Intelligent modulator management system and method based on Internet of Things

Abstract

The invention discloses an intelligent modulator management system and method based on the Internet of Things, and relates to the technical field of optical communication, and the method comprises the steps: collecting operation data in the operation process of a modulator, and building a dynamically updated data window based on a time sequence; performing difference calculation on the operation data at the current moment and the operation data in the data windows, extracting data with differences in a plurality of data windows, and predicting a modulation depth prediction value of the modulator at the next moment according to the extracted data; recursively generating a plurality of predicted values at future moments based on the modulation depth predicted value at the next moment, and constructing a modulation prediction sequence; comparing the modulation prediction sequence with a modulation data target value, and calculating to obtain a disturbance index for measuring the stability of the system; and calculating an adjustment value for controlling the modulation depth based on the disturbance index and the modulation depth predicted value at the next moment.

Description

Technical Field

[0001] The present invention relates to the field of optical communication technology, and in particular to an Internet of Things-based modulator intelligent management system and method. Background Art

[0002] In modern optical communication systems, optical modulators are key components whose stable operation directly impacts the system's signal quality and communication stability. Precise control of modulation depth is crucial for ensuring signal quality, especially in fields such as optical communications and spectral analysis. Fluctuations in modulation depth or deviations from the target can cause signal distortion, impacting system performance. Real-time monitoring of modulation depth trends and precise adjustments can effectively ensure system stability and signal quality. Currently, optical modulator control methods primarily rely on traditional feedback control algorithms, such as PID control and fuzzy control. While these methods can achieve certain regulation effects, they suffer from slow response, low precision, and poor adaptability in dynamically changing environments. Traditional methods rely on fixed regulation rules and struggle to respond to changes in device operating conditions in real time. In particular, they are prone to inaccurate or lagging regulation in situations with significant environmental disturbances. Therefore, there is an urgent need for an intelligent management system and method for modulators based on the Internet of Things to solve the above problems. Summary of the Invention

[0003] The purpose of the present invention is to provide a modulator intelligent management system and method based on the Internet of Things to solve the problems raised in the prior art.

