Multi-reference clock synchronization control method integrating frequency taming and energy efficiency optimization

Abstract

The invention discloses a multi-reference clock synchronization control method integrating frequency taming and energy efficiency optimization, and relates to the field of communication service system time synchronization. According to the method, dynamic characteristics of a reference source are sequentially extracted, a confidence coefficient model is constructed for optimal source selection, self-adaptive controller adjustment is completed through a bionic activation mechanism, a frequency health degree function is introduced for system state judgment, and closed-loop modeling and control of the whole process from source selection and tuning to state evaluation are achieved. According to the method, the frequency tracking stability and the adaptive capacity of the clock system in a multi-source environment are improved, and meanwhile, the observability and the energy efficiency response mechanism of the frequency control process are enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of communication service system time synchronization, in particular to a multi-reference clock synchronization control method combining frequency taming and energy efficiency optimization. BACKGROUND

[0002] In the existing multi-reference clock synchronization system, a single preferred reference source is usually used for local clock synchronization. Common synchronization methods include 1PPS (pulse per second), PTP (precision time protocol), B code (IRIG-B), and TOD (time and date) reference signals. By calculating the deviation between the reference signal and the local clock, a frequency adjustment control signal is output to achieve frequency tracking and locking of the local oscillator. Some systems use PID control structure for feedback calibration, or introduce Kalman filter and PLL (phase-locked loop) technology to improve the short-term stability and locking accuracy of the local clock. In addition, some synchronization systems will switch based on the historical performance of the reference source, but the switching strategy mainly depends on static scoring or manual configuration parameters, and the response sensitivity and stability are insufficient.

[0003] However, the above-mentioned existing synchronization control method has the following defects: first, it lacks unified modeling capability for multiple heterogeneous reference signals, cannot fully extract the time-frequency characteristic information of the reference source, and leads to unstable source selection mechanism; second, the traditional control method is mostly fixed parameter structure, which cannot dynamically and adaptively adjust the controller response according to the system state, resulting in that the frequency taming efficiency and locking accuracy are not ideal when the system faces reference source fluctuation; third, the existing scheme generally lacks quantitative evaluation mechanism of frequency health degree, which cannot judge whether the current oscillator running state reaches the taming target in real time, affecting the robustness and energy management ability of the system. SUMMARY

[0004] The present application proposes a multi-reference clock synchronization control method combining frequency taming and energy efficiency optimization, which introduces a source selection mechanism based on dynamic feature extraction and confidence driving of the reference source, an adaptive frequency taming control strategy of bionic activation model, and an energy efficiency evaluation method of frequency health degree function, solving the technical bottlenecks of low synchronization accuracy, unstable response and uncontrollable power consumption caused by single evaluation dimension of reference source, rigid configuration of control parameters, and undeterminable frequency locking state in the prior art.

[0005] The multi-reference clock synchronization control method combining frequency taming and energy efficiency optimization comprises the following steps:

[0006] S1. Synchronize the acquisition of time data from multiple external reference signal inputs, digitize each reference signal, and extract the dynamic characteristics of each reference signal, including instantaneous synchronization deviation, frequency offset, and frequency stability factor, using a time difference analysis mechanism. Combine the dynamic characteristics to form a time-frequency dynamic characteristic vector for each reference source;

[0007] Specifically, this step synchronously acquires multiple heterogeneous reference signals (such as 1PPS, PTP, B code, and TOD) and extracts a multi-dimensional dynamic characteristic vector containing instantaneous synchronization deviation, frequency offset, and frequency stability factor using a digital time difference analysis mechanism. Compared to traditional methods that rely on a single path or single feature extraction, this step integrates multiple dynamic characteristics, improving the overall perception of the time-frequency state of the reference source, which helps to more accurately reflect the real dynamic behavior of the reference source and improve the quality of the basic data for synchronization control.

[0008] S2. Based on the time-frequency dynamic characteristic vector, construct a reference source confidence evaluation model using a confidence function. Select the reference source with the highest confidence at the current time as the main synchronization reference source, calculate the derivative term for the corresponding reference source based on the main synchronization reference source, and construct a control error vector combining the instantaneous synchronization deviation and frequency offset of the corresponding reference source;

[0009] Specifically, this step constructs a reference source confidence evaluation model using the time-frequency dynamic characteristic vector and the confidence function, dynamically selects the main synchronization reference source with the highest confidence, and then constructs a control error vector based on the derivative term. Compared to traditional fixed main reference sources or simple weighted average strategies, the confidence model can reflect the time-frequency quality and stability of the reference source in real time, effectively avoid synchronization errors caused by abnormal single paths, and enhance the robustness of the system and the ability to adapt to diverse reference signal environments.

