Simultaneous power and data transfer over common bus using talkative bidirectional flyback converters for battery equalization Authors: Jiting Wang, Bingkun Zeng, Lingyu Li, Jiande Wu, Yan Deng, Xiangning He, B. T. Maharaj, Ramesh C. Bansal Published: 2026 ABSTRACT Cell inconsistency is an inherent challenge for lithium-ion batteries during long-term operation, necessitating the use of battery equalization systems (BESs). In such systems, real-time data communication is essential for ensuring accurate system coordination and effective equalization performance. This paper proposes a distributed BES based on the talkative power converter (TPC) technique, employing a bidirectional flyback converter as the core equalization unit. In the proposed approach, the converter not only transfers energy among distributed battery modules but also modulates data signals for communication. This dual-function design eliminates the need for additional communication hardware or dedicated signal transmission circuits, thereby reducing the overall system complexity. On this basis, a distributed equalization system and its communication channel model are developed. The system noise sources are then analyzed and a noise suppression strategy is proposed to improve the communication reliability. Furthermore, a practical communication protocol tailored to the distributed architecture is designed. Finally, a hardware prototype is constructed and tested, achieving real-time data communication at a rate of 8.33 kb/s. The results verify the feasibility of using switching-ripple-based power and signal dual modulation for communication within distributed BESs. Introduction With the widespread adoption of electric vehicles and renewable energy, energy storage technologies have attracted significant attention from both academia and industry1. Owing to advantages such as high energy density, low self-discharge rate, and favorable temperature characteristics, lithium-ion batteries are extensively used in the energy storage systems2,3. However, the long-term operation of a lithium-ion battery pack inevitably leads to inconsistencies in the State of Charge (SOC) among individual cells4, preventing the complete utilization of the stored charge and causing potential safety hazards5. To address these issues, battery packs are typically equipped with a battery equalization system (BES). A battery pack can be organized into multiple cells or several modules, each consisting of a number of cells. By balancing the SOC of the individual cells or modules, the BES effectively prevents overcharging and over-discharging, thereby extending the overall lifespan of the battery pack6–8. According to the system architecture, BESs can generally be classified into centralized and distributed structures. In distributed architectures, each cell or module is equipped with a dedicated equalizer board, and reliable real-time communication is required to exchange SOC information among different cells or modules. Therefore, the design of a safe, stable, and cost-effective communication scheme has become a critical challenge for distributed BESs9. Communication schemes in distributed BESs can generally be classified into independent-channel and shared-channel schemes, depending on whether power and data are transmitted over the same physical path. Wired communication is widely adopted in BESs, and architectures based on the Controller Area Network (CAN) bus have been extensively investigated and successfully implemented10,11. Although the daisy-chain topology features a simpler circuit structure compared with CAN-based solutions12, both approaches require dedicated communication modules in addition to the power equalization hardware. As the number of cells or modules increases, the system inevitably demands a large number of communication modules and interconnecting cables. The resulting wiring complexity not only increases manufacturing cost but also degrades system reliability due to the increased risk of connection failures. In contrast to independent-channel schemes, shared-channel schemes feature lower hardware complexity, simplifying the system and reducing costs. Power Line Communication (PLC) was applied in BESs13–16. For instance, Reference17 established a small-scale PLC network for cylindrical batteries. However, traditional PLC schemes require additional signal modulation circuits, and the coupling between power and signals can lead to mutual interference, degrading the communication quality. Power electronic converters inherently possess the capability to transmit data without requiring additional hardware or structural modifications. The technique of embedding information into the power waveform generated by a switched-mode converter is known as talkative power conversion (TPC), which is commonly realized through power and signal dual modulation (PSDM). By integrating power regulation and data transmission within a single modulation framework, PSDM eliminates the need for dedicated communication hardware or separate signal paths, thereby significantly reducing system complexity. The feasibility and effectiveness of this approach have been demonstrated in a wide range of applications18–20. Recent studies have further investigated the application of TPC in BES. Reference21 applied TPC to the equalization circuits of series-connected battery strings, enabling simultaneous power and data transfer. Building on this work22, further refined both the equalization and communication strategies, achieving improvements in communication reliability and equalization speed. However, these schemes are primarily designed for cell-level equalization using a daisy-chain topology and are not suitable for module-level equalization. Fig. 1Structure of BES. Structure of BES. This paper proposes a module-level BES scheme whose system architecture is illustrated in Fig. 1. A string of battery modules is connected in series to form a battery pack, and each battery module is interfaced with a Battery Equalization Unit (BEU). Each BEU is implemented using a bidirectional flyback converter integrated with the TPC technique, referred to as a talkative bidirectional flyback converter. All BEUs are connected to a common DC bus supplied by a bidirectional power source. By employing the TPC technique, the power bus is multiplexed as a communication channel, thereby enabling simultaneous power and data transfer. The main contributions of this paper