A hierarchical intelligent control system for power time synchronization

Abstract

The present invention belongs to the field of intelligent control technology. The present invention discloses a hierarchical intelligent control system for power time synchronization, including a clock source management layer, which is used to manage N time sources and provide reference time signal data; the N time sources are divided into a main time source and a backup time source, and a time source monitoring mechanism is introduced to automatically switch to the backup time source when a main time source fails; a clock layer, which is used to receive the reference time signal data through the clock and synchronize it to each power device; a multi-time source acquisition layer, which is used to collect reference time signal data, equipment status data and environmental monitoring data from the N time sources; a cross-domain coupling layer, which processes the equipment status data and the environmental monitoring data respectively, and obtains the corresponding equipment status feature data set and the environmental monitoring feature data set; and accurate control of time synchronization is achieved through the division of labor and cooperation of each module.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and more specifically, to an electric power time synchronization hierarchical intelligent control system. Background Art

[0002] Patent publication number CN110708133A discloses a method and device for clock synchronization and time synchronization in a system based on FPGA, the method comprising: receiving a reference clock sent by a master subsystem through FPGA, and outputting a first clock with the same frequency as the reference clock to update the system clock of the slave subsystem; sampling bit information based on the first clock with the same frequency as the reference clock, and parsing the data frame corresponding to the sampled bit information; extracting the system time of the master subsystem according to the parsed data frame; adding the line delay time to the system time of the master subsystem, and outputting the first time synchronized with the system time of the master subsystem in real time to update the system time of the slave subsystem. This application realizes system clock synchronization and system time synchronization, and the implementation method is simple and reliable.

[0003] The existing time synchronization control system lacks analysis of noise and high-frequency details in the signal, and cannot extract valuable features from complex signals; it does not combine cross-correlation analysis for matching calculation, and cannot further efficiently extract time signal features to ensure accurate synchronization between the reference time signal and each subsystem; it does not have the ability of adaptive optimization, and cannot dynamically adjust processing parameters according to different signal characteristics and noise environments to improve the flexibility and accuracy of signal analysis; it cannot comprehensively evaluate the impact of various equipment states and environmental changes on the reference time signal, and the accuracy and robustness of the time synchronization system are insufficient; it does not analyze the potential impact of equipment status and environmental monitoring data on the characteristics of the reference time signal, and capture the deep relationship between different features, which leads to insufficient accuracy of the prediction model; the accuracy of feature selection is not high, so the model wastes energy on processing feature data that has no influence; it cannot overcome the multicollinearity problem that may exist in traditional regression models, making the system's performance in time synchronization prediction unstable.

[0004] In view of this, the present invention proposes a power time synchronization hierarchical intelligent control system to solve the above problems. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: a power time synchronization hierarchical intelligent control system, comprising:

[0006] The clock source management layer is used to manage N time sources and provide reference time signal data; N time sources are divided into primary time sources and backup time sources, and a time source monitoring mechanism is introduced to automatically switch to the backup time source when the primary time source fails;

[0007] The clock layer is used to receive the reference time signal data through the clock and synchronize it to each power device;

[0008] The multi-time source acquisition layer is used to collect reference time signal data, equipment status data and environmental monitoring data from N time sources;

[0009] The cross-domain coupling layer processes the device status data and the environmental monitoring data respectively to obtain the corresponding device status feature data set and the environmental monitoring feature data set; decomposes the reference time signal data through wavelet transform, adjusts the scale factor in the wavelet transform, calculates the statistical characteristics and matching degree of the time signal component, fuses the multi-scale feature data and the signal matching degree feature data, and obtains the reference time signal feature data set;

[0010] A partial minimum regression model is constructed, and the influence of the equipment state characteristic data set and the environmental monitoring characteristic data set on the reference time signal characteristic data set is evaluated by constraining the partial minimum regression coefficient to obtain the influence characteristic data set;

[0011] The error prediction layer obtains the time synchronization error prediction model based on the influencing feature data set training, and predicts the time synchronization error; compares the predicted time synchronization error with the preset time synchronization error threshold to determine whether clock correction is required;

[0012] The correction control layer maintains the current time synchronization state if clock correction is not required; if clock correction is required, a correction instruction is generated through the power time synchronization intelligent control terminal to adjust the clock's time synchronization state and synchronize it to each power device.

[0013] Furthermore, the N types of time sources adopt a multi-source selection mode signal source for mutual redundant backup, specifically including satellite signals, upper-level ground link time codes and hot standby signals.

[0014] Furthermore, the reference time signal data includes reference time signal, time signal strength, signal frequency, time signal source, time signal delay and signal noise; the equipment status data includes time source status, clock operation status, power equipment operation status, equipment operation time and load status; the environmental monitoring data includes temperature, humidity, electromagnetic interference and ionospheric delay.

[0015] Furthermore, the method for processing the equipment status data and the environment monitoring data separately includes:

[0016] The equipment status data and environmental monitoring data are cleaned, noise is removed and standardized; the periodic fluctuation characteristics and statistical feature data in the equipment status data and environmental monitoring data are extracted respectively through Fourier transform and statistical feature calculation; the periodic fluctuation characteristics extracted from the equipment status data are merged with the statistical feature data to obtain the equipment status feature data set; the periodic fluctuation characteristics extracted from the environmental monitoring data are merged with the statistical feature data to obtain the environmental monitoring feature data set.

