LED lamp drive circuit control method for preventing stroboscopic effect
By adopting low-voltage start-up mechanism, overvoltage protection and adaptive fuzzy control, quantum optimization and machine learning algorithms in the LED lamp driving circuit, the strobe effect problems caused by power supply fluctuations and load changes are solved, and efficient and stable light is achieved Output and system reliability.
Patent Information
- Application Number
- CN202411267522.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-11
- Publication Date
- 2024-11-22
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing LED lamp driving circuits have insufficient response speed when power supply fluctuations and load changes, poor anti-interference ability, high cost and complexity, and difficult to dynamically adjust, resulting in the occurrence of strobe effects.
Using low voltage start-up mechanism, overvoltage and overcurrent protection functions, combined with adaptive fuzzy control, quantum optimization and machine learning algorithms, graph convolutional neural networks are used through multi-band filtering technology, empirical modal decomposition and Hilbert transformation. Quantum phase interference technology, real-time monitoring and dynamic adjustment of driving parameters, predicting and eliminating strobe effects caused by power supply fluctuations.
It realizes high stability of LED lamp light output, strong system reliability and real-time response capabilities, significantly reduces the occurrence of strobe effects, and improves the overall performance and user experience of the system.
Smart Images

Figure CN119012455A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of lighting technology, and in particular to a method for controlling an LED lamp driving circuit for preventing a stroboscopic effect. Background Art
[0002] The main defects include the following aspects:
[0003] Insufficient response speed: Traditional methods such as linear regulators, switching power supplies and constant current drives have insufficient response speed when the power supply fluctuates greatly, making it difficult to quickly adjust the drive parameters, resulting in the occurrence of stroboscopic effects.
[0004] Poor anti-interference ability: Although technologies such as high-frequency switching and PWM dimming have improved efficiency and dimming effects, the high-frequency noise and electromagnetic interference introduced have an adverse effect on the stable output of LED lamps.
[0005] High cost and complexity: Although digital control systems can achieve precise regulation, the high system complexity and cost limit their popularity in large-scale applications.
[0006] Limited dynamic adjustment capability: Existing technologies make it difficult to achieve dynamic real-time adjustment of power supply fluctuations and load changes, and are unable to effectively eliminate the stroboscopic effect caused by them.
[0007] Therefore, how to provide a method for controlling an LED lamp driving circuit for preventing the stroboscopic effect is an urgent problem to be solved by those skilled in the art. Summary of the invention
[0008] One object of the present invention is to propose a method for controlling a LED lamp driving circuit for preventing the stroboscopic effect. The present invention adopts a low voltage starting mechanism, overvoltage and overcurrent protection functions, and combines adaptive fuzzy control, quantum optimization and machine learning algorithms, and describes in detail a method for dynamically adjusting driving parameters, predicting and eliminating the stroboscopic effect caused by power supply fluctuations, which has the advantages of high light output stability, strong system reliability and strong real-time response capability.
[0009] According to an embodiment of the present invention, a method for controlling an LED lamp driving circuit for preventing a stroboscopic effect comprises the following steps:
[0010] S1, providing input power and converting the input power into a driving voltage suitable for LED lamps;
[0011] S2, detect the fluctuation of input power supply, obtain real-time voltage and current signals and perform filtering processing;
[0012] S3, process voltage and current signals through multi-band filtering technology, extract fluctuation information of different frequency components, perform empirical mode decomposition and Hilbert transform, generate instantaneous frequency and amplitude, construct a composite feature matrix, use graph convolutional neural network to perform convolution operation and feature fusion, and obtain the final fusion signal;
[0013] S4, transmitting the processed fusion signal to the controller, which calculates the corresponding compensation signal according to the preset stroboscopic effect elimination algorithm and the adaptive graph neural network combined with the variational autoencoder for modeling;
[0014] S5. Through high-frequency pulse width modulation technology, the compensation signal is applied to the driving voltage to adjust the input current of the LED lamp;
[0015] S6. Real-time monitoring of the output light intensity of LED lamps, acquisition of output light signals, wavelet transform denoising, extraction of dynamic characteristics using phase space reconstruction technology, calculation of the stroboscopic index through quantum phase interference, calculation of the stability index of light output, input into the adaptive quantum compensation control module, and dynamic adjustment of driving parameters;
[0016] S7, feeding back the output light signal to the controller and comparing it with a preset light output standard;
[0017] S8, the controller dynamically adjusts the compensation parameters of the compensation circuit according to the comparison result to further optimize the driving voltage;
[0018] S9. Design a low-voltage starting mechanism to enable LED lamps to start smoothly under low voltage conditions and optimize the stroboscopic effect caused by low voltage.
[0019] Optionally, the S2 specifically includes:
[0020] S21, using the voltage sensor to detect the real-time voltage value of the input power supply and obtain the voltage signal V in ;
[0021] S22, using the current sensor to detect the real-time current value of the input power supply and obtain the current signal I in ;
[0022] S23, the detected voltage signal V in and the current signal I in Perform analog-to-digital conversion and convert it into a digital signal V digital and I digital ;
[0023] S24, the digital signal V digital and I digital Input to the signal processing module in the controller;
[0024] S25, the signal processing module pre-processes the input digital signal, filters out high-frequency noise, and obtains a purified voltage signal V filtered and the current signal I filtered ;
[0025] S26, using a joint analysis algorithm based on fractional Fourier transform and Hilbert-Huang transform to analyze the purified voltage signal V filtered and the current signal I filtered The specific method is as follows:
[0026] Perform fractional Fourier transform to obtain a time-frequency domain signal table;
[0027] The Hilbert-Huang transform is used to perform instantaneous frequency analysis on the fractional Fourier transform results and extract the instantaneous frequency components:
[0028]
[0029] Among them, IMF k (t) is the intrinsic mode function, θ k (t) is the instantaneous phase, FrFT is the Fourier transform, Hilbert-Huang transform HHT;
[0030] The fusion signal F is obtained by joint analysis:
[0031]
[0032] Among them, α k and β k is the weight coefficient, which is adjusted according to actual needs. FrFT is Fourier transform, Hilbert-Huang transform HHT;
[0033] S27, transmitting the fused signal F to the computing module of the controller.
[0034] Optionally, the S3 specifically includes:
[0035] S31, the purified voltage signal V filtered and the current signal I filtered Perform empirical mode decomposition to decompose the signal into a series of intrinsic mode function IMF sequences, and obtain the intrinsic mode function IMF sequence of the voltage signal And the intrinsic mode function IMF sequence of the current signal
[0036]
[0037] Where n is the total number of IMFs, r V (t) is the residual signal remaining after decomposition, which represents the part of the signal that has not been decomposed. is the residual signal remaining after decomposition, indicating the part of the signal that has not been decomposed. One of the ingredients;
[0038] S32, for each eigenmode function and Perform Hilbert transform to obtain the instantaneous frequency and amplitude of each modal function, thereby generating the instantaneous frequency of the voltage and current signals and and instantaneous amplitude and
[0039] S33, using the instantaneous frequency and amplitude data obtained in step S32, construct a composite characteristic matrix T, each row of which contains the instantaneous frequency and amplitude of the voltage and current signals, and calculates the instantaneous phase difference Δθ k (t) and the instantaneous amplitude ratio R k (t):
[0040]
[0041] Among them, the instantaneous phase difference Δθ k (t) is calculated as follows:
[0042]
[0043] Among them, Δθ k (t): The instantaneous phase difference between the voltage and current signals of the kth IMF, which represents the phase difference between the two signals at a certain moment. The instantaneous phase of the kth IMF of the voltage signal, The instantaneous phase of the kth IMF of the current signal
[0044] Instantaneous amplitude ratio R k (t) is calculated as follows:
[0045]
[0046] Among them, R k (t): The instantaneous amplitude ratio of the voltage and current signals of the kth IMF, which indicates the relative strength of the two signals at a certain moment, The instantaneous amplitude of the kth IMF of the voltage signal, The instantaneous amplitude of the kth IMF of the current signal;
[0047] S34, performing multi-scale embedded graph neural network analysis on the composite feature matrix T constructed in step S33, and extracting the multi-scale graph structural features of the signal through a multi-scale embedding method:
[0048] Convert the composite feature matrix T into a multi-scale graph structure representation Gs , where graph structures of different scales represent local and global relationships of different features, and the adjacency matrix A of each scale graph structure is i And the feature map matrix H i :
[0049]
[0050] H i =f(A i ,T);
[0051] Among them, T i and T j represents the i-th and j-th rows of the feature vector T, σ is the similarity scale parameter, and f(A i ,T is based on the adjacency matrix A i and the complex relationship function of the feature matrix T, ∥T i -T j ∥ 2 The square of the Euclidean distance between two eigenvectors, A ij : An element in the adjacency matrix of a graph, representing the similarity between the i-th node and the j-th node in the graph;
[0052] S35. Perform convolution operation on multi-scale feature maps through graph convolutional neural network to extract global features H of each scale map s ;
[0053] S36, the global features H of each scale graph structure extracted in step S35 s Fusion, using the attention mechanism to perform weighted combination of features of different scales to obtain the final fusion signal F;
[0054] S37. Transmit the fusion signal F obtained in step S36 to the calculation module of the controller.
