Intelligent Battery Management in a Hybrid Photovoltaic Using Fuzzy Logic System

Abstract

LiFePO₄ batteries need a battery management system (BMS) to improve performance, extend their lifespan, and maintain safety by utilizing advanced monitoring, control, and optimization techniques. This paper presents the design, development, and implementation of an intelligent battery management system (i-BMS) that integrates the real-time monitoring and control of batteries. The system was extensively tested using multiple datasets, and the results show that the system was able to maintain battery temperature within the set range, balance the cell voltages, and distribute energy according to load prioritization. It uses a fuzzy logic system approach to effectively manage farm energy requirements. Additionally, the proposed method embedded a three-level load prioritization algorithm woven into the fuzzy rule set to allocate energy dynamically among essential, regular, and non-essential loads.

1. Introduction

Driven by the imperative to mitigate climate change, there is an increasingly global shift to renewable energy sources, a movement inspired by the United Nations’ 2030 Sustainable Development Goal to provide clean, green, sustainable, and affordable energy for all [1]. Nevertheless, despite their abundance and cost-effectiveness, renewable resources remain inherently weather-dependent and intermittent [2]. This variability underscores the critical need for efficient energy storage to maintain a consistent power supply [3]. According to Statista 2025, it is expected that by year 2031, batteries in the solar energy systems will have the second largest market share, amounting to 154.31 billion US dollars (refer to Figure 1). Among the various battery technologies available, lithium iron phosphate (LiFePO₄) has emerged as the preferred choice, due to its superior performance and enhanced safety [4]. Today, this rising demand for LiFePO₄ batteries is widely observed both in the automative and renewable energy industries [5].

A LiFePO₄ battery has high power density, longer life span, better thermal stability, and low self-discharge rate, and is environmentally friendly, robust, and reliable, making it suitable for solar power systems. Although it offers numerous advantages, it also has some drawbacks, such as low cell voltage, typically 3.2 V, and a relatively flat discharge voltage [6]. The low cell voltage implies more batteries connected in a series to meet the 12 V, 24 V, 36 V, or more system design. Thus, a battery management system (BMS) is needed.

In any energy storage system, BMS is used for energy optimization, grid stabilization, peak shaving, and safety [7]. Primarily, the functionalities include (1) measurement and control; (2) charge and discharge control; (3) thermal control; (4) communications; (5) diagnostics, data logging, and authentication; and (6) protection [8]. It has also evolved from no management to advanced management [9]:

• Low Management—This management level primarily involves the manual measurement and periodic monitoring of battery parameters, typically suited for single-cell lead–acid batteries. It necessitates regular manual maintenance to ensure proper operation and safety.
• Basic Management—Provides the real-time online monitoring of essential external battery parameters, offering protection against overcharging and over-discharging scenarios. This level is suitable for managing small sets of batteries, ensuring improved reliability and lifespan through basic protective functionalities.
• Enhanced Management—Explicitly designed for onboard battery systems in electric vehicles (EVs), this advanced management level encompasses comprehensive battery modeling, precise state estimation, sophisticated fault diagnosis, and optimized control strategies. It is tailored to handle battery modules, significantly enhancing performance, safety, and overall system efficiency.

The continued development of battery technology has led to intelligent battery management systems. This emerging technology incorporates artificial intelligence, big data, IoT, wireless sensing network (WSN), cloud computing, and predictive models [10]. Hence, this convergence of digital technologies led to the development of this study.

The primary objective of this study was to design and develop an intelligent battery management system for hybrid photovoltaic (PV) technology, using computational intelligence (CI) techniques to optimize energy distribution and utilization. Specifically, the research aimed to

• Develop a comprehensive monitoring system that tracks and analyzes energy generation, distribution, and priority-based load management in real time;
• Design and implement a real-time intelligent battery management framework using advanced sensors integrated with computational intelligence algorithms to enhance energy distribution efficiency and optimize utilization;
• Evaluate the overall performance of the intelligent battery management system (i-BMS) by comparing the energy generated versus energy demand utilization.

This study is presented as follows: Section 2 discusses the related studies and works on battery management systems; Section 3 covers the system overview and fuzzy logic application; Section 4 includes MATLAB R2024b simulation; Section 5 shows the results and discussion; and, lastly, Section 6 presents the conclusion and recommendations.

2. Review of Related Studies and Works

Recent studies show significant development of BMS in different applications. It has dramatically improved and has impacted on applications such as hybrid photovoltaic (PV) systems. The Scopus keyword search (2021–present) for “Battery AND Management AND System” shows a rapidly expanding field centered on topics like state-of-charge (SOC) estimation, state-of-health (SOH) monitoring, and advanced control strategies [11,12,13]. Figure 2 emphasizes battery management systems as a highly interdisciplinary area, linking battery health and thermal management with AI and ML, IoT integration, and cybersecurity concerns [14,15].

2.1. BMS: An Overview

A system that has software and hardware elements designed to ensure that batteries function within the safe operating region [16]. BMS combines monitoring, controlling, and fault diagnosis to optimize battery operation, specifically in applications that require stable, reliable, and efficient energy storage. Numerous studies have been conducted to further improve BMS’s performance from zero to an intelligent, adaptive type of battery management. The published articles from the IEEE Xplore digital library related to intelligent BMS numbered 5792 conferences, 1144 journals, 98 magazines, 41 books, 34 early access articles, and 10 standards.

