A wireless sensor dynamic voltage regulation method and system based on load current prediction

Abstract

The present application relates to the technical field of power control, and discloses a wireless sensor dynamic voltage regulation method and system based on load current prediction, the method comprising: performing adaptive sliding window processing on load current sampling values to obtain a load current sequence; constructing a multi-dimensional load feature vector according to the statistical distribution characteristics and the frequency domain wavelet energy characteristics of the difference sequence in the load current sequence; inputting the multi-dimensional load feature vector into a preset load mode expert rule library to identify a load change mode and a load current trend and output a certainty score; matching a corresponding voltage regulation scheme in a pre-stored voltage configuration strategy; generating a control signal according to the voltage regulation scheme and the certainty score; adjusting the duty cycle of the control signal to obtain an optimized control signal; monitoring the area error integral of the actual load current trajectory and the prediction trend envelope line, and adjusting the feature matching degree threshold and the overall duty cycle; and the present application can improve the efficiency of wireless sensor dynamic voltage regulation.

Description

Technical Field

[0001] This invention relates to the field of power control technology, and in particular to a method and system for dynamic voltage regulation using wireless sensors based on load current prediction. Background Technology

[0002] Existing technologies have significant shortcomings in the processing of load current data from wireless sensors. They do not perform dynamic adaptive window optimization on the sampled load current values, but only use a fixed-size window for serialization. This cannot adapt to the instantaneous dynamic characteristics of the current, resulting in the generated current sequence easily missing key change details or containing too much invalid sampled data, failing to truly reflect the current change pattern. At the same time, feature extraction only focuses on single-dimensional information, without combining the statistical distribution characteristics of the difference sequence and the frequency domain wavelet energy characteristics to construct a multi-dimensional feature vector. This makes it difficult to comprehensively capture the temporal change pattern and frequency domain characteristics of the current, and the generated feature data lacks representativeness. The basic data provided for subsequent load pattern recognition and current trend prediction is of low quality, directly affecting the accuracy of voltage regulation decisions.

[0003] Existing technologies have significant shortcomings in the strategy matching and closed-loop optimization stages of dynamic voltage regulation for wireless sensors. When matching voltage regulation schemes, they fail to comprehensively consider the characteristic parameters of load change patterns and current trends for quantitative calculation, relying solely on experience or single parameter matching, which cannot accurately select a scheme suitable for the current load state. After generating control signals, they do not dynamically adjust the duty cycle based on the transient characteristics of current trends, nor do they perform ripple suppression processing on the signals, resulting in the control signals being unable to accurately drive multi-phase power supplies and easily causing voltage output fluctuations. Furthermore, they lack dynamic monitoring and parameter correction mechanisms for the actual regulation effect, failing to calculate the area error integral between the actual current trajectory and the predicted trend envelope to evaluate regulation accuracy, and are unable to adaptively adjust the feature matching degree threshold and overall duty cycle of load pattern recognition based on error conditions. After long-term use, the voltage regulation accuracy continues to decline, making it difficult to meet the power supply stability and energy efficiency requirements of wireless sensors. Summary of the Invention

[0004] This invention provides a method and system for dynamic voltage regulation of wireless sensors based on load current prediction, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a method for dynamic voltage regulation of wireless sensors based on load current prediction, comprising:

[0006] S1. Perform dynamic adaptive sliding window processing on the load current sample values ​​to obtain the load current sequence of the load current sample values;

[0007] S2. Based on the statistical distribution characteristics and frequency domain wavelet energy characteristics of the differential sequence in the load current sequence, construct a multidimensional load feature vector of the load current sequence;

[0008] S3. Input the multidimensional load feature vector into the preset load mode expert rule base to identify the current load change mode and load current trend, and output the deterministic score of the load current trend.

[0009] S4. Based on the load change pattern and the load current trend, match the voltage regulation scheme corresponding to the pre-stored voltage configuration strategy;

[0010] S5. Based on the voltage regulation scheme and the deterministic score, generate the control signal for the multi-phase switching strategy;

[0011] S6. Based on the transient characteristics of the load current trend, dynamically adjust the duty cycle of the control signal to obtain an optimized control signal for the multiphase power supply.

[0012] S7. Execute the optimized control signal, monitor the area error integral between the actual load current trajectory and the predicted trend envelope, and adaptively adjust the feature matching degree threshold and the overall duty cycle of the preset load mode expert rule base according to the area error integral.

[0013] In a preferred embodiment, the step of performing dynamic adaptive sliding window processing on the load current sample values ​​to obtain the load current sequence of the load current sample values ​​includes:

[0014] The instantaneous dynamic characteristics of the load current sample values ​​are analyzed to obtain the dynamic change characteristics of the load current sample values;

[0015] The sampling window is optimized in real time based on the dynamic change characteristics to obtain the optimized sliding window size of the sampling window;

[0016] The load current sample values ​​are serialized based on the optimized sliding window size to obtain the load current sequence of the load current sample values.

[0017] In a preferred embodiment, constructing a multidimensional load feature vector for the load current sequence based on the statistical distribution characteristics and frequency domain wavelet energy characteristics of the differential sequence in the load current sequence includes:

[0018] The load current sequence is subjected to trend quantization processing to obtain the load current difference sequence of the load current sequence;

[0019] Extract the statistical distribution characteristics of the load current differential sequence;

[0020] The load current sequence is subjected to a multi-scale wavelet transform to obtain the multi-scale wavelet coefficients of the load current sequence.

[0021] Energy distribution analysis is performed on the multi-scale wavelet coefficients to obtain the frequency domain wavelet energy characteristics of the load current sequence;

[0022] The statistical distribution features and the frequency domain wavelet energy features are fused to obtain the multidimensional load feature vector of the load current sequence.

[0023] In a preferred embodiment, the step of inputting the multidimensional load feature vector into a preset load pattern expert rule base to identify the current load change pattern and load current trend, and simultaneously outputting a deterministic score for the load current trend, includes:

[0024] The multidimensional load feature vector is matched with the feature templates in the preset load mode expert rule base to obtain the feature matching degree set of the load current sequence;

[0025] Load pattern recognition is performed on the feature matching degree set to obtain the load change pattern recognition result of the feature matching degree set;

[0026] Based on the load change pattern recognition results, trend prediction is performed to obtain the load current trend of the load current sequence;

[0027] A deterministic evaluation of the distribution characteristics of the feature matching degree set is performed to obtain a deterministic score of the load current trend.

[0028] In a preferred embodiment, matching the voltage regulation scheme corresponding to the pre-stored voltage configuration strategy based on the load change pattern and the load current trend includes:

[0029] Statistical distribution analysis is performed on the load change pattern to obtain the pattern characteristic parameters of the load change pattern;

[0030] The load current trend is decomposed into trend components to obtain the trend characteristic parameters of the load current trend;

[0031] The pattern feature parameters and the trend feature parameters are fused to obtain the comprehensive feature parameters of the load current sequence;

[0032] Calculate the matching score between the integrated characteristic parameters and the reference characteristic parameters in the pre-stored voltage configuration strategy, wherein the formula for calculating the matching score is:

[0033] ;

[0034] In the formula, For the matching score, These are the weight coefficients for pattern features. For similarity calculation function, The weighting coefficients for trend characteristics. The mode feature parameters, The reference mode parameters for the pre-stored voltage configuration strategy, These are trend characteristic parameters. The baseline trend parameters for the pre-stored voltage configuration strategy;

[0035] Based on the matching score, the voltage regulation scheme with the highest matching score is selected as the corresponding voltage regulation scheme in the voltage configuration strategy.

[0036] In a preferred embodiment, the step of generating the control signal for the multi-phase switching strategy based on the voltage regulation scheme and the deterministic score includes:

[0037] The voltage regulation scheme is analyzed to obtain the target output voltage and regulation rate of the voltage regulation scheme;

[0038] Based on the deterministic score, the number of phase activations for the multi-phase switching strategy is determined;

[0039] Based on the adjustment rate and the number of phase activations, a multi-phase clock signal for the multi-phase switching strategy is generated;

[0040] The target output voltage and the multi-phase clock signal are integrated into the primary control signal of the multi-phase switching strategy;

[0041] The primary control signal is modulated by duty cycle to obtain the control signal for the multi-phase switching strategy.