[0010] S3. According to the bionic mechanism of the activation model, automatically adjust the controller gain parameters based on the error vector, and output a voltage control signal using the adjusted gain parameters. Adjust the output frequency of the oscillator through the voltage control signal, output the tuned voltage, and return the parameter information at the current time;

[0011] Specifically, this step introduces the bionic mechanism of the activation model to simulate the response of neurons to nonlinear adaptive adjustment of controller gain parameters, achieving dynamic control gain adjustment driven by the error vector. Traditional controllers mostly use fixed or manually adjusted proportional-integral gain. This step uses a self-evolution mechanism to intelligently optimize the controller parameters online, improving the response speed and stability of frequency tracking. At the same time, feedback of the tuned voltage and parameter information realizes closed-loop regulation, enhancing control precision and system adaptability.

[0012] S4. Comprehensive modeling is performed on the state vectors of each link, a frequency health degree function is defined by weighting, and health degree evaluation is performed based on the calculated function value.

[0013] Specifically, by comprehensively modeling the state vectors of each link, a weighted frequency health degree function is defined, and quantitative evaluation of the overall operation health state of the system is realized. Compared with the traditional method of only focusing on a single error indicator, this step can multi-dimensionally integrate the key states of the system to form a more scientific and comprehensive health degree indicator, which is convenient for early warning of system abnormalities and guiding energy efficiency optimization, and improves the safety and efficiency of system operation.

[0014] The beneficial effects of the application are:

[0015] The application sequentially extracts the dynamic characteristics of the reference source, constructs a confidence model for optimal source selection, completes adaptive controller adjustment through a bionic activation mechanism, and introduces a frequency health degree function for system state determination, realizing the whole-process closed-loop modeling and control from source selection, tuning to state evaluation. This method improves the frequency tracking stability and adaptive ability of the clock system in a multi-source environment, and enhances the observability and energy efficiency response mechanism of the frequency control process. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 A method flowchart of a multi-reference clock synchronization control method fusing frequency taming and energy efficiency optimization is provided for the embodiments of the application. DETAILED DESCRIPTION

[0017] The technical solutions of the application will be described in further detail below with reference to the drawings, but the protection scope of the application is not limited to the following description.

[0018] In order to make the purpose, technical solutions and advantages of the application clearer and more apparent, the application will be further described in detail in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the application, and are not used to limit the application, that is, the described embodiments are only a part of the embodiments of the application, but not all the embodiments. The components of the embodiments of the application described and shown in the drawings herein can be arranged and designed in various different configurations.

[0019] Therefore, the following detailed description of the embodiments of the application provided in the accompanying drawings is not intended to limit the scope of the application claimed, but merely represents selected embodiments of the application. Based on the embodiments of the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the application. It should be noted that the relational terms such as "first" and "second" and the like are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations.

[0020] Moreover, the terms "comprising", "containing" or any other variant thereof are intended to cover non-exclusive inclusions, so that a process, method, article or mechanical device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such a process, method, article or mechanical device. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of another identical element in the process, method, article or mechanical device comprising the element.

[0021] The features and performances of the application are further described in detail below in combination with embodiments.

[0022] Embodiment one

[0023] As Figure 1 A multi-reference clock synchronization control method combining frequency taming and energy efficiency optimization includes the following steps:

[0024] S1. Synchronize the collection of time data from multiple external reference signal inputs, digitize each reference signal, extract the dynamic characteristics of each reference signal by combining the time difference analysis mechanism, the dynamic characteristics include instantaneous synchronization deviation, frequency offset and frequency stability factor, and jointly construct the time-frequency dynamic characteristic vector of each reference source;

[0025] S2. Based on the time-frequency dynamic characteristic vector, construct a reference source confidence evaluation model by combining the confidence function; select the reference source with the highest confidence at the current time as the main synchronization reference source, calculate the derivative term of the corresponding reference source according to the main synchronization reference source, and construct the control error vector combining the instantaneous synchronization deviation and frequency offset of the corresponding reference source;

[0026] S3. According to the bionic mechanism of the activation model, automatically adjust the gain parameters of the controller combining the error vector, and output the voltage control signal combining the adjusted gain parameters; adjust the oscillator output frequency through the voltage control signal, output the tuning voltage, and return the parameter information at the current time;

[0027] S4. Comprehensive modeling of each link state vector, defining a frequency health function by weighting, and based on the calculated function value, health evaluation is carried out.