Figure 3 summarizes the comparison between conventional versus intelligent BMS. Conventional BMS focuses on basic voltage, current, and temperature monitoring, using fixed thresholds to trigger control actions. This traditional approach tends to be more reactive: decisions are made after certain limits are reached rather than anticipating issues before they occur. As a result, conventional BMS may be less efficient and can lead to a reduced overall battery lifespan [17].

In contrast, intelligent BMS incorporates advanced monitoring and control strategies, often powered by AI or machine learning. These systems utilize adaptive thresholds that change in real time based on operating conditions, enabling proactive decision-making. An intelligent BMS can improve battery performance, reduce energy losses, and prolong the lifespan by analyzing patterns and predicting potential failures or performance declines. This more sophisticated approach also enhances safety protocols and communication capabilities, ultimately resulting in a more reliable and efficient energy storage solution [18].

BMS Core Functions

Designing a BMS is difficult since it involves simultaneous functions of several integrated systems to achieve higher battery efficiency. BMS functions include monitoring, estimation, control, protection, fault diagnosis, and communication. Some of these BMS functions need sensors to measure and control battery parameters.

• Monitoring and Sensing—A smart BMS constantly checks the battery’s condition using built-in sensors. When batteries are connected in a series, they lose their balance through repeated overcharging and over-discharging cycles. This scenario might lead to cell voltage imbalance and overheating, eventually shortening battery lifespan in the long run [19]. With this, voltage sensors are used to monitor the voltage level on each cell so it stays within safe limits and balanced [20]. Temperature sensors look for heat build-up to prevent thermal runaway. Current sensors track how much charge flows in and out, accurately showing battery use [21]. These measurements let the BMS protect the pack and extend its life.
• State Estimation—It refers to the informational backbone of BMS, including SOC, SOH, state of energy (SOE), state of power (SOP), state of safety (SOS), state of temperature (SOT), and remaining useful life (RUL) [22]. Accurate battery state estimates are critical for optimal charging control, temperature, and overall battery health.
• Control and Protection—The BMS can detect and keep the batteries from experiencing over-voltage (OV), under-voltage (UV), over-current (OC), short-circuit (SC), over-temperature (OT), under-temperature (UT) status and prevent further damage [23].
• Diagnosis—It is vital for the optimal operation of batteries and is useful for identifying and detecting whenever a fault has occurred in the cell or pack. It can be used for state estimation [24], data monitoring and analysis [25], machine learning, and advanced algorithms [26], fault detection and isolation [27,28].

2.2. Smart Farming Technologies

Smart farming (SF) uses cutting-edge technologies and strategies in agriculture, including (1) data acquisition technologies, (2) data analysis and evaluation technologies, and (3) precision application technologies.

Several studies highlight key technologies used in smart farming and early adopters. They present the rapid adoption and growing importance of PV technology, IoT, and AI in agricultural settings, while also categorizing each application.

Various modern technologies are transforming traditional agricultural practices: IoT sensors for soil and crop monitoring [29,30], drones and UAVs for crop mapping and health assessment [31,32], and cloud computing for remote oversight [33], which provides vital data on farm operations. Geospatial technologies further enable precise field mapping [34], allowing farmers to tailor interventions more accurately.

From the data analysis side, machine learning enhances yield prediction and disease detection [35,36], while big data analytics improve resource management [37]. Robotics and automation streamline tasks like planting and harvesting [38], and blockchain guarantees transparency and traceability throughout the supply chain [39,40]. Finally, renewable energy sources offer cost-effective and eco-friendly power solutions [35,36], making agriculture more sustainable and efficient.

2.3. Hybrid Photovoltaic Systems

The term “photovoltaic” is derived from the Latin words “photo”, meaning “light”, and “voltaic”, meaning “energy”, which together form the word “photovoltaic” [41]. When a sufficient amount of light energy strikes the solar panel, it knocks out free electrons residing at the outermost valence shell. The panel then goes into a conduction state, producing a current.

Photovoltaic (PV) technology harnesses renewable energy by arranging panels, charge controllers, batteries, and inverters. It can be implemented as an off-grid, grid-connected, or hybrid system. However, weather variations and battery degradation can compromise the reliability of a continuous power supply. Factors such as solar radiation, shading, and dust accumulation notably impact the overall system performance [42].

Figure 4 shows a hybrid PV system setup. When sunlight energy strikes the solar panel, it converts this light into electrical energy. The solar charge controller protects the battery from an overcharge state by controlling the flow of current [43]. Batteries serve as energy storage, while the inverter converts the DC signal into a usable AC signal compatible with the AC loads.

2.4. Synthesis, Research Gap, and Novel Contributions

Although fuzzy logic battery management systems (BMSs) have been studied extensively, no previous work has combined dual critical capabilities in a single, real-time hardware implementation:

• Millisecond scale fuzzy i-BMS control running directly on embedded hardware;
• Three-tier load prioritization (essential/regular/non-essential) integrated into the rule base.

To close this gap, the present study introduces a unified framework with two key advances:

• Integrated i-BMS architecture that couples photovoltaic input, battery bank, and utility grid while sampling voltage, current, SOC, and cell temperature at a millisecond resolution;
• Embedded three-level load prioritization algorithm woven into the fuzzy rule set to allocate energy dynamically among essential, regular, and non-essential loads.

