A self-adapting power management and power supply circuit system for composite broadband micro-vibration energy harvesting

Abstract

The application discloses a kind of composite broadband microvibration energy acquisition adaptive power management and power supply circuit system, comprising: transducer signal separation acquisition module, makes piezoelectric and electromagnetic two-way signal from source physical isolation;First, second impedance matching rectifier module, extract band energy characteristics from respective rectification output ripple, dynamically adjust equivalent input impedance in feedforward mode;Power fusion module, by controlled MOSFET realizes the one-way low-loss merging of two-way direct current power;Hybrid energy storage and charging path scheduling module, according to bus power level automatically switches between sequencing charging state;Transient power distribution and load state detection module, to pure hardware timing distinguishes network matching load current transient characteristics, realizes power supply trigger and shutdown.The application makes impedance matching response and vibration spectrum change synchronous, avoids double-source cross interference and energy storage boundary oscillation, is independent of digital protocol, realizes the autonomous perception and precision power distribution of load task.

Description

Technical Field

[0001] This invention relates to the technical field of a circuit device or system for power supply or distribution, and more particularly to an adaptive power management and power supply circuit system for composite broadband micro-vibration energy harvesting. Background Technology

[0002] Predictive maintenance of industrial robot joint reducers is a key aspect of achieving intelligent manufacturing. Its foundation lies in long-term online monitoring of various state parameters within the reducer, such as temperature and vibration. Due to the extremely confined space within the joint and the continuous rotation and pitching motion, wired power supply faces risks of cable entanglement and wear, while battery power supply presents challenges such as replacement difficulties and potential environmental pollution. Harvesting energy from the broadband micro-vibrations generated by the reducer's own operation to provide battery-free self-powered wireless sensor nodes embedded in the joint has become a promising technological approach for engineering applications. The vibrations in these scenarios are formed by the superposition of multiple excitation sources, including gear meshing, bearing rolling, and flexible deformation, with a frequency spectrum spanning from tens to thousands of hertz, and typical acceleration amplitudes ranging from 0.1g to 2g. Due to the real-time changes in robot joint posture, load, and rotation speed, the vibration spectrum exhibits significant non-steady-state and multi-peak characteristics on a time scale of seconds. Limited by small amplitude and electromechanical energy conversion efficiency, the average power that can be extracted is only on the order of hundreds of microwatts. At this power level, the power supply system must have extremely low self-loss characteristics and be able to adapt to the high-speed time-varying spectrum in order to support the intermittent high power consumption requirements of sensing acquisition and wireless radio frequency transmission.

[0003] In existing technologies, power management solutions for vibration energy harvesting mainly fall into two categories: general-purpose energy harvesting integrated circuits and single-frequency resonant harvesting power supply devices for rotating machinery. General-purpose energy harvesting integrated circuits are designed based on quasi-steady-state vibration, resulting in a large maximum power point tracking loop response time constant, making it difficult to track rapid frequency shifts, leading to impedance mismatch and a sharp decline in energy capture efficiency. Single-frequency resonant harvesting devices achieve higher conversion efficiency by aligning the resonant frequency of a piezoelectric cantilever beam with a known fixed dominant frequency. However, when faced with broadband, multi-peak vibrations with continuously drifting dominant frequencies, they can only capture... The energy obtained is within an extremely narrow frequency band, and most of the usable vibration energy is dissipated. The above-mentioned schemes usually adopt a single-channel input structure and do not differentiate between the high voltage and high impedance characteristics of piezoelectric transducers and the low voltage and low impedance characteristics of electromagnetic transducers. If multiple sources are connected in parallel, it will cause energy loss. On the energy storage side, existing hybrid energy storage management mostly adopts voltage threshold floating charging logic, which lacks adaptability to the microwatt-level intermittent energy harvesting rhythm. On the load side, the scheme that relies on microcontrollers for voltage monitoring and enable scheduling will have milliwatt-level standby power consumption that will offset most of the collected energy in micro-energy harvesting scenarios. Summary of the Invention

[0004] This invention overcomes the shortcomings of the prior art and provides an adaptive power management and power supply circuit system for composite broadband micro-vibration energy harvesting.

[0005] To achieve the above objectives, the technical solution adopted by this invention is: an adaptive power supply circuit system for composite broadband micro-vibration energy harvesting, comprising: The transducer signal separation and acquisition module is used to acquire the high-voltage AC signal generated by the piezoelectric transducer in response to broadband micro-vibration and the low-voltage AC signal generated by the electromagnetic transducer in response to broadband micro-vibration, respectively. The first impedance matching rectifier module is connected to the transducer signal separation and acquisition module. It is used to receive high-voltage AC signals, extract the first ripple characteristics from the voltage waveform after active rectification, dynamically adjust the first equivalent input impedance in a feedforward manner, and output the first DC power. The second impedance matching rectifier module is connected to the transducer signal separation and acquisition module. It is used to receive low-voltage AC signals, extract the second ripple characteristics from the voltage waveform after active rectification, dynamically adjust the second equivalent input impedance in a feedforward manner, and output the second DC power. The power fusion module is connected to the output terminals of the first impedance matching rectifier module and the second impedance matching rectifier module respectively. It is used to combine the first DC power and the second DC power into a common energy storage bus in a unidirectional low-loss manner and block the reverse current that may be generated at either output terminal. The hybrid energy storage and charging path scheduling module is connected to the common energy storage bus and includes a supercapacitor unit, a battery unit and a charging sequencer. The charging sequencer is used to automatically switch between multiple preset sequential charging states according to the input power level of the common energy storage bus in order to control the differentiated energy buffer storage of the supercapacitor unit and the battery unit. The transient power distribution and load status detection module is connected to the hybrid energy storage and charging path scheduling module and a load, respectively. It enables transient power supply to the load when the energy storage status of the supercapacitor unit exceeds the power supply threshold and the transient characteristics of the load current obtained through non-intrusive load status sensing meet the preset task readiness criteria; and shuts off the power supply path to the load when the transient characteristics of the load current meet the preset task completion criteria during the power supply maintenance period.

[0006] In a preferred embodiment of the present invention, the first impedance matching rectifier module and the second impedance matching rectifier module each include: An active full-bridge rectifier uses controlled semiconductor switching devices to convert the polarity of AC signals; The ripple feature extraction unit includes a peak detector and a transconductance amplifier. The peak detector tracks the upper envelope waveform of the rectified output voltage, and the transconductance amplifier converts the envelope voltage into a control current. A variable impedance network is formed by a MOSFET operating in the linear region connected in parallel with a fixed resistor. The control current changes the gate-source bias voltage of the MOSFET to adjust the equivalent parallel impedance. The variable impedance network adjustment range of the first impedance matching rectifier module is adapted to the source internal resistance range of the piezoelectric transducer, and the variable impedance network adjustment range of the second impedance matching rectifier module is adapted to the source internal resistance range of the electromagnetic transducer.

[0007] In a preferred embodiment of the present invention, the power fusion module includes two gate drive comparators and two low on-resistance MOSFETs driven by them respectively. Each gate drive comparator includes a pair of voltage comparators. The non-inverting input terminal of the voltage comparator is connected to the anode, the inverting input terminal is connected to the cathode, and the output terminal is connected to the gate of the corresponding MOSFET. When the anode voltage is higher than the cathode voltage and the difference exceeds the preset bias voltage, the MOSFET is turned on; when the anode voltage is lower than the cathode voltage, the MOSFET is turned off to block the reverse current from the common energy storage bus to the corresponding impedance-matched rectifier module.

[0008] In a preferred embodiment of the present invention, the charging sequencer includes a voltage window detector and a switch array; The voltage window detector includes multiple hysteresis comparators to define at least three power ranges: a low-energy range below a first threshold, a high-energy range above a second threshold, and a saturation range where the supercapacitor voltage reaches a preset saturation upper limit. The switch array switches between the following sequential charging states based on the output of the voltage window detector: charging only the supercapacitor unit, charging both the supercapacitor unit and the battery unit simultaneously, and constant current charging from the supercapacitor unit to the battery unit.

[0009] This invention provides an adaptive power management method for composite broadband micro-vibration energy harvesting, comprising: S1. Acquire the high-voltage AC signal generated by the piezoelectric transducer in response to broadband micro-vibration and the low-voltage AC signal generated by the electromagnetic transducer in response to broadband micro-vibration. S2. Actively rectify the high-voltage AC signal, extract the first ripple characteristic from the high-voltage AC signal, dynamically adjust the first equivalent input impedance, and output the first DC power; simultaneously, actively rectify the low-voltage AC signal, extract the second ripple characteristic from the low-voltage AC signal, dynamically adjust the second equivalent input impedance, and output the second DC power. S3. Combine the first DC power and the second DC power onto a common energy storage bus; S4. Receive energy from the common energy storage bus and automatically and without oscillation switch between sequential charging states based on the input power of the common energy storage bus to complete energy buffer storage; wherein, the sequential charging states include: when the bus power is lower than the first threshold, only the energy is stored in the supercapacitor; when the bus power is higher than the second threshold, the energy is stored in both the supercapacitor and the battery; when the supercapacitor voltage reaches the preset saturation upper limit, the supercapacitor switches to constant current charging to the battery; S5. Monitor the energy storage status of the supercapacitor in real time and continuously detect the transient current waveform on the load power supply circuit using a non-inductive sampling resistor; when the supercapacitor voltage exceeds the power supply threshold and the transient current waveform matches the task ready characteristic mode, trigger the voltage regulator to provide transient power supply to the load. S6. During the power supply maintenance period, continuously monitor the load current. When a match with the task completion characteristic pattern is detected, immediately shut off all power supply paths between the voltage regulator and the load, and restore the system to standby state.

