Ball valve remote intelligent control method, system, device and medium based on internet of things

Abstract

The application discloses a ball valve remote intelligent control method, system, equipment and medium based on Internet of Things, and relates to the field of industrial control.The method comprises the following steps: collecting running data such as the opening degree, torque and pressure of a ball valve, carrying out denoising and complementing preprocessing on the data, eliminating invalid information and reserving effective data, performing optimal state estimation on the preprocessed data by combining a sliding window with a motion prior factor constraint, generating a control instruction by adopting a PID control algorithm and combining a theoretical torque model, transmitting the calibrated control instruction to an actuator through an Internet of Things communication module, collecting execution feedback data in real time through a sensor, comparing the control target to calculate a deviation, dynamically adjusting the control instruction, forming a closed-loop optimization, and realizing precise and stable remote intelligent control.

Description

Technical Field

[0001] This invention relates to the field of industrial control, and more specifically, to a method, system, device, and medium for remote intelligent control of ball valves based on the Internet of Things. Background Technology

[0002] In the field of industrial control, ball valves, as the core control element of fluid transportation systems, are widely used in various industrial pipeline networks such as chemical, municipal, and petroleum industries. Their control accuracy and stability directly affect the operating efficiency and safety of the entire pipeline network. With the rapid development of Internet of Things (IoT) technology, traditional manual and ordinary electric ball valves have gradually been upgraded to remote intelligent control. Some IoT-based ball valve control solutions have emerged in the existing technology. By collecting operating data through sensors and using communication modules to achieve remote command transmission, the problems of low efficiency and high labor intensity of traditional manual operation have been initially solved.

[0003] However, in practical use, it still has some shortcomings, such as: First, the data preprocessing is imperfect. The collected data such as opening degree, torque, and pressure are easily affected by external interference and are not effectively denoised and completed, resulting in a large deviation in subsequent state estimation; Second, the state estimation method is simplistic. It does not combine motion prior factor constraints and only processes data through simple algorithms, making it difficult to obtain accurate ball valve operating states; Third, the control command generation lacks an effective calibration mechanism and is not verified in conjunction with the theoretical torque model, which easily leads to commands exceeding the actuator's operating range; Fourth, the closed-loop control is imperfect. The feedback adjustment is lagging and cannot dynamically adapt to changes in pipeline network conditions, resulting in low control accuracy and slow response. It is difficult to meet the high precision and high stability requirements of industrial production for remote intelligent control of ball valves. At the same time, some solutions also have problems such as complex structure and inconvenient maintenance. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, the present invention provides a remote intelligent control method, system, device, and medium for ball valves based on the Internet of Things, and solves the problems mentioned in the background art through the following solutions.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a remote intelligent control method, system, device, and medium for ball valves based on the Internet of Things, comprising:

[0006] S1: Deploy multi-source sensors at corresponding locations on the ball valve body and pipeline network, collect multi-source sensor output data, form raw time series data, and perform preprocessing to obtain observation sequence data;

[0007] S2: Based on the observed sequence data, establish the mapping relationship between opening degree and torque, and construct a theoretical torque model by combining pressure influence parameters; based on the theoretical torque model, integrate the mechanical characteristics of the ball valve to directly construct the dynamic prior constraint equation, and extract the motion prior factor from the dynamic prior constraint equation;

[0008] S3: Based on the observed sequence data, construct an observation probability model, calculate the observation residuals and generate robust weighting factors, construct an observation model with robust correction based on the robust weighting factors, and output robust observation constraints;

[0009] S4: Unify the motion prior factors with the robust observation constraints, construct a factor graph, embed the robust weight factors into the factor graph, and construct a joint objective function;

[0010] S5: Set a fixed-length sliding window, optimize the state sequence in the factor graph and update the window content, solve the joint objective function iteratively, determine the iteration termination condition based on the convergence criterion, and output the optimal state estimate.

[0011] S6: Generate control commands based on the optimal state estimate and the preset control target, transmit the control commands to the ball valve actuator, collect the actuator feedback state and compare it with the control target to form a closed-loop control circuit, and continuously correct the control commands.