[0042] In a preferred embodiment, dynamically adjusting the duty cycle of the control signal based on the transient characteristics of the load current trend to obtain an optimized control signal for the multiphase power supply includes:

[0043] Transient characteristics are extracted from the load current trend to obtain the transient characteristic parameters of the load current trend;

[0044] Based on the transient characteristic parameters, the severity of load changes is evaluated to obtain a transient intensity index of the load current trend;

[0045] The duty cycle adjustment direction and amplitude are determined based on the transient intensity index, and the duty cycle correction strategy of the control signal is obtained.

[0046] The pulse width of the control signal is adjusted according to the duty cycle correction strategy to obtain the primary optimized signal of the multiphase power supply.

[0047] The primary optimization signal is subjected to ripple suppression processing to obtain the optimized control signal of the multiphase power supply.

[0048] In a preferred embodiment, executing the optimized control signal, monitoring the area error integral between the actual load current trajectory and the predicted trend envelope, and adaptively adjusting the feature matching degree threshold and overall duty cycle of the preset load mode expert rule base based on the area error integral includes:

[0049] Calculate the area error integral between the actual load current trajectory and the predicted trend envelope;

[0050] Based on the integral of the area error, a prediction accuracy evaluation index for dynamic voltage regulation is generated.

[0051] The feature matching degree threshold and the overall duty cycle are adjusted simultaneously based on the prediction accuracy evaluation index.

[0052] In a preferred embodiment, the formula for calculating the area error integral is:

[0053] ;

[0054] In the formula, Integral the area error. For integration operations, This refers to the actual load current trajectory. The predicted trend envelope, For time infinitesimal elements, It is the absolute value symbol.

[0055] To address the aforementioned problems, the present invention also provides a wireless sensor dynamic voltage regulation system based on load current prediction, the system comprising:

[0056] A sliding window processing module is used to perform dynamic adaptive sliding window processing on the load current sample values ​​to obtain the load current sequence of the load current sample values.

[0057] The feature extraction module is used to construct a multidimensional load feature vector of the load current sequence based on the statistical distribution characteristics and frequency domain wavelet energy characteristics of the differential sequence in the load current sequence.

[0058] The pattern recognition module is used to input the multi-dimensional load feature vector into a preset load pattern expert rule base, identify the current load change pattern and load current trend, and output a deterministic score of the load current trend.

[0059] The strategy matching module is used to match the voltage regulation scheme corresponding to the pre-stored voltage configuration strategy according to the load change pattern and the load current trend.

[0060] The control signal generation module is used to generate control signals for the multi-phase switching strategy based on the voltage regulation scheme and the deterministic score.

[0061] The duty cycle adjustment module is used to dynamically adjust the duty cycle of the control signal based on the transient characteristics of the load current trend, so as to obtain an optimized control signal for the multiphase power supply.

[0062] An adaptive adjustment module is used to execute the optimized control signal, monitor the area error integral between the actual load current trajectory and the predicted trend envelope, and adaptively adjust the feature matching degree threshold and the overall duty cycle of the preset load mode expert rule base according to the area error integral.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] 1. This invention provides high-quality data support and decision-making basis for dynamic voltage regulation of wireless sensors through refined load current processing and accurate pattern recognition. It performs dynamic adaptive sliding window processing on the sampled load current values, optimizing the window size based on the instantaneous dynamic characteristics of the current to generate a load current sequence that accurately reflects the current change pattern. Then, it constructs a multi-dimensional load feature vector using the statistical distribution characteristics of the differential sequence and the energy characteristics of frequency domain wavelets. This vector is input into a preset load pattern expert rule base to identify load change patterns and current trends, and outputs a deterministic score. This comprehensively captures the temporal patterns and potential characteristics of the load current, ensuring that subsequent voltage regulation schemes can accurately match load requirements, thus improving the foresight and adaptability of voltage regulation from the data source.

[0065] 2. This invention significantly improves the efficiency and stability of voltage regulation for wireless sensors by employing scientific voltage strategy matching, dynamic duty cycle adjustment, and closed-loop optimization mechanisms. Based on load variation patterns and current trends, an optimal voltage regulation scheme is matched using a weighted similarity formula, and a multi-phase switching control signal is generated by combining deterministic scoring. The duty cycle is dynamically adjusted based on the transient characteristics of the current trend, and ripple is suppressed to obtain an optimized control signal. During execution, the area error integral between the actual current trajectory and the predicted trend envelope is monitored, and the feature matching degree threshold and overall duty cycle are adaptively adjusted accordingly. This forms a complete process of "data processing – pattern recognition – strategy matching – adjustment optimization – closed-loop correction," significantly improving the accuracy and response speed of voltage regulation, effectively reducing the energy consumption of wireless sensors, and extending their service life. Attached Figure Description

[0066] Figure 1This is a flowchart illustrating a method for dynamic voltage regulation of a wireless sensor based on load current prediction, provided in an embodiment of the present invention.

[0067] Figure 2 A functional block diagram of a wireless sensor dynamic voltage regulation system based on load current prediction is provided in an embodiment of the present invention.

[0068] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0069] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0070] This application provides a method for dynamic voltage regulation of wireless sensors based on load current prediction. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for dynamic voltage regulation of wireless sensors based on load current prediction can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cluster of cloud servers. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0071] Reference Figure 1 The diagram shown is a flowchart illustrating a method for dynamic voltage regulation of a wireless sensor based on load current prediction, according to an embodiment of the present invention. In this embodiment, the method for dynamic voltage regulation of a wireless sensor based on load current prediction includes:

[0072] S1. Perform dynamic adaptive sliding window processing on the load current sample values ​​to obtain the load current sequence of the load current sample values;

[0073] In this embodiment of the invention, the step of performing dynamic adaptive sliding window processing on the load current sample values ​​to obtain the load current sequence of the load current sample values ​​includes:

[0074] The instantaneous dynamic characteristics of the load current sample values ​​are analyzed to obtain the dynamic change characteristics of the load current sample values;

[0075] The sampling window is optimized in real time based on the dynamic change characteristics to obtain the optimized sliding window size of the sampling window;

[0076] The load current sample values ​​are serialized based on the optimized sliding window size to obtain the load current sequence of the load current sample values.

[0077] Specifically, the instantaneous dynamic characteristics of the load current sampling values ​​are analyzed to obtain the dynamic change characteristics of the load current sampling values. Specifically, the load current sampling values ​​are continuously collected within a preset time period, and adjacent sampling values ​​are compared one by one to determine whether the change direction between adjacent sampling values ​​is increasing or decreasing. At the same time, the change amount between adjacent sampling values ​​is calculated, and the number of times the change amount is greater than a preset change threshold and the continuous change duration of the change direction are counted within the preset time period. The overall distribution of the change amount, the number of times the change amount is greater than the preset change threshold, and the continuous change duration of the change direction are integrated to form a dynamic change characteristic that can reflect the change law of the load current sampling values ​​within the instantaneous range.

[0078] Furthermore, the sampling window is optimized in real time based on the dynamic change characteristics to obtain the optimized sliding window size. Specifically, the sampling window size is adjusted according to the overall distribution of the change in the dynamic change characteristics, the number of times the change exceeds the preset change threshold, and the duration of continuous change in the direction of change. If the dynamic change characteristics show that the overall change in the load current sampling value is small, the number of times the change exceeds the preset change threshold is small, and the duration of continuous change in the direction of change is long, it indicates that the instantaneous dynamic characteristics of the load current sampling value are relatively stable. In this case, the initial size of the sampling window is increased to include more sampling values.

[0079] Furthermore, if the dynamic change characteristics show that the overall change in the load current sample value is large, the number of times the change exceeds the preset change threshold is large, and the duration of continuous change in the direction of change is short, it indicates that the instantaneous dynamic characteristics of the load current sample value are relatively volatile. In this case, the initial size of the sampling window is reduced to accurately capture instantaneous changes. After such adjustment, the optimized sliding window size of the sampling window is obtained.

[0080] Furthermore, the load current sample values ​​are serialized based on the optimized sliding window size to obtain the load current sequence of the load current sample values. Specifically, the load current sample values ​​are collected in chronological order, and the optimized sliding window size is used as a fixed length. Starting from the first load current sample value, a set of consecutive sample values ​​equal to the number of samples in the optimized sliding window size is selected as a set of sample values ​​within a window.