[0010] In a preferred embodiment of the present invention, the extraction of the first ripple feature and the second ripple feature in step S2 respectively includes: Track the upper envelope waveform of the corresponding active rectified output voltage, and convert the voltage change of the upper envelope waveform into control current through transconductance amplification; The equivalent parallel impedance of the variable impedance network, which consists of a MOSFET and a fixed resistor connected in parallel, is changed by the control current to adjust the equivalent input impedance of the corresponding transducer.

[0011] In a preferred embodiment of the present invention, the method for merging the first DC power and the second DC power to a common energy storage bus in step S3 is as follows: The first DC power and the second DC power are respectively connected to the anodes of two MOSFETs driven by voltage comparators, and the cathodes of the two MOSFETs are connected to a common energy storage bus. When the anode voltage of either path is lower than the voltage of the common energy storage bus, the corresponding MOSFET is turned off, blocking the reverse current from the bus to that path.

[0012] In a preferred embodiment of the present invention, the switching between sequential charging states in step S4 is implemented based on a hysteresis comparator: The upper and lower thresholds of the first hysteresis comparator define the boundary between the low-energy region and the abundant region, and the upper and lower thresholds of the second hysteresis comparator define the entry and exit conditions of the saturation region. When the supercapacitor voltage reaches the preset saturation upper limit, the constant current charging path from the supercapacitor to the battery is closed, and the supercapacitor energy is transferred to the battery with a preset constant current until the supercapacitor voltage falls back to below the lower threshold of the saturation region.

[0013] In a preferred embodiment of the present invention, the task-ready feature pattern in step S5 includes at least a combination of the following temporal logic conditions: An initial surge current pulse with a rise time shorter than a preset value was detected within the first time window; A working current plateau was detected within the second time window, with a duration exceeding a preset value. Subsequently, the current was detected to have dropped to a level below the preset static maintenance threshold.

[0014] In a preferred embodiment of the present invention, the task completion feature pattern in step S6 includes at least a combination of the following temporal logic conditions: The load current is detected to drop below a preset low current threshold and remain below a preset time period. After the preset time period, a negative pulse group with a pulse width less than the preset value and an amplitude exceeding a multiple of the previous normal load current peak value is detected; After confirming the task completion characteristic mode, turn off the enable terminal of the voltage regulator and disconnect the switch between the voltage regulator output terminal and the load to completely electrically isolate the load from the circuit.

[0015] This invention addresses the shortcomings of the prior art and has the following beneficial effects: (1) This invention directly extracts the analog feedforward signal reflecting the energy distribution of the vibration frequency band from the rectified output ripple and injects it into their respective impedance matching control loops. This solves the tracking error problem caused by the limited response bandwidth of the traditional a posteriori feedback regulation based on output power detection when the vibration spectrum inside the joint reducer of an industrial robot changes abruptly within milliseconds. The feedforward mechanism shortens the response path of impedance regulation from a multi-stage delay link to the intrinsic response time of the ripple extraction circuit, so that the equivalent input impedance of the transducer can change synchronously with the optimal load point under the current vibration main frequency, eliminating the inherent hysteresis mismatch loss of the a posteriori regulation. Combined with the independent feedforward optimization architecture of the piezoelectric and electromagnetic channels, the two transducers can lock the maximum power point in their respective response frequency bands, so that the mid-high frequency components and low frequency components in the broadband vibration energy are efficiently captured.

[0016] (2) The present invention uses two low on-resistance MOSFETs driven by gate drive comparators to form a unidirectional low-loss power fusion network, which solves the problem of channel clamping or Schottky diode energy consumption when two DC power sources with different characteristics are connected in parallel; the network physically blocks reverse current loads, avoids cross interference, reduces conduction loss, ensures that the advantages of dual-source collaborative acquisition are converted into usable DC power, and prevents energy leakage or heat dissipation.

[0017] (3) This invention solves the problems of energy storage element life decay and excessive switching energy overhead caused by frequent oscillations at the intermittent micro-current boundary caused by traditional voltage threshold comparators through sequential charging scheduling logic based on hysteresis window grading and input power level; the charging sequencer automatically switches between three states according to the micro-energy region, abundant region and saturation region of the bus voltage: supercapacitor charging only, supercapacitor and battery parallel charging, and supercapacitor constant current charging to battery, so that the supercapacitor can quickly respond to fluctuating power, and the battery, as a long-term storage, only intervenes when energy is sufficient or the supercapacitor is full, which fits the mismatch between intermittent energy supply and pulse load. The hysteresis window provides hysteresis, eliminates boundary oscillation and switching power consumption, improves energy capture efficiency and extends the service life of energy storage elements. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Figure 1 This is a system overall architecture block diagram of a preferred embodiment of the present invention; Figure 2 This is a flowchart of the steps of a preferred embodiment of the present invention; Figure 3 This is a flowchart illustrating the internal working principle of the impedance matching rectifier module according to a preferred embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein. Therefore, the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0021] Application Overview: This application addresses the specific scenario of the internal structure of an industrial robot joint reducer, characterized by confined space, vibration energy exhibiting wide-band, multi-peak, low-amplitude characteristics, and rapidly time-varying spectrum, resulting in extremely limited power extraction. In existing technologies, the rectification and maximum power point tracking circuits employ a posteriori feedback control based on output power detection. The control loop passively corrects the equivalent input impedance by sensing DC current disturbances, adjusting for changes lagging behind the energy spectrum. When the vibration characteristic frequency undergoes abrupt changes, the impedance matching state deviates for a period, leading to a loss in extraction efficiency.

[0022] For a dual-source acquisition structure that uses both piezoelectric and electromagnetic transducers, if they are directly connected in parallel to the same rectifier bus without characteristic decoupling, the high voltage output of the piezoelectric transducer will clamp the output terminal of the electromagnetic transducer in a non-conductive state, and the energy it carries cannot be extracted, with some energy being consumed internally. In the power distribution stage, the method of relying on the digital handshake signal issued by the load for power supply triggering results in a tight coupling between the power supply system and a specific sensor model, losing the versatility of the basic energy platform.

[0023] The shortcomings of existing technologies lie in the fact that control and scheduling are based on the assumption of steady-state or quasi-steady-state changes in energy state. The maximum power point tracking loop is constrained by power sampling filters and stability conditions, limiting its response bandwidth to below 30Hz~90Hz. However, abrupt changes in joint vibration energy distribution can occur within milliseconds, and the tracking error generated by feedback regulation is a fundamental flaw that cannot be eliminated through parameter optimization. While connecting the piezoelectric and electromagnetic transducers to independent universal energy harvesting circuits can avoid direct clamping, their respective maximum power point tracking controls remain a posteriori and slow, still unable to match rapid spectrum changes. At the energy storage management level, charging switching driven by ordinary voltage threshold comparators is prone to frequent oscillations near the charge / discharge threshold when facing discontinuous energy flows, causing accelerated degradation of the energy storage element's cycle life, and the energy consumption of the switching process itself is too high. Furthermore, if a microcontroller is used to schedule the load power supply in a periodic wake-up manner, its sleep current and wake-up operation energy account for too large a proportion in the average power budget of 3μW~9μW, which will cause the net energy accumulation of the system to be negative, and ultimately it will be unable to maintain the intermittent operation of the sensing node.

[0024] This invention employs an adaptive power management and power supply circuit system that combines vibration energy frequency band feedforward prediction with hardware sensing of load current transient characteristics, without relying on a posteriori feedback adjustment or digital protocol dependence. It achieves predictive impedance matching by extracting analog feedforward signals reflecting the vibration frequency band energy distribution from the output ripple of two rectifiers and injecting them into their respective maximum power point tracking control loops. Simultaneously, it constructs a hardware-sequential three-state charging scheduling logic adapted to the micro-energy harvesting rhythm on the energy storage side. Furthermore, it introduces a purely hardware-implemented load current transient waveform classification mechanism at the power distribution end, automatically identifying the load task status in a digital protocol-independent manner to complete the timing control of sudden power supply and shutdown.

[0025] Exemplary method: like Figure 2 As shown, an adaptive power management method for composite broadband micro-vibration energy harvesting includes: S1. Acquire the high-voltage AC signal generated by the piezoelectric transducer in response to broadband micro-vibration and the low-voltage AC signal generated by the electromagnetic transducer in response to broadband micro-vibration. S2. Actively rectify the high-voltage AC signal, extract the first ripple characteristic from the high-voltage AC signal, dynamically adjust the first equivalent input impedance, and output the first DC power; simultaneously, actively rectify the low-voltage AC signal, extract the second ripple characteristic from the low-voltage AC signal, dynamically adjust the second equivalent input impedance, and output the second DC power. S3. Combine the first DC power and the second DC power onto a common energy storage bus; S4. Receive energy from the common energy storage bus and automatically and without oscillation switch between sequential charging states based on the input power of the common energy storage bus to complete energy buffer storage; wherein, the sequential charging states include: when the bus power is lower than the first threshold, only the energy is stored in the supercapacitor; when the bus power is higher than the second threshold, the energy is stored in both the supercapacitor and the battery; when the supercapacitor voltage reaches the preset saturation upper limit, the supercapacitor switches to constant current charging to the battery; S5. Monitor the energy storage status of the supercapacitor in real time and continuously detect the transient current waveform on the load power supply circuit using a non-inductive sampling resistor; when the supercapacitor voltage exceeds the power supply threshold and the transient current waveform matches the task ready characteristic mode, trigger the voltage regulator to provide transient power supply to the load. S6. During the power supply maintenance period, continuously monitor the load current. When a match with the task completion characteristic pattern is detected, immediately shut off all power supply paths between the voltage regulator and the load, and restore the system to standby state.

