A multi-voltage level self-adaptive dc power supply control method and system

Abstract

The present application relates to a kind of multi-voltage level self-adapting DC in-house power supply control method and system, method includes: by applying characteristic test current sequence to each voltage level loop to establish crosstalk coupling coefficient matrix;In the normal operating state of multi-voltage level loop, the original current signal of each loop is synchronously collected;Based on the crosstalk coupling coefficient matrix, the original current signal is cascaded decoupling processing, separates crosstalk noise and generates the real current estimate of each loop;Real current estimate is recursively least square filtering processing, inhibits the estimated element shock caused by load mutation, generates the real current reconstruction value of dynamic optimization;The current reconstruction value of each loop is compared with threshold value, and the current reconstruction value of four loops is accumulated as total load current, to generate the charge-discharge control instruction of energy storage module.

Description

Technical Field

[0001] This invention relates to the field of power supply control technology, specifically to a method and system for controlling DC power supply to a household with multi-voltage level self-adaptation. Background Technology

[0002] In traditional AC-to-DC power supply models, household appliances frequently need to switch between AC and DC, resulting in energy loss and a higher risk of electric shock. With advancements in power electronics technology and the widespread adoption of DC appliances, the DC-to-DC power supply method, which converts grid AC power to safe voltage levels and directly supplies it to end-user devices, has gained increasing attention and promotion. This method aims to improve household power safety and energy efficiency by providing multiple DC voltage output ports to meet the power needs of different types of appliances. Building on this, to meet the varying operating voltage requirements of loads such as lighting, information equipment, and low-power devices, many power supply systems have further adopted multi-voltage-level self-adaptive mechanisms, simultaneously outputting DC power at multiple safe voltage levels while using a shared reference ground.

[0003] While the aforementioned methods can improve power safety and system complexity to some extent, in actual operation, because multiple DC / DC converters share a single reference ground loop, the high-frequency switching ripple generated during the operation of each converter will crosstalk with each other through common ground impedance and spatial coupling paths. This results in the current sampling value of any voltage level loop being mixed with interference components from other loops. Existing technologies mainly focus on energy dispatch and management strategies for multi-voltage-level DC distribution networks, neglecting the crosstalk problem between multiple current detection signals under the condition of a shared reference ground. They lack systematic identification of this coupling effect and corresponding decoupling and correction methods, thus failing to accurately extract the true current information of each output port. Summary of the Invention

[0004] Based on this, the purpose of the present invention is to provide a multi-voltage level self-adaptive DC power supply control method and system capable of cross-loop coupling interference identification and cascaded decoupling for multi-voltage level DC power supply systems with a shared reference ground.

[0005] The objective of this invention is achieved through the following solution:

[0006] In a first aspect, the present invention provides a multi-voltage-level self-adaptive DC power supply control method for residential use, comprising the following steps:

[0007] S1: Apply a preset characteristic test current sequence to the four voltage level circuits in sequence, and simultaneously sample the current of all four voltage level circuits. Then, perform correlation analysis on the sampled current signals and the characteristic test current sequence, calculate the transmission mapping relationship of the coupling effect between each circuit, and generate the crosstalk coupling coefficient matrix.

[0008] S2: Under the normal power supply operation of multiple voltage level circuits, the original current signals of the four voltage level circuits are simultaneously sampled at high frequency to obtain the original current sampling vector containing crosstalk noise.

[0009] S3: Based on the crosstalk coupling coefficient matrix, the original current sampling vector is decoupled in a cascade manner. The original current sampling vector is separated into the target loop current principal component and the crosstalk noise estimation component through frequency band decomposition. The residual coupling interference between each loop is removed from the target loop current principal component by a generalized inverse transform with regularization correction, and the true current estimation value of each voltage level loop is generated.

[0010] S4: Perform recursive minimum variance filtering on the true current estimate, recursively update the current state estimate using the true current estimate from the previous moment and the sampling residual from the current moment, suppress the oscillation of the current estimation operator caused by load changes, and generate the dynamically optimized true current reconstruction value.

[0011] S5: Perform threshold comparison and accumulation processing on the real current reconstruction value. Compare the current reconstruction value of each circuit with the preset overcurrent protection threshold and leakage protection threshold circuit by circuit and cycle by cycle. Accumulate the current reconstruction values ​​of the four circuits to obtain the total load current and generate the charging and discharging control command of the energy storage module. The charging and discharging control command is used to start charging during the off-peak period and to start discharging during the peak period, emergency power supply period or when the energy storage capacity is sufficient.

[0012] In one embodiment, S1 of the multi-voltage level self-adaptive DC power supply control method provided by the present invention specifically includes the following steps:

[0013] S11: Apply a preset characteristic test current sequence to the four voltage level circuits in sequence, and synchronously collect the transient response current signal of each circuit at the sampling frequency. Use the test current sequence as the input vector and the transient response current signal as the output vector to build a training dataset.

[0014] S12: The training dataset is processed for online system identification based on the recursive least squares adaptive filtering algorithm. A parameter recursion mechanism with an adaptive forgetting factor is introduced. The weight distribution of new and old data is balanced by adjusting the size of the forgetting factor in real time. Singular value decomposition is used to correct the ill-conditioned problem of the observation matrix in the identification process. The coupling transmission coefficient between each loop is estimated recursively from the input and output data, and the crosstalk coupling coefficient matrix is ​​constructed.

[0015] S13: Perform matrix structure analysis on the crosstalk coupling coefficient matrix, calculate the condition number of the matrix and the confidence interval of each element, remove abnormal estimates caused by the step edge of the test current during the identification process, and arrange the corrected matrix elements in the loop order to generate a crosstalk coupling coefficient matrix that has been verified for accuracy.

[0016] In one embodiment, step S3 of the multi-voltage-level self-adaptive DC power supply control method provided by the present invention specifically includes the following steps:

[0017] S31: Perform adaptive variational mode decomposition on each channel component of the original current sampling vector. Use the nominal switching frequency and corresponding multiples of the DC / DC converter as the initial center frequency of the mode component. Iteratively optimize the number of modes and penalty factor parameters by using the relative entropy between the original signal and the intrinsic mode function. Separate the original sampling signal of each channel into the main current component of the corresponding loop and the estimated crosstalk noise component coupled from other loops. Calculate the center frequency distribution of the intrinsic mode function components to generate the target loop signal component set and the crosstalk noise component set.

[0018] S32: Perform time-domain waveform reconstruction processing on the main current components of each channel in the target loop signal component set to generate an intermediate current vector after the first stage of decoupling, and perform integral statistical processing on the signal energy in the crosstalk noise component set to calculate the variance distribution of crosstalk noise in each loop and generate a crosstalk noise energy estimation vector.

[0019] S33: Using the crosstalk coupling coefficient matrix as the observation matrix, the intermediate current vector as the sampling vector, and the variance in the crosstalk noise energy estimation vector as the regularization penalty factor, a ridge regression objective function is constructed in combination with the weight matrix. The optimal estimate of the true current is obtained by minimizing the weighted sum of squared residuals with regularization constraints, and the true current estimate of each voltage level circuit is generated.

[0020] In one embodiment, step S4 of the multi-voltage-level self-adaptive DC power supply control method provided by the present invention specifically includes the following steps:

[0021] S41: Construct a state-space model based on the Kalman filter algorithm, take the first sampling point of the real current estimate as the initial estimate of the state vector, establish a linear observation equation with the crosstalk coupling coefficient matrix as the observation matrix, and use the pre-stored noise statistics parameters as the initial noise covariance matrix of the filter to generate the initial state and initial error covariance matrix of the Kalman filter.

