Environment adaptive LED constant current energy-saving control method

Abstract

This invention relates to the field of lighting control technology, specifically disclosing an environment-adaptive LED constant current energy-saving control method. The method includes: arranging a dielectric-photoelastic sensing array, a magnetostrictive-leakage flux detection coil, and a broadband impedance spectrum detection unit on the phosphor encapsulation layer, the inductor core, and the output bus, respectively; simultaneously acquiring three types of timing signals; extracting dielectric-magnetic-electric coupling degradation feature vectors; inputting the feature vectors into an energy-saving-lifetime co-optimization model; correcting the target luminous flux based on the dielectric relaxation offset; constructing a core loss-efficiency mapping surface to solve for the minimum energy consumption operating point; generating a constant current target value offset sequence and a PWM chopper phase-frequency co-transfer strategy; and simultaneously adjusting the duty cycle, switching frequency, and gate impedance through a feedforward-feedback composite adjustment network, and feeding back the current ripple spectrum and dielectric relaxation offset to update the feature vector weights. This invention achieves multi-physics field co-sensing and energy-saving-lifetime co-optimization.

Description

Technical Field

[0001] This invention relates to the field of lighting control technology, and more specifically to an environmentally adaptive LED constant current energy-saving control method. Background Technology

[0002] Current LED constant current drive technologies generally use photoresistors or temperature sensors for single environmental parameter feedback, maintaining a constant output current through simple proportional-integral-derivative closed-loop regulation. However, as LED lighting systems develop towards higher power and longer lifespan, the limitations of traditional sensing solutions are becoming increasingly apparent.

[0003] In the phosphor encapsulation layer, YAG:Ce³ + Under long-term thermoelectric coupling stress, phosphors experience dielectric polarization drift, leading to Stokes shift and quantum efficiency decay. Current technologies, relying solely on point temperature monitoring, struggle to accurately detect the precursors to luminous efficiency degradation indicated by dielectric relaxation shift. In power inductor cores, grid harmonic injection and magnetostriction effects couple to cause abnormal shifts in the ferrite hysteresis loop, distorting the air gap leakage flux density distribution with aging. Traditional drive strategies fail to establish a joint sensing mechanism for magnetostriction harmonics and leakage flux crosstalk, making it impossible to simultaneously suppress core loss increments during constant current regulation. At the drive power supply output bus, the equivalent series resistance of the filter electrolytic capacitor increases with lifetime. Low-frequency impedance increments and high-frequency resonant peak shifts characterize the circuit's electrical aging state. Current technologies lack broadband impedance spectrum detection methods, making it difficult to incorporate the frequency domain evolution of bus impedance into energy-saving optimization decisions.

[0004] Furthermore, existing PWM chopping strategies only focus on duty cycle adjustment and do not consider the coupling relationship between the magnetostrictive vibration phase of the magnetic core and the chopping phase, which leads to the intensification of electromagnetic-mechanical stress in a specific phase and accelerates the degradation of the magnetic core.

[0005] In summary, existing technologies have failed to establish a multi-physics field coupling sensing mechanism that combines dielectric, magnetic, and electrical fields with a collaborative optimization mechanism for energy saving and lifespan, making it difficult to simultaneously achieve constant current accuracy, energy efficiency, and drive power supply lifespan. Summary of the Invention

[0006] The purpose of this invention is to provide an environmentally adaptive LED constant current energy-saving control method to solve the problem that existing LED constant current driving technology relies on single temperature or photosensitive feedback, which makes it difficult to coordinate the identification of phosphor dielectric polarization degradation, magnetostriction harmonic vibration of magnetic core and bus impedance aging state, resulting in a difficulty in balancing energy saving and lifespan.

[0007] To solve the above-mentioned technical problems, the present invention specifically provides the following technical solution: An environmentally adaptive LED constant current energy-saving control method includes the following steps: S1. A dielectric-photoelastic sensing array, a magnetostrictive-leakage flux detection coil, and a broadband impedance spectrum detection unit are respectively arranged on the phosphor encapsulation layer, the inductor core, and the output bus. Three types of time-series signals are acquired simultaneously, and the dielectric-magnetic-electric coupling degradation feature vector is extracted. S2. Input the dielectric-magnetic-electric coupling degradation feature vector into the energy-saving-lifetime co-optimization model, and generate a constant current target value offset sequence and a PWM chopper phase-frequency co-transfer strategy based on the degree of dielectric relaxation offset. S3. Based on the constant current target value offset sequence and the PWM chopper phase-frequency collaborative migration strategy, the duty cycle, switching frequency and gate impedance are synchronously adjusted through the feedforward-feedback composite adjustment network. The current ripple spectrum and dielectric relaxation offset are then fed back to update the dielectric-magnetic-electric coupling degradation characteristic vector to form a closed loop.

[0008] As a preferred embodiment of the present invention, S1 specifically includes: S11. A dielectric-optic elastic coupling sensor array is attached to the surface of the LED phosphor encapsulation layer, a magnetostrictive-leakage flux coupling detection coil is arranged on the air gap sidewall of the power inductor core, and a broadband impedance spectrum detection unit is arranged at a preset node of the drive power output bus. S12. After the dimming command is issued, three detection channels are started simultaneously to continuously capture dielectric relaxation timing signal, magnetostriction harmonic-leakage flux crosstalk timing signal and broadband impedance spectrum timing signal. S13. Input the captured timing signal into the frequency domain-time domain joint decoupling module, extract the phosphor dielectric relaxation offset and polarization loss factor, magnetostriction amplitude index and leakage flux density distribution characteristics, low-frequency impedance increment and high-frequency resonance peak offset from it, and encode them into dielectric-magnetic-electric coupling degradation feature vector.

[0009] As a preferred embodiment of the present invention, S11 specifically includes: S111. A dielectric-optical-elastic coupling sensing array is attached to the surface of a phosphor encapsulation layer using thermally conductive insulating adhesive. The sensing array consists of multiple interdigitated electrode units arranged in a coplanar waveguide structure, and the surface of each interdigitated electrode unit is coated with a coating similar to YAG:Ce³. + A sensitive thin film with matching phosphor dielectric properties, and the attachment range covers the central luminescent area and the edge transition area of ​​the phosphor encapsulation layer; S112. The magnetostrictive-leakage flux coupling detection coil is fixed to the air gap sidewall of the power inductor core. The detection coil consists of an excitation winding and an induction winding wound on a high permeability core. The excitation winding is used to apply low-frequency alternating excitation to excite magnetostrictive vibration, and the induction winding is used to pick up harmonic changes and air gap leakage flux induced electromotive force. S113. Connect the broadband impedance spectrum detection unit to the preset detection node of the drive power supply output bus. The detection unit includes a broadband current injection probe connected in series with the bus and a broadband voltage pickup probe connected in parallel with the bus, which are used to inject a sweep frequency excitation signal and synchronously acquire a voltage response signal, respectively.

[0010] As a preferred embodiment of the present invention, S13 specifically includes: S131. The dielectric relaxation timing signal is converted to the time-frequency domain by short-time Fourier transform. The center frequency offset of the dielectric relaxation peak is extracted as the phosphor dielectric relaxation offset, and the peak value of the dielectric loss tangent is extracted as the polarization loss factor. S132. The magnetostrictive harmonic-leakage flux crosstalk timing signal is decomposed into multiple frequency band components by wavelet packet decomposition. The amplitude is extracted from the fundamental frequency component as the magnetostrictive amplitude index, and the energy spectral density is extracted from the harmonic component as the leakage flux density distribution characteristic. S133. The broadband impedance spectrum time-series signal is decomposed into the slope of the real part of the impedance in the low-frequency band as the low-frequency impedance increment, and the resonant frequency offset in the high-frequency band as the high-frequency resonant peak offset, by an impedance fitting algorithm. S134. The phosphor dielectric relaxation offset, polarization loss factor, magnetostriction amplitude index, leakage flux density distribution characteristics, low-frequency impedance increment and high-frequency resonance peak offset are arranged in a preset dimension order and timestamp is added. They are then encoded together into a one-dimensional feature vector, which serves as the dielectric-magnetic-electric coupling degradation feature vector.

[0011] As a preferred embodiment of the present invention, S2 specifically includes: S21. Obtain the dielectric-magnetic-electric coupling degradation feature vector, and simultaneously read the ambient light intensity data and the current measured luminous flux data at the output terminal to determine the current adjustment equivalent required to transition from the current output luminous flux level to the target luminous flux. S22. Input the dielectric-magnetic-electric coupling degradation feature vector into the energy-saving-lifetime co-optimization model, and correct the target luminous flux prediction value based on the phosphor dielectric relaxation offset and polarization loss factor. When the dielectric relaxation offset exceeds the threshold, lower the target luminous flux setting value to compensate for luminous efficiency attenuation. S23. Construct a core loss-efficiency mapping surface based on the magnetostriction amplitude index and leakage flux density distribution characteristics, and solve for the minimum energy consumption operating point of the drive circuit. S24. Based on the output of the optimization model, dynamically generate the constant current target value offset sequence and the PWM chopper phase-frequency collaborative migration strategy, and use the two as the current setting target value and the power switch execution parameters, respectively.

[0012] As a preferred embodiment of the present invention, S22 specifically includes: S221. Read the phosphor dielectric relaxation offset and polarization loss factor in the dielectric-magnetic-electric coupling degradation feature vector, and simultaneously obtain the preset dielectric aging alarm threshold and phosphor luminous efficacy decay mapping table. S222. Compare the dielectric relaxation offset with the dielectric aging alarm threshold. If the dielectric relaxation offset is lower than the alarm threshold, keep the current target luminous flux prediction value unchanged. If the dielectric relaxation offset reaches or exceeds the alarm threshold, determine the current luminous efficacy attenuation coefficient of the phosphor by looking up the value of the polarization loss factor in the table. S223. Based on the luminous efficacy attenuation coefficient, recalibrate the target luminous flux energy-saving operating point. Under the premise of ensuring that it is not lower than the minimum lighting requirement constraint, lower the target luminous flux prediction value to the set value corresponding to the energy-saving operating point.