[0081] Furthermore, the window is slid forward one sample value position in chronological order, and the same number of consecutive sample values ​​are selected as the sample value set in the next window. This sliding selection process is repeated until all sample value sets in the windows are arranged in chronological order, forming a load current sequence composed of multiple sample value sets in chronological order.

[0082] In summary, by capturing the details of instantaneous current changes, optimizing the window to fit the current rhythm, reducing information loss and invalid sampling, high-quality data is provided for subsequent processing.

[0083] In summary, the generated load current sequence accurately reflects current changes, providing support for multimodal trend evolution and current prediction, and ensuring the rationality of voltage regulation schemes.

[0084] In summary, it eliminates the need for manual window presetting, automatically adapts to sensor current fluctuations, improves the adaptability of voltage regulation methods, and ensures stable sensor operation and energy efficiency.

[0085] S2. Based on the statistical distribution characteristics and frequency domain wavelet energy characteristics of the differential sequence in the load current sequence, construct a multidimensional load feature vector of the load current sequence;

[0086] In this embodiment of the invention, constructing a multidimensional load feature vector for the load current sequence based on the statistical distribution characteristics and frequency domain wavelet energy characteristics of the differential sequence in the load current sequence includes:

[0087] The load current sequence is subjected to trend quantization processing to obtain the load current difference sequence of the load current sequence;

[0088] Extract the statistical distribution characteristics of the load current differential sequence;

[0089] The load current sequence is subjected to a multi-scale wavelet transform to obtain the multi-scale wavelet coefficients of the load current sequence.

[0090] Energy distribution analysis is performed on the multi-scale wavelet coefficients to obtain the frequency domain wavelet energy characteristics of the load current sequence;

[0091] The statistical distribution features and the frequency domain wavelet energy features are fused to obtain the multidimensional load feature vector of the load current sequence.

[0092] Specifically, the load current sequence is subjected to trend quantification processing to obtain the load current difference sequence. Specifically, according to the time sequence of the sample value sets of each window in the load current sequence, the sample value sets in two adjacent windows are selected in turn. The sample value at the corresponding position in the latter window is subtracted from the sample value at the corresponding position in the former window to obtain a set of differences. Each pair of adjacent windows is calculated in this way to obtain a set of differences. The differences calculated by all the pairs of adjacent windows are arranged in the order of the windows to form a load current difference sequence that can quantify the trend of the load current sequence between different time windows.

[0093] Furthermore, the statistical distribution characteristics of the load current differential sequence are extracted. Specifically, all differences in the load current differential sequence are analyzed one by one, the maximum and minimum values ​​of all differences are counted, the average value of all differences is calculated, and the deviation of each difference from the average value is calculated by comparing all differences with the average value. Then, an index reflecting the dispersion of the differences is calculated through these deviations. At the same time, the number of differences greater than the average value and less than the average value, as well as the distribution of differences in different intervals, are counted. These statistically obtained information such as the maximum value, minimum value, average value, dispersion index, and interval distribution are integrated to form the statistical distribution characteristics of the load current differential sequence.

[0094] Furthermore, a multi-scale wavelet transform is performed on the load current sequence to obtain the multi-scale wavelet coefficients of the load current sequence. Specifically, a specified wavelet basis function is selected, and the sampled values ​​of each window in the load current sequence are taken as the processing object. First, the sampled values ​​in the first window are convolved with the wavelet basis function to obtain the low-frequency coefficients and high-frequency coefficients corresponding to that window. The low-frequency coefficients reflect the overall trend of the sampled values, and the high-frequency coefficients reflect the detailed changes of the sampled values. Then, the low-frequency coefficients are convolved again with the same wavelet basis function to obtain the next level of low-frequency coefficients and high-frequency coefficients. This process is repeated until the preset scale level is reached. After that, the same multi-scale convolution operation is performed on the sampled values ​​in all windows. The low-frequency coefficients and high-frequency coefficients obtained in all windows at each scale level are summarized to form the multi-scale wavelet coefficients of the load current sequence.

[0095] Furthermore, energy distribution analysis is performed on the multi-scale wavelet coefficients to obtain the frequency domain wavelet energy characteristics of the load current sequence. Specifically, for the high-frequency and low-frequency coefficients at each scale level in the multi-scale wavelet coefficients, the square value of each coefficient is calculated. The square values ​​of all coefficients at the same scale level are added together to obtain the energy value corresponding to that scale level. The corresponding energy value for each scale level is calculated in this way. At the same time, the proportion of the energy value of each scale level to the total energy value of all scale levels is calculated. The energy values ​​of each scale level and the proportion of each energy value to the total energy are arranged in the order of scale level to form the frequency domain wavelet energy characteristics that can reflect the energy distribution of the load current sequence at different frequency scales.

[0096] Furthermore, the statistical distribution features and frequency domain wavelet energy features are fused to obtain a multidimensional load feature vector of the load current sequence. Specifically, the information such as the maximum value, minimum value, average value, dispersion index and number of interval distributions contained in the statistical distribution features are combined with the information such as the energy values ​​and energy proportions of each scale level contained in the frequency domain wavelet energy features. These are arranged in a preset order to ensure that each feature information has a unique position. After the arrangement is completed, an ordered feature set containing all feature information is formed. This ordered feature set is the multidimensional load feature vector of the load current sequence.

[0097] In summary, by quantifying the trend of load current sequence changes to obtain a differential sequence, and combining it with the extracted statistical distribution characteristics, the current change pattern is clearly presented, providing an accurate basis for subsequent analysis.

[0098] In summary, the frequency domain wavelet energy features obtained by multi-scale wavelet transform and energy analysis can capture the characteristics of current at different frequency scales, supplement information not covered by statistical features, and enrich the feature dimensions.

[0099] In summary, the fused multidimensional load feature vector comprehensively reflects the load current characteristics, providing reliable feature support for subsequent matching voltage configuration strategies and control signal generation, and ensuring the accuracy of dynamic voltage regulation.

[0100] S3. Input the multidimensional load feature vector into the preset load mode expert rule base to identify the current load change mode and load current trend, and output the deterministic score of the load current trend.

[0101] In this embodiment of the invention, the step of inputting the multidimensional load feature vector into a preset load pattern expert rule base to identify the current load change pattern and load current trend, and simultaneously outputting a deterministic score for the load current trend, includes:

[0102] The multidimensional load feature vector is matched with the feature templates in the preset load mode expert rule base to obtain the feature matching degree set of the load current sequence;

[0103] Load pattern recognition is performed on the feature matching degree set to obtain the load change pattern recognition result of the feature matching degree set;

[0104] Based on the load change pattern recognition results, trend prediction is performed to obtain the load current trend of the load current sequence;

[0105] A deterministic evaluation of the distribution characteristics of the feature matching degree set is performed to obtain a deterministic score of the load current trend.

[0106] Specifically, the multidimensional load feature vector is matched with feature templates in the preset load mode expert rule base to obtain a feature matching degree set of the load current sequence. First, the contents of the preset load mode expert rule base are defined. The base stores feature templates for various typical load modes. Each template contains standard parameters and feature weights for the statistical distribution characteristics and frequency domain wavelet energy characteristics of the corresponding mode. Then, all feature parameters are extracted from the multidimensional load feature vector. According to the parameter order of the feature template, each parameter of the vector is compared with the corresponding standard parameter of the template to obtain the fit degree of a single parameter. Then, the fit degree is multiplied by the parameter weight to obtain a weighted fit value. After all parameters of a template are processed, all weighted fit values ​​are added to obtain the overall fit degree of the vector with the template. The above operation is repeated for all templates in the base. All overall fit degrees are sorted by template type to form a feature matching degree set.

[0107] Furthermore, load pattern recognition is performed on the feature matching degree set to obtain the load change pattern recognition result of the feature matching degree set. Specifically, a preset matching threshold is first retrieved from the library. This threshold is determined through a large number of sample tests. Then, the feature matching degree set is traversed to filter out the overall matching degree greater than the threshold. If there are multiple matching degrees, the template corresponding to the maximum value is selected, and its labeled pattern is used as the preliminary result. If there is only one matching degree, it is directly used as the preliminary result. If there is none, it is marked as unknown. Then, the verification rules corresponding to the preliminary result are used to verify whether the core features in the vector meet the pattern requirements. After the verification is passed, it is determined as the load change pattern recognition result.

