Monitoring method and system based on digital twin remote intelligent protection device

Abstract

The invention provides a monitoring method and system based on a digital twin remote intelligent protection device, and the method comprises the steps: carrying out the synchronous processing and feature extraction of operation data, generating cross-device space-time alignment data, and enabling the operation data to be multi-source heterogeneous data collected from a plurality of devices; based on the space-time alignment data, constructing a multi-dimensional association relationship among the plurality of devices, the multi-dimensional association relationship being used for representing a potential fault propagation path among the devices; performing simulation verification on the multi-dimensional association relationship through a digital twinborn model, and determining root cause equipment and a propagation path of the implicit association fault; and generating a cooperative control instruction according to the root cause device and the propagation path, wherein the cooperative control instruction is used for indicating the root cause device to execute protection operation. By adopting the method, the hidden fault missing detection problem caused by energy transfer or control dependence between the devices can be explored based on the multi-dimensional incidence relation between the devices.

Description

Technical Field

[0001] The present invention belongs to the technical field of digital twins, and particularly relates to a monitoring method and system for a digital twin remote intelligent protection device. Background Art

[0002] With the rapid development of industrial Internet of Things and digital twin technologies, the device status monitoring system has gradually evolved from single-point threshold alarm to multi-dimensional predictive maintenance. Existing technologies usually implement single-device anomaly detection based on sensor data and machine learning algorithms, and simulate device behavior through digital twin models. However, in complex industrial scenarios, devices are closely coupled due to mechanical connection, energy transfer, or control logic. A local fault in a single device may trigger systemic risks through cross-device association paths. Due to the lack of dynamic modeling ability for implicit associations between devices, traditional methods can only rely on single-point thresholds or static rules to judge anomalies, unable to quantify the propagation paths of faults among multiple devices, and difficult to distinguish the root cause device from secondary affected devices, resulting in frequent false alarms and missed detections, and unable to trace the root cause device. In addition, problems such as time-space asynchrony of multi-source heterogeneous data, cross-protocol communication barriers, and sensitivity to dynamic environmental parameters further limit the accuracy and real-time performance of multi-device collaborative monitoring.

[0003] However, the current monitoring method for digital twin remote intelligent protection devices has the problem of insufficient ability to detect cross-device implicit association faults, and it is difficult to meet the active protection requirements for implicit association faults in industrial scenarios. Summary of the Invention

[0004] Based on this, it is necessary to provide a monitoring method and system for a digital twin remote intelligent protection device aiming at the above technical problems, which can discover the problem of missed detection of implicit faults caused by energy transfer or control dependence between devices based on the multi-dimensional association relationship between devices.

[0005] In a first aspect, the present application provides a monitoring method for a digital twin remote intelligent protection device, including:

[0006] Performing synchronization processing and feature extraction on operation data to generate spatio-temporal alignment data across devices, where the operation data is multi-source heterogeneous data collected from multiple devices;

[0007] Based on the spatio-temporal alignment data, constructing a multi-dimensional association relationship between multiple devices, where the multi-dimensional association relationship is used to represent potential fault propagation paths between devices;

[0008] Verifying the multi-dimensional association relationship through a digital twin model to determine the root cause device and propagation path of the implicit association fault;

[0009] Generate collaborative control instructions according to the root cause device and the propagation path, where the collaborative control instructions are used to instruct the root cause device to perform protection operations.

[0010] In a possible embodiment, perform synchronization processing and feature extraction on the operation data to generate spatio-temporal alignment data across devices, including:

[0011] Collect vibration signals through vibration sensors deployed on multiple devices;

[0012] Perform wavelet denoising processing on the vibration signals to generate denoised vibration signals;

[0013] Through the dynamic time window alignment algorithm, unify the timestamps of the operation data of multiple devices to the same time sequence reference to generate spatio-temporal alignment data.

[0014] In a possible embodiment, based on the spatio-temporal alignment data, construct multi-dimensional association relationships between multiple devices, including:

[0015] Based on the mechanical connection structure between devices, construct a physical transmission topology relationship;

[0016] Perform coherence analysis on the denoised vibration signals to generate an energy transfer relationship between devices;

[0017] Analyze the control signals of the devices to generate a control dependency relationship;

[0018] Fuse the physical transmission topology relationship, the energy transfer relationship, and the control dependency relationship to generate a multi-dimensional association relationship.

[0019] In a possible embodiment, use a digital twin model to simulate and verify the multi-dimensional association relationship to determine the root cause device and the propagation path of the latent associated fault, including:

[0020] Analyze the operation data between multiple devices based on the information entropy theory, calculate the causal influence intensity, and determine candidate device pairs based on a decision threshold, where the candidate device pairs include a first device and a second device;

[0021] Inject the fault parameters of the first device in the candidate device pair into the digital twin model, simulate the operation state of the second device in the candidate device pair, and obtain simulation data;

[0022] When the error between the simulation data and the actual monitoring data of the second device is less than the preset tolerance, determine the first device as the root cause device and generate a propagation path.

[0023] In a possible embodiment, when generating collaborative control instructions according to the root cause device and the propagation path, it further includes:

[0024] Construct a collaborative decision-making model based on device health score and propagation path risk;

[0025] Optimize the collaborative decision-making model through a reinforcement learning algorithm to generate collaborative control instructions that maximize the system's benefits;

[0026] Send the collaborative control instructions to the root cause device and obtain the return signal, where the return signal is used to characterize the execution effect of the control instructions; update the digital twin model parameters based on the return signal.