[0013] As a preferred embodiment of the present invention, S23 specifically includes: S231. Read the magnetostriction amplitude index and leakage flux density distribution characteristics in the dielectric-magnetic-electric coupling degradation feature vector, and simultaneously obtain the switching frequency and peak current value of the current constant current drive output. S232. The magnetostriction amplitude exponent is mapped to the eddy current loss component of the magnetic core, and the leakage flux density distribution characteristics are mapped to the hysteresis loss component of the magnetic core. The two are superimposed to construct a two-dimensional response surface of the total loss of the magnetic core with the switching frequency as the abscissa and the peak current as the ordinate. S233. Traverse the two-dimensional response surface to search for the point of minimum loss, and take the switching frequency and peak current corresponding to the minimum point as the lowest energy consumption operating point of the drive circuit.

[0014] As a preferred embodiment of the present invention, S3 specifically includes: S31. Based on the constant current target value offset sequence and the PWM chopper phase-frequency coordinated migration strategy, the duty cycle parameter, switching frequency parameter and gate drive impedance matching parameter of the power switch gate drive signal are synchronously adjusted through the feedforward-feedback composite adjustment network. S32. During the correction operation, continuously monitor the output current ripple spectrum characteristics and the dielectric relaxation offset update amount of the phosphor encapsulation layer surface; S33. The output current ripple spectrum characteristics and dielectric relaxation offset update amount are transmitted back to the frequency domain-time domain joint decoupling module in step S1 in real time to update the frequency domain weight allocation and time domain correlation coefficient of each physical parameter in the dielectric-magnetic-electric coupling degradation feature vector.

[0015] As a preferred embodiment of the present invention, S31 specifically includes: S311. Receive the constant current target value offset sequence and the PWM chopper phase-frequency co-transfer strategy, convert the constant current target value offset sequence into a current setting reference value, and interpret the PWM chopper phase-frequency co-transfer strategy into a switching frequency control word and a phase offset. S312. The current setting reference value is sent to the feedforward path of the feedforward-feedback composite regulation network and directly mapped to the initial duty cycle parameter of the power switch gate drive signal; at the same time, the difference between the actual current sample value and the current setting reference value is sent to the feedback path and generated as a duty cycle correction value after PID compensation. S313. The initial duty cycle parameter is superimposed with the duty cycle correction amount to obtain the final duty cycle. Combined with the switching frequency control word and phase offset, it is synchronously applied to the gate drive circuit of the power switch. At the same time, the output impedance of the drive circuit is adjusted according to the gate drive impedance matching parameter so that the actual current tracks the current target trajectory indicated by the constant current target value offset sequence.

[0016] As a preferred embodiment of the present invention, S33 specifically includes: S331. Compare the output current ripple spectrum characteristics with the preset ripple tolerance spectrum template. If the amplitude of a certain harmonic component exceeds the upper limit of the template, increase the frequency domain weight coefficient of the frequency band corresponding to the harmonic in the broadband impedance spectrum analysis; otherwise, maintain the current weight allocation. S332. Perform Kalman fusion between the updated dielectric relaxation offset and the dielectric relaxation offset in the current eigenvector to update the relaxation time constant and loss factor prediction value in the time domain correlation coefficient. S333. Inject the updated frequency domain weight allocation and time domain correlation coefficient into the frequency domain-time domain joint decoupling module of step S1 for feature extraction during the next round of dimming command execution.

[0017] Compared with the prior art, the present invention has the following advantages: 1. By synchronously sensing multiple physical fields such as phosphor dielectric-photoelasticity, magnetic core magnetostriction-leakage flux, and bus broadband impedance, the limitations of traditional single feedback are overcome, and the collaborative identification of dielectric-magnetic-electric coupling degradation state is achieved, significantly improving the dimension of environmental adaptive sensing.

[0018] 2. Based on the dielectric-magnetic-electric coupling degradation feature vector, the luminous flux energy-saving operating point is calibrated according to the dielectric relaxation offset to delay light decay. The core loss-efficiency mapping surface is constructed to solve the minimum energy consumption operating point, and a PWM phase-frequency migration strategy is generated to avoid the core vibration peak, so as to achieve the synergistic optimization of luminous efficiency compensation, energy efficiency tracking and core stress suppression.

[0019] 3. The duty cycle, switching frequency and gate impedance are synchronously corrected through a feedforward-feedback composite adjustment network, and the current ripple spectrum and dielectric relaxation offset are fed back to update the eigenvector weights, forming a complete closed loop of sensing-optimization-drive-feedback, ensuring control accuracy and energy efficiency stability throughout the entire life cycle. Attached Figure Description

[0020] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating the method described in Embodiment 1 of the present invention.

[0022] Figure 2 This is a framework diagram of the system described in Embodiment 2 of the present invention. Detailed Implementation

[0023] 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.

[0024] The concepts involved in this application will first be described with reference to the accompanying drawings. It should be noted that the following descriptions of various concepts are only for the purpose of making the content of this application easier to understand and do not constitute a limitation on the scope of protection of this application; furthermore, the embodiments and features in the embodiments of this application can be combined with each other unless otherwise specified. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0025] Example 1 like Figure 1 As shown, the present invention provides an environmentally adaptive LED constant current energy-saving control method, comprising the following steps: S1. A dielectric-photoelastic sensing array, a magnetostrictive-leakage flux detection coil, and a broadband impedance spectrum detection unit are respectively arranged on the phosphor encapsulation layer, the inductor core, and the output bus to simultaneously acquire three types of time-series signals and extract the dielectric-magnetic-electric coupling degradation feature vector; specifically including: S11. Arrangement and installation of multiphysics sensing units, specifically as follows: S111. The arrangement of the phosphor encapsulation layer dielectric-optical-elastic coupling sensing array, specifically: The dielectric-optic elastic coupling sensor array is uniformly attached to the surface of the LED phosphor encapsulation layer using thermally conductive insulating adhesive. This thermally conductive insulating adhesive has both thermal conductivity and electrical insulation properties, ensuring a stable thermal coupling and electrical isolation interface between the sensor array and the phosphor encapsulation layer. Specifically, silicon-based thermally conductive gel or polyimide thermally conductive tape is used, with a thermal conductivity of not less than 1.5 W / m·K and a breakdown voltage of not less than 3 kV / mm.

[0026] The dielectric-optical-elastic coupling sensing array consists of multiple interdigital electrode units arranged periodically in a coplanar waveguide structure. The finger width and interfinite spacing of each interdigital electrode unit are designed according to the radio frequency signal transmission characteristics to ensure stable impedance matching characteristics within the target frequency band. A sensitive thin film is coated onto the surface of each interdigital electrode unit using a vacuum evaporation process. The temperature coefficient of the dielectric constant of this sensitive thin film is related to that of YAG:Ce³. + The dielectric relaxation characteristics of the phosphor are matched to ensure a sensitive response to changes in dielectric polarization of the phosphor encapsulation layer under thermal stress. Specifically, barium strontium titanate (BST) or lead zirconate titanate (PZT) ferroelectric thin films are used, with their Curie temperatures matching those of YAG:Ce³. + The phosphor's thermal quenching threshold temperature is matched, and the film thickness ranges from 100 nm to 500 nm; YAG:Ce³ + Yttrium aluminum garnet (YAG) is short for yttrium aluminum garnet doped with trivalent cerium ions. It is currently the most widely used phosphor material in commercial white LEDs.

[0027] The sensor array covers the central light-emitting area and the edge transition area of ​​the phosphor encapsulation layer. The central light-emitting area corresponds to the main light-emitting path of the LED chip, and the edge transition area corresponds to the area where the phosphor coating thickness gradually changes. The spatial distribution information of the dielectric properties of the phosphor is obtained by zoning.

[0028] S112. Arrangement of magnetostriction-leakage flux coupling detection coil on the air gap side of the inductor core, specifically: The magnetostrictive-leakage flux coupling detection coil is fixed to the air gap sidewall of the power inductor core using a high-temperature resistant adhesive, so that the geometric center of the detection coil is aligned with the central axis of the air gap, ensuring the maximum coupling coefficient for the air gap leakage flux.

[0029] The detection coil employs a split-winding structure, with excitation and induction windings wound separately on a core frame made of high-permeability ferrite or amorphous alloy. The two windings are isolated by an insulating layer to reduce mutual capacitance interference. A low-frequency alternating excitation source is connected to the excitation winding, applying a low-frequency alternating magnetic field with a frequency no higher than 1 / 10 of the normal operating switching frequency of the power inductor inside the inductor core. This excites the magnetostrictive vibration effect of the core material, causing mechanical deformation of the core during alternating magnetization. Specifically, the frequency range of the low-frequency alternating excitation source is 100 Hz to 10 kHz (classical value 1 kHz), and the excitation current amplitude range is 10 mA to 200 mA (classical value 50 mA), ensuring that the excitation frequency is far lower than the normal operating switching frequency of the power inductor and higher than the lower limit of the core's mechanical resonant frequency. The induction winding is used to synchronously pick up the harmonic changes in magnetic flux density caused by the vibration of the magnetic core and the induced electromotive force of the leakage flux from the air gap. By analyzing the harmonic amplitude and spectrum distribution of the output signal of the induction winding, the coupling degradation characteristics of magnetic core hysteresis loss and air gap leakage flux are extracted.

