Remote control method and system for waste rubber thermal cracking treatment based on internet of things

Abstract

The application provides a remote control method and system for waste rubber thermal cracking treatment based on the Internet of Things, relates to the technical field of the Internet of Things, and comprises collecting thermal cracking treatment equipment operation state data and performing space-time correlation analysis, constructing a multi-scale time axis to identify rapid fluctuation and slow drift characteristics, and generating a layered state representation. According to the state quantity, local instant and remote collaborative control tasks are divided, and the instruction generation and execution path are determined through a dynamic permission allocation mechanism. Based on the deviation evolution track of the execution feedback and the state quantity, a two-way correction mechanism is constructed to adaptively reconstruct the permission allocation proportion and the time axis scale parameter. The application realizes intelligent grading and dynamic resource allocation of control tasks, and improves the response speed and control accuracy of the continuous thermal cracking treatment system.

Description

Technical Field

[0001] This invention relates to Internet of Things (IoT) technology, and more particularly to a remote control method and system for the pyrolysis treatment of waste rubber based on IoT. Background Technology

[0002] In the field of remote control of continuous pyrolysis processes for waste rubber and waste plastics based on the Internet of Things (IoT), existing technologies typically rely on centralized monitoring and control architectures. The conventional approach involves collecting operating parameters, such as temperature, pressure, and material flow, through a sensor network deployed locally on the continuous pyrolysis equipment. This data is then uploaded to a remote cloud control center via an IoT gateway. The cloud center centrally processes and analyzes the aggregated data, generates control commands based on pre-set control models or algorithms, and then distributes them to the on-site actuators via the network. This model aims to achieve centralized management and remote operation of the continuous pyrolysis process, reducing reliance on on-site manual labor. Another common approach is to employ a relatively fixed edge computing and cloud collaboration framework. Simple, latency-sensitive computing tasks are deployed on edge nodes close to the equipment, while complex, big data-intensive analysis and decision-making tasks are retained in the cloud.

[0003] The pyrolysis of waste rubber and waste plastics is a complex physicochemical process, characterized by both rapid dynamic fluctuations and long-term slow drifts. Existing centralized or fixed hierarchical control architectures struggle to effectively analyze this multi-timescale coupled state evolution behavior. Uploading data to the cloud for centralized processing introduces significant communication latency, causing a lag in the system's response to sudden, rapid fluctuations in equipment performance, potentially missing optimal control opportunities and impacting process stability and safety. Furthermore, fixed task division and permission allocation mechanisms lack flexibility, failing to adaptively adjust the generation location and execution path of control decisions based on the dynamic changes in the real-time state and network conditions of the continuous pyrolysis process. This results in insufficient robustness and efficiency of the overall control system when dealing with complex operating conditions. Summary of the Invention

[0004] This invention provides a remote control method and system for continuous pyrolysis treatment of waste rubber based on the Internet of Things, which can solve the problems in the prior art.

[0005] A first aspect of the present invention provides a remote control method for pyrolysis processing based on the Internet of Things, comprising: The system collects operational status data of distributed pyrolysis equipment, performs spatiotemporal correlation analysis on the operational status data, constructs a multi-scale time axis of state evolution, identifies rapid fluctuation characteristics and slow drift characteristics, and generates hierarchical state representation based on the coupling relationship between the rapid fluctuation characteristics and the slow drift characteristics to obtain instantaneous response state quantity and trend prediction state quantity. The priority of the control task is determined based on the real-time response state quantity and the trend prediction state quantity, and the control task is divided into local real-time control task and remote collaborative control task of the pyrolysis production line. Based on the execution time requirements of the local real-time control task and the remote collaborative control task, the generation location and execution path of the control command are determined through the dynamic permission allocation mechanism between the edge node and the cloud center. The control commands are transmitted to the distributed pyrolysis processing equipment through the execution path. Based on the deviation evolution trajectory of the status feedback data after execution, the instantaneous response status quantity, and the trend prediction status quantity, a two-way correction mechanism between the edge node and the cloud center is constructed. The allocation ratio of the dynamic permission allocation mechanism and the scale division parameters of the multi-scale time axis are adaptively reconstructed.

[0006] By constructing a multi-scale time axis for state evolution, rapid fluctuation characteristics and slow drift characteristics are identified. Based on the coupling relationship between the rapid fluctuation characteristics and the slow drift characteristics, a hierarchical state representation is generated, resulting in instantaneous response state quantities and trend prediction state quantities, including: A fast-scale time axis and a slow-scale time axis are constructed for the operating status data. Rapid fluctuation characteristics are identified by capturing instantaneous state jump information on the fast-scale time axis, and slow drift characteristics are identified by tracking the long-term evolution trajectory of the state on the slow-scale time axis. A modulation mapping model of the fast fluctuation feature on the slow drift feature is established. By analyzing the dependence between the amplitude change of the fast fluctuation feature and the evolution stage of the slow drift feature, the transfer coupling characteristics between scales are extracted to obtain the coupling transfer function. Based on the coupling transfer function, the rapid fluctuation feature and the slow drift feature are projected onto a unified state representation space, and an instantaneous response dimension and a trend prediction dimension are constructed in the state representation space. Based on the orthogonal decomposition relationship between the instant response dimension and the trend prediction dimension, instant response state variables and trend prediction state variables are generated.

[0007] A modulation mapping model of the rapid fluctuation feature on the slow drift feature is established. By analyzing the dependency between the amplitude change of the rapid fluctuation feature and the evolution stage of the slow drift feature, the inter-scale transfer coupling characteristics are extracted, and the coupling transfer function is obtained, including: The amplitude of the rapid fluctuation feature is demodulated, the amplitude envelope curve of the rapid fluctuation feature is extracted, and a time synchronization correspondence is established between the amplitude envelope curve and the evolution trajectory of the slow drift feature. Based on the time synchronization correspondence, the local gradient and curvature of the amplitude envelope curve at different evolution positions of the slow drift feature are calculated, and the deformation mode of the amplitude envelope curve as the slow drift feature evolves is identified. The evolution trajectory of the slow drift feature is divided into an accelerated evolution interval, a uniform evolution interval, and a decelerated evolution interval according to the state space gradient. The deformation mode features of the amplitude envelope curve are statistically analyzed in each evolution interval to establish a modulation mapping model. The modulation mapping model is used to quantify the dependence of the amplitude change of the fast fluctuation feature on the evolution stage of the slow drift feature, and the transmission relationship of the dependence strength between different evolution intervals is extracted to obtain the inter-scale transmission coupling characteristics. The coupling characteristic is expressed as a response function of the amplitude of the rapid fluctuation feature to the evolution stage of the slow drift feature, thus obtaining the coupling transfer function.

[0008] The priority of the control task is determined based on the real-time response state quantity and the trend prediction state quantity, and the control task is divided into local real-time control tasks and remote collaborative control tasks for the pyrolysis production line, including: The urgency of the immediate response state quantity is quantified by evaluating the deviation magnitude and rate of change of the immediate response state quantity to obtain an urgency score; the complexity of the trend prediction state quantity is quantified by evaluating the number of associated parameters and evolution uncertainty involved in the evolution path of the trend prediction state quantity to obtain a complexity score. Based on the urgency score and the complexity score, a priority evaluation matrix for the control task is constructed. The priority evaluation matrix is ​​calculated by weighted combination of the urgency score and the complexity score to obtain the priority value. The control tasks are classified according to the priority values. Control tasks with an urgency score exceeding the first classification threshold and a complexity score below the second classification threshold are classified as local real-time control tasks, and control tasks with a complexity score exceeding the third classification threshold are classified as remote collaborative control tasks.

[0009] Based on the execution time requirements of the local real-time control task and the remote collaborative control task, the generation location and execution path of the control command are determined through a dynamic permission allocation mechanism between edge nodes and the cloud center, including: For local real-time control tasks, timeliness requirements are analyzed, the coupling relationship between calculation latency and communication latency is quantified, and local timeliness constraints are obtained; for remote collaborative control tasks, timeliness requirements are analyzed, the cascading effect of data transmission latency and algorithm deduction latency is quantified, and remote timeliness constraints are obtained. Based on the local time constraint value and the remote time constraint value, a dynamic permission allocation mechanism is established between edge nodes and the cloud center. The dynamic permission allocation mechanism monitors the computing resource utilization rate of edge nodes and the network transmission latency of the cloud center, and combines the matching degree between the local time constraint value and the available latency margin of edge nodes and the matching degree between the remote time constraint value and the end-to-end latency of the cloud center to dynamically adjust the allocation ratio of control permissions. Based on the allocation ratio of the dynamic permission allocation mechanism, the generation location and execution path of the control command are determined.