[0026] In order to transform this invention into a practically working circuit system, several problems need to be solved.

[0027] The frequency band energy characteristics upon which feedforward control relies must be extracted from the rectified ripple in real time and converted into a continuous impedance adjustment at the cost of low power consumption. The information carried in the ripple is essentially a weak analog disturbance superimposed on the DC bias, with both the frequency components and amplitude changing rapidly with the joint motion state. If the power consumption of the extraction circuit itself exceeds 30nA~90nA, or if the conversion delay from the appearance of the ripple to the completion of the impedance adjustment exceeds 5ms, then the feedforward prediction degenerates into hysteresis feedback, losing its advantage over traditional a posteriori control.

[0028] The piezoelectric and electromagnetic channels each perform impedance optimization based on independent feedforward signals. However, the two DC power sources eventually converge into the same common energy storage bus. When the voltage of one channel drops due to a sudden drop in vibration energy, if the fusion network cannot block the reverse current within nanoseconds, or if the voltage disturbance generated during the blocking process is reverse-coupled to the regulation loop of the other channel, it will trigger a cross-modulation effect, causing the dual-source independent optimization to degenerate into mutual interference. The overall extraction efficiency is actually lower than that of the single-source scheme.

[0029] Furthermore, the pure hardware perception of the load task status must simultaneously meet the dual requirements of universality and nanoampere power consumption. The current transient characteristics of different models of wireless sensor nodes have tolerance ranges in pulse width, amplitude and timing. If the characteristic logic sequence fixed by the timing discrimination network is set too narrow, it will lose the ability to cover the target load set. If it is set too wide, it will increase the risk of false triggering. In addition, an analog-to-digital converter or a high-precision clock source is introduced to improve the discrimination accuracy, but the power consumption of the discrimination circuit itself exceeds the energy budget.

[0030] like Figure 3 As shown, based on the above requirements, this invention takes S1 as the physical starting point of the entire process. By acquiring the high-voltage AC signal generated by the piezoelectric transducer in response to broadband micro-vibration and the low-voltage AC signal generated by the electromagnetic transducer in response to broadband micro-vibration, the two transducers with very different characteristics are physically separated from the signal source. This allows their complete time-frequency information to independently enter the subsequent feedforward extraction and impedance matching stages, thus building an independent signal path foundation for subsequent steps that is not affected by the clamping interference on the other side.

[0031] In step S1, the high-voltage AC signal refers to the time-varying voltage signal generated at the output terminal of the piezoelectric transducer when it is subjected to mechanical stress, based on the piezoelectric effect, which has a high peak open-circuit voltage and a large output impedance. Inside the joint reducer of an industrial robot, the vibration spectrum covers 30Hz~3000Hz. The piezoelectric transducer is more sensitive to the mid-to-high frequency vibration components of 500Hz~3000Hz. Its output signal can reach a peak open-circuit voltage of 3V~25V, but the short-circuit current that can be output is only 10μA~500μA, and the source internal resistance is usually in the range of 50kΩ~500kΩ.

[0032] Low-voltage AC signal refers to a time-varying voltage signal generated by an electromagnetic transducer based on the principle of electromagnetic induction, which has a low peak open-circuit voltage and low output impedance when the permanent magnet and the coil move relative to each other. The electromagnetic transducer is more sensitive to low-frequency vibration components of 30Hz to 200Hz, and its output signal's peak open-circuit voltage is usually in the range of 20mV to 300mV, and its source internal resistance is in the range of 5Ω to 200Ω.

[0033] High-voltage AC signals and low-voltage AC signals have fundamental differences in voltage amplitude, impedance level, and response frequency band, and belong to heterogeneous power supply outputs.

[0034] The reason for acquiring both signals simultaneously in S1 is that the vibration energy of the industrial robot joint reducer is not concentrated in a single frequency or a single frequency band, but is distributed across a wide frequency band consisting of the low-frequency joint rotation fundamental frequency, the mid-frequency gear meshing harmonics, and the high-frequency bearing characteristic frequency. Using only a single type of transducer, whether it is using only a piezoelectric transducer to discard low-frequency energy or using only an electromagnetic transducer to discard mid- and high-frequency energy, will result in a significant loss of the total extractable energy. Acquiring both signals simultaneously is a prerequisite for achieving wideband vibration energy coverage from a physical perspective.

[0035] In a specific embodiment, the output terminal of the piezoelectric transducer is connected to the AC input port of the first impedance matching rectifier module via a shielded cable, and the output terminal of the electromagnetic transducer is connected to the AC input port of the second impedance matching rectifier module via an independent shielded cable. The two signals are physically separated from the transducer output terminals, each having its own independent signal loop and not sharing any nodes. This separate acquisition method physically avoids the high output voltage of the piezoelectric transducer directly acting on the output terminal of the electromagnetic transducer, thus avoiding the mutual clamping phenomenon caused by voltage difference when the two types of transducers are directly connected in parallel.

[0036] In existing technologies, when a system uses both piezoelectric and electromagnetic transducers, their outputs are typically connected in parallel and then connected to the same rectifier circuit. Since the open-circuit voltage of the piezoelectric transducer is usually much higher than that of the electromagnetic transducer, the conduction condition of the rectifier bridge is determined by the side with the higher voltage. The output of the electromagnetic transducer is clamped at a voltage level lower than required for its own rectification and conduction most of the time, preventing the collected energy from being transferred to subsequent circuits. Instead, it is consumed internally as reactive circulating current. Alternatively, separate universal energy harvesting circuits can be configured for each transducer. While this solves the voltage clamping problem caused by direct parallel connection, it does not change the limitation that subsequent processing relies on a posteriori feedback control. Furthermore, it increases the circuit board area and static power consumption, making it unsuitable for applications with extremely limited internal space.

[0037] S1 physically isolates the two transducers at the initial signal acquisition stage, ensuring that the high-voltage and low-voltage AC signals retain their original amplitude, frequency, and phase information without being affected by the other side. This provides an undistorted signal source for subsequent independent ripple feature extraction and impedance matching adjustment in S2, tailored to the characteristics of each signal. Based on this, S2 rectifies and impedance-matches the two AC signals with different characteristics, converting unstable AC energy into DC power usable for subsequent energy storage and power distribution, while simultaneously addressing impedance mismatch caused by rapid changes in the vibration spectrum.

[0038] Step S2 performs active rectification on the high-voltage AC signal, extracts the first ripple characteristic from the high-voltage AC signal, dynamically adjusts the first equivalent input impedance, and outputs the first DC power; at the same time, it performs active rectification on the low-voltage AC signal, extracts the second ripple characteristic from the low-voltage AC signal, dynamically adjusts the second equivalent input impedance, and outputs the second DC power.

[0039] Active rectification refers to a circuit technology that uses controlled semiconductor switching devices to replace traditional passive diodes to convert the polarity of AC signals. An active rectifier detects the polarity of the AC input voltage and turns on the corresponding switching devices during the positive and negative half-cycles, converting a bipolar AC signal into a unidirectional DC signal. Compared to full-bridge rectification using diodes, the voltage drop of the switching devices in an active rectifier during conduction is less than the forward voltage drop of the PN junction of a diode, ranging from 20mV to 50mV, while the forward voltage drop of a Schottky diode is in the range of 200mV to 600mV. Therefore, under typical operating currents of 100μA to 500μA for low-energy harvesting, active rectifiers can avoid significant energy loss due to the rectification process.

[0040] The first ripple characteristic refers to the time-varying characteristic quantity extracted from the voltage waveform obtained after active rectification of the high-voltage AC signal, reflecting the frequency band energy distribution of the original vibration signal; it is composed of parameters such as the amplitude envelope, frequency components, and rate of change of the ripple component of the rectified output voltage; since the high-voltage AC signal directly originates from the electromechanical conversion of the joint reducer vibration by the piezoelectric transducer, its amplitude and frequency change in real time with the vibration state, and the above-mentioned change information is retained in the voltage ripple after rectification; the first equivalent input impedance refers to the equivalent parallel impedance value seen from the AC input port of the first impedance matching rectifier module into the circuit, which can be adjusted in the range of 50kΩ~500kΩ through a variable impedance network, and determines the actual load impedance connected to the output terminal of the piezoelectric transducer.

[0041] The second ripple characteristic refers to the time-varying characteristic quantity that reflects the frequency band energy distribution of the original vibration signal, extracted from the voltage waveform obtained after the low-voltage AC signal is actively rectified; the second equivalent input impedance refers to the equivalent parallel impedance value seen from the AC input port of the second impedance matching rectifier module into the circuit, which can also be adjusted in the range of 5Ω~200Ω through a variable impedance network; although the first ripple characteristic and the second ripple characteristic are extracted on the same principle, they correspond to the vibration frequency band and signal characteristics of their respective channels and are independent of each other.

[0042] In S2, ripple characteristics are extracted from both the high-voltage and low-voltage AC signals. The piezoelectric and electromagnetic transducers respond to vibrations in different frequency bands, and the frequency band energy distribution information carried by their output signals reflects the vibration state changes in the mid-to-high frequency and low-frequency bands, respectively. If features are extracted from only one signal while the other is adjusted using a fixed impedance or solely relying on output power feedback, it is impossible to ensure that both transducers operate independently near their maximum power points. Furthermore, the vibration spectrum of the joint reducer may experience dominant frequency drift and multi-peak energy redistribution on a second-scale timescale. If impedance matching adjustment lags behind these changes, energy extraction efficiency will be lost.