[0030] S113. Arrangement of the output bus broadband impedance spectrum detection unit, specifically: A wideband impedance spectrum detection unit is connected to a preset detection node on the output bus of the drive power supply. This preset detection node is located between the output filter capacitor of the constant current drive circuit and the LED load, and is used to obtain the wideband impedance characteristics of the drive circuit at the load end.

[0031] The wideband impedance spectrum detection unit consists of two parts: a wideband current injection probe and a wideband voltage pickup probe. The wideband current injection probe is connected in series to the positive terminal of the output bus. It integrates a wideband power amplifier and a coupling transformer to inject a sweep excitation signal with constant amplitude and linearly varying frequency into the output bus. The sweep range of this excitation signal is 10 Hz to 10 MHz, with a classic value of 100 Hz to 1 MHz. The sweep rate is ten octaves per second or linearly increasing by 100 kHz per second. The injected current amplitude is constant at 10 mA to 100 mA, with a classic value of 50 mA. The wideband voltage pickup probe is connected in parallel between the positive and negative terminals of the same preset detection node. Its front end is equipped with a differential amplifier and an anti-aliasing filter to simultaneously acquire the voltage response signal across the bus while injecting the sweep excitation signal.

[0032] S12. Synchronous acquisition of multi-channel timing signals, specifically: After the dimming command is issued, the central controller synchronously outputs trigger pulses to the three detection channels, initiating a high-precision data acquisition sequence. The three detection channels share the same clock reference source, ensuring that the dielectric relaxation timing signal, magnetostriction harmonic-leakage flux crosstalk timing signal, and broadband impedance spectrum timing signal are strictly aligned on the time axis. This provides a synchronous data foundation for subsequent frequency-time domain joint decoupling. Specifically: S121. Under the control of a synchronous trigger signal, each interdigital electrode unit of the dielectric-optical-elastic coupling sensing array continuously samples the dielectric polarization response of the phosphor encapsulation layer surface at a preset sampling frequency. Each interdigital electrode unit measures the dynamic change of capacitance between its interdigital strips over time, converting the dielectric relaxation process of the phosphor under the coupling of thermal stress and electric field into a capacitance fluctuation signal. After conditioning by a front-end charge amplifier and a low-pass filter, a dielectric relaxation timing signal is formed. The preset sampling frequency ranges from 1kHz to 50kHz, with a classic value of 10kHz.

[0033] S122. Under the control of the synchronous trigger signal, the excitation winding of the magnetostrictive-leakage flux coupling detection coil starts a low-frequency alternating excitation output, establishing a low-frequency alternating magnetization field inside the power inductor core to excite magnetostrictive vibration; the induction winding synchronously enters the sampling state, continuously recording the magnetostrictive harmonic changes caused by the core vibration and the superposition response of the induced electromotive force of the air gap leakage flux at the same preset sampling frequency as the dielectric channel. After bandpass filtering and differential amplification, a magnetostrictive harmonic-leakage flux crosstalk timing signal is formed.

[0034] S123. Under the control of the synchronous trigger signal, the wideband impedance spectrum detection unit starts the frequency sweep sequence of its wideband current injection probe, injecting a constant amplitude AC excitation current into the output bus of the drive power supply point by point from low frequency to high frequency; the wideband voltage pickup probe synchronously collects the voltage response at both ends of the bus at each frequency point, extracts the voltage amplitude and phase information of each frequency point through the lock-in amplifier, and forms a wideband impedance spectrum timing signal after impedance conversion.

[0035] S13. Frequency-time domain joint decoupling and degradation feature coding, specifically: The captured time-series signal is input into a frequency-time joint decoupling module. This module is configured with a dielectric relaxation time-frequency analysis submodule, a magnetic-leakage flux harmonic decoupling submodule, a broadband impedance spectrum decomposition submodule, and a multi-physics feature fusion coding submodule. Specifically: S131. Dielectric relaxation time-frequency analysis submodule processing and phosphor degradation feature extraction, specifically: The dielectric relaxation time-frequency analysis submodule receives the dielectric relaxation time-series signal from the dielectric-optical-elastic coupling sensing array. This submodule first performs a short-time Fourier transform on the dielectric relaxation time-series signal, mapping the one-dimensional time-series data to a time-frequency joint representation space to obtain the spectral distribution of the dielectric response over time. It then locates the center frequency trajectory of the dielectric relaxation peak in the time-frequency spectrum and calculates the phosphor dielectric relaxation offset by comparing the difference between the current center frequency and the initial calibration frequency. Simultaneously, it extracts the dielectric loss tangent value at each moment from the time-frequency spectrum and takes its peak value within the observation window as the polarization loss factor. These two parameters together characterize the degree of dielectric polarization degradation of the phosphor encapsulation layer under thermo-electric coupling stress.

[0036] S132. Processing of the magnetic-leakage flux harmonic decoupling submodule and extraction of core degradation features, specifically: The magneto-leakage flux harmonic decoupling submodule receives the magnetostriction harmonic-leakage flux crosstalk timing signal from the magnetostriction-leakage flux coupling detection coil. This submodule first performs wavelet packet decomposition on the input crosstalk timing signal, dividing the signal into multiple orthogonal frequency band components. It then extracts the signal amplitude from the fundamental frequency component corresponding to the low-frequency excitation, normalizes it, and uses it as the magnetostriction amplitude index to quantify the intensity of the magnetostriction vibration of the magnetic core. The energy spectral density distribution is calculated in the harmonic components of the corresponding harmonic frequency band and used as the leakage flux density distribution characteristic to characterize the spatial diffusion and spectral leakage degree of the air gap leakage flux. These two parameters together characterize the magnetic performance degradation state of the power inductor core under electromagnetic-mechanical coupling stress.

[0037] S133. Broadband impedance spectrum decomposition submodule processing and bus degradation feature extraction, specifically: The broadband impedance spectrum decomposition submodule receives the broadband impedance spectrum timing signal from the broadband impedance spectrum detection unit. This submodule first performs an impedance fitting algorithm on the broadband impedance spectrum timing signal, decomposing the measured impedance spectrum data into a superposition model of resistive and inductive components. Specifically, the impedance fitting algorithm uses a rational function vector fitting algorithm or a least-squares-based equivalent circuit parameter extraction algorithm to fit the measured impedance spectrum into a second-order RLC equivalent circuit model composed of equivalent series resistance, equivalent series inductance, and parasitic parallel capacitance. In the low-frequency band, the slope parameter of the real part of the impedance changing with frequency is extracted and used as the low-frequency impedance increment to characterize the growth trend of the equivalent series resistance of the output bus and filter capacitor in the low-frequency domain. In the high-frequency band, the resonant frequency corresponding to the zero-crossing point of the imaginary part of the impedance is located. By comparing the difference between the current resonant frequency and the initial calibrated resonant frequency, the high-frequency resonant peak offset is calculated to characterize the drift degree of high-frequency parasitic parameters. These two parameters together characterize the electrical performance degradation state of the drive power supply output circuit under electro-thermal coupling stress.

[0038] S134. Multiphysics Feature Fusion Encoding Submodule Processing and Degenerate Feature Vector Generation, specifically: The multi-physics feature fusion encoding submodule receives the phosphor dielectric relaxation offset and polarization loss factor output from the dielectric relaxation time-frequency analysis submodule, the magnetostriction amplitude index and leakage flux density distribution characteristics output from the magnetic-leakage flux harmonic decoupling submodule, and the low-frequency impedance increment and high-frequency resonant peak offset output from the broadband impedance spectrum decomposition submodule. This submodule sequentially arranges these six physical parameters into a one-dimensional array, appending a timestamp of the current data acquisition cycle to the beginning of the array, collectively encoding them to form a standardized one-dimensional feature vector. This vector, as the dielectric-magnetic-electric coupling degradation feature vector, is output to the energy-saving-lifetime co-optimization model, providing a multi-physics fusion state representation for the subsequent generation of constant current energy-saving strategies.

[0039] S2. Input the dielectric-magnetic-electric coupling degradation feature vector into the energy-saving-lifetime co-optimization model, and generate a constant current target value offset sequence and a PWM chopper phase-frequency co-transfer strategy based on the degree of dielectric relaxation offset; specifically including: S21. Multi-source data fusion and current adjustment equivalent calculation, specifically: S211. The central controller establishes a communication link with the frequency domain-time domain joint decoupling module through the internal high-speed data bus, obtains the dielectric-magnetic-electric coupling degradation feature vector from the output of the module according to the preset reading cycle, and temporarily stores the vector in the controller's local register for subsequent calculations; the preset reading cycle ranges from 0.1s to 5s, with a classic value of 1s.

[0040] S212. Synchronously activate the ambient light sensor located on the side wall of the light outlet of the lamp, so that it continuously collects the changes in the light intensity of the external environment of the lamp at a fixed sampling interval to form ambient light intensity data; wherein the fixed sampling interval ranges from 10ms to 500ms, with a classic value of 100ms.

[0041] Simultaneously, the luminous flux detection unit integrated into the backlight channel of the LED light source module is activated. This detection unit receives the light emitted from the LED module through a silicon photodiode array, and outputs the measured luminous flux data at the current output end after transimpedance amplification and analog-to-digital conversion.

[0042] S213. Perform an arithmetic difference calculation between the target luminous flux value set by the user or preset by the system and the measured luminous flux data at the current output terminal to obtain the luminous flux difference required to transition from the current output luminous flux level to the target luminous flux; wherein the preset target luminous flux value ranges from 20% to 100% of the rated luminous flux, with a classic value of 80%.