[0022] S42: Update the state estimate at each sampling time. Take the real current reconstruction value output after Kalman filtering correction at the previous sampling time as the starting point for state prediction at the current sampling time. Generate the prior state estimate at the current time after linear mapping operation of the state transition matrix. Then, superimpose the error covariance matrix of the previous sampling time with the process noise covariance matrix after mapping through the state transition matrix to generate the prior error covariance matrix at the current time.

[0023] S43: The original current sampling vector at the current moment is mapped by the observation matrix and then the difference is calculated with the prior state estimate to generate the sampling residual vector at the current moment. The Kalman gain matrix is ​​calculated using the prior error covariance matrix and the initial noise covariance matrix. The prior state estimate is corrected and compensated by weighting the sampling residual vector with the Kalman gain matrix, and the posterior state estimate at the current sampling moment is generated as the real current reconstruction value after dynamic optimization.

[0024] In one embodiment, the Kalman gain matrix calculation formula for the multi-voltage level self-adaptive DC power supply control method provided by the present invention is as follows:

[0025]

[0026] in, For the first The regularized dynamic Kalman gain matrix of the step. For the first The prior error covariance matrix of the step, This is the crosstalk coupling coefficient matrix. To measure the noise covariance matrix, For the first The sampling residual covariance matrix of the step. The preset regularization smoothing parameters are: This is the trace operation of a matrix.

[0027] In one embodiment, step S5 of the multi-voltage-level self-adaptive DC power supply control method provided by the present invention specifically includes the following steps:

[0028] S51: Perform cycle-by-cycle sliding window sampling on the actual current reconstruction value of each voltage level circuit, compare the current reconstruction value of each sampling point in the window with the preset overcurrent protection threshold, record the duration of continuous overcurrent protection threshold, and generate overcurrent warning flags for each circuit.

[0029] S52: Compare the actual current reconstructed value of each sampling point in the window with the preset leakage protection threshold, record the duration of continuous exceedance of the leakage protection threshold, and generate leakage warning signs for each circuit.

[0030] S53: The real current reconstruction values ​​of the four voltage level circuits are vector-added according to the sampling time to obtain the instantaneous waveform sequence of the total load current. The instantaneous waveform sequence is then low-pass filtered to eliminate the high-frequency spikes introduced by load switching and generate a smooth total load current time sequence.

[0031] S54: Performs joint logic judgment processing on the smooth total load current time series, real-time grid electricity price period, energy storage battery state of charge, and overcurrent warning flags and leakage warning flags of each circuit. If any circuit has an overcurrent warning flag or leakage warning flag, the energy storage module is prohibited from entering the discharge mode and is kept in charging or standby state, generating a discharge prohibition command. If all circuits have no overcurrent warning flags or leakage warning flags, the corresponding charging control command or discharging control command is generated based on the comparison result between the real-time grid electricity price period and the energy storage battery state of charge. The charging and discharging control command is used to start charging during the off-peak period and to start discharging during the peak period, emergency power supply period, or period when the energy storage capacity is sufficient.

[0032] Secondly, the present invention provides a multi-voltage level self-adaptive DC power supply control system for residential use, which is configured with the following modules:

[0033] The loop coupling coefficient calibration module is used to apply a preset characteristic test current sequence to the four voltage level loops in sequence, and simultaneously sample the current of all four voltage level loops. It also performs correlation analysis on the sampled current signals and the characteristic test current sequence, calculates the transmission mapping relationship of the coupling effect between each loop, and generates a crosstalk coupling coefficient matrix.

[0034] The multi-circuit synchronous sampling module is used to simultaneously perform high-frequency synchronous sampling of the original current signals of four voltage level circuits under normal power supply operation conditions, and obtain the original current sampling vector containing crosstalk noise.

[0035] The cross-loop decoupling correction module is used to perform cascaded decoupling on the original current sampling vector based on the crosstalk coupling coefficient matrix. The original current sampling vector is separated into the target loop current principal component and the crosstalk noise estimation component through frequency band decomposition. The residual coupling interference between each loop is removed from the target loop current principal component by a generalized inverse transform with regularization correction, and the true current estimation value of each voltage level loop is generated.

[0036] The current dynamic filtering and reconstruction module is used to perform recursive minimum variance filtering on the real current estimate. It recursively updates the current state estimate using the real current estimate from the previous moment and the sampling residual from the current moment, suppresses the oscillation of the current estimation operator caused by load changes, and generates the dynamically optimized real current reconstruction value.

[0037] The energy storage collaborative control module is used to perform threshold comparison and accumulation processing on the real current reconstruction value. It compares the current reconstruction value of each circuit with the preset overcurrent protection threshold and leakage protection threshold circuit by circuit and cycle by cycle, and accumulates the current reconstruction values ​​of the four circuits to obtain the total load current. It generates the charging and discharging control command of the energy storage module. The charging and discharging control command is used to start charging during the off-peak period and to start discharging during the peak period, emergency power supply period, or period when the energy storage capacity is sufficient.

[0038] Thirdly, this application provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements any of the above-mentioned multi-voltage level self-adaptive DC power supply control methods.

[0039] Fourthly, this application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements any of the above-mentioned multi-voltage level self-adaptive DC power supply control methods.

[0040] In summary, the multi-voltage-level self-adaptive DC power supply control method provided in this application can accurately describe the crosstalk transmission characteristics between circuits of different voltage levels via a shared reference ground by sequentially applying characteristic test current sequences and establishing a crosstalk coupling coefficient matrix. Based on this, a cascaded decoupling process is performed on the original current sampling vector under normal operating conditions. First, the target current principal component and crosstalk noise component are separated through frequency band decomposition. Then, a generalized inverse transform with regularization correction is used to remove residual coupling interference, effectively eliminating the cross-contamination of current detection by common ground noise from other circuits, allowing the true current estimates of each circuit to escape the root cause of crosstalk misjudgment. Further... The estimated real current value is processed by recursive minimum variance filtering. By recursively correcting the state estimate from the previous moment with the current sampling residual, the technical effect of suppressing the oscillation of the estimation operator caused by load changes can be achieved, thereby outputting a stable and timely real current reconstruction value. Finally, based on the reconstruction value, the overcurrent and leakage thresholds of each loop are compared, and the currents of the four loops are accumulated to generate the charging and discharging control command of the energy storage module. This enables intelligent scheduling of off-peak charging, peak discharging and emergency power supply while ensuring the independent safety protection of each power unit. This is used to complete the adaptive dynamic calibration and safe energy management of the multi-voltage power supply system under the condition of a shared reference ground.

[0041] To better understand and implement this invention, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description

[0042] Figure 1 A flowchart illustrating a multi-voltage-level self-adaptive DC power supply control method provided in this application embodiment;

[0043] Figure 2 This is a schematic diagram of a multi-voltage-level self-adaptive DC power supply control system provided in another embodiment of this application. Detailed Implementation

[0044] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.

[0045] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0046] In one embodiment, such as Figure 1 As shown, a multi-voltage level self-adaptive DC power supply control method is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0047] S1: Apply a preset characteristic test current sequence to the four voltage level circuits in sequence, and simultaneously sample the current of all four voltage level circuits. Then, perform correlation analysis on the sampled current signals and the characteristic test current sequence, calculate the transmission mapping relationship of the coupling effect between each circuit, and generate a crosstalk coupling coefficient matrix.