[0010] The control commands are transmitted to the distributed pyrolysis processing equipment through the execution path. Based on the deviation evolution trajectory between the executed state feedback data and the instantaneous response state quantity and the trend prediction state quantity, a two-way correction mechanism between the edge node and the cloud center is constructed, including: The control command is transmitted to the distributed pyrolysis processing equipment through the execution path, and the status feedback data generated by the distributed pyrolysis processing equipment after executing the control command is received. Extract the actual response state quantity from the state feedback data, calculate the instantaneous deviation between the actual response state quantity and the instantaneous response state quantity, and the trend deviation between the actual response state quantity and the trend prediction state quantity. Based on the time-dimensional change sequence of the instantaneous deviation value and the trend deviation value, a deviation evolution trajectory is constructed, and feature decomposition is performed on the deviation evolution trajectory to identify the transient response deviation component and the steady-state tracking deviation component in the deviation evolution trajectory. Based on the transient response deviation component and the steady-state tracking deviation component, a two-way correction mechanism between the edge node and the cloud center is constructed.

[0011] Based on the time-series changes of the instantaneous deviation value and the trend deviation value, a deviation evolution trajectory is constructed. Feature decomposition is performed on the deviation evolution trajectory to identify the transient response deviation component and the steady-state tracking deviation component, including: The time-series sampling of the change sequences of instantaneous deviation values ​​and trend deviation values ​​in the time dimension is performed, and the change sequences of instantaneous deviation values ​​and trend deviation values ​​are segmented by setting a time window to obtain deviation sequence segments of multiple time periods. The deviation sequence segments from multiple time periods are continuously spliced ​​together on the time axis. By fitting the continuous change trend of the deviation sequence segments, a deviation evolution trajectory is constructed. The deviation evolution trajectory includes the instantaneous deviation evolution curve corresponding to the instantaneous deviation value and the trend deviation evolution curve corresponding to the trend deviation value. The deviation evolution trajectory is decomposed in the frequency domain, and the instantaneous deviation evolution curve and the trend deviation evolution curve are respectively decomposed into a set of frequency components; Frequency components whose frequency changes exceed the frequency boundary threshold are extracted from the set of frequency components as transient response deviation components, and frequency components whose frequency changes are below the frequency boundary threshold are extracted as steady-state tracking deviation components.

[0012] A second aspect of the present invention provides a remote control system for waste rubber pyrolysis treatment based on the Internet of Things, comprising: The state analysis unit is used to collect the operating state data of the distributed pyrolysis processing equipment, perform spatiotemporal correlation analysis on the operating state data, identify rapid fluctuation characteristics and slow drift characteristics by constructing a multi-scale time axis of state evolution, and generate hierarchical state representation based on the coupling relationship between the rapid fluctuation characteristics and the slow drift characteristics to obtain the instantaneous response state quantity and trend prediction state quantity. The task division unit is used to determine the priority of the control task based on the real-time response state quantity and the trend prediction state quantity, and to divide the control task into local real-time control tasks and remote collaborative control tasks of the pyrolysis production line. The permission allocation unit is used to determine the generation location and execution path of control commands through a dynamic permission allocation mechanism between edge nodes and the cloud center, based on the execution time requirements of the local real-time control task and the remote collaborative control task. The feedback correction unit is used to transmit the control command to the distributed pyrolysis processing equipment through the execution path, and construct a two-way correction mechanism between the edge node and the cloud center based on the deviation evolution trajectory of the post-execution status feedback data, the instantaneous response status quantity and the trend prediction status quantity, and adaptively reconstruct the allocation ratio of the dynamic permission allocation mechanism and the scale division parameters of the multi-scale time axis.

[0013] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0014] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0015] The beneficial effects of this application are as follows: This method enables refined perception and forward-looking prediction of the operating status of distributed pyrolysis processing equipment. Through spatiotemporal correlation analysis and multi-scale time axis construction, it effectively separates and couples the rapid fluctuation characteristics and slow drift characteristics of the equipment status, thereby generating a hierarchical status representation that includes both immediate response status variables and trend prediction status variables. This process significantly improves the depth and breadth of status monitoring, making the capture of instantaneous anomalies and long-term performance degradation trends more accurate and timely.

[0016] Based on hierarchical state representation, intelligent classification and precise scheduling of control tasks are achieved. By automatically classifying local real-time control tasks and remote collaborative control tasks according to the properties of state variables and determining their priorities, it ensures that critical and urgent control needs can be responded to quickly and efficiently. This effectively optimizes the allocation of control resources, avoids congestion or delays in control commands, and enhances the system's agility in responding to emergencies.

[0017] By employing a dynamic permission allocation mechanism between edge nodes and the cloud center, optimized decisions regarding the generation location and execution path of control commands are achieved. This mechanism can flexibly distribute computational and decision-making loads between the edge and the cloud based on the timeliness requirements of control tasks, ensuring the real-time performance of local control while fully utilizing the global optimization and complex computing capabilities of the cloud. This ultimately improves the overall response speed and decision-making quality of the control system. Attached Figure Description

[0018] Figure 1 This is a schematic flowchart of a remote control method for waste rubber pyrolysis treatment based on the Internet of Things, according to an embodiment of the present invention. Figure 2 This is a flowchart of the deviation evolution trajectory recognition and component extraction processing according to an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0021] Figure 1 This is a schematic flowchart of a remote control method for waste rubber pyrolysis treatment based on the Internet of Things, according to an embodiment of the present invention. Figure 1 As shown, the method includes: The system collects operational status data of distributed pyrolysis equipment, performs spatiotemporal correlation analysis on the operational status data, constructs a multi-scale time axis of state evolution, identifies rapid fluctuation characteristics and slow drift characteristics, and generates hierarchical state representation based on the coupling relationship between the rapid fluctuation characteristics and the slow drift characteristics to obtain instantaneous response state quantity and trend prediction state quantity. The priority of the control task is determined based on the real-time response state quantity and the trend prediction state quantity, and the control task is divided into local real-time control task and remote collaborative control task of the pyrolysis production line. Based on the execution time requirements of the local real-time control task and the remote collaborative control task, the generation location and execution path of the control command are determined through the dynamic permission allocation mechanism between the edge node and the cloud center. The control commands are transmitted to the distributed pyrolysis processing equipment through the execution path. Based on the deviation evolution trajectory of the status feedback data after execution, the instantaneous response status quantity, and the trend prediction status quantity, a two-way correction mechanism between the edge node and the cloud center is constructed. The allocation ratio of the dynamic permission allocation mechanism and the scale division parameters of the multi-scale time axis are adaptively reconstructed.

[0022] In one optional implementation, by constructing a multi-scale time axis for state evolution, rapid fluctuation characteristics and slow drift characteristics are identified, and a hierarchical state representation is generated based on the coupling relationship between the rapid fluctuation characteristics and the slow drift characteristics, resulting in instantaneous response state quantities and trend prediction state quantities, including: A fast-scale time axis and a slow-scale time axis are constructed for the operating status data. Rapid fluctuation characteristics are identified by capturing instantaneous state jump information on the fast-scale time axis, and slow drift characteristics are identified by tracking the long-term evolution trajectory of the state on the slow-scale time axis. A modulation mapping model of the fast fluctuation feature on the slow drift feature is established. By analyzing the dependence between the amplitude change of the fast fluctuation feature and the evolution stage of the slow drift feature, the transfer coupling characteristics between scales are extracted to obtain the coupling transfer function. Based on the coupling transfer function, the rapid fluctuation feature and the slow drift feature are projected onto a unified state representation space, and an instantaneous response dimension and a trend prediction dimension are constructed in the state representation space. Based on the orthogonal decomposition relationship between the instant response dimension and the trend prediction dimension, instant response state variables and trend prediction state variables are generated.