3. Materials and Method

3.1. Dataset

The dataset comprises high-resolution operational records from the i-BMS battery voltage and temperature monitoring platform installed on a hybrid PV smart farm testbed. Continuous measurements were logged once per minute over 11 consecutive days—from 19 September 2024 to 29 September 2024—yielding approximately 4700 time-stamped entries.

At each time stamp, the following variables were captured:

• Battery Pack Health

1. Individual cell voltages (V1–V16) of the 16 series connected LiFePO₄ cells, expressed in volts (V);

2. Individual cell surface temperatures (T1–T16) of the same cells, expressed in degrees Celsius (°C).

• Load Category Power Metrics

Essential, regular, and non-essential loads consist of an aggregate line current (Iₑ), line voltage (Vₑ), and instantaneous active power (Pₑ).

In summary, each data point stores 16 cell voltages, 16 cell temperatures, and 9 load parameters, a total of 41 features per minute. The resulting time series dataset supports detailed analyses such as cell-level balancing, thermal management, and prioritized demand-side energy optimization across all load types.

3.2. System Calibration

Calibration of the voltage sensor follows this regression equation:

V_calibrated = 0.3402V_{sensor_raw} + 1.7876, (volts)

(1)

The model summary has a standard error of the regression (S) of 0.0480807, an R-squared (R²) of 86.32%, an R-squared adjusted of 85.99, and an R-squared predicted of 83.17%. This S value is within the accuracy resolution of the voltage sensor used in this study. The high R² value means an excellent result for sensor calibration, where some noise is unavoidable. The predicted R-squared confirms that the calibration should work on readings, not in the training set.

Calibration of the temperature sensor follows this regression equation:

T_calibrated = 1.008T_{sensor_raw} − 0.2405, (Celsius)

(2)

With a regression standard error of 0.361 °C and an R² of 99.69% (adjusted R² = 99.67%), the calibration model captures virtually all the variation in the reference readings. In practical terms, each calibrated temperature measurement deviates from the reference thermometer by only about ±0.36 °C on average, confirming that the sensors are highly accurate across the tested range.

3.3. Site Climate and Environmental Conditions

The battery bank was housed in a roofed, well-shaded battery shed in Morong, Rizal, Philippines (14.5181° N, 121.2390° E). The ambient conditions during the test phase (19–29 September 2024) data were extracted from gridded surface meteorology taken from NASA POWER v 2.4.14. Averaged over 11 days, the site experienced the following:

• A mean daily maximum air temperature (T2M_MAX) of 29.6 °C;
• A mean daily minimum temperature (T2M_MIN) of 24.7 °C;
• An average 2 m relative humidity of 88.5%.

These figures indicate an optimal operating temperature and a highly humid environment that remained well below the LiFePO₄ manufacturer’s critical thermal threshold, thereby ensuring that the observed battery behavior primarily reflects the BMS control logic rather than external heat stress [44].

3.4. System Architecture Overview of i-BMS

A conventional hybrid-PV installation typically includes PV modules, a solar charge controller, a battery bank, and an inverter. The proposed system upgrades this architecture by embedding the i-BMS driven by a fuzzy logic controller that supervises and optimizes the energy flow between the batteries, the utility grid, and the loads. This smart farm integrates several subsystems, including irrigation, lighting, computer vision, CCTV surveillance, cooling systems, aquaponics, and energy management. These units draw their primary power from a PV array with battery storage, while the utility grid automatically covers any remaining demand. The authors developed the i-BMS, which was deployed in a smart farm, in this study.

Figure 5 depicts the complete i-BMS architecture for the hybrid PV system, comprising four key control units—the solar management control unit (SMCU), battery management control unit (BMCU), load management control unit (LMCU), and the fuzzy logic engine—as well as load prioritization logic and a web-based monitoring dashboard. The smart farm’s primary power source is the SMCU, which comprises a PV array, charge controller, and inverter. The BMCU, consisting of the battery bank, voltage, current, and temperature sensors, ensures that cell operation remains within safe limits. The LMCU allocates energy to loads based on their assigned priority levels. The fuzzy logic controller is at the heart of the system, processing inputs from all units to regulate charge, discharge, and load switching commands in real time. Finally, every metric, including cell temperatures and voltages, pack currents, and load usage by load category, is streamed live to a web interface for remote visualization and analysis.

To make real-time decisions, the i-BMS continuously acquires the variables listed in

Table 1. Data acquired by the fuzzy logic engine.

i-BMS Parameter	Purpose
Cell temperature	Detect thermal stress and prevent thermal runaway
Cell voltage	Track cell balance and health status
Pack current	Compute charge/discharge rates
SOC	Estimate remaining capacity
Instantaneous load demand	Match supply with priority loads

.

Based on those inputs, the fuzzy logic controller issues four high-level commands, which are as follows:

• PV coupling—Connect or isolate the PV array from the charge controller;
• Battery dispatch—Engage or disconnect batteries for charging or discharging;
• Load prioritization—Energize or defer loads according to their priority class;
• Grid interchange—Transfer the bus to or from the utility grid via an automatic transfer switch (ATS).

The i-BMS subsystem consists of the central processing unit (CPU), BMCU, SMCU, LMCU, and ATSCU.