[0043] Based on the luminous flux and current characteristic curves of the LED light source module as specified at the factory, the difference in luminous flux is mapped to the corresponding current adjustment equivalent. This equivalent, expressed in mA, represents the increment or decrease in driving current required to maintain the target illumination level. The current adjustment equivalent, along with the dielectric-magnetic-electric coupling degradation feature vector and ambient light intensity data, together form the basic input parameter set of the energy-saving-lifetime co-optimization model.

[0044] S22. Dielectric aging determination and target luminous flux correction are as follows: The dielectric-magnetic-electric coupling degradation feature vector is input into the energy-saving-lifetime co-optimization model. This model incorporates a dielectric aging analytical judgment submodule, a luminous efficacy attenuation correction calibration submodule, an energy-saving operating point calibration unit, a core loss mapping construction submodule, and a strategy generation unit. The energy-saving-lifetime co-optimization model uses a multi-objective particle swarm optimization algorithm or the NSGA-II multi-objective evolutionary algorithm as the solution kernel, with LED luminous flux maintenance rate, driver power supply conversion efficiency, and core hysteresis loss increment as constraints, and minimizing the total system energy consumption as the optimization objective. Decision variables include constant current target value offset, PWM switching frequency, and chopper phase offset. A Pareto front solution set is generated through iterative optimization, from which the lowest energy consumption solution satisfying the minimum lighting requirement constraint is selected as the model output. Specifically: S221. Dielectric aging analysis and determination, specifically: The dielectric aging analysis and determination submodule reads the dielectric-magnetic-electric coupling degradation feature vector from the local register of the central controller through the standard data interface, and analyzes the vector to extract the phosphor dielectric relaxation offset in the first dimension and the polarization loss factor in the second dimension.

[0045] Simultaneously, this submodule retrieves a preset dielectric aging alarm threshold from the model's built-in non-volatile parameter storage area. This threshold is based on YAG:Ce³. + The dielectric relaxation limit drift of phosphor under long-term operating conditions at rated junction temperature was experimentally calibrated, with the value ranging from 5% to 25% of the initial center frequency of the phosphor's dielectric relaxation peak, and the classical value being 15%.

[0046] The extracted phosphor dielectric relaxation offset is compared with the dielectric aging alarm threshold. If the dielectric relaxation offset is lower than the alarm threshold, the phosphor dielectric properties are determined to be in the normal degradation range, and a normal status flag is output to the luminous efficiency attenuation correction calibration submodule. If the dielectric relaxation offset reaches or exceeds the alarm threshold, the dielectric aging alarm state is determined, and the aging alarm flag and the current polarization loss factor value are output to the luminous efficiency attenuation correction calibration submodule, triggering the subsequent luminous efficiency attenuation compensation process.

[0047] S222. Optical efficacy attenuation mapping and coefficient determination, specifically: The luminous efficacy attenuation correction calibration submodule receives the status flags and parameter data output from the dielectric aging analysis and judgment submodule: When a normal status indicator is received, this submodule directly keeps the current target optical flux prediction value unchanged and outputs it to the subsequent strategy generation unit.

[0048] When an aging alarm is received, this submodule retrieves the phosphor luminous efficacy decay mapping table from the model's built-in storage area. This mapping table uses the polarization loss factor as the index key and the luminous efficacy decay coefficient as the mapping result. Through linear interpolation, it looks up the table to determine the phosphor's luminous efficacy decay coefficient under the current thermoelectric coupling stress based on the precise value of the current polarization loss factor. This coefficient is defined as the normalized ratio of the phosphor's current quantum efficiency to its initial calibration state, ranging from 0 to 1. A smaller coefficient indicates a more severe degree of luminous efficacy decay. A classic value illustration of the phosphor luminous efficacy decay mapping table is shown in Table 1 below. This mapping table is based on YAG:Ce³ + Experimental calibration of dielectric polarization loss and quantum efficiency decay of phosphors under different thermoelectric coupling stresses was established.

[0049] The determined luminous efficacy attenuation coefficient and aging alarm indicator are output together to the energy-saving operating point calibration unit.

[0050] S223. Energy-saving operating point calibration and target value are lowered, specifically: After receiving the luminous efficacy attenuation coefficient, the energy-saving operating point calibration unit recalibrates the target luminous flux energy-saving operating point based on this coefficient; in specific implementation: The theoretical compensated luminous flux is obtained by multiplying the rated target luminous flux by the luminous efficacy attenuation coefficient. This theoretical compensated luminous flux is then compared with the system's preset minimum lighting requirement constraint value, which is then clamped at an upper limit. The preset minimum lighting requirement constraint value is 10% to 80% of the rated target luminous flux, with a classic value of 30%. If the theoretical compensated luminous flux is higher than the minimum lighting requirement constraint value, the predicted target luminous flux value is lowered to the energy-saving operating point setting value corresponding to that theoretical compensated luminous flux. If the theoretical compensated luminous flux is lower than or equal to the minimum lighting requirement constraint value, the predicted target luminous flux value is maintained at the minimum lighting requirement constraint value and is not lowered further. The final determined target luminous flux setting value is output to the strategy generation unit as the basic constraint condition for the subsequent generation of the constant current target value offset sequence, achieving a lifetime synergistic optimization goal of reducing driving current stress and delaying phosphor thermal quenching rate by moderately lowering the luminous flux.

[0051] S23. Core loss mapping construction and minimum energy consumption operating point solution are as follows: S231. Reading core degradation characteristics and obtaining operating parameters, specifically: The core loss mapping construction submodule sequentially extracts the magnetostriction amplitude index in the third dimension and the leakage flux density distribution characteristics in the fourth dimension. The magnetostriction amplitude index represents the normalized amount of the mechanical vibration amplitude generated by the power inductor core under low-frequency alternating excitation, and the leakage flux density distribution characteristics represent the degree of energy diffusion of leakage flux at the air gap edge in the frequency domain.

[0052] Meanwhile, this submodule obtains the switching frequency and peak current value of the current constant current drive output in real time through the feedback interface of the drive circuit control unit. The switching frequency is directly read from the current configuration value of the carrier frequency register inside the PWM controller, and the peak current value is obtained by inputting the digital result of the analog-to-digital converter after differential amplification by the instrumentation amplifier through the precision current sampling resistor connected in series with the constant current output bus.

[0053] The magnetostriction amplitude index, leakage flux density distribution characteristics, switching frequency, and peak current value are cached together in the local computation buffer of the core loss mapping construction submodule, serving as the basic input dataset for subsequent construction of the mapping relationship between core loss and efficiency.

[0054] S232. Mapping of eddy current hysteresis loss components and construction of two-dimensional response surfaces, specifically: The core loss mapping construction submodule calls the preset eddy current loss conversion coefficient to map the magnetostriction amplitude index to the eddy current loss component of the power inductor core. This eddy current loss conversion coefficient is experimentally calibrated under standard operating conditions based on the volume resistivity of the core ferrite material and the lamination thickness. It reflects the active power consumed by the eddy currents induced by the alternating magnetic flux caused by magnetostriction vibration inside the core. The value ranges from 0.1 W / unit to 10.0 W / unit, with a classic value of 2.0 W / unit. The unit dimension corresponds to the dimensionless normalized quantity of the magnetostriction amplitude index and leakage flux density distribution characteristics.

[0055] Simultaneously, a preset hysteresis loss conversion coefficient is invoked to map the leakage flux density distribution characteristics to the hysteresis loss component of the magnetic core. This hysteresis loss conversion coefficient is experimentally calibrated based on the coercivity and maximum permeability of the magnetic core material under standard operating conditions. It reflects the magnetization work consumed by the irreversible movement of the domain walls caused by the air gap leakage flux. The value ranges from 0.05 W / unit to 5.0 W / unit, with a classical value of 1.0 W / unit. The unit dimension corresponds to the dimensionless normalized quantity of the magnetostriction amplitude exponent and the leakage flux density distribution characteristics.

[0056] The total core loss is obtained by performing a scalar algebraic superposition of the eddy current loss component and the hysteresis loss component. With the switching frequency as the abscissa, the peak current as the ordinate, and the total core loss as the response height, the total core loss is calculated point by point within the allowable adjustment range of the switching frequency and the allowable output range of the peak current according to the principle of equal interval sampling. The surface gaps between the sampling points are filled by the bilinear interpolation algorithm to construct a continuous and smooth two-dimensional response surface of the total core loss.

[0057] S233. Minimum loss traversal search and determination of the lowest energy consumption operating point, specifically: The core loss mapping construction submodule performs a gridded search on the constructed two-dimensional response surface of the total core loss with a preset frequency step along the horizontal axis and a preset current step along the vertical axis. The preset frequency step ranges from 0.5 kHz to 5 kHz, with a classic value of 1 kHz. The preset current step ranges from 10 mA to 200 mA, with a classic value of 50 mA.

[0058] During the traversal, the total core loss value at each grid node is read point by point and compared with the currently recorded minimum loss value in real time. If the current node loss value is smaller, the minimum loss record is updated and the corresponding switching frequency and peak current coordinates of the node are saved. After the full surface traversal is completed, a second local fine search is performed on the neighborhood of the minimum value point to eliminate pseudo-extreme points caused by interpolation errors.

[0059] Finally, the switching frequency and peak current coordinates corresponding to the verified minimum point are output to the strategy generation unit as the lowest energy consumption operating point of the drive circuit. This lowest energy consumption operating point represents the optimal combination of switching frequency and peak current parameters that minimizes the total loss of the power inductor core under the current magnetostriction and leakage flux degradation state of the core.

