its capacity and efficiency and even safety problems. The term cycle life refers to the total number of times a battery can be discharged or charged before it is replaced [72]. Nonlinearity in battery degradation can be traced to a variety of causes, such as SOC, high temperature, depth of discharge (DOD), and charge or discharge current rate [73], as shown in Figure 5. One of the issues contributing to the short lifespan of Li-Ion batteries, for example, is the highly utilized DOD, which tends to significantly reduce the total number of cycles [74,75].

**Figure 5.** Relationship between battery capacity and SOC, DOD, and cycle life Li-Ion battery [38].

The remaining useful life (RUL) and state of health (SOH) are the most critical factors in predicting Li-Ion battery degeneration. Generally, usage capacity, energy, and accessible power, which diminish with battery age, influence SOH and RUL [76]. Although SOH tests detect a decrease in performance, they also prevent potential accidents [77]. The accuracy with which one may anticipate the RUL of a given battery capacity relies on several factors, and the most important is the ability to calculate the SOH. Managing discharge problems, improved performance, and optimized operation requires precise and reliable prediction algorithms to determine a battery SOH and RUL.

SOH refers to the percentage of a battery cell's capacity that is still usable and used to quantify the entire aging degree. This value is expressed as a percentage [78] and ideally, the SOH of the new battery should be 100%. The decreasing trend of SOH is due to the accelerated aging of the battery, which is one of the reasons of the increased cycle times. When the state of health reaches the failure threshold, the battery becomes ineffective [79]. The formula for SOH is written in Equation (1).

$$\text{SOH}(t) = \frac{\text{C}\_t}{\text{C}\_0} \tag{1}$$

where *Ct* and *C*<sup>0</sup> denote the *t*-th cycle and initial battery capacity. The maximum capacity of the battery tends to drop in accordance with the number of times it is cycled, with continuous increase in the battery's internal resistance. Generally, a battery fails when its internal impedance increases to a level that is twice as high as its initial impedance. Several performance parameters, such as power and the number of charge and discharge cycles, can also be used to define SOH. Further studies must utilize a wide variety of methods or models to estimate SOH, such as the use of direct measurement and indirect analysis. By measuring the standard aging characteristic parameters of the battery, the direct measurement technique determines the value of its current capacity, internal resistance, cycle times, etc. This is the technique through which the values of the current state's identifying parameters are determined. Examples of direct measurements are counting ampere hours, cycle numbers, measuring internal resistance and impedance. The indirect analysis consists of obtaining the SOH value by estimation based on online observable data from health indicators that have a high link with the performance and characteristic parameter degradation that occurs with the SOH condition. Model-based analysis, datadriven analysis, and hybrid analysis are examples of indirect analysis [80].

Wei J et al. [81] monitored the estimated diagnosis of battery SOH with three stages. In the initial stage, a particle filter (PF) technique was initiated, followed by the execution of a procedure to update the particle's time. The support vector regression (SVR) model was also used to estimate the capacity in each battery cycle number in the second stage. This SVR model is trained with characteristics collected from sensor data during constantvoltage (CV) charging mode at cycle number, to determine the charged capacity. The third stage updated the particle constitutes, which can be resampled based on their normalized importance weights. In accordance with the PF-based estimator, the anticipated capacity at the cycle number is considered as a Gaussian distribution, whose variance and mean are obtained. SOH is further defined as the ratio between the capacity of a new battery and the expected capacity. In general, the SOH estimation flowchart can be seen in the flowchart in Figure 6.