[0027] In a possible embodiment, analyze the operation data between multiple devices based on the information entropy theory, calculate the causal influence intensity, and determine the candidate device pairs based on the decision threshold, including:

[0028] Calculate the transfer entropy value between devices based on the historical state sequence of the devices, where the transfer entropy value is used to quantify the information gain of the state of the first device on the future state of the second device;

[0029] Dynamically adjust the decision threshold of the causal influence intensity according to the causal strength distribution of the device pairs in the historical fault data;

[0030] Add the device pairs with transfer entropy values exceeding the decision threshold to the candidate device pair list to generate candidate device pairs.

[0031] In a possible embodiment, optimize the collaborative decision-making model through a reinforcement learning algorithm to generate collaborative control instructions that maximize the system's benefits, including:

[0032] Define the reward function of the collaborative decision-making model, where the reward function is obtained based on the balance relationship between system risk and production capacity;

[0033] Iteratively update the action value function through the Q-learning algorithm to optimize the control strategy of the collaborative decision-making model;

[0034] Generate collaborative control instructions according to the control strategy, where the collaborative control instructions include isolation instructions for the root cause device and derating operation parameters for associated devices, and the associated devices are the devices on the propagation path of the root cause device;

[0035] Obtain the feedback signal, where the feedback signal is used to indicate the digital twin model to update the association relationship parameters.

[0036] In a second aspect, the present application also provides a monitoring system based on a digital twin remote intelligent protection device, including:

[0037] A data preprocessing module for synchronizing and extracting features from the operation data to generate spatio-temporal aligned data across devices, where the operation data is multi-source heterogeneous data collected from multiple devices;

[0038] An association establishment module, configured to construct multi-dimensional association relationships between multiple devices based on spatio-temporal alignment data, where the multi-dimensional association relationships are used to represent potential fault propagation paths between devices;

[0039] A root cause determination module, configured to perform simulation verification on the multi-dimensional association relationships through a digital twin model, and determine the root cause device and propagation path of latent associated faults;

[0040] A collaborative control module, configured to generate collaborative control instructions according to the root cause device and propagation path, where the collaborative control instructions are used to instruct the root cause device to perform protection operations.

[0041] Thirdly, the present application also provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the above-mentioned monitoring method of the digital twin-based remote intelligent protection device is implemented.

[0042] Fourthly, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned monitoring method of the digital twin-based remote intelligent protection device is implemented.

[0043] For the above-mentioned monitoring method and system of the digital twin-based remote intelligent protection device, by synchronously processing and feature extracting multi-source heterogeneous operation data collected from multiple devices, spatio-temporal alignment data across devices is generated; based on this spatio-temporal alignment data, multi-dimensional association relationships between multiple devices are further constructed, and the multi-dimensional association relationships can represent potential fault propagation paths between devices; through simulation verification of these multi-dimensional association relationships by a digital twin model, the root cause device and its propagation path of latent associated faults can be accurately determined; collaborative control instructions are generated according to the determined root cause device and propagation path, instructing the root cause device to perform protection operations. The above technical solutions can deeply explore and solve the problem of missed detection of latent faults caused by energy transfer or control dependence between devices based on the multi-dimensional relationships between devices, thereby improving the fault diagnosis accuracy and protection ability of the monitoring of the intelligent protection device. Description of the Drawings

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for the description of the embodiments or related technologies. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0045] Figure 1 It is a flowchart of a monitoring method of a digital twin-based remote intelligent protection device provided by an embodiment of the present invention;

[0046] Figure 2 This is a schematic structural diagram of a monitoring system for a digital twin-based remote intelligent protection device provided by an embodiment of the present invention. Specific implementation manners

[0047] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0048] First, a brief introduction is given to the nouns involved in the embodiments of the present application.

[0049] The Dynamic Time Warping (DTW) algorithm is an algorithm used to measure the similarity between two time series, especially suitable for cases where the lengths of the time series are different or there are stretching and shrinking changes on the time axis. It uses dynamic programming to establish an optimal alignment path between time series, allowing local stretching and compression on the time axis, and thus calculates the minimum cumulative distance between the two series. This method is widely used in fields such as speech recognition, gesture recognition, and financial data analysis, and can effectively solve the non-linear change problem of time series data in the time dimension, providing a powerful tool for the classification, clustering, and similarity analysis of time series.

[0050] Coherence analysis is a statistical method used to study the frequency correlation between two signals. By calculating the phase and amplitude correlation degrees of the two signals at different frequency components, it quantifies the linear dependence relationship between them. The result is usually represented by a coherence function, with a value range between 0 and 1. The closer the value is to 1, the stronger the correlation between the two signals at the corresponding frequency components. This method is widely used in fields such as signal processing, physics, and biomedical engineering. For example, analyzing the synchronization between electroencephalogram signals and studying the correlation of economic time series helps to reveal the potential coupling mechanism and interaction mode between signals.

[0051] The collaborative decision-making model is a theoretical framework for dealing with decision-making problems jointly participated by multiple agents or multiple stakeholders. It designs a decision-making plan that can coordinate the interests of all parties and optimize the overall benefit by integrating the preferences, objectives, and constraints of different agents. This model usually uses methods such as mathematical modeling, optimization algorithms, and game theory to analyze the cooperation and competition relationships between agents and seeks solutions to achieve group optimality or Nash equilibrium in a complex and changing decision-making environment. The collaborative decision-making model is widely used in fields such as supply chain management, intelligent transportation systems, and multi-agent systems, and can effectively improve decision-making efficiency, promote the rational allocation of resources, and the collaborative operation of the system.