[0023] To address the simultaneous rapid and slow dynamic behaviors of complex industrial systems during continuous pyrolysis of waste rubber, a multi-scale time axis decomposition method is employed to achieve refined identification of state characteristics. The original sampled time series is reconstructed according to different time resolutions. The sampling interval for the fast-scale time axis is set to 1 to 3 times the baseline period to capture abrupt state changes caused by equipment start-up and shutdown, load switching, and external disturbances. The sampling interval for the slow-scale time axis is set to 50 to 200 times the baseline period to track long-term evolution processes such as equipment aging, seasonal environmental changes, and slow changes in raw material composition. On the fast-scale time axis, the rate of state change between adjacent sampling points is calculated. When the absolute value of the rate of change exceeds a set threshold, it is marked as an instantaneous jump event. Feature parameters such as the time of occurrence, amplitude, and duration of the jump are extracted to form a rapid fluctuation feature vector. On the slow-scale time axis, a sliding window smoothing filter method is used to eliminate short-term disturbances while preserving the overall trend of state change. The smoothed curve is then piecewise linearly fitted, and feature parameters describing the evolution trajectory, such as slope and curvature, are extracted for each time period to form a slow drift feature vector.

[0024] To establish the intrinsic relationship between the two scale characteristics, a modulation mapping model of fast fluctuation characteristics on slow drift characteristics is constructed. The evolution trajectory on the slow-scale time axis is divided into several stages, each corresponding to a specific system health state or operating mode. The distribution characteristics of fast fluctuation characteristics within each stage are statistically analyzed, including fluctuation frequency, average amplitude, and peak occurrence probability. By comparing the statistical differences of fast fluctuation characteristics in different evolution stages, it is found that the intensity of fast fluctuations is closely related to the long-term state of the system. In the early stages when the system is relatively healthy, the amplitude of fast fluctuations is small and uniformly distributed; as the slow drift characteristics deteriorate, the amplitude of fast fluctuations gradually increases and exhibits an asymmetric distribution. A regression relationship is established between the amplitude of fast fluctuations and the evolution stage indicators of slow drift characteristics, using piecewise functions or nonlinear mapping functions to describe this dependency. Simultaneously, the energy distribution of fast fluctuation characteristics in different frequency bands is analyzed and correlated with the rate of change of slow drift characteristics to extract the energy transfer law between scales. When the rate of change of slow drift characteristics accelerates, the energy proportion of fast fluctuation characteristics in the high-frequency band increases significantly, indicating an enhanced coupling strength within the system. By quantifying this transfer characteristic, a transfer function describing the dynamic coupling between scales is obtained. The input of this function is the current state value and rate of change of the slowly drifting feature, and the output is the expected distribution parameters of the rapidly fluctuating feature.

[0025] The extracted rapid fluctuation and slow drift features are projected onto a unified multidimensional state representation space. The dimensional design of this space needs to accommodate both instantaneous dynamic information and long-term trend information. The weight matrix of the projection transformation is determined by the aforementioned coupling transfer function, ensuring that the projected feature vectors retain the key information of the original features while eliminating redundancy and noise. In the state representation space, the components along the instantaneous response dimension primarily reflect the system's rapid response capability to current inputs or disturbances, and their magnitude is positively correlated with the amplitude and frequency of the rapid fluctuation features. The components along the trend prediction dimension primarily reflect the evolution direction of the system's future state, and their magnitude is positively correlated with the slope and acceleration of the slow drift features. By adjusting the weighting coefficients in the projection transformation, the information distribution ratio between the two dimensions can be controlled, making the instantaneous response dimension sensitive to short-term disturbances, while the trend prediction dimension is sensitive to long-term changes.

[0026] To ensure the independence between the immediate response dimension and the trend prediction dimension, an orthogonal decomposition method is used to process the projected eigenvectors. First, the basis vectors corresponding to the two dimensions are calculated, requiring that the inner product of these two basis vectors is 0, satisfying mathematical orthogonality. By solving a constrained optimization problem, the amount of state information captured by each dimension is maximized while maintaining orthogonality. The projected eigenvectors are then decomposed along the two orthogonal basis vectors to obtain the values ​​of the immediate response state quantity and the trend prediction state quantity. The time series of the immediate response state quantity exhibits high-frequency oscillation characteristics, reflecting the degree of deviation of the system's current operating state in real time; the time series of the trend prediction state quantity exhibits smooth change characteristics, predicting the system's evolution trend over a future period. The combination of these two state quantities provides a comprehensive description of the system state, including both immediately observable surface phenomena and deeper changes that require long-term accumulation to manifest. In practical applications, abrupt changes in the immediate response state quantity can be used to determine whether immediate control measures are needed, and the trend prediction state quantity can be used to formulate preventative maintenance plans, achieving hierarchical and multi-timescale precise management of the system's operating state.

[0027] In one optional implementation, a modulation mapping model of the fast fluctuation feature on the slow drift feature is established. By analyzing the dependency between the amplitude change of the fast fluctuation feature and the evolution stage of the slow drift feature, the inter-scale transfer coupling characteristics are extracted, and the coupling transfer function is obtained, including: The amplitude of the rapid fluctuation feature is demodulated, the amplitude envelope curve of the rapid fluctuation feature is extracted, and a time synchronization correspondence is established between the amplitude envelope curve and the evolution trajectory of the slow drift feature. Based on the time synchronization correspondence, the local gradient and curvature of the amplitude envelope curve at different evolution positions of the slow drift feature are calculated, and the deformation mode of the amplitude envelope curve as the slow drift feature evolves is identified. The evolution trajectory of the slow drift feature is divided into an accelerated evolution interval, a uniform evolution interval, and a decelerated evolution interval according to the state space gradient. The deformation mode features of the amplitude envelope curve are statistically analyzed in each evolution interval to establish a modulation mapping model. The modulation mapping model is used to quantify the dependence of the amplitude change of the fast fluctuation feature on the evolution stage of the slow drift feature, and the transmission relationship of the dependence strength between different evolution intervals is extracted to obtain the inter-scale transmission coupling characteristics. The coupling characteristic is expressed as a response function of the amplitude of the rapid fluctuation feature to the evolution stage of the slow drift feature, thus obtaining the coupling transfer function.

[0028] After separating the rapid fluctuation characteristics from the slow drift characteristics, it is necessary to analyze the interaction and modulation relationship between the two. This modulation relationship reflects the energy transfer and state coupling mechanism between characteristics at different time scales, which is of great significance for understanding the multi-scale dynamic behavior of the system.

[0029] The rapid fluctuation characteristics are demodulated using amplitude, and their analytic signal representation is obtained using the Hilbert transform method. The amplitude envelope curve of the rapid fluctuation characteristics is extracted by calculating the magnitude of the analytic signal. This envelope curve reflects the energy change trend of the rapid fluctuation component, eliminating details of high-frequency oscillations while preserving the overall evolution trajectory of the amplitude. After extracting the amplitude envelope curve, it is placed on the same time axis as the evolution trajectory of the slow drift characteristics, establishing a strict time synchronization correspondence. This correspondence ensures that the amplitude envelope value at each moment accurately matches the corresponding slow drift characteristic state, providing a foundation for subsequent correlation analysis. During the synchronization process, the two curves need to be timestamped to eliminate errors caused by differences in sampling frequency or time offsets.

[0030] Based on the established time synchronization correspondence, the local dynamic characteristics of the amplitude envelope curve are calculated along the evolution trajectory of the slow drift feature. Multiple key points on the evolution trajectory of the slow drift feature are selected, and the first derivative of the amplitude envelope curve is calculated at each point using the central difference method. This derivative value characterizes the local gradient of the amplitude envelope curve, reflecting the rate of amplitude change. The second derivative is further calculated to obtain curvature information. The curvature value describes the degree of bending of the amplitude envelope curve; positive curvature indicates convex deformation, and negative curvature indicates concave deformation. By analyzing the gradient and curvature combinations at different evolution positions, the deformation patterns of the amplitude envelope curve as it evolves with the slow drift feature are identified. Common deformation patterns include synchronous enhancement, where the amplitude envelope curve increases with the rise of the slow drift feature; reverse suppression, where the amplitude envelope curve decreases with the rise of the slow drift feature; and phase lag, where the change of the amplitude envelope curve is delayed relative to the slow drift feature.