[0055] Optionally, the S4 specifically includes:
[0056] S41, inputting the fusion signal F obtained in step S3 into the prediction module of the controller;
[0057] S42. Use adaptive graph neural network combined with variational autoencoder to model historical data and real-time monitoring information, and train the prediction model AGNN-VAE model ,The model is trained with the fusion signal F as input feature to capture the complex spatiotemporal correlations in the data;
[0058] S43. Use the trained prediction model AGNN-VAE model Predict future power supply fluctuations and flicker conditions and generate prediction signals P(t);
[0059] S44, input the predicted signal P(t) into the topological quantum state control algorithm module, and calculate the corresponding compensation signal (C(t). The specific process is to use the high-dimensional topological quantum state mapping function The prediction signal is converted into a signal representation in the topological space and integrated with the adaptive weight function Γ(ω) to generate a compensation signal:
[0060]
[0061] C(t) compensation signal is the signal obtained after topological quantum state control of the predicted signal. -1 The inverse transform operation is used to map the signal in the topological space back to the time or frequency domain, the Ω integral domain represents the range in the frequency or topological space, and the Γ(ω) adaptive weight function is used to weight the signal at different frequencies or in different parts of the topological space. High-dimensional topological quantum state mapping function, which converts the predicted signal P(t) into a signal representation in the topological space, ω frequency or variables in the topological space, is used to represent different components of the signal;
[0062] S45, performing multi-scale random modulation on the compensation signal C(t) through a fractional-order Lévy flight operator. This step uses the compensation signal as input and performs multi-scale random modulation on the compensation signal C(t) through a fractional-order Lévy flight fusion operator. Process the signal to optimize its dynamic characteristics:
[0063]
[0064] Among them, Z k is the random variable of the compensation signal, μ k is the expected value of the signal, Γ is the Gamma function, The fractional-order Lévy flight operator is used to perform multi-scale random modulation on the signal. The α fractional-order parameter controls the scale of the Lévy flight process, and the β stable distribution parameter determines the randomness and distribution characteristics of the signal. FLFO The compensated signal after processing by the fractional-order Lévy flight operator, It is the Gamma function, which is a generalized factorial function in mathematics and appears in many probability density functions of continuous distributions. k The random variable of the compensation signal represents the specific realization of the signal in the random process, μ k The expected value of the signal, used to normalize the random variable Z k ;
[0065] S46, the compensation signal C after multi-scale random modulation FLFO (t) Input to the driver circuit, the driver circuit adjusts the input current I of the LED lamp in and the driving voltage Vin , the complex dynamic characteristics of flicker suppression and compensation signals optimize the stability and response speed of LED lamps;
[0066] S47, real-time monitoring of the adjusted drive current I in and the driving voltage V in The monitoring data is fed back to the controller to dynamically adjust the topological quantum state mapping function and fractional Lévy flight operators parameters for real-time optimization.
[0067] Optionally, the S5 specifically includes:
[0068] S51, inputting the compensation signal C(t) obtained in step S4 into an adaptive fractional-order Fourier transform module, and converting the compensation signal into a high-frequency modulation signal M(t) through fractional-order Fourier transform;
[0069] S52, inputting the high-frequency modulation signal M(t) into the topological quantum optimization module, and further optimizing the modulation signal through the topological quantum optimization algorithm;
[0070] S53, the optimized modulation signal M opt (t) Input to the quantum state modulation module, and use quantum state superposition and interference effects to perform multi-band modulation on the modulation signal:
[0071]
[0072] in, represents the multi-band modulation function, ω n is the center frequency of each frequency band, MQSM(t): the signal after multi-band modulation (t) multi-band modulation function, which controls the modulation effect of each frequency band, M opt The optimized modulation signal is input into the quantum state modulation module for processing. Represents the modulation of the signal in the frequency domain, through different frequencies ω n To achieve multi-band modulation;
[0073] S54, the multi-band modulated signal M QSM (t) Input to the LED drive circuit to adjust the input current I in (t) and input voltage V in (t) duty cycle;
[0074] S55, the modulation signal M is filtered by an adaptive quantum perturbation filter QSM (t) Perform high-frequency filtering to eliminate high-frequency noise in the signal;
[0075] S56, the filtered modulation signal M filtered(t) Feedback to the LED driving circuit to further optimize the modulation process of input current and input voltage;
[0076] S57, dynamically adjust the parameters of the PWM module through the adaptive quantum controller to optimize the effective transmission of the compensation signal and the stability of the light output:
[0077]
[0078] Among them, C AQC represents the adaptive quantum control operator, Θ opt is the optimized PWM parameter, Θ init is the initial PWM parameter, The integral of the filtered modulation signal is used to calculate the PWM adjustment amplitude;
[0079] S58. Use machine learning algorithms to analyze historical data and real-time monitoring information, continuously optimize the modulation strategy of the PWM module, and optimize the light output stability of LED lamps under different working conditions.
[0080] Optionally, the S6 specifically includes:
[0081] S61. Install a high-precision photoelectric sensor at the output end of the LED lamp. The sensor is used to monitor the output light intensity of the LED lamp in real time. The output light intensity signal is L(t).
[0082] S62, input the output light intensity signal L(t) collected by the photoelectric sensor into the light signal processing module, pre-process the light intensity signal through a denoising algorithm based on wavelet transform, and obtain a denoised light intensity signal L filtered (t);
[0083] S63, the denoised light intensity signal L filtered (t) is input into the nonlinear time series analysis module, and the phase space reconstruction technology is used to extract the dynamic characteristics X(t) and reconstruction trajectory R(t) of the light intensity signal. Through phase space reconstruction, the denoised signal L filtered (t) is converted into dynamic features X(t) and reconstructed trajectory R(t), which represent the complex dynamic behavior of the signal;
[0084] S64, inputting the reconstructed dynamic characteristics X(t) into the topological quantum phase detection module, and calculating the stroboscopic index FI(t) of the light intensity signal by quantum phase interference technology;
[0085] S65, inputting the reconstructed trajectory R(t) into the topological structure analysis module, and calculating the stability index SI(t) of the light output by the topological structure analysis algorithm;
[0086] S66, input the stroboscopic index FI(t) and the light output stability index SI(t) into the adaptive quantum compensation control module, and dynamically adjust the working parameters of the LED driving circuit through the quantum fuzzy logic control algorithm:
[0087]
[0088] in, represents the quantum fuzzy logic control operator, Θ opt is the optimized driving parameter;
[0089] The adaptive quantum compensation control module uses the flicker index FI(t) and the stability index SI(t) to dynamically adjust the driving parameter Θ opt , optimize the working status of LED lamps;
[0090] S67, storing the real-time monitoring data L(t), the stroboscopic index FI(t)) and the light output stability index SI(t) into a historical data database;
[0091] S68. Historical data and real-time data are input into the adaptive model optimization module through the quantum machine learning algorithm to continuously optimize the stroboscopic effect detection algorithm and the light output stability analysis algorithm. The quantum machine learning algorithm uses historical data and real-time data to continuously update and optimize the model.