[0048] Specifically, the system determines characteristic test current sequences for four voltage level circuits. These sequences are pseudo-random sequences with good orthogonality, and the frequency bands of the characteristic test current sequences for each circuit do not overlap, covering the main frequency band of high-frequency switching ripple in the DC / DC converter and avoiding mutual interference between test signals from different circuits. The system uses a synchronous triggering mechanism to sequentially apply the preset characteristic test current sequences to the four voltage level circuits. Only one circuit is applied to at a time, while the other three circuits are in an idle standby state, completing the test of all four circuits in a cyclical manner. The system uses a synchronous clock module to control the current detection modules of the four circuits to sample simultaneously. The sampling duration is three times the duration of the characteristic test current sequence applied to each circuit, ensuring complete capture of the changing characteristics of the test current and crosstalk signals. After the current detection modules acquire the current signals, they are converted into digital signals by an ADC converter and transmitted to the system control core.

[0049] The system performs correlation analysis on the sampled current signals and the characteristic test current sequences of the corresponding circuits. Using cross-correlation analysis, it calculates the cross-correlation coefficient between the target circuit with applied test current and the sampled signals of the other three non-target circuits. Simultaneously, it calculates the autocorrelation coefficient of the target circuit's sampled signal. The crosstalk coupling coefficient between the target circuit and each non-target circuit is obtained by comparing the cross-correlation coefficient to the autocorrelation coefficient. The system repeats this process sequentially, testing and analyzing all four circuits as target circuits, ultimately generating a crosstalk coupling coefficient matrix. In this matrix, elements represent the crosstalk coupling strength between circuits, diagonal elements represent the transmission coefficient of the circuit's own current signal, and off-diagonal elements represent the degree of crosstalk between different circuits.

[0050] S2: Under the normal power supply operation of multiple voltage level circuits, the original current signals of the four voltage level circuits are simultaneously sampled at high frequency to obtain the original current sampling vector containing crosstalk noise.

[0051] Specifically, when the multi-voltage-level DC power supply system enters normal operation, the four voltage-level circuits output corresponding voltages according to load requirements. The DC / DC converters in each circuit operate normally. The current in each circuit not only contains the target current component required for load operation but also includes crosstalk noise components transmitted from other circuits through common ground impedance and spatial coupling paths. The system uses a synchronous clock module to control the current detection modules of the four circuits to perform high-frequency synchronous sampling. The sampling frequency is more than ten times higher than the switching frequency of the DC / DC converter, satisfying the Nyquist sampling theorem, avoiding aliasing of the sampled signals, and ensuring that crosstalk noise components are not lost.

[0052] During synchronous sampling, the current detection modules of the four loops simultaneously acquire the current signals of each loop. After being converted into digital signals by an ADC converter, the system control core integrates them into an original current sampling vector according to the loop sequence. Each element in the vector corresponds to a single sampled current value of a loop at a specific voltage level. The sampling process is continuous, and the original current sampling vector is updated in real time according to the sampling period, providing continuous raw data support for subsequent real-time decoupling and control. The system ensures through continuous sampling that the original current sampling vector can fully reflect the real-time changes of the current in each loop, guaranteeing the integrity of the actual current information after decoupling.

[0053] S3: Based on the crosstalk coupling coefficient matrix, the original current sampling vector is decoupled in a cascade manner. The original current sampling vector is separated into the target loop current principal component and the crosstalk noise estimation component through frequency band decomposition. The residual coupling interference between each loop is removed from the target loop current principal component by a generalized inverse transform with regularization correction, and the true current estimation value of each voltage level loop is generated.

[0054] Specifically, the system employs a cascaded decoupling strategy, achieving crosstalk noise removal in two steps. First, it separates the target current principal component from the crosstalk noise estimation component through frequency band decomposition. Then, it removes residual coupling interference through a generalized inverse transform with regularization correction, ultimately obtaining the true current estimates for each loop. The system performs frequency band decomposition on each element in the original current sampling vector, using a wavelet decomposition algorithm and selecting appropriate wavelet basis functions to distinguish the target current principal component from the crosstalk noise estimation component. The target loop current principal component corresponds to the low-frequency band, which is mainly the frequency range of the load operating current; the crosstalk noise estimation component corresponds to the high-frequency band, which is consistent with the frequency band of the characteristic test current sequence, mainly consisting of high-frequency switching ripple crosstalk generated by the operation of the DC / DC converters in each loop.

[0055] The system achieves initial separation of the two types of components through frequency band decomposition, obtaining the target current principal component vector and the estimated crosstalk noise component vector for each loop. Subsequently, based on the crosstalk coupling coefficient matrix, the system performs a generalized inverse transform with regularization correction on the target current principal component vector to remove residual coupling interference between loops. The system introduces a regularization correction term to improve the ill-conditioned nature of the crosstalk coupling coefficient matrix, enhancing the stability and accuracy of the inverse transform. The system takes the target current principal component vector as input and multiplies it by the generalized inverse matrix with regularization correction to obtain the true current estimate for each loop after removing residual coupling interference. This estimate minimizes the influence of crosstalk from other loops and closely approximates the true load current of each loop.

[0056] S4: Perform recursive minimum variance filtering on the true current estimate, recursively update the current state estimate using the true current estimate from the previous moment and the sampling residual from the current moment, suppress the oscillation of the current estimation operator caused by load changes, and generate the dynamically optimized true current reconstruction value.

[0057] Specifically, household loads exhibit abrupt changes, causing oscillations in the actual current estimate. The recursive minimum variance filter, based on adaptive filtering theory, recursively updates the current state estimate using the previous time-to-time actual current estimate and the current sampling residual. This allows for rapid tracking of dynamic changes in load current, suppressing oscillations and improving the stability and accuracy of current estimation. The system sets initial filtering parameters, using the no-load current values ​​of each circuit as the initial actual current estimate. A forgetting factor is set to balance filtering stability and tracking speed. The value of the forgetting factor must balance filtering stability with the speed of tracking load abrupt changes, ensuring the system can respond quickly during load abrupt changes while maintaining stable filtering performance.

[0058] Furthermore, the system calculates the sampling residual at the current moment, which is obtained by the difference between the product of the principal component of the target loop current at the current moment, the crosstalk coupling coefficient matrix, and the true current estimate at the previous moment. The system updates the true current estimate at the current moment through a recursive minimum variance formula, which includes the calculation of the filter gain matrix and the covariance matrix. The filter gain matrix is ​​calculated from the previous moment's covariance matrix, the transpose of the crosstalk coupling coefficient matrix, and the inverse of the correlation matrix. The covariance matrix is ​​calculated from the identity matrix, the filter gain matrix, the crosstalk coupling coefficient matrix, the previous moment's covariance matrix, and the reciprocal of the forgetting factor. Through the above recursive process, the system dynamically optimizes the true current estimate at each moment, suppressing the oscillation of the current estimation operator caused by sudden load changes, and finally generates a dynamically optimized true current reconstruction vector, which can accurately reflect the true operating current of the loop load at each voltage level.

[0059] S5: Perform threshold comparison and accumulation processing on the real current reconstruction value. Compare the current reconstruction value of each circuit with the preset overcurrent protection threshold and leakage protection threshold circuit by circuit and cycle by cycle. Accumulate the current reconstruction values ​​of the four circuits to obtain the total load current and generate the charging and discharging control command of the energy storage module. The charging and discharging control command is used to start charging during the off-peak period and to start discharging during the peak period, emergency power supply period or when the energy storage capacity is sufficient.

