My research journey into battery management systems began with a fundamental observation about the growing demands placed on modern energy storage solutions. As the world shifts toward electrified transportation, the role of high-performance and safe battery systems has never been more critical. The electric vehicle battery represents the heart of this technological revolution, yet its full potential remains untapped due to challenges in safety, reliability, and lifespan. This prompted me to dedicate my research to developing a comprehensive energy management system tailored specifically for power lithium-ion battery packs, with the goal of unlocking their capabilities while ensuring dependable operation.
1. Introduction and Research Motivation
In the context of a rapidly evolving energy landscape, the electric vehicle battery has emerged as a pivotal component in the global transition toward sustainable transportation. Throughout my work, I have consistently focused on the fundamental challenges that limit the practical deployment of lithium-ion technology in demanding applications. The primary obstacles include safety concerns, cycle life degradation, and the substantial costs associated with battery pack replacement. These issues directly impact the economic viability and user acceptance of electric vehicles. I recognized early in my research that the performance of a battery pack is not solely determined by its electrochemical properties but is significantly influenced by how effectively it is managed during operation. Without an intelligent management system, even the most advanced cells cannot deliver their optimal performance over an extended period.
I undertook this research with the objective of designing and implementing an energy management system that could actively monitor, protect, and balance lithium-ion battery packs. The system I envisioned would provide real-time information about the state of charge (SOC), protect against hazardous conditions, and implement strategies to maximize the operational lifespan of the pack. This focus aligns with the broader industry goal of making electric vehicle battery technology reliable enough for mass adoption. The significance of this work lies not only in the specific circuit designs and software algorithms I developed but also in the comprehensive approach to addressing the interconnected challenges of measurement, estimation, and protection.
1.1 Current State of Power Battery Technology
My literature review revealed a dynamic landscape in power battery development. Several technologies compete for dominance in the electric vehicle battery market, each with distinct advantages and limitations. The following table, which I constructed based on my research, summarizes the key characteristics of the primary battery types available for traction applications:
| Battery Type | Energy Density (Wh/kg) | Cycle Life | Safety | Environmental Impact | Cost |
|---|---|---|---|---|---|
| Lead-Acid | 30-50 | 300-500 | Good | High Pollution | Low |
| Ni-MH | 60-120 | 500-1000 | Good | Low Pollution | Medium |
| Li-Ion | 150-250 | 1000+ | Requires Protection | Clean | High |
| Fuel Cell | 350+ | N/A | Hydrogen Sensitive | Clean | Very High |
The lead-acid battery, although mature and cost-effective, suffers from limited energy density and significant environmental concerns due to lead content. Nickel-metal hydride batteries, which have been widely used in hybrid applications, offer better performance but still fall short of the energy requirements for long-range pure electric vehicles. The electric vehicle battery market has increasingly turned toward lithium-ion technology as the preferred solution. This shift is driven by the superior electrochemical characteristics of lithium, the lightest metal. Fuel cells represent a promising long-term option, yet the infrastructure and hydrogen delivery challenges remain substantial barriers to immediate adoption.
Through a careful analysis of battery characteristics that I conducted in my preliminary research, I arrived at a clear conclusion regarding the superior attributes of lithium-ion batteries. Their high operating voltage, exceptional energy density, and prolonged cycle life make them the most suitable choice for contemporary electric vehicle battery systems. Furthermore, the absence of the memory effect addresses a critical usability concern that afflicts older technologies. These advantages are not minor incremental improvements but represent a fundamental qualitative leap in storage capability. The remarkable energy density of lithium-ion cells directly enables the driving range that consumers expect from modern electric vehicles.
1.2 Essential Functions of Battery Management Systems
My analysis of existing energy management systems highlighted several essential functions that must be integrated into a comprehensive solution for electric vehicle battery packs. These functions together form the backbone of safe and efficient battery operation:
(1) Safety Protection: This is the non-negotiable foundation of any battery management system. The system must continuously monitor critical parameters and intervene immediately when dangerous conditions are detected. This includes protection against overcharging, over-discharging, over-current events, and thermal runaway scenarios. In my design, I implemented multiple layers of protection to ensure redundancy and reliability in safety-critical situations.
(2) Real-time State Monitoring: Accurate measurement of voltage, current, and temperature forms the basis for all system decisions. The integrity of these measurements directly determines the effectiveness of protective actions and the accuracy of SOC estimation. I recognized that improving measurement precision, particularly for voltage and current, is fundamental to advancing battery management technology. The precision of current sensing directly impacts the SOC calculation, which is central to user experience.