[0028] Further, the step S1 specifically comprises the following sub-steps:

[0029] S101. Receive the 1PPS, PTP, B code and TOD multi-heterogeneous clock reference signals, and uniformly convert them into a digital timestamp signal stream;

[0030] S102. For each reference source, time difference analysis is carried out to extract the instantaneous offset between the local clock; through the continuous time difference data in the sliding window, the instantaneous frequency offset of the reference source is calculated;

[0031] S103. In the same window, based on multiple instantaneous offsets and / or instantaneous frequency offsets, the frequency stability factor of the reference source is calculated;

[0032] S104. The three dynamic characteristic values extracted from each reference source are combined into a set of time-frequency dynamic characteristic vectors as output.

[0033] Specifically, the above steps are specifically designed to realize the time information synchronous acquisition and feature extraction of multiple external heterogeneous reference sources. First, through the interface module, multiple clock reference signals are received, including 1PPS (one pulse per second), PTP (precision time protocol), B code (Beidou time encoding) and TOD (Time of Day) and other different formats of clock signals, and these heterogeneous signals are uniformly converted into standardized digital timestamp stream to ensure the consistency of subsequent processing. Subsequently, high-precision time difference analysis is carried out between each reference source and the local clock, and the instantaneous offset of the reference source relative to the local clock is extracted, and the time difference change is continuously recorded in the sliding time window, and the current frequency offset of the reference source is calculated. Further, based on multiple instantaneous offset values or frequency offset values in the same window, the fluctuation index is calculated, and the frequency stability factor representing the long-term frequency stability capability is constructed, such as through variance, triangular wave detection or Allan variance. Finally, the three dynamic characteristics of each reference source, i.e. instantaneous synchronization deviation, frequency offset and frequency stability factor, are combined to form a time-frequency dynamic characteristic vector, which is a digital description of the current time-frequency characteristics of the reference source, for subsequent synchronization reference source evaluation.

[0034] Further, in the step S103, when calculating the frequency stability factor of the reference source through multiple instantaneous frequency offsets, Allan variance is used for calculation, and the specific flow is represented as:

[0035]

[0036] wherein, the S i (t) represents the frequency stability factor of the i-th reference source at time t, the k represents the time index variable, the represents the Allan variance, the τ represents the observation time interval, the N represents the sampling point number, the k represents the time index variable, the represents the frequency offset of the i-th reference source at the k-th sampling time, the represents the frequency offset of the i-th reference source at the k+1-th sampling time.

[0037] Further, in the step S103, when the frequency stability factor of the reference source is calculated by the plurality of instantaneous offsets, then the MTIE variance is used for calculation, and the specific flow is represented as:

[0038]

[0039] wherein, the S i (t) represents the frequency stability factor of the i-th reference source at time t, the k represents the time index variable, the represents the instantaneous time difference between the i-th reference source and the local clock at the k-th sampling, the mean(Δt i ) represents the time difference mean of the reference source, the max k represents the maximum value of the instantaneous offset within the time window k.

[0040] It should be noted that the above embodiments respectively propose to calculate the frequency stability factor by using the Allan variance and the MTIE variance, wherein the Allan variance is a statistical method for evaluating the random fluctuation characteristics of the oscillator or clock frequency stability. Based on the variance calculation of the frequency offset in the continuous adjacent time interval, by calculating the second-order difference statistics of the frequency change at different sampling times τ, the type and characteristics of the frequency noise are reflected. And the MTIE is an index for describing the maximum change range of the clock time error (time interval error), which reflects the maximum time deviation amplitude of the clock within a given observation window, and the maximum peak-valley difference value of the measured time error is calculated by the sliding window statistics, which is used to evaluate the maximum time jitter or deviation of the system. The Allan variance is used to evaluate the random fluctuation and stability characteristics of the clock frequency, and focuses on analyzing the type of frequency noise and the frequency stability; while the MTIE focuses on the maximum change range of the clock time error, and emphasizes the maximum time deviation of the synchronization system within a certain time window, which is used to evaluate the extreme time drift and synchronization quality of the system. Both of them evaluate the clock performance from two different angles of frequency fluctuation and maximum amplitude of time error. The person skilled in the art can select according to the actual needs.

[0041] Specifically, the frequency stability factor is used to reflect the "smoothness" and "predictability" of the clock frequency change, avoid the increase of synchronization error caused by frequency mutation or severe jitter, and also be called frequency drift rate.

[0042] Further, the step S2 specifically comprises the following sub-steps:

[0043] S201. Constructing a confidence function according to the time-frequency dynamic feature vector, measuring the time-frequency quality and credibility of the reference source at the current time, and outputting the confidence of the reference source;

[0044] S202. Mapping the confidence of the reference source to a relative evaluation model, and selecting the reference source with the highest confidence as the main synchronization reference source, and taking the main synchronization reference source as the reference input of the control error modeling;

[0045] S203. Calculating the derivative term of the corresponding reference source according to the main synchronization reference source and the local clock, and constructing a control error vector combining the instantaneous synchronization deviation and frequency offset of the corresponding reference source.