[0060] Specifically, the system presets overcurrent protection thresholds and system leakage protection thresholds for each voltage level circuit. The overcurrent protection threshold is set according to the rated current of each circuit, and the leakage protection threshold complies with household electrical safety standards. The system compares the actual current reconstruction value of each circuit with the threshold on a circuit-by-circuit and sampling-by-sampling-cycle basis. If the actual current reconstruction value of a circuit exceeds its preset overcurrent protection threshold for a reasonable number of sampling cycles, the system immediately generates an overcurrent protection signal, controls the switching device of that circuit to open, and stops supplying power to that circuit until the fault is cleared and manually reset.

[0061] Furthermore, the system determines whether leakage exists by calculating the vector sum of the reconstructed real current values ​​of the four circuits. If the absolute value of the vector sum exceeds the leakage protection threshold for a reasonable number of sampling periods, a leakage protection signal is immediately generated, cutting off the system's main power supply to ensure personal and equipment safety. The system arithmetically sums the reconstructed real current values ​​of the four circuits to obtain the total load current, which is used to determine the current household electricity load status. The system obtains the remaining power of the energy storage module through its power detection mechanism, and the measurement of the remaining power must reach a reasonable accuracy. The system combines the total load current, the remaining energy storage power, and the preset load period to generate charging and discharging control commands for the energy storage module. During off-peak load periods, the system controls the energy storage module to start charging, with the charging current and voltage set according to the characteristics of the energy storage module, until the remaining power reaches a reasonable level and charging stops. During peak load periods, the system controls the energy storage module to start discharging, with the discharging current dynamically adjusted according to the total load current. During emergency power supply periods, the system controls the energy storage module to supply power to important loads separately. When the remaining energy storage power reaches a reasonable level and it is not an off-peak load period, the system controls the energy storage module to start discharging. The charging and discharging control commands are transmitted to the energy storage charging and discharging controller through a preset transmission bus to achieve precise control of the charging and discharging status of the energy storage module.

[0062] In summary, the multi-voltage-level self-adaptive DC power supply control method provided in this application can accurately describe the crosstalk transmission characteristics between circuits of different voltage levels via a shared reference ground by sequentially applying characteristic test current sequences and establishing a crosstalk coupling coefficient matrix. Based on this, a cascaded decoupling process is performed on the original current sampling vector under normal operating conditions. First, the target current principal component and crosstalk noise component are separated through frequency band decomposition. Then, a generalized inverse transform with regularization correction is used to remove residual coupling interference, effectively eliminating the cross-contamination of current detection by common ground noise from other circuits, allowing the true current estimates of each circuit to escape the root cause of crosstalk misjudgment. Further... The estimated real current value is processed by recursive minimum variance filtering. By recursively correcting the state estimate from the previous moment with the current sampling residual, the technical effect of suppressing the oscillation of the estimation operator caused by load changes can be achieved, thereby outputting a stable and timely real current reconstruction value. Finally, based on the reconstruction value, the overcurrent and leakage thresholds of each loop are compared, and the currents of the four loops are accumulated to generate the charging and discharging control command of the energy storage module. This enables intelligent scheduling of off-peak charging, peak discharging and emergency power supply while ensuring the independent safety protection of each power unit. This is used to complete the adaptive dynamic calibration and safe energy management of the multi-voltage power supply system under the condition of a shared reference ground.

[0063] In one embodiment, S1 of the multi-voltage level self-adaptive DC power supply control method provided by the present invention specifically includes the following steps:

[0064] S11: Apply a preset characteristic test current sequence to the four voltage level circuits in sequence, and synchronously collect the transient response current signal of each circuit at the sampling frequency. Use the test current sequence as the input vector and the transient response current signal as the output vector to form a training dataset.

[0065] Specifically, the system sequentially applies a preset characteristic test current sequence to four voltage level circuits. The application process proceeds in a preset order, with only one voltage level circuit receiving the characteristic test current sequence at a time. The other three voltage level circuits remain in their initial operating state. The application process only begins after the current circuit has completed its application, and this cycle continues until all four voltage level circuits have received their characteristic test current sequences. The system synchronously acquires the transient response current signals of each voltage level circuit at a preset sampling frequency. The acquisition process starts and stops synchronously with the application of the characteristic test current sequence, ensuring that the acquired transient response current signals completely correspond to the application process of the characteristic test current sequence.

[0066] Furthermore, when acquiring transient response current signals for each circuit, the system continuously acquires the transient response current signals for each voltage level circuit without interrupting the acquisition process, ensuring the continuity and integrity of the acquired data. The system uses the characteristic test current sequence applied to each circuit as an input vector, and the corresponding acquired transient response current signals of that circuit and the other three circuits as output vectors. Each application process corresponds to a set of input and output vectors. The system integrates all the input and output vectors corresponding to all application processes to form a training dataset. The training dataset contains the input and output data when the characteristic test current sequences are applied to each of the four voltage level circuits.

[0067] S12: The training dataset is processed for online system identification based on the recursive least squares adaptive filtering algorithm. A parameter recursion mechanism with an adaptive forgetting factor is introduced. The weight distribution of new and old data is balanced by adjusting the size of the forgetting factor in real time. Singular value decomposition is used to correct the ill-conditioned problem of the observation matrix in the identification process. The coupling transmission coefficient between each loop is estimated recursively from the input and output data to construct the crosstalk coupling coefficient matrix.

[0068] Specifically, the system invokes the recursive least squares adaptive filtering algorithm to perform online system identification processing on the assembled training dataset. During online system identification, the system does not perform offline data preprocessing; instead, it directly processes the input and output data in the training dataset in real time to achieve real-time estimation of the coupling transfer coefficients. The system introduces a parameter recursion mechanism with an adaptive forgetting factor, embedded in the recursive least squares adaptive filtering algorithm, to adjust the weight distribution of new and old data in the training dataset.

[0069] The system monitors the changing trends of data in the training dataset in real time and adjusts the forgetting factor to ensure a reasonable weight distribution between new and old data during parameter estimation. This avoids excessive influence of old data on current parameter estimation while ensuring that new data reflects the current state of the system promptly. The system utilizes singular value decomposition (SVD) to process the observation matrix during online system identification, addressing its ill-conditioned nature and implementing regularization correction to avoid parameter estimation bias caused by ill-conditioned observation matrices. Through these processes, the system recursively estimates the coupling transfer coefficients between loops at each voltage level from the input and output data of the training dataset. Based on these coupling transfer coefficients, a crosstalk coupling coefficient matrix is ​​constructed, with each element corresponding to the coupling transfer relationship between the loops.

[0070] S13: Perform matrix structure analysis on the crosstalk coupling coefficient matrix, calculate the condition number of the matrix and the confidence interval of each element, remove abnormal estimates caused by the step edge of the test current during the identification process, and arrange the corrected matrix elements in the loop order to generate a crosstalk coupling coefficient matrix that has been verified for accuracy.

[0071] Specifically, the system performs matrix structure analysis on the constructed crosstalk coupling coefficient matrix, calculating the condition number of the matrix through matrix operations. The condition number reflects the numerical stability and ill-conditioned nature of the matrix, providing a basis for matrix accuracy verification. Simultaneously, the system calculates the confidence interval for each element in the crosstalk coupling coefficient matrix, determining the reasonable value range for each element and judging the reasonableness of the estimation based on the confidence interval. Based on the calculated confidence interval, the system filters out abnormal estimates in the crosstalk coupling coefficient matrix. These abnormal estimates originate from interference generated by the step edge of the test current during the application of the characteristic test current sequence, leading to deviations in the identification process. The system removes the filtered abnormal estimates and corrects the matrix elements after removing the abnormal values, ensuring that the corrected matrix elements conform to the actual situation of inter-loop coupling transmission. The system arranges the corrected matrix elements according to a preset order of the four voltage level loops, ensuring a one-to-one correspondence between the matrix elements and each voltage level loop, forming a crosstalk coupling coefficient matrix that has undergone accuracy verification. This matrix accurately reflects the crosstalk coupling relationship between each loop.

