A valve remote control flow and abnormal shutdown processing method and device and medium

Abstract

This invention discloses a method, device, and medium for remote flow control and abnormal shutdown handling of valves, comprising: collecting multi-dimensional time-series sensor data of valve operation to construct a multivariate time-series dataset; inputting the multivariate time-series dataset into an industrial time-series large model, the time-series large model extracting time-series features through a multi-scale attention mechanism, calculating the time-series feature offset by comparing it with a standard operating condition feature library, and outputting a decision result; when the abnormal handling decision instruction is an emergency shutdown instruction, a hardware security isolation unit operating independently of the time-series large model receives the emergency shutdown instruction, performs a legality check, and performs a dual-channel vote between the emergency shutdown instruction and the backup shutdown signal output by an independently operating backup shutdown decision unit; if the check passes and the vote passes, the shutdown execution signal is transmitted to the valve actuator; if the check fails or the vote fails, the shutdown is refused and a warning message is sent.

Description

Technical Field

[0001] This invention discloses a method, device, and medium for remote flow control and abnormal shutdown handling of valves, belonging to the technical field of remote flow control and abnormal shutdown handling of valves. Background Technology

[0002] Currently, most remote flow control and abnormal shutdown handling for fluid pipeline valves employs a fixed-parameter PID controller combined with a single threshold over-limit alarm. While this traditional approach meets basic control requirements under steady-state conditions, it suffers from significant drawbacks such as response lag, insufficient control precision, and weak adaptive capabilities when facing complex conditions including drastic load fluctuations, changes in medium parameters, and multivariate coupling interference. Traditional single-threshold comparison-based anomaly detection methods struggle to identify gradual, latent anomalies like slow leakage, valve jamming and wear, and continuous flow accuracy deviations, relying solely on preset thresholds for shutdown triggering, which is prone to false alarms or missed alarms. Furthermore, research on the application of deep learning and model predictive control in valve flow control often focuses on single technical aspects, lacking an integrated system solution encompassing data perception, time-series intelligent modeling, precise flow control decision-making, and safe shutdown execution. Although large-scale models possess powerful time-series modeling and intelligent decision-making capabilities, their direct application in industrial valve control presents safety risks such as decision-making randomness and inference uncertainty. Therefore, organically combining the intelligent decision-making advantages of large-scale time-series models with the safety and reliability of industrial control is a critical technical problem urgently needing to be solved in this field.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This invention provides a method, device, and medium for remote flow control and abnormal shutdown of valves, which at least solves the technical problems of poor adaptability to complex working conditions, difficulty in identifying hidden faults, and difficulty in balancing large-scale intelligent decision-making with industrial control security in traditional valve control schemes.

[0005] According to one aspect of the present invention, in order to achieve the above objective, a method for remote flow control and abnormal shutdown handling of a valve is provided, comprising:

[0006] Collect multi-dimensional time-series sensor data of valve operation and construct a multivariate time-series dataset;

[0007] The multivariate time series dataset is input into the industrial time series large model. The time series large model extracts time series features through a multi-scale attention mechanism, calculates the time series feature offset by comparing it with the standard operating condition feature library, and outputs decision results, including: optimal valve opening adjustment parameters, valve operating status classification results, and abnormal handling decision instructions.

[0008] When the abnormal handling decision instruction is an emergency shutdown instruction, the hardware security isolation unit, which operates independently of the timing model, receives the emergency shutdown instruction, performs a legality check, and conducts a dual-channel vote between the emergency shutdown instruction and the backup shutdown signal output by the independently operating backup shutdown decision unit. If the check passes and the vote passes, the shutdown execution signal is transmitted to the valve actuator. If the check fails or the vote fails, the shutdown is refused and a warning message is sent.

[0009] Furthermore, the legality verification includes verifying whether the current upstream pressure of the valve is higher than the preset safe shut-off pressure threshold, whether the current instantaneous flow rate of the pipeline is lower than the preset shut-off flow rate lower limit, and whether the current valve drive current is within the preset normal operating current range.

[0010] Dual-channel voting is hardware-level voting. The hardware safety isolation unit is only allowed to output a shutdown execution signal when both the emergency shutdown command and the backup shutdown signal are valid.

[0011] It also includes watchdog forced shutdown: the process of the valve actuator performing an emergency shutdown action is timed independently. If no feedback on completion is received within the preset safe shutdown time, the watchdog is triggered to force shutdown. The watchdog bypasses the main control program and directly drives the valve actuator to cut off power so that the valve is reset to the fully closed position.

[0012] Furthermore, the temporal large-scale model extracts temporal features through a multi-scale attention mechanism, including:

[0013] The input multivariate time series matrix is ​​decomposed into trend components, periodic components and residual components.

[0014] Attention calculations are performed on the trend component, periodic component, and residual component with different receptive fields to capture the changing patterns of operating conditions at different time scales.

[0015] The components after multi-scale attention calculation are fused to generate a multivariate temporal fusion feature vector.

[0016] Furthermore, the valve operating status classification results include normal status, minor abnormal status, early warning abnormal status, and emergency fault status;

[0017] When the valve operating status classification result is a slightly abnormal state, the abnormal handling decision instruction is a fine-tuning correction instruction, which automatically calibrates the operating condition by adaptively correcting the valve opening.

[0018] When the valve operating status classification result is an abnormal warning state, the abnormal handling decision instruction is an alarm warning instruction, which triggers a remote alarm and locks the valve regulation authority.

[0019] When the valve operating status classification result is an emergency fault state, the abnormal handling decision instruction is an emergency shutdown instruction.

[0020] Furthermore, after inputting the multivariate time series dataset into the large industrial time series model, it also includes:

[0021] The optimal valve opening adjustment parameters output by the time series large model are validated according to industrial rules. The validation checks whether the single adjustment amplitude exceeds the preset single maximum adjustment amplitude threshold and whether the adjustment speed exceeds the preset maximum adjustment speed threshold. If they exceed the threshold, the amplitude is limited according to the threshold.

[0022] The optimal valve opening adjustment parameters output at multiple consecutive time steps are subjected to anti-jitter filtering to eliminate random fluctuation components and generate control parameters adapted to the valve actuator.