[0004] To achieve the above object, the present invention provides the following technical solution: a modulator intelligent management method based on the Internet of Things, the method comprising the following steps: S100 collects operating data during the operation of the modulator and establishes a dynamically updated data window based on the time series; Furthermore, the specific steps for establishing a dynamically updated data window based on time series are as follows: S101 starts from the initial sampling time t1 and continuously collects the modulation depth data M(t), driving voltage data V(t) and chip temperature data T(t) of the modulator during operation at n moments. The sampling frequency of collecting the modulation depth, driving voltage and chip temperature should meet the Nyquist sampling theorem. The operation data at each sampling time ti are collected, i∈[1,n], and an operation data set Pti={M(ti),V(ti),T(ti)} is constructed, where M(ti), V(ti) and T(ti) respectively represent the modulation depth data, driving voltage data and chip temperature data at the sampling time ti. The operation data set is stored according to the acquisition timestamp to construct a time series data set P={Pt1, Pt2, Pt3, ..., Ptn}, where Pt1, Pt2, Pt3, ... and Ptn respectively represent the operation data sets constructed at the 1st, 2nd, 3rd, ... and nth acquisition timestamps, and tn represents the current real-time sampling time. The modulator is an interdigitated segmented silicon-based modulator consisting of multiple independently adjustable modulation segments. Each modulation segment contains an array of interdigitated electrodes, with the electrodes arranged orthogonally between adjacent segments. Each modulation segment is physically separated by an isolation slot. The modulation units are arranged along the optical waveguide structure, and each segment adjusts the phase of the optical signal under voltage control. The modulator adopts an adjustable segmented design, dynamically controlling the activation state of each segment to achieve different bit outputs, supporting multi-channel independent modulation. Each segment is equipped with an RC matching network to ensure optimized electric field distribution and reduce losses. By adopting a distributed interdigital structure and a discrete phase modulation unit in a collaborative design, the technical bottlenecks of the traditional continuous modulation area in terms of process tolerance and control accuracy are broken through while maintaining high-speed modulation characteristics. On the input-to-output path, due to processing defects, the doping gain of each segment is not exactly the same. By combining the interdigital segmented modulation structure, a segmented phase modulator with reconfigurable working state is realized. The phase difference can be changed directly through drive control, crosstalk can be suppressed, and modulation linearity can be optimized. The modulator has multiple configurable modulation segments, and the modulation capability of each segment is independent and adjustable. The modulation units can work independently, or be connected in series or parallel in a set manner to form a multi-segment modulation structure. The multi-segment modulation units are independently controlled by the electro-optical phase modulator to achieve high-resolution dynamic adjustment of the refractive index. It has dynamic response characteristics and does not require the introduction of a phase compensation zone. The process error is dispersed to all segments, improving robustness. S102 constructs a sliding window Wt of length n. The length n of the sliding window can be configured based on the response time of the modulator, computing resources, and target accuracy. To ensure real-time performance and accuracy, the update frequency of the sliding window can be set to the same as the data acquisition frequency. The structure of the sliding window is defined as Wt={Pt1, Pt2, Pt3, ..., Ptn}. When the operating data set Ptn+1 at time tn+1 is collected, Ptn+1 is inserted at the end of the sliding window, and the operating data set at time t1 is removed to construct a new sliding window. Through the sliding window mechanism, this solution ensures real-time data collection and updates, significantly improving the system's real-time performance, response speed, prediction accuracy, and control stability. Furthermore, the sliding window's adaptive nature effectively adapts to environmental changes, reduces the impact of outdated data, and optimizes the system's computational efficiency and stability. S200 calculates the difference between the current operating data and the operating data in the data window, extracts data in a plurality of data windows with differences, and predicts the modulation depth prediction value of the modulator at the next moment based on the extracted data; Furthermore, the specific steps for predicting the modulation depth prediction value at the next moment are: S201 calculates the driving voltage change ΔVn at time tn and the chip temperature change ΔTn at time tn. The calculation formula is: , Construct the state feature vector S(tn) at the reaction time tn, expressed as S(tn)=[M(tn),M(tn-1),ΔVn,ΔTn]; S202 constructs the state feature vector S(tj) at the time tj in the sliding window Wt, j∈[1,n-1], and calculates the difference dj(tn) between the state at time tn and the state at time tj. The calculation formula is: , Where ω1 represents the weight of the driving voltage change, and ω2 represents the weight of the chip temperature change. The weight coefficients ω1 and ω2 are adjusted according to the actual working conditions to reflect the relative importance of the voltage change and temperature change on the system. The differences between the state characteristics from time t1 to tn-1 and the state characteristics at time tn are collected to construct the difference set D={d1(tn), d2(tn), d3(tn), ..., dn-1(tn)}, where d1(tn), d2(tn), d3(tn), ..., and dn-1(tn) represent the differences between time t1, t2, t3, ..., and tn-1 and time tn, respectively. S203 sorts the differences in set D in a forward order and extracts the first k difference values. The value of k is recommended to be 5% to 10% of n, balancing computational complexity and sample diversity. A screening difference value set D*={d1,d2,d3,...,dk} is constructed, where d1,d2,d3,... and dk represent the first, second, third,... and kth difference values ​​extracted, respectively. The linear interpolation weight ωdx of the xth difference value dx is calculated using the following formula: ; S204 calculates the corresponding modulation depth data change ΔMx for the state feature vector S(tx) corresponding to the x-th difference value at time tx, using the following formula: , The modulation depth prediction value M^(tn+1) at time tn+1 is calculated based on the modulation depth data change at the time corresponding to the difference value in the set D* and the corresponding linear interpolation weight. The calculation formula is: ; Through difference calculation and dynamic weighted prediction, the accuracy of modulation depth prediction and control stability are effectively improved. By comparing historical data with real-time data, combined with linear interpolation and weighting strategies, the system's adaptability to changing environments is optimized. At the same time, difference screening ensures computational efficiency. The combined application of these technical means enables the optical modulator control system to maintain efficient and stable operation under complex and dynamically changing operating conditions. S300 recursively generates predicted values ​​at several future moments based on the modulation depth prediction value at the next moment and the latest modulation change direction, thereby constructing a modulation prediction sequence. Furthermore, the specific steps of constructing the modulation prediction sequence are: S301 generates the modulation depth prediction value corresponding to the next q moments based on the modulation depth prediction value at time tn+1 through a recursive formula. For the modulation depth prediction value M^(tn+g) corresponding to the g-th moment in the future, g∈[2,q], the recursive process is: , Where M^(tn+g-1) represents the modulation depth prediction value corresponding to the g-1th moment in the future, and α is the trend