[0108] Furthermore, based on the load change pattern recognition results, trend prediction is performed to obtain the load current trend of the load current sequence. Specifically, the trend rule corresponding to the recognition result is first retrieved from the library. The rule includes the historical current change patterns and characteristics under the pattern and their correspondence with the trend. Then, the change amplitude of the statistical distribution characteristics in the vector and the frequency distribution of the wavelet energy characteristics in the frequency domain are analyzed. Combined with the rule, the current operating stage is determined. If the change amplitude increases and the proportion of high-frequency energy increases, it is judged as an upward stage, and the subsequent trend is upward. If the change amplitude decreases and the proportion of low-frequency energy increases, it is judged as a downward stage, and the subsequent trend is downward. If the change amplitude is stable and the energy proportion is balanced, it is judged as a stable stage, and the subsequent trend is stable. The integrated stage and trend form the load current trend.

[0109] Furthermore, a deterministic assessment of the distribution characteristics of the feature matching degree set is performed to obtain a deterministic score for the load current trend. Specifically, the maximum and second largest values ​​in the feature matching degree set are first identified, and the difference between them is calculated. Then, the average value of the matching degrees other than the maximum value is calculated. If the maximum value is much larger than the average value and the difference between it and the second largest value is large, it indicates that the distribution is concentrated and the matching uniqueness is strong. If the difference between the maximum value and the average value is small or the difference between it and the second largest value is small, it indicates that the distribution is dispersed and the uniqueness is weak. Then, according to the scoring criteria in the library, a score is assigned based on the degree of distribution concentration. Concentration results in a higher score, and dispersion results in a lower score. This score is the deterministic score.

[0110] In summary, by matching multidimensional features with rule base templates and combining matching degree set analysis, load change patterns can be accurately identified, providing a clear basis for subsequent voltage regulation.

[0111] In summary, the system predicts current trends based on identified load patterns, outputs deterministic scores simultaneously, ensures the reliability of trend prediction results, and supports voltage strategy formulation.

[0112] In summary, it provides accurate load information and trend references for dynamic voltage regulation of wireless sensors, improves the adaptability and reliability of voltage regulation, and ensures the energy efficiency and stable operation of sensors.

[0113] S4. Based on the load change pattern and the load current trend, match the voltage regulation scheme corresponding to the pre-stored voltage configuration strategy;

[0114] In this embodiment of the invention, the step of matching the voltage regulation scheme corresponding to the pre-stored voltage configuration strategy based on the load change pattern and the load current trend includes:

[0115] Statistical distribution analysis is performed on the load change pattern to obtain the pattern characteristic parameters of the load change pattern;

[0116] The load current trend is decomposed into trend components to obtain the trend characteristic parameters of the load current trend;

[0117] The pattern feature parameters and the trend feature parameters are fused to obtain the comprehensive feature parameters of the load current sequence;

[0118] Calculate the matching score between the integrated characteristic parameters and the reference characteristic parameters in the pre-stored voltage configuration strategy, wherein the formula for calculating the matching score is:

[0119] ;

[0120] In the formula, For the matching score, These are the weight coefficients for pattern features. For similarity calculation function, The weighting coefficients for trend characteristics. The mode feature parameters, The reference mode parameters for the pre-stored voltage configuration strategy, These are trend characteristic parameters. The baseline trend parameters for the pre-stored voltage configuration strategy;

[0121] Based on the matching score, the voltage regulation scheme with the highest matching score is selected as the corresponding voltage regulation scheme in the voltage configuration strategy.

[0122] Specifically, the load change patterns are statistically analyzed to obtain the pattern characteristic parameters. First, the load current change data set corresponding to the load change pattern is identified. This set includes the load current sampling values ​​and change records within a certain period after the load change pattern identification result is determined. Then, this data set is sorted out, and the proportion of the load current in different change amplitude ranges is statistically analyzed. The number of consecutive times the load current change direction is maintained is recorded, and the frequency of the same change amplitude is calculated. At the same time, the peak occurrence time of the load current change and the interval between adjacent peaks are extracted. These statistically obtained information such as the proportion of change amplitude range, the number of consecutive change directions, the change amplitude frequency, the peak time, and the peak interval are integrated to form the pattern characteristic parameters that can accurately characterize the core characteristics of the load change pattern.

[0123] Furthermore, the load current trend is decomposed into trend components to obtain trend characteristic parameters. Specifically, the complete time series data corresponding to the load current trend is first obtained. This data records the changes in load current values ​​in chronological order. Then, the time series data is divided into multiple continuous subsequences at fixed time intervals. The average value of the load current in each subsequence is calculated. The difference between the average values ​​of adjacent subsequences is used to determine whether the change component type of each subsequence is rising, falling, or stable. The duration of each change component type is counted, and the average rise rate of the rising component and the average fall rate of the falling component are calculated. At the same time, the dominant change component type in the entire trend is determined. The obtained information such as change component type, duration of each type, rise and fall rates, and dominant component type are organized to form trend characteristic parameters that reflect the key characteristics of the load current trend.

[0124] Furthermore, the mode characteristic parameters and trend characteristic parameters are fused to obtain the comprehensive characteristic parameters of the load current sequence. Specifically, the characteristic items contained in the mode characteristic parameters and trend characteristic parameters are sorted out first, and the characteristic items that repeatedly represent the same characteristic are removed. All characteristic items with unique characterization significance in the two types of parameters are retained. Then, according to the preset characteristic sorting rules, the characteristic items such as the proportion of change range and the number of consecutive changes in the change direction in the mode characteristic parameters are arranged in sequence with the characteristic items such as the change component type and the duration of each type in the trend characteristic parameters, ensuring that each characteristic item has a unique position after arrangement. After the arrangement is completed, all characteristic items are uniformly standardized to make the characterization form of different characteristic items consistent. The set of characteristic items after sorting and standardization is the comprehensive characteristic parameters of the load current sequence.

[0125] Further, the matching score between the comprehensive feature parameters and the reference feature parameters in the pre-stored voltage configuration strategy is calculated. Specifically, it is first determined that the pre-stored voltage configuration strategy contains multiple voltage regulation schemes, each of which corresponds to a set of reference feature parameters. The feature items of the reference feature parameters completely correspond to the feature items of the comprehensive feature parameters. Then, feature items related to the mode features are extracted from the comprehensive feature parameters as mode feature parameters, and corresponding reference mode parameters are extracted from the reference feature parameters. By comparing the degree of fit between the corresponding feature items in the mode feature parameters and the reference mode parameters, the mode feature similarity is obtained. Then, feature items related to the trend features are extracted from the comprehensive feature parameters as trend feature parameters, and corresponding reference trend parameters are extracted from the reference feature parameters. By comparing the degree of fit between the trend feature parameters and the corresponding feature items in the reference trend parameters, the trend feature similarity is obtained. Then, according to the preset mode feature weight and trend feature weight, the mode feature similarity is multiplied by the mode feature weight, and the trend feature similarity is multiplied by the trend feature weight. The two products are added together to obtain the matching score between the comprehensive feature parameter and the reference feature parameter. The above calculation operation is performed on the reference feature parameters of all voltage regulation schemes to obtain multiple matching scores.

[0126] Furthermore, based on the matching score, the voltage regulation scheme with the highest matching score is selected as the corresponding voltage regulation scheme in the voltage configuration strategy. Specifically, the matching scores of all voltage regulation schemes are first collected, and these scores are compared one by one to find the matching score with the largest value. The voltage regulation scheme corresponding to the largest score is then determined. Next, the historical application records of the voltage regulation scheme are retrieved to check the application effect of the scheme under the same or similar load scenarios in the past. It is confirmed that the application effect meets the voltage regulation requirements of the current scenario. If it meets the requirements, the voltage regulation scheme is directly determined as the current corresponding voltage regulation scheme. If the historical application records show that there are details that need to be adjusted, the details are fine-tuned while keeping the core content of the scheme unchanged, and then the scheme is determined as the current corresponding voltage regulation scheme.

[0127] Specifically, the source of the mode feature parameters is obtained by statistically analyzing the load change patterns. Specifically, the load current change data set corresponding to the load change pattern is first identified. This set contains the load current sampling values ​​and change records within a certain period after the load change pattern recognition result is determined. After sorting out this data set, the percentage of time the load current is in different change amplitude ranges, the number of consecutive times the change direction is maintained, and the frequency of the same change amplitude appearing are statistically analyzed. The peak occurrence time of the load current change and the interval between adjacent peaks are extracted. These statistical information are integrated to form the mode feature parameters.