are summarized as follows:A modulation strategy based on Frequency Hopping-Differential Phase Shift Keying (FH-DPSK) is applied to embed information into the converter’s switching ripple. This method utilizes distinct switching frequencies for communication and power control, encoding data in the phase difference between consecutive ripple groups. The receiver identifies the communication status by detecting specific frequency component on the bus and performing demodulation without requiring precise phase synchronization from the transmitter.A distributed BES is constructed based on the proposed BEUs. Unlike traditional architectures, this system eliminates the need for dedicated modulation/demodulation hardware and separate signal cabling. This approach significantly simplifies system design, reduces manufacturing costs, and enhances overall system density.A channel model for the distributed BES is established. Furthermore, the system’s noise sources are comprehensively analyzed, and corresponding noise suppression strategies are developed to ensure real-time communication. A practical protocol is designed to ensure orderly data exchange in the distributed system. A modulation strategy based on Frequency Hopping-Differential Phase Shift Keying (FH-DPSK) is applied to embed information into the converter’s switching ripple. This method utilizes distinct switching frequencies for communication and power control, encoding data in the phase difference between consecutive ripple groups. The receiver identifies the communication status by detecting specific frequency component on the bus and performing demodulation without requiring precise phase synchronization from the transmitter. A modulation strategy based on Frequency Hopping-Differential Phase Shift Keying (FH-DPSK) is applied to embed information into the converter’s switching ripple. This method utilizes distinct switching frequencies for communication and power control, encoding data in the phase difference between consecutive ripple groups. The receiver identifies the communication status by detecting specific frequency component on the bus and performing demodulation without requiring precise phase synchronization from the transmitter. A distributed BES is constructed based on the proposed BEUs. Unlike traditional architectures, this system eliminates the need for dedicated modulation/demodulation hardware and separate signal cabling. This approach significantly simplifies system design, reduces manufacturing costs, and enhances overall system density. A distributed BES is constructed based on the proposed BEUs. Unlike traditional architectures, this system eliminates the need for dedicated modulation/demodulation hardware and separate signal cabling. This approach significantly simplifies system design, reduces manufacturing costs, and enhances overall system density. A channel model for the distributed BES is established. Furthermore, the system’s noise sources are comprehensively analyzed, and corresponding noise suppression strategies are developed to ensure real-time communication. A practical protocol is designed to ensure orderly data exchange in the distributed system. A channel model for the distributed BES is established. Furthermore, the system’s noise sources are comprehensively analyzed, and corresponding noise suppression strategies are developed to ensure real-time communication. A practical protocol is designed to ensure orderly data exchange in the distributed system. This paper is organized into the following sections. Section 2 introduces the information flow within the BES and details the modulation and demodulation processes for information in the bidirectional flyback converter. Section 3 describes the construction of the distributed BES, analyzes noise and its suppression strategies, and presents the communication protocol for the system. Section 4 provides experimental validation of the proposed communication method and the equalization system. Communication of battery equalization system Power and information flow in battery equalization system In the proposed BES architecture shown in Fig. 1, the bus functions simultaneously as a power transmission medium and a communication channel. A bidirectional power supply establishes the DC bus and enables energy input, output, and temporary storage within the BES. During power transfer, energy is discharged from high-SOC battery modules through BEUs to the bidirectional power source. When necessary, the stored energy is redistributed via the common bus and BEUs to battery modules with lower SOC, thereby achieving energy balancing across the battery modules. By employing the PSDM technique, information is embedded into the switching ripple generated by the transmitting BEU and injected into the common bus concurrently with power transfer. The underlying principle of PSDM lies in the multiple degrees of freedom inherent in a PWM waveform, namely the duty cycle, switching frequency, and phase. In this scheme, the duty cycle is exclusively allocated for power regulation, while the switching frequency and phase are modulated to convey information, enabling simultaneous power and data transmission over the same bus. Figure 2 illustrates the information flow within the BES. At the transmitter, the binary data is used to modulate the triangular carrier wave of the PWM generator. This modulation alters the frequency and phase of the gate drive signals, which subsequently induces corresponding variations in the converter’s input voltage or current ripple. Consequently, digital information is embedded directly into the power bus. The receiving BEUs filter the bus voltage to extract the ripple component at the designated communication frequency. This signal is then processed by specific demodulation algorithms to recover the transmitted data. Fig. 2Information flow in BES. Information flow in BES. FH-DPSK modulation and data transmission in bidirectional flyback converter Power transmission in bidirectional flyback converterThe bidirectional flyback converter evolves from the conventional topology by replacing the secondary diode with an active switch. This modification enables bidirectional power flow by regulating the current direction while maintaining consistent voltage polarity. Power transmission in bidirectional flyback converter Power transmission in bidirectional flyback converter Depending on the direction of average current, the converter