[0023] Furthermore, it also includes online iterative optimization of large time-series models:

[0024] During system operation, new operating condition data are continuously collected, and an importance score matrix is ​​maintained for each network parameter of the time series model. The importance score is calculated based on the contribution of the parameter to the prediction accuracy in different historical operating conditions.

[0025] When new operating condition data arrives, elastic regularization constraints are applied to parameters whose importance scores are higher than a preset threshold to limit their update magnitude.

[0026] Based on the operating condition type identifier, historical samples different from the current new operating condition type are extracted from the historical data repository, and a mixed playback batch is constructed with the current new operating condition samples. The joint loss is calculated based on the mixed playback batch to update the parameters.

[0027] Furthermore, online iterative optimization also includes:

[0028] Calculate the average prediction error of the time series model within the most recent preset number of time steps. When the average prediction error exceeds the preset performance alarm threshold, it is determined that the model has experienced performance degradation.

[0029] Trigger model version rollback, reverting the parameters of the time series model to the latest checkpoint version before the performance degradation occurred, and pausing incremental learning until an external recovery confirmation instruction is received.

[0030] Furthermore, after collecting multi-dimensional time-series sensor data on valve operation, the process also includes:

[0031] Calculate the timestamp deviation based on the actual sampling frequency of each sensor, and use the asynchronous data packet timestamp linear interpolation method with the system clock as the reference to align the data of each sensor to a unified time reference.

[0032] When sensor data is missing, the following steps are performed in sequence: compensation for maintaining the valid value of the previous moment, compensation based on the moving average of historical data from the same period of the same type of sensor, and compensation based on the predicted value of the forward time series data of the missing period based on the time series large model. When the pairwise deviation of the compensation values ​​of the three compensation methods exceeds the preset deviation threshold, a sensor fault alarm is triggered.

[0033] When constructing a multivariate time series dataset, the current operating condition type identifier code is embedded synchronously.

[0034] According to one embodiment of the present invention, a valve remote flow control and abnormal shutdown handling device is also provided, characterized in that it includes:

[0035] The data acquisition module is used to collect multi-dimensional time-series sensor data of valve operation and construct a multivariate time-series dataset;

[0036] The decision module is used to input the multivariate time series dataset into the industrial time series large model. The time series large model extracts time series features through a multi-scale attention mechanism, calculates the time series feature offset by comparing it with the standard operating condition feature library, and outputs decision results. The decision results include: optimal valve opening adjustment parameters, valve operating status classification results, and abnormal handling decision instructions.

[0037] The isolation module is used to receive the emergency shutdown command, which operates independently of the time-series large model, when the abnormal handling decision command is an emergency shutdown command. The unit performs a legality check on the emergency shutdown command and conducts a dual-channel vote with the backup shutdown signal output by the independently operating backup shutdown decision unit. If the check and vote pass, the shutdown execution signal is transmitted to the valve actuator. If the check or vote fails, the shutdown is refused and a warning message is sent.

[0038] According to another aspect of the present invention, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the valve remote flow control and abnormal shutdown handling methods of various embodiments of the present invention.

[0039] In this embodiment of the invention, a multivariate time-series dataset is constructed by collecting multi-dimensional time-series sensor data of valve operation. The multi-scale attention mechanism of an industrial time-series large-scale model is used to extract time-series features, and the time-series feature offset is calculated by comparing it with a standard operating condition feature library. This solves the technical problem of traditional solutions being unable to accurately identify progressive latent anomalies. By integrating the optimal valve opening adjustment parameters, valve operating status classification results, and anomaly handling decision commands into a single time-series large-scale model, the flow control accuracy and adaptive adjustment capability under complex operating conditions are effectively improved. More importantly, by setting up a hardware security isolation unit independent of the time-series large-scale model's operation, independent legality verification is performed when the large model outputs an emergency shutdown command, and the large model's commands are synchronized with the backup... By using the signal from the shutdown decision maker for hardware-level dual-channel voting and setting up a watchdog forced shutdown mechanism, a hierarchical control architecture of "large model intelligent decision-making + hardware safe execution" is realized. This architecture not only leverages the intelligent decision-making advantages of capturing long-term dependencies and extracting multi-variable coupled features from the large-series model, but also completely avoids the safety risks caused by the uncertainty of large model inference through hardware-level safety isolation and dual-channel redundant voting. It balances the level of control intelligence with the high reliability and real-time requirements of industrial sites, significantly improving the accuracy of remote valve flow control, the sensitivity of anomaly identification, and the safety of emergency shutdown. It is adaptable to various industrial pipeline valve operating conditions such as oil and gas, water, chemical, and HVAC, and has broad engineering application value. Attached Figure Description

[0040] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0041] Figure 1 This is a flowchart of a valve remote flow control and abnormal shutdown handling method according to one embodiment of the present invention;

[0042] Figure 2 This is a structural block diagram of a valve remote flow control and abnormal shutdown processing device according to one embodiment of the present invention. Detailed Implementation

[0043] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0044] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0045] According to an embodiment of the present invention, a method for remote flow control and abnormal shutdown of a valve is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer device such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0046] This method embodiment can be executed in an electronic device or similar computing device that includes a memory and a processor. Taking data processing as an example, the data processing may include one or more processors (processors may include, but are not limited to, central processing units (CPUs), graphics processing units (GPUs), digital signal processing (DSP) chips, microcontroller units (MCUs), field-programmable gate arrays (FPGAs), neural network processors (NPUs), tensor processors (TPUs), artificial intelligence (AI) type processors, etc.) and a memory for storing data. Optionally, the above data processing may also include transmission devices, input / output devices, and display devices for communication functions. Those skilled in the art will understand that the above structural description is merely illustrative and does not limit the structure of the above data processing. For example, the data processing may include more or fewer components than described above, or have a different configuration than described above.