Table 2. i-BMS subsystem architecture.

i-BMS Subsystem Module	Function
CPU	Aggregates sensor data, executes fuzzy rules, and broadcasts control signals to all subordinate units.
BMCU	Measures cell temperature, voltage, current, and SOC; manages balancing circuits and charge relays.
SMCU	Reads PV power output and governs PV to controller connection through a dedicated relay.
LMCU	Track consumption for three loads—essential, regular, and non-essential—and switch each via separate relays.
ATSCU	Commands the ATS to toggle the DC bus between the PV system and the utility grid.

summarizes the function of each subsystem.

Together, these coordinated modules allow for the i-BMS to allocate solar energy efficiently, maintain battery health, guarantee uninterrupted power for essential loads, and call on the grid only when necessary, all without manual intervention.

3.5. Application of Fuzzy Logic in i-BMS

The i-BMS primarily aims to track the battery’s condition and allocate power to the smart farm based on predetermined priority levels. The system automatically transitions to utility grid power if the battery lacks sufficient energy to maintain continuous farm load operations. Figure 6 illustrates the proposed i-BMS that utilizes fuzzy logic to optimize battery operations. The system processes multiple input parameters through an intelligent decision-making model to regulate charging and discharging.

Input parameters include SOC expressed as a percentage, cell voltage (V_cell), and cell temperature (T_cell). Individual battery voltage should be balanced regularly to extend the battery lifespan [45], whereas cell temperature monitoring is essential to avoid thermal runaway [46,47].

The i-BMS controls battery functions based on its fuzzy logic-based decisions, such as charging, discharging, and utility grid connection. Charging will be initiated when SOC is low, and temperature conditions are safe. Discharging will be initiated when load demand is high and SOC is sufficient to supply power to smart farm loads. Lastly, the system can intelligently decide whether to store or distribute energy based on demand and available resources.

Table 3. Input linguistic variables of i-BMS.

No.	Input Linguistic Variable	Conditions			Unit
1	SOC	Low 0–30	Moderate 20–80	High 70–100	%
2	Vcell	Low 2.0–2.5	Moderate 2.3–3.5	High 3.2–3.6	V
3	Tcell	Low 0–20	Moderate 15–45	High 40–60	°C

shows the input linguistic variables and their respective conditions for an i-BMS using fuzzy logic. The SOC is classified into low (0–30%), moderate (20–80%), and high (70–100%). V_cell is one of the indicators of battery health status. This system is subcategorized into low (0–2.5 V), moderate (2.3–3.5 V), and high (3.2–3.6 V). T_cell is categorized as low (0–20 °C), moderate (15–45 °C), and high (40–60 °C). This variable helps the i-BMS regulate energy distribution efficiently based on consumption patterns.

Figure 7 displays the plots of the three fuzzy input variables—SOC, Vcell, and Tcell —categorized as low, moderate, and high using trapezoidal and triangular membership functions.

SOC is divided into three conditions:

• Low, where batteries are almost depleted;
• Moderate, where the battery is partially charged;
• High, indicating a battery close to full capacity.

The V_cell has also three linguistic levels:

• Low, which corresponds to an undervoltage region where the battery risks deep discharge;
• Moderate refers to the typical operational range; and
• High is the near or at maximum charge voltage, signifying a fully charged cell.

T_cell is subcategorized as follows:

• Low is the level where battery performance may be reduced but not dangerously high;
• Moderate pertains to the optimal band for regular operation;
• High temperatures are approaching the upper safety threshold, potentially risking battery degradation.

Table 4. Output linguistic variables of i-BMS.

Parameter	Description
	1	2	3
Battery Status	C	ID	D
Load Prioritization	E	E + R	All

presents the output linguistic values used in the i-BMS, which utilize fuzzy logic to determine appropriate charging or discharging actions based on SOC, V_cell, and T_cell, presented as follows:

Figure 8 depicts the output-side membership function diagram used by the fuzzy controller to classify the i-BMS charging regime. Three triangular membership functions define the linguistic states: charge (C), idle (ID), and discharge (D). When the cell temperature rises above its safe limit, the controller forces the system into the idle state, disconnects the battery, and draws power from the utility grid to keep the farm running. Once the temperature falls back to the low or moderate bands, the battery is returned online, and the controller selects either charging or discharging according to the real-time SOC, V_cell, and T_cell.

Load allocation decisions are made simultaneously through three priority levels—essential (E), essential + regular (E + R), and all loads—ensuring that limited energy is routed first to critical services. At every control step, the fuzzy inference engine evaluates the three inputs, assigns graded degrees of membership to each of the three charging states, and defuzzifies the result into a single actionable command for the battery system.

Table 5. Fuzzy logic rules of i-BMS.

Rule	Input Linguistic Variable IF			Output Status THEN
	SOC	Vcell	Tcell	Battery Status	Load Prioritization
1	Low	Low	Low	C	E
2	Low	Low	Mod	C	E
3	Low	Low	High	ID	E
4	Low	Mod	Low	C	E
5	Low	Mod	Mod	C	E
6	Low	Mod	High	ID	E
7	Low	High	Low	C	E
8	Low	High	Mod	C	E
9	Low	High	High	ID	E
10	High	Low	Low	D	All
11	High	Low	Mod	D	All
12	High	Low	High	ID	E
13	High	Mod	Low	D	All
14	High	Mod	Mod	D	All
15	High	Mod	High	ID	E
16	High	High	Low	D	All
17	High	High	Mod	D	All
18	High	High	High	ID	E
19	Mod	Low	Low	C	E + R
20	Mod	Low	Mod	C	E
21	Mod	Low	High	ID	E
22	Mod	Mod	Low	D	All
23	Mod	Mod	Mod	D	All
24	Mod	Mod	High	ID	E
25	Mod	High	Low	D	All
26	Mod	High	Mod	D	All
27	Mod	High	High	ID	E

presents a fuzzy rule-based system for determining the i-BMS’s battery status. The decision is based on three input conditions: SOC, V_cell, and T_cell. These inputs are processed using IF-THEN rules to determine the appropriate battery status and load prioritization. For instance, if SOC,V_cell, and T_cell are low, the system will only direct the battery to the charging state and supply energy to essential loads.