[0012] The IoT-based remote intelligent control system for ball valves is characterized by comprising:

[0013] Multi-source sensor acquisition module: Deployed in the ball valve body and corresponding locations in the pipeline network, it includes multi-source sensors and IoT terminals to collect operating data such as ball valve opening degree, torque, and pressure, and outputs standardized observation sequence data after preprocessing;

[0014] Model building and factor extraction module: Receives observation sequence data, establishes data mapping relationships, constructs a theoretical torque model, extracts motion prior factors, and generates robust observation constraints;

[0015] Robust optimization module: Constructs an observation probability model based on observation sequence data, calculates observation residuals and generates robust weighting factors, constructs an observation model with robust correction, and outputs robust observation constraints;

[0016] Factor graph optimization module: Receives motion prior factors and robust observation constraints, constructs a factor graph, embeds robust weights into the observation factors, and outputs the optimal state estimate through iterative solution;

[0017] Closed-loop control module: Receives the optimal state estimate, generates control commands and transmits them to the actuator, and simultaneously collects execution feedback status, compares it with the preset target, and continuously corrects the control commands;

[0018] IoT communication module: Responsible for the two-way transmission of data transmission, control commands, and feedback information between modules, ensuring the coordinated operation of the entire system.

[0019] An electronic device, characterized in that it includes a processor and a memory, the memory storing a computer program, which, when executed by the processor, implements the method according to any one of claims 1 to 7.

[0020] A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, which, when executed by a processor, implements the method of any one of claims 1 to 7.

[0021] The technical effects and advantages of this invention are as follows:

[0022] This invention achieves accurate state estimation by denoising and completing preprocessing of collected data such as opening degree, torque, and pressure, combined with sliding window and motion prior factor constraints, significantly improving data accuracy and state estimation precision. It employs a PID control algorithm combined with a theoretical torque model to generate and calibrate control commands, avoiding command anomalies and ensuring the safe and stable operation of the actuator. An IoT communication module enables error-free, low-latency command transmission, coupled with real-time feedback and closed-loop optimization mechanisms, allowing dynamic adjustment of control parameters and effectively solving the problems of lag and insufficient precision in traditional control. Furthermore, this invention features a reasonable structure and convenient operation, enabling remote intelligent control of ball valves, reducing manual labor intensity, improving pipeline network operating efficiency, and adapting to various industrial scenarios, thus possessing high application value. Attached Figure Description

[0023] Figure 1 This is a flowchart of a remote intelligent control method for ball valves based on the Internet of Things (IoT).

[0024] Figure 2 This is a structural diagram of a remote intelligent control system for ball valves based on the Internet of Things (IoT).

[0025] Figure 3 This is a schematic diagram of the electronic device structure of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] refer to Figures 1-2The IoT-based remote intelligent control method for ball valves shown includes:

[0028] S1: Deploy multi-source sensors at corresponding locations on the ball valve body and pipeline network, collect multi-source sensor output data, form raw time series data, and perform preprocessing to obtain observation sequence data;

[0029] Further explanation is needed in this embodiment: the data acquisition components include a ball valve opening sensor, a torque sensor, a pipeline pressure sensor, and a temperature sensor. The opening sensor is installed on the actuator of the ball valve body, the torque sensor is installed on the valve stem of the ball valve body, the pressure sensors are installed at the inlet and outlet of the ball valve, and the temperature sensors are installed at key monitoring points in the pipeline network. The original time-series data includes several unit-time rotation angle values ​​of the ball valve stem collected by the opening sensor, several unit-time torque values ​​of the valve stem collected by the torque sensor, several unit-time medium pressure values ​​of the ball valve inlet and outlet collected by the pressure sensor, and several unit-time temperature values ​​of the medium in the pipeline network collected by the temperature sensor. The preprocessed observation sequence data is used to remove abnormal data with large deviation values. The removal method is conventional noise reduction and normalization processing, which is not specifically limited in this embodiment.