[0047] The memory can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the valve remote flow control and abnormal shutdown handling method in this embodiment of the invention. The processor executes various functional applications and data processing by running the computer program stored in the memory, thereby realizing the aforementioned valve remote flow control and abnormal shutdown handling method. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to a mobile terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0048] The transmission device is used to receive or send data via a network. Specific examples of the network mentioned above may include a wireless network provided by the mobile terminal's communication provider. In one example, the transmission device includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0049] Display devices can be, for example, touchscreen liquid crystal displays (LCDs) and touch displays (also referred to as "touchscreens" or "touch displays"). The LCD allows users to interact with the user interface of the mobile terminal. In some embodiments, the mobile terminal has a graphical user interface (GUI), which allows users to interact with the GUI through finger contact and / or gestures on a touch-sensitive surface. Optional human-computer interaction functions include: creating web pages, drawing, word processing, creating electronic documents, playing games, video conferencing, instant messaging, sending and receiving emails, call interfaces, playing digital video, playing digital music, and / or web browsing, etc. Executable instructions for performing the above human-computer interaction functions are configured / stored in one or more processor-executable computer program products or readable storage media.

[0050] Figure 1 According to one embodiment of the present invention, a method for remote flow control and abnormal shutdown of a valve includes:

[0051] Step S110: Collect multi-dimensional time-series sensor data of valve operation and construct a multivariate time-series dataset. The specific content is as follows:

[0052] In step S110, a multi-type sensor array is deployed at key locations of the target valve and its pipeline to achieve synchronous acquisition of valve operation status data across all dimensions. Specifically, an angular displacement sensor or potentiometer-type position sensor is installed on the valve body to acquire valve opening signals in real time; pressure transmitters are installed upstream and downstream of the valve to acquire inlet and outlet pressures; an ultrasonic flow meter or vortex flow meter is installed in the downstream straight pipe section (typically 15 to 20 times the pipe diameter) to acquire instantaneous flow rate; a platinum resistance temperature sensor is installed along the pipeline medium flow path to acquire medium temperature; and a current detection module is connected in series in the valve actuator's drive circuit to acquire actuator operating current. The sampling period of each sensor is set according to the dynamic response characteristics of the monitored parameters. The preferred sampling period for pressure and flow rate is 1 second / time, for temperature and valve opening is 5 seconds / time, and for actuator current is 0.5 seconds / time, to ensure sufficient capture of changes in each physical quantity.

[0053] Raw data collected by each sensor is transmitted to the edge data acquisition terminal via fieldbus (such as Modbus-RTU, Profibus-DP) or industrial wireless protocols (such as WirelessHART, ZigBee). Due to differences in sampling frequency, clock source, and communication delay among different sensors, the data from each channel is offset on the time axis. Therefore, time alignment is required after data aggregation. Specifically, using the system clock of the edge data acquisition terminal as a unified time reference, a precise time protocol is used to synchronize and calibrate the clocks of each sensor node, aligning the local clocks of all nodes to the same time reference. The preferred time synchronization accuracy is that the clock deviation of each node does not exceed 1 millisecond. For timestamp misalignment caused by different sampling frequencies, an asynchronous data packet timestamp linear interpolation method is used to uniformly interpolate the data from each sensor onto the same standard time grid, forming a time-aligned multivariate sensor data sequence.

[0054] After time alignment, the data sequence is preprocessed to remove outliers and suppress random noise. Specifically, an adaptive Kalman filter algorithm is used to filter the data from each channel, and a first-order inertial state transition model for each parameter is established. The covariance matrix of process noise and measurement noise is estimated online using the covariance matching method of the innovation sequence, enabling adaptive adjustment of the filtering parameters. When the deviation between the measured value and the filtered predicted value at a certain moment exceeds the dynamic threshold calculated based on the statistical characteristics of the innovation, the measured value is determined to be an outlier and removed, replaced by the filtered predicted value. Simultaneously, for data loss due to instantaneous sensor failure or communication packet loss, a multi-level compensation strategy is adopted: first, compensation is performed using the valid value of the sensor at the previous moment; if continuous loss exceeds a preset number (preferably 3 sampling periods), compensation is switched to a moving average based on historical data from the same period of the same type of sensor; if valid data still cannot be obtained, compensation is performed based on the predicted value of the forward time series data for the missing period using a large time series model. When the pairwise deviations of the compensation values ​​from the three compensation methods exceed a preset deviation threshold, a persistent sensor failure is determined, triggering a sensor fault alarm.

[0055] After the above preprocessing, a sliding window multivariate time series matrix is ​​constructed according to a preset time window length. Specifically, the time window length N is set (preferably 1440, corresponding to minute-level data for 24 hours), and the current moment is taken as the end point of the window. Sensing data of each dimension from the past N time steps are extracted to form an N-row, M-column multivariate time series matrix, where M corresponds to seven core feature dimensions: valve opening, instantaneous pipeline flow rate, inlet pressure, outlet pressure, medium temperature, actuator operating current, and remotely received flow setting command. During matrix construction, the current operating condition type identifier is simultaneously encoded and embedded into the matrix's additional channels. This operating condition type identifier includes steady-state operation, start-stop regulation, load switching, and abnormal fluctuation conditions, enabling the time series model to distinguish data distribution differences under different operating conditions and improve cross-condition generalization ability. Finally, the constructed standardized multivariate time series dataset is used as the input feature of the time series model. This dataset is stored in the time series database of the edge data acquisition terminal, supporting real-time inference calls and historical data tracing of the model.

[0056] Through the above data sensing process, a complete data link from raw sensor signals to a standardized time series matrix is ​​realized, providing a high-quality, highly complete, and time-consistent multivariate time series data foundation for accurate decision-making in subsequent large-scale time series models.

[0057] Step S120: Input the multivariate time-series dataset into the industrial time-series large model. The time-series large model extracts time-series features through a multi-scale attention mechanism, calculates the time-series feature offset by comparing it with the standard operating condition feature library, and outputs the decision results. The decision results include: optimal valve opening adjustment parameters, valve operating status classification results, and abnormal handling decision instructions, as detailed below:

[0058] In step S120, the input feature tensor of the industrial time series model is constructed. The previously collected and preprocessed multivariate time series dataset is used to construct a sliding window multivariate time series matrix according to a preset time window length N. The rows of this matrix correspond to time steps, and the columns correspond to seven feature dimensions: valve opening, instantaneous pipeline flow rate, inlet pressure, outlet pressure, medium temperature, actuator operating current, and remotely set flow rate value. Using the current moment as the end point of the window, data from the past N time steps are extracted to form a standardized input feature tensor. The industrial time series model preferably uses the Timer time series model or the PatchTST time series Transformer model. Compared to traditional Long Short-Term Memory (LSTM) networks and Gated Recurrent Units (GRUs), these models have stronger long-term time-series dependency capture capabilities and multivariate coupling feature extraction capabilities, and can accurately fit the nonlinear dynamic coupling relationship between opening, flow rate, and pressure in valve flow control scenarios.

