Service life tests for storage batteries used in islanded power systems with renewable energy sources

We investigated the service life of storage batteries to provide recommendations on the design of energy storage systems used in islanded energy systems based on renewable power sources. The service life of maintenance-free, sealed lead-acid batteries produced by absorbed glass mat (AGM) technology was determined by endurance tests carried out by repeated charge/discharge cycles according to specified load profiles, implemented at a specialized Chroma Test System station. Three battery load profiles were simulated: one for the standard DC charge/discharge mode, and two for the charge/discharge modes from renewable energy sources. To this end, the actual data obtained from monitoring the operating modes of a wind power plant were used. It was found that the battery service life depends on the intensity of stress factors. Among them, the throughput factor has the most pronounced influence on the battery lifespan. To extend the service life of storage batteries, it is proposed to separate the charge/discharge modes in time. For batteries operated on renewable energy profiles, this approach decreases time intervals between full charges and at low battery levels, which increases the battery service life by 14%. A solution to designing an energy storage system for microgrids was proposed, which consists in the use of a combined double-circuit energy storage unit. An experimental prototype of such a unit with a power of 15 kW was developed. The use of a combined energy storage unit in the microgrid system: increases the battery service life by 20–30% compared to analogues; improves the static and dynamic stability of the local energy system with a response time of no more than 50 ms towards power change; allows a fuel replacement level of at least 25%; reduces the electricity cost by 25–30%.


INTRODUCTION
Islanded energy systems based on various generating units, such as microgrids 1 , are promising technologies for producing electrical energy. Microgrid technology was initially aimed at increasing the energy efficiency and environmental friendliness of autonomous power supply systems that incorporate diesel generator sets (DGS). Therefore, the developers relied on the use of renewable energy sources (RES), both wind power plants (WWP) and photovoltaic installations.
Microgrids can operate either in gridconnected or islanded mode. According to the Navigant Research analytical company, in 2018, the annual production of islanded and gridconnected microgrids amounted to 1,231 MW and 1,463 MW, respectively, with the total sales volume exceeding USD 3 billion. By 2027, the market demand for such microgrids is predicted to reach 4,230 MW and 11,576 MW, respectively, which will require investments of about USD 30 billion 2 .
Microgrid equipment has a large potential in Russia, whose eastern and northern areas are characterized by a low population density and a poorly-developed transport infrastructure. Relia-ble power supply in decentralized regions is an urgent state task, confirmed by a number of legal and regulatory documents 3,4 .
Microgrids incorporate energy storage systems, which significantly improve energy efficiency by increasing the installed capacity utilization factor of renewable energy installations, reducing the DGS operating hours, and, accordingly, decreasing fuel and service costs 5 .
Storage batteries (SB) as energy storage systems satisfy the requirements of microgrids in terms of power range and energy storage duration. Despite significant progress in the industry of electrochemical power sources [2,3], leadacid SBs are mainly used for storing energy in microgrids due to their good value-for-money characteristics. In this work, we also consider this type of SBs.
The expenses involved with energy storage are quite significant, amounting from 25 to 60% of the total cost of an energy system [4,5]. Therefore, it is important to increase the SB service life, which is typically lower than that of other microgrid components [6]. To that end, the main factors affecting the lifespan of SBs when used in microgrids should be identified. It should be noted that the operating modes of SBs incor-___________________________________ 1 Off-grid renewable energy solutions to expand electricity access: an opportunity not to be missed. International Renewable Energy Agency. Available from: https://www.irena.org/publications/2019/Jan/Off-grid-renewable-energysolutions-to-expand-electricity-to-access-An-opportunity-not-to-be-missed [Accessed 17th February 2021] / Off-grid renewable energy solutions to expand electricity access: an opportunity not to be missed // International Renewable Ene rgy Agency [Электронный ресурс]. URL: https://www.irena.org/publications/2019/Jan/Off-grid-renewable-energysolutions-to-expand-electricity-to-access-An-opportunity-not-to-be-missed (17.02.2021 porated in microgrids differ significantly from those in direct current (DC) power supply systems (DCS).