[0092] Optionally, the S7 specifically includes:
[0093] S71, inputting the real-time output light intensity signal monitored in step S6 into a comparison module, the comparison module compares the real-time output light intensity signal with a preset light output standard, and generates an error signal, the error signal reflects the difference between the current light output and the target standard;
[0094] S72, input the error signal to the adaptive fuzzy control module, process the error signal through the fuzzy logic algorithm, generate a preliminary control signal, the fuzzy control module combines the machine learning algorithm, updates the fuzzy rules and membership functions to optimize the control accuracy, and the preliminary control signal is used to guide the subsequent optimization adjustment;
[0095] S73, inputting the preliminary control signal into the quantum optimization module, optimizing the control signal using an optimization algorithm based on quantum tunneling effect and quantum superposition state, and generating a final control signal, wherein the quantum optimization module optimizes efficiency and accuracy by searching for an optimal path in a high-dimensional quantum state space;
[0096] S74, feeding back the final control signal to the LED driving circuit, adjusting the driving current and driving voltage through the adaptive current and voltage regulation technology, optimizing the output light intensity of the LED lamp to stabilize it within the preset light output standard, and dynamically adjusting the driving parameters using the adaptive genetic algorithm, so that the adjusted current and voltage optimize the stability of the light output;
[0097] S75, real-time monitoring of the adjusted output light intensity, and feeding back the data to the adaptive fuzzy control module and the quantum optimization module for further parameter adjustment and optimization. The feedback mechanism is based on the deep reinforcement learning algorithm, autonomous learning and optimization adjustment strategy;
[0098] S76, storing the real-time monitoring data, error signal, preliminary control signal and final control signal in a quantum distributed database;
[0099] S77. Use historical data and real-time monitoring data to periodically train and update the adaptive fuzzy control module and quantum optimization module, continuously optimize the control algorithm, and record each training and update process by integrating blockchain technology.
[0100] Optionally, the S8 specifically includes:
[0101] S81, inputting the compensation signal optimized in step S7 into a dynamic compensation module, which dynamically adjusts the compensation parameters based on the real-time monitored output light intensity and current signal to adapt to the current working state;
[0102] S82. In the dynamic compensation module, the fuzzy control algorithm is used to analyze and adjust the compensation signal in real time to generate a dynamic compensation signal C. dynamic (t), the fuzzy control algorithm continuously optimizes the parameters of the compensation signal by comparing the actual monitoring data with the preset standards;
[0103] S83, the dynamic compensation signal C dynamic (t) is input to the adaptive neural network module. The adaptive neural network further optimizes the compensation signal by training historical data and real-time data to generate the final compensation signal C final (t), the neural network model uses historical data as a training set and real-time data as a correction basis, and continuously adjusts the network weights to optimize the accuracy of the signal;
[0104] S84, in the adaptive neural network module, the dynamic compensation signal is trained and optimized using the back propagation algorithm and the recursive neural network, and the final compensation signal C final (t) It can more accurately match the actual needs and suppress the stroboscopic effect. The back propagation algorithm adjusts the network parameters according to the error gradient, and the recursive neural network processes the time series data to make the compensation signal have time correlation;
[0105] S85, the final compensation signal C final (t) is fed back to the LED driving circuit, and the driving current I in (t and driving voltage V in (t), optimize the output light intensity of the LED lamp to stabilize it within the preset light output standard, and compensate the signal C final (t) Directly act on the driving circuit to adjust the current and voltage to optimize the light intensity stability;
[0106] S86, real-time monitoring of the adjusted output light intensity, and feeding back the data to the dynamic compensation module and the adaptive neural network module, through further parameter adjustment and optimization, dynamically adapting to power supply fluctuations and load changes, the monitoring data provides a real-time reference for subsequent adjustments, so that the system can respond to changes quickly;
[0107] S87, storing the real-time monitoring data, dynamic compensation signal and final compensation signal in a historical data database;
[0108] S88. Utilize historical data and real-time monitoring data to periodically train and update the adaptive neural network module and dynamic compensation module, and continuously optimize the control algorithm. By integrating reinforcement learning and evolutionary algorithms, the system achieves self-optimization and evolution to improve overall performance. The reinforcement learning algorithm optimizes strategies through trial and error and reward mechanisms, and the evolutionary algorithm simulates the natural selection process to find the optimal solution.
[0109] Beneficial Effects
[0110] High light output stability: The present invention uses multi-band filtering technology, empirical mode decomposition and Hilbert transform method to accurately extract information on different frequency components in voltage and current signals, and combines graph neural network and quantum optimization technology to adjust the driving voltage and current in real time, ensuring the stability of the light output of LED lamps under various working conditions and significantly reducing the occurrence of stroboscopic effect.
[0111] Strong system reliability: The present invention integrates a low-voltage starting mechanism and overvoltage and overcurrent protection functions. By real-time monitoring of input voltage and driving current, it dynamically adjusts driving parameters to effectively protect the safe operation of LED lamps under conditions of power fluctuations and load changes, thereby improving the overall reliability of the system.
[0112] Strong real-time response capability: Adaptive fuzzy control and machine learning algorithms are used to dynamically predict power fluctuations and flicker based on real-time monitoring data and historical data, and adjust drive parameters in advance to prevent the occurrence of flicker effects. Quantum optimization technology further improves the accuracy and response speed of control signals, enabling the system to quickly respond to power fluctuations and load changes.
[0113] Intelligent regulation: This invention uses machine learning algorithms to analyze and model historical data and real-time monitoring information. Through the fusion model of adaptive topological convolutional network and quantum generative adversarial network, it predicts power supply fluctuations and flicker, optimizes the drive parameter settings, and improves the intelligence level of the system.
[0114] Efficient startup and protection: Design a low-voltage startup mechanism to ensure that LED lamps start smoothly under low voltage conditions and prevent the stroboscopic effect caused by low voltage. At the same time, the integrated overvoltage and overcurrent protection functions adjust the driving current and voltage in real time through the topological quantum state protection algorithm and quantum state adaptive adjustment algorithm to protect LED lamps from damage.
[0115] Energy saving and environmental protection: By optimizing the control strategy of the drive circuit, the energy waste caused by the stroboscopic effect is reduced, the overall energy efficiency of the system is improved, and the requirements of energy saving and environmental protection are met. BRIEF DESCRIPTION OF THE DRAWINGS
[0116] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:
[0117] Figure 1 A summary drawing of the LED lamp driving circuit control method for preventing stroboscopic effect proposed by the present invention;
[0118] Figure 2 The flowchart of claim 3 of the LED lamp driving circuit control method for preventing the stroboscopic effect proposed by the present invention;
[0119] Figure 3 This is the flow chart of claim 4 of the LED lamp driving circuit control method for preventing the stroboscopic effect proposed by the present invention. DETAILED DESCRIPTION
[0120] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.