[0017] Furthermore, the method for acquiring the reference time signal feature data set includes:

[0018] S51, preset reference time signal data is f(c)={c 1 ,...,c i ,...,c n}; c is the reference time signal variable; c i is the i-th reference time signal variable, i=1,2,...,n; n is the index of the reference time signal variable;

[0019] S52, applying a wavelet transform formula to decompose the reference time signal data into low-frequency and high-frequency time signal components; the wavelet transform formula is: Among them, W f (a, b) are the wavelet transform coefficients of the reference time signal data f(c) at scale a and position b; is the conjugate complex number of the wavelet basis function; a is the scale factor; b is the translation factor; dc is the signal sampling time interval;

[0020] The scale factor a is limited by the scale factor adjustment formula; the scale factor adjustment formula is specifically: Among them, a 0 is the initial scale factor; L signal is the length of the time signal; F signal is the signal frequency; γ noise is the signal noise;

[0021] Calculate statistical features for each time signal component, including energy, mean, and variance; collect statistical features of all time signal components to obtain multi-scale feature data;

[0022] S53, preset a reference time signal y(c′), calculate a cross-correlation function between the low-frequency and high-frequency time signal components and the reference signal, and quantify the signal matching degree between the low-frequency and high-frequency time signal components and the reference time signal through the cross-correlation function;

[0023] By maximizing the cross-correlation function, the corresponding time delay τ is obtained max, the signal matching degree between the low-frequency and high-frequency time signal components and the reference time signal is calculated according to the matching degree formula to obtain the signal matching degree characteristic data; the matching degree formula is: Where ρ is the signal matching degree; R(τ max ) is the cross-correlation function at time delay τ max The value at R x(c) (0) is the value of the autocorrelation function of the time signal component x(c) when τ = 0; R y(c′) (0) is the value of the autocorrelation function of the reference time signal y(c′) at τ = 0;

[0024] S54, concatenating and fusing the multi-scale feature data and the signal matching feature data to obtain a reference time signal feature data set.

[0025] Furthermore, the method for obtaining the impact feature data set includes:

[0026] S61, taking the device state feature data set, the environment monitoring feature data set and the reference time signal feature data set as samples; merging the device state feature data set and the environment monitoring feature data set in the samples into an input matrix X; each row in the input matrix represents a device state feature data set and an environment monitoring feature data set in a sample, and each column in the input matrix represents a feature in the device state feature data set and the environment monitoring feature data set;

[0027] The reference time signal feature data set in the sample is defined as a target matrix Y; each row in the target matrix represents a reference time signal feature data set in a sample; each column in the target matrix represents a feature in the reference time signal feature data set;

[0028] S62, constructing latent variables to maximize the covariance between the input matrix X and the target matrix Y; performing a linear transformation on the input matrix X to obtain a latent variable T = XW′; wherein W′ is a weight matrix; performing a linear transformation on the target matrix Y to obtain a latent variable U = YC; wherein C is a regression coefficient matrix;

[0029] S63. Construct a partial minimum regression model to predict the target matrix Y through the latent variable T; the partial minimum regression model is: Among them, Q is the partial minimum regression coefficient matrix; m is the number of samples; y e is the target value of the e-th sample; k′ is the number of latent variables T; t ed is the dth latent variable of the eth sample; q d is the partial minimum regression coefficient, which is used to associate the dth latent variable with the target matrix Y; d is the index of the latent variable; e is the index of the sample;

[0030] S64. Constrain the partial minimum regression coefficient in the partial minimum regression model to obtain the constrained partial minimum regression coefficient, calculate the absolute value of the constrained partial minimum regression coefficient, obtain the degree of influence of the equipment status feature data set and the environmental monitoring feature data set on the reference time signal feature data set, preset an influence degree threshold, and select the feature composition corresponding to the partial minimum regression coefficient whose influence degree is greater than the influence degree threshold to influence the feature data set.

[0031] Furthermore, the method for constraining the partial minimum regression coefficient in the partial minimum regression model includes:

[0032] The partial minimum regression coefficient in the partial minimum regression model is constrained by the regression coefficient restriction formula; the regression coefficient restriction formula is: Among them, q ′ d is the partial minimum regression coefficient after restriction; is the maximum absolute value of all partial minimum regression coefficients; r is the number of feature types in the evaluation equipment status feature dataset and the environmental monitoring feature dataset.

[0033] Furthermore, the method for acquiring a time synchronization error prediction model by training the influencing feature data set includes:

[0034] The data set is divided into a training set, a validation set and a test set, and a time synchronization error prediction model is constructed; the time synchronization error prediction model includes an input layer, a GRU layer, a fully connected layer and an output layer; the input layer of the time synchronization error prediction model is used to input the historical impact feature data set, and the number of neurons in the input layer matches the number of features of the historical impact feature data set; the time synchronization error prediction model is a gated recurrent unit model;

[0035] Use mean absolute error as the loss function; use training set data to train the model and minimize the loss function through Adam optimizer; use validation set to evaluate the performance of the model and tune the model's hyperparameters until the preset number of iterations is reached; use test set to evaluate the performance of the model in the prediction task and input the current influencing feature data set into the trained time synchronization error prediction model to obtain the time synchronization error.

[0036] Furthermore, the method of comparing the predicted time synchronization error with a preset time synchronization error threshold to determine whether clock correction is required includes:

[0037] If the predicted time synchronization error is greater than or equal to a preset time synchronization error threshold, it is determined that clock correction is required;

[0038] If the predicted time synchronization error is less than a preset time synchronization error threshold, it is determined that clock correction is not required.

[0039] Furthermore, the method of generating a correction instruction through the power time synchronization intelligent control terminal, adjusting the time synchronization state of the clock, and synchronizing to each power device includes:

[0040] The real-time time synchronization error of each power device is collected to obtain the time synchronization error data; the PID control signal is calculated based on the time synchronization error data, and the clock deviation amount that needs to be adjusted is determined according to the calculated PID control signal; the power time synchronization intelligent control terminal generates a correction instruction according to the acquired clock deviation amount that needs to be adjusted, and automatically adjusts the time synchronization state of the clock; the adjusted clock time synchronization state is synchronized to each power device.