[0059] The specific process by which the large-scale time-series model extracts time-series features through a multi-scale attention mechanism is as follows: The model first performs a time-series decomposition operation on the input multivariate time-series matrix, separating the original time-series signal into trend components, periodic components, and residual components. The trend component reflects the overall trend of valve operating conditions over a longer period, such as the slow increase or decrease in flow rate due to changes in pipeline load; the periodic component reflects the regular fluctuations in valve operation, such as periodic load changes like shift changes and differences in gas consumption between day and night; and the residual component captures random disturbances and noise components in the time-series data. For these three components, the model uses attention heads with different receptive fields for parallel computation. The trend component uses a large receptive field attention to capture long-cycle variation patterns, the periodic component uses a medium receptive field attention to match the cycle length, and the residual component uses a small receptive field attention to preserve local detail features. After each attention head independently completes its self-attention computation, the attention features at different scales are weighted and concatenated through a multi-head attention fusion layer to generate a fused feature vector containing multi-scale time-series information.

[0060] After obtaining the multi-scale time-series fused feature vector, the time-series large model compares this feature vector with a standard operating condition feature library to calculate the time-series feature offset. The standard operating condition feature library is a set of benchmark features constructed during the model training phase based on massive historical time-series data collected from valves under normal operating conditions, through feature extraction and cluster analysis. This feature library covers standard feature vectors and their allowable fluctuation ranges under different steady-state and transitional operating conditions. During real-time inference, the model calculates the similarity between the fused feature vector at the current moment and the closest operating condition template in the standard operating condition feature library, using cosine similarity or Mahalanobis distance as the similarity metric. The calculated similarity value is converted into a time-series feature offset after nonlinear mapping. This offset is a normalized continuous value used to quantify the degree of deviation of the current operating condition from the standard operating condition. The closer the offset is to 0, the closer the operating condition is to the standard state; the larger the offset, the higher the degree of abnormality of the operating condition.

[0061] The time-series large-scale model, based on extracted multi-scale time-series fusion features and calculated time-series feature offsets, outputs integrated decision results across four dimensions through layer-by-layer mapping of a fully connected classification head and a regression head. First, the optimal valve opening adjustment parameter, a continuous value, is output by the regression head to determine the optimal valve opening adjustment required to bring the pipeline flow rate close to the remotely set target value under current operating conditions. Simultaneously, a feedforward compensation parameter is output for potential pressure fluctuations during the adjustment process, achieving feedforward-based precise flow control. Second, the valve operating status classification result, a discrete category label, is output by the classification head to indicate the current valve status category, including four levels: normal, slightly abnormal, warning abnormal, and emergency fault. Third, the abnormal handling decision command, corresponding to the state classification result, includes three types: fine-tuning correction command, alarm warning command, and emergency shutdown command. The model automatically matches and outputs the corresponding handling command based on the valve operating status classification result. Fourth, the predicted flow and pressure values ​​for future preset time steps are output by the model's prediction head to predict operating condition trends, providing a forward-looking basis for flow control decisions. The outputs of the above four dimensions are interconnected and generated synchronously, realizing integrated intelligent decision-making of "state perception - trend prediction - parameter solution - disposal decision".

[0062] Through the above intelligent decision-making process, the time-series large model can not only accurately solve the flow control parameters adapted to the current complex working conditions, but also simultaneously complete the identification of abnormal states and hierarchical handling decisions. This avoids the decision delay and information loss problems caused by the separate operation of the prediction module, control module and diagnosis module in the traditional solution, and greatly improves the intelligence level and decision-making efficiency of remote valve flow control.

[0063] Step S140: When the abnormal handling decision instruction is an emergency shutdown instruction, the hardware security isolation unit, which operates independently of the time-series large model, receives the emergency shutdown instruction, performs a legality check, and conducts a dual-channel vote between the emergency shutdown instruction and the backup shutdown signal output by the independently operating backup shutdown determiner. If the check passes and the vote passes, the shutdown execution signal is transmitted to the valve actuator. If the check fails or the vote fails, the shutdown is refused and a warning message is sent, as detailed below:

[0064] In step S140, when the abnormal handling decision instruction is an emergency shutdown instruction, the hardware security isolation unit, which operates independently of the timing model, receives the emergency shutdown instruction, performs a legality check, and performs a dual-channel vote between the emergency shutdown instruction and the backup shutdown signal output by the independently operating backup shutdown determiner. If the check passes and the vote passes, the shutdown execution signal is transmitted to the valve actuator. If the check fails or the vote fails, the shutdown is refused and a warning message is sent. The specific implementation process is as follows:

[0065] First, it should be noted that the hardware security isolation unit is physically independent of the intelligent decision-making module of the runtime large-scale model. Specifically, the large-scale model is deployed in an edge industrial control computer or server, while the hardware security isolation unit is implemented using an independent programmable logic controller or security relay module. The two communicate unidirectionally or bidirectionally through isolated data interfaces, but the operation of the hardware security isolation unit does not depend on the software state of the large-scale model. Internally, the hardware security isolation unit integrates a microprocessor or logic gate array independent of the large-scale model's inference logic, equipped with independent memory and clock sources. Its power supply lines are independent of the intelligent decision-making module's power supply lines, ensuring that the hardware security isolation unit can still maintain normal monitoring and decision-making functions in the event of software anomalies in the large-scale model, system crashes, or communication interruptions.