[0121] refer to Figure 1-3 , a method for controlling an LED lamp driving circuit for preventing a stroboscopic effect, comprising the following steps:
[0122] S1, providing input power and converting the input power into a driving voltage suitable for LED lamps;
[0123] S2, detect the fluctuation of input power supply, obtain real-time voltage and current signals and perform filtering processing;
[0124] S3, process voltage and current signals through multi-band filtering technology, extract fluctuation information of different frequency components, perform empirical mode decomposition and Hilbert transform, generate instantaneous frequency and amplitude, construct a composite feature matrix, use graph convolutional neural network to perform convolution operation and feature fusion, and obtain the final fusion signal;
[0125] S4, transmitting the processed fusion signal to the controller, which calculates the corresponding compensation signal according to the preset stroboscopic effect elimination algorithm and the adaptive graph neural network combined with the variational autoencoder for modeling;
[0126] S5. Through high-frequency pulse width modulation technology, the compensation signal is applied to the driving voltage to adjust the input current of the LED lamp;
[0127] S6. Real-time monitoring of the output light intensity of LED lamps, acquisition of output light signals, wavelet transform denoising, extraction of dynamic characteristics using phase space reconstruction technology, calculation of the stroboscopic index through quantum phase interference, calculation of the stability index of light output, input into the adaptive quantum compensation control module, and dynamic adjustment of driving parameters;
[0128] S7, feeding back the output light signal to the controller and comparing it with a preset light output standard;
[0129] S8, the controller dynamically adjusts the compensation parameters of the compensation circuit according to the comparison result to further optimize the driving voltage;
[0130] S9. Design a low-voltage starting mechanism to enable LED lamps to start smoothly under low voltage conditions and optimize the stroboscopic effect caused by low voltage.
[0131] In this implementation, S2 specifically includes:
[0132] S21, using the voltage sensor to detect the real-time voltage value of the input power supply and obtain the voltage signal V in ;
[0133] S22, using the current sensor to detect the real-time current value of the input power supply and obtain the current signal I in ;
[0134] S23, the detected voltage signal V in and the current signal I in Perform analog-to-digital conversion and convert it into a digital signal V digital and I digital ;
[0135] S24, the digital signal V digital and I digital input to the signal processing module in the controller;
[0136] S25, the signal processing module pre-processes the input digital signal, filters out high-frequency noise, and obtains a purified voltage signal V filtered and the current signal I filtered ;
[0137] S26, using a joint analysis algorithm based on fractional Fourier transform and Hilbert-Huang transform to analyze the purified voltage signal V filtered and the current signal I filtered The specific method is as follows:
[0138] Perform fractional Fourier transform to obtain a time-frequency domain signal table;
[0139] The Hilbert-Huang transform is used to perform instantaneous frequency analysis on the fractional Fourier transform results and extract the instantaneous frequency components:
[0140]
[0141]
[0142] Among them, IMF k (t) is the intrinsic mode function, θ k (t) is the instantaneous phase, FrFT is the Fourier transform, Hilbert-Huang transform HHT;
[0143] The fusion signal F is obtained by joint analysis:
[0144]
[0145] Among them, α k and β k is the weight coefficient, which is adjusted according to actual needs. FrFT is Fourier transform, Hilbert-Huang transform HHT;
[0146] S27, transmitting the fused signal F to the computing module of the controller.
[0147] In this implementation, S3 specifically includes:
[0148] S31, the purified voltage signal V filtered and the current signal I filtered Perform empirical mode decomposition to decompose the signal into a series of intrinsic mode function IMF sequences, and obtain the intrinsic mode function IMF sequence of the voltage signal And the intrinsic mode function IMF sequence of the current signal
[0149]
[0150] Where n is the total number of IMFs, r V(t) is the residual signal remaining after decomposition, which represents the part of the signal that has not been decomposed. is the residual signal remaining after decomposition, indicating the part of the signal that has not been decomposed. One of the ingredients;
[0151] S32, for each eigenmode function and Perform Hilbert transform to obtain the instantaneous frequency and amplitude of each modal function, thereby generating the instantaneous frequency of the voltage and current signals and and instantaneous amplitude and
[0152] S33, using the instantaneous frequency and amplitude data obtained in step S32, construct a composite characteristic matrix T, each row of which contains the instantaneous frequency and amplitude of the voltage and current signals, and calculates the instantaneous phase difference Δθ k (t) and the instantaneous amplitude ratio R k (t):
[0153]
[0154] Among them, the instantaneous phase difference Δθ k (t) is calculated as follows:
[0155]
[0156] Among them, Δθ k (t): The instantaneous phase difference between the voltage and current signals of the kth IMF, which represents the phase difference between the two signals at a certain moment. The instantaneous phase of the kth IMF of the voltage signal, The instantaneous phase of the kth IMF of the current signal
[0157] Instantaneous amplitude ratio R k (t) is calculated as follows:
[0158]
[0159] Among them, R k (t): The instantaneous amplitude ratio of the voltage and current signals of the kth IMF, which indicates the relative strength of the two signals at a certain moment, The instantaneous amplitude of the kth IMF of the voltage signal, The instantaneous amplitude of the kth IMF of the current signal;
[0160] S34, performing multi-scale embedded graph neural network analysis on the composite feature matrix T constructed in step S33, and extracting the multi-scale graph structural features of the signal through a multi-scale embedding method:
[0161] Convert the composite feature matrix T into a multi-scale graph structure representation G s , where graph structures of different scales represent local and global relationships of different features, and the adjacency matrix A of each scale graph structure is i And the feature map matrix H i :
[0162]
[0163] H i =f(A i ,T);
[0164] Among them, T i and T j represents the i-th and j-th rows of the feature vector T, σ is the similarity scale parameter, and f(A i ,T is based on the adjacency matrix A i and the complex relationship function of the feature matrix T, ∥T i -T j ∥ 2 The square of the Euclidean distance between two eigenvectors, A ij : An element in the adjacency matrix of a graph, representing the similarity between the i-th node and the j-th node in the graph;
[0165] S35. Perform convolution operation on multi-scale feature maps through graph convolutional neural network to extract global features H of each scale map s ;
[0166] S36, the global features H of each scale graph structure extracted in step S35 s Fusion, using the attention mechanism to perform weighted combination of features of different scales to obtain the final fusion signal F;
[0167] S37. Transmit the fusion signal F obtained in step S36 to the calculation module of the controller.
[0168] In this implementation manner, the S4 specifically includes:
[0169] S41, inputting the fusion signal F obtained in step S3 into the prediction module of the controller;
[0170] S42. Use adaptive graph neural network combined with variational autoencoder to model historical data and real-time monitoring information, and train the prediction model AGNN-VAE model ,The model is trained with the fusion signal F as input feature to capture the complex spatiotemporal correlations in the data;
[0171] S43. Use the trained prediction model AGNN-VAE modelPredict future power supply fluctuations and flicker conditions and generate prediction signals P(t);
[0172] S44, input the predicted signal P(t) into the topological quantum state control algorithm module, and calculate the corresponding compensation signal (C(t). The specific process is to use the high-dimensional topological quantum state mapping function The prediction signal is converted into a signal representation in the topological space and integrated with the adaptive weight function Γ(ω) to generate a compensation signal:
[0173]
[0174] C(t) compensation signal is the signal obtained after topological quantum state control of the predicted signal. -1 The inverse transform operation is used to map the signal in the topological space back to the time or frequency domain, the Ω integral domain represents the range in the frequency or topological space, and the Γ(ω) adaptive weight function is used to weight the signal at different frequencies or in different parts of the topological space. High-dimensional topological quantum state mapping function, which converts the predicted signal P(t) into a signal representation in the topological space, ω frequency or variables in the topological space, is used to represent different components of the signal;
[0175] S45, performing multi-scale random modulation on the compensation signal C(t) through a fractional-order Lévy flight operator. This step uses the compensation signal as input and performs multi-scale random modulation on the compensation signal C(t) through a fractional-order Lévy flight fusion operator. Process the signal to optimize its dynamic characteristics:
[0176]
[0177] Among them, Z k is the random variable of the compensation signal, μ k is the expected value of the signal, Γ is the Gamma function, The fractional-order Lévy flight operator is used to perform multi-scale random modulation on the signal. The α fractional-order parameter controls the scale of the Lévy flight process, and the β stable distribution parameter determines the randomness and distribution characteristics of the signal. FLFO The compensated signal after processing by the fractional-order Lévy flight operator, It is the Gamma function, which is a generalized factorial function in mathematics and appears in many probability density functions of continuous distributions. k The random variable of the compensation signal represents the specific realization of the signal in the random process, μ k The expected value of the signal, used to normalize the random variable Z k ;
[0178] S46, the compensation signal C after multi-scale random modulation FLFO(t) Input to the driver circuit, the driver circuit adjusts the input current I of the LED lamp in and the driving voltage V in , the complex dynamic characteristics of flicker suppression and compensation signals optimize the stability and response speed of LED lamps;
[0179] S47, real-time monitoring of the adjusted drive current I in and the driving voltage V in The monitoring data is fed back to the controller to dynamically adjust the topological quantum state mapping function and fractional Lévy flight operators parameters for real-time optimization.