[0043] Therefore, S2 adopts a feedforward regulation mechanism based on ripple characteristics, which synchronously extracts the frequency band information carried by the signal itself during the rectification process and directly uses it for impedance adjustment, bypassing the delay path of the traditional scheme of first rectifying and storing energy, then detecting the output power, and then feeding back to adjust the impedance.

[0044] In a specific embodiment, S2 includes two parallel signal processing paths. The first path processes high-voltage AC signals, and the second path processes low-voltage AC signals. The two processing logics are symmetrical, but the parameter settings are determined independently based on the characteristics of their respective transducers.

[0045] In the first path, the high-voltage AC signal enters the active full-bridge rectifier of the first impedance matching rectifier module. The active full-bridge rectifier consists of four N-channel MOSFET switches and their gate drive logic. The drive logic includes a pair of polarity comparators and corresponding level shifting circuits. The polarity comparators continuously monitor the zero-crossing point of the high-voltage AC signal and switch the MOSFET pair in the diagonal direction to conduct each time the polarity flips. The on-resistance of the MOSFET is selected in the range of 20mΩ to 100mΩ. Under a typical operating current of 100μA to 500μA, the on-state voltage drop can be controlled in the range of 5μV to 50μV. Compared with the forward voltage drop of 200mV to 600mV of Schottky diodes, the rectification loss is reduced by 10 to 100 times. For the piezoelectric transducer output with an open-circuit voltage peak of 3V to 25V but a short-circuit current in the range of 10μA to 500μA, the proportion of energy consumed by the rectification stage itself in the total extractable energy is reduced from 10% to 20% to less than 1%.

[0046] During the rectification process, a ripple feature extraction path is drawn from the output side of the active full-bridge rectifier; the ripple feature extraction unit consists of a peak detector and a transconductance amplifier.

[0047] The peak detector, consisting of an operational amplifier with ultra-low bias current, a precision diode, and a holding capacitor, tracks the upper envelope waveform of the rectified output voltage; the envelope waveform reflects the variation pattern of the vibration signal amplitude. The envelope signal is fed into a transconductance amplifier, which converts the input voltage variation into an output current variation. This output current directly controls a variable impedance network. The variable impedance network consists of a MOSFET operating in the linear region and a fixed resistor connected in parallel. The fixed resistor has a value of 500kΩ. The output current of the transconductance amplifier changes the gate-source bias voltage of the MOSFET, thereby changing the equivalent resistance between its drain and source, and consequently, the first equivalent input impedance.

[0048] When the vibration energy is concentrated in the mid-to-high frequency range and the amplitude is large, the control current output by the ripple feature extraction unit increases, and the equivalent parallel resistance of the variable impedance network decreases accordingly, so that the load impedance seen by the piezoelectric transducer moves closer to the optimal load point under its current vibration frequency. Conversely, when the vibration energy weakens or the frequency shifts down, the control current decreases and the equivalent parallel resistance increases to adapt to the new optimal load conditions.

[0049] The second path processes low-voltage AC signals using the same principle, but the circuit parameters are adapted to the low-voltage and low-impedance characteristics of the electromagnetic transducer. The low-voltage AC signal enters the active full-bridge rectifier of the second impedance matching rectifier module. The MOSFET on-resistance of the second path active full-bridge rectifier is selected in the range of 10mΩ to 50mΩ to match the low source internal resistance characteristics of the electromagnetic transducer. The parameter ranges of the ripple feature extraction unit and the variable impedance network are also adjusted accordingly. The fixed resistance is 200Ω, and the adjustment range of the second equivalent input impedance is set from 5Ω to 200Ω to adapt to the optimal load impedance range of the electromagnetic transducer.

[0050] The first DC power is output from the first impedance matching rectifier module, and the second DC power is output from the second impedance matching rectifier module. The two DC power outputs are combined in the subsequent S3.

[0051] Since the control quantity of impedance regulation comes directly from the frequency band characteristics extracted from the rectified output ripple of each channel, rather than from the sampling and feedback calculation of the output DC power, the delay between the control signal and the change in vibration state is compressed to the order of the response time of the ripple extraction circuit, realizing rapid tracking of sudden changes in the vibration spectrum and avoiding energy extraction efficiency loss caused by impedance mismatch; the impedance regulation of the two channels is completely independent and does not cross each other, serving the optimal operating point of their respective transducers, and is not affected by voltage or power fluctuations of the other channel.

[0052] Rectification and impedance matching in vibration energy harvesting circuits typically rely on output power detection feedback. First, the transducer output is rectified, and the input power is estimated by measuring the rate of change of the energy storage capacitor voltage. Then, the equivalent impedance of the rectifier is adjusted. However, the adjustment signal needs to pass through multiple stages, resulting in limited response bandwidth. This leads to lag in impedance adjustment when the vibration spectrum changes abruptly, preventing complete energy extraction. For dual-transducer systems, while independent circuits avoid voltage clamping, a posteriori feedback control still exhibits hysteresis, and the static power consumption and board area of ​​the two circuits increase the burden when space is limited.

[0053] S2 extracts the frequency band energy characteristics directly from the rectified ripple and drives impedance adjustment in a feedforward manner, changing the impedance matching response path from output feedback to input feedforward. This synchronizes the timing of the adjustment action with the change in vibration state, overcoming the inherent tracking error caused by response lag in existing a posteriori feedback schemes. Both DC power sources are in a highly efficient extraction state under the current vibration conditions, but power backflow and cross-interference due to voltage differences must be avoided during the merging process to ensure the complete transfer of the independent optimization results of the dual sources to subsequent energy storage and distribution stages. Step S3 merges the first and second DC power sources onto a common energy storage bus.

[0054] The common energy storage bus refers to the DC power aggregation node that connects to the input of the hybrid energy storage and charging path scheduling module. The node carries all the output power from the first impedance matching rectifier module and the second impedance matching rectifier module, and provides charging energy to the subsequent supercapacitor units and battery units. Power merging refers to the circuit operation of merging the output of two independent power sources into the same electrical node in a unidirectional low-loss manner. When the output voltage of any power source is lower than the voltage of the common energy storage bus, the corresponding circuit must not draw reverse current from the bus.

[0055] The power merging stage is set in S3 because the voltage amplitude and power level of the first DC power and the second DC power change independently and are not synchronized. The piezoelectric transducer may output a high voltage instantaneously under medium and high frequency vibration excitation, so that the voltage of the first DC power reaches 3V~15V or more; while the voltage of the second DC power output by the electromagnetic transducer under low frequency vibration is only 50mV~500mV.

[0056] If the two outputs are directly connected to the same node without processing, when the voltage of the first DC power is higher than the voltage of the second DC power, the current will flow from the high voltage side to the low voltage side through the common connection point. This will cause the output of the second impedance matching rectifier module to be subjected to reverse bias, and the MOSFET parasitic body diode of its internal active rectifier may be turned on, forming a reverse current path. The reverse current not only consumes the extracted energy, but may also interfere with the normal operation of the ripple characteristic extraction unit of the second impedance matching rectifier module, and cannot accurately reflect the current vibration state of the electromagnetic transducer.

[0057] In a specific embodiment, power combining is accomplished through a power fusion module. The power fusion module consists of two gate drive comparators and two low on-resistance MOSFETs driven by them respectively; a first DC power is connected to the anode of the first gate drive comparator, a second DC power is connected to the anode of the second gate drive comparator, and the cathodes of the two gate drive comparators are connected to a common energy storage bus.

[0058] Each gate drive comparator contains a pair of voltage comparators. The non-inverting input of the voltage comparator is connected to the anode, the inverting input is connected to the cathode, and the output is connected to the gate of the corresponding MOSFET. To avoid MOSFET mis-turn-on due to common-mode interference that may exist in the two output voltages, a preset bias voltage of 20mV is set. This bias voltage value can be adjusted within the range of 10mV to 50mV according to the common-mode interference level in the actual circuit. When the anode voltage is higher than the cathode voltage and the difference exceeds the preset bias voltage, the comparator outputs a high level, the MOSFET turns on, and current flows from the anode to the cathode. When the anode voltage is lower than the cathode voltage, the comparator flips, the MOSFET turns off quickly, and the reverse current path from the common energy storage bus to the corresponding impedance-matched rectifier module is blocked.

[0059] In the power fusion module, the on-resistance of the MOSFET is in the range of 10mΩ to 50mΩ. Within the typical operating current range of 100μA to 50mA, the conduction loss is controlled to be less than 10μW. The turn-off response time of the MOSFET is determined by the push-pull capability of the gate drive circuit, and the state switching can be completed within 50ns. When the output voltage of a certain path suddenly drops, the corresponding MOSFET will turn off in a very short time to avoid the formation of reverse current.

[0060] After the two DC power sources are combined to a common energy storage bus via the power fusion module, the voltage on the bus is determined by the higher voltage of the two inputs, while the current is contributed by both. When one of the inputs has a lower voltage, the gate drive comparator of the corresponding circuit automatically blocks its connection to the bus, temporarily ceasing power delivery to the bus. However, the ripple characteristic extraction and impedance regulation within the impedance matching rectifier module continue to operate independently, unaffected by the bus side status. When the output voltage of the corresponding circuit rises again to exceed the existing voltage on the bus due to changes in vibration conditions, the gate drive comparator re-turns on the corresponding MOSFET, resuming power delivery.

[0061] The aforementioned construction method achieves unidirectional low-loss merging of two DC power sources. In terms of circuit structure, the power fusion module replaces the traditional parallel connection of two Schottky diodes with a combination of two MOSFETs and two gate drive comparators, reducing conduction losses from 50μW~500μW in the diode solution to 5μW~50μW. Functionally, the power fusion module physically prevents either output from becoming a reverse current load due to relatively low voltage, avoiding energy loss caused by cross-conduction in a dual-source system.