inference coefficient. The trend inference coefficient needs to be selected according to the dynamic characteristics of the system. A higher value can make the prediction more sensitive to recent changes, while a lower value can improve the stability of the prediction and avoid over-response to short-term fluctuations. It can be set through experiments or online optimization. S302 collects the modulation depth prediction values ​​corresponding to q future moments and constructs a prediction sequence M^={M^(tn+1),M^(tn+2),M^(tn+3),...,M^(tn+q)}, where M^(tn+1), M^(tn+2), M^(tn+3),... and M^(tn+q) represent the modulation depth prediction values ​​corresponding to the 1st, 2nd, 3rd,...th and qth future moments, respectively; The predicted modulation depth at future moments is used to generate a prediction sequence for multiple moments. The recursive process can effectively capture the changing trend of the modulation depth. The recursive formula can combine the trends of historical prediction values ​​and current prediction values ​​to generate a multi-step prediction sequence, ensuring that the system can identify and adjust the modulation depth at future moments in advance. This recursive process can capture the trend changes of the system at different time scales, providing an accurate prediction basis for subsequent voltage regulation and disturbance assessment. S400 compares the modulation prediction sequence with the modulation data target value and calculates the disturbance index to measure the system stability; Furthermore, the specific steps to obtain the disturbance index for measuring system stability are as follows: S401 collects actual modulation depth data for the next q moments, and calculates the deviation ΔM^ between the modulation depth prediction value at the gth future moment and the modulation depth target value M* for the next q moments. The calculation formula is: , Aggregate the deviation values ​​of the next q moments to construct an error value set W={ΔM^1,ΔM^2,ΔM^3,...,ΔM^q}, where ΔM^1,ΔM^2,ΔM^3,... and ΔM^q represent the deviation values ​​between the modulation depth prediction value and the modulation depth data target value at the 1st, 2nd, 3rd,... and qth moments in the future, respectively. Extract the maximum value of the data in the set W as the maximum deviation value ΔMmax; S402 calculates the cumulative error Esun between the modulation depth prediction value and the actual modulation depth data deviation value at the next q moments, and the calculation formula is: ; S403 calculates the trend change ΔMtrend between the modulation depth prediction value and the modulation depth data target value at the qth moment in the future. The calculation formula is: ; S404 constructs a comprehensive disturbance index Dn based on the maximum deviation value, cumulative error and trend change. The construction formula is: , Among them, β1, β2, and β3 are weight coefficients. β1, β2, and β3 can be configured according to the requirements of system design and application scenarios. Generally, the maximum deviation ΔMmax has a larger weight because it directly reflects the difference between the system modulation depth and the target value, which may lead to system instability. By accurately calculating the modulation depth deviation, cumulative error, and trend change at future moments, and combining factors such as maximum deviation, cumulative error, and trend change, a comprehensive disturbance index is constructed to effectively evaluate the stability and adaptability of the system. By flexibly configuring weight coefficients, the system can dynamically adjust the control strategy according to different operating conditions, identify and respond to potential disturbances and errors in advance, and ensure that the modulation depth always approaches the target value, thereby improving the system's robustness, accuracy, and adaptability. S500 calculates an adjustment value for controlling the modulation depth based on the disturbance index and the modulation depth prediction value at the next moment; Furthermore, the specific steps for calculating the adjustment value for controlling the modulation depth are: S501 calculates the response gain value Kt of the driving voltage data to the modulation depth data based on the modulation depth data at time tn and the predicted value of the modulation depth data at time tn+1 in the future, combined with the driving voltage data at time tn. The calculation formula is: , Where ε is a constant to prevent zero minima; S502 calculates the driving voltage data adjustment value ΔVd that needs to be applied based on the trend change amount, the comprehensive disturbance index and the response gain. The calculation formula is: ; In order to ensure the system's adaptability to different environmental changes, the voltage regulation amount will be adaptively adjusted according to the real-time disturbance index Dn. In the case of high disturbance, the regulation amount may increase to quickly restore system stability; in the case of low disturbance, the regulation amount will decrease to avoid over-adjustment and waste of energy. Extract the maximum rated driving voltage Vmax and the minimum rated driving voltage Vmin of the modulator, and set the adjustment range of the driving voltage data adjustment amount to Vmin≤V(tn)+ΔVd≤Vmax; S503 calculates the final output drive voltage adjustment value V(tn+1) at the next tn+1th moment based on the drive voltage data adjustment value, and the calculation formula is: , Where Vmax represents the maximum rated drive voltage of the modulator, Vmin represents the minimum rated drive voltage of the modulator, and the clip function indicates that the final output drive voltage is between the maximum rated drive voltage and the minimum rated drive voltage; Through adaptive voltage regulation and disturbance assessment, it achieves fast and stable recovery under different environmental changes, optimizes the system's energy efficiency and response speed, and ensures that the drive voltage is always maintained within the rated range, avoiding over-adjustment and energy waste; In order to better implement the above method, a modulator intelligent management system based on the Internet of Things is proposed. The system includes a data acquisition module, a state prediction module, a sequence recursion module, a disturbance assessment module and a regulation module. The data acquisition module is used to collect the operating data of the modulator during operation and establish a dynamically updated data window; The data acquisition unit includes a data acquisition unit and a window management unit; The data acquisition unit is used to collect the operating data of modulation depth, driving voltage and chip temperature and store them according to the timestamp; The window management unit is used to insert new data and remove old data through a sliding window mechanism to build a dynamically updated data window; The state prediction module is used to predict the modulation depth prediction value at the next moment based on the data in the data window that is significantly different from the real-time data; The state prediction module includes a feature calculation unit and a prediction generation unit; The feature calculation unit is used to calculate the voltage and temperature changes, construct the state feature vector and calculate the difference value with the historical state; The prediction generation unit is used to screen similar historical states and calculate the next moment prediction value based on the weighted modulation depth change; The sequence recursion module is used to recursively generate prediction values ​​for several future moments and construct a modulation prediction sequence; The disturbance assessment module is used to calculate the disturbance index to measure the stability of the system; The adjustment module is used to calculate the adjustment value for controlling the modulation depth and output it to the modulator; The adjustment module includes a gain calculation unit and an adjustment control unit; The gain calculation unit is used to calculate the response gain of the driving voltage to the modulation depth; The regulation control unit is used to calculate the voltage regulation amount based on the disturbance index and the predicted deviation, and output the final voltage through the safety range constraint.