[0072] In one embodiment, step S3 of the multi-voltage-level self-adaptive DC power supply control method provided by the present invention specifically includes the following steps:

[0073] S31: Perform adaptive variational mode decomposition on each channel component of the original current sampling vector. Use the nominal switching frequency and corresponding multiples of the DC / DC converter as the initial center frequency of the mode component. Iteratively optimize the number of modes and penalty factor parameters by using the relative entropy between the original signal and the intrinsic mode function. Separate the original sampling signal of each channel into the main current component of the corresponding loop and the estimated crosstalk noise component coupled from other loops. Calculate the center frequency distribution of the intrinsic mode function components to generate the target loop signal component set and the crosstalk noise component set.

[0074] Specifically, the system extracts the nominal switching frequency and its corresponding harmonics of each DC / DC converter, using them as the initial center frequencies of the modal components to provide an initial reference for mode decomposition. The system constructs the objective function for adaptive variational mode decomposition, expressed as:

[0075]

[0076] In the formula, Represents each intrinsic mode function, Indicates the center frequency of each modal component. Represents the Dirac function, Represents the original sampled signal. This represents the penalty factor.

[0077] Furthermore, the system calculates the relative entropy between the original signal and each intrinsic mode function (EMF). Through iterative changes in the relative entropy, it optimizes the number of modes and the penalty factor parameters, enabling the decomposed modal components to accurately distinguish between the principal current component and crosstalk noise components. Through the above decomposition process, the system separates the original sampled signal of each channel into the principal current component of the corresponding loop and the estimated crosstalk noise component coupled from other loops. The system calculates the center frequency distribution of each EMF component, and based on the center frequency distribution, selects the EMFs corresponding to the principal current component of the corresponding loop and the EMFs corresponding to the crosstalk noise. These are then integrated to generate the target loop signal component set and the crosstalk noise component set.

[0078] S32: Perform time-domain waveform reconstruction processing on the principal current components of each channel in the target loop signal component set to generate an intermediate current vector after first-stage decoupling. Perform integral statistical processing on the signal energy in the crosstalk noise component set to calculate the variance distribution of crosstalk noise in each loop and generate a crosstalk noise energy estimation vector.

[0079] Specifically, the system performs time-domain waveform reconstruction processing on the principal current components of each channel in the target loop signal component set. The system extracts the intrinsic mode functions (EMFs) corresponding to each channel in the target loop signal component set, and achieves time-domain waveform reconstruction through the superposition of the mode components. The reconstruction formula is as follows:

[0080]

[0081] In the formula, This represents the time-domain waveform of the reconstructed principal current component of the i-th channel. This represents the set of intrinsic mode function indices corresponding to the i-th channel in the target loop signal component set. This represents the intrinsic mode function of the corresponding index. Through the above reconstruction process, the system generates an intermediate current vector after the first stage of decoupling. This intermediate current vector contains the current signals of each channel after initial crosstalk removal. Simultaneously, the system performs integral statistical processing on the signal energy in the crosstalk noise component set, calculating the energy integral of the crosstalk noise component corresponding to each channel, and then calculates the variance distribution of the crosstalk noise for each loop based on the energy integral results. The system integrates the variances of the crosstalk noise for each loop in channel order to generate a crosstalk noise energy estimation vector, which quantifies the intensity of the crosstalk noise in each loop.

[0082] S33: Using the crosstalk coupling coefficient matrix as the observation matrix, the intermediate current vector as the sampling vector, and the variance in the crosstalk noise energy estimation vector as the regularization penalty factor, a ridge regression objective function is constructed in combination with the weight matrix. The optimal estimate of the true current is obtained by minimizing the weighted sum of squared residuals with regularization constraints, and the true current estimate of each voltage level circuit is generated.

[0083] Specifically, the system uses the crosstalk coupling coefficient matrix as the observation matrix, the intermediate current vector as the sampling vector, and the variance in the crosstalk noise energy estimation vector as the regularization penalty factor. A weight matrix is ​​also introduced to construct a ridge regression objective function. The objective function expression is:

[0084]

[0085] In the formula, Represents the true current estimation vector. Represents the crosstalk coupling coefficient matrix. This represents the intermediate current vector. Denotes the weighted norm. Represents the weight matrix. This represents the i-th variance in the crosstalk noise energy estimation vector. This represents the i-th element in the true current estimation vector.

[0086] The system solves the ridge regression objective function by minimizing the weighted sum of squared residuals with regularization constraints to obtain the optimal estimate of the true current. During the solution process, the system uses a weight matrix to balance the influence of current signals from each channel and a regularization penalty factor to suppress estimation bias caused by ill-conditioned observation matrices, ensuring the stability of the estimation results. Through the above solution process, the system generates true current estimates for each voltage level loop. These estimates have been sufficiently filtered out of crosstalk interference between loops and can accurately reflect the actual load current of each loop.

[0087] In one embodiment, step S4 of the multi-voltage-level self-adaptive DC power supply control method provided by the present invention specifically includes the following steps:

[0088] S41: Construct a state-space model based on the Kalman filter algorithm, take the first sampling point of the real current estimate as the initial estimate of the state vector, establish a linear observation equation with the crosstalk coupling coefficient matrix as the observation matrix, and use the pre-stored noise statistical parameters as the initial noise covariance matrix of the filter to generate the initial state and initial error covariance matrix of the Kalman filter.

[0089] Specifically, the system constructs a state-space model based on the Kalman filter algorithm. This model includes state equations and observation equations, used to describe the dynamic changes of the real current signal and the observation process. The state equations constructed by the system are as follows:

[0090]

[0091] In the formula, This is the state estimate at the k-th sampling time. This is the predicted state value at the (k+1)th sampling time. This is the state transition matrix, used to characterize the propagation relationship of state vectors between adjacent sampling times. The process noise input matrix, Let be the process noise vector, which follows a zero-mean Gaussian distribution.

[0092] Furthermore, the system extracts the first sampling point of the true current estimate and uses it as the initial estimate of the state vector, providing a starting reference for the iterative calculation of the Kalman filter. The system uses the crosstalk coupling coefficient matrix as the observation matrix to establish a linear observation equation, which is:

[0093]

[0094] In the formula, Let be the original current sampling vector at the k-th sampling time. This is the crosstalk coupling coefficient matrix. The measurement noise vector follows a zero-mean Gaussian distribution. The system calls pre-stored noise statistics parameters as the initial noise covariance matrix of the filter. The process noise covariance matrix is ​​determined by the statistical properties of the process noise vector, and the measurement noise covariance matrix is ​​determined by the statistical properties of the measurement noise vector. Through the above initialization process, the system generates the initial state and initial error covariance matrix of the Kalman filter. The initial error covariance matrix is ​​used to characterize the uncertainty of the initial state estimate.

[0095] S42: Update the state estimate at each sampling time. Take the real current reconstruction value output after Kalman filtering correction at the previous sampling time as the starting point for state prediction at the current sampling time. Generate the prior state estimate at the current time after linear mapping operation of the state transition matrix. Then, superimpose the error covariance matrix of the previous sampling time with the process noise covariance matrix after mapping through the state transition matrix to generate the prior error covariance matrix at the current time.