[0052] Based on the above-mentioned noun explanations, the implementation environment of the monitoring method for the digital twin remote intelligent protection device provided in the embodiments of the present application is described. Schematically, this implementation environment includes: sensors, processors, and terminals. Among them, the processor is signal-connected to the sensors and terminals through network signals; the sensors are deployed in the device group, and the sensors can be temperature and humidity sensors, vibration sensors, optical cameras, etc.; the processor includes but is not limited to a central processing unit (CPU), a graphics processing unit (GPU), an artificial intelligence chip, etc., which are not limited here.

[0053] Combined with the above-mentioned noun explanations and implementation environment, the application scenarios of the embodiments of the present application are described. The monitoring method for the digital twin remote intelligent protection device provided in the embodiments of the present application can be applied to, but is not limited to, the following scenarios:

[0054] In the field of industrial automation, the stable operation of equipment is crucial for production efficiency and product quality. Through the above-mentioned monitoring method for the digital twin remote intelligent protection device, the operation data of multiple devices in the factory can be monitored in real time; after synchronously processing and feature extracting the operation data, spatio-temporal alignment data across devices is generated, and then a multi-dimensional correlation relationship between devices is constructed; using the digital twin model to simulate and verify these relationships can accurately identify potential fault propagation paths between devices, discover hidden faults in advance, and generate collaborative control instructions, so as to achieve timely protection of key devices, reduce downtime, and improve production efficiency.

[0055] In an automobile manufacturing factory, multiple welding robots, transfer robotic arms, and assembly equipment need to cooperate. If one of the welding robots generates abnormal vibration due to gear wear, the vibration may be transmitted to the adjacent transfer robotic arm through the equipment base, resulting in a decrease in its positioning accuracy. Through this technical solution, the vibration and current signals of each device can be collected in real time, a physical connection and energy transfer relationship model can be constructed, and combined with digital twin simulation verification, the fault source can be quickly located as the welding robot rather than the transfer robotic arm; and instructions are issued to isolate the faulty robot and start the standby equipment to avoid the entire production line from stopping and ensure production efficiency.

[0056] In the field of medical and health, digital twin technology can be used for remote monitoring and diagnosis. By collecting various physiological data of patients (such as heart rate, blood pressure, blood sugar, etc.) and spatio-temporally aligning them with the operation data of medical equipment, a multi-dimensional correlation relationship between the patient's health status and the equipment is constructed; using the digital twin model to simulate and analyze these relationships can discover potential health problems in advance, such as abnormal monitoring data caused by equipment failures or hidden changes in the patient's physical indicators, and the generated collaborative control instructions can notify medical staff in a timely manner to take measures to achieve precise medical treatment and remote monitoring.

[0057] Schematically, the monitoring method of the digital twin-based remote intelligent protection device provided in the embodiments of the present application can also be applied to other application scenarios. Only examples are given here, and the specific application scenarios are not limited.

[0058] In an exemplary embodiment, as Figure 1 shown, a monitoring method of a digital twin-based remote intelligent protection device is provided. In this embodiment, an example is given where this method is applied to the terminal in the foregoing implementation environment. It can be understood that this method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps 101 to step 104:

[0059] Step 101, perform synchronization processing and feature extraction on the operation data to generate cross-device spatio-temporal alignment data, where the operation data is multi-source heterogeneous data collected from multiple devices.

[0060] Specifically, perform synchronization processing and feature extraction on the multi-source heterogeneous operation data collected from multiple devices. Exemplarily, the operation data includes, but is not limited to, information such as the temperature, pressure, and vibration frequency of the device, which come from different types of sensors and monitoring devices. Through data synchronization processing, data with different timestamps and formats are aligned to the same time axis, and key features are extracted to generate cross-device spatio-temporal alignment data.

[0061] Step 102, based on the spatio-temporal alignment data, construct multi-dimensional association relationships between multiple devices, where the multi-dimensional association relationships are used to characterize potential fault propagation paths between devices.

[0062] Specifically, based on the spatio-temporal alignment data, construct multi-dimensional association relationships between multiple devices. The association relationships are determined by analyzing the data correlation, physical connection, and functional dependence between devices, and can characterize potential fault propagation paths between devices. For example, if the failure of a certain device may affect other devices through mechanical connection or signal transmission, this potential propagation path will be identified and recorded. By constructing multi-dimensional association relationships, the interaction between devices can be comprehensively understood, and the hidden fault propagation links between devices can be accurately captured, breaking through the limitations of single-dimensional analysis.

[0063] Step 103, through the digital twin model, simulate and verify the multi-dimensional association relationships to determine the root cause device and propagation path of the hidden associated faults.

[0064] Specifically, by simulating the operating states and interaction relationships of devices, the dynamic behaviors between devices can be accurately reproduced. Through the simulation and analysis of multi-dimensional correlation relationships, the root cause devices of hidden correlation faults and their propagation paths can be determined. Combining digital twin dynamic simulation with actual data verification can avoid the pseudo-causal misjudgment of traditional data-driven methods and improve the credibility of root cause location.

[0065] Step 104: Generate a collaborative control instruction based on the root cause device and its propagation path. The collaborative control instruction is used to instruct the root cause device to perform a protection operation.

[0066] Specifically, based on the determined root cause device and its propagation path, a collaborative control instruction is generated. This instruction is used to instruct the root cause device to perform corresponding protection operations, such as adjusting operating parameters, starting standby devices, or triggering alarms. The generation of the collaborative control instruction takes into account the interdependent relationships between devices to ensure that the protection measures can effectively prevent the further propagation of faults, achieve the rapid isolation of cross-device faults and the minimization of system-level risks, and ensure production continuity and device safety.