[0180] In this implementation manner, S5 specifically includes:
[0181] S51, inputting the compensation signal C(t) obtained in step S4 into an adaptive fractional-order Fourier transform module, and converting the compensation signal into a high-frequency modulation signal M(t) through fractional-order Fourier transform;
[0182] S52, inputting the high-frequency modulation signal M(t) into the topological quantum optimization module, and further optimizing the modulation signal through the topological quantum optimization algorithm;
[0183] S53, the optimized modulation signal M opt (t) Input to the quantum state modulation module, and use quantum state superposition and interference effects to perform multi-band modulation on the modulation signal:
[0184]
[0185] in, represents the multi-band modulation function, ω n is the center frequency of each frequency band, MQSM(t): the signal after multi-band modulation (t) multi-band modulation function, which controls the modulation effect of each frequency band, M opt The optimized modulation signal is input into the quantum state modulation module for processing. Represents the modulation of the signal in the frequency domain, through different frequencies ω n To achieve multi-band modulation;
[0186] S54, the multi-band modulated signal M QSM (t) Input to the LED drive circuit to adjust the input current I in (t) and input voltage V in (t) duty cycle;
[0187] S55, the modulation signal M is filtered by an adaptive quantum perturbation filter QSM (t) Perform high-frequency filtering to eliminate high-frequency noise in the signal;
[0188] S56, the filtered modulation signal M filtered (t) Feedback to the LED driving circuit to further optimize the modulation process of input current and input voltage;
[0189] S57, dynamically adjust the parameters of the PWM module through the adaptive quantum controller to optimize the effective transmission of the compensation signal and the stability of the light output:
[0190]
[0191] Among them, C AQC represents the adaptive quantum control operator, Θ opt is the optimized PWM parameter, Θ init is the initial PWM parameter, The integral of the filtered modulation signal is used to calculate the PWM adjustment amplitude;
[0192] S58. Use machine learning algorithms to analyze historical data and real-time monitoring information, continuously optimize the modulation strategy of the PWM module, and optimize the light output stability of LED lamps under different working conditions.
[0193] In this implementation manner, S6 specifically includes:
[0194] S61. Install a high-precision photoelectric sensor at the output end of the LED lamp. The sensor is used to monitor the output light intensity of the LED lamp in real time. The output light intensity signal is L(t).
[0195] S62, input the output light intensity signal L(t) collected by the photoelectric sensor into the light signal processing module, pre-process the light intensity signal through a denoising algorithm based on wavelet transform, and obtain a denoised light intensity signal L filtered (t);
[0196] S63, the denoised light intensity signal L filtered (t) is input into the nonlinear time series analysis module, and the phase space reconstruction technology is used to extract the dynamic characteristics X(t) and reconstruction trajectory R(t) of the light intensity signal. Through phase space reconstruction, the denoised signal L filtered (t) is converted into dynamic features X(t) and reconstructed trajectory R(t), which represent the complex dynamic behavior of the signal;
[0197] S64, inputting the reconstructed dynamic characteristics X(t) into the topological quantum phase detection module, and calculating the stroboscopic index FI(t) of the light intensity signal by quantum phase interference technology;
[0198] S65, inputting the reconstructed trajectory R(t) into the topological structure analysis module, and calculating the stability index SI(t) of the light output by the topological structure analysis algorithm;
[0199] S66, input the stroboscopic index FI(t) and the light output stability index SI(t) into the adaptive quantum compensation control module, and dynamically adjust the working parameters of the LED driving circuit through the quantum fuzzy logic control algorithm:
[0200]
[0201] in, represents the quantum fuzzy logic control operator, Θ opt is the optimized driving parameter;
[0202] The adaptive quantum compensation control module uses the flicker index FI(t) and the stability index SI(t) to dynamically adjust the driving parameter Θ opt , optimize the working status of LED lamps;
[0203] S67, storing the real-time monitoring data L(t), the stroboscopic index FI(t)) and the light output stability index SI(t) into a historical data database;
[0204] S68. Historical data and real-time data are input into the adaptive model optimization module through the quantum machine learning algorithm to continuously optimize the stroboscopic effect detection algorithm and the light output stability analysis algorithm. The quantum machine learning algorithm uses historical data and real-time data to continuously update and optimize the model.
[0205] In this implementation manner, the S7 specifically includes:
[0206] S71, inputting the real-time output light intensity signal monitored in step S6 into a comparison module, the comparison module compares the real-time output light intensity signal with a preset light output standard, and generates an error signal, the error signal reflects the difference between the current light output and the target standard;
[0207] S72, input the error signal to the adaptive fuzzy control module, process the error signal through the fuzzy logic algorithm, generate a preliminary control signal, the fuzzy control module combines the machine learning algorithm, updates the fuzzy rules and membership functions to optimize the control accuracy, and the preliminary control signal is used to guide the subsequent optimization adjustment;
[0208] S73, inputting the preliminary control signal into the quantum optimization module, optimizing the control signal using an optimization algorithm based on quantum tunneling effect and quantum superposition state, and generating a final control signal, wherein the quantum optimization module optimizes efficiency and accuracy by searching for an optimal path in a high-dimensional quantum state space;
[0209] S74, feeding back the final control signal to the LED driving circuit, adjusting the driving current and driving voltage through the adaptive current and voltage regulation technology, optimizing the output light intensity of the LED lamp to stabilize it within the preset light output standard, and dynamically adjusting the driving parameters using the adaptive genetic algorithm, so that the adjusted current and voltage optimize the stability of the light output;
[0210] S75, real-time monitoring of the adjusted output light intensity, and feeding back the data to the adaptive fuzzy control module and the quantum optimization module for further parameter adjustment and optimization. The feedback mechanism is based on the deep reinforcement learning algorithm, autonomous learning and optimization adjustment strategy;
[0211] S76, storing the real-time monitoring data, error signal, preliminary control signal and final control signal in a quantum distributed database;
[0212] S77. Use historical data and real-time monitoring data to periodically train and update the adaptive fuzzy control module and quantum optimization module, continuously optimize the control algorithm, and record each training and update process by integrating blockchain technology.