3.6. The i-BMS Electrical Circuit Design

In Figure 9, a 220 VAC-to-5 VDC switch mode supply powers the ESP32 via its Vin pin and feeds two 16-channel analog multiplexers that sit on the same 5VDC supply and ground. Four shared GPIO lines (D14, D26, D25, D32) provide the S3–S0 address bits, while GPIO 17 independently enables or disables each multiplexer so that only one connects to the signal bus at a time. The active multiplexer routes the selected cell tap (C0–C15) to the ESP32′s ADC input on GPIO 34, allowing a single ADC channel to scan up to 16 cell voltages in sequence. Detailed pin assignments for the voltage sensor, including relay connections, voltage and temperature sensor connections, and the current sensor interface, are shown in Figure 10, Figure 11, Figure 12, Figure 13, Figure 14 and Figure 15.

3.7. On-Site Deployment of the Hybrid PV Systems with the i-BMS

Figure 16 illustrates the overall system setup. The experiment platform operates on a hybrid PV system, where the utility grid provides backup energy. The system consists of several key components:

• Solar charge controller (top right)—Regulates the voltage and current coming from the solar panels to prevent overcharging [48];
• Automatic transfer switch (middle section)—Controls the transfer of energy sources between solar and utility grid [49];
• Inverter (center, blue box)—Converts the stored DC power from the batteries into ac [50];
• Battery bank (Refer to Figure 17)—Stores the energy collected from the solar panels for later use, ensuring power availability even when sunlight is not present [51].

Figure 17 presents the on-site deployment of the i-BMS, showcasing the LMCU, the cell balancing module, the i-BMS control board, and the associated battery banks.

Sixteen pieces of LiFePO₄ batteries were deployed on-site in an eight–series–two-parallel (8S-2P) configuration (refer to Figure 18a), with a 24-V, 404 Ah capacity system design. Sixteen voltage and temperature sensors were used to measure and monitor individual cells. Two current sensors were deployed at the load input and output terminals. The cell balancing module monitors and regulates the voltage of each cell to ensure the batteries remain balanced [52]. This i-BMS uses an active balancing circuit in Figure 18b.

Table 6. LiFePO₄ battery specification.

Parameter	Typical Value
Nominal Capacity	202 Ah
Nominal Voltage	3.2 V
Internal Resistance	≤0.4 mΩ
Max. Charging Current	0.5 C (Continuous), 1 C (30 s)
Charging Voltage	3.65 V
Max. Discharging Current	0.5 C (Continuous), 1 C (30 s)
Discharging Cut-off Voltage	2.5 V (>0 °C), 2.0 V (≤0 °C)
Operating temperature	Charging: −20 °C~55 °C
	Discharging: −30 °C~55 °C
Dimension	54 × 173 × 205 mm
Cycle life	≥2000 cycles

presents the summary of the battery specification used in this study.

The LMCU (refer to Figure 18c) prioritizes whether the farm’s load demands are essential, regular, or non-essential. The system responds based on the priority level. It also validates the state of the batteries before supplying the energy demand.

The load prioritization categories are as follows:

• Essential loads—These loads are mission-critical and require uninterrupted power supply due to their operational requirements. They include life support systems for aquaculture tanks and security systems such as CCTV. They are last to shed, first to restore.
• Regular loads—These are operationally necessary loads, and short outages are tolerable, leaving no severe damage or impact on system operations. Loads include irrigation pumps, data logging servers, and IoT gateways.
• Non-essential loads—These loads are usually not vital to smart farm operations and thus pertain to convenience or discretionary loads. They may operate only when surplus energy exists. This may include electric vehicles, decorative lighting, or household loads.

Furthermore,

Table 7. Sensors, power electronics, and control modules utilized in the i-BMS.

Device	Model	Parameter	Typical Value
Voltage sensor	Voltage Detection Sensor Module 25 V Sensor	Input Voltage range	DC 0 to 25 V
		Voltage detection range	DC 0.02445 V to 25 V
		Analog Voltage resolution	0.00489 V (5 V/1023)
Temperature sensor	DS18B20	Supply Voltage
		Min	+3.0
		Max	+5.5
		Thermometer Error
		−10 °C to +85 °C
		Max	±0.5
PZEM-004T AC communication module	PZEM-004T	Voltage
		Range	80~260 V
		Resolution	0.1 V
		Accuracy	0.5%
		Current
		Range	0~100 A
		Active Power
		Range	0~23 kW
		Starting Measure	0.4 W
		Resolution	0.1 W
		Accuracy	0.5%
		Power factor
		Range	0~1.00
		Resolution	0.1
		Accuracy	1%

summarizes the sensors and control modules utilized in this study, along with their operating range, resolution, and accuracy.