[0046] Specifically, the above embodiment first takes the time-frequency dynamic feature vector of each reference source constructed in S1 as input, designs a confidence function containing weight factor, fluctuation sensitive factor and trend response mechanism, measures the performance of each reference source in terms of synchronization quality, stability and credibility at the current time from multiple dimensions, and outputs the confidence score of each reference source; in the relative evaluation model, the confidence values of all reference sources are normalized to build a ranking mechanism, and the reference source with the highest confidence at the current time is automatically selected as the main synchronization reference source, ensuring that the system always tracks the signal source with the best quality. After selecting the main reference source, the state derivative (i.e. frequency change rate or trend term) of the local clock is calculated based on the state of the main reference source and the local clock, and a control error vector is constructed combining the current synchronization deviation and frequency offset of the local clock. The vector accurately reflects the deviation between the current synchronization state and the target state, and is the core input of the control tuning process.

[0047] Further, in the step S201, the confidence function is specifically represented as:

[0048]

[0049] Wherein, the W i (t) represents the confidence score of the i-th reference source, and the a1, a2, a3 represent the weight coefficients corresponding to the instantaneous synchronization deviation, the instantaneous synchronization deviation, and the frequency stability factor, respectively. respectively represent the nonlinear mapping functions corresponding to the instantaneous synchronization deviation, the instantaneous synchronization deviation, and the frequency stability factor, respectively, for transforming each error value into a confidence metric value, and the Δt i(t) represents the instantaneous synchronization deviation of the i-th reference source at the current time, and the Δf i (t) represents the frequency offset of the i-th reference source at the current time, and the S i (t) represents the frequency stability factor of the i-th reference source.

[0050] Further, in the step S202, the confidence of the reference source is uniformly mapped to the relative evaluation model, and the specific process is as follows: the confidence of the reference source is normalized, that is,

[0051]

[0052] Wherein, the represents the normalized confidence value of the i-th reference source, the n represents the total number of calculated reference source confidence, and the j represents the index of the calculated reference source confidence.

[0053] Further, the step S3 specifically includes the following sub-steps:

[0054] S301. The error vector is taken as the model input, and a bionic mechanism is introduced to simulate the neuron response mechanism to construct an activation model. The model converts the error information into the response strength of the system through a nonlinear response function, and outputs the activation value corresponding to the error vector;

[0055] S302. The activation value is mapped to the adaptive gain coefficient of the controller, and the instantaneous time difference and the historical integral error are respectively applied to the proportional gain and the integral gain to generate a control signal for frequency tuning;

[0056] S303. The generated control signal is transmitted to the voltage tuning port of the local VCXO or OCXO oscillator, and the output frequency of the oscillator is adjusted according to the control signal, so as to gradually reduce the instantaneous synchronization deviation and the frequency offset between the local clock and the main reference source, and continuously output the tuning voltage.

[0057] Further, in the step S303, the instantaneous synchronization deviation, the frequency offset and the derivative term between the oscillator and the main synchronization reference source are also monitored. When the instantaneous synchronization deviation, the frequency offset and the derivative term are all within the set convergence threshold in a fixed time window, it indicates that the system is in a frequency domestication stable state at this time, and the parameter information of the current time is returned.

[0058] Specifically, the above embodiment introduces the neuron activation mechanism in bionics to convert the control error vector into the controller response strength. The error vector is first input into a nonlinear activation model to simulate the response process of neurons under stimulation, and the typical form is sigmoidal, ReLU or Gaussian activation function, and the output represents the activation value of the response strength. This activation value is then used to map the adaptive gain coefficient of the controller, where the instantaneous synchronization deviation is adjusted through the proportional channel, and the integral component of the frequency offset is adjusted through the integral channel, which together constitute the PI control strategy of the controller. Finally, according to the gain coefficient and the error, a voltage control signal is generated, which is sent to the tuning end of the VCXO or OCXO oscillator to realize fine tuning of the local output frequency, drive it to gradually lock to the frequency of the main reference source, and form a closed-loop frequency tracking mechanism. In addition, the synchronization deviation, frequency offset and derivative term between the main synchronization reference source are continuously monitored during the control process, and if these indicators are less than the convergence threshold within the set time window, it is determined that the system enters the frequency taming stable state, and the parameter information at the current time is returned in real time, including the control error vector, the tuning voltage, the current gain coefficient and the synchronization state identifier, to provide basic data for frequency health assessment.