[0060] S24. Constant current strategy generation and execution parameter output, specifically: Based on the output of the optimization model, a constant current target value offset sequence and a PWM chopper phase-frequency co-transfer strategy are dynamically generated, and the two are used as the current setting target value and the power switch execution parameters, respectively.

[0061] S241. The strategy generation unit receives the target luminous flux setting value corrected in step S22 and the minimum energy consumption operating point solved in step S23, and starts the collaborative calculation process. First, it reads the current adjustment equivalent cached in step S21, and performs a weighted fusion operation on the current adjustment equivalent and the peak current corresponding to the minimum energy consumption operating point. After fusion, a constant current target value offset sequence is generated. The weighting coefficient of the current adjustment equivalent ranges from 0.2 to 0.7, with a classic value of 0.4. The weighting coefficient of the peak current corresponding to the minimum energy consumption operating point ranges from 0.3 to 0.8, with a classic value of 0.6. The sum of the two weighting coefficients is always 1. When the ambient light intensity is lower than a preset threshold... Furthermore, when the low-frequency impedance increment exceeds the aging warning value, the current adjustment equivalent weighting coefficient shifts to the lower limit, and the peak current weighting coefficient at the lowest energy consumption operating point shifts to the upper limit, making the constant current target value more focused on the optimal solution of magnetic core energy efficiency; conversely, the current adjustment equivalent weighting coefficient shifts to the upper limit, prioritizing the rapid tracking of the target light flux; the preset threshold for ambient light intensity is 10% to 30% of the full-scale output value of the ambient light intensity sensor, with a classic value of 20%, corresponding to insufficient indoor natural lighting or nighttime operation; the aging warning value is defined as 50% to 300% of the initial calibrated low-frequency impedance value of the drive power supply output bus, with a classic value of 150% of the initial calibrated low-frequency impedance value.

[0062] The constant current target value offset sequence is in the form of a discrete time series, representing the point-by-point correction trajectory from the current drive current value to the target current value. It serves as the reference current source for outputting the current setting target value of this dimming operation to the constant current drive circuit, and is used to set the digital input code value of the digital-to-analog converter.

[0063] S242. Synchronously, the strategy generation unit reads the switching frequency corresponding to the lowest power consumption operating point and reads the magnetostriction amplitude index in the dielectric-magnetic-electric coupling degradation feature vector as the phase compensation feedback quantity. The PWM chopping phase offset is calculated based on the magnetostriction amplitude index. Specifically: the magnetostriction amplitude index in the dielectric-magnetic-electric coupling degradation feature vector is read, multiplied by a preset phase conversion coefficient to obtain a phase offset reference value, and then this reference value is superimposed with the current peak phase of the magnetostriction harmonic sequence. This causes the PWM carrier trigger edge to lag or lead the peak phase of the magnetostriction vibration by a calculated electrical angle offset, thereby weakening the electromagnetic-mechanical coupling resonance. The preset phase conversion coefficient ranges from 30° to 90° / unit, with a classic value of 60° / unit.

[0064] The frequency shift is calculated based on the switching frequency corresponding to the lowest energy consumption operating point. Specifically: the switching frequency corresponding to the lowest energy consumption operating point is read, and the difference between it and the actual switching frequency of the current constant current drive output is calculated to obtain the frequency difference. The frequency difference is divided by the preset number of shift steps to obtain the frequency shift increment for each step. The frequency shift increment is accumulated point by point according to the discrete time sequence to form a frequency shift sequence that smoothly transitions from the current switching frequency to the target switching frequency. The time interval between each step is an integer multiple of the PWM control cycle, with a value range of 1 ms to 10 ms, and a classic value of 5 ms, to ensure that the output current is continuous without sudden changes during the switching frequency shift. The preset number of shift steps ranges from 5 steps to 50 steps, with a classic value of 20 steps.

[0065] The PWM chopping phase offset and frequency shift are combined and encoded to generate a PWM chopping phase-frequency co-shift strategy. This strategy is output as an execution parameter to the gate driver to drive the power switch and is used to adjust the carrier frequency and trigger phase of the gate drive signal of the power switch.

[0066] The constant current target value offset sequence and the PWM chopper phase-frequency collaborative migration strategy together constitute the final output result of step S2, which is transmitted to step S3 via the data bus as the reference input command for the feedforward-feedback composite regulation network to implement multi-objective collaborative correction.

[0067] S3. Based on the constant current target value offset sequence and the PWM chopper phase-frequency coordinated migration strategy, the duty cycle, switching frequency, and gate impedance are synchronously adjusted through a feedforward-feedback composite adjustment network. The current ripple spectrum and dielectric relaxation offset are then fed back to update the dielectric-magnetic-electric coupling degradation characteristic vector, forming a closed loop; specifically including: S31. Feedforward-feedback composite regulation and multi-parameter collaborative correction, specifically: S311. Policy reception and instruction interpretation, specifically: The feedforward-feedback composite regulation network receives the constant current target value offset sequence and the PWM chopper phase-frequency collaborative migration strategy output from step S2 via an internal high-speed data bus. The network's built-in instruction parsing unit extracts the digital input code of the current point from the constant current target value offset sequence, multiplies the digital input code by the full-scale reference voltage of the digital-to-analog converter, and divides it by the maximum quantization code value of the digital-to-analog converter to obtain the analog voltage setpoint. This analog voltage setpoint is then divided by the product of the resistance value of the precision current sampling resistor and the gain coefficient of the current sampling amplifier to obtain the corresponding analog current setpoint, forming a current setting reference value sequence. This sequence is then written into the reference value buffer register of the feedforward path. The gain coefficient of the current sampling amplifier is collaboratively set based on the LED drive current range, the resistance value of the precision current sampling resistor, and the full-scale input voltage of the subsequent analog-to-digital converter, with a value range of 10 to 500 and a classic value of 100.

[0068] Simultaneously, the instruction parsing unit interprets the PWM chopper phase-frequency coordinated migration strategy, extracts the frequency migration target field and phase offset field from the strategy data frame, converts the frequency migration target field into a switching frequency control word after quantization encoding, and writes it into the frequency configuration register of the PWM carrier generator; and converts the phase offset field into a phase offset amount after angle digital encoding and writes it into the phase accumulator register of the PWM carrier generator.

[0069] The aforementioned current setting reference value, switching frequency control word, and phase offset together constitute the basic instruction set for gate drive adjustment, awaiting call from the feedforward and feedback arithmetic units.

[0070] S312. Feedforward mapping and feedback compensation operations, specifically: The feedforward-feedback composite regulation network reads the current setting reference value from the reference value buffer register. Based on the known circuit parameters of the power switch's on-resistance, output filter inductance, and input DC bus voltage, it directly maps the current setting reference value to the initial duty cycle parameter of the power switch's gate drive signal using the steady-state duty cycle conversion formula of the buck constant current topology. Specifically: the current setting reference value is multiplied by the equivalent series resistance of the output circuit to obtain the resistive voltage drop of the output circuit at the target current; the factory-calibrated forward voltage drop value of the LED light source module string is read, and the resistive voltage drop is added to the forward voltage drop value to obtain the total output voltage requirement required to maintain the target current; this total output voltage requirement is divided by the input DC bus voltage, and the resulting ratio is the initial duty cycle parameter of the power switch's gate drive signal. This initial duty cycle parameter characterizes the theoretical conduction duty cycle required to achieve the target current under ideal, undisturbed operating conditions.

[0071] Simultaneously, the feedback path obtains the actual current sampling value through a precision current sampling resistor connected in series with the constant current output bus. After differential amplification and analog-to-digital conversion by an instrumentation amplifier, the digitized actual current value is obtained. The arithmetic difference between this actual current value and the current setting reference value is calculated to obtain the current tracking error. This current tracking error is fed into a proportional-integral-derivative (PID) compensation network, sequentially undergoing proportional coefficient multiplication in the proportional stage, error accumulation and summation in the integral stage, and error change rate calculation in the derivative stage. The outputs of the three stages are then algebraically superimposed to generate the duty cycle correction. Specifically, in the proportional stage, the current tracking error is multiplied by a preset proportional coefficient to obtain the proportional stage output value. The value of this proportional coefficient ranges from 0.001 / A. -1 up to 0.1 / A -1 The classic value is 0.01 / A. -1In the integration stage, the current tracking error of the current sampling period is multiplied by the sampling period of the feedback path, and then multiplied by a preset integration coefficient to obtain the current integration increment. The value of the integration coefficient ranges from 0.0001 / A / s to 0.01 / A / s, with a classic value of 0.001 / A / s. This integration increment is added to the cumulative integration value saved in the previous sampling period, and the updated cumulative integration value is the output value of the integration stage. In the differentiation stage, the current tracking error of the current sampling period is subtracted from the current tracking error of the previous sampling period to obtain the error change. This error change is divided by the sampling period and then multiplied by a preset differentiation coefficient to obtain the output value of the differentiation stage. The value of the differentiation coefficient ranges from 1 / 100000s / A to 1 / 1000s / A, with a classic value of 1 / 10000s / A. Finally, the output values ​​of the proportional stage, the integration stage, and the differentiation stage are arithmetically added together, and the result is the duty cycle correction. This duty cycle correction amount represents the amount of duty cycle fine-tuning required to eliminate the deviation between the actual current and the target current.