[0066] When the abnormal handling decision instruction is an emergency shutdown instruction, it is transmitted from the intelligent decision module to the hardware security isolation unit through an isolated data interface. Upon receiving the instruction, the hardware security isolation unit first triggers its internal interrupt service routine, temporarily storing the emergency shutdown instruction in a separate instruction cache register, and simultaneously starts a hardware timer to monitor the response time of subsequent stages. Subsequently, the hardware security isolation unit calls its built-in industrial valve safety operation rule library to perform a legality check on the received emergency shutdown instruction. This industrial valve safety operation rule library is pre-stored in the read-only memory chip of the hardware security isolation unit. This rule library is constructed based on the target valve's equipment specifications, process safety operating procedures, and industry safety standards, and includes multi-dimensional constraints for valve safety operation. The legality check specifically includes three parallel check channels: the first channel checks whether the current upstream pressure of the valve is higher than a preset safety shutdown pressure threshold. Specifically, the hardware safety isolation unit reads the pressure transmitter signal installed on the upstream pipe section of the valve in real time, compares the measured pressure value with the safety shutdown pressure threshold stored in the rule base. If the upstream pressure is higher than the threshold, it indicates that the pipeline is in an overpressure danger state, the shutdown condition is met, and this channel passes the verification. If the upstream pressure does not exceed the threshold but the emergency shutdown command still requires shutdown, further judgment is needed in conjunction with other channels. The second channel verifies whether the current pipeline instantaneous flow rate is lower than the preset shutdown flow rate lower limit. Specifically, the hardware safety isolation unit reads the downstream flow meter signal in real time. If the measured flow rate is lower than the shutdown flow rate lower limit, it indicates that the pipeline may be in a low flow or interrupted flow state. Shutdown will not cause any additional impact beyond the supply interruption, and this channel passes the verification. The third channel verifies whether the current valve drive current is within the preset normal operating current range. Specifically, the hardware safety isolation unit reads the current detection module signal on the valve actuator drive circuit. If the drive current is within the normal operating range, it indicates that the actuator's electrical state is normal, the shutdown action can be reliably executed, and this channel passes the verification. The validity check is considered passed when all three verification channels pass the check; the validity check is considered failed when any one channel fails the check. It should be noted that the three verification channels can be independently configured as enabled or bypass modes to adapt to the differentiated needs of different valve types and application scenarios.

[0067] Based on the successful legality verification, the hardware security isolation unit initiates a dual-channel voting mechanism. This dual-channel voting is hardware-level voting, the core of which lies in performing hardware logic AND operations on the emergency shutdown command from the intelligent decision-making channel of the time-series large model and the backup shutdown signal from a completely independently operating backup shutdown determiner. The backup shutdown determiner is a third set of safety monitoring devices independent of the time-series large model and the hardware security isolation unit. It adopts a completely different judgment principle from the traditional large model to achieve redundant backup of shutdown decisions. Preferably, the backup shutdown determiner is a hardware comparator circuit based on a single threshold logic. This circuit monitors key process parameters in real time through the hardware comparator, such as whether the pipeline pressure exceeds the preset physical limit threshold, whether the flow rate drops sharply, etc. When the measured parameter exceeds the preset physical limit threshold, the backup shutdown determiner outputs a valid backup shutdown signal. The threshold setting of the backup shutdown determiner is determined based on physical safety boundaries such as the pressure bearing limit of the pipeline material and the mechanical strength limit of the valve, which is completely different from the intelligent judgment logic of the aforementioned time-series large model, thereby avoiding the risk of simultaneous failure of both channels due to the same source of failure. At the hardware implementation level, the emergency shutdown command of the timing model and the backup shutdown signal of the backup shutdown decision unit are respectively connected to the dual-channel voting subunit of the hardware safety isolation unit through two physically isolated signal cables. This voting subunit is implemented using a dual-input AND gate logic hardware circuit. The AND gate circuit only outputs a valid shutdown execution level signal when both input signals are valid, for example, both are high. This hardware logic AND operation does not involve any software judgment and is completed entirely by the hardware gate circuit in nanoseconds, fundamentally eliminating the delay and logic error risks that may be introduced by software decision-making.

[0068] If the legality verification passes and the dual-channel voting passes, the hardware safety isolation unit transmits the shutdown execution signal to the valve actuator through its output interface. Specifically, the shutdown execution signal is amplified by the output drive circuit of the hardware safety isolation unit and transmitted to the emergency shutdown input terminal of the valve actuator through a shielded cable. For electric actuators, this shutdown execution signal triggers the emergency braking circuit inside the actuator, causing the motor to stop immediately and activating the mechanical brake device to keep the valve in the current fully closed position. For pneumatic or hydraulic actuators, this shutdown execution signal triggers the solenoid valve to switch, causing the actuator's air or hydraulic pressure to reverse and drive the valve core to move rapidly to the shutdown position. The shutdown execution signal is also recorded in the non-volatile memory inside the hardware safety isolation unit, forming a shutdown event log. This log contains information such as the time of shutdown triggering, the trigger source (i.e., the large model instruction or the backup decision unit, or both simultaneously), and a snapshot of the values ​​of each key parameter before shutdown, for subsequent accident tracing and analysis.

[0069] If the legality verification fails or the dual-channel voting fails, the hardware security isolation unit refuses to execute the shutdown action and sends an early warning message to the remote operation and maintenance platform. Specifically, when the legality verification fails, it means that the emergency shutdown command output by the large model does not meet the security constraints in the industrial safety operation rule base. The hardware security isolation unit determines the emergency shutdown command as an untrusted command and does not execute it. At the same time, it generates an early warning data packet containing the reason code for the verification failure and sends it to the operation and maintenance monitoring center via an independent communication link to notify the operation and maintenance personnel to conduct manual review. This independent communication link is preferably a 4G or 5G wireless communication module or a wired redundant network that is physically isolated from the main control network. When the dual-channel voting fails, it means that the backup shutdown determiner has not output a valid backup shutdown signal, and only the timing large model requests shutdown via a single channel. In this case, the hardware security isolation unit determines the large model command as a questionable command and suspends its execution, while triggering the system self-test program. The self-testing procedure includes checking whether there are any output anomalies in the timing model, checking whether the threshold comparison circuit of the backup shutdown decision unit is working properly, and checking the line integrity of the two signal channels. It generates detailed early warning information and uploads it to the operation and maintenance platform for operation and maintenance personnel to make a decision on whether to manually perform shutdown after comprehensive analysis.