operates in three modes: forward power transfer mode (i1 > 0), backward power transfer mode (i1 < 0), and a transitional mode (i1 ≈ 0). Complementary gate signals are applied to the primary switch Q₁ and the secondary switch Q₂ to ensure one switch is always conducting, thereby eliminating the discontinuous conduction mode. Fig. 3Two operational stages during forward power transfer mode. (a) t0-t1 (b) t1-t0+Ts Two operational stages during forward power transfer mode. (a) t0-t1 (b) t1-t0+Ts Figure 3 illustrates the two operational stages in forward power transfer mode. During the interval t0-t1, Q1 is turned on while Q2 remains off. The primary current i1 increases linearly. Given the primary-side magnetizing inductance Lm, the rising slope of the primary current i1 is:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{{\Delta {i_1}}}{{\Delta t}}=\frac{{{V_1}}}{{{L_{\mathrm{m}}}}}$$\end{document} During the interval t1-t0+Ts, Q1 turns off while Q2 turns on. To maintain flux continuity, the magnetizing current flows through Q2. Consequently, the secondary current i2 decreases linearly. With the secondary-side equivalent inductance defined as n²Lₘ where n denotes the transformer turns ratio, the current decay slope is:2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{{\Delta {i_2}}}{{\Delta t}}=\frac{{{V_{\mathrm{o}}}}}{{{n^2}{L_{\mathrm{m}}}}}$$\end{document} Similarly, for both backward power transfer (i1 < 0) and transitional (i1 ≈ 0) modes, the rates of change of i1 and i2 remain consistent with Eqs. (1) and (2). Fig. 4Operational waveform of bidirectional flyback converter in different modes. Operational waveform of bidirectional flyback converter in different modes. Figure 4 shows the key operational waveforms of the bidirectional flyback converter. The diagram illustrates the gate drive signals ug1 and ug2 for Q1 and Q2, the primary and secondary currents i1 and i2, the transformer magnetic flux φm, and the switching period Ts. Fig. 5The process of data modulation and transmission in battery equalization system. The process of data modulation and transmission in battery equalization system. 2)Data modulation in bidirectional flyback converter 2)Data modulation in bidirectional flyback converter Data modulation in bidirectional flyback converter The principle of data modulation and transmission using the BEU is detailed in Fig. 5. The transmitter performs digital modulation on the standard PWM power carrier. By altering the carrier’s frequency and phase according to the transmission data, a composite power-signal carrier is generated. This composite carrier is compared with the output of the PI controller to generate the gate drive signals. Consequently, the digital information is embedded into the frequency and phase of the switching ripple at the converter’s input and output ports, achieving simultaneous power conversion and data transmission. Conventional digital modulation schemes include Amplitude Shift Keying (ASK), Frequency Shift Keying (FSK), Phase Shift Keying (PSK), and Differential Phase Shift Keying (DPSK). DPSK encodes data using the phase difference between two consecutive carrier groups. However, in the distributed BMS where multiple BEUs share the common DC bus, the standard DPSK scheme maintains the same switching frequency for transmitters. Consequently, the switching ripple from non-transmitting BEUs causes significant interference to the demodulation process. To mitigate this, a hybrid Frequency Hopping-Differential Phase Shift Keying (FH-DPSK) strategy is adopted. This strategy simultaneously uses frequency and phase to modulate the power carrier, ensuring robust communication in the multi-converter environment. Fig. 6FH-DPSK modulation process for the power carrier. (Taking N = 4 as an example) FH-DPSK modulation process for the power carrier. (Taking N = 4 as an example) A typical FH-DPSK modulation process is illustrated in Fig. 6. This scheme employs two switching frequencies to identify the communication state: a baseline frequency f0 during idle periods, and a communication frequency f1 during data transmission. Data is transmitted in symbols, with each symbol consisting of N switching cycles. FH-DPSK demodulation is typically performed using a sliding window technique. By selecting the demodulation window length Tg to satisfy the orthogonality condition in Eq. (3), the interference caused by switching ripples at non-communication frequencies can be effectively eliminated23.3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${T_{\mathrm{g}}}={k_0}*\frac{1}{{{f_0}}}={k_1}*\frac{1}{{{f_1}}},{\text{ }}{k_0},{k_1} \in Z$$\end{document} In binary FH-DPSK, the first group of switching cycles serves as a reference phase. Subsequent data is encoded in the phase difference Δφ between consecutive carrier groups. A phase difference Δφ = 0 represents data “0”, while Δφ = π corresponds to data “1”. This differential encoding enables non-coherent demodulation, eliminating the need for carrier phase synchronization at the receiver. As shown in Fig. 6, the FH-DPSK scheme uses both frequency and phase to achieve data transmission. This scheme preserves the duty cycle of the drive signals, thus maintaining the steady-state output of the BEU. However, variations in the frequency and phase of the drive signals affect the charging and discharging processes of inductors and capacitors in the circuit, introducing output transient disturbances. Fig. 7Improved FH-DPSK modulation process (Taking N = 4 as an example). Improved FH-DPSK modulation process (Taking N = 4 as an example). Theoretical analysis indicates that disturbances induced by frequency shifting are generally smaller than the nominal output voltage ripple and can be neglected. In contrast, direct phase shifting introduces significant voltage spikes, often exceeding the switching ripple magnitude. In the improved FH-DPSK process, phase switching is achieved indirectly through the introduction of transitional carriers, resulting in the entire communication process only changing the power carrier frequency. By altering the frequency of several power carriers, indirect modification of the power carrier phase is accomplished without requiring abrupt phase jumps, as illustrated in Fig. 7. This approach minimizes the impact of data transmission on power quality. Table 1 summarizes the key parameters utilized in this work. Table 1Parameters of data transmitting process.ParameterValueParameterValueSwitching