[0005] Compared with the prior art, the present invention has the following beneficial effects: 1. Through real-time disturbance feedback regulation and adaptive voltage regulation, the system can quickly recover stability under high disturbance conditions and avoid over-regulation under low disturbance conditions, thereby improving the system's stability and responsiveness and ensuring that the modulator can operate efficiently under various environmental conditions; 2. Dynamically adjust the voltage, especially reducing the adjustment amount in low-disturbance conditions, to avoid over-adjustment and thus avoid unnecessary energy waste. This ensures the performance of the entire system while significantly improving energy efficiency and extending the service life of the equipment. 3. By combining historical data with future forecasts and calculating response gains, adjustments are made more precise. Trend deduction and disturbance assessment mechanisms ensure that the system responds quickly to actual environmental changes, thereby optimizing the system's response speed and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] Figure 1 This is a structural diagram of a modulator intelligent management system based on the Internet of Things of the present invention; Figure 2 This is a flow chart of a modulator intelligent management method based on the Internet of Things according to the present invention; Figure 3 This is a schematic diagram of an embodiment of a modulator intelligent management method based on the Internet of Things of the present invention. DETAILED DESCRIPTION

[0007] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0008] Example: Figure 1-Figure 3 As shown, the present invention provides a technical solution, a modulator intelligent management method based on the Internet of Things, the method comprising the following steps: S100 collects operating data during the operation of the modulator and establishes a dynamically updated data window based on the time series; Furthermore, the specific steps for establishing a dynamically updated data window based on time series are as follows: S101 starts from the initial sampling time t1 and continuously collects the modulation depth data M(t), driving voltage data V(t) and chip temperature data T(t) of the modulator during operation at n moments. The sampling frequency of collecting the modulation depth, driving voltage and chip temperature should meet the Nyquist sampling theorem. The operation data at each sampling time ti are collected, i∈[1,n], and an operation data set Pti={M(ti),V(ti),T(ti)} is constructed, where M(ti), V(ti) and T(ti) respectively represent the modulation depth data, driving voltage data and chip temperature data at the sampling time ti. The operation data set is stored according to the acquisition timestamp to construct a time series data set P={Pt1, Pt2, Pt3, ..., Ptn}, where Pt1, Pt2, Pt3, ... and Ptn respectively represent the operation data sets constructed at the 1st, 2nd, 3rd, ... and nth acquisition timestamps, and tn represents the current real-time sampling time. The modulator is an interdigitated segmented silicon-based modulator consisting of multiple independently adjustable modulation segments. Each modulation segment contains an array of interdigitated electrodes, with the electrodes arranged orthogonally between adjacent segments. Each modulation segment is physically separated by an isolation slot. The modulation units are arranged along the optical waveguide structure, and each segment adjusts the phase of the optical signal under voltage control. The modulator adopts an adjustable segmented design, dynamically controlling the activation state of each segment to achieve different bit outputs, supporting multi-channel independent modulation. Each segment is equipped with an RC matching network to ensure optimized electric field distribution and reduce losses. By adopting a distributed interdigital structure and a discrete phase modulation unit in a collaborative design, the technical bottlenecks of the traditional continuous modulation area in terms of process tolerance and control accuracy are broken through while maintaining high-speed modulation characteristics. On the input-to-output path, due to processing defects, the doping gain of each segment is not exactly the same. By combining the interdigital segmented modulation structure, a segmented phase modulator with reconfigurable working state is realized. The phase difference can be changed directly through drive control, crosstalk can be suppressed, and modulation linearity can be optimized. The modulator has multiple configurable modulation segments, and the modulation capability of each segment is independent and adjustable. The modulation units can work independently, or be connected in series or parallel in a set manner to form a multi-segment modulation structure. The multi-segment modulation units are independently controlled by the electro-optical phase modulator to achieve high-resolution dynamic adjustment of the refractive index. It has dynamic response characteristics and does not require the introduction of a phase compensation zone. The process error is dispersed to all segments, improving robustness. S102 constructs a sliding window Wt of length n. The length n of the sliding window can be configured based on the response time of the modulator, computing resources, and target accuracy. To ensure real-time performance and accuracy, the update frequency of the sliding window can be set to the same as the data acquisition frequency. The structure of the sliding window is defined as Wt={Pt1, Pt2, Pt3, ..., Ptn}. When the operating data set Ptn+1 at time tn+1 is collected, Ptn+1 is inserted at the end of the sliding window, and the operating data set at time t1 is removed to construct a new sliding window. Through the sliding window mechanism, this solution ensures real-time data collection and updates, significantly improving the system's real-time performance, response speed, prediction accuracy, and control stability. Furthermore, the sliding window's adaptive nature effectively adapts to environmental changes, reduces the impact of outdated data, and optimizes the system's computational efficiency and stability. S200 calculates the difference between the current operating data and the operating data in the data window, extracts data in a plurality of data windows with differences, and predicts the modulation depth prediction value of the modulator at the next moment based on the extracted data; Furthermore, the specific steps for predicting the modulation depth prediction value at the next moment are: S201 calculates the driving voltage change ΔVn at time tn and the chip temperature change ΔTn at time tn. The calculation formula is: , Construct the state feature vector S(tn) at the reaction time tn, expressed as S(tn)=[M(tn),M(tn-1),ΔVn,ΔTn]; S202 constructs the state feature vector S(tj) at the time tj in the sliding window Wt, j∈[1,n-1], and calculates the difference dj(tn) between the state at time tn and the state at time tj. The calculation formula is: , Where ω1 represents the weight of the driving voltage change, and ω2 represents the weight of the chip temperature change. The weight coefficients ω1 and ω2 are adjusted according to the actual working conditions to reflect the relative importance of the voltage change and temperature change on the system. The differences between the state characteristics from time t1 to tn-1 and the state characteristics at time tn are collected to construct the difference set D={d1(tn), d2(tn), d3(tn), ..., dn-1(tn)}, where d1(tn), d2(tn), d3(tn), ..., and dn-1(tn) represent the differences between time t1, t2, t3, ..., and tn-1 and time tn, respectively. S203 sorts the differences in set D in a forward order and extracts the first k difference values. The value of k is recommended to be 5% to 10% of n, balancing computational complexity and sample diversity. A screening difference value set D*={d1,d2,d3,...,dk} is constructed, where d1,d2,d3,... and dk represent the first, second, third,... and kth difference values ​​extracted, respectively. The linear interpolation weight ωdx of the xth difference value dx is calculated using the following formula: ; S204 calculates the corresponding modulation depth data change ΔMx for the state feature vector S(tx) corresponding to the x-th difference value at time tx, using the following formula: , The modulation depth prediction value M^(tn+1) at time tn+1 is calculated based on the modulation depth data change at the time corresponding to the difference value in the set D* and the corresponding linear interpolation weight. The calculation formula is: ; Through difference calculation and dynamic weighted prediction, the accuracy of modulation depth prediction and control stability are effectively improved. By comparing historical data with real-time data, combined with linear interpolation and weighting strategies, the system's adaptability to changing environments is optimized. At the same time, difference screening ensures computational efficiency. The combined application of these technical means enables the optical modulator control system to maintain efficient and stable operation