[0072] S313. Multi-parameter coordinated application and impedance matching adjustment, specifically: The feedforward-feedback composite control network performs an arithmetic superposition of the initial duty cycle parameter output from the feedforward path and the duty cycle correction amount output from the feedback path to obtain the final duty cycle command value. The network loads this final duty cycle command value into the duty cycle register of the PWM comparator, and simultaneously loads the switching frequency control word decoded in step S311 into the frequency configuration register of the PWM carrier generator, and loads the phase offset into the phase offset register of the PWM carrier generator. This enables the PWM controller to synchronously output a pulse width modulation signal with a specified duty cycle, switching frequency, and carrier phase. This pulse width modulation signal, after level conversion and power amplification by the gate driver, is applied to the gate electrode of the power switch transistor, controlling the turn-on and turn-off timing of the power switch transistor.

[0073] Simultaneously, based on preset gate drive impedance matching parameters, the network adjusts the output impedance of the gate drive circuit by adjusting the resistance value of the adjustable resistor array inside the output stage of the gate driver or the external series gate resistor. This achieves impedance matching between the leading and trailing edge slopes of the gate drive signal and the input capacitance and parasitic inductance of the power switch, suppressing voltage overshoot and ringing during the switching transient process. Consequently, the actual current output by the constant current drive accurately tracks the current target trajectory indicated by the constant current target value offset sequence. The preset gate drive impedance matching parameters are determined by looking up a table or using linear interpolation based on the product of the current switching frequency and the input capacitance of the power switch. This ensures that the gate loop time constant is within the critical damping range, with a value ranging from 1Ω to 100Ω, and a classic value of 10Ω.

[0074] S32. Multiphysics state monitoring during the correction process, specifically: Throughout the multi-parameter collaborative correction operation in step S31, the central controller synchronously starts two independent monitoring channels and shares the same clock reference: The first monitoring channel acquires the actual current signal through a precision current sampling resistor connected in series with the constant current drive output bus. After the signal is suppressed by the front-end anti-aliasing filter to suppress high-frequency aliasing interference, it is sent to the high-speed analog-to-digital converter for digital processing. The converted digital signal is input to the digital signal processing unit, and the output current ripple spectrum features are extracted frame by frame through the fast Fourier transform algorithm. The amplitude and phase information of each harmonic component are recorded point by point to form a spectrum feature data frame.

[0075] The second monitoring channel continuously acquires dielectric polarization response signals through a dielectric-optical-elastic coupling sensor array attached to the surface of the phosphor encapsulation layer. After impedance transformation and signal conditioning by a front-end charge amplifier, the signal is sent to an analog-to-digital converter. The center frequency of the current dielectric relaxation peak is extracted by short-time Fourier transform. The difference between this center frequency and the initial calibration center frequency is calculated to obtain the dielectric relaxation offset update amount.

[0076] The aforementioned output current ripple spectrum characteristics and dielectric relaxation offset update amount serve as the closed-loop feedback data source for step S3, are cached in real time in the shared data buffer of the central controller, and wait to be fed back to step S1 through an interrupt triggering mechanism.

[0077] S33. Closed-loop backpropagation and feature vector weight update, specifically: S331. Frequency domain weight allocation update, specifically: The frequency-time domain joint decoupling module incorporates a frequency domain weight allocation and update unit. This unit receives the output current ripple spectrum characteristics returned from step S32 via a high-speed data bus. First, this unit retrieves a preset ripple tolerance spectrum template from the non-volatile parameter storage area. This template specifies the upper limit of the allowable amplitude for each harmonic component. Specifically, the upper limit of the allowable amplitude for the fundamental component is 5% to 30% of the rated output current, with a classic value of 15%; the upper limit of the allowable amplitude for the second harmonic component is 3% to 20% of the rated output current, with a classic value of 8%; and the upper limit of the allowable amplitude for the third and higher harmonic components is 1% to 10% of the rated output current, with a classic value of 5%.

[0078] Subsequently, the output current ripple spectrum characteristics are compared with the ripple tolerance spectrum template at each frequency point, and the frequency coordinates corresponding to each harmonic component are traversed. If the measured amplitude of a certain harmonic component exceeds the upper limit of the template, it is determined that there is abnormal ripple energy accumulation in that frequency band. At this time, the frequency domain weight coefficient of that frequency band is increased by a preset incremental step size, which is 10% to 50% of the current weight, with a classic value of 20%, until the weight of that frequency band reaches the upper limit saturation value. The upper limit saturation value is 200% to 500% of the initial weight of that frequency band, with a classic value of 300%, to prevent the excessive expansion of the weight of a single frequency band from causing the degradation features of other frequency bands to be submerged. This makes the feature extraction in the next S1 step more focused on the abnormal frequency band, improving the sensitivity to bus impedance aging and harmonic resonance. If the measured amplitude does not exceed the upper limit of the template, the weight allocation of the current frequency band remains unchanged.

[0079] After completing the full frequency band traversal, the updated frequency domain weight coefficient matrix is ​​written into the weight configuration register inside the module, waiting to be loaded and applied in the next round of feature extraction.

[0080] S332. Update of temporal correlation coefficients, specifically: a. The frequency-time domain joint decoupling module has a built-in time-domain correlation coefficient update unit, which receives the dielectric relaxation offset update from step S32. This unit reads the existing dielectric relaxation offset from the current dielectric-magnetic-electric coupling degradation feature vector as a priori estimate, and uses the updated dielectric relaxation offset as the observation. It then performs optimal weighted fusion of the two through Kalman fusion computation. Internally, this unit presets the process noise covariance and the observation noise covariance. The process noise covariance characterizes the uncertainty of the priori estimate due to the randomness of phosphor aging, while the observation noise covariance characterizes the uncertainty caused by the measurement noise of the dielectric-optical-elastic coupling sensor array. Specifically, this unit calculates the Kalman gain for the current period by calculating the ratio of the process noise covariance to the observation noise covariance, combined with the estimation error covariance of the previous period. The magnitude of the Kalman gain reflects the reliability of the observation relative to the priori estimate; the smaller the observation noise, the larger the gain. This unit subtracts the prior estimate from the observed value to obtain the deviation, multiplies the deviation by the Kalman gain to obtain the optimal correction, and then adds the optimal correction to the prior estimate to obtain the optimal estimate of the fused dielectric relaxation offset. At the same time, this unit updates the estimation error covariance matrix based on the Kalman gain and saves the updated error covariance to an internal register for use in the next sampling cycle.

[0081] b. Based on the fused dielectric relaxation offset, the time-domain correlation coefficient update unit extracts historical data points of the dielectric relaxation offset from the most recent consecutive sampling periods from the internal circular data buffer to construct a time-series sample set. This unit assumes that the dielectric relaxation offset evolves exponentially with time, and takes a natural logarithmic transformation on each data point in the time-series sample set to convert the exponential decay relationship into a linear decay relationship. The unit then uses the least squares method to fit a straight line to the transformed linear data points, solves for the slope coefficient of the fitted line, and takes the negative reciprocal of this slope coefficient to obtain the updated relaxation time constant. This relaxation time constant characterizes the characteristic time scale required for the phosphor's dielectric polarization response to decay from the excited state to the steady-state polarization state, reflecting the degree to which phosphor lattice defects hinder dipole orientation polarization under current thermal stress. The unit writes the updated relaxation time constant into the time-domain correlation coefficient storage area, replacing the old value from the previous period.

[0082] c. The time-domain correlation coefficient update unit calls the internally pre-stored coupled mapping model of dielectric relaxation offset and polarization loss factor. This model is established based on Debye relaxation theory or empirical curves calibrated at the factory, describing the evolution trend of polarization loss factor as the dielectric relaxation offset increases. This unit inputs the fused and updated dielectric relaxation offset into this coupled mapping model, which includes four functional layers: input interface layer, Debye relaxation core mapping layer, phosphor defect correction layer, and output interface layer.

[0083] The input interface layer receives the dielectric relaxation offset from the time-domain correlation coefficient update unit after Kalman fusion. The offset is normalized to convert it into a percentage offset relative to the initial calibration center frequency. At the same time, the validity of the value is checked, and abnormal jump values ​​caused by instantaneous failure of the sensor array are removed to ensure that the data entering the core mapping layer is within a reasonable evolution range.

[0084] The Debye relaxation core mapping layer incorporates a dielectric response transfer function based on Debye relaxation theory. This transfer function uses the dielectric relaxation offset as an input variable and calculates the real and imaginary parts of the dielectric constant under the current aging state through a complex dielectric constant frequency domain distribution model. In practice, the dielectric relaxation offset is substituted into the relaxation time constant term in the Debye relaxation equation to solve for the ratio of the imaginary to real parts of the complex dielectric constant at the characteristic frequency. This ratio is then used to derive the theoretical polarization loss factor benchmark value after arctangent calculation, characterizing the intrinsic energy dissipation level of phosphor dielectric polarization under ideal lattice conditions.

[0085] Phosphor defect correction layer with built-in YAG:Ce³ +A correction coefficient library for phosphor lattice defect density and trap energy level distribution is constructed. This library is divided into multiple nonlinear correction intervals based on the dielectric relaxation shift range at different aging stages. When the dielectric relaxation shift falls into a certain interval, the model automatically calls the defect correction coefficients for that interval to perform nonlinear weighted correction on the theoretical baseline value output by the Debye relaxation core mapping layer. The correction process compensates for lattice distortion, oxygen vacancy accumulation, and Ce³⁺. + The additional dipole orientation polarization energy dissipation caused by changes in the local environment of the luminescent center makes the output result closer to the actual thermoelectric coupling degradation state of the phosphor encapsulation layer.

[0086] The output interface layer outputs the value processed by the defect correction layer as the theoretical expected value of the polarization loss factor.