[0096] Specifically, the system updates the state estimate at each sampling moment, performing state prediction and error covariance prediction during the time update process, providing prior information for subsequent measurement updates. The system extracts the actual current reconstruction value output after Kalman filtering correction from the previous sampling moment and uses it as the starting point for state prediction at the current sampling moment, ensuring the continuity and correlation of state prediction. The system inputs this state prediction starting point into the state transition matrix, and through linear mapping operations on the state transition matrix, generates the prior state estimate for the current moment. The calculation of the prior state estimate follows the formula:

[0097]

[0098] In the formula, This is the prior state estimate at the (k+1)th sampling time. Let be the posterior state estimate at the k-th sampling time. Let be the mathematical expectation of the process noise vector.

[0099] Furthermore, the system extracts the error covariance matrix from the previous sampling time, performs a linear mapping operation on it using the state transition matrix, and obtains the mapped error covariance matrix. The system then superimposes this mapped error covariance matrix with the process noise covariance matrix to generate the prior error covariance matrix for the current time. The calculation of the prior error covariance matrix follows the formula:

[0100]

[0101] In the formula, Let be the prior error covariance matrix at the (k+1)th sampling time. Let be the posterior error covariance matrix at the k-th sampling time. This is the transpose of the state transition matrix. The process noise covariance matrix is... This is the transpose of the process noise input matrix. After the time update is completed, the system obtains the prior state estimate and prior error covariance matrix at the current moment, providing a basis for the correction calculation in the measurement update phase.

[0102] S43: The original current sampling vector at the current moment is mapped by the observation matrix and then the difference is calculated with the prior state estimate to generate the sampling residual vector at the current moment. The Kalman gain matrix is ​​calculated using the prior error covariance matrix and the initial noise covariance matrix. The prior state estimate is corrected and compensated by weighting the sampling residual vector with the Kalman gain matrix, and the posterior state estimate at the current sampling moment is generated as the real current reconstruction value after dynamic optimization.

[0103] Specifically, the system performs measurement updates, correcting the prior state estimate using current observation information to generate a dynamically optimized true current reconstruction value. The system extracts the original current sampling vector at the current moment and performs a linear mapping operation on it using the observation matrix to obtain the mapped observation vector. The system then subtracts the mapped observation vector from the prior state estimate to generate the sampling residual vector at the current moment. The calculation of the sampling residual vector follows the formula:

[0104]

[0105] In the formula, Let be the sampling residual vector at the (k+1)th sampling time. Let be the original current sampling vector at the (k+1)th sampling time. This is the result of mapping the prior state estimates to the observation matrix.

[0106] Preferably, the system calculates the Kalman gain matrix using the prior error covariance matrix and the initial noise covariance matrix. The formula for calculating the Kalman gain matrix is ​​as follows:

[0107]

[0108] in, For the first The regularized dynamic Kalman gain matrix of the step. For the first The prior error covariance matrix of the step, This is the crosstalk coupling coefficient matrix. To measure the noise covariance matrix, For the first The sampling residual covariance matrix of the step. The preset regularization smoothing parameters are: This is the trace operation of the matrix. The system uses the Kalman gain matrix to weight the sampled residual vector to obtain the correction compensation amount. This correction compensation amount is then superimposed with the prior state estimate to generate the posterior state estimate at the current sampling time. This posterior state estimate is the real current reconstruction value after dynamic optimization, which can effectively suppress the oscillation of the current estimation operator caused by load changes and accurately reflect the actual load current of each voltage level circuit.

[0109] In one embodiment, step S5 of the multi-voltage-level self-adaptive DC power supply control method provided by the present invention specifically includes the following steps:

[0110] S51: Perform cycle-by-cycle sliding window sampling on the actual current reconstruction value of each voltage level circuit, compare the current reconstruction value of each sampling point in the window with the preset overcurrent protection threshold, record the duration of continuous overcurrent protection threshold, and generate overcurrent warning flags for each circuit.

[0111] Specifically, the system performs cycle-by-cycle sliding window sampling of the actual current reconstruction values ​​for each voltage level circuit. The length of the sliding window matches the current sampling period, and the sampling process is continuous. The sampling data within the sliding window is updated once per sampling period to ensure that the data within the window reflects the current changes in the current of the current circuit in real time. The system extracts the current reconstruction value of each sampling point within the sliding window and compares it with a preset overcurrent protection threshold. The comparison is performed sequentially according to the sampling point order, ensuring no sampling point is missed. The system records the comparison result for each sampling point. When the current reconstruction value exceeds the overcurrent protection threshold, a timing mechanism is activated to record the duration of continuous exceedance of the overcurrent protection threshold. The system constructs a judgment formula for the overcurrent warning flag, expressed as:

[0112]

[0113] In the formula, This is the overcurrent warning flag for the i-th voltage level circuit at the k-th sampling time. Let be the duration for which the i-th voltage level circuit continuously exceeds the overcurrent protection threshold at the k-th sampling time. The overcurrent warning time threshold is defined above. Based on the formula above, the system judges the continuous timeout length for each voltage level circuit. When the continuous timeout length reaches the preset threshold, an overcurrent warning flag is generated as valid; when the continuous timeout length does not reach the preset threshold, the overcurrent warning flag is generated as invalid. The system integrates the overcurrent warning flags for each voltage level circuit in circuit order to complete the generation of overcurrent warning flags for each circuit.

[0114] S52: Compare the actual current reconstructed value of each sampling point in the window with the preset leakage protection threshold, record the duration of continuous leakage protection threshold, and generate leakage warning signs for each circuit.

[0115] Specifically, the system continues the cycle-by-cycle sliding window sampling method to sample the actual current reconstruction values ​​of circuits at each voltage level. The sliding window setting is consistent with step S51 to ensure the consistency and continuity of the sampled data. The system extracts the actual current reconstruction value of each sampling point within the sliding window and compares it with a preset leakage protection threshold. The comparison is performed synchronously with the sampling process, and the comparison result of each sampling point is obtained in real time. The system monitors the comparison results in real time. When the actual current reconstruction value exceeds the leakage protection threshold, an independent timing mechanism is activated to record the duration for which the circuit continuously exceeds the leakage protection threshold. The timing process does not interfere with the overcurrent timing to ensure timing accuracy. The system constructs a judgment formula for the leakage warning sign, the expression of which is:

[0116]

[0117] in, This is a step function that outputs 1 when the input value is greater than 0, and 0 otherwise. This is the leakage warning flag for the i-th voltage level circuit at the k-th sampling time. This represents the reconstructed current value of the i-th loop at the m-th sampling time. The leakage current protection threshold, This represents the number of sampling points within the sliding window. The sampling period is specified above. The system calculates the leakage current warning flag for each circuit using the formula above. When the calculation result reaches the preset leakage current warning threshold, the leakage current warning flag is valid; otherwise, it is invalid. The system generates leakage current warning flags for each circuit, which, together with the overcurrent warning flags, form a safety protection and early warning system.

[0118] S53: Perform vector addition on the real current reconstructed values ​​of the four voltage level circuits at the sampling time to obtain the instantaneous waveform sequence of the total load current. Then, perform low-pass filtering on the instantaneous waveform sequence to eliminate the high-frequency spikes introduced by load switching and generate a smooth total load current time sequence.