(3) Cell Equalization: When cells are connected in series, their individual characteristics naturally diverge over time due to manufacturing tolerances and thermal gradients within the pack. These differences, if left unaddressed, will progressively degrade the capacity of the entire electric vehicle battery pack, as the weakest cell limits the performance of the whole string. My research focused on implementing an effective equalization strategy to maintain all cells at similar states, thereby maximizing the usable capacity of the pack.
(4) Thermal Management: Temperature exerts a profound effect on battery performance and aging. The system must include provisions for temperature monitoring and potentially active cooling to maintain the cells within their optimal operating window. In my design, I incorporated temperature sensing to support this function and inform the SOC algorithm.
Despite the progress made in battery management research, I identified significant gaps in current implementations. The precision of data acquisition modules often fails to meet the requirements for accurate SOC estimation, particularly under dynamic load conditions. Additionally, the challenges of cell balancing become increasingly pronounced as pack sizes grow. My thesis sought to address these specific shortcomings through a combination of improved hardware design and more sophisticated estimation algorithms.
2. Methodologies for Remaining Capacity and State-of-Charge Estimation
The concept of State-of-Charge (SOC) is the central variable in battery management, providing a normalized measure of the remaining energy in the cell. It serves as the intuitive interface between the physical battery and the user. A accurate SOC value enables the driver of an electric vehicle battery system to make informed decisions about range management and charging schedules. Without a reliable SOC, the usefulness of the battery is fundamentally compromised.
2.1 Definitions and Fundamental Concepts
Electrochemically, SOC is formally defined as the ratio of the remaining charge capacity to the nominal capacity of the battery. The foundational equations are as follows:
$$SOC = \frac{Q_{c}}{C_{I}} \qquad (2-1)$$
where \(Q_{c}\) represents the residual capacity of the battery and \(C_{I}\) represents the capacity discharged at a constant current \(I\). An alternative expression in terms of the cumulative discharged charge Q is:
$$SOC = 1 – \frac{Q}{C_{I}} \qquad (2-2)$$
In this convention, SOC = 1 denotes a fully charged battery, while SOC = 0 signifies discharge to the specified cut-off voltage. However, my research indicated that this classic definition has inherent limitations when subjected to variable load conditions. The fundamental issue arises from the fact that the usable capacity of a lithium-ion cell is not constant but depends on the discharge rate and temperature. To improve robustness, I adopted a modified definition that incorporates the effect of discharge rate and temperature on capacity, accounting for the fact that a battery may possess residual energy even when SOC is reported as zero.
2.2 Factors Influencing Remaining Capacity
A comprehensive understanding of the factors that affect remaining capacity is essential for developing accurate SOC estimation algorithms. I systematically analyzed the five primary factors that have the most significant impact:
(1) Discharge Current: The relationship between discharge current and capacity is an inverse one; higher currents lead to lower efficiency and reduced overall deliverable capacity. This is due to increased internal losses and polarization effects at elevated rates, a key consideration for high-power electric vehicle battery applications. The battery’s ability to deliver charge is influenced by its internal electrochemical kinetics. This is fundamental to understanding the limitations of lab-based capacity tests versus real-world performance.
(2) Temperature: Temperature affects capacity in a strongly positive manner. Elevated temperatures enhance chemical activity and reaction rates, leading to greater usable capacity. Conversely, low temperatures significantly reduce capacity due to reduced ionic conductivity and slower kinetics. This variation poses a significant challenge for SOC estimation, especially in climates with substantial temperature swings. The temperature compensation factor in any SOC algorithm is critical.
(3) Self-Discharge: Even when a battery is not in use, it gradually loses charge due to internal parasitic chemical reactions. The self-discharge rate for lithium-ion batteries is relatively low compared to other chemistries, but it still needs to be considered, especially for long-term storage scenarios. This effect is often overlooked in simpler SOC models but is crucial for accurate long-term state estimation.
(4) Cycle Life: As a battery is repeatedly cycled, its available capacity steadily declines due to material degradation within the electrodes and electrolyte. Over time, the active lithium inventory decreases, and internal resistance increases, leading to a reduction in usable capacity and power capability. This aging process must be accounted for in the estimation algorithm to maintain accuracy over the lifespan of the electric vehicle battery pack.
(5) Self-Restoration: The battery exhibits a self-recovery effect after the removal of a load, where its terminal voltage gradually approaches an equilibrium open-circuit value. This phenomenon is critical for the open-circuit voltage method. Ignoring this transient behavior can introduce significant errors into SOC estimation.