[0087] d. This unit reads the polarization loss factor from the current dielectric-magnetic-electric coupling degradation feature vector as the current actual value, and performs a difference calculation between the theoretical expected value and the current actual value. If the absolute value of the difference exceeds a preset tolerance threshold, it indicates a mismatch between the actual loss factor and the dielectric relaxation state. In this case, the unit recalibrates the predicted loss factor value based on the inverse function relationship of the coupling mapping model and the dielectric relaxation offset, restoring the synchronous evolution relationship between the predicted loss factor value and the dielectric relaxation offset. If the difference does not exceed the tolerance, the original predicted loss factor value remains unchanged, and the corrected predicted loss factor value is finally written into the time-domain correlation coefficient storage area. The preset tolerance threshold ranges from 0.01 to 0.05, with a classic value of 0.02.

[0088] S333. Parameter injection and closed-loop initialization, specifically: The frequency-time joint decoupling module injects the frequency-domain weight coefficient matrix updated in step S331 and the time-domain correlation coefficient updated in step S332 into the module's initialization parameter area. This initialization parameter area is located in a non-volatile storage medium and is used to save the final state configuration of this round of closed-loop operation.

[0089] When the next dimming command is issued, the frequency-time joint decoupling module directly reads the updated frequency domain weight allocation and time domain correlation coefficients from the initialization parameter area. These serve as the initial weights and correlation benchmarks for the frequency-time joint decoupling operation of the new round of dielectric relaxation timing signal, magnetostriction harmonic-leakage flux crosstalk timing signal, and broadband impedance spectrum timing signal. This enables real-time reverse driving of the sensing front-end feature extraction strategy by the output current ripple spectrum characteristics and dielectric relaxation offset update, forming an environmental adaptive constant current energy-saving closed loop covering the entire process of sensing, optimization, driving, and data transmission.

[0090] Example 2 like Figure 2As shown, an environment-adaptive LED constant current energy-saving control system is used to implement an environment-adaptive LED constant current energy-saving control method, comprising: The dielectric-magnetic-electric coupling sensing module is implemented using a multi-channel synchronous acquisition board based on an FPGA or ARM Cortex-M series microcontroller. It is used to attach a dielectric-optic elastic coupling sensing array to the surface of the LED phosphor encapsulation layer, arrange a magnetostrictive-leakage flux coupling detection coil on the air gap sidewall of the power inductor core, and arrange a broadband impedance spectrum detection unit at a preset detection node on the output bus of the drive power supply. It synchronously acquires dielectric relaxation timing signals, magnetostrictive harmonic-leakage flux crosstalk timing signals, and broadband impedance spectrum timing signals, and extracts dielectric-magnetic-electric coupling degradation feature vectors through frequency-time domain joint decoupling.

[0091] The energy-saving and lifespan co-optimization module, connected to the dielectric-magnetic-electric coupling sensing module, is deployed in the embedded software layer of the microcontroller. It is used to obtain the dielectric-magnetic-electric coupling degradation feature vector, correct the target luminous flux setting value based on the phosphor dielectric relaxation offset and the polarization loss factor, construct the core loss-efficiency mapping surface based on the magnetostriction amplitude index and leakage flux density distribution characteristics, solve the minimum energy consumption operating point, and dynamically generate the constant current target value offset sequence and PWM chopper phase-frequency co-transfer strategy. The multi-target drive collaborative correction and closed-loop feedback module, connected to the energy-saving-lifetime collaborative optimization module, is implemented by a dedicated LED driver chip integrating a gate driver and a PWM controller. It is used to synchronously adjust the duty cycle parameters, switching frequency parameters, and gate drive impedance matching parameters of the power switch gate drive signal through a feedforward-feedback composite adjustment network based on the constant current target value offset sequence and the PWM chopping phase-frequency collaborative migration strategy. It also feeds back the output current ripple spectrum characteristics and dielectric relaxation offset update to the dielectric-magnetic-electric coupling sensing module to update the frequency domain weight allocation and time domain correlation coefficient of the dielectric-magnetic-electric coupling degradation feature vector, forming an environmentally adaptive constant current energy-saving closed loop.

[0092] The three modules are interconnected via SPI or CAN bus.

[0093] As can be seen from the above description, the embodiments of the present invention achieve the following technical effects: By attaching a dielectric-optical elastic coupling sensing array to the phosphor encapsulation layer, arranging a magnetostrictive-leakage flux coupling detection coil on the air gap side of the inductor core, and arranging a broadband impedance spectrum detection unit on the output bus, three types of timing signals—dielectric relaxation, magnetostrictive harmonic-leakage flux crosstalk, and broadband impedance—are simultaneously acquired. The dielectric-magnetic-electric coupling degradation feature vector is extracted through joint frequency-time domain decoupling. This overcomes the limitation of traditional LED driving relying solely on single temperature or photosensitive feedback, and achieves multi-physics field collaborative identification of phosphor thermoelectric coupling degradation, magnetic core electromagnetic-mechanical coupling degradation, and bus electrical aging status. This significantly improves the environmental adaptive sensing dimension and the accuracy of degradation status assessment.

[0094] The dielectric-magnetic-electric coupling degradation feature vector is input into the energy-saving-lifetime co-optimization model. Based on the phosphor dielectric relaxation offset, the target luminous flux energy-saving operating point is actively calibrated and appropriately lowered to compensate for luminous efficiency decay and delay the phosphor thermal quenching rate. At the same time, a core loss-efficiency mapping surface is constructed based on the magnetostriction amplitude exponent and leakage flux density distribution characteristics to solve for the minimum energy consumption operating point. A PWM chopper phase-frequency co-transfer strategy is generated to ensure that the chopper phase avoids the magnetostriction vibration peak. This achieves multi-objective co-optimization of luminous efficiency compensation, energy efficiency tracking, and core stress suppression, effectively balancing energy saving and drive power supply lifespan.

[0095] By synchronously adjusting the duty cycle, switching frequency, and gate drive impedance matching parameters of the power switch gate drive signal through a feedforward-feedback composite regulation network, the constant current output accurately tracks the target trajectory and reduces transient electromagnetic interference during switching. At the same time, the output current ripple spectrum characteristics and dielectric relaxation offset update are fed back to the sensing front end in real time to update the frequency domain weight allocation and time domain correlation coefficient of the dielectric-magnetic-electric coupling degradation feature vector. This allows the feature extraction strategy to adaptively evolve with the circuit aging process, forming an environmentally adaptive constant current energy-saving closed loop covering the entire process of sensing, optimization, driving, and feedback, ensuring control accuracy and energy efficiency stability throughout the entire life cycle.

[0096] The embodiments and / or implementation methods described above are merely preferred embodiments and / or implementation methods for implementing the technology of the present invention, and are not intended to limit the implementation methods of the technology of the present invention in any way. Any person skilled in the art may make some modifications or alterations to other equivalent embodiments without departing from the scope of the technical means disclosed in the present invention, but these should still be regarded as the technology or embodiments that are substantially the same as the present invention.

[0097] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. The above descriptions are only preferred embodiments of this application. It should be noted that due to the limitations of written expression, while there are objectively infinite specific structures, those skilled in the art can make several improvements, modifications, or changes without departing from the principles of this application, and can also combine the above technical features in an appropriate manner. These improvements, modifications, changes, or combinations, or the direct application of the inventive concept and technical solution to other situations without modification, should all be considered within the scope of protection of this application.

Claims

1. An environmentally adaptive LED constant current energy-saving control method, characterized in that, include: A dielectric-photoelastic sensing array, a magnetostrictive-leakage flux detection coil, and a broadband impedance spectrum detection unit are respectively arranged on the phosphor encapsulation layer, the inductor core, and the output bus. Three types of time-series signals are acquired simultaneously, and the dielectric-magnetic-electric coupling degradation feature vector is extracted. The dielectric-magnetic-electric coupling degradation feature vector is input into the energy-saving-lifetime co-optimization model, and a constant current target value offset sequence and PWM chopper phase-frequency co-transfer strategy are generated based on the degree of dielectric relaxation offset. Based on the constant current target value offset sequence and the PWM chopper phase-frequency collaborative migration strategy, the duty cycle, switching frequency and gate impedance are synchronously adjusted through the feedforward-feedback composite adjustment network, and the current ripple spectrum and dielectric relaxation offset are back-transmitted to update the dielectric-magnetic-electric coupling degradation feature vector. The dielectric-photoelastic sensing array, magnetostrictive-leakage flux detection coil, and broadband impedance spectrum detection unit are respectively arranged on the phosphor encapsulation layer, the inductor core, and the output bus to simultaneously acquire three types of time-series signals and extract the dielectric-magnetic-electric coupling degradation feature vector, specifically including: A dielectric-optic elastic coupling sensor array is attached to the surface of the LED phosphor encapsulation layer, a magnetostrictive-leakage flux coupling detection coil is arranged on the air gap sidewall of the power inductor core, and a broadband impedance spectrum detection unit is arranged at a preset node of the drive power output bus. After the dimming command is issued, three detection channels are started simultaneously to continuously capture dielectric relaxation timing signal, magnetostriction harmonic-leakage flux crosstalk timing signal and broadband impedance spectrum timing signal. The captured time-series signal is input into the frequency-time domain joint decoupling module, from which the phosphor dielectric relaxation offset and polarization loss factor, magnetostriction amplitude index and leakage flux density distribution characteristics, low-frequency impedance increment and high-frequency resonant peak offset are extracted and encoded into dielectric-magnetic-electric coupling degradation feature vector. The dielectric-magnetic-electric coupling degradation feature vector is input into the energy-saving-lifetime co-optimization model. Based on the degree of dielectric relaxation offset, a constant current target value offset sequence and a PWM chopper phase-frequency co-transfer strategy are generated, specifically including: The dielectric-magnetic-electric coupling degradation feature vector is obtained, and the ambient light intensity data and the current measured luminous flux data at the output terminal are read simultaneously to determine the current adjustment equivalent required to transition from the current output luminous flux level to the target luminous flux. The dielectric-magnetic-electric coupling degradation feature vector is input into the energy-saving-lifetime co-optimization model. The target luminous flux prediction value is corrected based on the phosphor dielectric relaxation offset and the polarization loss factor. When the dielectric relaxation offset exceeds the threshold, the target luminous flux setting value is lowered. Based on the magnetostriction amplitude index and leakage flux density distribution characteristics, a core loss-efficiency mapping surface is constructed to solve for the minimum energy consumption operating point of the drive circuit. Based on the output of the optimization model, a constant current target value offset sequence and a PWM chopper phase-frequency co-transfer strategy are dynamically generated, and the two are used as the current setting target value and the power switch execution parameters, respectively.