[0213] In this implementation manner, S8 specifically includes:
[0214] S81, inputting the compensation signal optimized in step S7 into a dynamic compensation module, which dynamically adjusts the compensation parameters based on the real-time monitored output light intensity and current signal to adapt to the current working state;
[0215] S82. In the dynamic compensation module, the fuzzy control algorithm is used to analyze and adjust the compensation signal in real time to generate a dynamic compensation signal C. dynamic (t), the fuzzy control algorithm continuously optimizes the parameters of the compensation signal by comparing the actual monitoring data with the preset standards;
[0216] S83, the dynamic compensation signal C dynamic (t) is input to the adaptive neural network module. The adaptive neural network further optimizes the compensation signal by training historical data and real-time data to generate the final compensation signal C final (t), the neural network model uses historical data as a training set and real-time data as a correction basis, and continuously adjusts the network weights to optimize the accuracy of the signal;
[0217] S84, in the adaptive neural network module, the dynamic compensation signal is trained and optimized using the back propagation algorithm and the recursive neural network, and the final compensation signal C final (t) It can more accurately match the actual needs and suppress the stroboscopic effect. The back propagation algorithm adjusts the network parameters according to the error gradient, and the recursive neural network processes the time series data to make the compensation signal have time correlation;
[0218] S85, the final compensation signal C final (t) is fed back to the LED driving circuit, and the driving current I in (t and driving voltage V in (t), optimize the output light intensity of the LED lamp to stabilize it within the preset light output standard, and compensate the signal C final (t) Directly act on the driving circuit to adjust the current and voltage to optimize the light intensity stability;
[0219] S86, real-time monitoring of the adjusted output light intensity, and feeding back the data to the dynamic compensation module and the adaptive neural network module, through further parameter adjustment and optimization, dynamically adapting to power supply fluctuations and load changes, the monitoring data provides a real-time reference for subsequent adjustments, so that the system can respond to changes quickly;
[0220] S87, storing the real-time monitoring data, dynamic compensation signal and final compensation signal in a historical data database;
[0221] S88. Utilize historical data and real-time monitoring data to periodically train and update the adaptive neural network module and dynamic compensation module, and continuously optimize the control algorithm. By integrating reinforcement learning and evolutionary algorithms, the system achieves self-optimization and evolution to improve overall performance. The reinforcement learning algorithm optimizes strategies through trial and error and reward mechanisms, and the evolutionary algorithm simulates the natural selection process to find the optimal solution.
[0222] Embodiment 1:
[0223] In order to verify the feasibility of the present invention in implementation, the present invention is applied to a large shopping mall in a certain city, where a large number of LED lamps are installed to meet the lighting needs of different areas. However, due to the fluctuation of the shopping mall's power supply system and the change of load, the existing LED lamps often have a stroboscopic effect, which causes customers to feel uncomfortable during the shopping process and affects the shopping experience. Especially during holidays, the shopping mall has a large passenger flow, the power supply load fluctuation is more obvious, and the stroboscopic problem is more serious. In addition, the stroboscopic effect also has a negative impact on the image quality of the shopping mall's monitoring system, increasing the difficulty of management.
[0224] In order to eliminate the stroboscopic effect and improve the light output stability of the LED lamp and the overall reliability of the system, the shopping mall decided to adopt the LED lamp driving circuit control method for preventing the stroboscopic effect proposed in the present invention.
[0225] The LED lamp drive circuit control system of the present invention is installed in the main lighting areas of the shopping mall, including supermarkets, clothing stores, dining areas and leisure areas. The LED lamps in each area are connected to the controller and centrally managed and controlled by the controller. After the system is installed, preliminary debugging is carried out, including detecting the voltage and current fluctuations of the input power supply, setting the low voltage startup mechanism and overvoltage and overcurrent protection parameters.
[0226] The controller starts to monitor the power fluctuation of the shopping mall in real time, collects voltage and current signals in various areas, and processes these signals through multi-band filtering technology to extract fluctuation information of different frequency components. Through empirical mode decomposition and Hilbert transform, the instantaneous frequency and amplitude of voltage and current signals are generated to construct a composite feature matrix.
[0227] Based on the extracted feature matrix, the controller calculates the corresponding compensation signal in real time through adaptive fuzzy control and quantum optimization algorithm, and applies the compensation signal to the driving voltage through high-frequency pulse width modulation technology to adjust the input current of the LED lamp. The output light intensity of the LED lamp is monitored in real time, and the output light signal is fed back to the controller and compared with the preset light output standard, and the parameters of the compensation circuit are dynamically adjusted to further optimize the driving voltage.
[0228] The controller uses machine learning algorithms to predict possible power fluctuations and flicker based on historical data and real-time monitoring information, and adjusts the drive parameters in advance to prevent the occurrence of flicker effects. The system regularly trains and updates the adaptive topology convolutional network and quantum generative adversarial network models to ensure that the system can respond to new power fluctuation patterns in a timely manner.
[0229] In order to verify the beneficial effects of the present invention, the shopping mall carried out detailed data recording and analysis within one month. The following are the specific data:
[0230] Table 1. Data power supply fluctuation of LED lamp drive circuit control method embodiment to prevent stroboscopic effect
[0231]
[0232]
[0233] Table 2. Stroboscopic effect records
[0234] Time period Average number of flashes (times / hour) Maximum strobe intensity (%) Before installing the system 30 15 After installing the system 5 3
[0235] Table 3. Customer satisfaction survey
[0236] Time period Number of respondents Discomfort complaint rate (%) Before installing the system 1000 25 After installing the system 1000 5
[0237] Table 4. Surveillance system image quality
[0238] Time period Number of image quality issues per week Before installing the system 5 After installing the system 1
[0239] The power fluctuation of the shopping mall has been significantly improved, with the peak voltage reduced from 245V to 240V, the minimum voltage increased from 210V to 215V, and the average voltage fluctuation frequency reduced from 50 times / hour to 20 times / hour. At the same time, the stroboscopic effect of LED lamps has been significantly reduced, with the average stroboscopic frequency reduced from 30 times / hour to 5 times / hour, and the maximum stroboscopic intensity reduced from 15% to 3%. Customer satisfaction has also been greatly improved, with the discomfort complaint rate reduced from 25% to 5%. In addition, the number of image quality problems in the monitoring system has been reduced from 5 times per week to 1 time. These data demonstrate the significant effect of the present invention in improving light output stability, system reliability and user experience.
[0240] It can be seen from the above data that the LED lamp driving circuit control method for preventing the stroboscopic effect of the present invention has shown significant beneficial effects in practical applications. The power supply fluctuation of the shopping mall has been significantly improved, the stroboscopic effect has been significantly reduced, customer satisfaction has been greatly improved, and the image quality of the monitoring system has been effectively guaranteed. The present invention realizes efficient, stable and intelligent control of LED lamps by comprehensively applying low-voltage starting mechanism, overvoltage and overcurrent protection functions, adaptive fuzzy control, quantum optimization and machine learning algorithms, and significantly improves the overall performance and reliability of the system.
[0241] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.
Claims
1. A method for controlling an LED lamp driving circuit for preventing a stroboscopic effect, characterized in that: The steps include: S1, providing input power and converting the input power into a driving voltage suitable for LED lamps; S2, detect the fluctuation of input power supply, obtain real-time voltage and current signals and perform filtering processing; S3, process voltage and current signals through multi-band filtering technology, extract fluctuation information of different frequency components, perform empirical mode decomposition and Hilbert transform, generate instantaneous frequency and amplitude, construct a composite feature matrix, use graph convolutional neural network to perform convolution operation and feature fusion, and obtain the final fusion signal; S4, transmitting the processed fusion signal to the controller, which calculates the corresponding compensation signal according to the preset stroboscopic effect elimination algorithm and the adaptive graph neural network combined with the variational autoencoder for modeling; S5. Through high-frequency pulse width modulation technology, the compensation signal is applied to the driving voltage to adjust the input current of the LED lamp; S6. Real-time monitoring of the output light intensity of LED lamps, acquisition of output light signals, wavelet transform denoising, extraction of dynamic characteristics using phase space reconstruction technology, calculation of the stroboscopic index through quantum phase interference, calculation of the stability index of light output, input into the adaptive quantum compensation control module, and dynamic adjustment of driving parameters; S7, feeding back the output light signal to the controller and comparing it with a preset light output standard; S8, the controller dynamically adjusts the compensation parameters of the compensation circuit according to the comparison result to further optimize the driving voltage; S9. Design a low-voltage starting mechanism to enable LED lamps to start smoothly under low voltage conditions and optimize the stroboscopic effect caused by low voltage.