frequency during non-communication periods \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{f}_{0}$$\end{document}100 kHzCarrier cycle during communication T112µsSwitching frequency during communication f183.3 kHzTransitional carrier cycle \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{T}_{2}/{T}_{3}$$\end{document}14µs/10µsSymbol duration \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{T}_{\mathrm{B}}$$\end{document}120µsCommunication rate8.33 kb/s Table 1Parameters of data transmitting process.ParameterValueParameterValueSwitching frequency during non-communication periods \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{f}_{0}$$\end{document}100 kHzCarrier cycle during communication T112µsSwitching frequency during communication f183.3 kHzTransitional carrier cycle \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{T}_{2}/{T}_{3}$$\end{document}14µs/10µsSymbol duration \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{T}_{\mathrm{B}}$$\end{document}120µsCommunication rate8.33 kb/s Parameters of data transmitting process. According to Table 1, the switching cycle during communication is T₁=1/f₁=12µs, and the symbol duration is TB=120µs, meaning N = TB/T₁=10. Figure 8 illustrates the phase switching process with transitional carriers for N = 10. When transmitting data “0”, both the preceding and subsequent carrier groups consist of 10 carrier cycles with a period of 12µs, resulting in no phase change. When transmitting data “1”, the preceding carrier group comprises 7 carrier cycles with 12µs period plus three transitional carrier cycles that enable the phase transition. Fig. 8Phase switching process of carrier with transitional carriers. (N = 10) Phase switching process of carrier with transitional carriers. (N = 10) Receive and demodulation of data As described in Sect. 2.2, the transmitter employs FH-DPSK modulation to encode information into the frequency and phase of the power carrier. In the proposed distributed BES architecture, all BEUs are connected in parallel to the common DC bus. Accordingly, the receiver retrieves the transmitted data by sensing the voltage or current ripple on the DC bus, followed by filtering and demodulation, as shown in Fig. 9. Fig. 9The process of data demodulation in battery equalization system. The process of data demodulation in battery equalization system. The fundamental component of the primary-side current shown in Fig. 4 serves as the primary component for communication. By applying Fourier series expansion to the primary-side current i₁(t), fundamental component i₁₁(t) can be expressed as:4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${i_{11}}(t)={a_1}\cos ({\omega _1}t)+{b_1}\sin ({\omega _1}t)$$\end{document}5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${a_1}=\frac{2}{{{T_1}}}\int_{{{t_0}}}^{{{t_0}+{T_1}}} {{i_1}} (t)\cos \left( {{\omega _1}t} \right){\mathrm{d}}t$$\end{document}6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${b_1}=\frac{2}{{{T_1}}}\int_{{{t_0}}}^{{{t_0}+{T_1}}} {{i_1}(t)\sin \left( {{\omega _1}t} \right){\mathrm{d}}t}$$\end{document} where ω₁ is the angular frequency of the fundamental component. The filtered signal sampled by the receiver contains both the target signal from the transmitter and the background switching noise from non-transmitting BEUs operating at frequency f0. Neglecting noise effects, the filtered signal s(t) can be represented as:7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{gathered}s(t)=Z_{de}*i_{11}(t)+A_{0}\cos(2{\pi}f_{0}t+{\varphi}_{0})\\=A_{1}\cos(2{\pi}f_{1}t+\varphi_{1})+A_{0}\cos(2{\pi}f_{0}t+\varphi_{0})\end{gathered}$$\end{document} where Zde represents the equivalent gain of the sampling circuit, and the term A0cos(2πf0t + φ0) denotes the aggregate switching ripple interference from non-communicating BEUs. To extract the amplitude and phase of the target signal at f1, a demodulation algorithm is employed as illustrated in Fig. 10. Fig. 10The information extraction process from the filtered signal. The information extraction process from the filtered signal. The signal s(t) is decomposed into in-phase and quadrature components. The in-phase component s1(t) of s(t) is given by:8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{gathered} {s_1}(t)=s(t)*\cos (2\pi{f_1}t) \\ =A_{1}\cos(2{\pi}f_{1}t+\varphi_{1})*\cos (2\pi{f_1}t)+A_{0}\cos(2{\pi}f_{0}t+\varphi_{0})*\cos (2\pi{f_1}t) \\ \end{gathered}$$\end{document} The quadrature component s2(t) of s(t) is given by:9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{gathered} {s_2}(t)=s(t)*s{\mathrm{in}}(2\pi{f_1}t) \\ =A_{1}\cos(2{\pi}f_{1}t+\varphi_{1})*\sin(2\pi{f_1}t)+A_{0}\cos(2{\pi}f_{0}t+\varphi_{0})*\sin (2\pi{f_1}t)\\ \end{gathered}$$\end{document} In the FH-DPSK method, the switching ripple at frequency f0 and f1 is orthogonal within the demodulation window, thereby mitigating the interference caused by non-communicating BEUs.10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\int_{0}^{{{T_{\mathrm{g}}}}} {{A_0}\cos (2\pi{f_0}t+{\varphi _0})*\cos (2\pi{f_1}t){\mathrm{d}}t} =0$$\end{document}11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\int_{0}^{{{T_{\mathrm{g}}}}} {{A_0}\cos (2\pi{f_0}t+{\varphi _0})*\sin (2\pi{f_1}t){\mathrm{d}}t} =0$$\end{document} This condition holds if the window duration Tg is selected as a common multiple of 1/f0 and 1/f1. Integrating s1(t) and s2(t) over one demodulation window yields:12\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\int_{0}^{{{T_{\mathrm{g}}}}} {{s_1}} (t){\mathrm{d}}t=\int_{0}^{{{T_{\mathrm{g}}}}} {\frac{{{A_1}}}{2}} \cos {\varphi _1}dt=\frac{{{A_1}{T_g}}}{2}{\mathrm{cos}}{\varphi _1}$$\end{document}13\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\int_{0}^{{{T_{\mathrm{g}}}}} {{s_2}} (t){\mathrm{d}}t=\int_{0}^{{{T_{\mathrm{g}}}}} {( - \frac{{{A_1}}}{2}{\mathrm{sin}}{\varphi _1})} dt= - \frac{{{A_1}{T_{\mathrm{g}}}}}{2}{\mathrm{sin}}{\varphi _1}$$\end{document} Thus, the amplitude A1 of the frequency component at f1 in the filtered signal s(t) satisfies:14\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$A_{1}^{2}=\frac{4}{{T_{{\mathrm{g}}}^{2}}}\left[ {{{\left( {\int_{0}^{{{T_{\mathrm{g}}}}} {{s_1}} (t){\mathrm{d}}t} \right)}^2}+{{\left( {\int_{0}^{{{T_{\mathrm{g}}}}} {{s_2}} (t){\mathrm{d}}t} \right)}^2}} \right]$$\end{document} The phase φ1 of the frequency component at f1 in the filtered signal s(t) satisfies:15\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathrm{tan}}{\varphi _1}= - \frac{{\int_{0}^{{{T_{\mathrm{g}}}}} {{s_2}} (t){\mathrm{d}}t}}{{\int_{0}^{{{T_{\mathrm{g}}}}} {{s_1}} (t){\mathrm{d}}t}}$$\end{document} Equation (12) to (15) provide the method for determining the amplitude A1 and phase φ1 of the filtered signal s(t) at the communication frequency f1. Based on this, the demodulation process for FH-DPSK is illustrated in Fig. 11. Fig. 11Demodulation flow of FH-DPSK. (taking N = 10 as an example) Demodulation flow of FH-DPSK. (taking N = 10 as an example) The receiver samples the filtered signal value s(th) at the current time instant th. The sliding demodulation window is defined over the interval [th−Tg, th]. By applying the demodulation algorithm to the filtered signal s(t) within this window, the real-time amplitude and phase of s(t) at the communication frequency f1 can be extracted. The specific data demodulation sequence is divided into the following four operational stages. Stage I (Idle and Detection): When no BEU is transmitting, the amplitude A1 of the signal at frequency f1 remains below a preset threshold Am. Consequently, the receiver stays in the idle state. Once any BEU starts transmission, the amplitude A1 continuously increases until time t1, when it reaches the threshold Am. The receiver identifies a valid transmission start event. Stage II (Reference Phase Extraction): Following detection, the demodulation window aligns to the interval [t2 ~ t2+Tg] as shown in Fig. 11. The algorithm calculates the amplitude and phase of the signal at f1. This initial phase value is stored as the reference phase for the subsequent differential decoding. Stage III (Data Decoding): The demodulation window slides forward by a step length equal to the symbol duration TB. The algorithm processes the window [t2 + TB, t2 + TB+Tg] to extract the new phase. The receiver calculates the phase difference Δφ relative to the previous window. A Δφ approaching 0 is decoded as data “0”, while a Δφ approaching π is decoded as data “1”. This process repeats for every TB shift to recover the continuous data stream. Stage IV (Termination): After the transmitter ends transmission, at time t4, the amplitude A1 falls below Am. At this point, the receiver concludes that the transmission has ended and returns to Stage I. Construction and analysis of distributed battery equalization system Analysis of channel and noise The proposed distributed BES is composed of one bidirectional power source and multiple BEUs, with each BEU connected to the respective battery module. Based on the BEU structure shown in Fig. 5, its communication equivalent model when operating as the transmitter is illustrated in Fig. 12(a). Fig. 12Communication equivalent model of transmitter and receiver. (a) Transmitter. (b) Receiver. Communication equivalent model of transmitter and receiver. (a) Transmitter. (b) Receiver. The signal strength transmitted by the transmitter is determined by the primary-side current of its transformer. Therefore, the main power circuit can be modeled as an input capacitor Cin in parallel with a current source isignal whose magnitude equals the transmitter’s primary-side current. In Fig. 12(a), Zde denotes the input impedance of the demodulation module, and inoise represents the interference introduced by the transmitter’s power control. When the BEU operates as receiver, it can be modeled as an impedance Zin in parallel with an input current source iin, as shown in Fig. 12(b). To determine Zin, the converter’s switching states are analyzed as follows. Fig. 13Approximate input impedance of the receiver with primary-side switch on. Approximate input impedance of the receiver with primary-side switch on. When the receiver’s primary-side switch Q1 is turned on, the transformer secondary side is effectively open-circuited, as shown in Fig. 13(a). Neglecting the on-resistance of Q1, the circuit simplifies to the parallel combination of Cin and the magnetizing inductance Lm, as depicted in Fig. 13(b). At the carrier frequency f1, the capacitive reactance of Cin is significantly lower than the inductive reactance of Lm. Consequently, the inductive branch can be neglected, as depicted in Fig. 13(c). The input impedance of the receiver’s main power circuit under this condition is approximately:16\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Z_{\mathrm{in\_on}} \approx \frac{1}{sC_{\mathrm{in}}} \parallel sL_{\mathrm{m}} \approx \frac{1}{sC_{\mathrm{in}}}$$\end{document} When Q1 is turned off and Q2 is turned on, the primary side is open-circuited. Assuming the infinite off-resistance of Q1, the input impedance Zin−off is determined by only Cin:17\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Z_{\mathrm{in\_off}} \approx \frac{1}{sC_{\mathrm{in}}}$$\end{document} Based on Eqs. (16) and (17), the receiver’s input impedance Zin can be approximated as the capacitive reactance of Cin near the communication frequency f1. Fig. 14Communication channel model of the BES. Communication channel model of the BES. For a distributed BES containing M BEUs (one transmitter and M-1 receivers), the communication channel model is illustrated in Fig. 14. In this model, subscripts j denotes the parameters (iin, inoise, Cin, and Zde) of the jth receiver. igau represents the inherent channel background noise, i.e., white noise, and inoise represents the switching noise contributed by each BEU, respectively. According to this communication channel model, the total equivalent impedance Zsum of the communication channel is given by:18\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Z_{{\mathrm{sum}}}}=\frac{1}{{\sum\limits_{{j=1}}^{M} {s{C_{{\mathrm{in}}j}}} +\sum\limits_{{j=1}}^{M} {{{\frac{1}{Z}}_{{\mathrm{del}}}}} }}$$\end{document} The signal usum received by the receivers is:19\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${u_{{\mathrm{sum}}}}=\left( {{i_{{\mathrm{signal}}}}+\sum\limits_{{j=1}}^{M} {{i_{{\mathrm{noise}}j}}} +\sum\limits_{{j=2}}^{M} {{i_{{\mathrm{in}}j}}} +{i_{{\mathrm{gau}}}}} \right){Z_{{\mathrm{sum}}}}=\frac{{{i_{{\mathrm{signal}}}}+\sum\limits_{{j=1}}^{M} {{i_{{\mathrm{noise}}j}}} +\sum\limits_{{j=2}}^{M} {{i_{{\mathrm{in}}j}}} +{i_{{\mathrm{gau}}}}}}{{\sum\limits_{{j=1}}^{M} {s{C_{{\mathrm{in}}j}}} +\sum\limits_{{j=1}}^{M} {\frac{1}{{{Z_{{\mathrm{de}}j}}}}} }}$$\end{document} usum primarily consists of the communication signal usignal from the transmitter and the noise unoise in channel. The