[0041] The technical effects and advantages of the electric power time synchronization hierarchical intelligent control system of the present invention are as follows:

[0042] The present invention decomposes the reference time signal into time signal components of different frequency bands through wavelet transform, effectively processes the noise and high-frequency details in the signal, performs signal analysis in the time and frequency domains, can extract valuable features from complex signals, and performs matching degree calculation on different frequency bands in combination with cross-correlation analysis, further efficiently extracts time signal features, and ensures accurate synchronization between the reference time signal and each subsystem; the scale factor adjustment formula and matching degree formula adopted have the ability of adaptive optimization, can dynamically adjust processing parameters according to different signal characteristics and noise environments, and improve the flexibility and accuracy of signal analysis;

[0043] By combining the device state feature data set with the environmental monitoring feature data set, the impact of various device states and environmental changes on the reference time signal can be comprehensively evaluated to ensure the accuracy and robustness of the time synchronization system; by constructing latent variables to maximize the covariance between the input matrix and the target matrix, the potential impact of device state and environmental monitoring data on the reference time signal characteristics can be extracted, and the deep relationship between different features can be captured, thereby enhancing the accuracy of the prediction model; by calculating the absolute value of the regression coefficient, setting the influence degree threshold, and selecting features with an influence degree greater than the threshold to construct the influence feature data set, redundant features are effectively reduced, the accuracy of feature selection is improved, and the model can focus on processing those feature data that are truly influential; by using the partial minimum regression model, through regression analysis of the latent variables, the target matrix (i.e., the reference time signal feature data set) can be effectively predicted under multicollinearity conditions, overcoming the multicollinearity problem that may exist in traditional regression models, so that the system can perform time synchronization prediction more stably. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A schematic diagram of the structure of a power time synchronization hierarchical intelligent control system of the present invention;

[0045] Figure 2 A schematic diagram of a flow chart of a power time synchronization hierarchical intelligent control method of the present invention;

[0046] Figure 3 This is a flow chart of the clock correction method provided by the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] Embodiment 1

[0049] See also Figure 1 and Figure 3 As shown, this embodiment further describes a power time synchronization hierarchical intelligent control system, including:

[0050] The power time synchronization hierarchical intelligent control system is developed based on the high-precision time synchronization requirements of the power system. The power system is a time-related system, and the changes in its various parameters (such as voltage, current, phase angle, power angle) are all based on the waveform of the time axis. Therefore, ensuring that each device and subsystem in the power system can operate under a unified time reference is crucial for the stable operation and fault analysis of the power system. Although power time synchronization technology has been widely used, there are still some problems that need to be solved in practical applications:

[0051] Time synchronization accuracy: With the development of power systems, the requirements for time synchronization accuracy are getting higher and higher. Especially in smart grids, it is necessary to ensure that the time synchronization accuracy between various devices and subsystems is better than 1 microsecond to ensure reliable power distribution management and fault analysis. Therefore, how to improve time synchronization accuracy is an important issue that needs to be solved at present.

[0052] Time synchronization reliability issues: There are various interference factors in the power system (such as electromagnetic interference, equipment failure, etc.), which may affect the reliability of time synchronization. If the time synchronization system fails or has errors, it may cause equipment malfunction or data misreporting in the power system, thus affecting the stable operation of the power system. Therefore, how to improve the reliability of time synchronization is also an important issue that needs to be solved at present.

[0053] Time synchronization management issues: With the increase in the number and complexity of equipment in power systems, time synchronization management has become increasingly difficult. How to effectively monitor and manage the time synchronization status in power systems and promptly detect and handle time synchronization errors or failures is a key issue that needs to be addressed.

[0054] In order to effectively solve the above problems, the present invention proposes a power time synchronization hierarchical intelligent control system, comprising:

[0055] The clock source management layer is used to manage N time sources and provide reference time signal data; N time sources are divided into primary time sources and backup time sources, and a time source monitoring mechanism is introduced to automatically switch to the backup time source when the primary time source fails;

[0056] The clock layer is used to receive the reference time signal data through the clock and synchronize it to each power device;

[0057] The multi-time source acquisition layer is used to collect reference time signal data, equipment status data and environmental monitoring data from N time sources;

[0058] The cross-domain coupling layer processes the device status data and the environmental monitoring data respectively to obtain the corresponding device status feature data set and the environmental monitoring feature data set; decomposes the reference time signal data through wavelet transform, adjusts the scale factor in the wavelet transform, calculates the statistical characteristics and matching degree of the time signal component, fuses the multi-scale feature data and the signal matching degree feature data, and obtains the reference time signal feature data set;

[0059] A partial minimum regression model is constructed, and the influence of the equipment state characteristic data set and the environmental monitoring characteristic data set on the reference time signal characteristic data set is evaluated by constraining the partial minimum regression coefficient to obtain the influence characteristic data set;

[0060] The error prediction layer obtains the time synchronization error prediction model based on the influencing feature data set training, and predicts the time synchronization error; compares the predicted time synchronization error with the preset time synchronization error threshold to determine whether clock correction is required;

[0061] The correction control layer maintains the current time synchronization state if clock correction is not required; if clock correction is required, a correction instruction is generated through the power time synchronization intelligent control terminal to adjust the clock's time synchronization state and synchronize it to each power device.

[0062] N types of time sources use a multi-source selection mode for mutual redundant backup of signal sources, including satellite signals, upper-level ground link time codes and hot standby signals; among them, satellite signals mainly rely on Beidou satellite signals, and the time synchronization benchmark of the position is obtained by receiving satellite signals; the upper-level ground link time code is provided through a wired network, and the time code information is received from the upper-level system or time management center; and the hot standby signal receives the backup time source information through the hot backup system, ensuring rapid switching when the main time source fails to maintain the continuity of time synchronization.

[0063] The reference time signal data includes the reference time signal, time signal strength, signal frequency, time signal source, time signal delay and signal noise; the equipment status data includes the time source status, clock operation status, power equipment operation status, equipment operation time and load status; the environmental monitoring data includes temperature, humidity, electromagnetic interference and ionospheric delay.

[0064] Methods for processing equipment status data and environmental monitoring data separately include:

[0065] The equipment status data and environmental monitoring data are cleaned, noise is removed and standardized; the periodic fluctuation characteristics and statistical feature data in the equipment status data and environmental monitoring data are extracted respectively through Fourier transform and statistical feature calculation; the periodic fluctuation characteristics extracted from the equipment status data are merged with the statistical feature data to obtain the equipment status feature data set; the periodic fluctuation characteristics extracted from the environmental monitoring data are merged with the statistical feature data to obtain the environmental monitoring feature data set.