2.3 Comparative Review of Common SOC Estimation Methods
I conducted an extensive review of the literature, which identified several principal techniques for SOC estimation. Each offers different trade-offs between accuracy, complexity, and practicality. My analysis is summarized in Table 2:
| Method | Advantages | Disadvantages | Appropriate Use Cases |
|---|---|---|---|
| Discharge Test | High accuracy | Time-consuming, offline process required | Battery maintenance, lab calibration |
| Ah Counting | Simple to implement, direct relationship to charge | Suffers from accumulation error, needs known initial SOC | Online estimation for many applications |
| Open-Circuit Voltage (OCV) | Relatively accurate for stable states, no integration required | Impractical for dynamic load scenarios, requires long rest periods | Initial SOC determination, calibration points |
| Internal Resistance | Can be indicative at low SOC | Measurement is difficult, highly variable with temperature and current | Assistive method for SOC refinement |
| Kalman Filter | Optimal state estimation in noise, handles dynamic conditions | Computationally intensive, model-dependent | Dynamic conditions in hybrid electric vehicles |
The discharge test method, while accurate, is unsuitable for online monitoring in an electric vehicle battery application due to its offline nature and the requirement to interrupt normal operation. The ampere-hour (Ah) counting method, also known as Coulomb counting, is commonly used for online estimation. It integrates the current flowing in and out of the battery. The mathematical representation is:
$$SOC(t) = SOC_0 – \frac{1}{C_N} \int_0^t \eta \, i(t) \, dt \qquad (2-3)$$
where \(SOC_0\) is the initial SOC, \(C_N\) is the nominal capacity, \(\eta\) is the Coulombic efficiency, and \(i(t)\) is the instantaneous current. The simplicity of this method is appealing, but its sensitivity to current measurement errors means that errors can accumulate over time, leading to significant drift if the initial SOC is not accurately known. The open-circuit voltage method provides an alternative for establishing the initial state by mapping a well-rested battery’s terminal voltage to its SOC. However, its reliance on a stable and rested condition makes it unsuitable for dynamic load profiles.
Given the specific demands of the electric vehicle battery environment, I concluded that the most effective strategy is to combine the strengths of both Ah counting and OCV measurement. My proposed method integrates these two approaches within a correctional framework that accounts for temperature, cycle life, and charge-discharge efficiency. This hybrid strategy leverages the practicality of Ah counting for continuous online estimation while using OCV measurements to periodically correct for drift and establish reliable initial conditions.
2.4 Proposed SOC Estimation Strategy
My proposed SOC estimation framework decomposes the total SOC into two primary components: an initial value and the dynamic contribution. The initial baseline is derived using the OCV method when the battery is in a stable, rested state. This initialization is critical, as any error in the starting point will propagate through the subsequent Coulomb counting. For the dynamic phase, the algorithm integrates the real-time current. The enhanced mathematical model I developed is:
$$SOC(t) = K_0 SOC_{ocv} + (1 – K_0) SOC’_{last} + K_t K_c \frac{\int_0^t \eta \, i(t) \, dt}{Q_N} \qquad (2-4)$$
where:
- \(SOC_{ocv}\) is the SOC obtained from the open-circuit voltage, used when the system is stable.
- \(SOC’_{last}\) is the SOC from the last shutdown, used during transient conditions when the battery is not fully rested.
- \(K_0\) is a stability constant; when the open-circuit voltage has reached a steady state, \(K_0 = 1\), otherwise \(K_0 = 0\).
- \(K_t\) is a temperature correction coefficient, which I calibrated for room temperature (set to 1 in initial tests).
- \(K_c\) is the cycle life correction factor, representing the aging state of the battery.
- \(\eta\) is the efficiency factor dependent on the charge or discharge rate.
- \(i(t)\) is the instantaneous current, positive for charging, negative for discharging.
- \(Q_N\) is the nominal capacity of the battery pack.
In practical implementation, the treatment of dynamic current profiles is critical. My algorithm digitizes the current-time relationship into discrete intervals. The charge transferred in each interval \(k\) is calculated using:
$$Q_k = Q_{k-1} + \Delta Q_k = Q_{k-1} + i_k \cdot \Delta t \qquad (2-5)$$
where \(i_k\) is the current at the \(k\)-th sample. This numerical integration allows the system to handle fluctuating load currents, which are common during the operation of an electric vehicle battery system, reflecting the varying power demands of the electric motor. The accuracy of this method is contingent upon the precision of the current sensing hardware, which I addressed in the system design.
3. System Hardware Design and Implementation
A robust hardware platform is indispensable for realizing the algorithms and protective functions of the battery management system. My design philosophy centered on creating a modular, reliable, and maintainable system that directly interfaces with the battery pack. The system is powered directly by the lithium-ion battery pack itself, eliminating the need for a separate supply, which simplifies the system architecture.