communication signal usignal is expressed as:20\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${u_{{\mathrm{signal}}}}={i_{{\mathrm{signal}}}}{Z_{{\mathrm{sum}}}}=\frac{{{i_{{\mathrm{signal}}}}}}{{\sum\limits_{{j=1}}^{M} {s{C_{{\mathrm{in}}j}}} +\sum\limits_{{j=1}}^{M} {\frac{1}{{{Z_{{\mathrm{de}}j}}}}} }}$$\end{document} The channel noise signal unoise is expressed as:21\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${u_{{\mathrm{noise}}}}=(\sum\limits_{{j=1}}^{M} {{i_{{\mathrm{noise}}j}}} +\sum\limits_{{j=2}}^{M} {{i_{{\mathrm{in}}j}}} +{i_{{\mathrm{gau}}}}){Z_{{\mathrm{sum}}}}=\frac{{\sum\limits_{{j=1}}^{M} {{i_{{\mathrm{noise}}j}}} +\sum\limits_{{j=2}}^{M} {{i_{{\mathrm{in}}j}}} +{i_{{\mathrm{gau}}}}}}{{\sum\limits_{{j=1}}^{M} {s{C_{{\mathrm{in}}j}}} +\sum\limits_{{j=1}}^{M} {\frac{1}{{{Z_{{\mathrm{de}}j}}}}} }}$$\end{document} The system noise primarily includes the fundamental switching ripple and its higher-order harmonics, high-frequency switching transients, low-frequency control jitter, and background white noise. Since the communication frequency employed in this work is 83.3 kHz, radiated noise in the BES can be neglected, and only conducted noise is considered. The switching ripple generated by non-transmitting BEUs constitutes the dominant noise source, classified as power noise. The noise consists of a fundamental component at frequency f0 and its harmonics at kf0. As derived in Sect. 2.3, by selecting the demodulation window length Tg that satisfies the orthogonality condition, the impact of the component at f0 and kf0 can be eliminated. Noise induced by switching transients refers to oscillations caused by parasitic parameters during switching transitions. The frequency of this noise fds is significantly higher than the communication frequency and can be effectively suppressed by a low-pass filter. Steady-state jitter stems from limited controller resolution or limit cycles in digital control loop. For converters employing PI/PID control, the frequency of such noise is typically below half the switching frequency. The cutoff frequencies of filters should satisfy:22\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{{{f_1}}}{2}{\mathrm{<}}{f_{{\mathrm{HPF}}}}{\mathrm{<}}{f_1}{\mathrm{<}}{f_{{\mathrm{LPF}}}}{\mathrm{<}}{f_{{\mathrm{ds}}}}$$\end{document} where fHPF represents the cutoff frequency of high-pass filter and fLPF represents that of low-pass filter. The system’s background noise is modeled as Additive White Gaussian Noise from the thermal noise of resistive elements and semiconductor devices. The amplitude probability of this noise follows a Gaussian distribution. The power spectral density Pn(f) (W/Hz) of Gaussian white noise is:23\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\mathrm{P}}_{\mathrm{n}}}(f)={n_0}(0f_{\mathrm{LPF}} \end{cases}$$\end{document} In summary, by selecting the suitable demodulation window length Tg, the interference from switching ripple at f0 and its higher-order harmonics can be eliminated. Furthermore, high-frequency parasitic ringing, low-frequency control jitter, and out-of-band white noise can be substantially suppressed by the hardware band-pass filters. Communication protocol of battery equalization system To ensure robust data transmission among BEUs, a specific data frame structure is defined, as depicted in Fig. 15. Fig. 15Frame format used for communication. Frame format used for communication. The length and meaning of each frame segment are listed in Table 2. The standard data frame consists of four parts: start-of-frame (SOF), Transmitter Address, Data Payload, and Frame Check Sequence (FCS). Each frame has a fixed length of 5 bytes, with each byte containing eight binary symbols. Due to the shared bus architecture, the frame transmitted by any BEU is broadcast to all BEUs on the DC bus. However, if FCS check was not passed, the frame will be discarded. Table 2The length and meaning of each part of the data frame.PartLength(Byte)MeaningSOF1Start of the frame.Transmitter Address1Address of transmitter BEU.Data2SOC of the locally managed battery module.FCS1Checksum for data integrity verification by the receiver. Table 2The length and meaning of each part of the data frame.PartLength(Byte)MeaningSOF1Start of the frame.Transmitter Address1Address of transmitter BEU.Data2SOC of the locally managed battery module.FCS1Checksum for data integrity verification by the receiver. The length and meaning of each part of the data frame. Since all BEUs share the common DC bus and use the same switching frequency f1 when acting as transmitters, a collision avoidance mechanism is essential. A practical communication protocol is adopted. Taking a distributed BES with three BEUs as an example, the timing sequence is illustrated in Fig. 16b. One complete communication cycle, with a duration Tc, is divided into three time slots labeled Tc1, Tc2, and Tc3 for three BEUs to transmit. Fig. 16Communication process for the BES. (a) Numbering of BEUs, (b) Composition of communication time slots. Communication process for the BES. (a) Numbering of BEUs, (b) Composition of communication time slots. Time slot Tc1 is divided into two stages. During Tc11, BEU #1 broadcasts a data frame containing the local information, which is received by BEUs #2 and #3. The Tc12 is an idle period following the transmission to prevent signal overlap. Similarly, Time slots Tc2,Tc3 consist of the transmitting stages Tc21,Tc31 and the idle stages Tc22,Tc32. Upon completion of the final slot Tc3, the cycle restarts from the Tc1 slot. Generally, for a system containing M BEUs, the duration of one communication cycle is expressed as:25\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${T_{\mathrm{c}}}=k{T_{\mathrm{B}}}M+\sum\limits_{{j=1}}^{M} {{T_{{\mathrm{c}}j2}}}$$\end{document} where k denotes the number of symbols in a frame and TB is the symbol duration. The primary data payload exchanged is the SOC of the battery modules. SOC is defined as the ratio of the remaining capacity to the rated capacity:26\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$SOC(t)=\frac{{Q(t)}}{{{Q_0}}} \times 100\%$$\end{document} where SOC(t) is the SOC at time t, Q(t) is the remaining capacity at time t, and Q0 is the rated capacity. To ensure estimation accuracy, this work employs a hybrid approach combining the Open-Circuit Voltage (OCV) method and the coulomb counting method for SOC