[0128] Furthermore, the source of the reference mode parameters is the pre-stored voltage configuration strategy. Specifically, the pre-stored voltage configuration strategy contains multiple voltage regulation schemes, each of which corresponds to a set of reference characteristic parameters. The reference mode parameters are the reference information corresponding to the mode characteristic parameters extracted from each set of reference characteristic parameters. These information are standard characteristic representations summarized from a large amount of test data of typical load modes.

[0129] Furthermore, the trend characteristic parameters are obtained by decomposing the load current trend into trend components. Specifically, the complete time series data corresponding to the load current trend is first obtained, and then divided into multiple continuous subsequences at fixed time intervals. The average value of the load current in each subsequence is calculated, and the type of change component is determined by the difference between the average values ​​of adjacent subsequences. The duration of each type is counted, the average rate of rise and fall components is calculated, the type of dominant change component is determined, and this information is organized to form the trend characteristic parameters.

[0130] Furthermore, the source of the benchmark trend parameters is the pre-stored voltage configuration strategy. Specifically, the benchmark information corresponding to the trend feature parameters is extracted from the benchmark feature parameters corresponding to each voltage regulation scheme in the pre-stored voltage configuration strategy. This information is a standard trend feature representation form summarized from historical data of typical load current trends.

[0131] Furthermore, the weighting coefficients of the pattern features and the trend features are derived from the pre-stored voltage configuration strategy. Specifically, the weighting coefficients are determined by calibrating a large amount of experimental data based on the degree of influence of the pattern features and trend features on the matching of voltage regulation schemes under different load scenarios, and are stored in the pre-stored voltage configuration strategy to distinguish the importance of the two features in the matching process.

[0132] Furthermore, by calculating the similarity between the mode feature parameters and the baseline mode parameters, and the similarity between the trend feature parameters and the baseline trend parameters, and weighting them with their respective weight coefficients, the two weighted results are added together to obtain a matching score. This score quantifies the degree of matching between the comprehensive feature parameters and the baseline feature parameters in the pre-stored voltage configuration strategy. The higher the score, the more suitable the comprehensive feature parameters are for the voltage regulation scheme corresponding to the baseline feature parameters, providing a quantitative basis for selecting the optimal voltage regulation scheme in the future.

[0133] Furthermore, the method of using the similarity calculation function is to first clarify the feature items contained in the pattern feature parameters and the baseline pattern parameters, as well as the feature items contained in the trend feature parameters and the baseline trend parameters, to ensure that the feature items of the two sets of comparison parameters are completely corresponding. Then, the degree of fit of the corresponding feature items is compared one by one, and the degree of fit is converted into a value between 0 and 1. This value is the similarity. The closer the value is to 1, the higher the degree of fit between the two sets of parameters. The closer the value is to 0, the lower the degree of fit.

[0134] Furthermore, the formula shows that the higher the similarity between the pattern feature parameters and the baseline pattern parameters, the greater the weighted result of this part and the higher the matching score, given that the weight coefficient of the pattern features is fixed; conversely, the lower the similarity, the smaller the corresponding weighted result and the lower the matching score.

[0135] Furthermore, the formula trend is that the higher the similarity between the trend feature parameter and the benchmark trend parameter, the greater the weighted result of this part when the weight coefficient of the trend feature is fixed, and the matching score will also increase accordingly; conversely, the lower the similarity, the smaller the corresponding weighted result, and the matching score will also decrease accordingly.

[0136] Furthermore, the formula trend is that, under the condition that the two similarities are fixed, the larger the weight coefficient of the pattern feature, the higher the proportion of the weighted result corresponding to the pattern feature similarity in the matching score, and the greater the influence of the similarity on the matching score; the larger the weight coefficient of the trend feature, the higher the proportion of the weighted result corresponding to the trend feature similarity in the matching score, and the greater the influence of the similarity on the matching score.

[0137] In summary, analyzing load change patterns, decomposing trend components to obtain characteristic parameters, and then integrating them into comprehensive characteristic parameters fully reflects the key characteristics of the load, laying an accurate foundation for matching voltage schemes.

[0138] In summary, the score is calculated using a weighted similarity formula, taking into account both pattern and trend feature weights, to ensure that the matching results are objective and reasonable, and to avoid adaptation bias caused by a single feature.

[0139] In summary, selecting the voltage regulation scheme based on the highest matching score quickly finds a solution that suits the current load condition, ensuring the accuracy of dynamic voltage regulation of wireless sensors and improving sensor energy efficiency and operational stability.

[0140] S5. Based on the voltage regulation scheme and the deterministic score, generate the control signal for the multi-phase switching strategy;

[0141] In this embodiment of the invention, the step of generating the control signal for the multi-phase switching strategy based on the voltage regulation scheme and the deterministic score includes:

[0142] The voltage regulation scheme is analyzed to obtain the target output voltage and regulation rate of the voltage regulation scheme;

[0143] Based on the deterministic score, the number of phase activations for the multi-phase switching strategy is determined;

[0144] Based on the adjustment rate and the number of phase activations, a multi-phase clock signal for the multi-phase switching strategy is generated;

[0145] The target output voltage and the multi-phase clock signal are integrated into the primary control signal of the multi-phase switching strategy;

[0146] The primary control signal is modulated by duty cycle to obtain the control signal for the multi-phase switching strategy.

[0147] Specifically, the voltage regulation scheme is analyzed to obtain the target output voltage and regulation rate. First, a complete technical document of the voltage regulation scheme is obtained, containing the specific execution parameters and operating specifications. Then, the final target parameter for voltage regulation is extracted from the document; this parameter is the target output voltage. Next, the rate requirement for adjusting the voltage from the current value to the target output voltage is extracted. This rate requirement specifies the fixed amplitude of voltage change per unit time. By interpreting and transforming the rate requirement, a regulation rate that can directly guide execution is obtained. The extracted target output voltage and the transformed regulation rate are then organized and recorded to form the analysis results.

[0148] Furthermore, based on the deterministic score, the number of phase activations for the multi-phase switching strategy is determined. Specifically, a pre-stored rule corresponding to the score and the number of phases is first retrieved. This rule is established based on a large amount of experimental data and clarifies the optimal number of phase activations corresponding to different deterministic scores. Then, the obtained deterministic score is compared one by one with the score intervals in the rule to find the score interval that perfectly matches the current deterministic score. The number of phase activations corresponding to this score interval is then extracted. If the deterministic score is high, it indicates that the reliability of the load current trend is high, and a larger number of phases are activated to improve the regulation accuracy and response speed. If the deterministic score is low, it indicates that the reliability of the load current trend is low, and a smaller number of phases are activated to ensure regulation stability. The determined number of phases is the number of phase activations for the multi-phase switching strategy.

[0149] Furthermore, based on the adjustment rate and the number of phase activations, a multi-phase clock signal for a multi-phase switching strategy is generated. Specifically, the total number of phase channels corresponding to the number of phase activations is first determined, with each phase channel corresponding to an independent clock signal. Then, the period of the clock signal is determined according to the adjustment rate. The faster the adjustment rate, the shorter the period of the clock signal to achieve rapid switching; the slower the adjustment rate, the longer the period of the clock signal to ensure smooth adjustment. Then, according to the timing requirements of phase switching, a different clock signal start time is assigned to each phase channel to ensure that the clock signals of each channel are staggered in time and do not overlap. By setting the period and start time of each channel, multiple independent and time-ordered phase clock signals are generated, i.e., multi-phase clock signals.

[0150] Furthermore, the target output voltage and multiple phase clock signals are integrated into the primary control signal of the multi-phase switching strategy. Specifically, the framework structure of the primary control signal is first constructed, which includes two core parts: a voltage command module and a clock drive module. Then, the target output voltage is written into the voltage command module as a voltage command, clarifying the voltage output standard that each phase channel needs to achieve during operation. Next, the multiple phase clock signals are connected to the clock drive module in the corresponding channel order to provide timing control basis for each phase channel. The voltage command module and the clock drive module are connected through a signal bus so that the voltage command and clock signal can be transmitted and executed synchronously. The integrated signal combination is the primary control signal.