3.8. Real-Time i-BMS Monitoring System

Figure 19 shows the mobile dashboard of the t-BMS, which is used for monitoring the battery status. In contrast, Figure 20, when viewed on a laptop or desktop, displays a navigation panel on the left side that lists various monitoring and management functions, including battery monitoring, load monitoring, and others. Within the main window, there are two bar charts: (1) battery voltage monitoring, which displays the 16 individual voltages in bars, and (2) battery temperature monitoring, a bar chart that shows temperature readings for the same set of batteries.

This dashboard provides a comprehensive view of key battery parameters across multiple cells, enabling a quick assessment of the health and status of each battery in the system.

3.9. i-BMS Data Transmission

The i-BMS consists of three Wi-Fi nodes: an ESP32 for multiplexed cell voltage and DS18B20 temperature measurements, an ESP8266 with an ACS712 Hall sensor for pack current readings, and a second ESP8266 equipped with dual PZEM-004T meters, with a relay for AC-side monitoring and load control. Each board communicates directly with an SQL database over plain HTTP. Sensor data are sent as HTTP GET requests to designated endpoints, and control commands are retrieved by polling another URL. When a load is toggled in the web app, the change is written to the SQL table. On the next polling cycle, the ESP8266 reads the updated flag and flips its GPIO relay accordingly. This direct HTTP → SQL → HTTP loop—is implemented entirely on the free hosting account and delivers real-time telemetry and remote load control.

4. Fuzzy Logic-Based i-BMS Learning Structure

MATLAB was the primary platform for designing, simulating, and validating i-BMS configurations. A classical Mamdani-type fuzzy inference system (FIS) was implemented, following the standard workflow of fuzzification → rule evaluation → aggregation → defuzzification. As illustrated in Figure 21 (Property Editor snapshot), the controller applies the minimum (min) operator for both the AND condition and rule implication, the maximum (max) operator for the OR condition and rule aggregation, and the centroid technique for defuzzification. The FIS accepts three input variables and produces two distinct outputs, yielding a complete rule base of 27 (3³) rules that comprehensively cover all possible operating states of the battery system.

Figure 22 illustrates that the i-BMS controller with SOC = 83.8%, V_cell = 2.8 V, and T_cell = 36.7 °C produces fuzzy outputs of battery status = 2.58 and load prioritization = 2.58. On the scale defined for this Mamdani FIS, a value near three corresponds to the Discharge–-High state and the All-Loads category. Hence, the i-BMS interprets these inputs as follows:

• The battery is healthy within its safe operating limits;
• Ample stored energy is available to power the entire farm—essential, regular, and non-essential loads—without drawing from the grid.

Thus, the controller authorizes full discharge to satisfy total demand while continuously monitoring temperature and cell voltage to prevent unsafe conditions.

Figure 23 illustrates the control surface maps produced by the Mamdani Type-1 i-BMS. For each input pair plotted on the X- and Y-axes (SOC × T_cell, SOC × V_cell, and T_cell × V_cell), the Z-axis shows the following:

• Battery operating mode—charge, idle, or discharge—in Figure 23a,c,e;
• Load prioritization level—essential only, essential + regular, or all loads—in Figure 23b,d,f.

These surfaces visualize how the controller simultaneously decides the pack’s dispatch state and the farm’s load mix across every combination of state-of-charge, cell voltage, and temperature.

In Figure 23a, the blue–turquoise trough at z = 0–1 corresponds to a low-SOC region (≤25%). Here, the fuzzy controller explicitly issues a charge command—overriding cell temperature considerations—to restore the battery’s reserve. Figure 23b shows a similar trough: while the pack is being replenished, the system restricts power delivery to essential loads only, shedding all non-critical demand until the SOC rises above the recovery threshold. A moderate SOC, around 25–60%, and safe T_cell (<≈45 °C) sloping face, colored blue, where z rises toward 2, is when the pack is allowed to discharge at a moderate rate; the higher the SOC, the stronger the discharge command. Also, the same corresponding face is where the system gradually enables regular loads in addition to the essential ones. A high SOC of greater than 60% and safe T_cell, the broad yellow plateau approximately z = 2.3–2.4, occurs when the controller authorizes full-rate discharge. A similar graph (Figure 23b) also shows a yellow plateau, enabling all three load categories. Lastly, when the T_cell surpasses 45 °C, both control surfaces flatten into a distinctive teal shelf at z = 1.5. This plateau activates the FIS’s temperature override safeguard, idle state, and essential loads only. No further energy exchange resumes until the temperature falls back into the safe operating range.

Figure 23c–f confirms that the controller behaves consistently across the remaining input planes:

• Battery status surfaces (Figure 23c: V_cell × SOC, Figure 23e: V_cell × T_cell) progress smoothly from charge through idle to discharge as the operating point moves from low-energy/low-voltage zones to high-energy thermally safe regions, provided the point lies inside the permissible envelope;
• Load prioritization companions (Figure 23d,f) mirror this progression, sequentially enabling essential, then regular, and finally non-essential loads in lock-step with the battery’s dispatch state.

Taken together, these maps demonstrate that i-BMS applies a uniform rule set to every combination of inputs, ensuring predictable, coordinated decisions throughout the safe operating area.

5. Results and Discussion

i-BMS and traditional BMS were implemented in a real physical system. Its operation is remotely monitored in real time via a website.