3.1 Overall System Architecture
The architecture I selected is anchored by a Microchip PIC18F452 microcontroller, a choice driven by its robust feature set, ease of use, and sufficient processing capability. The system block diagram illustrates the core functional components I designed:
- Sensing Modules: For precise measurement of cell voltages, pack current, and temperature.
- Protection Module: Dedicated circuits for overcharge, over-discharge, and overcurrent cut-offs.
- Equalization Module: To maintain voltage balance across the series-connected cells.
- Display and Control Module: For providing information to the user and receiving input via a keypad.
- Communication Module: A serial interface for potential data logging or external monitoring.
Functionally, the system was designed to fulfill four main requirements: real-time data acquisition, SOC estimation and display, comprehensive battery safety protection, and cell voltage equalization. The robustness of this architecture stems from its modularity, allowing for each component to be developed, tested, and optimized independently before integration.
3.2 Microcontroller Selection and Core Circuitry
I selected the PIC18F452 microcontroller for its high performance and integrated peripherals. The microcon troller features 32 KB of flash memory and 256 bytes of EEPROM, which is sufficient for my intended program and calibration data. Furthermore, the on-chip 10-bit analog-to-digital converter (ADC) with multiple channels is a crucial feature, as it simplifies the interface for analog sensors. The device operates at up to 40 MHz, ensuring quick instruction execution. The instructions are single-cycle, improving throughput efficiency. For a management system, low power consumption is also an important factor, a feature this MCU supports through various low-power states.
The minimal peripheral circuit includes a stable power supply. Given the MCU’s 5V supply requirement, I utilized an LM7805 linear regulator to convert the battery pack’s voltage to a stable +5V. This approach is straightforward and effective for low-power applications. To ensure adequate power supply rejection, I incorporated both 100 nF and 10 µF decoupling capacitors near the power pins of the MCU. I also designed a simple resistor-capacitor external reset circuit connected to the MCLR pin to ensure proper startup and manual reset capabilities. An 8 MHz crystal oscillator is connected to the OSCI and OSC2 pins, providing the necessary clock source for the microcontroller core. This frequency dictates that each instruction cycle lasts 0.5 microseconds, a good compromise between speed and power consumption.
I arranged the I/O ports for specific functions to keep the layout logical and the circuit board design straightforward. Port A pins were allocated for analog input signal conditioning. Port B was used for the keypad interface, and Port C for the temperature sensor and the display driver interface. This allocation kept the digital communication lines closer together and separated from the analog input channels.
3.3 Design of Voltage and Current Acquisition Circuits
Accurate data acquisition is the linchpin of the energy management system. The total pack voltage was measured using a potential divider network. Due to the maximum pack voltage being around 30V, and the ADC reference being 5V, a precise voltage division is required. I used two precision resistors with 0.1% tolerance to achieve this reduction. The relationship defining the measured voltage is:
$$V_{ADC} = V_{pack} \cdot \frac{R_{bottom}}{R_{top} + R_{bottom}} \qquad (3-1)$$
For individual cell voltage monitoring, I devised a more elaborate circuit. Direct measurement of series-connected cells is complicated by the fact that the positive terminal of one cell is the negative terminal of the adjacent cell, creating a common-mode voltage issue. I addressed this by designing a sampling circuit using a general-purpose operational amplifier, the LM358, and a P-channel MOSFET, the F9630. The circuit in the figure acts as a level shifter, converting the cell voltage into a ground-referenced signal that can be safely read by the ADC. The amplifier operates to force the MOSFET source voltage such that the voltage across a known low-side resistor equals the cell voltage. This approach enables accurate measurement of each cell without requiring complex differential instrumentation amplifiers, providing a cost-effective and practical solution.
Current measurement is equally crucial for accurate SOC tracking. I selected a direct sensing method, placing a precision shunt resistor directly in the circuit. A resistance of 1 ohms with a power rating of 5W was chosen as a compromise between a sufficient voltage drop for sensing and minimized power loss. The current-residue voltage (\(V_i\)) across the shunt is then processed by a specific signal conditioning circuit. The core of this circuit is an absolute value amplifier based on an op-amp (LM324). The circuit’s transfer function was characterized as:
$$V_{out} = 5 \cdot | – V_i | \qquad (3-2)$$
This configuration is elegant as it allows the ADC to read both charge and discharge currents as positive voltages. For discharging, the current flow is in one direction, and for charging, it is in the opposite direction. Without such a circuit, the ADC would only measure half the possible range. I used precision resistors with a tolerance of 1% to guarantee the accuracy of this operational amplifier circuit. The entire processing chain was designed to be linear, maintaining a proportional relationship between the true current and the ADC input voltage.