[0031] To more precisely characterize the regional differences in modulation relationships, the evolution trajectory of the slow drift feature is divided into intervals according to the state-space gradient. The time derivative of the evolution trajectory is calculated, and the evolution process is divided into three types of intervals based on the magnitude of the derivative: the accelerated evolution interval corresponds to a period with a large and continuously increasing absolute value of the derivative, representing a rapid transition phase of the system state; the uniform evolution interval corresponds to a period with a relatively stable absolute value of the derivative, representing a smooth progression phase of the system state; and the deceleration evolution interval corresponds to a period with a gradually decreasing absolute value of the derivative, representing a transition phase where the system state tends to stabilize. After completing the interval division, the deformation mode characteristics of the amplitude envelope curve are statistically analyzed within each evolution interval. For the accelerated evolution interval, the average gradient value, maximum curvature value, and gradient sign change frequency of the amplitude envelope curve within this interval are statistically analyzed; for the uniform and deceleration evolution intervals, the same statistical indicators are used for feature extraction. By comparing the differences in deformation mode characteristics within different evolution intervals, a modulation mapping model is established. This model describes how the amplitude changes of the rapid fluctuation feature respond to different evolution stages of the slow drift feature.

[0032] The dependence strength is quantified using an established modulation mapping model, defined as the ratio of the change in amplitude envelope curve to the change in slow drift characteristics. This ratio is calculated for each evolution interval; a larger ratio indicates a stronger dependence of the amplitude change of rapid fluctuation characteristics on the evolutionary stage of slow drift characteristics. Further analysis of the transmission relationship of dependence strength across different evolution intervals reveals the transfer pattern of dependence strength from accelerated evolution intervals to uniform evolution intervals and then to decelerated evolution intervals. A significant jump in dependence strength during interval transitions indicates the existence of strong coupling transmission; a smooth transition indicates continuous coupling transmission. By tracing the evolution trajectory of dependence strength, the inter-scale transmission coupling characteristics are extracted, revealing how energy and information flow across different time scales.

[0033] The extracted transfer coupling characteristics are expressed as a mathematical response function, with the evolutionary stage of the slow drift feature as the independent variable and the amplitude change rate of the fast fluctuation feature as the dependent variable. The response function can be piecewise, with different functional expressions in different evolution intervals, respectively characterizing the modulation patterns of the acceleration, constant velocity, and deceleration stages. By fitting measured data points within each interval, the parameters of the response function are determined, yielding the complete expression of the coupled transfer function. This coupled transfer function can predict the trend of the amplitude change of the fast fluctuation feature under a given slow drift feature evolution state, providing a quantitative tool for state prediction and control of multi-scale systems.

[0034] In one optional implementation, the priority of the control task is determined based on the immediate response state quantity and the trend prediction state quantity, and the control task is divided into local immediate control tasks and remote collaborative control tasks for the pyrolysis production line, including: The urgency of the immediate response status quantity is quantified, and an urgency score is obtained by evaluating the deviation magnitude and rate of change of the immediate response status quantity. The complexity of the trend prediction state quantity is quantified by evaluating the number of associated parameters and evolution uncertainty involved in the evolution path of the trend prediction state quantity, and a complexity score is obtained. Based on the urgency score and the complexity score, a priority evaluation matrix for the control task is constructed. The priority evaluation matrix is ​​calculated by weighted combination of the urgency score and the complexity score to obtain the priority value. The control tasks are classified according to the priority values. Control tasks with an urgency score exceeding the first classification threshold and a complexity score below the second classification threshold are classified as local real-time control tasks, and control tasks with a complexity score exceeding the third classification threshold are classified as remote collaborative control tasks.

[0035] In multi-level intelligent control systems, the rational allocation of control tasks directly affects the system's response efficiency and decision-making quality. By comprehensively evaluating the instantaneous response state variables and trend prediction state variables, intelligent division of control tasks between local execution and remote collaboration can be achieved.

[0036] For the urgency quantification of immediate response state variables, deviation data between the current state parameters and the desired operating state is collected. Taking temperature control as an example, when the deviation between the actual temperature value and the set temperature value exceeds the specified range, the absolute value of the deviation is calculated as the deviation amplitude index. Simultaneously, time-series data of state variables are acquired through continuous sampling, and the ratio of the difference between adjacent state variables to the time interval is calculated to obtain the rate of change index. The deviation amplitude reflects the severity of the current state deviating from the normal range, while the rate of change reflects the speed of state deterioration. The deviation amplitude is mapped to a scoring range of 0 to 10, with a higher score for a larger deviation; similarly, the rate of change is mapped to a corresponding scoring range, with a higher score for a faster change. By weighted summing of the deviation amplitude score and the rate of change score, where the deviation amplitude weight can be set to 0.6 and the rate of change weight to 0.4, a comprehensive urgency score is obtained. This quantification method can accurately identify control tasks requiring immediate intervention.

[0037] To quantify the complexity of trend prediction state variables, it's necessary to analyze the multi-dimensional information involved in the state evolution process. First, the number of associated parameters affecting the state variable's evolution is statistically analyzed, including directly influencing parameters and indirectly coupled parameters. For example, in industrial production line control, product quality status is influenced by multiple parameters such as raw material composition, processing temperature, processing time, and ambient humidity. A greater number of associated parameters indicates more factors needing to be considered in control decisions, thus increasing complexity. The number of associated parameters is normalized and mapped to a scoring range of 0 to 10. Simultaneously, the uncertainty of the evolution path is assessed by analyzing the dispersion of state evolution in historical data and calculating the standard deviation or coefficient of variation as an uncertainty indicator. Higher uncertainty in the evolution path means more difficult to accurately predict future states, requiring stronger computational and analytical capabilities. This uncertainty indicator is also mapped to a scoring range of 0 to 10. A weighted combination of the associated parameter quantity score and the evolution uncertainty score is then performed, with each score having a weight of 0.5, resulting in a comprehensive complexity score.

[0038] A two-dimensional priority evaluation matrix is ​​constructed based on urgency and complexity scores. This matrix maps control tasks into a two-dimensional evaluation space, with urgency scores on the horizontal axis and complexity scores on the vertical axis. Priority values ​​are calculated using a weighted combination method, with weight coefficients for urgency and complexity set according to actual application requirements. In scenarios requiring rapid response, the urgency weight can be set to 0.7 and the complexity weight to 0.3; in scenarios requiring fine-grained decision-making, the weight can be adjusted to 0.4 for urgency and 0.6 for complexity. The priority value for each control task is obtained by multiplying the urgency score by its weight coefficient and then adding the complexity score multiplied by its weight coefficient. This value comprehensively reflects the urgency and processing difficulty of the task.

[0039] Control tasks are intelligently categorized based on priority scores, with a threshold of 7.5 for the first category, 4.0 for the second, and 6.5 for the third. When a control task's urgency score exceeds 7.5 and its complexity score is below 4.0, it indicates that the task requires immediate processing but has relatively simple decision-making logic, suitable for immediate control on local edge devices. These tasks include simple threshold judgments and single-parameter adjustments, where local devices possess sufficient computing resources and response speed. When a control task's complexity score exceeds 6.5, regardless of urgency, it is classified as a remote collaborative control task. These tasks involve multi-parameter coupling analysis, complex model calculations, or require access to cloud-based knowledge bases. Local devices struggle to make high-quality decisions independently, necessitating uploading task data to a remote control center for collaborative processing leveraging the center's powerful computing capabilities and global information. For tasks with urgency scores below 7.5 and complexity scores between 4.0 and 6.5, dynamic allocation is performed based on the system's current load and communication status, prioritizing local processing to reduce communication latency, and requesting remote collaborative support only when local resources are insufficient.

[0040] This hierarchical control task division mechanism achieves a balanced optimization between response speed and decision quality. Local real-time control tasks respond within milliseconds, avoiding control lag caused by communication delays; remote collaborative control tasks fully utilize cloud computing power and data resources to improve decision accuracy in complex scenarios. The classification threshold can be adaptively adjusted based on performance statistics in actual applications. When local device performance improves, the second classification threshold is appropriately increased to handle more tasks; when network communication quality improves, the third classification threshold is appropriately decreased to increase the proportion of remote collaborative tasks, ensuring the system always operates in its optimal working state.