Figure 24a plots the voltage of each cell at successive time stamps (colored markers) with no balancing circuitry in place. As the test progressed, the pack visibly drifted: the highest charged cells climbed to about 3.13–3.55 V, whereas the weakest sagged to roughly 2.68–2.90 V. These disparities persisted, showing that passive equalization alone could not bring the cells to a standard level. In sharp contrast, Figure 24b captures the same pack after enabling the i-BMS balancing algorithm. The out-of-tolerance cells were pulled back into line, and all voltages rapidly converged within a tight window, demonstrating effective active balancing across the stack.

As summarized in

Table 8. Descriptive statistics of individual cell voltages for a 16-cell LiFePO₄ battery pack.

Battery No.	N	Mean	SE Mean	Std Dev	Min	Q1	Median	Q3	Max
1	4706	3.20778	0.001195	0.081993	2.75	3.15	3.20	3.28	3.79
2	4705	3.23062	0.001379	0.094606	2.72	3.16	3.23	3.32	3.78
3	4722	3.17050	0.001915	0.131613	2.69	3.09	3.16	3.29	3.75
4	4718	3.22873	0.001132	0.077756	2.78	3.16	3.22	3.30	3.90
5	4714	3.12503	0.001788	0.122780	2.60	3.02	3.10	3.24	3.60
6	4714	3.16185	0.001266	0.086973	2.73	3.10	3.16	3.24	3.53
7	4715	3.15788	0.001342	0.092181	2.43	3.08	3.15	3.24	3.70
8	4711	3.13437	0.001640	0.112619	2.55	3.04	3.15	3.24	3.52
9	4684	3.19580	0.001195	0.164421	2.02	3.16	3.23	3.30	3.80
10	4715	3.24445	0.001379	0.077446	2.80	3.18	3.24	3.32	3.62
11	4706	3.17577	0.001915	0.066561	2.72	3.13	3.16	3.24	3.46
12	4715	3.05919	0.001132	0.175828	2.54	2.91	3.05	3.23	3.57
13	2748	3.36663	0.001788	0.195578	2.89	3.25	3.30	3.45	3.99
14	4724	3.01741	0.001266	0.222009	2.41	2.79	2.99	3.24	3.66
15	4715	3.24098	0.001342	0.061240	2.93	3.20	3.24	3.29	3.55
16	4708	3.23071	0.001640	0.075105	2.78	3.17	3.23	3.29	3.61

, the i-BMS uses the voltage readings from the 16 voltage sensors to show the minimum voltage, maximum voltage, and standard deviation for each battery. This experiment used around 4700 datasets.

The following are the findings about the performance of the cell balancing circuit:

• There is a sufficient dataset for the experiment.
• The mean for all the batteries is within the safe region. However, the minimum cell voltage for batteries numbered 7, 9, and 14 is below the lower cut-off voltage.
• The maximum voltage for batteries numbered 1–4, 7,9, and 13–14 goes beyond the upper threshold voltage.
• The pack is not well balanced: the difference between the highest and lowest mean cell voltages is ~0.35 V.
• Several cells have experienced unsafe voltage extremes (>3.65 V or <2.5 V), which can shorten life or pose safety risks.
• Cells 9, 12, 13, and 14 stand out for their high variability; Cell 14 is the weakest, and Cell 13 is the most overcharged.
• To restore uniformity and pack longevity, a balancing routine, stricter charge/discharge cut-offs, or replacing the outlier cells is advisable.

These findings are further confirmed by the boxplot shown in Figure 25, where there are several outliers(*) for almost all the batteries.

Table 9. Descriptive statistics of individual cell temperature for a 16-cell LiFePO₄ battery pack.

Battery No.	N	Mean	SE Mean	Std Dev	Min	Q1	Median	Q3	Max
1	4706	26.9717	0.021036	1.44312	18.94	25.81	26.75	28.12	29.87
2	4705	27.2164	0.027771	1.90493	24.31	25.56	26.87	28.62	32.13
3	4722	27.3226	0.020485	1.40766	24.94	26.19	27.13	28.44	30.31
4	4718	26.9871	0.020545	1.41121	24.63	25.81	26.81	28.12	29.94
5	4714	26.7711	0.020895	1.43462	24.37	25.56	26.56	27.94	30.00
6	4714	27.0127	0.019921	1.36778	24.69	25.88	26.81	28.06	29.75
7	4715	27.1802	0.021284	1.46149	24.75	25.94	26.94	28.38	30.37
8	4711	27.2619	0.021034	1.44371	24.87	26.06	27.00	28.44	30.31
9	4684	27.0942	0.023246	1.59095	24.69	25.75	26.75	28.3625	31.00
10	4715	27.1392	0.023740	1.63019	24.69	25.75	26.81	28.44	31.00
11	4706	27.2880	0.020209	1.38641	25.00	26.12	27.06	28.44	30.31
12	4715	27.2454	0.018574	1.27540	25.06	26.19	27.06	28.25	29.69
13	2748	26.8862	0.028988	1.51961	24.63	25.56	26.69	28.12	30.94
14	4724	27.1918	0.024216	1.66443	24.63	25.75	26.81	28.50	31.31
15	4715	27.1583	0.019245	1.32149	24.94	26.06	27.00	28.25	29.81
16	4708	26.9070	0.019101	1.31065	24.69	25.81	26.75	27.94	29.56

shows that the cells remained within the safe thermal band: the overall mean operating temperature was 27.10 °C, with a minimum of 18.94 °C and a maximum of 31.31 °C, which is inside the −20 to 60 °C range recommended for LiFePO₄ batteries. The boxplot in Figure 26 reinforces this finding, with its tight interquartile range and absence of outliers across all cells confirming consistently stable temperature behavior during the entire test.