[0059] Further, in the step S302, the instantaneous time difference and the historical integral error are respectively applied to the proportional gain and the integral gain to generate a control signal for frequency tuning, and the specific process includes:

[0060] S3021. Initial controller proportional gain and integral gain

[0061] S3022. According to the response vector, the adaptive adjustment is as follows:

[0062]

[0063] Wherein, the K p (t) represents the proportional control gain at time t, the K i (t) represents the integral control gain at time t, the κ p represents the amplification adjustment coefficient of the proportional gain, the κ i represents the amplification adjustment coefficient of the integral gain, and the Θ1(t), Θ2(t) respectively represent the neural activation response output related to the error vector for adjusting the gain.

[0064] Further, the specific generation process of the tuning voltage in the step S303 is as follows: the adjusted gain is applied to the synchronization error to generate the tuning voltage, that is: Wherein, the V tune (t) represents the voltage tuning control signal output at time t, in volts, and the represents the instantaneous synchronization deviation between the master synchronization reference source and the local clock at time t, in seconds or nanoseconds, which is set according to the actual situation, and the represents the instantaneous time deviation between the master synchronization reference source i * and the local clock at a certain time τ in history, and the represents the integral of the master reference source synchronization deviation from the system startup to the current time.

[0065] Further, in step S303, adjusting the own output frequency according to the control signal is specifically represented as: f osc (t) = f0+ β·V tune (t); the gradual reduction of the instantaneous synchronization deviation and the frequency offset between the local clock and the master reference source is specifically represented as: and wherein, the f osc (t) represents the frequency output by the local voltage-controlled oscillator VCXO or OCXO at time t, the f0represents the free-running frequency of the oscillator, the β represents the tuning sensitivity coefficient of the oscillator, in hertz per volt, and the V tune (t) represents the voltage tuning control signal output at time t, in volts, and the lim t→∞ represents the limit when time tends to infinity, which is used to describe the final behavior of the system after a long time, i.e., whether to reach a stable state, and the represents the frequency of the i * th selected master synchronization reference source at time t, in hertz.

[0066] Further, the parameter information includes the error vector, the tuning voltage, the adaptive gain coefficient, and the system state, and the system state includes the frequency taming stable state and the frequency taming unstable state.

[0067] Further, in step S301, the error vector is taken as the model input, and an activation model is constructed by combining the introduction of a bionic mechanism to simulate the neuron response mechanism, and the specific process is represented as:

[0068] Θ(t) = σ(W·E(t) + b);

[0069] wherein, the Θ(t) represents the system response value after the neural activation, i.e., the activation value, the W represents the weight coefficient of the error vector, the σ represents the activation function, the E(t) represents the error vector, and the b represents the bias term for adjusting the center deviation of the neural response.

[0070] Further, the step S4 specifically includes the following sub-steps:

[0071] S401. Constructing a state vector of the current time according to the instantaneous synchronization deviation, the frequency offset, the frequency stability factor, the derivative term and the tuning voltage;

[0072] S402. Defining a comprehensive weighted health degree function according to the state vector of the current time and calculating the corresponding health degree;

[0073] S403. Classifying the calculated health degree according to the set health degree classification interval.

[0074] Further, the weighted health degree function in step S402 is specifically represented as:

[0075]

[0076] Specifically, the above embodiment constructs a unified state vector set based on the state variables of each link in the system operation process, and defines a frequency health degree function based thereon to quantitatively evaluate the time-frequency stability and energy efficiency state of the system. The state vector includes multi-dimensional quantities such as synchronization deviation, frequency offset, frequency derivative, tuning voltage, tuning voltage change rate, adaptive gain value, etc., which comprehensively reflect the operation characteristics of the system. These state parameters are substituted into the weighted health degree function, the function form is a polynomial combination or a linear weighting structure, and a scalar index representing the current health state of the local clock system is output. The lower the value, the healthier the system, indicating that the frequency is stable, the synchronization is tight, and the control is smooth; otherwise, it indicates that the system is in an unstable or disordered state. The health degree function value is further mapped to different level labels such as "excellent", "normal", "warning" and "fault", which are used as self-diagnosis signals of system operation, for control strategy switching, reference source reselection, energy efficiency mode adjustment and other upper logic calls. The evaluation value and the state label are written into the sliding window history sequence in real time, constituting a time sequence tracking mechanism, which assists in identifying the trend deviation and energy consumption anomaly under long-time operation, and provides key support for frequency taming and energy efficiency optimization of the entire clock system.

[0077] wherein the Q f (t) represents the frequency health degree of the system at the current time, w1, w2, w3, w4 and w5 represent the weight coefficients corresponding to the instantaneous synchronization deviation, the frequency offset, the frequency stability factor, the derivative term and the tuning voltage respectively, ε(t) represents the instantaneous synchronization deviation between the main reference source and the local clock, Δf(t) represents the current frequency offset, f'(t) represents the derivative term, σ(t) represents the current frequency stability factor, V(t) represents the tuning voltage, and represents the derivative term, σ f (t) represents the current frequency stability factor, V tune (t) represents the tuning voltage, and represents the change rate of the tuning voltage.