[0151] Furthermore, the primary control signal is subjected to duty cycle modulation to obtain the control signal for the multi-phase switching strategy. Specifically, the voltage command and clock signal in the primary control signal are first acquired, and the ideal duty cycle range is determined according to the target output voltage requirements. The duty cycle is the ratio of the high-level duration within one cycle of the clock signal to the total duration of the cycle. Then, the clock signal in the primary control signal is adjusted through the duty cycle modulation circuit to adjust the high-level duration to match the ideal duty cycle. This ensures that when the adjusted clock signal drives the phase channel, it can output a voltage that meets the target requirements, while maintaining the timing relationship of the multi-phase clock signals. The combination of the modulated voltage command and the modulated clock signal is the control signal for the multi-phase switching strategy.

[0152] In summary, the analytical voltage regulation scheme obtains the target output voltage and regulation rate, providing precise direction for control signal generation and avoiding regulation deviations. The number of phase activations is determined based on deterministic scoring; more phases improve accuracy at high confidence levels, while fewer phases maintain stability at low confidence levels, ensuring regulation adaptability.

[0153] In summary, by combining the adjustment rate and the number of phases to generate a clock signal, integrating the voltage and clock signals and modulating the duty cycle, the output control signal can accurately drive multi-phase switching, support the efficient execution of dynamic voltage regulation of wireless sensors, and ensure stable operation and energy efficiency of the sensors.

[0154] S6. Based on the transient characteristics of the load current trend, dynamically adjust the duty cycle of the control signal to obtain an optimized control signal for the multiphase power supply.

[0155] In this embodiment of the invention, the step of dynamically adjusting the duty cycle of the control signal based on the transient characteristics of the load current trend to obtain an optimized control signal for the multiphase power supply includes:

[0156] Transient characteristics are extracted from the load current trend to obtain the transient characteristic parameters of the load current trend;

[0157] Based on the transient characteristic parameters, the severity of load changes is evaluated to obtain a transient intensity index of the load current trend;

[0158] The duty cycle adjustment direction and amplitude are determined based on the transient intensity index, and the duty cycle correction strategy of the control signal is obtained.

[0159] The pulse width of the control signal is adjusted according to the duty cycle correction strategy to obtain the primary optimized signal of the multiphase power supply.

[0160] The primary optimization signal is subjected to ripple suppression processing to obtain the optimized control signal of the multiphase power supply.

[0161] Specifically, transient characteristics of the load current trend are extracted to obtain transient feature parameters of the load current trend. First, the full-time time series data corresponding to the load current trend is obtained. This data records the real-time value of the load current at millisecond intervals. Then, the data is analyzed segment by segment to find the intervals where the current value changes rapidly. The start and end times of each interval are marked as abrupt change points. The difference between the current values ​​before and after the abrupt change point is calculated to obtain the abrupt change amplitude. The number of abrupt change points per minute is counted as the abrupt change frequency. At the same time, the duration for which the current value remains stable after each abrupt change and the difference between the maximum and minimum current values ​​during the stable phase are recorded as the fluctuation range. The information of abrupt change points, abrupt change amplitude, abrupt change frequency, stable duration, and fluctuation range are systematically organized to form the transient feature parameters of the load current trend.

[0162] Furthermore, based on transient characteristic parameters, the severity of load changes is assessed to obtain a transient intensity index of the load current trend. Specifically, a pre-stored transient assessment rule set is first retrieved. This rule set was established through a large amount of experimental data from different load scenarios, clarifying the correspondence between parameters such as mutation amplitude and mutation frequency and the severity of load changes. Then, the extracted transient characteristic parameters are compared item by item with the standard parameters in the rule set. If the mutation amplitude is large, the mutation frequency is high, the stability time is short, and the fluctuation range is large, the load change is judged to be severe. If the mutation amplitude is small, the mutation frequency is low, the stability time is long, and the fluctuation range is small, the load change is judged to be gentle. The judgment result is assigned to a preset severity level, and the level is then converted into a fixed characterization value. This value is the transient intensity index of the load current trend.

[0163] Furthermore, the duty cycle adjustment direction and amplitude are determined based on the transient intensity index to obtain the duty cycle correction strategy for the control signal. Specifically, a pre-set intensity index and adjustment strategy comparison table is first obtained. This table clarifies the duty cycle adjustment direction and specific adjustment amplitude corresponding to different transient intensity indices. Then, the corresponding entry for the current transient intensity index in the comparison table is found, and the adjustment direction and amplitude within the entry are extracted. If the transient intensity index is high, it indicates that the load changes drastically, and the adjustment direction is to increase the duty cycle with a larger amplitude to quickly improve the power output capacity to match the load demand. If the transient intensity index is low, it indicates that the load changes gradually, and the adjustment direction is to decrease the duty cycle with a smaller amplitude to avoid excessive output leading to increased energy consumption. The adjustment direction, amplitude, and execution sequence are organized into directly executable operation specifications to form the duty cycle correction strategy.

[0164] Furthermore, the pulse width of the control signal is adjusted according to the duty cycle correction strategy to obtain the primary optimized signal of the multiphase power supply. Specifically, the original pulse width data of the control signal is first read through a signal analysis device to determine the fixed duration of the high level and the signal period in the original signal. Then, the pulse width adjustment amount is calculated according to the adjustment direction and amplitude in the duty cycle correction strategy. If the duty cycle needs to be increased, the high level duration is extended by the adjustment amount; if the duty cycle needs to be decreased, the high level duration is shortened by the adjustment amount. During the adjustment process, the signal period duration is kept constant. At the same time, the timing interval of the multiple phase control signals is kept consistent through a phase synchronization tool to avoid phase overlap or gaps. After adjustment, the pulse width is confirmed to meet the requirements by signal detection to obtain the primary optimized signal of the multiphase power supply.

[0165] Furthermore, the primary optimization signal undergoes ripple suppression processing to obtain the optimized control signal for the multiphase power supply. Specifically, the primary optimization signal is first connected to a ripple suppression system, which includes a filtering unit and a signal calibration unit. The filtering unit adopts a passive low-pass filter structure, utilizing the impedance characteristics of capacitors and inductors to filter high-frequency ripple components in the signal while retaining the core control waveform. Then, the signal calibration unit performs amplitude calibration on the filtered signal, adjusting the amplitudes of the high and low levels to preset standard values, while correcting the distortion at the waveform edges to ensure that the rising and falling edges remain steep and without trailing. After processing, the signal ripple content is detected by a ripple detector. Once it is confirmed to meet the preset standard, the signal becomes the optimized control signal for the multiphase power supply.

[0166] In summary, extracting the transient characteristics of load current trends and assessing the degree of change provides a precise basis for duty cycle adjustment, avoiding blind adjustment.

[0167] In summary, the direction and magnitude of duty cycle adjustment are determined based on transient intensity to match the rhythm of load changes, ensuring the responsiveness and stability of voltage regulation.

[0168] In summary, by adjusting the pulse width and suppressing ripple, the generated optimized control signal can accurately drive the multiphase power supply, support the efficient execution of dynamic voltage regulation of wireless sensors, and improve the energy efficiency and operational stability of the sensors.

[0169] S7. Execute the optimized control signal, monitor the area error integral between the actual load current trajectory and the predicted trend envelope, and adaptively adjust the feature matching degree threshold and the overall duty cycle of the preset load mode expert rule base according to the area error integral.

[0170] In this embodiment of the invention, executing the optimized control signal, monitoring the area error integral between the actual load current trajectory and the predicted trend envelope, and adaptively adjusting the feature matching degree threshold and overall duty cycle of the preset load mode expert rule base based on the area error integral includes:

[0171] Calculate the area error integral between the actual load current trajectory and the predicted trend envelope;

[0172] Based on the integral of the area error, a prediction accuracy evaluation index for dynamic voltage regulation is generated.

[0173] The feature matching degree threshold and the overall duty cycle are adjusted simultaneously based on the prediction accuracy evaluation index.

[0174] The formula for calculating the integral of the area error is:

[0175] ;

[0176] In the formula, Integral the area error. For integration operations, This refers to the actual load current trajectory. The predicted trend envelope, For time infinitesimal elements, It is the absolute value symbol.

[0177] Specifically, the area error integral between the actual load current trajectory and the predicted trend envelope is calculated. This involves first acquiring load current data in real time during the execution of the optimized control signal using a high-precision current sensor, then arranging the data sequentially according to the acquisition time to form a continuous actual load current trajectory. Simultaneously, the predicted trend envelope previously generated based on the load current trend is retrieved from the system storage module. This envelope consists of the upper and lower bounds of the predicted load current, which together define the predicted range of current variation.