[0067] The above-mentioned monitoring method and system for a digital twin remote intelligent protection device synchronously processes and extracts features from multi-source heterogeneous operating data collected from multiple devices to generate cross-device spatio-temporal alignment data. Based on this spatio-temporal alignment data, a multi-dimensional correlation relationship between multiple devices is further constructed. This multi-dimensional correlation relationship can represent potential fault propagation paths between devices. Through the simulation and verification of these multi-dimensional correlation relationships using a digital twin model, the root cause devices of hidden correlation faults and their propagation paths can be accurately determined. A collaborative control instruction is generated based on the determined root cause device and propagation path to instruct the root cause device to perform a protection operation. The above technical solution can deeply explore and solve the problem of missed detection of hidden faults caused by energy transfer or control dependence between devices based on the multi-dimensional relationships between devices, thereby improving the fault diagnosis accuracy and protection ability of the monitoring of intelligent protection devices.

[0068] In a possible embodiment, synchronously processing and extracting features from the operating data to generate cross-device spatio-temporal alignment data may include:

[0069] Step 201: Collect vibration signals through vibration sensors deployed on multiple devices.

[0070] Specifically, vibration sensors deployed at key monitoring points of multiple devices are used to collect vibration signals during the operation of the devices in real time. The vibration sensors can adopt a high-frequency sampling mode to obtain original vibration waveform data containing device mechanical state characteristics (such as impacts, friction, and imbalances).

[0071] Step 202: Perform wavelet denoising processing on the vibration signals to generate denoised vibration signals.

[0072] Specifically, wavelet threshold denoising is performed on the collected original vibration signals. Exemplarily, the Daubechies wavelet basis function can be selected to perform multi-scale decomposition on the signals, and high-frequency noise components (such as electromagnetic interference and environmental vibration noise) are filtered out by the hard threshold method, while retaining the low-frequency and intermediate-frequency components (such as gear meshing frequency and bearing fault characteristic frequency) that reflect the equipment fault characteristics, generating a denoised vibration signal that can be used for precise analysis.

[0073] Step 203: Through the dynamic time window alignment algorithm, the timestamps of the operation data of multiple devices are unified to the same time sequence reference to generate spatio-temporal alignment data.

[0074] Specifically, based on the time axis of the vibration signal of the master device, the phase offset of the slave device signal is calculated by the dynamic time warping (DTW) algorithm, and the slave device signal is interpolated and resampled within the sliding time window to eliminate the time sequence deviation caused by device clock asynchronization or communication delay, generating cross-device spatio-temporal alignment data to ensure strict synchronization of multi-device vibration signals in the time and space dimensions.

[0075] In a possible embodiment, based on the spatio-temporal alignment data, constructing multi-dimensional association relationships between multiple devices may include:

[0076] Step 301: Based on the mechanical connection structure between devices, construct a physical transmission topology relationship.

[0077] Exemplarily, based on the mechanical connection structure between devices (such as bolt fastening, coupling coupling, pipeline welding, etc.), the spatial positions and connection methods of the devices are extracted through 3D modeling software or industrial site surveying data, and according to the material properties (such as elastic modulus and damping coefficient) and geometric parameters (such as contact area and connection length) of the connecting components, the physical transmission stiffness coefficient between devices is calculated to construct a physical transmission topology relationship network, where the topology relationship network takes devices as nodes and the stiffness coefficient as the edge weight, characterizing the transmission path and intensity of mechanical vibration energy between devices. By quantifying the physical connection stiffness between devices, the potential transmission path of mechanical vibration energy is accurately reflected, providing a physical basis for cross-device fault propagation analysis.

[0078] Step 302: Perform coherence analysis on the denoised vibration signals to generate the energy transfer relationship between devices.

[0079] Specifically, perform coherence analysis on the denoised vibration signals to generate the energy transfer relationship between devices. Through coherence analysis, calculate the frequency correlation between the vibration signals of different devices, quantify the intensity and direction of energy transfer between devices, so as to reveal the potential associations formed between devices due to vibration energy transfer, and help identify the fault propagation paths that may be caused by energy coupling.

[0080] Step 303: Analyze the control signals of the devices to generate control dependency relationships.

[0081] Specifically, analyze the control signals of the devices (such as PLC program instructions, DCS system interlock logic), extract the control dependency rules between the devices, which may specifically include: interlock logic analysis, for example, identifying whether the fault shutdown signal of device A triggers the interlock shutdown of device B; priority determination, analyzing the priority order of the control instructions, for example, the emergency shutdown instruction overrides the normal operation instruction; timing correlation modeling, constructing the timing dependency graph of the control instructions, and marking the control response delay between the devices. Through the analysis of the control logic, reveal the fault conduction risks caused by program interlocks or timing dependencies between the devices, and supplement the correlation factors beyond the physical and energy dimensions.

[0082] Step 304: Integrate the physical transfer topology relationship, energy transfer relationship, and control dependency relationship to generate a multi-dimensional association relationship.

[0083] Specifically, the physical transfer topology relationship, energy transfer relationship, and control dependency relationship can be input into a graph fusion algorithm to generate a unified multi-dimensional association network. Use a graph convolutional network (GCN) to perform feature fusion on the three types of relationships, learn the cross-dimensional association weights. In the fused association network, through the shortest path algorithm (Dijkstra) or random walk model, output the propagation probability and path of the fault from the root cause device to the associated device. Through the multi-dimensional relationship fusion, comprehensively capture the physical, energy, and control logic links of the fault propagation between the devices, and improve the detection coverage rate and location accuracy of hidden association faults.