The concept of charge/discharge cycles and methods for determining the SB service life are defined in regulatory documents 6,7 . In accordance with these requirements 6,7 , each cycle consists of sequential DC charge/discharge steps, and the SB lifetime is determined by a number of cycles, following which the SB capacity is reduced to 80% compared to the nominal value. When SBs are applied as part of DCSs, the charge/discharge currents vary across the ranges and profiles recommended by the SB producers, ensuring their maximal service life. In such systems, the main factors affecting the SB service life are temperature and the depth of discharge. These dependencies are given in the SB technical specification, allowing determination of their service life when used in a DCS.
A characteristic feature of microgrids is a change in the generated and consumed power values over a wide range, which determines the presence of ripples in the SB charge/discharge currents. Operation of SBs in the impulse current mode leads to a rapid degradation of their performance and a reduction in their service life [7]. The stochastic nature of generation leads to incomplete SB charge/discharge cycles with significant time intervals between full charges and at low residual charge rates, which also diminishes the SB lifetime [8,9].
The main reasons for SB deterioration are physicochemical processes leading to electrolyte separation, water loss, electrode sulfur poisoning and corrosion, as well as loss of active mass. The intensity of these processes, or the deterioration rate, depends on both SB type and its operating conditions. SB deterioration mechanisms are described in [10,11]. The main factors affecting these processes are defined in [12,13]. A qualitative estimation of the effect of these factors on SB degradation is given in [13,14]. The aforementioned studies determined the main stress factors during SB operation, considerably accelerating degradation processes:  temperature;  сharge factor (CF);  throughput (Q thr );  discharge rate (DR);  time between full charge (TF);  time at low state of charge (TL);  partial cycling (PC). Determination of reliable quantitative relationships between stress factors, degradation mechanisms and the SB lifespan is a challenging research problem attracting widespread interest [15][16][17][18][19][20][21][22][23][24][25][26]. The high complexity of this problem is associated with the stochastic nature and mutual influence of stress factors affecting SB deterioration processes.
Mathematical modelling is the primary and, in many cases, sole way to predict the service life of SBs incorporated in microgrids. The majority of studies apply the following 3 types of models: performance or charge, voltage and lifetime. Performance models, where the SB service life is determined by the state of charge [15,16], are used most widely. Voltage models measure the voltage at the SB cleats, on the basis of which losses and the degradation degree are calculated [17]. Service life models rely on empirical relationships linking stress factors with the SB service life [18,19]. All these models can be used independently or integrated into a generalized SB model [20,21]. A comparison of various models is presented in [22,23]; the questions of practical application of SB models when selecting an optimal composition of microgrid equipment are considered in [16,24,25]. The studies discussed above aimed to develop SB deterioration models that reliably and accurately describe the effect of stress factors on the SB lifespan. However, a serious limitation to the practical implementation of RES-based microgrids is a lack of approaches to extending the SB service life and creating efficient energy storage systems.
In this work, we investigate the service life of SBs operated in microgrids to propose recommendations on the design of energy storage systems for use in islanded RES energy systems.

MATERIALS AND METHODS
Maintenance-free, sealed lead-acid batteries produced by AGM technology, CS3 battery trademark, GP 12120 model 8 were investigated. The main technical characteristics of CSB GP 12120 SB are given in tab. 1. Due to the high complexity involved with the mathematical description of electrochemical and thermal processes in SBs, direct experiment was chosen as the research method producing valid results.
The batteries were tested using a Chroma 17011 Test System (7208M-6-30 model). This system is used for testing battery cycle life by repeated charging/discharging according to the specified load profiles, allowing estimation of the battery capacity and internal resistance with an error of up to 0.02% 9 .