[0078] Embodiment Two

[0079] Further, as a preferred embodiment of the above-mentioned embodiment, for step S403, an exemplary setting mode is proposed:

[0080] When Q f (t) < θ1, it indicates that the state label of the system frequency at this moment is Excellent, i.e., the system frequency control is extremely stable, and all indicators are in the optimal interval;

[0081] When θ1≤Q f (t) < θ2, it indicates that the state label of the system frequency at this moment is Normal, i.e., the system stability is good, there is a slight deviation but within the control threshold range;

[0082] When θ2≤Q f (t) < θ3, it indicates that the state label of the system frequency at this moment is Degraded, i.e., the system frequency drift is slightly severe, and the tuning system has obvious instability;

[0083] When Q f (t) ≥ θ3, it indicates that the state label of the system frequency at this moment is Unhealthy, i.e., abnormal states such as synchronization loss, reference source failure, and serious control fluctuation occur.

[0084] Further, as a preferred embodiment of the above-mentioned embodiment, as a reference for whether the system continues to maintain the current control strategy, when the state label is Degraded or Unhealthy, a feedback process can be triggered:

[0085] Re-evaluate the reference source confidence (back to S2);

[0086] Start the controller gain retuning process (re-enter S301-S302 of S3);

[0087] Switch the reference source or enter the degraded operation mode in a serious state.

[0088] Further, as a preferred embodiment of the above-mentioned embodiment, a multi-reference clock synchronization control system integrating frequency taming and energy efficiency optimization is proposed, which specifically includes:

[0089] A signal acquisition and dynamic feature extraction module for receiving multiple heterogeneous clock signals and extracting time-frequency dynamic features;

[0090] A reference source confidence evaluation and control error modeling module for constructing a reference source confidence function, selecting a master synchronization reference source, and constructing a control error vector;

[0091] An adaptive activation control and frequency tuning module for adjusting the controller gain combined with a bionic neural activation mechanism, outputting a tuning control signal, and realizing frequency tracking and locking;

[0092] A frequency health comprehensive modeling and evaluation module is configured to comprehensively model the state vectors of each link, calculate a frequency health function, and perform health evaluation.

[0093] Further, the signal acquisition and dynamic feature extraction module includes:

[0094] A multi-path heterogeneous clock signal receiving and timestamp conversion unit is configured to receive 1PPS, PTP, B code, and TOD clock signals, and convert them into a digital timestamp signal stream.

[0095] A time difference analysis and instantaneous synchronization deviation and frequency offset extraction unit is configured to calculate instantaneous synchronization deviation and frequency offset based on a sliding window.

[0096] A frequency stability factor calculation unit is configured to calculate a frequency stability factor.

[0097] A time-frequency dynamic feature vector combination unit is configured to combine the above three dynamic features into a time-frequency dynamic feature vector.

[0098] Further, the reference source confidence evaluation and control error modeling module includes:

[0099] A confidence function construction unit is configured to calculate reference source confidence based on the time-frequency dynamic feature vector.

[0100] A confidence mapping and primary synchronization reference source selection unit is configured to select the reference source with the highest confidence as the primary synchronization reference source.

[0101] A control error vector construction unit is configured to calculate derivative terms and combine instantaneous synchronization deviation and frequency offset to construct a control error vector.

[0102] Further, the adaptive activation control and frequency tuning module includes:

[0103] An activation model construction and response calculation unit is configured to input the error vector into a biomimetic activation model and calculate an activation value.

[0104] A controller gain automatic adjustment unit is configured to map the activation value into proportional gain and integral gain coefficients to achieve adaptive adjustment.

[0105] An oscillator voltage control and frequency tuning unit is configured to output a tuning voltage signal to adjust the oscillator frequency, achieve frequency tracking, monitor frequency taming state, and return error vector, tuning voltage, adaptive gain, and system state information.

[0106] Further, the frequency health comprehensive modeling and evaluation module includes:

[0107] A state vector fusion unit is configured to collect and fuse state parameters of each link; and a frequency health degree function definition unit is configured to calculate a frequency health degree function by weighting;

[0108] A health grade division unit is configured to divide health grades according to the health degree function;

[0109] A health degree evaluation feedback unit is configured to feed back health state information and trigger a corresponding control strategy;

[0110] A historical health state management unit is configured to record and analyze historical health degree data, and realize trend monitoring and prediction.