under complex and dynamically changing operating conditions. S300 recursively generates predicted values ​​at several future moments based on the modulation depth prediction value at the next moment and the latest modulation change direction, thereby constructing a modulation prediction sequence. Furthermore, the specific steps of constructing the modulation prediction sequence are: S301 generates the modulation depth prediction value corresponding to the next q moments based on the modulation depth prediction value at time tn+1 through a recursive formula. For the modulation depth prediction value M^(tn+g) corresponding to the g-th moment in the future, g∈[2,q], the recursive process is: , Where M^(tn+g-1) represents the modulation depth prediction value corresponding to the g-1th moment in the future, and α is the trend inference coefficient. The trend inference coefficient needs to be selected according to the dynamic characteristics of the system. A higher value can make the prediction more sensitive to recent changes, while a lower value can improve the stability of the prediction and avoid over-response to short-term fluctuations. It can be set through experiments or online optimization. S302 collects the modulation depth prediction values ​​corresponding to q future moments and constructs a prediction sequence M^={M^(tn+1),M^(tn+2),M^(tn+3),...,M^(tn+q)}, where M^(tn+1), M^(tn+2), M^(tn+3),... and M^(tn+q) represent the modulation depth prediction values ​​corresponding to the 1st, 2nd, 3rd,...th and qth future moments, respectively; The predicted modulation depth at future moments is used to generate a prediction sequence for multiple moments. The recursive process can effectively capture the changing trend of the modulation depth. The recursive formula can combine the trends of historical prediction values ​​and current prediction values ​​to generate a multi-step prediction sequence, ensuring that the system can identify and adjust the modulation depth at future moments in advance. This recursive process can capture the trend changes of the system at different time scales, providing an accurate prediction basis for subsequent voltage regulation and disturbance assessment. S400 compares the modulation prediction sequence with the modulation data target value and calculates the disturbance index to measure the system stability; Furthermore, the specific steps to obtain the disturbance index for measuring system stability are as follows: S401 collects actual modulation depth data for the next q moments, and calculates the deviation ΔM^ between the modulation depth prediction value at the gth future moment and the modulation depth target value M* for the next q moments. The calculation formula is: , Aggregate the deviation values ​​of the next q moments to construct an error value set W={ΔM^1,ΔM^2,ΔM^3,...,ΔM^q}, where ΔM^1,ΔM^2,ΔM^3,... and ΔM^q represent the deviation values ​​between the modulation depth prediction value and the modulation depth data target value at the 1st, 2nd, 3rd,... and qth moments in the future, respectively. Extract the maximum value of the data in the set W as the maximum deviation value ΔMmax; S402 calculates the cumulative error Esun between the modulation depth prediction value and the actual modulation depth data deviation value at the next q moments, and the calculation formula is: ; S403 calculates the trend change ΔMtrend between the modulation depth prediction value and the modulation depth data target value at the qth moment in the future. The calculation formula is: ; S404 constructs a comprehensive disturbance index Dn based on the maximum deviation value, cumulative error and trend change. The construction formula is: , Among them, β1, β2, and β3 are weight coefficients. β1, β2, and β3 can be configured according to the requirements of system design and application scenarios. Generally, the maximum deviation ΔMmax has a larger weight because it directly reflects the difference between the system modulation depth and the target value, which may lead to system instability. By accurately calculating the modulation depth deviation, cumulative error, and trend change at future moments, and combining factors such as maximum deviation, cumulative error, and trend change, a comprehensive disturbance index is constructed to effectively evaluate the stability and adaptability of the system. By flexibly configuring weight coefficients, the system can dynamically adjust the control strategy according to different operating conditions, identify and respond to potential disturbances and errors in advance, and ensure that the modulation depth always approaches the target value, thereby improving the system's robustness, accuracy, and adaptability. S500 calculates an adjustment value for controlling the modulation depth based on the disturbance index and the modulation depth prediction value at the next moment; Furthermore, the specific steps for calculating the adjustment value for controlling the modulation depth are: S501 calculates the response gain value Kt of the driving voltage data to the modulation depth data based on the modulation depth data at time tn and the predicted value of the modulation depth data at time tn+1 in the future, combined with the driving voltage data at time tn. The calculation formula is: , Where ε is a constant to prevent zero minima; S502 calculates the driving voltage data adjustment value ΔVd that needs to be applied based on the trend change amount, the comprehensive disturbance index and the response gain. The calculation formula is: ; In order to ensure the system's adaptability to different environmental changes, the voltage regulation amount will be adaptively adjusted according to the real-time disturbance index Dn. In the case of high disturbance, the regulation amount may increase to quickly restore system stability; in the case of low disturbance, the regulation amount will decrease to avoid over-adjustment and waste of energy. Extract the maximum rated driving voltage Vmax and the minimum rated driving voltage Vmin of the modulator, and set the adjustment range of the driving voltage data adjustment amount to Vmin≤V(tn)+ΔVd≤Vmax; S503 calculates the final output drive voltage adjustment value V(tn+1) at the next tn+1th moment based on the drive voltage data adjustment value, and the calculation formula is: , Where Vmax represents the maximum rated drive voltage of the modulator, Vmin represents the minimum rated drive voltage of the modulator, and the clip function indicates that the final output drive voltage is between the maximum rated drive voltage and the minimum rated drive voltage; Through adaptive voltage regulation and disturbance assessment, it achieves fast and stable recovery under different environmental changes, optimizes the system's energy efficiency and response speed, and ensures that the drive voltage is always maintained within the rated range, avoiding over-adjustment and energy waste; In order to better implement the above method, a modulator intelligent management system based on the Internet of Things is proposed. The system includes a data acquisition module, a state prediction module, a sequence recursion module, a disturbance assessment module and a regulation module. The data acquisition module is used to collect the operating data of the modulator during operation and establish a dynamically updated data window; The data acquisition unit includes a data acquisition unit and a window management unit; The data acquisition unit is used to collect the operating data of modulation depth, driving voltage and chip temperature and store them according to the timestamp; The window management unit is used to insert new data and remove old data through a sliding window mechanism to build a dynamically updated data window; The state prediction module is used to predict the modulation depth prediction value at the next moment based on the data in the data window that is significantly different from the real-time data; The state prediction module includes a feature calculation unit and a prediction generation unit; The feature calculation unit is used to calculate the voltage and temperature changes, construct the state feature vector and calculate the difference value with the historical state; The prediction generation unit is used to screen similar historical states and calculate the next moment prediction value based on the weighted modulation depth change; The sequence recursion module is used to recursively generate prediction values ​​for several future moments and construct a modulation prediction sequence; The disturbance assessment module is used to calculate the disturbance index to measure the stability of the system; The adjustment module is used to calculate the adjustment value for controlling the modulation depth and output it to the modulator; The adjustment module includes a gain calculation unit and an adjustment control unit; The gain calculation unit is used to calculate the response gain of the driving voltage to the modulation depth; The regulation control unit is used to calculate the voltage regulation amount based on the disturbance index and the predicted deviation, and output the final voltage through the safety range constraint.