[0041] In one optional implementation, the generation location and execution path of control commands are determined through a dynamic permission allocation mechanism between edge nodes and the cloud center, based on the execution time requirements of the local real-time control task and the remote collaborative control task. The timeliness requirements of local real-time control tasks are analyzed, the coupling relationship between latency and communication latency is quantified, and local timeliness constraint values ​​are obtained. The timeliness requirements of remote collaborative control tasks are analyzed, the cascading effect of data transmission delay and algorithm deduction delay is quantified, and the remote timeliness constraint value is obtained. Based on the local time constraint value and the remote time constraint value, a dynamic permission allocation mechanism is established between edge nodes and the cloud center. The dynamic permission allocation mechanism monitors the computing resource utilization rate of edge nodes and the network transmission latency of the cloud center, and combines the matching degree between the local time constraint value and the available latency margin of edge nodes and the matching degree between the remote time constraint value and the end-to-end latency of the cloud center to dynamically adjust the allocation ratio of control permissions. Based on the allocation ratio of the dynamic permission allocation mechanism, the generation location and execution path of the control command are determined.

[0042] In industrial control systems employing edge-cloud collaboration, a refined timeliness requirement analysis mechanism is needed to address the varying timeliness characteristics of different types of control tasks. For local real-time control tasks, key time nodes are first identified, including the completion time of sensor data acquisition, the data reception time of edge nodes, the completion time of control algorithm calculation, and the actuator response time. Computational latency and communication latency are not simply additive but exhibit a coupling effect. Specifically, when edge nodes experience high processing loads, the queuing time of data in the input buffer prolongs the manifestation of communication latency. Simultaneously, competition for computing resources under high load conditions also leads to fluctuations in computational latency. By establishing a latency coupling model, collecting measured data within continuous time windows, and statistically analyzing the joint distribution characteristics of communication latency and computational latency, a 95% confidence level upper bound on latency is extracted as the local timeliness constraint value. This constraint value not only reflects the time requirement of a single task execution but also includes the system's latency tolerance margin under load fluctuations.

[0043] For remote collaborative control tasks, the timeliness requirements need to be analyzed to consider longer data transmission links and more complex algorithm derivation processes. Data transmission latency involves uplink latency from edge nodes to the cloud center, switching latency within the cloud data center, and downlink latency from the cloud to the edge. Network congestion, changes in routing hop count, and transmission protocol retransmission mechanisms all affect transmission latency. Algorithm derivation latency depends on the complexity of the optimization algorithm deployed in the cloud center, the size of the input data, and the resource scheduling strategy of the cloud computing cluster. There is a cascading effect between data transmission latency and algorithm derivation latency. Specifically, the uncertainty in data arrival time caused by transmission latency forces cloud algorithms to increase buffering time to ensure data integrity, thereby extending the startup time of algorithm derivation.

[0044] By constructing an end-to-end latency monitoring mechanism, a timestamp is added to each uploaded data packet at the edge node, and the data reception time, algorithm start time, calculation completion time, and result distribution time are recorded at the cloud center, thus fully tracking the entire collaborative control process. Based on historical monitoring data, a predictive model for latency cascading effects is established to comprehensively evaluate the end-to-end latency distribution under different network conditions and algorithm loads, and to determine the remote timeliness constraint value that meets the service quality requirements of remote collaborative control tasks.

[0045] When establishing a dynamic permission allocation mechanism between edge nodes and the cloud center, the core lies in real-time monitoring of system operation status and quantifying matching indicators. For edge nodes, a lightweight resource monitoring module is deployed to obtain CPU utilization, memory usage, and network interface traffic at millisecond sampling intervals. Calculating resource utilization not only reflects the current load level but also requires analyzing its changing trends. The variance of resource utilization over the most recent time period is statistically analyzed using a sliding window to determine whether the load is stable. For the cloud center, distributed network probe nodes are deployed to periodically send probe packets to the edge nodes, measuring round-trip latency and decomposing it into uplink and downlink latency. Monitoring network transmission latency requires eliminating abnormal jitter values ​​and using a median filtering method to extract stable latency features.

[0046] Matching degree calculation is a crucial step in dynamic permission allocation. The matching degree between the local time constraint value and the available latency margin of the edge node is quantified by calculating the ratio of the actual execution latency that the edge node can provide under the current resource occupancy state to the local time constraint value. When the edge node has sufficient resources, the actual available execution latency is much less than the time constraint value, resulting in a high matching degree, indicating that the edge node has the ability to undertake local real-time control tasks. When the edge node has limited resources, the actual execution latency approaches or even exceeds the time constraint value, resulting in a lower matching degree, and it is necessary to consider moving some tasks to the cloud. The matching degree between the remote time constraint value and the end-to-end latency of the cloud center is based on the sum of the measured network transmission latency and the latency inferred by the cloud algorithm, compared with the remote time constraint value. When the end-to-end latency is significantly less than the remote time constraint value, the cloud center has sufficient time margin to handle collaborative control tasks, resulting in a high matching degree. When network latency increases or cloud load rises, causing the end-to-end latency to approach the constraint value, the matching degree decreases. In this case, it is necessary to reduce the amount of tasks undertaken by the cloud and decentralize more control decisions to the edge nodes.

[0047] When dynamically adjusting the allocation ratio of control permissions, a two-way decision-making mechanism is established, setting a matching degree threshold range. When the local matching degree is higher than the upper threshold and the remote matching degree is lower than the lower threshold, the proportion of control tasks undertaken by edge nodes is increased, and some tasks that originally required cloud collaboration are changed to edge autonomous decision-making. When the local matching degree is lower than the lower threshold and the remote matching degree is higher than the upper threshold, the control permissions of the cloud center are increased, and local tasks with less stringent real-time requirements are moved to the cloud for execution, freeing up edge node resources to handle high-priority, real-time control tasks. When both matching degrees are within the threshold range, the current permission allocation ratio is maintained. To avoid system instability caused by frequent permission switching, a hysteresis mechanism is introduced, triggering an adjustment to the permission allocation ratio only when the matching degree change continues for more than a set time window.

[0048] Based on the allocation ratio determined by the dynamic permission allocation mechanism, the generation location and execution path of control commands are clearly defined. For control tasks assigned to edge nodes, control commands are generated locally at the edge nodes, and the execution path involves the edge nodes directly issuing commands to the field actuators without going through the cloud center, ensuring the shortest possible command delivery latency. For control tasks assigned to the cloud center, sensor data is uploaded to the cloud, control commands are generated after algorithm derivation in the cloud, and the execution path involves the cloud center issuing commands to the edge nodes via the communication network, and then the edge nodes forwarding them to the actuators. For complex control tasks requiring edge-cloud collaboration, edge nodes generate preliminary control commands to ensure basic response, while simultaneously uploading status data to the cloud for deep optimization. The optimized commands generated in the cloud are superimposed on the edge commands as corrections, forming a collaborative execution path that improves control accuracy while ensuring real-time performance. Through this dynamic mechanism, the system can flexibly adjust the edge-cloud division of labor according to actual operating conditions, achieving synergistic optimization of timeliness and computing resources.

[0049] In one optional implementation, the control command is transmitted to the distributed pyrolysis processing device through the execution path. Based on the deviation evolution trajectory between the executed state feedback data and the immediate response state quantity and the trend prediction state quantity, a two-way correction mechanism between the edge node and the cloud center is constructed, including: The control command is transmitted to the distributed pyrolysis processing equipment through the execution path, and the status feedback data generated by the distributed pyrolysis processing equipment after executing the control command is received. Extract the actual response state quantity from the state feedback data, calculate the instantaneous deviation between the actual response state quantity and the instantaneous response state quantity, and the trend deviation between the actual response state quantity and the trend prediction state quantity. Based on the time-dimensional change sequence of the instantaneous deviation value and the trend deviation value, a deviation evolution trajectory is constructed, and feature decomposition is performed on the deviation evolution trajectory to identify the transient response deviation component and the steady-state tracking deviation component in the deviation evolution trajectory. Based on the transient response deviation component and the steady-state tracking deviation component, a two-way correction mechanism between the edge node and the cloud center is constructed.