SB life tests were carried out according to the requirements for test methods established in regulatory documents 6,7 . During the experiments, the SB degradation rate was determined at various charge/discharge profiles. Three SB load profiles were simulated: one for the standard DC charge/discharge mode, and two for the charge/discharge modes when using a RES.
The maximum values of the charge and discharge current for all profiles were limited to 3.6 A and 2.4 A, respectively, which correspond to a 3-hour discharge current according to the battery passport. The maximum depth of discharge (DOD) was taken equal to 80%. The state of charge (SOC) and gas release values were continuously monitored during the experiments. The tests were carried out at a temperature of 24 °C and an air humidity of 60−70%.
The selected parameters of charge/discharge profiles correspond to severe SB operating modes, leading to the fastest possible degradation of their operating characteristics. Therefore, in all the experiments, we applied a used SB, in which 3 cells of a similar residual capacity were selected according to the results of preliminary tests.
The profile of RES1 was formed using the actual data obtained by monitoring the operating modes of a WPP VDM-2kW wind turbine produced by VDM-tekhnika LLC, installed at a wind farm in Dubna 10 . The WPP output power log data recorded at 1 second intervals were normalized to the required experimental conditions. The change in the SB charge/discharge current, corresponding to the RES1 profile, is shown in __________________________________ 8   .57 h. The profile of RES2 was formed based on the same initial data; however, the discharge and charge modes were separated in time while maintaining the overall power balance.
To ensure comparable experimental conditions during RES profile tests, the following limitations were set. When the voltage reached 2.35 V (full charge), the SB was transferred to the discharge section of the profile. When the voltage dropped to 1.8 V (DOD = 80%), the SB was transferred to the charge section of the profile.

DETERMINATION OF STRESS FACTORS
Since the ambient temperature was kept constant during the SB life tests, the temperature-related stress factors were not taken into account.
The charge factor (CF) is a dimensionless coefficient, its numerical values are determined through the ratio of the energy received and delivered by a SB, expressed in A • h per operation year: where Hthe Heaviside function; I bat -SB current, whose values are used to separate charge/discharge modes according to the given conditions: I bat > 0charge mode (charged), I bat < 0discharge mode (discharged). Depending on the microgrid design and the methods used for controlling the microgrid mode, the CF value can vary within a wide range. Thus, each specific current value will have a corresponding optimal CF value under a certain combination with other stress factors. The CF value can be used to assess the efficiency of the SB voltage regulation system, as well as possible deterioration mechanisms. Very low CF values lead to intensification of the sulfation and acid separation processes; moreover, an imbalance in the battery and cell charges in the SB bank is possible. Higher CF values promote corrosion and the loss of water and active mass [14].
The throughput factor Q thr is expressed through the normalized value of the SB annual discharge capacity: where C N -the rated SB capacitance, A • h. The effect of Q thr factor on the SB service life can only be considered in combination with two other PC and TL stress factors. In general, higher Q thr values contribute to the active mass The discharge rate factor is applied to assess the effect of high discharge currents on the SB lifetime. The function of discharge current distribution is used for calculating DR, where 1% of the capacity from the total SB throughput (Q thr ) is discharged. These distribution parameters are determined by dividing the entire range of discharge currents into M groups from the lowest I 1 to the highest I M current, with the subsequent determination of the corresponding frequency f M to construct a distribution histogram. By combining adjacent current groups in the total amount L, m groups with the I m average current value, the t m time interval and the f m distribution frequency are extracted from the original histogram. These groups satisfy the following condition: The DR value is determined by normalizing the average discharge current of all selected groups to the SB 10-hour rated discharge current I 10 : A practical example of defining DR is discussed in [17]. Higher DR values increase the electrolyte temperature, losses and selfdischarge current of a SB.
Time between full charge. SB average time between full charge (in hours) is determined by the following equation: where n 90% is the number of events per year that satisfy the condition: SOC (t) > 90% and SOC (t -Δt) ≤ 90%. The choice of SOC = 90% as a criterion when calculating this factor is explained by the fact that determination of actual SOC values at high SB charge levels involve significant errors [14].