[0111] Through the above functional modules, the system can realize real-time synchronous acquisition and dynamic feature extraction of multiple heterogeneous external reference clock signals, accurately evaluate the time-frequency quality and confidence of each reference source, intelligently select the optimal master synchronization reference source, and adaptively adjust the controller gain based on the bionic neural activation model to generate an accurate voltage control signal to drive the local oscillator to realize high-precision frequency tracking and locking. At the same time, the system can dynamically monitor the synchronization error state between the oscillator and the reference source, determine the frequency domestication stability, model the state parameters of each link in real time, calculate the frequency health degree, accurately evaluate the system operation health status, guarantee the stability and reliability of clock synchronization, and realize energy efficiency optimization and fault warning through health degree feedback to improve the overall performance and robustness of the system.

[0112] The above only describes the preferred embodiments of the present application, and it should be understood that the present application is not limited to the forms disclosed herein, and should not be considered as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the concepts described herein by the above teachings or related art or knowledge. Any modification and change made by those skilled in the art without departing from the spirit and scope of the present application shall be within the protection scope of the appended claims of the present application.

Claims

1. A multi-reference clock synchronization control method integrating frequency taming and energy efficiency optimization, characterized in that: The following steps are involved: S1. Synchronously collect time data from multiple external reference signal inputs, digitize each reference signal, and extract the dynamic characteristics of each reference signal in combination with a time difference analysis mechanism. The dynamic characteristics include instantaneous synchronization deviation, frequency offset, and frequency stability factor. The dynamic characteristics are combined to form a time-frequency dynamic feature vector for each reference source; S2. Based on the time-frequency dynamic feature vector and the confidence function, a reference source confidence assessment model is constructed. The reference source with the highest confidence at the current moment is selected as the primary synchronization reference source. The derivative term of the corresponding reference source is calculated based on the primary synchronization reference source. The control error vector is then constructed by combining the instantaneous synchronization deviation and frequency offset of the corresponding reference source. S3. Based on the bionic mechanism of the activation model and the error vector, the controller gain parameters are automatically adjusted and a voltage control signal is output based on the adjusted gain parameters. The oscillator output frequency is tuned using the voltage control signal, a tuning voltage is output, and the current parameter information is returned. S4. Comprehensively model the state vectors of each link, define the frequency health function by weighting, and perform health assessment based on the calculated function value.

2. The multi-reference clock synchronization control method integrating frequency taming and energy efficiency optimization according to claim 1, characterized in that: The step S1 specifically includes the following sub-steps: S101 receives multiple heterogeneous clock reference signals including 1PPS, PTP, B code and TOD, and converts them into a digital timestamp signal stream; S102. For each reference source, perform time difference analysis to extract the instantaneous offset between the local clock and the reference source; calculate the instantaneous frequency offset of the reference source by using the continuous time difference data within the sliding window; S103. In the same window, based on multiple instantaneous offsets and / or instantaneous frequency offsets, calculate the frequency stability factor of the reference source; S104. Combining the three dynamic eigenvalues ​​extracted from each reference source into a set of time-frequency dynamic eigenvectors as output.

3. The multi-reference clock synchronization control method integrating frequency taming and energy efficiency optimization according to claim 2, characterized in that: In step S103, when the frequency stability factor of the reference source is calculated using multiple instantaneous frequency offsets, the Allan variance is used for calculation. The specific process is as follows: Among them, the S i (t) represents the frequency stability factor of the i-th reference source at time t, represents the Allan variance, τ represents the observation time interval, N represents the number of sampling points, k represents the time index variable, represents the frequency offset of the i-th reference source at the k-th sampling moment, Indicates the frequency offset of the i-th reference source at the k+1-th sampling time.

4. The multi-reference clock synchronization control method integrating frequency taming and energy efficiency optimization according to claim 2, characterized in that: In step S103, when the frequency stability factor of the reference source is calculated using multiple instantaneous offsets, the MTIE variance is used for calculation. The specific process is as follows: Among them, the S i (t) represents the frequency stability factor of the i-th reference source at time t, k represents the time index variable, It represents the instantaneous time difference between the i-th reference source and the local clock at the k-th sampling time. The mean(Δt i ) represents the mean time difference of the reference source, the max k Indicates the maximum instantaneous offset value within time window k.

5. The multi-reference clock synchronization control method integrating frequency taming and energy efficiency optimization according to claim 1, characterized in that: The step S2 specifically includes the following sub-steps: S201. Constructing a confidence function based on the time-frequency dynamic feature vector, measuring the time-frequency quality and credibility of the reference source at the current moment, and outputting the confidence of the reference source; S202. The confidence levels of the reference sources are uniformly mapped to the relative evaluation model, and the reference source with the highest confidence level is selected as the main synchronization reference source, and the main synchronization reference source is used as the reference input for the control error modeling; S203. Calculate the derivative term of the corresponding reference source based on the master synchronization reference source and the local clock, and construct a control error vector in combination with the instantaneous synchronization deviation and frequency offset of the corresponding reference source.