[0050] After control commands are transmitted to the distributed pyrolysis equipment via the planned execution path, the equipment generates real-time status feedback data during execution. This feedback data includes actual measured values ​​of key process parameters such as pyrolysis furnace temperature, pressure, and material flow rate. When collecting this status feedback data, it is necessary to ensure that the data acquisition frequency matches the execution cycle of the control commands, typically set to 0.5 to 2 seconds, to ensure that the dynamic changes in the equipment status are captured. The status feedback data undergoes preliminary processing by the data acquisition module at the edge nodes to remove obvious outliers and noise interference, retaining only valid data that accurately reflects the equipment's operating status.

[0051] When extracting actual response state quantities from state feedback data, different extraction strategies are adopted for different types of process parameters. For temperature parameters, the average value within a continuous time window is extracted as the actual response state quantity; for pressure parameters, the combination of instantaneous peak value and steady-state value is extracted; for flow rate parameters, both cumulative flow rate and instantaneous flow velocity are extracted. This multi-dimensional state quantity extraction method can comprehensively reflect the actual response of the equipment to control commands.

[0052] When calculating the instantaneous deviation between the actual response state quantity and the instantaneous response state quantity, a point-to-point difference calculation method is used. Specifically, for the pyrolysis furnace temperature parameter, if the predicted instantaneous response state quantity temperature is 850℃, while the actual measured temperature is 847℃, the instantaneous deviation value is 3℃. For multiple monitoring points, the instantaneous deviation value for each point is calculated separately, and different weighting coefficients are assigned to each point according to its importance in the process flow, resulting in a weighted instantaneous deviation value. The weighting coefficients are set based on the degree of influence of process parameters on the final product quality; the temperature deviation weight for critical reaction areas is typically set to 0.8 to 1.0, and the weight for auxiliary areas is set to 0.3 to 0.5.

[0053] The calculation of trend deviation involves a comparative analysis of the actual response state quantity and the trend prediction state quantity. The trend prediction state quantity reflects the state evolution trend predicted over a future period based on historical operating data, typically with a prediction time span of 30 to 300 seconds. The trend deviation value is obtained by calculating the difference between the actual response state quantity at the same time point and the trend prediction state quantity. This deviation value reflects the degree of deviation between the actual operating trajectory of the equipment and the expected evolution trend. When the trend deviation value continues to increase, it indicates that the equipment's operating state is deviating from the expected stable trajectory, requiring timely adjustment of the control strategy.

[0054] Instantaneous and trend deviation values ​​are continuously recorded over time to form a sequence of deviation data. A time window of 60 to 600 seconds is set, and deviation data is recorded at fixed sampling intervals within this window, forming a deviation evolution trajectory. This trajectory visually demonstrates the dynamic evolution of the equipment's response characteristics over time. For multiple monitoring parameters, their respective deviation evolution trajectories are constructed and compared on the same time axis to identify the correlation response characteristics between different parameters.

[0055] When performing feature decomposition on the deviation evolution trajectory, a joint time-frequency domain analysis method is employed. In the time domain, abrupt change points and inflection points of the deviation curve are identified. Abrupt change points correspond to the instantaneous response stage after the device receives control commands, while inflection points correspond to the transition stage from transient response to steady-state tracking. In the frequency domain, spectral analysis of the deviation evolution trajectory is performed. High-frequency components primarily reflect the transient response characteristics of the device, while low-frequency components reflect the slow drift trend during steady-state tracking. By setting a cutoff frequency threshold, the deviation evolution trajectory is decomposed into high-frequency transient response deviation components and low-frequency steady-state tracking deviation components.

[0056] The transient response deviation component reflects the device's rapid response capability in the initial stage of receiving control commands. The amplitude of this component reflects whether the response speed meets expectations, while its oscillation characteristics reflect the system's damping characteristics and stability. When the peak value of the transient response deviation component exceeds a set threshold, or the oscillation period is too long, it indicates that the response speed configuration of the execution path is unreasonable, or the parameters of the control command are set too aggressively, requiring immediate correction at the edge node. Based on the characteristics of the transient response deviation component, the edge node autonomously adjusts the execution strength and rate parameters of the control command to suppress overshoot and response oscillations.

[0057] The steady-state tracking deviation component reflects the persistent deviation between the actual state variables and the target state variables after the equipment enters steady-state operation. The magnitude of this component characterizes the control accuracy, and its changing trend characterizes the long-term stability of the system. When the steady-state tracking deviation component exhibits a monotonically increasing trend, it indicates a systematic deviation between the process model parameters and the actual equipment characteristics, requiring in-depth calibration of the model parameters at the cloud center. The cloud center collects steady-state tracking deviation data reported by multiple edge nodes, performs global model identification and parameter optimization, and updates the key coefficients of the trend prediction model.

[0058] Based on the characteristic differences between transient response deviation components and steady-state tracking deviation components, a bidirectional correction mechanism between edge nodes and the cloud center is constructed. Edge nodes are responsible for rapidly correcting transient response deviations, completing the correction action on a millisecond to second timescale through localized control parameter adjustments. Specific correction strategies include adjusting the execution gain coefficient of control commands, modifying communication latency compensation parameters of execution paths, and optimizing the collaborative timing configuration between multiple devices. The correction decisions of edge nodes are based on a pre-defined fuzzy rule base or a simplified neural network model, eliminating the need for real-time communication with the cloud center and ensuring the real-time nature of the correction response.

[0059] The cloud center is responsible for correcting steady-state tracking deviations by deeply adjusting the process model, prediction model, and decision-making strategy through a global optimization algorithm. The cloud center aggregates and analyzes the deviation evolution data reported by each edge node on a timescale of minutes to hours, identifying systemic model errors and control strategy defects. Corrections include updating the calculation model parameters of the immediate response state variables, optimizing the time-series prediction weights of the trend prediction state variables, and adjusting the objective function trade-off coefficients in the control command generation strategy. After completing the corrections, the cloud center distributes the updated model parameters and strategy rules to each edge node, enabling the distributed application of the global optimization results. This two-way correction mechanism forms a collaborative closed loop of "rapid edge response and deep cloud optimization."

[0060] In one optional implementation, based on the time-series changes of the instantaneous deviation value and the trend deviation value, a deviation evolution trajectory is constructed. Feature decomposition is then performed on the deviation evolution trajectory to identify the transient response deviation component and the steady-state tracking deviation component within the trajectory, including: The time-series sampling of the change sequences of instantaneous deviation values ​​and trend deviation values ​​in the time dimension is performed, and the change sequences of instantaneous deviation values ​​and trend deviation values ​​are segmented by setting a time window to obtain deviation sequence segments of multiple time periods. The deviation sequence segments from multiple time periods are continuously spliced ​​together on the time axis. By fitting the continuous change trend of the deviation sequence segments, a deviation evolution trajectory is constructed. The deviation evolution trajectory includes the instantaneous deviation evolution curve corresponding to the instantaneous deviation value and the trend deviation evolution curve corresponding to the trend deviation value. The deviation evolution trajectory is decomposed in the frequency domain, and the instantaneous deviation evolution curve and the trend deviation evolution curve are respectively decomposed into a set of frequency components; Frequency components whose frequency changes exceed the frequency boundary threshold are extracted from the set of frequency components as transient response deviation components, and frequency components whose frequency changes are below the frequency boundary threshold are extracted as steady-state tracking deviation components.

[0061] like Figure 2 As shown, the method includes: In practical implementation, when sampling the time-series changes of instantaneous deviation and trend deviation values, an appropriate sampling interval needs to be set according to the time resolution of the data acquisition system. For example, in an industrial production line monitoring scenario, when the data update frequency of the acquisition device is 10 times per second, a sampling interval of 100 milliseconds can be set to ensure that every change in the deviation value is accurately recorded. During sampling, precise alignment of timestamps is maintained, and instantaneous deviation and trend deviation values ​​are recorded according to the same time scale, forming two parallel time-series datasets. For any missing data, linear interpolation is used to fill in the missing values, ensuring the continuity and integrity of the time series.