[0084] In a possible embodiment, a digital twin model is used to simulate and verify the multi-dimensional association relationship to determine the root cause device and propagation path of the hidden association fault, which may include:

[0085] Step 401: Analyze the operation data between multiple devices based on the information entropy theory, calculate the causal influence intensity, and determine candidate device pairs based on the determination threshold. The candidate device pairs include a first device and a second device.

[0086] Specifically, extract the operation state sequence of the first device (such as vibration amplitude, temperature change trend) and the state change of the second device in a future period of time. By calculating the information gain amount of the state of the first device on the future state of the second device, determine the causal association intensity between the two. According to the distribution characteristics of the causal intensity in the historical fault data, dynamically adjust the determination threshold, and screen the device pairs with causal intensity exceeding the threshold as candidate device pairs to generate a candidate list containing potential fault sources and affected devices. Objectively quantify the causal relationship between the devices through the information entropy theory, avoid misjudgment by human experience, and effectively screen out candidate pairs of high-probability fault propagation links to narrow the scope of simulation verification.

[0087] Step 402: Inject the fault parameters of the first device in the candidate device pair into the digital twin model, and simulate the operating state of the second device in the candidate device pair to obtain simulation data.

[0088] Specifically, inject the typical fault parameters of the first device in the candidate device pair, i.e., the candidate fault source (such as bearing wear, gear clearance deviation, insulation aging degree), into the digital twin model, and simulate the fault evolution process of this device in the virtual environment. Through multi-physical field coupling simulation (such as mechanical vibration transmission, thermodynamic diffusion, electromagnetic interference propagation), calculate the operating state data of the second device, i.e., the affected device, under the influence of fault propagation (such as vibration spectrum, temperature distribution curve, current waveform), generate a high-fidelity simulation data set, and strictly reproduce the operating environment parameters of the actual device (such as load, speed, ambient temperature) during the simulation process to ensure the comparability of the simulation results with the actual monitoring data. Through fault injection and multi-physical field simulation of the digital twin model, accurately simulate the fault propagation process, verify the causal relationship hypothesis of the candidate device pair, and provide a reliable virtual verification environment for root cause location.

[0089] Step 403: When the error between the simulation data and the actual monitoring data of the second device is less than the preset tolerance, determine the first device as the root cause device and generate a propagation path.

[0090] Specifically, conduct a multi-dimensional error comparison between the actual monitoring data of the second device (such as real-time vibration spectrum, temperature sensor readings) and the simulation data of the digital twin model, which can include: the energy error of the vibration signal in the characteristic frequency band, the root mean square deviation of the temperature change trend, and the response delay difference of the control signal. If all error terms are less than the preset tolerance threshold, determine the first device as the root cause device of the latent associated fault, and generate a complete propagation path chain from the root cause device to the affected device according to the energy transfer path, control logic trigger sequence, and physical connection topology in the simulation results. Through multi-dimensional error comparison and propagation path chain generation, clarify the fault source and propagation logic, provide an interpretable decision basis for targeted protection operations, and shorten the monitoring, fault diagnosis, and disposal cycle.

[0091] In a possible embodiment, generating a collaborative control instruction according to the root cause device and the propagation path may further include:

[0092] Step 501: Construct a collaborative decision-making model based on device health score and propagation path risk.

[0093] Specifically, a collaborative decision-making model is constructed based on the equipment health score and the propagation path risk. Among them, the equipment health score is calculated through real-time monitoring data (such as vibration amplitude, temperature deviation, current harmonic distortion rate), reflecting the degree of deterioration of the current operating state of the equipment; the propagation path risk is quantitatively evaluated according to the number, correlation strength, and criticality level of the equipment involved in the fault propagation path. The equipment health score and the propagation path risk value are weighted and fused to generate a system-level risk index, which is used as the input constraint condition of the collaborative decision-making model. Through multi-dimensional index fusion, a comprehensive evaluation of the individual health state of the equipment and the system-level risk propagation is realized, supporting the control decision-making of global optimization.

[0094] Step 502, optimize the collaborative decision-making model through the reinforcement learning algorithm to generate a collaborative control instruction that maximizes the system benefit.

[0095] Specifically, the collaborative decision-making model is optimized through the reinforcement learning algorithm to generate a collaborative control instruction that maximizes the system benefit. The reinforcement learning algorithm continuously adjusts the decision-making strategy by simulating the operating environment of the system to achieve the maximization of the overall system benefit. Through the dynamic exploration and exploitation mechanism of reinforcement learning, adaptive decision-making under complex working conditions can be realized, balancing the requirements of fault isolation and production continuity, and maximizing the overall system benefit.

[0096] Step 503, send the collaborative control instruction to the root cause equipment and obtain the return signal. The return signal is used to characterize the execution effect of the control instruction; update the digital twin model parameters based on the return signal.

[0097] Specifically, the optimized collaborative control instruction can be sent to the root cause equipment and related equipment through a low-latency industrial communication network to trigger the isolation, derating, or switching operations of the equipment. At the same time, the equipment status data after the execution of the instruction is collected in real time to generate a return signal for characterizing the control effect. Based on the return signal, the correlation relationship weight, fault propagation probability parameter, and equipment health score calculation coefficient in the digital twin model are corrected in reverse, forming a closed-loop optimization link of "decision-making - execution - feedback - update", improving the consistency between the model and the actual system, ensuring the accuracy and timeliness of subsequent decision-making, and realizing continuous self-optimizing intelligent protection.

[0098] In a possible embodiment, based on the information entropy theory, the operation data between multiple devices is analyzed to calculate the causal influence strength, and determining candidate device pairs based on a determination threshold may include:

[0099] Step 601, calculate the transfer entropy value between devices based on the historical state sequence of the devices. The transfer entropy value is used to quantify the information gain of the state of the first device on the future state of the second device.