estimation:27\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$SOC(t)=SOC(0)+\frac{1}{{{Q_0}}}\int_{0}^{t} {\eta {i_{\mathrm{o}}}} \operatorname{d} t$$\end{document} where SOC(0) is the initial SOC according to the OCV curve, η represents the coulombic efficiency, and i(τ) is the battery module’s equalization current, defined as positive during charging. Based on Eq. (27), each BEU can calculate the current SOC of its own battery module. By acquiring the SOC of other battery modules via communication, the equalization current for each unit is determined according to the SOC of all battery modules within the system. Experimental verification To validate the proposed strategy, a distributed BES prototype consisting of three BEUs was constructed, as shown in Fig. 17. Each BEU employs the talkative bidirectional flyback topology and manages a lithium-ion battery module. The key parameters of the prototype system are provided in Table 3. In the experimental setup, high-precision DC sources (ITECH IT6432/IT6433) are employed as the bidirectional power source and battery simulators, and an oscilloscope (Keysight DSO-X 3024T) is used for waveform acquisition. Each BEU is controlled by an MCU (STM32G431), in which a closed-loop PI control strategy is adopted to regulate the output current, enabling precise management of the equalization current for each battery module. Table 3The key parameters of the prototype system.ParametersValueBus voltage12 VRated charge/discharge current of each battery module0.8 ACapacity of each lithium-ion battery module20000mAhVoltage of each lithium-ion battery module12 VTransformer’s primary side magnetizing inductance100µHTransformer’s secondary side magnetizing inductance100µHTransformer turns ratio1:1Communication rate8.33 kb/sMOSFETsInfineon IRFR4620TRLPBF Table 3The key parameters of the prototype system.ParametersValueBus voltage12 VRated charge/discharge current of each battery module0.8 ACapacity of each lithium-ion battery module20000mAhVoltage of each lithium-ion battery module12 VTransformer’s primary side magnetizing inductance100µHTransformer’s secondary side magnetizing inductance100µHTransformer turns ratio1:1Communication rate8.33 kb/sMOSFETsInfineon IRFR4620TRLPBF The key parameters of the prototype system. Fig. 17Prototype of the distributed BES. (a) Overall experimental setup of the prototype system. (b) Bidirectional flyback BEU board. Prototype of the distributed BES. (a) Overall experimental setup of the prototype system. (b) Bidirectional flyback BEU board. In the constructed system, the switching frequency f0 during non-communication state was set to 100 kHz, while 83.3 kHz was adopted as the communication frequency f1. The selection of f0 and f1 is governed by (10) and (11) to ensure orthogonality between the two frequencies within the demodulation window. A communication rate of 8.33 kbps is achieved in the proposed system. While this bandwidth is sufficient for SOC equalization tasks, it may be perceived as relatively low for data-intensive monitoring applications, such as high-frequency cell voltage logging or detailed thermal profiling. Three experiments were conducted to evaluate the system performance. First, the feasibility of FH-DPSK communication was verified. BEU #1 continuously transmitted random data frames, while BEU #2 performed demodulation. The waveforms corresponding to a single transmitted frame are shown in Fig. 18. CH1 represents the DC bus voltage sampled and filtered before entering the microcontroller unit (MCU) of BEU #2, containing the information from the transmitter. CH2 shows the digital data stream transmitted by BEU #1, and CH3 depicts the demodulated signal at BEU #2. CH4 presents the Fast Fourier Transform (FFT) spectrum of the sampled CH1 signal. As shown in the spectrum, distinct frequency components at 83.3 kHz and 100 kHz are clearly observable, confirming the validity of the modulation principle. Fig. 18Test waveform when BEU #1 transmitting random data frame. (a) Long time scale. (b) Short time scale. Test waveform when BEU #1 transmitting random data frame. (a) Long time scale. (b) Short time scale. To evaluate the power quality of the shared DC bus during simultaneous power and data transmission, the total root-mean-square (RMS) value of the AC ripple was analyzed. The measurement results indicate that the total RMS value of the AC component on the bus voltage is 0.123 V, which accounts for 1.02% of the steady-state DC voltage. This moderate ripple level represents an optimal trade-off: it ensures that the communication signal maintains a sufficient amplitude for reliable data demodulation, while simultaneously guaranteeing that the overall power quality of the DC bus is not compromised. Fig. 19Communication waveform in larger system. Communication waveform in larger system. Furthermore, to validate the scalability of the proposed system, the experimental setup was expanded to include five BEUs. Figure 19 illustrates the waveforms during data transmission from BEU1 to BEU2 within the expanded system. CH1 represents the DC bus voltage sampled and filtered before entering the MCU of BEU #2, containing the information from the transmitter. CH2 shows the digital data stream transmitted by BEU #1, and CH3 depicts the demodulated signal at BEU #2. CH4 presents the FFT spectrum of the sampled CH1 signal. It is observed that although the ripple amplitude of the communication frequency component undergoes certain attenuation due to the increased number of parallel nodes, the demodulation results remain accurate. This further demonstrates the robustness and feasibility of the proposed architecture under expanded-scale conditions. Second, the communication reliability was evaluated. While BEU #1 transmitted random data frames, the amplitude and phase of the demodulated signals at BEU #2 were recorded. The resulting constellation diagram is shown in Fig. 20. The points are primarily clustered into two distinct regions, corresponding to symbols “0” and “1”. All demodulated points lie close to their theoretical phase values with substantial margin, and the clusters become tighter as the signal strength increases, demonstrating reliable demodulation performance. Fig. 20Demodulation result of BEU #2. (a) Low signal strength. (b) High signal strength. Demodulation result of BEU #2. (a) Low signal strength. (b) High signal strength. To quantitatively evaluate the communication reliability, a Bit Error Rate (BER) test was conducted. During the