Higher TF values activate the process of forming insoluble sulfur compounds, which cannot be converted back into active substances. This ultimately leads to an irreversible decrease in the SB capacity relative to its rated value.
The time at low state of charge. The TL factor is the SB service time expressed as a percentage over the calculated period (usually per year) at a SOC value < 35%: Long operation periods at low SOC levels accelerate sulfation and electrolyte decomposition. It should be noted that this stress factor has a significant negative impact on the service life of lead-acid SB, although its effect on other SB types is much smaller [14].
Partial cycling. The PC stress factor characterizes the weighted and averaged value of the battery discharge energy at different SOC levels to the total throughput, expressed as a percentage [17]: where the A -E coefficients represent the SB total discharge capacity in a specific SOC range, expressed as a percentage of the total annual discharge capacity.
The following SOC ranges are used to calculate the coefficient values in equation (7): 100−80% corresponds to coefficient A, 85−70% -to coefficient B, 70−55%to coefficient C, 55−40% -to coefficient D, and 40−0%to coefficient E. The other coefficients are determined in a similar manner. The effect of the PC factor on SB deterioration is manifested similarly to that of the TL stress factor. Fig. 2 shows a fragment of SB life tests registered for different load profiles. During the tests, the values of SB (I) charge/discharge current and cell voltage (V), as well as the calculated value of residual capacity (SOC) were recorded. The measurements were carried out with an interval of 1 s and were recorded in a spreadsheet for further processing. Fig. 2 presents the experimental data for 2 charge/discharge cycles according to the DC profile, as well as 1 operating cycle for each RES profile. Following 5 consecutive cycles according to the given load profiles, the SB residual capacity was measured 7 . The obtained experimental dependences of the change in the SB residual capacity (SOC) on operating cycle numbers (N cycle ) and operating time (t) are presented in fig. 3. The experiments showed that the evaluation of the service life of SBs operated according to RES profiles based on the number of charge/discharge cycles and (or) the calendar operation from the data provided by the technical specification can lead to serious errors. This can be explained by significant differences in the full charge/discharge cycles between the DC and RES profiles, both in terms of duration and throughput. Performance models are more efficient [15,16], allowing the residual battery life to be estimated by the amount of passed energy. Nevertheless, this method of determining the SB service life ignores the influence of other stress factors, thus leading to additional errors.

Fig. 3. Experimental dependences for the SB residual capacity on the number of operating cycles and operating time Рис. 3. Экспериментальные зависимости остаточной емкости аккумуляторных батарей от числа рабочих циклов и времени эксплуатации
Our experiments revealed a significant effect of pulsed charge currents on SB degradation. When the battery is charged by pulsed currents, the voltage increases to the limiting values of 2.35-2.4 V per cell. Further charging will lead to a sharp growth in temperature and boiling off the electrolyte with irreversible consequences in terms of SB degradation [7]. In the experiments carried out on RES profiles, when the voltage increased to 2.35 V, the tested cell was switched to the discharge mode, ensuring its protection from overcharging. Voltage control and charging current limitation allow the SB to be protected from overcharge modes; however, this leads to an incomplete battery charge and increases the operating time between full charge and at a low SOC level (see fig. 2). In addition, due to significant voltage ripples (for the RES1 profile, in particular), it is rather difficult to provide reliable and accurate control of the SOC value, which is essential for the effective operation of the microgrid energy storage mode control system.
The efficient recovery and accurate SOC control (see fig. 2) can be ensured by the standard two-tier DC charging mode based on the current-voltage method (DC profile). However, this requires the SB to be in a low-current charging mode (up to one percent of the rated capacity C N ) for a sufficiently long period of 5−7 h. The practical implementation of this SB charging mode in a RES-based microgrid is complicated by the excess electricity generated at such time intervals. The inability to consume this excess energy will lead to a drop in the RES ICUF and, accordingly, to deteriorated technical and economic characteristics of the entire energy system.