**Figure 6.** Block diagram of SOH estimation in general.

RUL refers to the information on the remaining life of a battery. It is imperative to change old and damaged batteries whose SOH has reached 0%, to guarantee the safety of the system and hence prevent problems [80,82]. The formula for RUL is written in Equation (2):

$$\text{RUL}(t) = t - t\_{\text{col}} \tag{2}$$

where *t* and *teol* denote the *t*-th and number of cycles remaining at the completion of a battery's life. It is difficult to compute the RUL of a battery due to several variables, such as its present health condition, historical data, and failure. Therefore, further study needs to be conducted on the prediction of batteries' RUL. Presently, there is no standard framework that is considered the optimal model for estimating RUL due to a lack of available data, model complexity, and system limitations. In general, RUL prediction methods can be categorized as physics-based, mathematical, data-driven, or hybrids [80].

Wei J et al. [81] also predicted the RUL of a battery using the SVR-based model using a flowchart as shown in Figure 7. Monitoring the prediction of RUL starts with developing a model that has been trained using extracted sensor data features and predicted capacity for SVR-based input models. Wei J. et al. applied the average degradation parameter to characterize the expected capacity distribution in this section. The result showed that RUL is considered then+1 after the predicted capacity has reached the EOL threshold.

**Figure 7.** Block diagram of RUL prediction in general.

The diagram in Figure 8 illustrates the connection between SOH, RUL, and the modeling of battery degradation. Some preliminary research developed a battery deterioration mechanism model using a framework that incorporated SOH and RUL [76]. The elements that influence general battery deterioration and failure were further explained in the SOH estimation model. Furthermore, its diagnostics and estimation help boost RUL battery modeling by determining how much time or cycles are left to attain 80% SOH. As a result, the reliable prediction of SOH and RUL is required for modeling battery deterioration behavior.

The SOH of a battery is measured in terms of its present ability to supply a certain quantity of energy in comparison to the initial capacity. At the same time, the RUL is helpful for monitoring the state of the battery and is also essential for executing operations that evaluate its degeneration. Due to the nonlinear nature of battery deterioration, it is necessary to have appropriate RUL estimations that are based on aging processes and suitable life models at various fading stages [76]. This entails calculating the time until a battery reaches its EOL. It tends to occur when the battery has reached the failure threshold. Moreover, the time left and the total number of charge-discharge cycles are considered [83]. The RUL estimation and degradation process are intimately linked to the working circumstances and dependability of Li-Ion batteries. Previous studies have reported that the successful prediction of the RUL prevents failure and timely functional maintenance without irreversibly harming the battery [84].

Scholars estimated the RUL using several different methodologies, as shown in Figure 9. These tend to be broken down into one of the four categories, namely based on physics, mathematics, data, or hybrid models. The amount of time a battery is going to be valuable is evaluated using a model-based technique. Therefore, a model that is representative of a battery application found in the real world, as well as an estimated algorithm used to predict voltage or other characteristics, needs to be developed. Empirical, analogous circuit and electrochemical models, including Kalman filters, are a few examples of the various methods that fall under this category. Data-driven RUL estimation is a prediction method that collects excess information and continues recording until battery health reaches its limit. Meanwhile, applying a hybrid model implies combining a model-based method with a data-driven model [76].

**Figure 9.** Classification method estimation RUL battery [76,82].

Table 9 reviews variables used to optimize BESS capacity size and placement with battery degradation models, which vary in different studies. Aside from the SOH and RUL models, preliminary research also used fading capacity and residual battery life for BESS optimization. Table 10 reviews the algorithm used for battery degradation models for BESS optimization.


**Table 9.** Review of variable features used in battery degradation models for optimization of BESS sizing and siting.

**Table 10.** Advantages and disadvantages of battery degradation algorithm for BESS optimization.


Battery lifespan is influenced by calendar and cycling aging. However, this is also determined by cycle or float lives [93]. Even though the computation of the BESS life value tends to be inaccurate, its datasheet is dependent on two limits, cycle and float lives. Both restrictions are measured in years, and when the BESS maximum life is equal to or exceeded by its float life, it is said to have a floating life equal to or exceeds its maximum life. The cycle life is represented as the maximum number of charge and discharge cycles that can occur prior to the BESS failing, and it varies depending on the technology of both the BES and the DOD [38].