[0009] Example: Starting from the initial time t1, the operating data of the modulator is collected, including the modulation depth, the driving voltage and the chip temperature. Assuming data is collected once per second at a frequency of 1Hz, the system collects the following first batch of data: modulation depth data M(t): 0.75(t1), 0.78(t2), 0.80(t3), ..., 0.85(tn); drive voltage data V(t): 3.5V(t1), 3.55V(t2), 3.60V(t3), ..., 3.8V(tn); and chip temperature data T(t): 35°C(t1), 35.2°C(t2), 35.4°C(t3), ..., 36.5°C(tn). These data form a data set P(ti) at each sampling instant. These data are sorted by timestamp to form a time series data set P, which is updated and stored in real time for subsequent calculations. The length of the sliding window is n, for example, assuming n = 10. This means that each data window processed by the system contains data from the most recent 10 sampling instants. Assuming new data is collected at time 11 (t11), the window removes the historical data from t1 to t10 and adds the data from time t11. The new sliding window becomes: Wt = {P(t2), P(t3), ..., P(t11)}. This sliding window ensures that the system processes the latest 10 data points in real time and updates the data promptly. Each time new data arrives, the system calculates the difference between the current data and the data in the sliding window. Assume that the modulation depth collected at time tn is 0.82, the voltage is 3.65V, and the temperature is 35.6°C. The difference between the current time and time t5 is: modulation depth 0.82-0.80, voltage 3.65V-3.60V, and temperature 35.6°C-35.4°C. After calculating the difference values, the system evaluates the significance of the differences and selects the top k most significant differences for subsequent modulation depth prediction. For each historical moment in the sliding window, the system calculates the difference between the current moment and the historical moment, sorts them by size, and selects the top k most representative differences. For example, suppose the difference values ​​from t1 to t10 are sorted and the top three most significant differences are selected. For these selected differences, the system calculates linear interpolation weights based on their similarity. Assuming the weights of the top three differences are 0.6, 0.3, and 0.1, respectively, the system applies these weights to the corresponding modulation depth changes and performs a weighted calculation. Through this weighted calculation, the system can predict the modulation depth at the next moment based on historical data and real-time data. For example, the system predicts the modulation depth at time tn+1 to be 0.83. Based on the predicted modulation depth value (0.83) at the next moment, the system recursively generates modulation depth predictions for the next q moments.Assuming q = 5, the predicted modulation depths for the next five moments are: M(tn+1)=0.83, M(tn+2)=0.84, M(tn+3)=0.85, M(tn+4)=0.86, M(tn+5)=0.87. The system aggregates these predicted values ​​to form a predicted sequence of future modulation depths. The system extracts the actual modulation depth data target value at future moments and calculates the deviation. By calculating the maximum deviation, cumulative error, and trend change, the system obtains the comprehensive disturbance index Dn. Based on the disturbance index Dn, the system calculates the adjustment value of the driving voltage. For example, assuming that the system calculates the adjustment value ΔVd = 0.1V based on Dn=0.015, then the adjustment value of the driving voltage is: V(tn+1)=3.65V+0.1V=3.75V.