[0066] The method for obtaining the reference time signal feature data set includes:

[0067] S51, preset reference time signal data is f(c)={c 1 ,...,c i ,...,c n}; c is the reference time signal variable; c i is the i-th reference time signal variable, i=1,2,...,n; n is the index of the reference time signal variable;

[0068] S52, applying a wavelet transform formula to decompose the reference time signal data into low-frequency and high-frequency time signal components; the wavelet transform formula is: Among them, W f (a, b) are the wavelet transform coefficients of the reference time signal data f(c) at scale a and position b; is the conjugate complex number of the wavelet basis function; a is the scale factor; b is the translation factor; dc is the signal sampling time interval; the scale factor a is limited by the scale factor adjustment formula; the scale factor adjustment formula is specifically: Among them, a 0 is the initial scale factor; L signal is the length of the time signal; F signal is the signal frequency; γ noise is the signal noise;

[0069] It should be noted that signal frequency and time signal length are the basic properties of signals. Signals of different frequencies and lengths have their features distributed on different time and frequency scales. For example, high-frequency signals change rapidly, so a smaller scale is needed to capture their details; low-frequency signals change slowly, so a larger scale can better show their overall trend. Signal noise is a key factor affecting signal quality and feature extraction. Noise will interfere with the judgment of the true characteristics of the signal. When the noise is large, in order to extract signal features more accurately, the scale factor needs to be adjusted.

[0070] Wavelet transform is an important method for signal feature extraction, and the scale factor is crucial to the effect of wavelet transform. Through the scale factor adjustment formula, the scale factor can be automatically adjusted according to the characteristics (frequency, length) and noise environment of different signals, so that the wavelet transform can better adapt to various signal conditions. For example, when processing reference time signal data from different time sources, these signals may have different frequencies, lengths and noise levels. By adjusting the scale factor, the wavelet transform can more effectively decompose the signal, extract valuable features, and improve the accuracy of signal analysis. Accurate scale factors enable wavelet transform to capture signal features more accurately in different frequency bands. In the power time synchronization system, accurate extraction of the characteristics of the reference time signal is crucial to the accuracy of time synchronization. By adjusting the scale factor through this formula, the signal can be analyzed more finely in the time and frequency domains, and the signal characteristics of different frequency components can be separated, avoiding the loss of important information or the introduction of noise interference due to improper scale selection, thereby optimizing the effect of reference time signal feature extraction and ensuring the accuracy of time synchronization.

[0071] From the perspective of signal processing principles, the design of the scale factor adjustment formula conforms to the characteristics of wavelet transform. Wavelet transform decomposes the signal through wavelet basis functions of different scales. The appropriate scale factor can make the wavelet basis function better match the signal characteristics. The formula adjusts the scale factor according to the signal frequency, length and noise, which can make the wavelet basis function more effectively adapt to the different frequency components of the signal at different scales, thereby achieving more accurate signal decomposition and feature extraction;

[0072] Compared with the existing technology, the technical effect is:

[0073] Enhanced flexibility of signal processing: Existing technologies may lack a mechanism to dynamically adjust the scale factor according to signal characteristics and noise environment, and usually adopt fixed scale factors or simple adjustment methods. The scale factor adjustment formula of this system can flexibly adjust the scale factor according to the frequency, length and noise conditions of the signal collected in real time, so that the system can better process various complex signals and improve the flexibility of signal processing. For example, in the power system, the signal characteristics and noise levels of different time sources may change at any time. This formula can enable the system to adapt to these changes in a timely manner, which may not be possible with existing technologies.

[0074] Improve time synchronization accuracy: By optimizing the scale factor, the system can more accurately extract the characteristics of the reference time signal and reduce the interference of noise on the signal characteristics. This is crucial for the power time synchronization system, because accurate signal feature extraction helps to improve the accuracy of time synchronization. In contrast, the existing technology may lead to inaccurate signal feature extraction due to unreasonable adjustment of the scale factor, which in turn affects the accuracy of time synchronization. In this system, more accurate time synchronization can ensure the coordinated operation of power equipment and reduce failures and errors caused by time asynchrony.

[0075] Improve the robustness of the system: The scale factor adjustment formula enables the system to automatically adjust the scale factor and maintain good signal processing capabilities when facing different signal environments. This adaptive capability improves the robustness of the system, allowing it to work stably in complex and changeable power environments, such as those subject to varying degrees of electromagnetic interference and signal frequency fluctuations. When facing similar situations, existing technologies may suffer from reduced signal processing effects and unstable system performance because the scale factor cannot adapt to environmental changes.

[0076] For example, assuming that the initial scale factor is 1, the signal frequency is 50 Hz, the signal length is 100, and the signal noise is 0.2; the scale factor can be obtained through the scale factor adjustment formula:

[0077] Calculate statistical features for each time signal component, including energy, mean, and variance; collect statistical features of all time signal components to obtain multi-scale feature data;

[0078] It should be noted that in the field of signal processing, energy is an important indicator to describe signal strength. The energy included in the statistical characteristics specifically refers to the energy spectral density of the time signal component, which reflects the energy distribution of the time signal component at each frequency;

[0079] S53, preset a reference time signal y(c′), calculate a cross-correlation function between the low-frequency and high-frequency time signal components and the reference signal, and quantify the signal matching degree between the low-frequency and high-frequency time signal components and the reference time signal through the cross-correlation function;

[0080] By maximizing the cross-correlation function, the corresponding time delay τ is obtained max , the signal matching degree between the low-frequency and high-frequency time signal components and the reference time signal is calculated according to the matching degree formula to obtain the signal matching degree characteristic data; the matching degree formula is: Where ρ is the signal matching degree; R(τ max ) is the cross-correlation function at time delay τ max The value at R x(c) (0) is the value of the autocorrelation function of the time signal component x(c) when τ = 0; R y(c′) (0) is the value of the autocorrelation function of the reference time signal y(c′) at τ = 0;

[0081] It should be noted that the cross-correlation function and the autocorrelation function are common technical means. The cross-correlation function can be used to measure the similarity between two signals, especially the matching degree when offset (delayed) over time; the autocorrelation function is a special case of the cross-correlation function, which represents the correlation between a signal and itself, that is, it measures the periodicity or repetitive pattern of the signal;

[0082] S54, concatenating and fusing the multi-scale feature data and the signal matching feature data to obtain a reference time signal feature data set.