| Parameter | Design Value | Transfer Function |
|---|---|---|
| Total Pack Voltage | 21-29.4 V | \(V_{ADC} = V_{pack} \times (R_2/(R_1+R_2))\) |
| Cell Voltage | 3-4.2 V | Level-shifted via LM358 & MOSFET |
| Current Sense | ±2 A | \(V_{out} = 5 \times I_{sense} \times R_s / 5\) |
| Temperature | -55 to +125°C | Digital output (DS18B20) |
3.4 Temperature Sensing Strategy
A thorough oversight of temperature is critical for battery safety and SOC compensation. I integrated a digital temperature sensor, the DS18B20, into my design. This sensor was selected for its distinct advantages: direct-to-digital output, which eliminates the need for calibration and external signal conditioning; high resolution, configurable up to 12 bits (0.0625°C); and a wide operating temperature range that spans the extremes of automotive applications. The “1-Wire” interface is a major benefit, as it requires only a single I/O pin for communication with the microcontroller. I used the external power supply mode, providing VDD with +5V, to ensure fast temperature conversions and reliable operation. This thermometer circuit is key for managing the battery’s thermal envelope.
Figure 1: Overview of the power battery system with integrated energy management hardware for electric vehicle battery applications.
3.5 Design of the Display and Keypad Interface
For user interaction, I developed a straightforward interface. The system’s core output is the digital display, which presents runtime data. I used the MAX7219 LED driver chip to manage a bank of four seven-segment displays. The MAX7219 is a crucial component because it handles the multiplexing tasks that would otherwise consume significant processor time. It communicates via a simple 3-wire serial interface (DATA, CLK, LOAD). I wrote a driver to send the appropriate BCD codes and select the correct display digit. The chip also offers a convenient method for adjusting the display brightness through an external resistor connected to the ISET pin. This reduces the number of I/O pins required and offloads the scanning task from the CPU.
The keypad is an array of independent push-button switches. I designed the key K1 to display the system voltage, K2 to display the current, and K3 to display the temperature. By default, the display shows the calculated SOC. This selective viewing allows the user to quickly check the essential operational parameters. In the software, I incorporated a debouncing algorithm to eliminate the mechanical chatter of the switch contacts.
In addition to the numerical display, I also developed a more visual indicator for battery energy levels using discrete LEDs. This system divides the battery capacity into four 25% segments. The circuit, built around a voltage comparator and a TL431 precision reference, illuminates a sequence of LEDs based on the pack voltage. This setup provides a quick visual gauge of the remaining energy, an intuitive feature for an electric vehicle battery pack.
3.6 Hardware-Based Protection Circuitry
Ensuring the safety of the lithium-ion battery pack is the highest priority in my design. I implemented a hardware protection circuit that offers a reliable defense independent of the main microcontroller. I selected the S-8261 series protection IC from Seiko Instruments. This chip is specifically designed for protecting single-cell lithium-ion batteries against overcharging, over-discharging, and overcurrent. Its strength lies in its precise internal voltage detection and integrated delay circuits, which distinguish between transient surges and genuine fault conditions.
To protect a multi-cell pack, I designed a cascaded circuit where each cell has its own S-8261 protection IC. The critical challenge in this topology is the lack of common ground between the cells. I solved this by using optocouplers (PC817) to transmit the digital fault signals (CO and DO outputs) from the high-side cell protection chips to the low-side control logic. This galvanic isolation is essential for the correct and safe functioning of the circuit.
The protection logic operates as follows:
- Overcharge Protection: When any protection IC detects its cell voltage exceeding the overcharge threshold (4.28V ± 0.025V), its CO output goes low. This activates the charging optocoupler, which in turn pulls the gate of the charge control MOSFET to ground, turning it off and interrupting the charging path. This stops any further current from flowing into the pack, preventing damage.
- Overdischarge Protection: Conversely, if the cell voltage falls below the discharge threshold (2.3V ± 0.05V), the DO output goes low. This turns off the discharge control MOSFET, isolating the load and preventing the cell from being deeply discharged. This is particularly important as deep discharge can irreversibly damage the cell and decrease its capacity.