[0010] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A modulator intelligent management method based on the Internet of Things, characterized by: S100 collects operating data during the operation of the modulator and establishes a dynamically updated data window based on the time series; S200 calculates the difference between the current operating data and the operating data in the data window, extracts data in a plurality of data windows with differences, and predicts the modulation depth prediction value of the modulator at the next moment based on the extracted data; S300 recursively generates predicted values ​​at several future moments based on the modulation depth predicted value at the next moment, and constructs a modulation prediction sequence; S400 compares the modulation prediction sequence with the modulation data target value and calculates the disturbance index to measure the system stability; S500 calculates an adjustment value for controlling the modulation depth based on the disturbance index and the modulation depth prediction value at the next moment.

2. The method for intelligent management of modulators based on the Internet of Things according to claim 1, characterized in that: The specific steps for establishing a dynamically updated data window based on a time series in S100 are as follows: S101 starts from the initial sampling time t1, continuously collects the modulation depth data M(t), driving voltage data V(t) and chip temperature data T(t) of the modulator during operation at n moments, collects the operation data at each sampling time ti, i∈[1,n], and constructs an operation data set Pti={M(ti),V(ti),T(ti)}, where M(ti), V(ti) and T(ti) respectively represent the modulation depth data, driving voltage data and chip temperature data at the sampling time ti, stores the operation data set according to the acquisition timestamp, and constructs a time series data set P={Pt1, Pt2, Pt3, ..., Ptn}, where Pt1, Pt2, Pt3, ... and Ptn respectively represent the operation data sets constructed at the 1st, 2nd, 3rd, ... and nth acquisition timestamps, and tn represents the current real-time sampling time; S102 constructs a sliding window Wt of length n. The structure of the sliding window is defined as Wt={Pt1, Pt2, Pt3, ..., Ptn}. When the operating data set Pt(n+1) at time tn+1 is collected, P(tn+1) is inserted into the end of the sliding window, and the operating data set at time t1 is removed to construct a new sliding window.

3. The method for intelligent management of modulators based on the Internet of Things according to claim 2, characterized in that: The specific steps of predicting the modulation depth prediction value at the next moment in S200 are: S201 calculates the driving voltage change ΔVn at time tn and the chip temperature change ΔTn at time tn. The calculation formula is: ， Construct the state feature vector S(tn) at the reaction time tn, expressed as S(tn)=[M(tn),M(tn-1),ΔVn,ΔTn]; S202 constructs the state feature vector S(tj) at the time tj in the sliding window Wt, j∈[1,n-1], and calculates the difference dj(tn) between the state at time tn and the state at time tj. The calculation formula is: ， Where ω1 represents the weight of the driving voltage change, ω2 represents the weight of the chip temperature change, and the differences between the state characteristics from t1 to tn-1 and the state characteristics at tn are collected to construct a difference set D={d1(tn), d2(tn), d3(tn), ..., dn-1(tn)}, where d1(tn), d2(tn), d3(tn), ..., and dn-1(tn) represent the differences between t1, t2, t3, ..., and tn-1 and tn, respectively. S203 arranges the differences in the set D in a forward order, extracts the first k difference values, and constructs a screening difference value set D*={d1,d2,d3,...,dk}, where d1,d2,d3,... and dk represent the first, second, third,... and kth difference values ​​extracted, respectively. Calculate the linear interpolation weight ωdx of the x-th difference value dx, and the calculation formula is: ； S204 calculates the corresponding modulation depth data change ΔMx for the state feature vector S(tx) corresponding to the x-th difference value at time tx, using the following formula: ， The modulation depth prediction value M^(tn+1) at time tn+1 is calculated based on the modulation depth data change at the time corresponding to the difference value in the set D* and the corresponding linear interpolation weight. The calculation formula is: 。 4. The method for intelligent management of modulators based on the Internet of Things according to claim 3, characterized in that: In S300, the specific steps of recursively generating predicted values ​​for several future moments and constructing a modulation prediction sequence are as follows: S301 generates the modulation depth prediction value corresponding to the next q moments based on the modulation depth prediction value at time tn+1 through a recursive formula. For the modulation depth prediction value M^(tn+g) corresponding to the g-th moment in the future, g∈[2,q], the recursive process is: ， Where M^(tn+g-1) represents the modulation depth prediction value corresponding to the g-1th moment in the future, and α is the trend deduction coefficient; S302 collects the modulation depth prediction values ​​corresponding to q future moments and constructs a prediction sequence M^={M^(tn+1),M^(tn+2),M^(tn+3),...,M^(tn+q)}, where M^(tn+1), M^(tn+2), M^(tn+3),... and M^(tn+q) represent the modulation depth prediction values ​​corresponding to the 1st, 2nd, 3rd,...th and qth future moments respectively.