[0100] Specifically, the historical state sequence of the first device is extracted as the input variable, and the future state sequence of the second device is taken as the output variable. The information gain of the state of the first device on the future state of the second device is analyzed through the information entropy theory. The transfer entropy value is calculated by statistically analyzing the difference between the joint probability distribution and the conditional probability distribution of the device states, quantifying the causal influence intensity of the first device on the second device. By using the transfer entropy value, the implicit causal relationship between devices is objectively measured, effectively identifying non-linear causal associations that are difficult to capture by traditional correlation analysis, and improving the accuracy of fault source localization.

[0101] Step 602: Dynamically adjust the decision threshold of the causal influence intensity according to the causal intensity distribution of device pairs in the historical fault data.

[0102] Specifically, the distribution range of the transfer entropy values of valid causal pairs in historical fault events can be statistically analyzed. Combining with the operating conditions of the current device group, an adaptive algorithm (such as sliding window statistics, exponentially weighted moving average) is used to dynamically set the decision threshold to ensure that the threshold is optimized in real time with the changes of device states and external environments. Through the dynamic threshold adjustment mechanism, the problem of insufficient adaptability of fixed thresholds under different operating conditions is solved, the false alarm rate and the missed detection rate are reduced, and the robustness of complex industrial scenarios is enhanced.

[0103] Step 603: Add the device pairs whose transfer entropy values exceed the decision threshold to the candidate device pair list to generate candidate device pairs.

[0104] Specifically, the transfer entropy values of all device pairs (such as device A and device B, device A and device C) in the device group are sorted, and the device pairs ranked in the top N (such as the top 10%) in terms of causal intensity are selected as high-priority candidate objects. The candidate device pair list contains the corresponding relationship between the first device (potential fault source) and the second device (affected device), and marks the transfer entropy value and the associated frequency band characteristics, providing input for subsequent simulation verification. Through threshold screening and priority sorting, the high-probability fault propagation path is focused, redundant computing resource consumption is reduced, and the fault diagnosis efficiency is improved.

[0105] In a possible embodiment, optimizing the collaborative decision-making model through a reinforcement learning algorithm to generate collaborative control instructions that maximize the system benefit may include:

[0106] Step 701: Define the reward function of the collaborative decision-making model, and the reward function is obtained based on the balance relationship between system risk and production capacity.

[0107] Specifically, the system risk value is calculated based on the health score of the devices on the propagation path, the fault propagation probability, and the device criticality level; the production capacity maintenance rate is comprehensively evaluated through the equipment operation efficiency, the production compliance rate, and the energy consumption index. By adjusting the weight coefficients of risk and production capacity (such as a risk weight of 0.7 and a production capacity weight of 0.3), the requirements for fault isolation and production continuity are dynamically balanced to generate a reward function for multi-objective optimization. By quantifying the balance relationship between risk and production capacity, it is ensured that the control strategy can maximize the production efficiency while ensuring the safety of the equipment, and avoid decision-making biases caused by single-objective optimization.

[0108] Step 702: Iteratively update the action value function through the Q-learning algorithm to optimize the control strategy of the collaborative decision-making model.

[0109] Specifically, the state space is defined as the combination of the device health score set and the propagation path risk level, and the action space is the set of operation instructions for isolating the root cause device, degrading the operation of associated devices, and switching standby devices. Actions are selected through the exploration-exploitation mechanism. After execution, the system state transition is observed and the immediate reward is calculated. The Q-value table is updated to approximate the optimal action value function. After multiple rounds of iteration, the optimal control strategy that maximizes the long-term reward is converged. Through the self-learning ability of reinforcement learning, it adapts to the decision-making requirements under complex dynamic working conditions, solves the rigidity problem of traditional rule engines, and improves the adaptability and global optimality of the control strategy.

[0110] Step 703: Generate collaborative control instructions according to the control strategy. The collaborative control instructions include the isolation instruction for the root cause device and the degradation operation parameters of the associated devices. The associated devices are the devices on the propagation path of the root cause device.

[0111] Specifically, a set of collaborative control instructions is generated according to the optimized control strategy. The instruction set includes: the physical isolation instruction for the root cause device, the degradation operation parameters of the associated devices on the propagation path, and the start timing instruction for the standby device. The instructions are sent to the target devices through the industrial Internet of Things protocol to ensure millisecond-level response delay and high-reliability execution. By precisely matching the device operation instructions for the fault propagation path, rapid isolation of the fault impact and active blocking of the propagation path are achieved, minimizing the spread of system-level risks.

[0112] Step 704: Obtain a feedback signal, which is used to indicate the digital twin model to update the association relationship parameters.

[0113] Specifically, the device status data after the execution of the real-time acquisition collaborative control instruction is collected, a feedback signal is generated and input into the digital twin model. Based on the actual execution effect in the feedback signal, the weight coefficients of the device association relationship, the fault propagation probability parameters, and the health score calculation rules in the digital twin model are corrected in reverse, realizing the dynamic adaptive update of the model parameters and improving the intelligent level of the remote protection of the digital twin model.