[0062] S3 uses a power fusion module to combine the first and second DC power into a common energy storage bus in a unidirectional, low-loss manner, where the energy extracted independently from two separate sources converges. However, the power supply on the common energy storage bus is not continuous and stable; instead, it exhibits intermittent and fluctuating characteristics depending on the vibration intensity of the joint reducer. Furthermore, the ratio of high-power duration to low-power interval dynamically changes with the robot's work cycle. Simultaneously, the power consumption mode of the subsequent wireless sensor nodes is a pulsed load characterized by long sleep periods and short bursts. This intermittent, low-power supply at the input end and the pulsed, high-power demand at the output end present a significant mismatch in both time scale and power magnitude, necessitating an intermediate link between them to buffer energy and smooth power fluctuations.

[0063] Step S4 receives energy from the common energy storage bus and automatically and without oscillation switches between sequential charging states based on the input power of the common energy storage bus to complete energy buffer storage. The sequential charging states include: when the bus power is lower than the first threshold, only energy is stored in the supercapacitor; when the bus power is higher than the second threshold, energy is stored in both the supercapacitor and the battery; and when the supercapacitor voltage reaches the preset saturation upper limit, the supercapacitor switches to constant current charging of the battery.

[0064] Hybrid energy storage units refer to energy storage arrays composed of supercapacitor units and battery units connected in parallel, with each unit playing a different energy storage role.

[0065] Supercapacitor units refer to energy storage elements that use the double-layer capacitor principle to store electrical energy. Their capacity ranges from 100μF to 1F, their equivalent series resistance is less than 100mΩ, their leakage current does not exceed 1μA under rated voltage, and they have the characteristics of high power density and charge-discharge cycle life of more than 500,000 times. They can absorb or release large currents in a very short time and are suitable for responding to power fluctuations.

[0066] A battery cell refers to a high-energy-density energy storage element that uses lithium ceramic solid electrolyte, with a capacity ranging from 500μAh to 2mAh, a charging cutoff voltage of 4.1V, and an allowable constant current charging range of 50μA to 500μA. It features high energy density and low self-discharge rate, making it suitable for long-term energy storage.

[0067] The differences between supercapacitor cells and battery cells in terms of charge / discharge rate, cycle life, and energy density determine their complementary relationship in a hybrid energy storage architecture: supercapacitor cells act as power buffers, quickly responding to input power fluctuations and instantaneous load demands; battery cells act as energy warehouses, storing surplus energy in a long-term stable manner.

[0068] Sequential charging state refers to the control logic by which the charging sequencer deterministically switches between three preset charging modes based on the current input power level of the common energy storage bus, rather than simply oscillating between charging and not charging based on a single voltage threshold.

[0069] The first threshold corresponds to a boundary where the input power is 100μW. When the input power on the common energy storage bus is less than 100μW, it indicates that the current vibration energy is at a low point and the extractable power is extremely limited.

[0070] The second threshold corresponds to a dividing boundary with an input power of 500μW. When the input power is higher than 500μW, it indicates that the vibration energy is abundant and the extractable power significantly exceeds the leakage current consumption of the supercapacitor unit.

[0071] The preset saturation upper limit refers to the state where the terminal voltage of the supercapacitor unit reaches 3.8V. At this point, the supercapacitor is close to its rated voltage, and continuing to charge it directly with a large current may shorten its lifespan. Furthermore, the input energy should be transferred to long-term storage.

[0072] Because the energy input on the common energy storage bus is significantly intermittent and fluctuating, S4 must employ sequential charging switching based on the input power level. If the supercapacitor and battery units are simply connected in parallel and directly to the common energy storage bus, during periods of extremely low input power, the bus voltage builds up slowly, and the battery unit's charging current may be uncertain due to insufficient voltage, causing repeated on / off switching of the switching transistors and generating unnecessary drive losses. When the input power suddenly increases, the supercapacitor unit, due to its low equivalent series resistance, will absorb most of the current, and the battery unit may remain in a partially charged state for an extended period due to insufficient charging current distribution. Furthermore, the supercapacitor and battery units have different charging and discharging characteristics; if a single charging management strategy is used, it cannot simultaneously meet the different needs of the supercapacitor unit's rapid response and the battery unit's long-term stability. Sequential charging is designed to adapt to the intermittent pulse rhythm of micro-energy harvesting, matching the energy buffering and storage process with the temporal distribution of the input energy.

[0073] In a specific embodiment, S4 is accomplished through a hybrid energy storage and charging path scheduling module, comprising three parts: a supercapacitor unit, a battery unit, and a charging sequencer. The charging sequencer consists of a voltage window detector and a switch array; the voltage window detector includes three hysteresis comparators, with the reference threshold generated by an ultra-low power bandgap reference source via a resistor divider network; the upper threshold of the first hysteresis comparator is set to 2.7V and the lower threshold to 2.5V, used to define the boundary between the low-energy region and the fully charged region; the upper threshold of the second hysteresis comparator is set to 3.8V and the lower threshold to 3.6V, used to define the entry and exit conditions of the saturation region.

[0074] The hysteresis window design provides hysteresis during state transitions, preventing frequent state jumps caused by minor voltage fluctuations near the boundary.

[0075] During operation, the charging sequencer uses voltage-power mapping logic based on the supercapacitor's state of charge to determine the power range, while also incorporating bus voltage change rate (dV / dt) detection to avoid misjudgment of voltage thresholds. The specific determination rules are as follows: When the supercapacitor voltage is below 2.5V, regardless of the current input power of the bus, it is determined to be a low-energy zone. Only the charging path from the common energy storage bus to the supercapacitor unit is closed, while the battery unit remains disconnected. When the supercapacitor voltage is between 2.5V and 2.7V, the dV / dt detection unit is activated to continuously calculate the bus voltage change rate: if dV / dt > 10mV / s, it is determined that the input power has exceeded the second threshold and entered the abundant region; if dV / dt ≤ 10mV / s, it remains in the low-energy region. When the voltage of the supercapacitor reaches or exceeds 2.7V, it automatically enters the charging zone and closes the charging path of the supercapacitor and the battery cell, so that the two energy storage elements can receive energy together. When the voltage of the supercapacitor rises above 3.8V, it enters the saturation region. At this time, the direct charging path of the bus is disconnected, and the constant current charging path from the supercapacitor to the battery is closed (constant current 20μA) to transfer the energy of the supercapacitor to the battery. When the supercapacitor voltage drops below 3.6V, it exits the saturation region and resumes the direct charging path from the bus to the energy storage unit.

[0076] The aforementioned dV / dt threshold of 10mV / s is set based on the charging time constant of the supercapacitor capacitance and equivalent series resistance. This can reliably distinguish between weak energy flow and abundant energy flow, eliminating the charging path oscillation caused by input fluctuations near the boundary due to the reliance on a voltage comparator.

[0077] The switch array consists of four single-pole single-throw analog switches, which respectively control the charging path from the common energy storage bus to the supercapacitor unit, the charging path from the common energy storage bus to the battery unit, the constant current charging path from the supercapacitor unit to the battery unit, and the protection disconnection path of the battery unit. The on / off state of each analog switch is directly driven by the combinational logic output of the comparator of the voltage window detector, without passing through a microcontroller or timing control unit, and the static power consumption is controlled below 100nA.

[0078] By using a charging sequencer in the hybrid energy storage and charging path scheduling module to classify and control the charging path switching between supercapacitor units and battery units, the hybrid energy storage unit achieves adaptive buffering for intermittent micro-energy. The supercapacitor unit, acting as a front-end fast-response energy storage, absorbs and releases power pulses with low loss, ensuring that the input energy on the common energy storage bus is captured in the shortest possible time. The battery unit, acting as a back-end long-term storage, only participates in charging when the input energy is sufficient or the supercapacitor unit is full, avoiding net energy loss due to charging management overhead during micro-energy phases. The introduction of a constant current transfer charging stage allows excess energy in the supercapacitor unit to be transferred at a rate adapted to the physical characteristics of the battery unit, improving the charge-discharge cycle life of the battery unit.

[0079] Step S4 avoids switching oscillations in the energy boundary region through hysteresis window grading and sequential state switching, reduces charging scheduling losses to nanoampere-level static current, and the supercapacitor-first architecture ensures that energy capture in the micro-energy stage is not affected by the power consumption of the battery management circuit, providing a reliable energy storage foundation with completed buffer accumulation for transient power distribution in S5.

[0080] S4 automatically switches between three sequential charging states based on the input power level, enabling the supercapacitor and battery units to complete energy buffering and storage at differentiated rhythms. The terminal voltage of the supercapacitor unit gradually increases with energy accumulation, providing an energy foundation for supplying power to the load. However, energy readiness is only one of the conditions for power supply. When to release electrical energy to the load also requires sensing the load's own state. The wireless sensing node does not continuously consume power, but only enters the data acquisition and burst transmission phase after power-on initialization. If power is forcibly supplied before the sensor has completed its internal reset or is in deep sleep, energy will be wasted in vain or even cause the sensor's power-on timing to be disordered. Therefore, S5 needs to simultaneously determine both the energy storage state and the load state, triggering power supply when both are satisfied.

[0081] Step S5 monitors the energy storage status of the supercapacitor unit in real time and continuously detects the transient current waveform on the load power supply circuit using a non-inductive sampling resistor; when the supercapacitor voltage exceeds the power supply threshold and the transient current waveform matches the task ready characteristic mode, the voltage regulator is triggered to provide transient power supply to the load.