[0178] Furthermore, using time as the horizontal axis and current value as the vertical axis, the actual load current trajectory and the predicted trend envelope are plotted on the same coordinate plane. The complete overlap of the two curves in the time dimension is determined, and this overlap is divided into multiple consecutive small time segments according to a fixed time step. For each small time segment, the area of ​​the closed region enclosed by the actual load current trajectory and the predicted trend envelope is calculated. If the actual current trajectory is inside the predicted envelope, the area is recorded as a positive value; if the actual current trajectory exceeds the predicted envelope, the area is recorded as a negative value. The area values ​​corresponding to all small time segments are summed, and the summation result is the area error integral between the actual load current trajectory and the predicted trend envelope.

[0179] Furthermore, based on the area error integral, a prediction accuracy evaluation index for dynamic voltage regulation is generated. Specifically, a pre-stored rule table mapping area error integrals to prediction accuracy levels is retrieved from the system configuration module. This rule table was determined after calibration using a large amount of experimental data under different load scenarios, clearly defining the range of different area error integral values ​​and their corresponding prediction accuracy levels. Next, the calculated area error integral value is compared one by one with the ranges in the rule table to find the interval that perfectly matches the current integral value, and the prediction accuracy level corresponding to that interval is extracted.

[0180] Furthermore, if the absolute value of the area error integral is small, it indicates a high degree of fit between the actual load current trajectory and the predicted trend envelope, corresponding to a high prediction accuracy level; if the absolute value of the area error integral is large, it indicates a significant deviation between the actual trajectory and the predicted envelope, corresponding to a low prediction accuracy level. The determined prediction accuracy level is then converted into a fixed quantized value, which serves as the prediction accuracy evaluation index for dynamic voltage regulation.

[0181] Furthermore, based on the prediction accuracy evaluation index, the feature matching degree threshold and the overall duty cycle are adjusted simultaneously. Specifically, a pre-set prediction accuracy evaluation index and adjustment parameter comparison table is retrieved from the system strategy library. This comparison table clarifies the adjustment amount of the feature matching degree threshold and the overall duty cycle corresponding to different evaluation index values.

[0182] Furthermore, based on the current prediction accuracy evaluation index, the corresponding adjustment parameter entries are found in the lookup table, and the feature matching degree threshold adjustment and overall duty cycle adjustment are extracted from the entries. If the evaluation index shows high prediction accuracy, it indicates that the feature matching effect of the preset load mode expert rule base is good. In this case, the feature matching degree threshold is appropriately increased by the adjustment amount to enhance the filtering accuracy of the rule base for load mode recognition. At the same time, according to the current stable state of the load, the overall duty cycle is adjusted to a value that can maintain a stable voltage output. If the evaluation index shows low prediction accuracy, it indicates that the matching effect of the rule base is poor and there is a pattern recognition bias. In this case, the feature matching degree threshold is appropriately decreased by the adjustment amount to expand the feature matching range and improve the adaptability to complex load modes. At the same time, according to the actual load current change requirements, the overall duty cycle is increased or decreased accordingly to improve the matching degree between voltage regulation and load requirements.

[0183] Furthermore, after the adjustment is completed, the updated feature matching degree threshold is written into the preset load mode expert rule base and saved, and the new overall duty cycle parameter is synchronized to the control signal generation unit to ensure that subsequent control signal generation can be executed based on the adjusted parameters.

[0184] Specifically, the actual load current trajectory is obtained by using a high-precision current sensor in the current monitoring module to collect the load current in real time during the execution of the optimized control signal. The collection process is carried out continuously at fixed time intervals, and the current value corresponding to each collection time point is recorded in chronological order to form a continuous current change data sequence, which is the actual load current trajectory.

[0185] Furthermore, the predicted trend envelope is obtained by retrieving the previously generated predicted trend envelope from the system's storage unit. This envelope is determined based on the load current trend, which is obtained by trend prediction through load change pattern recognition results. The predicted trend envelope consists of the upper and lower limits of the predicted load current, which together define the predicted range of load current changes.

[0186] Furthermore, the integration time range is determined by comparing the time span of the actual load current trajectory and the time span of the predicted trend envelope, and selecting the time period in which the two completely overlap as the integration time range. This ensures that within this time range, both the actual load current trajectory and the predicted trend envelope have corresponding values, providing a complete and matching time interval for subsequent calculations.

[0187] Furthermore, the integration time range is segmented, specifically by dividing the determined integration time range into multiple continuous and minute time segments according to a fixed time step. The duration of each time segment remains consistent. By subdividing the time range, the changes in the actual load current and the predicted trend envelope within each minute time segment can be approximated as stable, laying the foundation for accurately calculating the error area of ​​each segment.

[0188] Furthermore, the error area of ​​each tiny time segment is calculated. Specifically, for each tiny time segment, the current value corresponding to the actual load current trajectory and the current value corresponding to the predicted trend envelope are extracted, the difference between the two is calculated, and the absolute value of the difference is taken. This absolute value is then multiplied by the duration of the tiny time segment, and the result is the error area between the actual load current trajectory and the predicted trend envelope within that tiny time segment.

[0189] Furthermore, the error areas of all the small time segments are accumulated. Specifically, the error areas calculated from all the small time segments are accumulated one by one, and the total value obtained after accumulation is used as the area error integral between the actual load current trajectory and the predicted trend envelope. This value can fully reflect the overall deviation between the actual load current and the predicted trend envelope over the entire integration time range.

[0190] In summary, by using the area error integral formula, the deviation between the actual load current trajectory and the predicted trend envelope is integrated over time. The absolute value is used to eliminate positive and negative cancellation, objectively quantifying the overall degree of deviation between the two, providing reliable data for accuracy assessment, and avoiding the limitations of comparing a single time point.

[0191] In summary, the prediction accuracy evaluation index generated based on the integration results can intuitively reflect the recognition accuracy, current trend prediction reliability, and voltage regulation adaptability of the preset load mode expert rule base, clearly define the regulation effect level, and provide a clear direction for subsequent parameter adjustments.

[0192] In summary, by synchronously adjusting the feature matching threshold and overall duty cycle according to the evaluation indicators, the load mode matching range and multi-phase control signal output can be optimized, adapting to the dynamic fluctuations of sensor load, ensuring long-term accurate dynamic voltage regulation, reducing energy consumption, and improving sensor endurance and operational stability.

[0193] like Figure 2The diagram shown is a functional block diagram of a wireless sensor dynamic voltage regulation system based on load current prediction, provided in an embodiment of the present invention.

[0194] The wireless sensor dynamic voltage regulation system 100 based on load current prediction described in this invention can be installed in an electronic device. Depending on the functions implemented, the wireless sensor dynamic voltage regulation system 100 based on load current prediction may include a sliding window processing module 101, a feature extraction module 102, a pattern recognition module 103, a strategy matching module 104, a control signal generation module 105, a duty cycle adjustment module 106, and an adaptive adjustment module 107. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0195] In this embodiment, the functions of each module / unit are as follows:

[0196] The sliding window processing module 101 is used to perform dynamic adaptive sliding window processing on the load current sample value to obtain the load current sequence of the load current sample value.

[0197] The feature extraction module 102 is used to construct a multidimensional load feature vector of the load current sequence based on the statistical distribution characteristics and frequency domain wavelet energy characteristics of the differential sequence in the load current sequence.

[0198] The pattern recognition module 103 is used to input the multi-dimensional load feature vector into a preset load pattern expert rule base, identify the current load change pattern and load current trend, and output a deterministic score of the load current trend.

[0199] The strategy matching module 104 is used to match the voltage regulation scheme corresponding to the pre-stored voltage configuration strategy according to the load change mode and the load current trend.

[0200] The control signal generation module 105 is used to generate control signals for a multi-phase switching strategy based on the voltage regulation scheme and the deterministic score.

[0201] The duty cycle adjustment module 106 is used to dynamically adjust the duty cycle of the control signal based on the transient characteristics of the load current trend, so as to obtain an optimized control signal for the multiphase power supply.

[0202] The adaptive adjustment module 107 is used to execute the optimized control signal, monitor the area error integral between the actual load current trajectory and the predicted trend envelope, and adaptively adjust the feature matching degree threshold and the overall duty cycle of the preset load mode expert rule base according to the area error integral.