- Overcurrent and Short-Circuit Protection: The S-8261 monitors the voltage drop across a sensing resistor in the discharge path. If this voltage exceeds the overcurrent threshold, the discharge MOSFET is turned off after a predefined delay. For a short circuit, the delay is much shorter, typically less than 320µs, ensuring a quick response.
| Parameter | Specification |
|---|---|
| Overcharge Protection Voltage | 4.28 ± 0.025 V |
| Overcharge Delay Time | < 1.3 s |
| Overdischarge Protection Voltage | 2.3 ± 0.05 V |
| Overdischarge Delay Time | < 150 ms |
| Overcurrent Protection Value | 5 A ± 10% |
| Overcurrent Delay Time | < 10 ms |
| Short Circuit Delay Time | < 3 ms |
| Static Power Consumption | < 200 µA |
I also integrated a self-locking function using a quad NAND gate (MC14011). This is a sophisticated addition that prevents the discharge MOSFET from oscillating between on and off states when the battery voltage is hovering near the under-voltage threshold. This latching mechanism keeps the battery disconnected until the load is physically removed, preventing the “ON-OFF” cycling that can generate heat and stress the power MOSFETs. This safety circuit provides a robust hardware foundation for the battery pack.
3.7 Circuit Design for Charge Equalization
Cell imbalance is an inevitable issue in series-connected battery strings. During the charging process, this imbalance can lead to some cells being overcharged while others are still undercharged, which severely compromises safety and cycle life. I recognized that effective charge equalization is fundamental to maximizing the overall capacity of an electric vehicle battery pack. For this design, I focused on passive or dissipative equalization due to its simplicity, reliability, and low cost, making it suitable for the current and capacity levels of this project.
The dissipative equalization circuit I designed incorporates the protection IC’s overcharge signal to control a bypass path. The circuit’s topology includes the protection IC (S-8261), a P-channel MOSFET (SI4435), and a bypass resistor. In the normal state, the CO output is high, keeping the MOSFET in the off-state. When a cell reaches the charge termination voltage, the CO output goes low, which turns on the MOSFET. This creates a parallel path for current, diverting the charging current away from the cell. The energy is dissipated as heat in the bypass resistor, limiting the voltage across the cell and preventing it from exceeding the safe threshold. Once the cell voltage drops back to the release voltage, the CO signal returns high, turning off the MOSFET, and the charging continues. This switching action creates a controlled clamping effect at the target voltage.
For the bypass resistor, I selected a 68-ohm resistor with a power rating of 2W. This selection was based on a balance between the equalization speed and thermal dissipation. The maximum equalization current for a single cell was calculated as \(I = 4.2V / 68\Omega = 62mA\). The corresponding power dissipated as heat is \(P = I^2R = 0.26W\). This value is well within the resistor’s capability, ensuring reliable long-term operation without overheating. The simple external circuit implementation of this module allows a dedicated equalizer to be placed across each cell in the battery pack, providing a consistent and uniform balancing capability.
4. Software Architecture and Algorithm Implementation
The software is the “brain” of the management system, orchestrating the hardware to perform complex tasks. I structured the software following a modular architectural design, which enhances code organization, maintainability, and debuggability. The main program executes a deterministic sequence: initialization, data acquisition, algorithmic processing, and display updating. The flow diagram on the right shows that the software becomes a continuous loop that processes real-time data and responds to user input.
4.1 System Initialization
Upon power-up or reset, the software executes a series of initialization routines. This critical first step sets all necessary registers to their default states to ensure the hardware functions as intended. The key initialization steps are:
- I/O Port Initialization: Configuring the direction (input/output) and selecting the analog functions for the required pins of PIC18F452.
- Timer Initialization: Setting up the TMR0 and TMR1 timers. These timers are used for generating a periodic interrupt for the software scheduler and for precisely measuring time intervals for current integration.
- ADC Initialization: Configuring the analog-to-digital converter module with respect to clock selection, input channel multiplexing, and result formatting.
- Display Initialization: Sending the initialization commands to the MAX7219 driver, which includes setting the scan limit and the decode mode, and turning off the test mode.
This initialization phase ensures the system is in a known, stable state before entering the main control loop. This is essential to prevent any unexpected behavior during the initial power-up sequence.
4.2 Data Acquisition and Filtering
The precise measurement of voltage and current is paramount. The ADC module is repeatedly sampled to gather data for the voltage, current, and temperature inputs. To improve the stability of the readings and to mitigate the effects of noise and transient spikes, I preceded the estimation algorithm with a software filtering step. I implemented an averaging filter algorithm, a simple technique effective for random noise. The algorithm samples the signal n times and then computes the average value, as shown by:
$$\bar{V} = \frac{1}{n} \sum_{i=1}^{n} V_i \qquad (4-1)$$
In my design, I collected 8 consecutive samples and computed their average before using the result in the estimation algorithm. This averaging process smooths out minor fluctuations in the signal, leading to a more stable and accurate SOC calculation. For the DS18B20 temperature sensor, the software follows a specific 1-Wire communication protocol, which includes initialization, ROM commands, and function commands, to retrieve the digital temperature value.