[0070] In addition, the hardware safety isolation unit also has a built-in timeout watchdog subunit, which independently times the process of the valve actuator performing an emergency shutdown action. After the shutdown execution signal is issued, the watchdog subunit starts an independent timer. If no position arrival signal or execution completion signal is received from the actuator within the preset safe shutdown time, the watchdog subunit determines that the execution process has abnormally timed out and immediately triggers forced shutdown. The preferred value of this preset safe shutdown time is 1.5 times the valve's full stroke time. For example, if the valve's full stroke time is 15 seconds, the safe shutdown time is set to 22.5 seconds. During forced shutdown, the watchdog subunit bypasses the timing model, the main control logic of the hardware safety isolation unit, and the PLC main control program, and directly cuts off the drive power of the valve actuator through an independent solid-state relay, so that the actuator automatically returns to the preset safe position, i.e., the fully closed position, under the action of the spring reset mechanism. This watchdog forced shutdown mechanism serves as the last line of defense, ensuring that fail-safe functions can still be implemented in the event of multiple system anomalies, maximizing the safety of the pipeline system.

[0071] Through the aforementioned hardware security verification process, the intelligent decision-making of the time-series large model and the hardware-level secure execution are organically integrated. This not only leverages the advantages of the large model in accurate identification and intelligent decision-making under complex operating conditions, but also ensures the security, reliability, and real-time performance of shutdown decisions under any abnormal circumstances through multiple protection mechanisms, including a hardware security isolation unit independent of software logic, legality verification, dual-channel hardware voting, and watchdog forced shutdown. This effectively solves the security and reliability challenges of directly applying large models to the field of industrial control.

[0072] Based on steps S110 to S140 above, in this embodiment of the invention, a multivariate time-series dataset is constructed by collecting multi-dimensional time-series sensor data of valve operation. The multi-scale attention mechanism of an industrial time-series large-scale model is used to extract time-series features, and the time-series feature offset is calculated by comparing it with a standard operating condition feature library. This solves the technical problem that traditional solutions cannot accurately identify progressive latent anomalies. Furthermore, by integrating the optimal valve opening adjustment parameters, valve operating status classification results, and anomaly handling decision commands from the time-series large-scale model, the flow control accuracy and adaptive adjustment capability under complex operating conditions are effectively improved. More importantly, by setting up a hardware security isolation unit independent of the time-series large-scale model's operation, independent validity checks are performed when the large model outputs emergency shutdown commands. The system employs hardware-level dual-channel voting between the large model instructions and the backup shutdown decision-maker, along with a watchdog forced shutdown mechanism. This achieves a hierarchical control architecture of "intelligent decision-making by the large model + secure hardware execution." It leverages the intelligent decision-making advantages of capturing long-term dependencies and extracting multi-variable coupled features from the large-series model, while completely mitigating the security risks arising from the uncertainty of large model inference through hardware-level security isolation and dual-channel redundant voting. This approach balances the level of control intelligence with the high reliability and real-time requirements of industrial sites, significantly improving the accuracy of remote valve flow control, the sensitivity of anomaly identification, and the safety of emergency shutdown. It is adaptable to various industrial pipeline valve operating conditions in oil and gas, water, chemical, and HVAC industries, and has broad engineering application value.

[0073] The method of this invention includes a legality verification process that verifies whether the current upstream pressure of the valve is higher than a preset safe shut-off pressure threshold, whether the current instantaneous flow rate of the pipeline is lower than a preset shut-off flow rate lower limit, and whether the current valve drive current is within a preset normal operating current range. The dual-channel voting is a hardware-level voting process, where the hardware safety isolation unit is only allowed to output a shut-off execution signal when both the emergency shut-off command and the backup shut-off signal are valid. The method also includes a watchdog forced shut-off: the process of the valve actuator performing the emergency shut-off action is independently timed, and if no completion feedback is received within the preset safe shut-off time, the watchdog is triggered to force shut-off. The watchdog bypasses the main control program and directly drives the valve actuator to power off so that the valve is reset to the fully closed position.

[0074] Furthermore, the time-series large model extracts time-series features through a multi-scale attention mechanism, including: performing time-series decomposition on the input multivariate time-series matrix to separate trend components, periodic components, and residual components; performing attention calculations on the trend components, periodic components, and residual components with different receptive fields to capture the changing patterns of operating conditions at different time scales; and fusing the features of each component after multi-scale attention calculation to generate a multivariate time-series fusion feature vector.

[0075] Furthermore, the valve operating status classification results include normal status, minor abnormal status, early warning abnormal status, and emergency fault status. When the valve operating status classification result is a minor abnormal status, the abnormal handling decision instruction is a fine-tuning correction instruction, which automatically calibrates the operating condition by adaptively correcting the valve opening. When the valve operating status classification result is an early warning abnormal status, the abnormal handling decision instruction is an alarm warning instruction, which triggers a remote alarm and locks the valve regulation authority. When the valve operating status classification result is an emergency fault status, the abnormal handling decision instruction is an emergency shutdown instruction.

[0076] Furthermore, after inputting the multivariate time-series dataset into the industrial time-series large model, the process also includes: performing industrial rule verification on the optimal valve opening adjustment parameters output by the time-series large model, verifying whether the single adjustment amplitude exceeds the preset single maximum adjustment amplitude threshold and whether the adjustment speed exceeds the preset maximum adjustment speed threshold; if they exceed, the amplitude is limited according to the threshold; and performing anti-jitter filtering on the optimal valve opening adjustment parameters output by multiple consecutive time steps to remove random fluctuation components and generate control parameters adapted to the valve actuator.

[0077] Furthermore, it also includes online iterative optimization of the time series large model: during system operation, new operating condition data are continuously collected, and an importance score matrix is ​​maintained for each network parameter of the time series large model. The importance score is calculated based on the contribution of the parameter to the prediction accuracy in different historical operating conditions. When new operating condition data arrives, elastic regularization constraints are applied to parameters with importance scores higher than a preset threshold to limit their update magnitude. Based on the operating condition type identifier, historical samples with different types from the current new operating condition are extracted from the historical data repository, and mixed replay batches are constructed with the current new operating condition samples. The joint loss is calculated based on the mixed replay batches to update the parameters.

[0078] Furthermore, online iterative optimization also includes: calculating the average prediction error of the time series large model within the most recent preset number of time steps; when the average prediction error exceeds the preset performance alarm threshold, determining that the model has experienced performance degradation; triggering model version rollback, reverting the parameters of the time series large model to the latest checkpoint version before the performance degradation occurred, and pausing incremental learning until receiving an external recovery confirmation instruction.