[0083] The value range of the matching degree ρ is between [-1, 1] and has the following meanings:

[0084] When ρ = 1, it means that the reference time signal is completely consistent with the low-frequency and high-frequency time signal components in time and frequency, without any delay, frequency offset or noise;

[0085] When ρ = 0, it means that there is no similarity between the reference time signal and the low-frequency and high-frequency time signal components, and there is time offset or noise interference, resulting in a complete mismatch of the signals;

[0086] When ρ=-1, it indicates that the reference time signal and the low-frequency and high-frequency time signal components present a completely anti-phase relationship at all time points, which usually occurs when the signal is severely interfered or phase-reversed.

[0087] Assuming a time signal x(c) is received from a satellite clock and there is a local time signal y(c′) as a reference, the matching degree of the two signals is evaluated by cross-correlation analysis;

[0088] First, collect data points of two signals. Suppose we have the following data:

[0089] Received time signal x(c): [1,2,3,4,5,6,7,8,9,10];

[0090] Local time signal y(c′): [0.9, 2.1, 2.9, 4.1, 4.9, 6.1, 6.9, 8.1, 9.1, 9.9];

[0091] Calculate the time delay between two signals max The value of the cross-correlation function under τ is used to find the maximum value of the cross-correlation function R(τ max ); Assuming that the maximum value of the cross-correlation function occurs at τ = 0, calculate the value of the autocorrelation function of the received time signal x(c) and the local time signal y(c′) at τ = 0: R x(c) (0)=(1 2 +2 2 +...+10 2 )=385;R y(c′) (0) = (0.9 2 +2.1 2 +...+9.9 2 )≈361.82; according to R x(c) (0) and R y(c′) (0) Calculate the matching degree: Based on the matching degree, it can be concluded that the received time signal and the local time signal match very well. Although there are slight time delays and amplitude changes, they are still very similar overall.

[0092] The methods that affect the acquisition of feature data sets include:

[0093] S61, taking the device state feature data set, the environment monitoring feature data set and the reference time signal feature data set as samples; merging the device state feature data set and the environment monitoring feature data set in the samples into an input matrix X; each row in the input matrix represents a device state feature data set and an environment monitoring feature data set in a sample, and each column in the input matrix represents a feature in the device state feature data set and the environment monitoring feature data set;

[0094] The reference time signal feature data set in the sample is defined as a target matrix Y; each row in the target matrix represents a reference time signal feature data set in a sample; each column in the target matrix represents a feature in the reference time signal feature data set;

[0095] S62, constructing latent variables to maximize the covariance between the input matrix X and the target matrix Y; performing a linear transformation on the input matrix X to obtain a latent variable T = XW′; wherein W′ is a weight matrix; performing a linear transformation on the target matrix Y to obtain a latent variable U = YC; wherein C is a regression coefficient matrix;

[0096] S63. Construct a partial minimum regression model to predict the target matrix Y through the latent variable T; the partial minimum regression model is: Among them, Q is the partial minimum regression coefficient matrix; m is the number of samples; y e is the target value of the e-th sample; k′ is the number of latent variables T; t ed is the dth latent variable of the eth sample; q d is the partial minimum regression coefficient, which is used to associate the dth latent variable with the target matrix Y; d is the index of the latent variable; e is the index of the sample;

[0097] S64. Constrain the partial minimum regression coefficient in the partial minimum regression model to obtain the constrained partial minimum regression coefficient, calculate the absolute value of the constrained partial minimum regression coefficient, obtain the degree of influence of the equipment status feature data set and the environmental monitoring feature data set on the reference time signal feature data set, preset an influence degree threshold, and select the feature composition corresponding to the partial minimum regression coefficient whose influence degree is greater than the influence degree threshold to influence the feature data set.

[0098] For example, suppose you have the following dataset:

[0099] As shown in Table 1: The input matrix X is a 5×5 matrix containing 5 samples and 5 features;

[0100] As shown in Table 2: The target matrix Y is a 5×3 matrix containing 5 samples and 3 target features;

[0101] Table 1 Input matrix X (equipment status feature data set + environmental monitoring feature data set)

[0102]

[0103] Table 2 Target matrix Y (reference time signal feature data set)

[0104]

[0105] Use linear transformation to map the input matrix X to the latent variable space to obtain the latent variable matrix T. Similarly, the target matrix Y is also linearly transformed to obtain the latent variable matrix U; assuming that three latent variables are chosen, we need to solve the weight matrix W′ and the regression coefficient matrix;

[0106] The partial minimum regression coefficient in the partial minimum regression model is constrained by the regression coefficient restriction formula. Assume that the calculated latent variable matrix is: Then, the sum of the absolute values ​​of each column (each latent variable) can be calculated:

[0107] The first latent variable:

[0108] The second latent variable:

[0109] The third latent variable:

[0110] So the maximum absolute value of all partial minimum regression coefficients is Then the restricted partial minimum regression coefficient

[0111] After constraining the regression coefficients in the partial minimum regression model through the regression coefficient restriction formula, the absolute value of the restricted partial minimum regression coefficient is calculated. Assume that the calculated regression coefficient matrix is: These coefficients are then compared to the limiting values:

[0112] q 1 =0.3 (greater than 0.208, needs to be constrained to 0.208); q 2 =0.4 (greater than 0.208, needs to be constrained to 0.208); q 3 =0.2 (less than 0.208, no adjustment required), assuming the influence threshold is 1.5, select features whose influence is greater than the influence threshold; therefore, the influence feature data set will contain the first and second features.

[0113] Methods for constraining the partial minimum regression coefficients in the partial minimum regression model include:

[0114] The partial minimum regression coefficient in the partial minimum regression model is constrained by the regression coefficient restriction formula; the regression coefficient restriction formula is: Among them, q ′ d is the partial minimum regression coefficient after restriction; is the maximum absolute value of all partial minimum regression coefficients; r is the number of feature types in the evaluation equipment status feature dataset and the environmental monitoring feature dataset.