2. The environmentally adaptive LED constant current energy-saving control method according to claim 1, characterized in that, The process of attaching a dielectric-optical elastic coupling sensor array to the surface of the LED phosphor encapsulation layer, arranging a magnetostrictive-leakage flux coupling detection coil on the air gap sidewall of the power inductor core, and arranging a broadband impedance spectrum detection unit at a preset node of the drive power supply output bus specifically includes: A dielectric-optical-elastic coupling sensing array is attached to the surface of a phosphor encapsulation layer using thermally conductive insulating adhesive. The sensing array consists of multiple interdigitated electrode units arranged in a coplanar waveguide structure, and the surface of each interdigitated electrode unit is coated with a coating similar to YAG:Ce³. + A sensitive thin film with matching phosphor dielectric properties, and the attachment range covers the central luminescent area and the edge transition area of ​​the phosphor encapsulation layer; A magnetostrictive-leakage flux coupling detection coil is fixed to the air gap sidewall of a power inductor core. The detection coil consists of an excitation winding and an induction winding wound on a high permeability core. The excitation winding is used to apply low-frequency alternating excitation to excite magnetostrictive vibration, and the induction winding is used to pick up harmonic changes and the induced electromotive force of air gap leakage flux. A broadband impedance spectrum detection unit is connected to a preset detection node on the output bus of the drive power supply. The detection unit includes a broadband current injection probe connected in series with the bus and a broadband voltage pickup probe connected in parallel with the bus, which are used to inject a sweep frequency excitation signal and synchronously acquire a voltage response signal, respectively.

3. The environmentally adaptive LED constant current energy-saving control method according to claim 1, characterized in that, The captured time-series signal is input into the frequency-time domain joint decoupling module, from which the phosphor dielectric relaxation offset and polarization loss factor, magnetostriction amplitude index and leakage flux density distribution characteristics, low-frequency impedance increment and high-frequency resonant peak offset are extracted and encoded into a dielectric-magnetic-electric coupling degradation feature vector, specifically including: The dielectric relaxation timing signal is converted to the time-frequency domain by short-time Fourier transform. The center frequency offset of the dielectric relaxation peak is extracted as the phosphor dielectric relaxation offset, and the peak value of the dielectric loss tangent is extracted as the polarization loss factor. The magnetostrictive harmonic-leakage flux crosstalk time-series signal is decomposed into multiple frequency band components by wavelet packet decomposition. The amplitude is extracted from the fundamental frequency component as the magnetostrictive amplitude index, and the energy spectral density is extracted from the harmonic component as the leakage flux density distribution characteristic. The broadband impedance spectrum time series signal is decomposed into the slope of the real part of the impedance in the low frequency band as the low frequency impedance increment, and the resonant frequency offset in the high frequency band as the high frequency resonant peak offset by the impedance fitting algorithm. The phosphor dielectric relaxation offset, polarization loss factor, magnetostriction amplitude index, leakage flux density distribution characteristics, low-frequency impedance increment, and high-frequency resonance peak offset are arranged in a predetermined dimension order and marked with a timestamp. They are then encoded together into a one-dimensional feature vector, which serves as the dielectric-magnetic-electric coupling degradation feature vector.

4. The environmentally adaptive LED constant current energy-saving control method according to claim 1, characterized in that, The dielectric-magnetic-electric coupling degradation feature vector is input into the energy-saving-lifetime co-optimization model. The target luminous flux prediction value is corrected based on the phosphor dielectric relaxation offset and polarization loss factor. When the dielectric relaxation offset exceeds the threshold, the target luminous flux setting value is lowered, specifically including: Read the phosphor dielectric relaxation offset and polarization loss factor in the dielectric-magnetic-electric coupling degradation feature vector, and simultaneously obtain the preset dielectric aging alarm threshold and phosphor luminous efficacy decay mapping table. The dielectric relaxation offset is compared with the dielectric aging alarm threshold. If the dielectric relaxation offset is lower than the alarm threshold, the current target luminous flux prediction value remains unchanged. If the dielectric relaxation offset reaches or exceeds the alarm threshold, the current luminous efficacy attenuation coefficient of the phosphor is determined by looking up the value of the polarization loss factor in a table. Based on the luminous efficacy attenuation coefficient, the target luminous flux energy-saving operating point is recalibrated. Under the premise of ensuring that it is not lower than the minimum lighting requirement constraint, the predicted value of the target luminous flux is lowered to the set value corresponding to the energy-saving operating point.

5. The environmentally adaptive LED constant current energy-saving control method according to claim 4, characterized in that, The process involves constructing a core loss-efficiency mapping surface based on the magnetostriction amplitude exponent and leakage flux density distribution characteristics to solve for the minimum energy consumption operating point of the drive circuit. Specifically, this includes: Read the magnetostriction amplitude index and leakage flux density distribution characteristics in the dielectric-magnetic-electric coupling degradation feature vector, and simultaneously obtain the switching frequency and peak current value of the current constant current drive output; The magnetostriction amplitude exponent is mapped to the eddy current loss component of the magnetic core, and the leakage flux density distribution characteristics are mapped to the hysteresis loss component of the magnetic core. The two are superimposed to construct a two-dimensional response surface of the total loss of the magnetic core with the switching frequency as the abscissa and the peak current as the ordinate. The minimum loss point is searched by traversing the two-dimensional response surface, and the switching frequency and peak current corresponding to the minimum point are taken as the lowest energy consumption operating point of the drive circuit.

6. The environmentally adaptive LED constant current energy-saving control method according to claim 5, characterized in that, Based on the constant current target value offset sequence and the PWM chopper phase-frequency co-transfer strategy, the duty cycle, switching frequency, and gate impedance are synchronously adjusted through a feedforward-feedback composite adjustment network. The current ripple spectrum and dielectric relaxation offset are then fed back to update the dielectric-magnetic-electric coupling degradation feature vector, specifically including: Based on the constant current target value offset sequence and the PWM chopper phase-frequency coordinated migration strategy, the duty cycle parameter, switching frequency parameter and gate drive impedance matching parameter of the power switch gate drive signal are synchronously adjusted through the feedforward-feedback composite adjustment network. During the correction operation, the output current ripple spectrum characteristics and the dielectric relaxation offset update amount of the phosphor encapsulation layer surface are continuously monitored. The output current ripple spectrum characteristics and dielectric relaxation offset update are transmitted back to the frequency domain-time domain joint decoupling module in real time to update the frequency domain weight allocation and time domain correlation coefficient of each physical parameter in the dielectric-magnetic-electric coupling degradation feature vector.

7. The environmentally adaptive LED constant current energy-saving control method according to claim 6, characterized in that, Based on the constant current target value offset sequence and the PWM chopper phase-frequency co-transfer strategy, the duty cycle parameter, switching frequency parameter, and gate drive impedance matching parameter of the power switch gate drive signal are synchronously adjusted through a feedforward-feedback composite adjustment network, specifically including: Receive the constant current target value offset sequence and the PWM chopper phase-frequency co-transfer strategy, convert the constant current target value offset sequence into a current setting reference value, and interpret the PWM chopper phase-frequency co-transfer strategy into a switching frequency control word and a phase offset. The current setting reference value is sent to the feedforward path of the feedforward-feedback composite regulation network and directly mapped to the initial duty cycle parameter of the power switch gate drive signal; at the same time, the difference between the actual current sample value and the current setting reference value is sent to the feedback path and generated as a duty cycle correction value after PID compensation. The initial duty cycle parameter is superimposed with the duty cycle correction amount to obtain the final duty cycle. Combined with the switching frequency control word and phase offset, it is synchronously applied to the gate drive circuit of the power switch. At the same time, the output impedance of the drive circuit is adjusted according to the gate drive impedance matching parameter so that the actual current tracks the current target trajectory indicated by the constant current target value offset sequence.

8. The environmentally adaptive LED constant current energy-saving control method according to claim 7, characterized in that, The output current ripple spectrum characteristics and dielectric relaxation offset update are fed back to the frequency-time joint decoupling module in real time to update the frequency domain weight allocation and time domain correlation coefficient of each physical parameter in the dielectric-magnetic-electric coupling degradation feature vector, specifically including: The output current ripple spectrum characteristics are compared with the preset ripple tolerance spectrum template. If the amplitude of a certain harmonic component exceeds the upper limit of the template, the frequency domain weight coefficient of the frequency band corresponding to the harmonic is increased in the broadband impedance spectrum analysis; otherwise, the current weight allocation is maintained. The dielectric relaxation offset update is fused with the dielectric relaxation offset in the current feature vector using Kalman fusion to update the relaxation time constant and loss factor prediction values ​​in the time domain correlation coefficient. The updated frequency domain weight allocation and time domain correlation coefficient are injected into the frequency domain-time domain joint decoupling module for feature extraction during the next round of dimming command execution.

Images