[0079] Furthermore, after collecting multi-dimensional time-series sensor data on valve operation, the process also includes: calculating the timestamp deviation based on the actual sampling frequency of each sensor, aligning the sensor data to a unified time reference using an asynchronous data packet timestamp linear interpolation method with the system clock as the reference; when sensor data is missing, sequentially performing the following compensation methods: valid value retention compensation from the previous moment, moving average compensation based on historical data from the same period of the same type of sensor, and predicted value compensation based on the time-series large model for the missing time-series data; when the pairwise deviation of the compensation values ​​from the three compensation methods exceeds a preset deviation threshold, a sensor fault alarm is triggered; and when constructing a multivariate time-series dataset, the current operating condition type identifier code is embedded synchronously.

[0080] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0081] This invention also provides a valve remote flow control and abnormal shutdown processing device, which is used to implement the above embodiments and preferred embodiments, and will not be repeated as described previously. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0082] Figure 2 According to one embodiment of the present invention, a valve remote flow control and abnormal shutdown processing device includes:

[0083] The acquisition module 201 is used to acquire multi-dimensional time-series sensor data of valve operation and construct a multivariate time-series dataset;

[0084] The decision module 202 is used to input the multivariate time series dataset into the industrial time series large model. The time series large model extracts time series features through a multi-scale attention mechanism, calculates the time series feature offset by comparing with the standard operating condition feature library, and outputs decision results. The decision results include: optimal valve opening adjustment parameters, valve operating status classification results, and abnormal handling decision instructions.

[0085] The isolation module 203 is used to receive the emergency shutdown command, which operates independently of the time-series large model, when the abnormal handling decision command is an emergency shutdown command. The unit performs a legality check and conducts a dual-channel vote between the emergency shutdown command and the backup shutdown signal output by the independently operating backup shutdown decision unit. If the check and vote pass, the shutdown execution signal is transmitted to the valve actuator. If the check or vote fails, the shutdown is refused and a warning message is sent.

[0086] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0087] According to one embodiment of the present invention, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the above-described method for remote valve flow control and abnormal shutdown.

[0088] Optionally, in this embodiment, the processor can be configured to perform the following steps via a computer program:

[0089] Step S1: Collect multi-dimensional time-series sensor data of valve operation and construct a multivariate time-series dataset;

[0090] Step S2: Input the multivariate time series dataset into the industrial time series large model. The time series large model extracts time series features through a multi-scale attention mechanism, calculates the time series feature offset by comparing it with the standard operating condition feature library, and outputs the decision results, which include: optimal valve opening adjustment parameters, valve operating status classification results, and abnormal handling decision instructions.

[0091] Step S3: When the abnormal handling decision instruction is an emergency shutdown instruction, the hardware security isolation unit, which operates independently of the time-series large model, receives the emergency shutdown instruction, performs a legality check, and conducts a dual-channel vote between the emergency shutdown instruction and the backup shutdown signal output by the independently operating backup shutdown decision unit. If the check passes and the vote passes, the shutdown execution signal is transmitted to the valve actuator. If the check fails or the vote fails, the shutdown is refused and a warning message is sent.

[0092] According to one embodiment of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the storage medium is located to execute the above-described valve remote flow control and abnormal shutdown processing method.

[0093] Optionally, in this embodiment, the storage medium may be configured to store a computer program for performing the following steps:

[0094] Step S1: Collect multi-dimensional time-series sensor data of valve operation and construct a multivariate time-series dataset;

[0095] Step S2: Input the multivariate time series dataset into the industrial time series large model. The time series large model extracts time series features through a multi-scale attention mechanism, calculates the time series feature offset by comparing it with the standard operating condition feature library, and outputs the decision results, which include: optimal valve opening adjustment parameters, valve operating status classification results, and abnormal handling decision instructions.

[0096] Step S3: When the abnormal handling decision instruction is an emergency shutdown instruction, the hardware security isolation unit, which operates independently of the time-series large model, receives the emergency shutdown instruction, performs a legality check, and conducts a dual-channel vote between the emergency shutdown instruction and the backup shutdown signal output by the independently operating backup shutdown decision unit. If the check passes and the vote passes, the shutdown execution signal is transmitted to the valve actuator. If the check fails or the vote fails, the shutdown is refused and a warning message is sent.

[0097] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0098] According to one embodiment of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the above-described method for remote valve flow control and abnormal shutdown.

[0099] Optionally, in this embodiment, the above-mentioned computer program product can be configured as a computer program that performs the following steps:

[0100] Step S1: Collect multi-dimensional time-series sensor data of valve operation and construct a multivariate time-series dataset;

[0101] Step S2: Input the multivariate time series dataset into the industrial time series large model. The time series large model extracts time series features through a multi-scale attention mechanism, calculates the time series feature offset by comparing it with the standard operating condition feature library, and outputs the decision results, which include: optimal valve opening adjustment parameters, valve operating status classification results, and abnormal handling decision instructions.

[0102] Step S3: When the abnormal handling decision instruction is an emergency shutdown instruction, the hardware security isolation unit, which operates independently of the time-series large model, receives the emergency shutdown instruction, performs a legality check, and conducts a dual-channel vote between the emergency shutdown instruction and the backup shutdown signal output by the independently operating backup shutdown decision unit. If the check passes and the vote passes, the shutdown execution signal is transmitted to the valve actuator. If the check fails or the vote fails, the shutdown is refused and a warning message is sent.

[0103] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated here.

[0104] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0105] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between units or modules may be electrical or other forms.

[0106] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0107] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0108] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0109] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for remote flow control and abnormal shutdown handling of valves, characterized in that, include: Collect multi-dimensional time-series sensor data of valve operation and construct a multivariate time-series dataset; The multivariate time series dataset is input into the industrial time series large model. The time series large model extracts time series features through a multi-scale attention mechanism, calculates the time series feature offset by comparing it with the standard operating condition feature library, and outputs decision results. The decision results include: optimal valve opening adjustment parameters, valve operating status classification results, and abnormal handling decision instructions. When the abnormal handling decision instruction is an emergency shutdown instruction, the hardware security isolation unit, which operates independently of the timing model, receives the emergency shutdown instruction, performs a legality check, and conducts a dual-channel vote between the emergency shutdown instruction and the backup shutdown signal output by the independently operating backup shutdown decision unit. If the check passes and the vote passes, the shutdown execution signal is transmitted to the valve actuator. If the check fails or the vote fails, the shutdown is refused and a warning message is sent.