Tab. 2 presents the numerical values of the stress factors acting during SB operation for different load profiles. Fig. 4 shows the intensity of stress factors according by a gradation proposed in [14]: 1very low, 2low, 3medium, 4high, 5very high. A comparison of the SB operating conditions according to the DC and RES profiles reveals significant differences between all stress factors. It can be seen that the throughput factor has a decisive influence on the SB service life. An increase in Q thr values corresponds to a proportional acceleration of battery degradation; nevertheless, the battery life is determined by the cumulative impact of all stress factors. The obtained stress factors for the RES1 and RES2 profiles indicate a better retention of the SB service life when the charge/discharge modes are separated in time. The conducted tests demonstrated that the SB operating conditions according to these profiles are significantly different in terms of time between full charge (TF) and time at low state of charge (TL). These differences lead to a longer service life (by 10-14%) of SB, when operated by the RES2 profile even at higher Q thr values.
The obtained results allowed us to develop recommendations for selecting an appropriate SB capacity and designing an efficient energy storage system for RES-based microgrids: 1. In RES-based microgrids, deep-discharge SBs demonstrating good cyclic characteristics should be applied. Among lead-acid batteries,

QP 
where P max is the maximum electrical load.
3. The impact of high-amplitude pulse charge/discharge currents on the SB should be avoided, depending on the SB type. Thus, for lead-acid SBs, the charge/discharge currents should be limited to I bat = 0.3 • С N .
4. The SB operating modes with the residual capacity below the maximum allowable discharge level should be minimized or avoided. For lead-acid SBs, optimal DOD values range from 30% to 40%.
5. SBs should be operated at a constant temperature using thermal containers equipped with a heating/air conditioning system. 6. In cases where the energy storage circuit consists of series-parallel SB chains, balancing devices should be used. These recommendations minimize the influence of stress factors on the performance of SBs, thus ensuring their maximum possible service life.
On the basis of the results obtained, a novel technical solution for developing an energy storage system for a RES-based microgrid was proposed [26]. Fig. 5 describes the proposed solution, which is based on a combined energy storage unit (CES) consisting of a supercapacitor module (SC) and 2 identical storage units based on SBs, alternately operating in the charge/discharge mode and a ballast load (BL). Here, the energy storage is connected to the microgrid busbars to control the energy balance in the system by regulating the SB charge/discharge currents, supercapacitors and the ballast load. The application of 2 SB circuits makes it possible to implement the effective charging modes according to the "currentvoltage" method and a precise control of the residual capacity. The supercapacitor module mitigates power ripples over short periods of time, and, accordingly, optimizes the magnitude of charge/discharge currents. The proposed solution is expected to increase the SB reliability and service life, as well as the RES ICUF. In addition, the design of an autonomous inverter (AI) can be significantly simplified.
An experimental prototypes of CES with a rated power of 15 kW was manufactured at the VDM-Tekhnika LLC; its declared parameters were experimentally confirmed when operated as part of a wind photo-diesel microgrid ( fig. 6) 10 . The use of CES in a microgrid increases the SB service life by 20−30%, ensures the static and dynamic stability of the local energy system with a response time to power changes of no more than 50 ms, provides a fuel substitution level of at least 25% and reduces the electricity cost by 25−30%.

CONCLUSIONS
The conducted experimental studies allowed us to determine the numerical values of stress factors, which reflect the relationship between the operating conditions and service life of SBs used in microgrids. On the basis of a comparative analysis of the effect of various stress factors on the SB service life, recommendations were formulated for selecting an optimal battery capacity, as well as for developing an efficient energy storage system for RES-based microgrids.
A novel technical solution for the design of an energy storage system was proposed, implemented and experimentally tested. This solution provides for an increase in the energy efficiency of RES-based microgrids.