[0082] A non-inductive sampling resistor is a current sensing element with extremely low resistance and parasitic inductance, connected in series in the output circuit of a voltage regulator. Its resistance ranges from 5mΩ to 20mΩ, and its parasitic inductance is less than 1nH. A weak differential voltage proportional to the load current is generated across the non-inductive sampling resistor. This voltage is then amplified to extract the transient characteristics of the load current without significantly consuming power from the power supply circuit.

[0083] The transient current waveform refers to the simulated waveform of the current changing over time as the load is charged from zero voltage at its power supply pins after the power supply path is connected, and the internal circuit sequentially completes the power-on reset and initialization process. The waveform includes time-series segments such as the initial surge charging stage, the stable operating current stage, and the static maintenance stage. The current amplitude, duration, and rate of change of each segment constitute an identifiable combination of features.

[0084] The task-ready characteristic mode refers to the pre-defined transient waveform characteristic criteria of the load current when the wireless sensor node is powered on, initialized, and ready to perform data acquisition tasks. The task-ready characteristic mode is defined as a combination of a set of timing logic conditions: an initial surge current pulse that rises to the threshold within 50μs~200μs, corresponding to the charging process of the decoupling capacitor inside the sensor; a working current plateau lasting 1ms~5ms, corresponding to the startup and initialization program of the sensor microcontroller; and the current falling back to a static sustaining level below 100μA, corresponding to the sensor having completed initialization and entered the data acquisition state.

[0085] The above combination of characteristics represents the common current behavior of most low-power wireless sensor nodes during power-up, determined by the power management circuitry and microcontroller reset timing within the sensor.

[0086] The power supply threshold refers to the minimum terminal voltage that the supercapacitor unit must reach, which is set to 3.3V to ensure that the subsequent voltage regulator can stably output the 1.8V or 3.3V operating voltage required by the sensor, and that the voltage drop of the supercapacitor unit during the power supply pulse will not cause the output to become unstable.

[0087] The reason why S5 uses a dual-condition joint judgment instead of triggering power supply solely based on the energy storage voltage threshold is that the energy storage state and the load state are two independent variables in terms of timing.

[0088] Triggering power supply solely based on a voltage threshold may occur when the sensor is in deep sleep or under internal short-circuit protection. In such cases, the power supply pulse not only fails to drive the sensor to complete its effective task but also consumes energy as an invalid charging cycle due to the unprepared load. Furthermore, if the sensor relies solely on itself to wake up and draw power from the energy storage system, the quiescent current of the sensor during sleep becomes a continuous load on the energy storage system. Under a microwatt-level average power budget, this continuous consumption significantly reduces the net energy accumulation rate. The S5 architecture transforms the power supply triggering mechanism from a one-way voltage judgment to a collaborative judgment based on the readiness of both energy storage and the load, ensuring that each power supply pulse occurs when the sensor is indeed capable of performing its task.

[0089] In a specific embodiment, S5 relies on the transient power distribution and load status detection module. The energy storage monitoring section continuously monitors the terminal voltage of the supercapacitor unit using an ultra-low power voltage comparator. When the voltage exceeds 3.3V, it outputs an energy ready signal. The load detection section consists of a non-inductive sampling resistor, an instrumentation amplifier, and a timing discrimination network.

[0090] The non-inductive sampling resistor is connected in series between the output circuit of the voltage regulator and the load connection point. The differential voltage amplitude generated across the resistor is 1μV~2μV when the load current is 100μA, and 150μV~600μV when the load current is 30mA.

[0091] A weak differential voltage is fed into a zero-drift instrumentation amplifier with a gain of 100. The amplified signal is then simultaneously distributed to three high-speed comparator channels. The first comparator channel has a differentiating circuit at its input, with a threshold of 50mV, to capture initial surge pulses with a rise time shorter than 200μs. The second comparator channel has a low-pass filter at its input, with a threshold of 20mV, to identify operating current plateaus lasting longer than 1ms. The third comparator channel has a threshold of 10mV, to identify static sustaining levels with current below 100μA. The output level changes of the three comparators collectively constitute the real-time determination of each timing segment of the current transient waveform.

[0092] The timing discrimination network uses the outputs of three comparators as the inputs to the state machine. The state machine contains five asynchronous logic states: IDLE, WINDOW1, DETECT, WINDOW2, and READY. It consists of D flip-flops, a retrievable monostable multivibrator, and a five-stage gate delay chain. The state transition path solidifies the task-ready feature pattern.

[0093] IDLE is the idle state, waiting for the rising edge of the first comparator channel to trigger; When WINDOW1 captures an initial surge pulse with a rise time shorter than 200μs (corresponding to the first time window) in the first comparator channel, it triggers the first monostable multivibrator to generate a 500μs timing window, and the state jumps from IDLE to WINDOW1. The 500μs timing corresponds to the first time window, which is used to define the effective judgment range of the surge pulse. DETECT means that if, within the validity period of WINDOW1, the second comparator channel confirms that the load current remains above 20mV (corresponding to the operating current plateau) and is maintained for more than 1ms, the state will transition from WINDOW1 to DETECT. In this state, the circuit continuously monitors the output of the third comparator channel. When the third comparator channel detects that the current has dropped from the operating platform and triggers a flip, WINDOW2 activates the second monostable multivibrator to generate a 10ms timing window, and the state transitions from DETECT to WINDOW2. This 10ms timing corresponds to the second time window, used to determine whether the load current has stably dropped back to the static maintenance level (below 100μA) within a preset time. If, within the 10ms window of WINDOW2, the third comparator channel confirms that the current has dropped to the static maintenance level and remains stable, the state changes from WINDOW2 to READY, and a power supply enable pulse is output.

[0094] To eliminate comparator output glitches, a five-stage gate delay chain is inserted in the critical path of state transition. Each stage has a delay of 2μs, with a total delay of 10μs. The transition is considered valid only after the signal has stabilized for more than 10μs, thus avoiding false state transitions caused by narrow pulse interference.

[0095] The corresponding circuit implementation is as follows: the clock of the first D flip-flop is connected to the output of the first comparator, the data terminal is connected to a high level, and the reset is connected to the global shutdown signal; the first monostable multivibrator is triggered by the rising edge of the first comparator, with a timing of 500μs, and its output enables the clock of the second D flip-flop; the data terminal of the second D flip-flop is connected to the non-inverting output of the first D flip-flop; the second monostable multivibrator is triggered by the topping edge of the third comparator, with a timing of 10ms; the clock of the third D flip-flop is connected to the output of the third comparator, and the data terminal is connected to the non-inverting output of the second D flip-flop; a three-input AND gate combines the non-inverting outputs of the first, second, and third D flip-flops and the necessary window valid signals, and the output is debouncing through a five-stage gate delay chain and then latched to the fourth D flip-flop, finally generating a power enable pulse.

[0096] The power enable pulse is directly connected to the enable terminal of the voltage regulator. The voltage regulator employs a synchronous buck topology based on pulse frequency modulation. After the enable signal is active, it converts the 3.3V~5.0V input voltage of the supercapacitor unit into the stable output voltage required by the sensor. The output current capability is 50mA, with a conversion efficiency exceeding 90% within a load range of 10μA~50mA, and a quiescent current below 100nA. The voltage regulator provides transient power to the load during the active enable pulse period, with a pulse width set to 50ms, covering the time required for the sensor to complete one data acquisition and wireless transmission cycle.

[0097] Power supply triggering is no longer solely determined by the energy storage voltage, nor does it rely on the digital handshake signal emitted by the sensor. The analog characteristics of the load current transient waveform directly reflect the physical state of the sensor's internal power-on process. The timing discrimination network completes real-time identification of the physical state with nanoampere-level static power consumption. The logical AND operation of the energy ready signal and the load state discrimination signal ensures that the initiation time of the power supply pulse is aligned with the sensor's task ready time, avoiding invalid energy output when the sensor is not ready, and also avoiding the static consumption of the energy storage system due to the sensor's continuous standby.

[0098] The S5 is built with pure analog hardware to achieve autonomous blind perception of load status. The power management circuit can still accurately determine the moment when the sensor is fully powered on and initialized without relying on any digital protocol.

[0099] S5 simultaneously monitors the terminal voltage of the supercapacitor unit and the transient current waveform on the load power supply circuit. When energy storage is ready and a task-ready characteristic mode is detected, it triggers the voltage regulator to provide transient power to the load. After the power supply pulse is initiated, energy is continuously delivered from the supercapacitor unit to the wireless sensing node via the voltage regulator. The sensor sequentially completes data acquisition, processing, and wireless burst transmission. However, the termination time of the power supply pulse is not determined by a preset fixed duration. If the power supply is turned off prematurely before the transmission is completed, it will lead to data loss or abnormal sensor operation. If the power supply continues after the transmission is completed, the static current after the sensor enters deep sleep will become a continuous load on the energy storage system, consuming the stored energy. Therefore, S6 needs to continuously monitor the load status during the power supply period and execute the power supply path shutdown operation when it confirms that the sensor task has been completed and has entered sleep mode.

[0100] Step S6 continuously monitors the load current during the power supply maintenance period. When a match with the task completion characteristic pattern is detected, all power supply paths between the voltage regulator and the load are immediately shut off, and the system is restored to standby state.

[0101] The task completion characteristic mode refers to the pre-defined transient waveform characteristic criterion of the load current when the corresponding wireless sensor node completes the burst data transmission and enters a deep sleep state.

[0102] The task completion characteristic pattern is defined as a combination of a set of timing logic conditions: First, there is the normal load current activity during the power enable pulse validity period, including the current changes corresponding to the stages of sensor microcontroller operation, analog-to-digital converter sampling, and burst emission of RF power amplifier; then there is a low-current segment lasting more than 10ms and with a current below 50μA, which corresponds to the static holding current when the sensor has completed data transmission, shut down the RF module, and entered low-power standby; after the low-current segment, there may be a negative pulse group with a pulse width of less than 200μs and an amplitude more than 3 times that of the previous normal load current. The negative pulse group is caused by the charge discharge generated by the power gate transistor of the digital circuit module inside the sensor at the moment of shutdown.