Table 10. Fuzzy logic rules of intelligent battery management system.

SM-2024|Load Monitoring
ID	Voltage (V)	Current (A)	Power (W)	Time	Load Category
23680	239.00	0.27	53.9	15:01	Regular
23682	238.70	0.27	53.8	15:01	Regular
23685	238.20	0.27	53.6	15:03	Regular
23688	238.10	0.27	53.6	15:05	Regular
23689	238.40	0.26	37.4	15:06	Non-Essential
23690	237.90	0.27	39.4	15:08	Non-Essential
23691	237.10	0.11	15.6	15:08	Essential
23692	237.50	0.27	53.3	15:08	Regular
23693	237.80	0.27	53.5	15:12	Essential
23694	237.80	0.22	32.4	15:12	Non-Essential
23695	237.60	0.08	10.4	15:12	Regular
23696	237.30	0.27	53.8	15:14	Essential
23697	237.60	0.27	53.9	15:16	Essential
23698	238.20	0.21	29.3	15:19	Non-Essential
23700	237.60	0.27	53.8	15:19	Essential
23701	237.00	0.27	53.5	15:20	Essential
23702	238.70	0.18	25.2	15:20	Non-Essential
23704	237.40	0.17	24.1	15:22	Non-Essential
23706	237.50	0.27	53.8	15:23	Essential
23707	238.30	0.19	27.2	15:23	Non-Essential
23709	237.40	0.27	53.8	15:30	Essential
23710	237.90	0.27	53.9	15:33	Essential
23714	237.60	0.27	53.9	15:35	Essential
23719	237.70	0.27	53.5	15:43	Essential
23723	237.20	0.27	53.8	15:47	Essential
23727	237.00	0.27	53.5	15:53	Essential
23677	237.90	0.27	53.4	15:58	Regular

shows power is measured at around 53 to 54 watts for the regular load category. The essential and non-essential load categories draw the same amount of power at a particular time. The farm’s energy supply requirement was delivered based on its priority level. The three load categories show consistency and steady supply.

Based on Figure 27, essential loads remained relatively stable, indicating consistent demand. Regular loads showed a slight dip but stay close to the 53–54 W range over the monitored interval. Lastly, non-essential loads trended slightly downward in power usage, yet remained near the same overall level.

The percentage chart reveals that essential loads comprised the most significant portion of total consumption, followed by regular loads at roughly a quarter and non-essential loads at 18%. Overall, all three load categories displayed minimal fluctuation in power draw, with essential loads dominating the total share.

Sensors 1 to 8 were dipped into cold water with a food thermometer attached, as shown in Figure 28, to verify that the temperature sensors were working correctly.

Figure 29 compares the reference food thermometer reading with the i-BMS sensor data. At around 14:36, Temperature Sensors 1–8 were deliberately immersed in cold water, producing a sharp, synchronous drop that the i-BMS captured in real time. Temperature Sensors 9–16, which remained adhered to the battery surface, recorded virtually unchanged values over the same interval. The close correspondence between the reference instrument and Sensors 1–8 during the transient and the stability of Sensors 9–16 demonstrates the i-BMS module’s accuracy and consistency throughout the test.

In this research, the developed prototype accomplished the following:

• The system successfully balanced the voltage across the batteries, ensuring uniform energy distribution and reducing the risk of overcharge or undercharge conditions.
• This monitoring system remotely tracked battery voltage, temperature, and power distribution. By prioritizing power distribution based on predefined levels, the system maintained optimal performance under varying operational loads.
• This intelligent control system is effective and reliable in decision-making, handling uncertainties in the sensor data, and adapting to fluctuating load operating conditions without compromising performance.

Developing this i-BMS presented several challenges, which are as follows:

• Voltage sensors experienced frequent malfunctions, leading to replacement.
• Loose circuit connections resulted in repetitive system failures.
• The automatic transfer switch malfunctioned two times, resulting in replacement.
• Battery terminals began to show signs of rust.
• Batteries also experienced imbalances, most of the time.
• Several batteries exhibited bloating.
• The prepaid Wi-Fi limits the number of devices that can be connected simultaneously, hindering the whole system operation at the farm.

6. Conclusions and Recommendations

In this study, we successfully deployed the i-BMS with 16 temperature sensors, 16 voltage sensors, and two current sensors. The i-BMS balanced the voltage across the batteries to ensure uniform energy distribution, monitor battery parameters, and execute power distribution based on priority levels.

The i-BMS efficiently and effectively delivers the required energy demand of the smart farm through a fuzzy logic system for its management and control. The website lets users view the farm’s battery status, available energy, and load demands in real time.

Even though the prototype of the i-BMS exhibited significant improvements over traditional battery management approaches, there are still several areas where the system can be enhanced to optimize its performance, efficiency, and reliability fully. Future researchers may consider incorporating advanced machine learning models to improve real-time analysis to predict battery degradation. Another area of study that can be explored is developing a more robust fault detection algorithm that monitors battery health and anticipates potential safety issues.