[0119] Specifically, the system performs vector addition on the actual current reconstructed values ​​of the four voltage level circuits at each sampling time. Each sampling time corresponds to a set of actual current reconstructed values ​​for the four circuits. The system uses the four current reconstructed values ​​at the same sampling time as vector elements and performs vector superposition to obtain the instantaneous value of the total load current at that sampling time. The system performs vector addition sequentially according to the sampling time to obtain a series of instantaneous values ​​of the total load current. After integration, a sequence of instantaneous waveforms of the total load current is formed, which can reflect the real-time changes in the overall household electricity load.

[0120] Furthermore, the system performs low-pass filtering on the instantaneous waveform sequence of the total load current to eliminate high-frequency spikes introduced during load switching, ensuring the smoothness of the total load current signal. The system employs an improved low-pass filtering algorithm, with the following filtering formula:

[0121]

[0122] In the formula, This represents the total load current value after filtering at the k-th sampling time. These are dynamic filter coefficients that adaptively adjust as the instantaneous waveform sequence changes. This represents the total load current value after filtering at the (k-1)th sampling time. Let be the reconstructed value of the actual current of the i-th voltage level circuit at the k-th sampling time. The system performs point-by-point filtering on the instantaneous waveform sequence using the above filtering formula, filtering out high-frequency spike components to generate a smooth total load current time series, which can accurately reflect the changing trend of the overall power load.

[0123] S54: Performs joint logic judgment processing on the smooth total load current time series, real-time grid electricity price period, energy storage battery state of charge, and overcurrent warning flags and leakage warning flags of each circuit. If any circuit has an overcurrent warning flag or leakage warning flag, the energy storage module is prohibited from entering the discharge mode and is kept in charging or standby state, generating a discharge prohibition command. If all circuits have no overcurrent warning flags or leakage warning flags, the corresponding charging control command or discharging control command is generated based on the comparison result between the real-time grid electricity price period and the energy storage battery state of charge. The charging and discharging control command is used to start charging during the off-peak period and to start discharging during the peak period, emergency power supply period, or period when the energy storage capacity is sufficient.

[0124] Specifically, the system performs joint logic judgment processing on the smoothed total load current time series, real-time grid electricity price periods, energy storage battery state of charge, and overcurrent and leakage warning flags of each circuit. The joint logic judgment is performed sequentially according to preset logic rules, first judging the safety warning status, and then judging the energy storage control status. The system constructs a joint logic judgment formula, the expression of which is:

[0125]

[0126] In the formula, This is the energy storage control command at the k-th sampling time. For charging or discharging control commands. This is a command to prohibit discharge.

[0127] The system performs logical judgments using the above formulas. If any circuit has an overcurrent warning or leakage warning, the logical judgment result is a safety anomaly. The system prohibits the energy storage module from entering discharge mode, maintaining the current charging or standby state of the energy storage module and generating a discharge prohibition command to ensure the safety of the power supply system and the load. If all circuits have no overcurrent or leakage warnings, the logical judgment result is a safety normality. The system generates corresponding charging or discharging control commands based on a comparison between the real-time grid electricity price and the energy storage battery's state of charge. The charging and discharging control commands follow preset control logic, controlling the energy storage module to charge during off-peak load periods and discharging during peak load periods, emergency power supply periods, or periods with sufficient energy storage capacity. This achieves energy efficiency optimization and reliable operation of the power supply system. The system outputs the generated charging and discharging control commands to the control circuit of the energy storage module to complete the execution of the control commands.

[0128] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0129] Based on the same inventive concept, this application also provides a multi-voltage-level self-adaptive DC power supply control device for implementing the multi-voltage-level self-adaptive DC power supply control method described above. The solution provided by this device is similar to the implementation described in the above method. Therefore, the specific limitations of the one or more embodiments of the multi-voltage-level self-adaptive DC power supply control device provided below can be found in the limitations of the multi-voltage-level self-adaptive DC power supply control method described above, and will not be repeated here.

[0130] Preferably, such as Figure 2 As shown, the present invention provides a multi-voltage level self-adaptive DC power supply control system 600, which is configured with the following modules:

[0131] The loop coupling coefficient calibration module 610 is used to sequentially apply a preset characteristic test current sequence to four voltage level loops, simultaneously sample the current of all four voltage level loops, perform correlation analysis on the sampled current signal and the characteristic test current sequence, calculate the transmission mapping relationship of the coupling effect between each loop, and generate a crosstalk coupling coefficient matrix.

[0132] The multi-circuit synchronous sampling module 620 is used to simultaneously perform high-frequency synchronous sampling of the original current signals of four voltage level circuits under normal power supply operation of multiple voltage level circuits, and obtain the original current sampling vector containing crosstalk noise.

[0133] The cross-loop decoupling correction module 630 is used to perform cascaded decoupling on the original current sampling vector based on the crosstalk coupling coefficient matrix. The original current sampling vector is separated into the target loop current principal component and the crosstalk noise estimation component through frequency band decomposition. The residual coupling interference between each loop is removed from the target loop current principal component by a generalized inverse transform with regularization correction, and the true current estimation value of each voltage level loop is generated.

[0134] The current dynamic filtering and reconstruction module 640 is used to perform recursive minimum variance filtering on the real current estimate. It recursively updates the current state estimate using the real current estimate from the previous moment and the sampling residual from the current moment, suppresses the oscillation of the current estimation operator caused by load changes, and generates the dynamically optimized real current reconstruction value.

[0135] The energy storage collaborative control module 650 is used to perform threshold comparison and accumulation processing on the real current reconstruction value. It compares the current reconstruction value of each circuit with the preset overcurrent protection threshold and leakage protection threshold circuit by circuit and cycle by cycle, and accumulates the current reconstruction values ​​of the four circuits to obtain the total load current. It generates the charging and discharging control command of the energy storage module. The charging and discharging control command is used to start charging during the off-peak period and to start discharging during the peak period, emergency power supply period, or period when the energy storage capacity is sufficient.

[0136] In one embodiment, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described multi-voltage level self-adaptive DC power supply control method for household power supply.

[0137] In one embodiment, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described multi-voltage level self-adaptive DC power supply control method for residential use.

[0138] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0139] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0140] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A multi-voltage level self-adaptive DC power supply control method, characterized in that, Includes the following steps: S1: Apply a preset characteristic test current sequence to the four voltage level circuits in sequence, and simultaneously sample the current of all four voltage level circuits. Perform correlation analysis on the sampled current signals and the characteristic test current sequence, calculate the transmission mapping relationship of the coupling effect between each circuit, and generate a crosstalk coupling coefficient matrix. S2: Under the normal power supply operation of multiple voltage level circuits, the original current signals of the four voltage level circuits are simultaneously sampled at high frequency to obtain the original current sampling vector containing crosstalk noise. S3: Based on the crosstalk coupling coefficient matrix, the original current sampling vector is decoupled in a cascade manner. The original current sampling vector is separated into the target loop current principal component and the crosstalk noise estimation component through frequency band decomposition. The residual coupling interference between each loop is removed from the target loop current principal component by a generalized inverse transform with regularization correction, and the true current estimation value of each voltage level loop is generated. S4: Perform recursive minimum variance filtering on the real current estimate, recursively update the current state estimate using the real current estimate from the previous moment and the sampling residual from the current moment, suppress the oscillation of the current estimation operator caused by load changes, and generate the dynamically optimized real current reconstruction value. S5: Perform threshold comparison and accumulation processing on the real current reconstruction value, compare the current reconstruction value of each circuit with the preset overcurrent protection threshold and leakage protection threshold circuit by circuit and cycle by cycle, and accumulate the current reconstruction values ​​of the four circuits to obtain the total load current, and generate the charging and discharging control command of the energy storage module. The charging and discharging control command is used to start charging during the off-peak period and to start discharging during the peak period, emergency power supply period or period when the energy storage capacity is sufficient.