4.3 SOC Estimation Algorithm Implementation
The software implementation of the SOC estimation algorithm is the core intellectual property of my system. The algorithm starts by determining the initial SOC value. In my design, the system checks if the open-circuit voltage has reached a stable state. If the battery has been at rest for a sufficient period and the voltage is stable, the algorithm performs a lookup in the pre-programmed OCV-SOC table using the measured voltage. This table was derived from prior experiments.
Once the initial SOC is established, the program enters the Coulomb counting loop. During each timer interrupt, the current is sampled and integrated. The integration calculates the amount of charge added or removed from the battery. The real-time SOC is then updated by combining the initial SOC and the integrated charge. The logic for this is:
$$SOC_{now} = SOC_{start} + \frac{\int_{t_0}^{t_1} \eta \cdot i(t) \cdot dt}{Q_N}$$
The program code on the left shows the continuous nature of this process. The estimation is done in real-time, and the result is repeatedly output for display. I also coded special handling to update the battery’s cycle life counter, which adjusts the \(K_c\) factor, ensuring the system adapts to battery aging. The whole process is designed to run seamlessly in the background, constantly refining the SOC estimate as the primary output of the system.
5. Experimental Validation and Results Analysis
Following the theoretical design and implementation, I conducted extensive experimental validation to evaluate the performance of my energy management system. The experiments tested the core functionalities of SOC estimation, data acquisition accuracy, and the validity of the protection mechanisms. I will present the key experiments and their results here.
5.1 Experimental Set Design
My test bench consisted of a lithium-ion battery pack, the assembled management system hardware, a constant current electronic load for discharging, a chargeable power supply, and a personal computer for data logging. The specific battery cells used in the tests were model TSE-086760, supplied by a Chinese manufacturer. The configuration of the battery was 7 cells connected in series. The essential parameters of this battery pack are summarized in Table 5:
| Parameter | Specification |
|---|---|
| Model | TSE-086760 |
| Nominal Capacity | 2 Ah |
| Rated Voltage | 25.9 V |
| Maximum Charge Voltage | 29.4 V |
| Discharge Cut-off Voltage | 21 V |
| Maximum Continuous Charge Current | 2 A |
| Maximum Continuous Discharge Current | 2 A |
| Operating Temperature | Charge: 0 to 45°C; Discharge: -20 to 45°C |
5.2 Open-Circuit Voltage Characterization
The initial experiment was designed to map the Open-Circuit Voltage (OCV) to the State-of-Charge (SOC) relationship. This mapping is vital for the accurate initial SOC estimation. I performed the test in a stable thermal environment of 23°C. The method involved discharging the battery in steps with an extremely low current (0.05C), and allowing the battery to rest until its voltage stabilized. This resting period is essential to mitigate the effects of over-potential and to obtain the true electromotive force of the battery. The data collected enabled the creation of the characteristic curve.
My analysis of the OCV-SOC relationship yielded a characteristic curve that is typical for lithium-ion chemistries. The curve exhibits a flat region in the mid-SOC range. This region indicates a stable voltage plateau, which highlights the challenge of estimating SOC via voltage alone when the SOC is between 20% and 80%. The voltage changes very little for large changes in the SOC, making it difficult to estimate the SOC with high resolution. However, in the high (SOC > 80%) and low (SOC < 20%) ranges, the voltage changes dramatically. This steep slope in the voltage allows for a much more precise estimation of the SOC. This confirms why the OCV method is often favored for the initial calibration or during near-full or near-empty conditions, where its accuracy is highest.
5.3 Analysis of Charge and Discharge Profiles
Next, I evaluated the system’s ability to track the dynamic characteristics of the battery during charging and discharging. For the charge test, I used the standard Constant Current-Constant Voltage (CC-CV) method. The data shows that during the CC phase, the terminal voltage rises steadily, and the calculated SOC increases linearly with respect to the integrated current. When the pack reached the cut-off voltage of 29.4V, the system switched to CV mode. In this phase, the current decays exponentially as the battery approaches full saturation. This is a demanding test for the system due to the changing current. My recorded SOC curve remained smooth and monotonic, validating the correct implementation of the current integration algorithm.