[0115] When analyzing the impact of the equipment status feature dataset and the environmental monitoring feature dataset on the benchmark time signal feature dataset, it is necessary to measure the importance of each feature. Different features have different impacts on the target variable (the benchmark time signal feature dataset). Some features may have a greater impact and are key factors, while some features may have a smaller impact and may even be redundant. By restricting and analyzing the partial minimum regression coefficient, it is possible to determine which features actually have a significant impact on the benchmark time signal features, thereby screening out key features and building a more accurate model.

[0116] In actual data, there may be multicollinearity between equipment status features and environmental monitoring features, that is, there is a strong linear correlation between these features. When dealing with multicollinearity, the coefficient estimation of traditional regression models becomes unstable, resulting in a decrease in model prediction performance. This formula can alleviate the impact of multicollinearity to a certain extent by limiting the partial minimum regression coefficient, making the model more stable and reliable.

[0117] By calculating the maximum absolute value of all partial minimum regression coefficients and combining the number of samples and the number of feature types, the restricted partial minimum regression coefficient is obtained. Taking this limit value as the standard, the features with an influence greater than the threshold are selected to form the influence feature data set, which can effectively remove redundant features that have little influence on the benchmark time signal feature data set, improve the accuracy of feature selection, and allow the model to focus on processing the truly important feature data. The screened influence feature data set contains features that have a greater impact on the target variable. The model trained based on these features can more accurately capture the patterns in the data, thereby improving the model's ability to predict time synchronization errors, making the system's prediction of time synchronization status more accurate, and thus improving the accuracy and robustness of the power time synchronization system.

[0118] This formula adjusts the regression coefficient based on the statistical characteristics of the data, which is in line with statistical principles. In the partial minimum regression model, by reasonably restricting the coefficients, it is possible to better balance the contribution of each feature to the target variable, making the model more consistent with the inherent laws of the actual data and improving the model's interpretability and predictive ability.

[0119] Compared with the existing technology, the technical effect is:

[0120] More accurate feature selection: The existing technology may lack effective feature selection methods, or the selection process is relatively simple, and it is unable to fully consider the complex relationship between features and the actual impact on the target variable. The regression coefficient restriction formula of the present invention can comprehensively evaluate the impact of features, more accurately screen out key features, reduce the interference of redundant information on the model, and improve the efficiency and accuracy of the model.

[0121] Stronger model stability: Due to the shortcomings of the existing technology in dealing with multicollinearity, the model stability is poor. The present invention effectively alleviates the multicollinearity problem by limiting the partial minimum regression coefficient, making the model more stable when facing complex data and the prediction results more reliable. In the power time synchronization system, the time synchronization error can be predicted more stably, reducing the problem of inaccurate time synchronization caused by model instability.

[0122] Higher time synchronization system performance: Based on more accurate feature selection and stronger model stability, the method of the present invention can significantly improve the performance of the power time synchronization system. Accurate time synchronization error prediction helps to timely discover and adjust the time synchronization status, ensure the time synchronization accuracy between power equipment, and improve the overall operation efficiency and reliability of the power system, which is difficult to achieve with existing technologies.

[0123] The method of obtaining a time synchronization error prediction model by training an impact feature data set includes:

[0124] The data set is divided into a training set, a validation set and a test set, and a time synchronization error prediction model is constructed; the sample set is a subset of the data set, and each sample set includes a historical impact feature data set and a corresponding time synchronization error; the time synchronization error prediction model includes an input layer, a GRU layer, a fully connected layer and an output layer; the input layer of the time synchronization error prediction model is used to input the historical impact feature data set, and the number of neurons in the input layer matches the number of features in the historical impact feature data set;

[0125] The GRU layer is used to process the historical impact feature dataset, and the number of GRU layers and neurons are adjusted according to the complexity of the task; the fully connected layer is used to provide additional nonlinear transformation; the model output layer is used to output the time synchronization error, and the predicted value is output through a neuron, and the identity function is used as the activation function; the time synchronization error prediction model is a gated recurrent unit model;

[0126] Use mean absolute error as the loss function to measure the error between the model's predicted value and the actual value; use the training set data to train the model and minimize the loss function through the Adam optimizer; use the validation set to evaluate the performance of the model and tune the model's hyperparameters until the preset stopping condition is reached;

[0127] The test set is used to evaluate the performance of the model in the prediction task, and the current influencing feature dataset is input into the trained time synchronization error prediction model to obtain the time synchronization error.

[0128] The method of comparing the predicted time synchronization error with a preset time synchronization error threshold to determine whether clock correction is required includes:

[0129] If the predicted time synchronization error is greater than or equal to a preset time synchronization error threshold, it is determined that clock correction is required;

[0130] If the predicted time synchronization error is less than a preset time synchronization error threshold, it is determined that clock correction is not required.

[0131] The method of generating a correction instruction through a power time synchronization intelligent control terminal, adjusting the time synchronization state of a clock, and synchronizing to each power device includes:

[0132] The real-time time synchronization error of each power device is collected to obtain the time synchronization error data; the PID control signal is calculated based on the time synchronization error data, and the clock deviation amount that needs to be adjusted is determined according to the calculated PID control signal; the power time synchronization intelligent control terminal generates a correction instruction according to the acquired clock deviation amount that needs to be adjusted, and automatically adjusts the time synchronization state of the clock; the adjusted clock time synchronization state is synchronized to each power device.