[0203] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0204] The modules described as separate components may or may not be physically separate. The components shown as modules 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 embodiment according to actual needs.

[0205] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0206] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0207] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0208] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for dynamic voltage regulation of wireless sensors based on load current prediction, characterized in that, The method includes: S1. Perform dynamic adaptive sliding window processing on the load current sample values ​​to obtain the load current sequence of the load current sample values; S2. Based on the statistical distribution characteristics and frequency domain wavelet energy characteristics of the differential sequence in the load current sequence, construct a multidimensional load feature vector of the load current sequence; S3. Input the multidimensional load feature vector into the preset load mode expert rule base to identify the current load change mode and load current trend, and output the deterministic score of the load current trend. S4. Based on the load change pattern and the load current trend, match the voltage regulation scheme corresponding to the pre-stored voltage configuration strategy; S5. Based on the voltage regulation scheme and the deterministic score, generate the control signal for the multi-phase switching strategy; S6. Based on the transient characteristics of the load current trend, dynamically adjust the duty cycle of the control signal to obtain an optimized control signal for the multiphase power supply. S7. Execute the optimized control signal, monitor the area error integral between the actual load current trajectory and the predicted trend envelope, and adaptively adjust the feature matching degree threshold and the overall duty cycle of the preset load mode expert rule base according to the area error integral.

2. The method for dynamic voltage regulation of wireless sensors based on load current prediction as described in claim 1, characterized in that, The process of performing dynamic adaptive sliding window processing on the load current sample values ​​to obtain the load current sequence of the load current sample values ​​includes: The instantaneous dynamic characteristics of the load current sample values ​​are analyzed to obtain the dynamic change characteristics of the load current sample values; The sampling window is optimized in real time based on the dynamic change characteristics to obtain the optimized sliding window size of the sampling window; The load current sample values ​​are serialized based on the optimized sliding window size to obtain the load current sequence of the load current sample values.

3. The method for dynamic voltage regulation of wireless sensors based on load current prediction as described in claim 1, characterized in that, The step of constructing a multidimensional load feature vector for the load current sequence based on the statistical distribution characteristics and frequency domain wavelet energy characteristics of the differential sequence in the load current sequence includes: The load current sequence is subjected to trend quantization processing to obtain the load current difference sequence of the load current sequence; Extract the statistical distribution characteristics of the load current differential sequence; The load current sequence is subjected to a multi-scale wavelet transform to obtain the multi-scale wavelet coefficients of the load current sequence. Energy distribution analysis is performed on the multi-scale wavelet coefficients to obtain the frequency domain wavelet energy characteristics of the load current sequence; The statistical distribution features and the frequency domain wavelet energy features are fused to obtain the multidimensional load feature vector of the load current sequence.

4. The method for dynamic voltage regulation of wireless sensors based on load current prediction as described in claim 1, characterized in that, The step of inputting the multidimensional load feature vector into a preset load pattern expert rule base to identify the current load change pattern and load current trend, and simultaneously outputting a deterministic score for the load current trend, includes: The multidimensional load feature vector is matched with the feature templates in the preset load mode expert rule base to obtain the feature matching degree set of the load current sequence; Load pattern recognition is performed on the feature matching degree set to obtain the load change pattern recognition result of the feature matching degree set; Based on the load change pattern recognition results, trend prediction is performed to obtain the load current trend of the load current sequence; A deterministic evaluation of the distribution characteristics of the feature matching degree set is performed to obtain a deterministic score of the load current trend.

5. The method for dynamic voltage regulation of wireless sensors based on load current prediction as described in claim 1, characterized in that, The step of matching the voltage regulation scheme corresponding to the pre-stored voltage configuration strategy based on the load change pattern and the load current trend includes: Statistical distribution analysis is performed on the load change pattern to obtain the pattern characteristic parameters of the load change pattern; The load current trend is decomposed into trend components to obtain the trend characteristic parameters of the load current trend; The pattern feature parameters and the trend feature parameters are fused to obtain the comprehensive feature parameters of the load current sequence; Calculate the matching score between the integrated characteristic parameters and the reference characteristic parameters in the pre-stored voltage configuration strategy, wherein the formula for calculating the matching score is: ； In the formula, For the matching score, These are the weight coefficients for pattern features. For similarity calculation function, The weighting coefficients for trend characteristics. The mode feature parameters, The reference mode parameters for the pre-stored voltage configuration strategy, These are trend characteristic parameters. The baseline trend parameters for the pre-stored voltage configuration strategy; Based on the matching score, the voltage regulation scheme with the highest matching score is selected as the corresponding voltage regulation scheme in the voltage configuration strategy.

6. The method for dynamic voltage regulation of wireless sensors based on load current prediction as described in claim 1, characterized in that, The step of generating control signals for a multi-phase switching strategy based on the voltage regulation scheme and the deterministic score includes: The voltage regulation scheme is analyzed to obtain the target output voltage and regulation rate of the voltage regulation scheme; Based on the deterministic score, the number of phase activations for the multi-phase switching strategy is determined; Based on the adjustment rate and the number of phase activations, a multi-phase clock signal for the multi-phase switching strategy is generated; The target output voltage and the multi-phase clock signal are integrated into the primary control signal of the multi-phase switching strategy; The primary control signal is modulated by duty cycle to obtain the control signal for the multi-phase switching strategy.

7. The method for dynamic voltage regulation of wireless sensors based on load current prediction as described in claim 1, characterized in that, The method of dynamically adjusting the duty cycle of the control signal based on the transient characteristics of the load current trend to obtain an optimized control signal for the multiphase power supply includes: Transient characteristics are extracted from the load current trend to obtain the transient characteristic parameters of the load current trend; Based on the transient characteristic parameters, the severity of load changes is evaluated to obtain a transient intensity index of the load current trend; The duty cycle adjustment direction and amplitude are determined based on the transient intensity index, and the duty cycle correction strategy of the control signal is obtained. The pulse width of the control signal is adjusted according to the duty cycle correction strategy to obtain the primary optimized signal of the multiphase power supply. The primary optimization signal is subjected to ripple suppression processing to obtain the optimized control signal of the multiphase power supply.

8. The method for dynamic voltage regulation of wireless sensors based on load current prediction as described in claim 1, characterized in that, The process of executing the optimized control signal, monitoring the area error integral between the actual load current trajectory and the predicted trend envelope, and adaptively adjusting the feature matching degree threshold and overall duty cycle of the preset load mode expert rule base based on the area error integral includes: Calculate the area error integral between the actual load current trajectory and the predicted trend envelope; Based on the integral of the area error, a prediction accuracy evaluation index for dynamic voltage regulation is generated. The feature matching degree threshold and the overall duty cycle are adjusted simultaneously based on the prediction accuracy evaluation index.

9. The method for dynamic voltage regulation of a wireless sensor based on load current prediction as described in claim 8, characterized in that, The formula for calculating the integral of the area error is: ； In the formula, Integral the area error. For integration operations, This refers to the actual load current trajectory. The predicted trend envelope, For time infinitesimal elements, It is the absolute value symbol.

10. A wireless sensor dynamic voltage regulation system based on load current prediction, characterized in that, The system includes: A sliding window processing module is used to perform dynamic adaptive sliding window processing on the load current sample values ​​to obtain the load current sequence of the load current sample values. The feature extraction module is used to construct a multidimensional load feature vector of the load current sequence based on the statistical distribution characteristics and frequency domain wavelet energy characteristics of the differential sequence in the load current sequence. The pattern recognition module is used to input the multi-dimensional load feature vector into a preset load pattern expert rule base, identify the current load change pattern and load current trend, and output a deterministic score of the load current trend. The strategy matching module is used to match the voltage regulation scheme corresponding to the pre-stored voltage configuration strategy according to the load change pattern and the load current trend. The control signal generation module is used to generate control signals for the multi-phase switching strategy based on the voltage regulation scheme and the deterministic score. The duty cycle adjustment module is used to dynamically adjust the duty cycle of the control signal based on the transient characteristics of the load current trend, so as to obtain an optimized control signal for the multiphase power supply. An adaptive adjustment module is used to execute the optimized control signal, monitor the area error integral between the actual load current trajectory and the predicted trend envelope, and adaptively adjust the feature matching degree threshold and the overall duty cycle of the preset load mode expert rule base according to the area error integral.

Images