6. The multi-reference clock synchronization control method integrating frequency taming and energy efficiency optimization according to claim 5, characterized in that: In step S201, the confidence function is specifically expressed as: Wherein, the W i (t) represents the confidence score of the i-th reference source, α1, α2, and α3 represent the weight coefficients corresponding to the instantaneous synchronization deviation, the instantaneous synchronization deviation, and the frequency stability factor, respectively. Respectively represent the nonlinear mapping functions corresponding to the instantaneous synchronization deviation, the instantaneous synchronization deviation, and the frequency stability factor, which are used to transform each error value into a confidence metric value. The Δt i (t) represents the instantaneous synchronization deviation of the i-th reference source at the current moment, and the Δf i (k) represents the frequency offset of the i-th reference source at the current moment, and the S i (t) represents the frequency stability factor of the i-th reference source.

7. The multi-reference clock synchronization control method integrating frequency taming and energy efficiency optimization according to claim 5, characterized in that: In step S202, the confidence of the reference source is uniformly mapped to the relative evaluation model. The specific process is: the confidence of the reference source is normalized, that is: Among them, the represents the normalized confidence value of the i-th reference source, n represents the total number of calculated reference source confidences, and j represents the index of the calculated reference source confidence.

8. The multi-reference clock synchronization control method integrating frequency taming and energy efficiency optimization according to claim 1, characterized in that: The step S3 specifically includes the following sub-steps: S301. The error vector is used as the model input, and an activation model is constructed by introducing a bionic mechanism to simulate the neuron response mechanism. The model converts the error information into the system response intensity through a nonlinear response function and outputs the activation value corresponding to the error vector; S302. Mapping the activation value to the adaptive gain coefficient of the controller, and applying the instantaneous time difference and the historical integral error to the proportional gain and integral gain respectively, to generate a control signal for frequency tuning; S303. The generated control signal is transmitted to the voltage tuning port of the local VCXO or OCXO oscillator, and its own output frequency is adjusted according to the control signal, gradually reducing the instantaneous synchronization deviation and frequency offset between the local clock and the main reference source, and continuously outputting the tuning voltage.

9. The multi-reference clock synchronization control method integrating frequency taming and energy efficiency optimization according to claim 8, characterized in that: Step S303 also includes monitoring the instantaneous synchronization deviation, frequency offset, and derivative term between the oscillator and the main synchronization reference source. When, within a fixed time window, the instantaneous synchronization deviation, frequency offset, and derivative term are all within the set convergence threshold, it indicates that the system is in a frequency taming stable state at that moment, and the parameter information of the current moment is returned.

10. The multi-reference clock synchronization control method integrating frequency taming and energy efficiency optimization according to claim 9, characterized in that: The parameter information includes an error vector, a tuning voltage, an adaptive gain coefficient, and a system state, and the system state includes a frequency taming stable state and a frequency taming unstable state.

11. The multi-reference clock synchronization control method integrating frequency taming and energy efficiency optimization according to claim 1, characterized in that: In step S301, the error vector is used as the model input, and the activation model is constructed by introducing a bionic mechanism to simulate the neuron response mechanism. The specific process is as follows: Θ(t)=σ(W·E(t)+b); Wherein, Θ(t) represents the system response value after neural activation, that is, the activation value, W represents the weight coefficient of the error vector, σ represents the activation function, E(t) represents the error vector, and b represents the bias term, which is used to adjust the center offset of the neural response.

12. The multi-reference clock synchronization control method integrating frequency taming and energy efficiency optimization according to claim 1, characterized in that: The step S4 specifically includes the following sub-steps: S401. Construct the current state vector based on the instantaneous synchronization deviation, frequency offset, frequency stability factor, derivative term and tuning voltage; S402. Define a comprehensive weighted health function based on the current state vector and calculate the corresponding health; S403. Classify the calculated health level according to the set health level classification interval.

13. The multi-reference clock synchronization control method integrating frequency taming and energy efficiency optimization according to claim 1, characterized in that: In step S402, the weighted health function is specifically expressed as: Among them, the Q f (t) represents the frequency health of the system at the current moment, w1, w2, w3, w4, and w5 represent the weight coefficients corresponding to the instantaneous synchronization deviation, frequency offset, frequency stability factor, derivative term, and tuning voltage, respectively, ε(t) represents the instantaneous synchronization deviation between the main reference source and the local clock, Δf(t) represents the current frequency offset, and represents the derivative term, the σ f (t) represents the current frequency stability factor, the V tune (t) represents the tuning voltage, the Indicates the rate of change of the tuning voltage.

Images