5. The method for intelligent management of modulators based on the Internet of Things according to claim 4, characterized in that: The specific steps for obtaining the disturbance index for measuring system stability in S400 are as follows: S401 collects actual modulation depth data for the next q moments, and calculates the deviation ΔM^ between the modulation depth prediction value at the gth future moment and the modulation depth target value M* for the next q moments. The calculation formula is: ， Aggregate the deviation values ​​of the next q moments to construct an error value set W={ΔM^1,ΔM^2,ΔM^3,...,ΔM^q}, where ΔM^1,ΔM^2,ΔM^3,... and ΔM^q represent the deviation values ​​between the modulation depth prediction value and the modulation depth data target value at the 1st, 2nd, 3rd,... and qth moments in the future, respectively. Extract the maximum value of the data in the set W as the maximum deviation value ΔMmax; S402 calculates the cumulative error Esun between the modulation depth prediction value and the actual modulation depth data deviation value at the next q moments, and the calculation formula is: ； S403 calculates the trend change ΔMtrend between the modulation depth prediction value and the modulation depth data target value at the qth moment in the future. The calculation formula is: ； S404 constructs a comprehensive disturbance index Dn based on the maximum deviation value, cumulative error and trend change. The construction formula is: ， Among them, β1, β2 and β3 are weight coefficients.

6. The method for intelligent management of modulators based on the Internet of Things according to claim 5, characterized in that: The specific steps of calculating the adjustment value for controlling the modulation depth in S500 are: S501 calculates the response gain value Kt of the driving voltage data to the modulation depth data based on the modulation depth data at time tn and the predicted value of the modulation depth data at time tn+1 in the future, combined with the driving voltage data at time tn. The calculation formula is: ， Where ε is a constant to prevent zero minimum; S502 calculates the driving voltage data adjustment value ΔVd that needs to be applied based on the trend change amount, the comprehensive disturbance index and the response gain. The calculation formula is: ； Extract the maximum rated driving voltage Vmax and the minimum rated driving voltage Vmin of the modulator, and set the adjustment range of the driving voltage data adjustment amount to Vmin≤V(tn)+ΔVd≤Vmax; S503 calculates the final output drive voltage adjustment value V(tn+1) at the next tn+1th moment based on the drive voltage data adjustment value, and the calculation formula is: ， Wherein Vmax represents the maximum rated driving voltage of the modulator, Vmin represents the minimum rated driving voltage of the modulator, and the clip function represents that the final output driving voltage is between the maximum rated driving voltage and the minimum rated driving voltage.

7. A modulator intelligent management system based on the Internet of Things, applied to a modulator intelligent management method based on the Internet of Things as claimed in any one of claims 1 to 6, characterized in that: The system includes a data acquisition module, a state prediction module, a sequence recursion module, a disturbance assessment module and a regulation module; The data acquisition module is used to collect the operation data of the modulator during operation and establish a dynamically updated data window; The state prediction module is used to predict the modulation depth prediction value at the next moment based on the data in the data window that is significantly different from the real-time data; The sequence recursion module is used to recursively generate prediction values ​​of several future moments to construct a modulation prediction sequence; The disturbance assessment module is used to calculate a disturbance index for measuring system stability; The adjustment module is used to calculate an adjustment value for controlling the modulation depth and output the adjustment value to the modulator.

8. The modulator intelligent management system based on the Internet of Things according to claim 7, characterized in that: The data acquisition unit includes a window management unit; The data acquisition unit is used to collect modulation depth, drive voltage and chip temperature operation data and store them according to timestamps; The window management unit is used to insert new data and remove old data through a sliding window mechanism to build a dynamically updated data window.

9. The modulator intelligent management system based on the Internet of Things according to claim 8, characterized in that: The state prediction module includes a feature calculation unit and a prediction generation unit; The characteristic calculation unit is used to calculate the voltage and temperature changes, construct the state characteristic vector and calculate the difference value with the historical state; The prediction generation unit is used to screen similar historical states and calculate the next moment prediction value based on the weighted modulation depth change.

10. The modulator intelligent management system based on the Internet of Things according to claim 9, characterized in that: The adjustment module includes a gain calculation unit and an adjustment control unit; The gain calculation unit is used to calculate the response gain of the driving voltage to the modulation depth; The regulation control unit is used to calculate the voltage regulation amount according to the disturbance index and the prediction deviation, and output the final voltage through the safety range constraint.

Images