[0114] In summary, the monitoring method of the remote intelligent protection device based on digital twin provided by the embodiments of the present application eliminates the timing deviation of data acquisition between devices and extracts key features by performing spatio-temporal alignment processing on multi-source heterogeneous operation data; constructs physical transfer topological relationships, vibration energy transfer relationships, and control dependency relationships between devices based on the aligned data, and fuses them to generate a multi-dimensional association network model to comprehensively represent potential fault propagation links between devices; uses the digital twin model to dynamically simulate and verify the causal relationship of candidate fault sources, and accurately locates the root cause device and propagation path by comparing the error tolerance between the simulation data and the actual monitoring data; optimizes the collaborative control strategy based on the reinforcement learning algorithm, generates protection instructions that take into account risk suppression and production efficiency, and continuously updates the model parameters through closed-loop feedback. The above technical solutions solve the problem of undetected hidden faults that cannot be captured by traditional methods due to isolated analysis of device data, and are particularly suitable for complex fault scenarios caused by mechanical energy transfer, control logic coupling, or thermodynamic interaction between devices, improving the monitoring reliability and operation and maintenance efficiency of the remote intelligent protection device.

[0115] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or steps or stages in other steps.

[0116] Based on the same inventive concept, an embodiment of the present application further provides a monitoring system for a digital twin-based remote intelligent protection device for implementing the monitoring method of the digital twin-based remote intelligent protection device involved above. The implementation solutions provided by this system to solve problems are similar to those recorded in the above method. Therefore, the specific limitations in one or more embodiments of the monitoring system for the digital twin-based remote intelligent protection device provided below can refer to the limitations on the monitoring method for the digital twin-based remote intelligent protection device in the above text, and will not be elaborated here.

[0117] In an exemplary embodiment, as Figure 2 shown, a monitoring system 10 for a digital twin-based remote intelligent protection device is provided, including:

[0118] A data preprocessing module 11, configured to perform synchronization processing and feature extraction on the operation data to generate spatio-temporal alignment data across devices, where the operation data is multi-source heterogeneous data collected from multiple devices.

[0119] An association establishment module 12, configured to construct multi-dimensional association relationships between multiple devices based on the spatio-temporal alignment data, and the multi-dimensional association relationships are used to represent potential fault propagation paths between devices;

[0120] A root cause determination module 13, configured to perform simulation verification on the multi-dimensional association relationships through a digital twin model to determine the root cause device and propagation path of the hidden association fault;

[0121] A collaborative control module 14, configured to generate collaborative control instructions according to the root cause device and propagation path, and the collaborative control instructions are used to instruct the root cause device to perform protection operations.

[0122] In a possible embodiment, the data preprocessing module 11 may include:

[0123] A vibration signal acquisition unit 111, configured to acquire vibration signals through vibration sensors deployed on multiple devices.

[0124] A vibration signal denoising unit 112, configured to perform wavelet denoising processing on the vibration signals to generate denoised vibration signals.

[0125] A spatio-temporal alignment unit 113, configured to unify the timestamps of the operation data of multiple devices to the same time sequence reference through a dynamic time window alignment algorithm to generate spatio-temporal alignment data.

[0126] In a possible embodiment, the association establishment module 12 may include:

[0127] A physical topology construction unit 121, configured to construct a physical transmission topology relationship based on the mechanical connection structure between devices.

[0128] A coherence analysis unit 122 for performing coherence analysis on the noise-reduced vibration signal to generate an energy transfer relationship between devices.

[0129] A control dependence analysis unit 123 for analyzing the control signals of devices to generate a control dependence relationship.

[0130] A multi-dimensional fusion unit 124 for fusing the physical transfer topological relationship, the energy transfer relationship, and the control dependence relationship to generate a multi-dimensional association relationship.

[0131] In a possible embodiment, the root cause determination module 13 may include:

[0132] A causality analysis unit 131 for analyzing the operation data between multiple devices based on the information entropy theory, calculating the causality influence intensity, and determining candidate device pairs based on a determination threshold. The candidate device pairs include a first device and a second device.

[0133] A fault injection simulation unit 132 for injecting the fault parameters of the first device in the candidate device pair into the digital twin model, simulating the operation state of the second device in the candidate device pair, and obtaining simulation data.

[0134] A root cause judgment unit 133 for determining the first device as the root cause device and generating a propagation path when the error between the simulation data and the actual monitoring data of the second device is less than a preset tolerance.

[0135] In a possible embodiment, the system further includes:

[0136] A decision model construction module 15 for constructing a collaborative decision model based on the device health score and the propagation path risk.

[0137] A reinforcement learning module 16 for optimizing the collaborative decision model through a reinforcement learning algorithm to generate a collaborative control instruction that maximizes the system benefit.

[0138] An execution feedback module 17 for sending the collaborative control instruction to the root cause device and obtaining a return signal. The return signal is used to characterize the execution effect of the control instruction; updating the digital twin model parameters based on the return signal.

[0139] In a possible embodiment, the causality analysis unit 131 may include:

[0140] A transfer entropy calculation sub-unit 1311 for calculating the transfer entropy value between devices based on the historical state sequences of the devices. The transfer entropy value is used to quantify the information gain of the state of the first device on the future state of the second device.

[0141] A dynamic threshold adjustment sub-unit 1312 for dynamically adjusting the determination threshold of the causality influence intensity according to the causality intensity distribution of device pairs in the historical fault data.

[0142] A candidate list generation subunit 1313 is configured to add device pairs whose transfer entropy values exceed a determination threshold to a candidate device pair list, and generate candidate device pairs.

[0143] In a possible embodiment, the reinforcement learning module 16 may include:

[0144] A reward function definition unit 161 is configured to define a reward function for the collaborative decision-making model, and the reward function is obtained based on the balance relationship between system risk and production capacity.

[0145] An algorithm optimization unit 162 is configured to iteratively update the action value function through a Q-learning algorithm to optimize the control strategy of the collaborative decision-making model.