continuous operation of the BES, a total of 106 bits were transmitted between the BEUs. The measured BER was approximately 5*10− 6. This error rate experimentally verifies the effectiveness of the proposed FH-DPSK modulation scheme in a practical power conversion environment. Furthermore, signal attenuation in larger systems was emulated by adding parallel capacitors to the DC bus, confirming system stability under low signal strength conditions. Third, the equalization performance was evaluated using three BEUs (#1, #2, and #3), with the initial SOCs of their respective battery modules set to 50%, 60%, and 70%. The equalization current was regulated at ± 0.8 A. Following the designed communication protocol, the three BEUs exchanged information and performed energy balancing on their corresponding battery modules. At a specific moment during communication, BEU #1 broadcast a data frame containing the current SOC of its managed battery module. Both BEUs #2 and #3 received this frame. In Fig. 21(a), CH1 is sampled and filtered DC bus voltage inputting into the MCU of BEU #2. CH2 shows the data frame transmitted by BEU #1. Its structure is as follows: the first byte, 0 × 55, is the SOF; the second byte, 0 × 00, represents the transmitter address as BEU #1; the third and fourth bytes, 0 × 138 F, represent the SOC of BEU #1’s battery module as 50.07%; and the fifth byte, 0 × 0 C, represents the checksum byte. CH3 displays the data successfully demodulated by BEU #2. CH4 indicates the equalization current of BEU #2. After acquiring the SOC information from BEU #1, BEU #2 determined that the direction of its equalization current needed to be reversed. Consequently, the load current of BEU #2 was changed from − 0.8 A to 0.8 A under closed-loop control. Fig. 21Waveform during communication. (a) BEU #1 transmit and BEU #2 receive. (b) BEU #2 transmit and BEU #3 receive. Waveform during communication. (a) BEU #1 transmit and BEU #2 receive. (b) BEU #2 transmit and BEU #3 receive. Following the completion of data transmission by BEU #1 and after an idle period, the current time slot ended. In the subsequent time slot, data was transmitted by BEU #2 and received by BEUs #1 and #3. In Fig. 21(b), CH1 is sampled and filtered DC bus voltage inputting into the MCU of BEU #3. CH2 displays the data frame transmitted by BEU #2. Its structure is as follows: the first byte, 0 × 55, is the SOF; the second byte, 0 × 01, represents the transmitter address as BEU #2; the third and fourth bytes, 0 × 1770, represent the SOC of BEU #2’s battery module as 60.00%; and the fifth byte, 0 × 0 C, is the checksum byte. CH3 shows the data successfully demodulated by BEU #3. CH4 indicates the equalization current of BEU #3. After acquiring the SOC information of the battery module managed by BEU #2, BEU #3 determined that the direction of its equalization current needed to be reversed. Consequently, the load current of BEU #3 changed from 0.8 A to -0.8 A. The duration of one communication cycle can be calculated from Eq. (25). In this study, with k = 40, TB=120 us, and Tcj=20 us, the communication cycle Tc for the three-unit system is 0.074 s. Even if the system is expanded to 20 units, Tc remains as low as 0.496 s. Given the slow-varying nature of SOC, this sub-second latency ensures that the exchanged data accurately reflects the real-time status of the battery modules, meaning the communication delay has no perceptible impact on system coordination. During the battery equalization process, the SOC data from the battery simulator was used as a benchmark to perform an error analysis of the real-time SOC estimated by each BEU. The results indicate that, excluding the effects of battery capacity variations, the estimation error of our method is less than 2%. This confirms the effectiveness and accuracy of the SOC estimation strategy employed in this work. Fig. 22Real-time SOC of battery modules during equalization. (a) 3-unit prototype. (b) 10-unit simulation. Real-time SOC of battery modules during equalization. (a) 3-unit prototype. (b) 10-unit simulation. The long-term equalization results are plotted in Fig. 22a. The curves demonstrate that the SOCs of the three battery modules converge gradually. After approximately 150 min of operation, the inconsistencies are eliminated, and all modules reach a balanced state. This confirms the effectiveness of the system’s communication and equalization functionality. To verify system feasibility at large scale, a 10-unit simulation was conducted, with the results plotted in Fig. 22b. Battery module capacity and idle stage duration Tcj were appropriately modified to enhance simulation efficiency. Conclusion This paper presents a communication-integrated BES leveraging TPC technology. By implementing bidirectional flyback converters as TPCs, the proposed architecture enables simultaneous power transfer and data transmission without the need for dedicated communication cabling or separate circuitry. This integration substantially reduces hardware complexity and manufacturing costs, while preserving electrical isolation between battery modules. Building on these foundations, this paper introduces several innovations. Firstly, to ensure robust communication in a distributed environment, an FH-DPSK modulation scheme is employed. This approach provides strong immunity to switching ripple interference from parallel, non-communicating converters and enables non-coherent demodulation, eliminating the need for precise phase synchronization at the receiver. Secondly, a comprehensive channel model was also developed to characterize noise behavior and guide effective interference suppression strategies. In addition, a communication protocol was designed to coordinate orderly data exchange among the BEUs. Finally, a distributed BES prototype comprising three battery modules was constructed to validate the proposed strategy. Experimental results demonstrate real-time data transmission at a bit rate of 8.33 kb/s and an equalization power of 10 W per unit. While the achieved transmission rate is well-suited for monitoring slow-varying SOC during the equalization process, it remains a limitation for more demanding high-speed diagnostic tasks, representing a clear objective for future performance enhancements. The successful balancing of SOC among the battery modules confirms the effectiveness and practicality of the proposed approach for next-generation battery management systems.