[0062] When segmenting a changing sequence by setting a time window, the length of the time window needs to comprehensively consider the periodic characteristics of the deviation change and the actual application requirements. In a temperature control system, if the typical response time of temperature change is 30 to 60 seconds, the time window length can be set to 90 seconds to ensure that each window can cover the complete change cycle. A sliding window strategy is used for segmentation, with the window sliding step set to 50% of the window length, meaning the window moves forward by 45 seconds each time, with a 45-second overlap between adjacent windows. This overlap design avoids losing important deviation change information at window boundaries. The instantaneous deviation value sequence and trend deviation value sequence within each time window are extracted separately to obtain deviation sequence segments corresponding to that time period. For example, in the first window from 0 to 90 seconds, the instantaneous deviation values ​​of all sampling points within that time period are extracted to form the first instantaneous deviation sequence segment, and the trend deviation values ​​within the same time period are extracted to form the first trend deviation sequence segment. Subsequent time periods are processed sequentially according to the sliding window strategy, ultimately resulting in a set of deviation sequence segments for multiple time periods.

[0063] When continuously splicing deviation sequence segments from multiple time periods along a time axis, it is necessary to address the numerical continuity of adjacent segments in overlapping regions. For data points within overlapping regions, a weighted average method is used for fusion. The weight of data at the end of the previous window gradually decreases, while the weight of data at the beginning of the next window gradually increases, with a linear transition in weight changes. Specifically, at the i-th sampling point in the overlapping region, the weight of the data in the previous window is set to 1-i / N, and the weight of the data in the next window is set to i / N, where N represents the total number of sampling points in the overlapping region. This weighted fusion method eliminates abrupt changes at the segment splicing points, achieving a smooth transition in the deviation sequence. After splicing, the continuously spliced ​​deviation value sequence is fitted. Cubic spline interpolation is used to fit the continuous trend of instantaneous deviation values. This method can accurately reflect the local variation characteristics of deviation values ​​while ensuring a smooth curve. For fitting the trend deviation values, a piecewise polynomial fitting method is used. The choice between quadratic or cubic polynomials depends on the smoothness of the trend deviation value changes. Cubic polynomials are used in areas of rapid change to improve fitting accuracy, while quadratic polynomials are used in areas of gradual change to avoid overfitting. During the fitting process, the polynomial coefficients are determined using the least squares method to minimize the sum of squares of the deviations between the fitted curve and the actual sampling points. Through the above fitting process, an instantaneous deviation evolution curve describing the change of instantaneous deviation values ​​and a trend deviation evolution curve describing the change of trend deviation values ​​are constructed. The two curves together constitute a complete deviation evolution trajectory.

[0064] When performing frequency domain decomposition on the deviation evolution trajectory, a Fourier transform is used to convert the time-domain signal to the frequency domain. A Fast Fourier Transform (FFT) is applied to the instantaneous deviation evolution curve, decomposing it into a linear combination of sine and cosine waves of different frequencies. The resulting spectrum is a graph where the horizontal axis represents frequency magnitude and the vertical axis represents the amplitude of the corresponding frequency component. All frequency components with amplitudes exceeding the noise threshold are extracted from the spectrum; each frequency component contains three parameters: frequency value, amplitude, and phase. The same Fourier transform is applied to the trend deviation evolution curve to obtain the set of frequency components corresponding to the trend deviation. In the actual decomposition process, to improve frequency domain resolution, zero-padding is performed on the deviation evolution curve, adding zero-value data points at the end of the curve to ensure the total number of data points reaches an integer power of 2, facilitating efficient calculation of the FFT.

[0065] When extracting transient response deviation components and steady-state tracking deviation components from the frequency component set, a reasonable frequency boundary threshold needs to be set. The determination of the frequency boundary threshold is based on the dynamic response characteristics of the controlled object. For systems with fast response speeds, the transient process duration is short, and the corresponding frequency components are high; the frequency boundary threshold can be set to 5 Hz to 10 Hz. For large inertial systems with slow response speeds, the transient process duration is long, and the frequency boundary threshold can be reduced to 0.5 Hz to 2 Hz. Taking 3 Hz as an example, traversing the frequency component set of the instantaneous deviation evolution curve, all frequency components with a change frequency greater than 3 Hz are classified as transient response deviation components. These high-frequency components reflect the rapid fluctuation characteristics of the deviation in a short time, corresponding to the rapid response stage after the system is disturbed. Frequency components with a change frequency less than or equal to 3 Hz are classified as steady-state tracking deviation components. These low-frequency components reflect the slow change trend of the deviation, reflecting the long-term tracking performance of the system under stable operating conditions. The same classification process is performed on the frequency component set of the trend deviation evolution curve to extract its transient response deviation components and steady-state tracking deviation components. After extraction, an inverse Fourier transform can be performed on the transient response deviation component to reconstruct the pure transient response deviation signal, intuitively demonstrating the system's rapid response process after being disturbed. Similarly, an inverse transform is performed on the steady-state tracking deviation component to reconstruct the steady-state tracking deviation signal, clearly presenting the tracking error variation law of the system during long-term operation. Through this frequency domain eigenvalue decomposition method, a deep analysis of the deviation evolution trajectory is achieved, providing a technical foundation for subsequent differentiated control strategies for different deviation components.

[0066] A second aspect of the present invention provides a remote control system for waste rubber pyrolysis treatment based on the Internet of Things, comprising: The state analysis unit is used to collect the operating state data of the distributed pyrolysis processing equipment, perform spatiotemporal correlation analysis on the operating state data, identify rapid fluctuation characteristics and slow drift characteristics by constructing a multi-scale time axis of state evolution, and generate hierarchical state representation based on the coupling relationship between the rapid fluctuation characteristics and the slow drift characteristics to obtain the instantaneous response state quantity and trend prediction state quantity. The task division unit is used to determine the priority of the control task based on the real-time response state quantity and the trend prediction state quantity, and to divide the control task into local real-time control tasks and remote collaborative control tasks of the pyrolysis production line. The permission allocation unit is used to determine the generation location and execution path of control commands through a dynamic permission allocation mechanism between edge nodes and the cloud center, based on the execution time requirements of the local real-time control task and the remote collaborative control task. The feedback correction unit is used to transmit the control command to the distributed pyrolysis processing equipment through the execution path, and construct a two-way correction mechanism between the edge node and the cloud center based on the deviation evolution trajectory of the post-execution status feedback data, the instantaneous response status quantity and the trend prediction status quantity, and adaptively reconstruct the allocation ratio of the dynamic permission allocation mechanism and the scale division parameters of the multi-scale time axis.

[0067] A third aspect of the present invention provides a control system electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0068] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0069] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0070] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A remote control method for waste rubber pyrolysis treatment based on the Internet of Things, characterized in that, include: The system collects operational status data of distributed pyrolysis equipment, performs spatiotemporal correlation analysis on the operational status data, constructs a multi-scale time axis of state evolution, identifies rapid fluctuation characteristics and slow drift characteristics, and generates hierarchical state representation based on the coupling relationship between the rapid fluctuation characteristics and the slow drift characteristics to obtain instantaneous response state quantity and trend prediction state quantity. The priority of the control task is determined based on the real-time response state quantity and the trend prediction state quantity, and the control task is divided into local real-time control task and remote collaborative control task of the pyrolysis production line. Based on the execution time requirements of the local real-time control task and the remote collaborative control task, the generation location and execution path of the control command are determined through the dynamic permission allocation mechanism between the edge node and the cloud center. The control commands are transmitted to the distributed pyrolysis processing equipment through the execution path. Based on the deviation evolution trajectory of the status feedback data after execution, the instantaneous response status quantity, and the trend prediction status quantity, a two-way correction mechanism between the edge node and the cloud center is constructed. The allocation ratio of the dynamic permission allocation mechanism and the scale division parameters of the multi-scale time axis are adaptively reconstructed.