[0103] The above combination of features reflects the complete transition process of the sensor from an active working state to a dormant state. Its physical essence is the sequence of current changes on the power supply pins when the various functional modules inside the sensor are turned off in sequence.

[0104] Since the task duration of wireless sensor nodes is not a fixed constant, S6 must continuously monitor and shut down based on task completion characteristic patterns.

[0105] Factors such as the number of sensor channels for data acquisition, the sampling rate of the analog-to-digital converter, the data packet length of the RF link, and the number of retransmissions can all cause the duration of a single task to vary within the range of 10ms to 50ms. If a fixed timer shutdown method is used, setting the timer value too short may result in power failure before the transmission is completed, while setting it too long will generate unnecessary energy consumption. The static current of the sensor in deep sleep mode is between 1μA and 10μA. If the power supply path is maintained continuously after the task is completed, calculated at 10μA, the additional energy consumed per second is 33μJ. Under the condition of an average extractable power of 200μW, the above continuous consumption accounts for more than 16% of the net energy accumulation rate. By detecting the physical characteristics of task completion and shutting down immediately after confirmation, the continuous consumption can be reduced to the nanoampere level.

[0106] In a specific embodiment, S6 continues to operate during the power supply maintenance period through the transient power distribution and load status detection module. The non-inductive sampling resistor, instrumentation amplifier, and three comparator channels established in S5 continue to operate during the validity period of the power supply pulse, but after the state machine of the timing discrimination network outputs the power supply enable pulse, its internal logic transitions to the task completion detection state.

[0107] The timing discrimination network continuously monitors the output level of the third comparator channel. The threshold of the third comparator channel is set to 10mV, corresponding to a current detection threshold of approximately 100μA. When the third comparator channel confirms that the load current has dropped below 100μA and remains below 10ms for more than 10ms, the sustained state of the output level triggers a 10ms monostable multivibrator in the timing discrimination network. The timing period of the monostable multivibrator is set to 10ms, and on-chip low temperature coefficient capacitors and precision bias current sources are used to ensure timing accuracy within ±15%, which is sufficient to cover the brief current fluctuations that may occur between the completion of the task and the entry into sleep mode, avoiding false shutdown due to transient current glitches.

[0108] The 10ms monostable output signal enables the negative pulse detection path, which consists of an AC coupling capacitor, a high-speed comparator, and a 200μs monostable multivibrator. The AC coupling capacitor blocks the DC current component, transmitting only the AC component of the transient current change to the high-speed comparator. The inverting input of the high-speed comparator is biased at the negative threshold, which is dynamically set to three times the amplitude based on the previous normal load current peak value. When a negative pulse occurs and its amplitude exceeds this dynamic threshold, the comparator flips, triggering the 200μs monostable state. The output of the 200μs monostable state and the output of the 10ms monostable state are combined in a logic AND gate. When both are active simultaneously, the AND gate outputs a global shutdown signal.

[0109] After acquiring the global shutdown signal, the system shuts down the enable terminal of the voltage regulator, puts all the switches of the synchronous buck converter in the off state, and stops power output; it shuts down the PMOS switch on the power supply path, which is located between the output terminal of the voltage regulator and the load connection point, completely electrically isolates the load from the circuit, cuts off any possible leakage current path on the load side, including the subthreshold current of the electrostatic protection diode inside the sensor and the leakage current of the decoupling capacitor; it resets the state machine of the timing discrimination network and all monostable multivibrators, so that the load detection part of the system returns to the idle state, ready for the triggering of the next power supply cycle.

[0110] The standby power consumption of the circuit after shutdown includes only the bias current of the ultra-low power voltage comparator, the quiescent current of the gate drive comparator, and the comparator bias current of the charging sequencer, with the total standby power consumption controlled below 150nA. The shutdown time is determined by the physical state of the load itself, rather than a preset fixed timing, adapting to changes in task duration under different data lengths and transmission conditions.

[0111] S6 continuously detects and matches the transient characteristics of load current in the analog domain, identifying the physical moment when the sensor task is completed in a purely hardware manner. It achieves precise power cut-off and complete isolation of the load side without the need for sensor software or digital communication protocols.

[0112] Exemplary system: like Figure 1 As shown, an adaptive power supply circuit system for composite broadband micro-vibration energy harvesting includes: The transducer signal separation and acquisition module is used to acquire the high-voltage AC signal generated by the piezoelectric transducer in response to broadband micro-vibration and the low-voltage AC signal generated by the electromagnetic transducer in response to broadband micro-vibration, and to keep the two signals physically isolated from the source of acquisition. The first impedance matching rectifier module is connected to the transducer signal separation and acquisition module. It is used to receive the high-voltage AC signal and extract the first ripple feature from the voltage waveform after the signal is actively rectified. It dynamically adjusts the first equivalent input impedance in a feedforward manner and outputs the first DC power. The second impedance matching rectifier module is connected to the transducer signal separation and acquisition module. It is used to receive low-voltage AC signals and extract the second ripple characteristics from the voltage waveform after the signal is actively rectified. It dynamically adjusts the second equivalent input impedance in a feedforward manner and outputs the second DC power. The power fusion module is connected to the output terminals of the first impedance matching rectifier module and the second impedance matching rectifier module respectively. It is used to combine the first DC power and the second DC power into a common energy storage bus in a unidirectional low-loss manner and block the reverse current that may be generated at either output terminal. The hybrid energy storage and charging path scheduling module is connected to the common energy storage bus and includes a supercapacitor unit, a battery unit and a charging sequencer. The charging sequencer is used to automatically switch between multiple preset sequential charging states according to the input power level of the common energy storage bus in order to control the differentiated energy buffer storage of the supercapacitor unit and the battery unit. The transient power distribution and load status detection module is connected to the hybrid energy storage and charging path scheduling module and a load, respectively. When the energy storage status of the supercapacitor unit exceeds the power supply threshold and the transient current waveform on the load power supply circuit continuously detected by the non-inductive sampling resistor matches a preset task ready characteristic mode, the voltage regulator is triggered to provide transient power to the load. During the power supply maintenance period, when the transient current waveform is continuously detected to match a preset task completion characteristic mode, all power supply paths between the voltage regulator and the load are immediately shut off.

[0113] In a specific embodiment, the timing discrimination network in the transient power distribution and load status detection module is further refined, and the specific circuit structure and parameter selection of the asynchronous state machine are described in detail to illustrate that the present invention does indeed have the ability to autonomously identify loads without a digital controller.

[0114] Specifically, the timing discrimination network consists of four D flip-flops, two retrievable monostable multivibrators, a five-stage gate delay chain, and a three-input AND gate.

[0115] The clock input of the first D flip-flop is connected to the output of the first comparator channel, the data input is connected to a high level, and its reset terminal is connected to the global shutdown signal. The clock input of the second D flip-flop is connected to the output of the second comparator channel, and the data input is connected to the non-inverting output of the first D flip-flop. The clock input of the third D flip-flop is connected to the output of the third comparator channel, and the data input is connected to the non-inverting output of the second D flip-flop. The fourth D flip-flop is used as an output latch. Its data input is connected to the output of a three-input AND gate. Its clock is provided by a local 100kHz oscillator. The non-inverting output is the output node of the power enable pulse.

[0116] The two monostable multivibrators were set to timing periods of 500μs and 10ms, respectively.

[0117] The first monostable multivibrator is triggered by the rising edge of the first comparator channel. The output pulse enables the clock of the second D flip-flop and performs an AND logic operation with the output of the second monostable multivibrator. The second monostable multivibrator is triggered by the topping edge of the third comparator channel. Its 10ms timing period is used to define the maximum time interval allowed for the load current to drop from the operating platform to the static holding level.

[0118] Initially, all D flip-flops output low.

[0119] When the first comparator channel detects a rising surge pulse within 50μs, it outputs a positive transition, triggering the first monostable multivibrator to generate a 500μs window pulse, and simultaneously setting the first D flip-flop to a high level.

[0120] During the window pulse validity period, if the second comparator channel confirms that the load current is continuously higher than 20mV and the duration exceeds 1ms, it outputs a high level, setting the second D flip-flop to a high level. After that, when the sensor completes internal initialization and the load current begins to drop, the third comparator channel flips, triggering the second monostable multivibrator to generate a 10ms window pulse. If the third comparator channel confirms that the current has dropped to a static holding level below 100μA within the 10ms window, the third D flip-flop is set to a high level.

[0121] At this point, all three inputs of the three-input AND gate are high, originating from the Q terminals of the first, second, and third D flip-flops respectively, and the AND gate outputs a high level. This high level is latched by the fourth D flip-flop on the rising edge of the next local oscillator clock, and the Q terminal outputs a power-on enable pulse. This pulse is simultaneously connected to the enable pin of the synchronous buck regulator and the asynchronous reset pin of the first D flip-flop, causing the state machine to enter the power-on state and resetting the front-end discrimination path, preparing for subsequent task completion and detection.

[0122] After the power enable pulse lasts for 50ms, it is automatically set low by the internal counter. At this time, the output of the fourth D flip-flop flips low, and the system enters the task completion detection phase. Subsequent circuits are connected to the task completion feature matching network. The network structure is similar to the above path, but the matching feature is a low current segment exceeding 10ms followed by a negative pulse group. Once the matching is successful, a global shutdown signal is output to turn off the PMOS switch and reset all D flip-flops and monostable circuits. The static power consumption of the entire timing discrimination network does not exceed 200nA, and the discrimination delay is less than 2μs, meeting the dual constraints of speed and power consumption in micro-energy harvesting scenarios.