[0146] An instruction issuing unit 163 is configured to generate a collaborative control instruction according to the control strategy, and the collaborative control instruction includes an isolation instruction for the root cause device and a derating operation parameter for the associated device, and the associated device is a device on the propagation path of the root cause device.

[0147] A closed-loop update unit 164 is configured to obtain a feedback signal, and the feedback signal is used to instruct the digital twin model to update the association relationship parameter.

[0148] In an embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of a monitoring method of a digital twin remote intelligent protection device as described above are implemented.

[0149] In an embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0150] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can refer to the partial descriptions of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of the present disclosure. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0151] The above-described embodiments merely represent several implementation manners of the embodiments of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent for the embodiments of the application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the embodiments of the present application, several modifications and improvements can still be made, and these all fall within the protection scope of the embodiments of the present application.

Claims

1. A monitoring method for a digital twin-based remote intelligent protection device, characterized in that, The method includes: Performing synchronization processing and feature extraction on the operation data to generate spatio-temporal alignment data across devices, where the operation data is multi-source heterogeneous data collected from multiple devices; Based on the spatio-temporal alignment data, constructing multi-dimensional association relationships between the multiple devices, where the multi-dimensional association relationships are used to characterize potential fault propagation paths between devices; Performing simulation verification on the multi-dimensional association relationships through a digital twin model to determine the root cause device and propagation path of latent associated faults; Generating a collaborative control instruction according to the root cause device and the propagation path, where the collaborative control instruction is used to instruct the root cause device to perform a protection operation.

2. The method according to claim 1, characterized in that, The performing synchronization processing and feature extraction on the operation data to generate spatio-temporal alignment data across devices includes: Collecting vibration signals through vibration sensors deployed on multiple devices; Performing wavelet denoising processing on the vibration signals to generate denoised vibration signals; Through a dynamic time window alignment algorithm, unifying the timestamps of the operation data of the multiple devices to the same time sequence reference to generate the spatio-temporal alignment data.

3. The method according to claim 2, wherein The constructing multi-dimensional association relationships between the multiple devices based on the spatio-temporal alignment data includes: Based on the mechanical connection structure between devices, constructing a physical transfer topology relationship; Performing coherence analysis on the denoised vibration signals to generate an energy transfer relationship between devices; Analyzing the control signals of the devices to generate a control dependency relationship; Fusing the physical transfer topology relationship, the energy transfer relationship, and the control dependency relationship to generate the multi-dimensional association relationships.

4. The method according to claim 1, characterized in that The performing simulation verification on the multi-dimensional association relationships through a digital twin model to determine the root cause device and propagation path of latent associated faults includes: Analyzing the operation data between the multiple devices based on the information entropy theory, calculating the causal influence intensity, and determining candidate device pairs based on a determination threshold, where the candidate device pairs include a first device and a second device; Injecting the fault parameters of the first device in the candidate device pair into the digital twin model, simulating the operation state of the second device in the candidate device pair to obtain simulation data; When the error between the simulation data and the actual monitoring data of the second device is less than a preset tolerance, determining the first device as the root cause device and generating the propagation path.

5. The method according to claim 1, characterized in that, The generating a collaborative control instruction according to the root cause device and the propagation path further includes: Constructing a collaborative decision-making model based on device health score and propagation path risk; Optimizing the collaborative decision-making model through a reinforcement learning algorithm to generate the collaborative control instruction that maximizes the system benefit; Sending the collaborative control instruction to the root cause device and obtaining a return signal, where the return signal is used to characterize the execution effect of the control instruction; updating the digital twin model parameters based on the return signal.

6. The method according to claim 4, characterized in that The analyzing the operation data between the multiple devices based on the information entropy theory, calculating the causal influence intensity, and determining candidate device pairs includes: Calculating the transfer entropy value between devices based on the historical state sequence of the devices, where the transfer entropy value is used to quantify the information gain of the state of the first device on the future state of the second device; Dynamically adjust the determination threshold of the causal influence strength according to the causal strength distribution of device pairs in historical fault data; Add device pairs with transfer entropy values exceeding the determination threshold to the candidate device pair list to generate the candidate device pairs.

7. The method according to claim 5, characterized in that, Optimize the collaborative decision-making model through a reinforcement learning algorithm to generate the collaborative control instructions that maximize the system benefit, including: Define the reward function of the collaborative decision-making model, and the reward function is obtained based on the balance relationship between system risk and production capacity; Iteratively update the action value function through the Q-learning algorithm to optimize the control strategy of the collaborative decision-making model; Generate collaborative control instructions according to the control strategy, and the collaborative control instructions include isolation instructions for the root cause device and derating operation parameters of associated devices, and the associated devices are devices on the propagation path of the root cause device; Obtain a feedback signal, and the feedback signal is used to instruct the digital twin model to update the association relationship parameters.

8. A monitoring system for a digital twin-based remote intelligent protection device, characterized in that, Include: A data preprocessing module for synchronously processing and feature extracting the operation data to generate spatio-temporal aligned data across devices, and the operation data is multi-source heterogeneous data collected from multiple devices; An association establishment module for constructing a multi-dimensional association relationship between the multiple devices based on the spatio-temporal aligned data, and the multi-dimensional association relationship is used to represent potential fault propagation paths between devices; A root cause determination module for simulating and verifying the multi-dimensional association relationship through a digital twin model to determine the root cause device and propagation path of the hidden associated fault; A collaborative control module for generating collaborative control instructions according to the root cause device and the propagation path, and the collaborative control instructions are used to instruct the root cause device to perform protection operations.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method according to any one of claims 1 to 7.

Images