2. The valve remote flow control and abnormal shutdown handling method according to claim 1, characterized in that, The legality verification includes verifying whether the current upstream pressure of the valve is higher than the preset safe shut-off pressure threshold, whether the current instantaneous flow rate of the pipeline is lower than the preset shut-off flow rate lower limit, and whether the current valve drive current is within the preset normal operating current range. The dual-channel voting is a hardware-level voting mechanism. The hardware security isolation unit is only allowed to output the shutdown execution signal when both the emergency shutdown command and the backup shutdown signal are valid. It also includes a watchdog forced shutdown: the process of the valve actuator performing an emergency shutdown action is timed independently. If no feedback on completion is received within the preset safe shutdown time, the watchdog is triggered to force shutdown. The watchdog bypasses the main control program and directly drives the valve actuator to cut off power so that the valve is reset to the fully closed position.

3. The valve remote flow control and abnormal shutdown handling method according to claim 1, characterized in that, The aforementioned temporal large-scale model extracts temporal features through a multi-scale attention mechanism, including: The input multivariate time series matrix is ​​decomposed into trend components, periodic components and residual components. Attention calculations are performed on the trend component, periodic component, and residual component with different receptive fields to capture the working condition change patterns at different time scales. The components after multi-scale attention calculation are fused to generate a multivariate temporal fusion feature vector.

4. The valve remote flow control and abnormal shutdown handling method according to claim 1, characterized in that, The valve operating status classification results include normal status, minor abnormal status, warning abnormal status, and emergency fault status. When the valve operating status classification result is the slight abnormal state, the abnormal handling decision instruction is a fine-tuning correction instruction, which automatically calibrates the operating condition by adaptively correcting the valve opening. When the valve operating status classification result is the early warning abnormal state, the abnormal handling decision instruction is an alarm warning instruction, which triggers a remote alarm and locks the valve adjustment authority; When the valve operating status classification result is the emergency fault status, the abnormal handling decision instruction is the emergency shutdown instruction.

5. The valve remote flow control and abnormal shutdown handling method according to claim 1, characterized in that, After inputting the multivariate time series dataset into the industrial time series large model, the following is also included: The optimal valve opening adjustment parameters output by the time series large model are verified by industrial rules. The verification checks whether the single adjustment amplitude exceeds the preset single maximum adjustment amplitude threshold and whether the adjustment speed exceeds the preset maximum adjustment speed threshold. If they exceed the threshold, the amplitude is limited according to the threshold. The optimal valve opening adjustment parameters output at multiple consecutive time steps are subjected to anti-jitter filtering to remove random fluctuation components and generate control parameters adapted to the valve actuator.

6. The valve remote flow control and abnormal shutdown handling method according to claim 1, characterized in that, It also includes online iterative optimization of the aforementioned large-scale time-series model: During system operation, new operating condition data are continuously collected, and an importance score matrix is ​​maintained for each network parameter of the time series large model. The importance score is calculated based on the contribution of the parameter to the prediction accuracy in different historical operating conditions. When new operating condition data arrives, elastic regularization constraints are applied to parameters whose importance scores are higher than a preset threshold to limit their update magnitude. Based on the operating condition type identifier, historical samples different from the current new operating condition type are extracted from the historical data repository, and a mixed playback batch is constructed with the current new operating condition samples. The joint loss is calculated based on the mixed playback batch to update the parameters.

7. The valve remote flow control and abnormal shutdown handling method according to claim 6, characterized in that, The online iterative optimization also includes: Calculate the average prediction error of the time series model within the most recent preset number of time steps. When the average prediction error exceeds a preset performance alarm threshold, it is determined that the model has experienced performance degradation. Trigger model version rollback, revert the parameters of the time series model to the latest checkpoint version before the performance degradation occurred, and pause incremental learning until an external recovery confirmation instruction is received.

8. The valve remote flow control and abnormal shutdown handling method according to claim 1, characterized in that, After collecting multi-dimensional time-series sensor data on valve operation, the following is also included: Calculate the timestamp deviation based on the actual sampling frequency of each sensor, and use the asynchronous data packet timestamp linear interpolation method with the system clock as the reference to align the data of each sensor to a unified time reference. When sensor data is missing, the following steps are performed in sequence: compensation for maintaining the valid value of the previous moment, compensation based on the moving average of historical data from the same period of the same type of sensor, and compensation based on the predicted value of the forward time series data of the missing period based on the time series large model. When the pairwise deviation of the compensation values ​​of the three compensation methods exceeds the preset deviation threshold, a sensor fault alarm is triggered. When constructing the multivariate time series dataset, the current operating condition type identifier code is embedded synchronously.

9. A valve remote flow control and abnormal shutdown handling device, characterized in that, include: The data acquisition module is used to collect multi-dimensional time-series sensor data of valve operation and construct a multivariate time-series dataset; The decision module is used to input the multivariate time series dataset into the industrial time series large model. The time series large model extracts time series features through a multi-scale attention mechanism, calculates the time series feature offset by comparing it with the standard operating condition feature library, and outputs decision results. The decision results include: optimal valve opening adjustment parameters, valve operating status classification results, and abnormal handling decision instructions. The isolation module is used to receive the emergency shutdown command, which operates independently of the time-series large model, when the abnormal handling decision command is an emergency shutdown command. The unit performs a legality check on the emergency shutdown command and conducts a dual-channel vote with the backup shutdown signal output by the independently operating backup shutdown decision unit. If the check and vote pass, the shutdown execution signal is transmitted to the valve actuator. If the check or vote fails, the shutdown is refused and a warning message is sent.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored executable program, wherein, when the executable program is executed, it controls the device where the computer-readable storage medium is located to perform the valve remote flow control and abnormal shutdown processing method as described in any one of claims 1 to 8.

Images