2. The method according to claim 1, characterized in that, By constructing a multi-scale time axis for state evolution, rapid fluctuation characteristics and slow drift characteristics are identified. Based on the coupling relationship between the rapid fluctuation characteristics and the slow drift characteristics, a hierarchical state representation is generated, resulting in instantaneous response state quantities and trend prediction state quantities, including: A fast-scale time axis and a slow-scale time axis are constructed for the operating status data. Rapid fluctuation characteristics are identified by capturing instantaneous state jump information on the fast-scale time axis, and slow drift characteristics are identified by tracking the long-term evolution trajectory of the state on the slow-scale time axis. A modulation mapping model of the fast fluctuation feature on the slow drift feature is established. By analyzing the dependence between the amplitude change of the fast fluctuation feature and the evolution stage of the slow drift feature, the transfer coupling characteristics between scales are extracted to obtain the coupling transfer function. Based on the coupling transfer function, the rapid fluctuation feature and the slow drift feature are projected onto a unified state representation space, and an instantaneous response dimension and a trend prediction dimension are constructed in the state representation space. Based on the orthogonal decomposition relationship between the instant response dimension and the trend prediction dimension, instant response state variables and trend prediction state variables are generated.

3. The method according to claim 2, characterized in that, A modulation mapping model of the rapid fluctuation feature on the slow drift feature is established. By analyzing the dependency between the amplitude change of the rapid fluctuation feature and the evolution stage of the slow drift feature, the inter-scale transfer coupling characteristics are extracted, and the coupling transfer function is obtained, including: The amplitude of the rapid fluctuation feature is demodulated, the amplitude envelope curve of the rapid fluctuation feature is extracted, and a time synchronization correspondence is established between the amplitude envelope curve and the evolution trajectory of the slow drift feature. Based on the time synchronization correspondence, the local gradient and curvature of the amplitude envelope curve at different evolution positions of the slow drift feature are calculated, and the deformation mode of the amplitude envelope curve as the slow drift feature evolves is identified. The evolution trajectory of the slow drift feature is divided into an accelerated evolution interval, a uniform evolution interval, and a decelerated evolution interval according to the state space gradient. The deformation mode features of the amplitude envelope curve are statistically analyzed in each evolution interval to establish a modulation mapping model. The modulation mapping model is used to quantify the dependence of the amplitude change of the fast fluctuation feature on the evolution stage of the slow drift feature, and the transmission relationship of the dependence strength between different evolution intervals is extracted to obtain the inter-scale transmission coupling characteristics. The coupling characteristic is expressed as a response function of the amplitude of the rapid fluctuation feature to the evolution stage of the slow drift feature, thus obtaining the coupling transfer function.

4. The method according to claim 1, characterized in that, The priority of the control task is determined based on the real-time response state quantity and the trend prediction state quantity, and the control task is divided into local real-time control tasks and remote collaborative control tasks for the pyrolysis production line, including: The urgency of the immediate response state quantity is quantified by evaluating the deviation magnitude and rate of change of the immediate response state quantity to obtain an urgency score; the complexity of the trend prediction state quantity is quantified by evaluating the number of associated parameters and evolution uncertainty involved in the evolution path of the trend prediction state quantity to obtain a complexity score. Based on the urgency score and the complexity score, a priority evaluation matrix for the control task is constructed. The priority evaluation matrix is ​​calculated by weighted combination of the urgency score and the complexity score to obtain the priority value. The control tasks are classified according to the priority values. Control tasks with an urgency score exceeding the first classification threshold and a complexity score below the second classification threshold are classified as local real-time control tasks, and control tasks with a complexity score exceeding the third classification threshold are classified as remote collaborative control tasks.

5. The method according to claim 1, characterized in that, Based on the execution time requirements of the local real-time control task and the remote collaborative control task, the generation location and execution path of the control command are determined through a dynamic permission allocation mechanism between edge nodes and the cloud center, including: For local real-time control tasks, timeliness requirements are analyzed, the coupling relationship between calculation latency and communication latency is quantified, and local timeliness constraints are obtained; for remote collaborative control tasks, timeliness requirements are analyzed, the cascading effect of data transmission latency and algorithm deduction latency is quantified, and remote timeliness constraints are obtained. Based on the local time constraint value and the remote time constraint value, a dynamic permission allocation mechanism is established between edge nodes and the cloud center. The dynamic permission allocation mechanism monitors the computing resource utilization rate of edge nodes and the network transmission latency of the cloud center, and combines the matching degree between the local time constraint value and the available latency margin of edge nodes and the matching degree between the remote time constraint value and the end-to-end latency of the cloud center to dynamically adjust the allocation ratio of control permissions. Based on the allocation ratio of the dynamic permission allocation mechanism, the generation location and execution path of the control command are determined.

6. The method according to claim 1, characterized in that, The control commands are transmitted to the distributed pyrolysis processing equipment through the execution path. Based on the deviation evolution trajectory between the executed state feedback data and the instantaneous response state quantity and the trend prediction state quantity, a two-way correction mechanism between the edge node and the cloud center is constructed, including: The control command is transmitted to the distributed pyrolysis processing equipment through the execution path, and the status feedback data generated by the distributed pyrolysis processing equipment after executing the control command is received. Extract the actual response state quantity from the state feedback data, calculate the instantaneous deviation between the actual response state quantity and the instantaneous response state quantity, and the trend deviation between the actual response state quantity and the trend prediction state quantity. Based on the time-dimensional change sequence of the instantaneous deviation value and the trend deviation value, a deviation evolution trajectory is constructed, and feature decomposition is performed on the deviation evolution trajectory to identify the transient response deviation component and the steady-state tracking deviation component in the deviation evolution trajectory. Based on the transient response deviation component and the steady-state tracking deviation component, a two-way correction mechanism between the edge node and the cloud center is constructed.

7. The method according to claim 6, characterized in that, Based on the time-series changes of the instantaneous deviation value and the trend deviation value, a deviation evolution trajectory is constructed. Feature decomposition is performed on the deviation evolution trajectory to identify the transient response deviation component and the steady-state tracking deviation component, including: The time-series sampling of the change sequences of instantaneous deviation values ​​and trend deviation values ​​in the time dimension is performed, and the change sequences of instantaneous deviation values ​​and trend deviation values ​​are segmented by setting a time window to obtain deviation sequence segments of multiple time periods. The deviation sequence segments from multiple time periods are continuously spliced ​​together on the time axis. By fitting the continuous change trend of the deviation sequence segments, a deviation evolution trajectory is constructed. The deviation evolution trajectory includes the instantaneous deviation evolution curve corresponding to the instantaneous deviation value and the trend deviation evolution curve corresponding to the trend deviation value. The deviation evolution trajectory is decomposed in the frequency domain, and the instantaneous deviation evolution curve and the trend deviation evolution curve are respectively decomposed into a set of frequency components; Frequency components whose frequency changes exceed the frequency boundary threshold are extracted from the set of frequency components as transient response deviation components, and frequency components whose frequency changes are below the frequency boundary threshold are extracted as steady-state tracking deviation components.

8. A remote control system for waste rubber pyrolysis treatment based on the Internet of Things, used to implement the method of any one of claims 1-7, characterized in that, include: The state analysis unit is used to collect the operating state data of the distributed pyrolysis processing equipment, perform spatiotemporal correlation analysis on the operating state data, identify rapid fluctuation characteristics and slow drift characteristics by constructing a multi-scale time axis of state evolution, and generate hierarchical state representation based on the coupling relationship between the rapid fluctuation characteristics and the slow drift characteristics to obtain the instantaneous response state quantity and trend prediction state quantity. The task division unit is used to determine the priority of the control task based on the real-time response state quantity and the trend prediction state quantity, and to divide the control task into local real-time control tasks and remote collaborative control tasks of the pyrolysis production line. The permission allocation unit is used to determine the generation location and execution path of control commands through a dynamic permission allocation mechanism between edge nodes and the cloud center, based on the execution time requirements of the local real-time control task and the remote collaborative control task. The feedback correction unit is used to transmit the control command to the distributed pyrolysis processing equipment through the execution path, and construct a two-way correction mechanism between the edge node and the cloud center based on the deviation evolution trajectory of the post-execution status feedback data, the instantaneous response status quantity and the trend prediction status quantity, and adaptively reconstruct the allocation ratio of the dynamic permission allocation mechanism and the scale division parameters of the multi-scale time axis.

9. A control system electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images