2. The method according to claim 1, characterized in that, S1 includes: S11: Apply a preset characteristic test current sequence to the four voltage level circuits in sequence, and synchronously collect the transient response current signal of each circuit at the sampling frequency. Use the test current sequence as the input vector and the transient response current signal as the output vector to build a training dataset. S12: The training dataset is processed for online system identification based on the recursive least squares adaptive filtering algorithm. A parameter recursion mechanism with an adaptive forgetting factor is introduced. The weight distribution of new and old data is balanced by adjusting the size of the forgetting factor in real time. Singular value decomposition is used to correct the ill-conditioned problem of the observation matrix in the identification process. The coupling transmission coefficient between each loop is estimated recursively from the input and output data, and the crosstalk coupling coefficient matrix is ​​constructed. S13: Perform matrix structure analysis on the crosstalk coupling coefficient matrix, calculate the condition number of the matrix and the confidence interval of each element, remove abnormal estimates caused by the step edge of the test current during the identification process, and arrange the corrected matrix elements in the loop order to generate a crosstalk coupling coefficient matrix that has been verified for accuracy.

3. The method according to claim 1, characterized in that, S3 includes: S31: Perform adaptive variational mode decomposition on each channel component of the original current sampling vector. Use the nominal switching frequency and corresponding multiples of the DC / DC converter as the initial center frequency of the mode component. Iteratively optimize the number of modes and penalty factor parameters by using the relative entropy between the original signal and the intrinsic mode function. Separate the original sampling signal of each channel into the main current component of the corresponding loop and the estimated crosstalk noise component coupled from other loops. Calculate the center frequency distribution of the intrinsic mode function components to generate the target loop signal component set and the crosstalk noise component set. S32: Perform time-domain waveform reconstruction processing on the main current components of each channel in the target loop signal component set to generate an intermediate current vector after first-stage decoupling, and perform integral statistical processing on the signal energy in the crosstalk noise component set to calculate the variance distribution of crosstalk noise in each loop and generate a crosstalk noise energy estimation vector. S33: Using the crosstalk coupling coefficient matrix as the observation matrix, the intermediate current vector as the sampling vector, and the variance in the crosstalk noise energy estimation vector as the regularization penalty factor, a ridge regression objective function is constructed in combination with the weight matrix. The optimal estimate of the true current is obtained by minimizing the weighted sum of squared residuals with regularization constraints, thereby generating the true current estimate of each voltage level circuit.

4. The method according to claim 1, characterized in that, S4 includes: S41: Construct a state-space model based on the Kalman filter algorithm, take the first sampling point of the real current estimate as the initial estimate of the state vector, establish a linear observation equation with the crosstalk coupling coefficient matrix as the observation matrix, and use the pre-stored noise statistics parameters as the initial noise covariance matrix of the filter to generate the initial state and initial error covariance matrix of the Kalman filter. S42: Update the state estimate at each sampling time. Take the real current reconstruction value output after Kalman filtering correction at the previous sampling time as the starting point for state prediction at the current sampling time. Generate the prior state estimate at the current time after linear mapping operation of the state transition matrix. Then, superimpose the error covariance matrix of the previous sampling time with the process noise covariance matrix after mapping through the state transition matrix to generate the prior error covariance matrix at the current time. S43: The original current sampling vector at the current moment is mapped by the observation matrix and the prior state estimate is subtracted to generate the sampling residual vector at the current moment. The Kalman gain matrix is ​​calculated using the prior error covariance matrix and the initial noise covariance matrix. The prior state estimate is corrected and compensated by weighting the sampling residual vector with the Kalman gain matrix to generate the posterior state estimate at the current sampling moment as the dynamically optimized real current reconstruction value.

5. The method according to claim 4, characterized in that, The formula for calculating the Kalman gain matrix is ​​as follows: in, For the first The regularized dynamic Kalman gain matrix of the step. For the first The prior error covariance matrix of the step, This is the crosstalk coupling coefficient matrix. To measure the noise covariance matrix, For the first The sampling residual covariance matrix of the step. The preset regularization smoothing parameters are: This is the trace operation of a matrix.

6. The method according to any one of claims 1-5, characterized in that, S5 includes: S51: Perform cycle-by-cycle sliding window sampling on the actual current reconstruction value of each voltage level circuit, compare the current reconstruction value of each sampling point in the window with the preset overcurrent protection threshold, record the duration of continuous overcurrent protection threshold, and generate overcurrent warning flags for each circuit. S52: Compare the actual current reconstructed value of each sampling point in the window with the preset leakage protection threshold, record the duration of continuous exceedance of the leakage protection threshold, and generate leakage warning signs for each circuit. S53: Perform vector addition on the actual current reconstructed values ​​of the four voltage level circuits at the sampling time to accumulate the instantaneous waveform sequence of the total load current, and perform low-pass filtering on the instantaneous waveform sequence to eliminate the high-frequency spikes introduced by load switching, thereby generating a smooth total load current time sequence. S54: Perform joint logic judgment processing on the smoothed total load current time series, real-time grid electricity price period, energy storage battery state of charge, and the overcurrent warning flag and leakage warning flag of each circuit. If any circuit has an overcurrent warning flag or leakage warning flag, the energy storage module is prohibited from entering the discharge mode and is kept in charging or standby state, and a discharge prohibition command is generated. If all circuits have no overcurrent warning flag or leakage warning flag, a corresponding charging control command or discharging control command is generated based on the comparison result between the real-time grid electricity price period and the energy storage battery state of charge. The charging and discharging control command is used to start charging during the off-peak period and to start discharging during the peak period, emergency power supply period, or period when the energy storage capacity is sufficient.

7. A multi-voltage level self-adaptive DC power supply control system, characterized in that, The system includes: The loop coupling coefficient calibration module is used to sequentially apply a preset characteristic test current sequence to four voltage level loops, simultaneously sample the current of all four voltage level loops, perform correlation analysis between the sampled current signals and the characteristic test current sequence, calculate the transmission mapping relationship of the coupling effect between each loop, and generate a crosstalk coupling coefficient matrix. The multi-circuit synchronous sampling module is used to simultaneously perform high-frequency synchronous sampling of the original current signals of four voltage level circuits under normal power supply operation conditions, and obtain the original current sampling vector containing crosstalk noise. The cross-loop decoupling correction module is used to perform cascaded decoupling on the original current sampling vector based on the crosstalk coupling coefficient matrix. The original current sampling vector is separated into the target loop current principal component and the crosstalk noise estimation component through frequency band decomposition. The residual coupling interference between each loop is removed from the target loop current principal component by a generalized inverse transform with regularization correction, and the true current estimation value of each voltage level loop is generated. The current dynamic filtering and reconstruction module is used to perform recursive minimum variance filtering on the real current estimate, recursively update the current state estimate using the real current estimate from the previous moment and the sampling residual from the current moment, suppress the oscillation of the current estimation operator caused by load changes, and generate a dynamically optimized real current reconstruction value. The energy storage collaborative control module is used to perform threshold comparison and accumulation processing on the real current reconstruction value. It compares the current reconstruction value of each circuit with the preset overcurrent protection threshold and leakage protection threshold circuit by circuit and cycle by cycle, and accumulates the current reconstruction values ​​of the four circuits to obtain the total load current. It generates the charging and discharging control command of the energy storage module. The charging and discharging control command is used to start charging during the off-peak period and start discharging during the peak period, emergency power supply period, or period when the energy storage capacity is sufficient.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 6.

Images