[0123] The correspondence between the above four D flip-flops and monostable circuit and the five states of the state machine (IDLE, WINDOW1, DETECT, WINDOW2, READY) is as follows: the first D flip-flop latches the trigger condition for IDLE→WINDOW1, the second D flip-flop corresponds to the jump from WINDOW1 to DETECT, the third D flip-flop corresponds to the entry condition for DETECT→WINDOW2, and the fourth D flip-flop latches the determination result of the READY state; each stage of the five-stage gate delay chain has a delay of 2μs, with a total delay of 10μs, and is inserted between the output of the three-input AND gate and the data input terminal of the fourth D flip-flop.

[0124] By using a pure hardware load detection architecture, the problem of identifying the timing state of complex loads is transformed into analog domain timing logic matching of several key current transient characteristics, without the need to use analog-to-digital converters, microcontrollers or any form of programmable logic.

[0125] Based on the preferred embodiments of the present invention described above, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. An adaptive power supply circuit system for composite broadband micro-vibration energy harvesting, characterized in that, include: The transducer signal separation and acquisition module is used to acquire the high-voltage AC signal generated by the piezoelectric transducer in response to broadband micro-vibration and the low-voltage AC signal generated by the electromagnetic transducer in response to broadband micro-vibration, respectively. The first impedance matching rectifier module is connected to the transducer signal separation and acquisition module. It is used to receive high-voltage AC signals, extract the first ripple characteristics from the voltage waveform after active rectification, dynamically adjust the first equivalent input impedance in a feedforward manner, and output the first DC power. The second impedance matching rectifier module is connected to the transducer signal separation and acquisition module. It is used to receive low-voltage AC signals, extract the second ripple characteristics from the voltage waveform after active rectification, dynamically adjust the second equivalent input impedance in a feedforward manner, and output the second DC power. The power fusion module is connected to the output terminals of the first impedance matching rectifier module and the second impedance matching rectifier module respectively. It is used to combine the first DC power and the second DC power into a common energy storage bus in a unidirectional low-loss manner and block the reverse current that may be generated at either output terminal. The hybrid energy storage and charging path scheduling module is connected to the common energy storage bus and includes a supercapacitor unit, a battery unit and a charging sequencer. The charging sequencer is used to automatically switch between multiple preset sequential charging states according to the input power level of the common energy storage bus in order to control the differentiated energy buffer storage of the supercapacitor unit and the battery unit. The transient power distribution and load status detection module is connected to the hybrid energy storage and charging path scheduling module and a load, respectively. It enables transient power supply to the load when the energy storage status of the supercapacitor unit exceeds the power supply threshold and the transient characteristics of the load current obtained through non-intrusive load status sensing meet the preset task readiness criteria; and shuts off the power supply path to the load when the transient characteristics of the load current meet the preset task completion criteria during the power supply maintenance period.

2. The adaptive power supply circuit system for composite broadband micro-vibration energy harvesting according to claim 1, characterized in that: The first impedance matching rectifier module and the second impedance matching rectifier module each include: An active full-bridge rectifier uses controlled semiconductor switching devices to convert the polarity of AC signals; The ripple feature extraction unit includes a peak detector and a transconductance amplifier. The peak detector tracks the upper envelope waveform of the rectified output voltage, and the transconductance amplifier converts the envelope voltage into a control current. A variable impedance network is formed by a MOSFET operating in the linear region connected in parallel with a fixed resistor. The control current changes the gate-source bias voltage of the MOSFET to adjust the equivalent parallel impedance. The variable impedance network adjustment range of the first impedance matching rectifier module is adapted to the source internal resistance range of the piezoelectric transducer, and the variable impedance network adjustment range of the second impedance matching rectifier module is adapted to the source internal resistance range of the electromagnetic transducer.

3. The adaptive power supply circuit system for composite broadband micro-vibration energy harvesting according to claim 1, characterized in that: The power fusion module includes two gate drive comparators and two low on-resistance MOSFETs driven by them respectively. Each gate drive comparator includes a pair of voltage comparators. The non-inverting input terminal of the voltage comparator is connected to the anode, the inverting input terminal is connected to the cathode, and the output terminal is connected to the gate of the corresponding MOSFET. When the anode voltage is higher than the cathode voltage and the difference exceeds the preset bias voltage, the MOSFET is turned on; when the anode voltage is lower than the cathode voltage, the MOSFET is turned off to block the reverse current from the common energy storage bus to the corresponding impedance-matched rectifier module.

4. The adaptive power supply circuit system for composite broadband micro-vibration energy harvesting according to claim 1, characterized in that: The charging sequencer includes a voltage window detector and a switch array; The voltage window detector includes multiple hysteresis comparators to define at least three power ranges: a low-energy range below a first threshold, a high-energy range above a second threshold, and a saturation range where the supercapacitor voltage reaches a preset saturation upper limit. The switch array switches between the following sequential charging states based on the output of the voltage window detector: charging only the supercapacitor unit, charging both the supercapacitor unit and the battery unit simultaneously, and constant current charging from the supercapacitor unit to the battery unit.

5. An adaptive power management method for composite broadband micro-vibration energy harvesting, comprising an adaptive power supply circuit system for composite broadband micro-vibration energy harvesting according to any one of claims 1-4, characterized in that, include: S1. Acquire the high-voltage AC signal generated by the piezoelectric transducer in response to broadband micro-vibration and the low-voltage AC signal generated by the electromagnetic transducer in response to broadband micro-vibration. S2. Actively rectify the high-voltage AC signal, extract the first ripple characteristic from the high-voltage AC signal, dynamically adjust the first equivalent input impedance, and output the first DC power; simultaneously, actively rectify the low-voltage AC signal, extract the second ripple characteristic from the low-voltage AC signal, dynamically adjust the second equivalent input impedance, and output the second DC power. S3. Combine the first DC power and the second DC power onto a common energy storage bus; S4. Receive energy from the common energy storage bus and automatically and without oscillation switch between sequential charging states based on the input power of the common energy storage bus to complete energy buffer storage; wherein, the sequential charging states include: when the bus power is lower than the first threshold, only the energy is stored in the supercapacitor; when the bus power is higher than the second threshold, the energy is stored in both the supercapacitor and the battery; when the supercapacitor voltage reaches the preset saturation upper limit, the supercapacitor switches to constant current charging to the battery; S5. Monitor the energy storage status of the supercapacitor in real time and continuously detect the transient current waveform on the load power supply circuit using a non-inductive sampling resistor; when the supercapacitor voltage exceeds the power supply threshold and the transient current waveform matches the task ready characteristic mode, trigger the voltage regulator to provide transient power supply to the load. S6. During the power supply maintenance period, continuously monitor the load current. When a match with the task completion characteristic pattern is detected, immediately shut off all power supply paths between the voltage regulator and the load, and restore the system to standby state.

6. The adaptive power management method for composite broadband micro-vibration energy harvesting according to claim 5, characterized in that: The extraction of the first ripple feature and the second ripple feature in step S2 respectively includes: Track the upper envelope waveform of the corresponding active rectified output voltage, and convert the voltage change of the upper envelope waveform into control current through transconductance amplification; The equivalent parallel impedance of the variable impedance network, which consists of a MOSFET and a fixed resistor connected in parallel, is changed by the control current to adjust the equivalent input impedance of the corresponding transducer.

7. The adaptive power management method for composite broadband micro-vibration energy harvesting according to claim 5, characterized in that: The method for merging the first DC power and the second DC power into the common energy storage bus in step S3 is as follows: The first DC power and the second DC power are respectively connected to the anodes of two MOSFETs driven by voltage comparators, and the cathodes of the two MOSFETs are connected to a common energy storage bus. When the anode voltage of either path is lower than the voltage of the common energy storage bus, the corresponding MOSFET is turned off, blocking the reverse current from the bus to that path.

8. The adaptive power management method for composite broadband micro-vibration energy harvesting according to claim 5, characterized in that: The switching between sequential charging states in step S4 is implemented based on a hysteresis comparator: The upper and lower thresholds of the first hysteresis comparator define the boundary between the low-energy region and the abundant region, and the upper and lower thresholds of the second hysteresis comparator define the entry and exit conditions of the saturation region. When the supercapacitor voltage reaches the preset saturation upper limit, the constant current charging path from the supercapacitor to the battery is closed, and the supercapacitor energy is transferred to the battery with a preset constant current until the supercapacitor voltage falls back to below the lower threshold of the saturation region.

9. The adaptive power management method for composite broadband micro-vibration energy harvesting according to claim 5, characterized in that: The task-ready feature pattern described in step S5 includes at least a combination of the following temporal logic conditions: An initial surge current pulse with a rise time shorter than a preset value was detected within the first time window; A working current plateau was detected within the second time window, with a duration exceeding a preset value. Subsequently, the current was detected to have dropped to a level below the preset static maintenance threshold.

10. The adaptive power management method for composite broadband micro-vibration energy harvesting according to claim 5, characterized in that: The task completion feature pattern described in step S6 includes at least a combination of the following temporal logic conditions: The load current is detected to drop below a preset low current threshold and remain below a preset time period. After the preset time period, a negative pulse group with a pulse width less than the preset value and an amplitude exceeding a multiple of the previous normal load current peak value is detected; After confirming the task completion characteristic mode, turn off the enable terminal of the voltage regulator and disconnect the switch between the voltage regulator output terminal and the load to completely electrically isolate the load from the circuit.

Images