[0133] In this embodiment, the present invention decomposes the reference time signal into time signal components of different frequency bands through wavelet transform, effectively processes the noise and high-frequency details in the signal, performs signal analysis in the time and frequency domains, can extract valuable features from complex signals, and performs matching degree calculation on different frequency bands in combination with cross-correlation analysis, further efficiently extracts time signal features, and ensures accurate synchronization between the reference time signal and each subsystem; the scale factor adjustment formula and matching degree formula adopted have the ability of adaptive optimization, can dynamically adjust processing parameters according to different signal characteristics and noise environments, and improve the flexibility and accuracy of signal analysis;

[0134] By combining the device state feature data set with the environmental monitoring feature data set, the impact of various device states and environmental changes on the reference time signal can be comprehensively evaluated to ensure the accuracy and robustness of the time synchronization system; by constructing latent variables to maximize the covariance between the input matrix and the target matrix, the potential impact of device state and environmental monitoring data on the reference time signal characteristics can be extracted, and the deep relationship between different features can be captured, thereby enhancing the accuracy of the prediction model; by calculating the absolute value of the regression coefficient, setting the influence degree threshold, and selecting features with an influence degree greater than the threshold to construct the influence feature data set, redundant features are effectively reduced, the accuracy of feature selection is improved, and the model can focus on processing those feature data that are truly influential; by using the partial minimum regression model, through regression analysis of the latent variables, the target matrix (i.e., the reference time signal feature data set) can be effectively predicted under multicollinearity conditions, overcoming the multicollinearity problem that may exist in traditional regression models, so that the system can perform time synchronization prediction more stably.

[0135] Embodiment 2

[0136] See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, which provides a power time synchronization hierarchical intelligent control method, including:

[0137] S1. Manage N time sources and provide reference time signal data; divide N time sources into primary time sources and backup time sources, introduce a time source monitoring mechanism, and automatically switch to the backup time source when the primary time source fails;

[0138] S2, receiving reference time signal data through a clock and synchronizing it to each power device;

[0139] S3, collecting reference time signal data, equipment status data and environmental monitoring data from N time sources;

[0140] S4, respectively process the equipment status data and the environmental monitoring data to obtain the corresponding equipment status feature data set and the environmental monitoring feature data set; decompose the reference time signal data by wavelet transform, adjust the scale factor in the wavelet transform, calculate the statistical characteristics and matching degree of the time signal component, fuse the multi-scale feature data and the signal matching degree feature data, and obtain the reference time signal feature data set;

[0141] A partial minimum regression model is constructed, and the influence of the equipment state characteristic data set and the environmental monitoring characteristic data set on the reference time signal characteristic data set is evaluated by constraining the partial minimum regression coefficient to obtain the influence characteristic data set;

[0142] S5. A time synchronization error prediction model is obtained by training the influencing feature data set to predict the time synchronization error; the predicted time synchronization error is compared with a preset time synchronization error threshold to determine whether clock correction is required;

[0143] S6. If clock correction is not required, maintain the current time synchronization state; if clock correction is required, generate a correction instruction through the power time synchronization intelligent control terminal, adjust the time synchronization state of the clock, and synchronize it to each power device.

[0144] Since the electronic device introduced in this embodiment is an electronic device used to implement a power time synchronization hierarchical intelligent control system in the embodiment of this application, based on the power time synchronization hierarchical intelligent control system introduced in the embodiment of this application, the technical personnel of this field can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application is not introduced in detail here. As long as the technical personnel of this field implement the electronic device used in the power time synchronization hierarchical intelligent control system in the embodiment of this application, it belongs to the scope of protection of this application.

[0145] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0146] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A power time synchronization hierarchical intelligent control system, characterized in that: include: The clock source management layer is used to manage N time sources and provide reference time signal data; N time sources are divided into primary time sources and backup time sources, and a time source monitoring mechanism is introduced to automatically switch to the backup time source when the primary time source fails; The clock layer is used to receive the reference time signal data through the clock and synchronize it to each power device; The multi-time source acquisition layer is used to collect reference time signal data, equipment status data and environmental monitoring data from N time sources; The cross-domain coupling layer processes the device status data and the environmental monitoring data respectively to obtain the corresponding device status feature data set and environmental monitoring feature data set; Decomposing the reference time signal data by wavelet transform, adjusting the scale factor in the wavelet transform, calculating the statistical characteristics and matching degree of the time signal components, fusing the multi-scale feature data and the signal matching feature data, and obtaining the reference time signal feature data set; The method for acquiring the reference time signal feature data set includes: S51, preset reference time signal data is f(c)={c1,...,c i ,...,c n }; c is the reference time signal variable; c i is the i-th reference time signal variable, i=1,2,...,n; n is the index of the reference time signal variable; S52, applying a wavelet transform formula to decompose the reference time signal data into low-frequency and high-frequency time signal components; the wavelet transform formula is: Among them, W f (a, b) are the wavelet transform coefficients of the reference time signal data f(c) at scale a and position b; is the conjugate complex number of the wavelet basis function; a is the scale factor; b is the translation factor; dc is the signal sampling time interval; The scale factor a is limited by the scale factor adjustment formula; the scale factor adjustment formula is: Where a0 is the initial scale factor; L signal is the length of the time signal; F signal is the signal frequency; γ noise is the signal noise; Calculate statistical features for each time signal component, including energy, mean, and variance; collect statistical features of all time signal components to obtain multi-scale feature data; S53, preset a reference time signal y(c′), calculate a cross-correlation function between the low-frequency and high-frequency time signal components and the reference signal, and quantify the signal matching degree between the low-frequency and high-frequency time signal components and the reference time signal through the cross-correlation function; By maximizing the cross-correlation function, the corresponding time delay τ is obtained max , the signal matching degree between the low-frequency and high-frequency time signal components and the reference time signal is calculated according to the matching degree formula to obtain the signal matching degree characteristic data; the matching degree formula is: Where ρ is the signal matching degree; R(τ max ) is the cross-correlation function at time delay τ max The value at R x(c) (0) is the value of the autocorrelation function of the time signal component x(c) when τ = 0; R y(c′) (0) is the value of the autocorrelation function of the reference time signal y(c′) at τ = 0; S54, concatenating and fusing the multi-scale feature data and the signal matching feature data to obtain a reference time signal feature data set; A partial minimum regression model is constructed, and the influence of the equipment state characteristic data set and the environmental monitoring characteristic data set on the reference time signal characteristic data set is evaluated by constraining the partial minimum regression coefficient to obtain the influence characteristic data set; The error prediction layer obtains the time synchronization error prediction model based on the influencing feature data set training, and predicts the time synchronization error; compares the predicted time synchronization error with the preset time synchronization error threshold to determine whether clock correction is required; The correction control layer maintains the current time synchronization state if clock correction is not required; if clock correction is required, a correction instruction is generated through the power time synchronization intelligent control terminal to adjust the clock's time synchronization state and synchronize it to each power device.