2. The LED lamp driving circuit control method for preventing stroboscopic effect according to claim 1, characterized in that: The S2 specifically includes: S21, using the voltage sensor to detect the real-time voltage value of the input power supply and obtain the voltage signal V in ; S22, using the current sensor to detect the real-time current value of the input power supply and obtain the current signal I in ; S23, the detected voltage signal V in and the current signal I in Perform analog-to-digital conversion and convert it into a digital signal V digital and I digital ; S24, the digital signal V digital and I digital Input to the signal processing module in the controller; S25, the signal processing module pre-processes the input digital signal, filters out high-frequency noise, and obtains a purified voltage signal V filtered and the current signal I filtered ; S26, using a joint analysis algorithm based on fractional Fourier transform and Hilbert-Huang transform to analyze the purified voltage signal V filtered and the current signal I filtered The specific method is as follows: Perform fractional Fourier transform to obtain a time-frequency domain signal table; The Hilbert-Huang transform is used to perform instantaneous frequency analysis on the fractional Fourier transform results and extract the instantaneous frequency components: Among them, IMF k (t) is the intrinsic mode function, θ k (t) is the instantaneous phase, FrFT is the Fourier transform, Hilbert-Huang transform HHT; The fusion signal F is obtained by joint analysis: Among them, α k and β k is the weight coefficient, which is adjusted according to actual needs. FrFT is Fourier transform, Hilbert-Huang transform HHT; S27, transmitting the fused signal F to the computing module of the controller.
3. The LED lamp driving circuit control method for preventing stroboscopic effect according to claim 1, characterized in that: The S3 specifically includes: S31, the purified voltage signal V filtered and the current signal I filtered Perform empirical mode decomposition to decompose the signal into a series of intrinsic mode function IMF sequences, and obtain the intrinsic mode function IMF sequence of the voltage signal And the intrinsic mode function IMF sequence of the current signal Where n is the total number of IMFs, r V (t) is the residual signal remaining after decomposition, which represents the part of the signal that has not been decomposed. is the residual signal remaining after decomposition, indicating the part of the signal that has not been decomposed. One of the ingredients; S32, for each eigenmode function and Perform Hilbert transform to obtain the instantaneous frequency and amplitude of each modal function, thereby generating the instantaneous frequency of the voltage and current signals and and instantaneous amplitude and S33, using the instantaneous frequency and amplitude data obtained in step S32, construct a composite characteristic matrix T, each row of which contains the instantaneous frequency and amplitude of the voltage and current signals, and calculates the instantaneous phase difference Δθ k (t) and the instantaneous amplitude ratio R k (t): Among them, the instantaneous phase difference Δθ k (t) is calculated as follows: Among them, Δθ k (t): The instantaneous phase difference between the voltage and current signals of the kth IMF, which represents the phase difference between the two signals at a certain moment. The instantaneous phase of the kth IMF of the voltage signal, The instantaneous phase of the kth IMF of the current signal Instantaneous amplitude ratio R k (t) is calculated as follows: Among them, R k (t): The instantaneous amplitude ratio of the voltage and current signals of the kth IMF, which indicates the relative strength of the two signals at a certain moment, The instantaneous amplitude of the kth IMF of the voltage signal, A Ik (t): instantaneous amplitude of the kth IMF of the current signal; S34, performing multi-scale embedded graph neural network analysis on the composite feature matrix T constructed in step S33, and extracting the multi-scale graph structural features of the signal through a multi-scale embedding method: Convert the composite feature matrix T into a multi-scale graph structure representation G s , where graph structures of different scales represent local and global relationships of different features, and the adjacency matrix A of each scale graph structure is i And the feature map matrix H i : H i =f(A i ,T); Among them, T i and T j represents the i-th and j-th rows of the feature vector T, σ is the similarity scale parameter, and f(A i ,T is based on the adjacency matrix A i and the complex relationship function of the feature matrix T, ∥T i -T j ∥ 2 The square of the Euclidean distance between two eigenvectors, A ij : An element in the adjacency matrix of a graph, representing the similarity between the i-th node and the j-th node in the graph; S35. Perform convolution operation on multi-scale feature maps through graph convolutional neural network to extract global features H of each scale map s ; S36, the global features H of each scale graph structure extracted in step S35 s Fusion, using the attention mechanism to perform weighted combination of features of different scales to obtain the final fusion signal F; S37. Transmit the fusion signal F obtained in step S36 to the calculation module of the controller.
4. The LED lamp driving circuit control method for preventing stroboscopic effect according to claim 1, characterized in that: The S4 specifically includes: S41, inputting the fusion signal F obtained in step S3 into the prediction module of the controller; S42. Use adaptive graph neural network combined with variational autoencoder to model historical data and real-time monitoring information, and train the prediction model AGNN-VAE model ,The model is trained with the fusion signal F as input feature to capture the complex spatiotemporal correlations in the data; S43. Use the trained prediction model AGNN-VAE model Predict future power supply fluctuations and flicker conditions and generate prediction signals P(t); S44, input the predicted signal P(t) into the topological quantum state control algorithm module, and calculate the corresponding compensation signal (C(t). The specific process is to use the high-dimensional topological quantum state mapping function The prediction signal is converted into a signal representation in the topological space and integrated with the adaptive weight function Γ(ω) to generate a compensation signal: C(t) compensation signal is the signal obtained after topological quantum state control of the predicted signal. -1 The inverse transform operation is used to map the signal in the topological space back to the time or frequency domain, the Ω integral domain represents the range in the frequency or topological space, and the Γ(ω) adaptive weight function is used to weight the signal at different frequencies or in different parts of the topological space. High-dimensional topological quantum state mapping function, which converts the predicted signal P(t) into a signal representation in the topological space, ω frequency or variables in the topological space, is used to represent different components of the signal; S45, performing multi-scale random modulation on the compensation signal C(t) through a fractional-order Lévy flight operator. This step uses the compensation signal as input and performs multi-scale random modulation on the compensation signal C(t) through a fractional-order Lévy flight fusion operator. Process the signal to optimize its dynamic characteristics: Among them, Z k is the random variable of the compensation signal, μ k is the expected value of the signal, Γ is the Gamma function, The fractional-order Lévy flight operator is used to perform multi-scale random modulation on the signal. The α fractional-order parameter controls the scale of the Lévy flight process, and the β stable distribution parameter determines the randomness and distribution characteristics of the signal. FLFO The compensated signal after processing by the fractional-order Lévy flight operator, It is the Gamma function, which is a generalized factorial function in mathematics and appears in many probability density functions of continuous distributions. k The random variable of the compensation signal represents the specific realization of the signal in the random process, μ k The expected value of the signal, used to normalize the random variable Z k ; S46, the compensation signal C after multi-scale random modulation FLFO (t) Input to the driver circuit, the driver circuit adjusts the input current I of the LED lamp in and the driving voltage V in , the complex dynamic characteristics of flicker suppression and compensation signals optimize the stability and response speed of LED lamps; S47, real-time monitoring of the adjusted drive current I in and the driving voltage V in The monitoring data is fed back to the controller to dynamically adjust the topological quantum state mapping function and fractional Lévy flight operators parameters for real-time optimization.