The discharge tests were conducted at different constant current rates to assess the impact of the discharge current (C-rate) on the usable capacity and the performance of the management system under different loads. The voltage curves for the 0.2C, 0.4C, and 0.5C discharge rates are of particular interest. The experimental voltage curves indicate that a higher C-rate results in a lower terminal voltage for any given depth of discharge. This is a known effect of internal resistance and higher polarization losses. More importantly, it was observed that the total delivered capacity before reaching the cut-off voltage was greater at lower discharge rates.
| Discharge Rate | Discharge Current | Delivered Capacity | Relative Capacity |
|---|---|---|---|
| 0.2C | 400 mA | 1420 mAh | 100% |
| 0.4C | 800 mA | 1380 mAh | 97.2% |
| 0.5C | 1000 mA | 1229 mAh | 86.6% |
This capacity reduction demonstrates the rate-dependent nature of the battery, and these findings are consistently aligned with my theoretical analysis. This confirmation is critical because it justifies the need for the \(\eta\) (efficiency factor) in my SOC algorithm. If the algorithm assumed a constant capacity, it would introduce significant errors at higher discharge rates. These experiments proved that my algorithm’s compensation for this variable is necessary for accurate real-world performance.
5.4 Model Validation under Variable Load Conditions
Finally, I tested the system using a variable resistance load to simulate more realistic usage patterns. I used a 40 Ω load as a proxy for an average constant-power load. The sampled voltage data shows a gradual, steady decline, characteristic of a constant resistance discharge. The current data remained relatively flat during the test. The SOC curve displayed a nearly linear decrease over time, as expected. In this scenario, the current remained around 600mA, which resulted in nearly Coulombic efficiency. The system successfully captured the data in real-time, performing the A/D conversion and updating the SOC estimate without any observable lag or fault, demonstrating the viability of the practical application for an electric vehicle battery monitoring system.
6. Conclusion and Future Research Directions
In this thesis, I have successfully designed, implemented, and validated a comprehensive energy management system specifically tailored for power lithium-ion battery packs. The core objective was to improve the safety, reliability, and lifecycle of the electric vehicle battery. My research journey encompassed the full spectrum from theoretical modeling to practical hardware and software implementation. The key findings and conclusions of this work are summarized as follows:
(1) I developed a hybrid SOC estimation methodology. This approach judiciously combines the real-time practicality of the ampere-hour integration method with the stable, absolute reference provided by the open-circuit voltage method. My experimental verification confirmed that this hybrid method, which integrates temperature and cycle-life compensation, effectively addresses the limitations of each method when implemented individually, enhancing the robustness and reliability of SOC estimation across a wide range of operating conditions.
(2) The hardware system I built, based on the PIC18F452 master controller, proved to be a stable and efficient platform. The carefully designed signal conditioning circuits for voltage and current ensured good measurement accuracy. The application of an averaging filter in the software further reduced measurement noise, providing a reliable source of truth for the state estimation algorithms. The design of the system was both cost-effective and modular.
(3) The implementation of dedicated hardware protection and an equalization scheme using the S-8261 was successful in fulfilling the primary safety requirements. The circuit’s functionality demonstrated reliable overcharge, over-discharge, and overcurrent protection. The self-locking feature of the discharge control circuit provided a robust defense against a problematic oscillation that is common in batteries operating at low voltages, significantly contributing to system longevity. The layout and function of the protection circuit demonstrated the feasibility of using this approach in a multi-cell series configuration.
(4) Experimental validation confirmed the system’s functionality. The system efficiently logged the discharge data under various C-rates. The analysis of the capacity data, showing a clear dependency on discharge rate, verified the crucial role of the rate-correction factor in my SOC algorithm. Without it, the SOC data would have been misleading. The system successfully managed the battery pack during charge and discharge cycles, accurately reporting SOC and reacting as programmed to boundary conditions, proving its viability for real-world applications as a fundamental component of an electric vehicle battery management system.
Despite the successes of this project, I have identified areas for future improvements in this domain. The scale of the current design is limited to a 7-cell battery pack used in this study. Scaling to a much larger number of cells for a full-scale automotive application, which can involve hundreds of cells, would be a natural next step. The developed hardware architecture, especially the modular protection circuit, exhibits potential for memory expansion, but this would require more complex balancing and control strategies. Future work could also enhance the fidelity of the SOC algorithm by incorporating more advanced state-estimation techniques, such as an extended Kalman filter or a particle filter, which are more capable of handling nonlinearities and system noise than my current algorithmic correction method. Also, equalization circuits that transfer energy instead of dissipating it as heat would offer superior energy efficiency, though their complexity increases. I believe the foundational work laid out in this thesis provides a solid stepping stone toward a more advanced, complete, and intelligent power battery management solution, contributing to the eventual mass adoption of clean and efficient electric transportation.