2. The power time synchronization hierarchical intelligent control system according to claim 1 is characterized in that: The N time sources adopt a multi-source selection mode signal source mutual redundant backup, including satellite signals, upper-level ground link time codes and hot standby signals.

3. The power time synchronization hierarchical intelligent control system according to claim 2 is characterized in that: The reference time signal data includes reference time signal, time signal strength, signal frequency, time signal source, time signal delay and signal noise; the equipment status data includes time source status, clock operation status, power equipment operation status, equipment operation time and load status; the environmental monitoring data includes temperature, humidity, electromagnetic interference and ionospheric delay.

4. The power time synchronization hierarchical intelligent control system according to claim 3 is characterized in that: The method for processing the equipment status data and the environment monitoring data separately comprises: The equipment status data and environmental monitoring data are cleaned, noise is removed and standardized; the periodic fluctuation characteristics and statistical feature data in the equipment status data and environmental monitoring data are extracted respectively through Fourier transform and statistical feature calculation; the periodic fluctuation characteristics extracted from the equipment status data are merged with the statistical feature data to obtain the equipment status feature data set; the periodic fluctuation characteristics extracted from the environmental monitoring data are merged with the statistical feature data to obtain the environmental monitoring feature data set.

5. The electric power time synchronization hierarchical intelligent control system according to claim 4 is characterized in that: The method for obtaining the impact feature data set includes: S61, taking the device state feature data set, the environment monitoring feature data set and the reference time signal feature data set as samples; merging the device state feature data set and the environment monitoring feature data set in the samples into an input matrix X; each row in the input matrix represents a device state feature data set and an environment monitoring feature data set in a sample, and each column in the input matrix represents a feature in the device state feature data set and the environment monitoring feature data set; The reference time signal feature data set in the sample is defined as a target matrix Y; each row in the target matrix represents a reference time signal feature data set in a sample; each column in the target matrix represents a feature in the reference time signal feature data set; S62, constructing latent variables to maximize the covariance between the input matrix X and the target matrix Y; performing a linear transformation on the input matrix X to obtain a latent variable T = XW′; wherein W′ is a weight matrix; performing a linear transformation on the target matrix Y to obtain a latent variable U = YC; wherein C is a regression coefficient matrix; S63. Construct a partial minimum regression model to predict the target matrix Y through the latent variable T; the partial minimum regression model is: Among them, Q is the partial minimum regression coefficient matrix; m is the number of samples; y e is the target value of the e-th sample; k′ is the number of latent variables T; t ed is the dth latent variable of the eth sample; q d is the partial minimum regression coefficient, which is used to associate the dth latent variable with the target matrix Y; d is the index of the latent variable; e is the index of the sample; S64. Constrain the partial minimum regression coefficient in the partial minimum regression model to obtain the constrained partial minimum regression coefficient, calculate the absolute value of the constrained partial minimum regression coefficient, obtain the degree of influence of the equipment status feature data set and the environmental monitoring feature data set on the reference time signal feature data set, preset an influence degree threshold, and select the feature composition corresponding to the partial minimum regression coefficient whose influence degree is greater than the influence degree threshold to influence the feature data set.

6. The electric power time synchronization hierarchical intelligent control system according to claim 5, characterized in that: The method for constraining the partial minimum regression coefficient in the partial minimum regression model comprises: The partial minimum regression coefficient in the partial minimum regression model is constrained by the regression coefficient restriction formula; the regression coefficient restriction formula is: Among them, q ′ d is the partial minimum regression coefficient after restriction; is the maximum absolute value of all partial minimum regression coefficients; r is the number of feature types in the evaluation equipment status feature dataset and the environmental monitoring feature dataset.

7. The electric power time synchronization hierarchical intelligent control system according to claim 6, characterized in that: The method for acquiring a time synchronization error prediction model by training an influencing feature data set comprises: The data set is divided into a training set, a validation set and a test set, and a time synchronization error prediction model is constructed; the time synchronization error prediction model includes an input layer, a GRU layer, a fully connected layer and an output layer; the input layer of the time synchronization error prediction model is used to input the historical impact feature data set, and the number of neurons in the input layer matches the number of features of the historical impact feature data set; the time synchronization error prediction model is a gated recurrent unit model; Use mean absolute error as the loss function; use training set data to train the model and minimize the loss function through Adam optimizer; use validation set to evaluate the performance of the model and tune the model's hyperparameters until the preset number of iterations is reached; use test set to evaluate the performance of the model in the prediction task and input the current influencing feature data set into the trained time synchronization error prediction model to obtain the time synchronization error.

8. The electric power time synchronization hierarchical intelligent control system according to claim 7, characterized in that: The method of comparing the predicted time synchronization error with a preset time synchronization error threshold to determine whether clock correction is required includes: If the predicted time synchronization error is greater than or equal to a preset time synchronization error threshold, it is determined that clock correction is required; If the predicted time synchronization error is less than a preset time synchronization error threshold, it is determined that clock correction is not required.

9. The electric power time synchronization hierarchical intelligent control system according to claim 8, characterized in that: The method of generating a correction instruction through a power time synchronization intelligent control terminal, adjusting the time synchronization state of a clock, and synchronizing to each power device includes: The real-time time synchronization error of each power device is collected to obtain the time synchronization error data; the PID control signal is calculated based on the time synchronization error data, and the clock deviation amount that needs to be adjusted is determined according to the calculated PID control signal; the power time synchronization intelligent control terminal generates a correction instruction according to the acquired clock deviation amount that needs to be adjusted, and automatically adjusts the time synchronization state of the clock; the adjusted clock time synchronization state is synchronized to each power device.

Images