5. The LED lamp driving circuit control method for preventing stroboscopic effect according to claim 1, characterized in that: The S5 specifically includes: S51, inputting the compensation signal C(t) obtained in step S4 into an adaptive fractional-order Fourier transform module, and converting the compensation signal into a high-frequency modulation signal M(t) through fractional-order Fourier transform; S52, inputting the high-frequency modulation signal M(t) into the topological quantum optimization module, and further optimizing the modulation signal through the topological quantum optimization algorithm; S53, the optimized modulation signal M opt (t) Input to the quantum state modulation module, and use quantum state superposition and interference effects to perform multi-band modulation on the modulation signal: in, represents the multi-band modulation function, ω n is the center frequency of each frequency band, MQSM(t): the signal after multi-band modulation (t) multi-band modulation function, which controls the modulation effect of each frequency band, M opt The optimized modulation signal is input into the quantum state modulation module for processing. Represents the modulation of the signal in the frequency domain, through different frequencies ω n To achieve multi-band modulation; S54, the multi-band modulated signal M QSM (t) Input to the LED drive circuit to adjust the input current I in (t) and input voltage V in (t) duty cycle; S55, the modulation signal M is filtered by an adaptive quantum perturbation filter QSM (t) Perform high-frequency filtering to eliminate high-frequency noise in the signal; S56, the filtered modulation signal M filtered (t) Feedback to the LED driving circuit to further optimize the modulation process of input current and input voltage; S57, dynamically adjust the parameters of the PWM module through the adaptive quantum controller to optimize the effective transmission of the compensation signal and the stability of the light output: Among them, C AQC represents the adaptive quantum control operator, Θ opt is the optimized PWM parameter, Θ init is the initial PWM parameter, The integral of the filtered modulation signal is used to calculate the PWM adjustment amplitude; S58. Use machine learning algorithms to analyze historical data and real-time monitoring information, continuously optimize the modulation strategy of the PWM module, and optimize the light output stability of LED lamps under different working conditions.
6. The LED lamp driving circuit control method for preventing stroboscopic effect according to claim 1, characterized in that: The S6 specifically includes: S61. Install a high-precision photoelectric sensor at the output end of the LED lamp. The sensor is used to monitor the output light intensity of the LED lamp in real time. The output light intensity signal is L(t). S62, input the output light intensity signal L(t) collected by the photoelectric sensor into the light signal processing module, pre-process the light intensity signal through a denoising algorithm based on wavelet transform, and obtain a denoised light intensity signal L filtered 9t); S63, the denoised light intensity signal L filtered (t) is input into the nonlinear time series analysis module, and the phase space reconstruction technology is used to extract the dynamic characteristics X(t) and reconstruction trajectory R(t) of the light intensity signal. Through phase space reconstruction, the denoised signal L filtered (t) is converted into dynamic features X(t) and reconstructed trajectory R(t), which represent the complex dynamic behavior of the signal; S64, inputting the reconstructed dynamic characteristics X(t) into the topological quantum phase detection module, and calculating the stroboscopic index FI(t) of the light intensity signal by quantum phase interference technology; S65, inputting the reconstructed trajectory R(t) into a topological structure analysis module, and calculating the stability index SI(t) of the light output by a topological structure analysis algorithm; S66, input the stroboscopic index FI(t) and the light output stability index SI(t) into the adaptive quantum compensation control module, and dynamically adjust the working parameters of the LED driving circuit through the quantum fuzzy logic control algorithm: in, represents the quantum fuzzy logic control operator, Θ opt is the optimized driving parameter; The adaptive quantum compensation control module uses the flicker index FI(t) and the stability index SI(t) to dynamically adjust the drive parameter Θ opt , optimize the working status of LED lamps; S67, storing the real-time monitoring data L(t), the stroboscopic index FI(t)) and the light output stability index SI(t) into a historical data database; S68. Historical data and real-time data are input into the adaptive model optimization module through the quantum machine learning algorithm to continuously optimize the stroboscopic effect detection algorithm and the light output stability analysis algorithm. The quantum machine learning algorithm uses historical data and real-time data to continuously update and optimize the model.
7. The LED lamp driving circuit control method for preventing stroboscopic effect according to claim 1, characterized in that: The S7 specifically includes: S71, inputting the real-time output light intensity signal monitored in step S6 into a comparison module, the comparison module compares the real-time output light intensity signal with a preset light output standard, and generates an error signal, the error signal reflects the difference between the current light output and the target standard; S72, input the error signal to the adaptive fuzzy control module, process the error signal through the fuzzy logic algorithm, generate a preliminary control signal, and the fuzzy control module combines the machine learning algorithm to update the fuzzy rules and membership functions to optimize the control accuracy; S73, inputting the preliminary control signal into the quantum optimization module, optimizing the control signal using an optimization algorithm based on quantum tunneling effect and quantum superposition state, and generating a final control signal, wherein the quantum optimization module optimizes efficiency and accuracy by searching for an optimal path in a high-dimensional quantum state space; S74, feeding back the final control signal to the LED driving circuit, adjusting the driving current and driving voltage through the adaptive current and voltage regulation technology, optimizing the output light intensity of the LED lamp to stabilize it within the preset light output standard, and dynamically adjusting the driving parameters using the adaptive genetic algorithm to optimize the stability of the light output; S75, real-time monitoring of the adjusted output light intensity, and feeding back the data to the adaptive fuzzy control module and the quantum optimization module for further parameter adjustment and optimization. The feedback mechanism is based on the deep reinforcement learning algorithm, autonomous learning and optimization adjustment strategy; S76, storing the real-time monitoring data, error signal, preliminary control signal and final control signal in a quantum distributed database; S77. Use historical data and real-time monitoring data to periodically train and update the adaptive fuzzy control module and quantum optimization module, continuously optimize the control algorithm, and record each training and update process by integrating blockchain technology.
8. The LED lamp driving circuit control method for preventing stroboscopic effect according to claim 1, characterized in that: The S8 specifically includes: S81, inputting the compensation signal optimized in step S7 into a dynamic compensation module, which dynamically adjusts the compensation parameters based on the real-time monitored output light intensity and current signal to adapt to the current working state; S82, in the dynamic compensation module, the fuzzy control algorithm is used to analyze and adjust the compensation signal in real time to generate a dynamic compensation signal C dynamic (t), the fuzzy control algorithm continuously optimizes the parameters of the compensation signal by comparing the actual monitoring data with the preset standards; S83, the dynamic compensation signal C dynamic (t) is input to the adaptive neural network module. The adaptive neural network further optimizes the compensation signal by training historical data and real-time data to generate the final compensation signal C final (t), the neural network model uses historical data as a training set and real-time data as a correction basis, and continuously adjusts the network weights to optimize the accuracy of the signal; S84, in the adaptive neural network module, using a back propagation algorithm and a recursive neural network to train and optimize the dynamic compensation signal, the back propagation algorithm adjusts the network parameters according to the error gradient, and the recursive neural network processes the time series data; S85, the final compensation signal C final (t) is fed back to the LED driving circuit, and the driving current I in (t and driving voltage V in (t), optimize the output light intensity of the LED lamp to stabilize it within the preset light output standard, and compensate the signal C final (t) Directly act on the driving circuit to adjust the current and voltage to optimize the light intensity stability; S86, real-time monitoring of the adjusted output light intensity, and feeding back the data to the dynamic compensation module and the adaptive neural network module, through further parameter adjustment and optimization, dynamically adapting to power supply fluctuations and load changes, the monitoring data provides a real-time reference for subsequent adjustments, and optimizes the system's ability to respond to changes; S87, storing the real-time monitoring data, dynamic compensation signal and final compensation signal in a historical data database; S88. Use historical data and real-time monitoring data to periodically train and update the adaptive neural network module and dynamic compensation module, continuously optimize the control algorithm, and self-optimize and evolve by integrating reinforcement learning and evolutionary algorithms.
Citation Information
Cited By
Area array light source stroboscopic detection and compensation method and system
CN119357620A
Array light source flicker detection and compensation method and system
CN119357620B
Digital power supply error feedback information generation method, digital power supply control method, system and device and computer equipment
CN119986447A
Anti-interference connector for automobile electronic system
CN120613615A