[0030] S2: Based on the observed sequence data, establish the mapping relationship between opening degree and torque, and construct a theoretical torque model by combining pressure influence parameters; based on the theoretical torque model, integrate the mechanical characteristics of the ball valve to directly construct the dynamic prior constraint equation, and extract the motion prior factor from the dynamic prior constraint equation;

[0031] The observation sequence data is retrieved, and ball valve opening, torque, inlet pressure, and outlet pressure data samples are precisely selected and extracted. These samples are then organized according to their corresponding timestamps to form a correlated dataset. During this process, timestamp matching is verified, and samples with misaligned timestamps or missing data are removed. The least squares method is used to fit the opening and torque data samples in the correlated dataset. Samples meeting the requirements are selected, and these two sets of data are fitted with the opening data as the independent variable and the torque data as the dependent variable. After fitting, the correlation of the fitting results is verified, and an allowable range for fitting error is set. The fitted value of each data sample is compared with the actual observed value, and data samples with fitting errors exceeding the allowable range are removed. The remaining valid data samples are refitted, and the fitting-verification-removal process is repeated until the fitting results stabilize. Finally, the mapping relationship between opening and torque is determined, and the torque benchmark value corresponding to each opening interval is identified.

[0032] Select ball valve inlet and outlet pressure data samples from the associated dataset, calculate the inlet and outlet pressure difference for each corresponding sample, and statistically analyze the actual torque changes corresponding to different pressure differences. Through comparative analysis, establish the correlation between the pressure difference and the ball valve driving force increment. Based on the correlation, determine the pressure-driving force correlation coefficient, clarify the specific change in the ball valve driving force increment when the pressure difference changes by one unit, and complete the extraction of pressure influence parameters. Pre-obtain the inherent parameters of the ball valve (including core parameters such as medium density and valve stem size), use the torque benchmark value as the no-load torque, and combine the inlet and outlet pressure difference, the pressure-driving force correlation coefficient, and the obtained ball valve inherent parameters to construct a preliminary theoretical torque model. Extract the opening and pressure data from the observation sequence, substitute them one by one into the theoretical torque model for calculation, compare the calculation results with the actual torque data in the observation sequence, and adjust the pressure-driving force correlation coefficient according to the comparison deviation until the deviation between the model calculation value and the actual torque value is within the preset allowable range, thus completing the construction of the theoretical torque model.

[0033] The mechanical characteristic parameters of the ball valve are obtained in advance, including the valve stem moment of inertia, the friction coefficient of the seal, and the elastic modulus of the valve body material. Based on the theoretical torque model, the obtained mechanical characteristic parameters of the ball valve are integrated into the constraint logic to establish a priori dynamic constraint equations. By calculating the valve stem moment of inertia, the friction coefficient of the seal, and the elastic modulus of the valve body material, the allowable range of torque and the allowable range of rotational acceleration are determined. The two allowable ranges are integrated into the constraint equations to clarify the upper and lower limits of torque and rotational acceleration during the operation of the ball valve.

[0034] The observation sequence data is retrieved, and the opening, pressure, and torque data are substituted into the constraint equations one by one. Based on the timestamp order of the observation sequence data, the dynamic constraint states at different times are solved sequentially. A numerical iteration method is used to calculate the optimal torque constraint value and rotational acceleration constraint value that satisfy the boundary conditions. Statistical analysis is performed on the obtained torque constraint value and rotational acceleration constraint value at each time point to determine the torque limit value and rotational acceleration limit value under all working conditions. The minimum and maximum values ​​of the torque limit value, the minimum and maximum values ​​of the rotational acceleration limit value are extracted from the solution results. These four parameters are used as prior motion factors, categorized and stored according to parameter type, and properly labeled after storage.

[0035] S3: Based on the observed sequence data, construct an observation probability model, calculate the observation residuals and generate robust weighting factors, construct an observation model with robust correction based on the robust weighting factors, and output robust observation constraints;

[0036] The observed sequence data is retrieved, and data entries identified by the data identifier field are filtered out as opening degree, torque, pressure, and temperature. Pressure data needs to be further differentiated into inlet and outlet pressure data. The specific values ​​of each data entry are extracted according to the field order of "opening degree - torque - inlet pressure - outlet pressure - temperature". All extracted data are then sorted in ascending order by timestamp to form an observed data subset. The integrity of the data subset is verified, and entries are removed based on the following criteria: missing timestamps, empty data fields, invalid identifiers, and entries inconsistent with the preprocessed observed sequence data. After removal, the data subset is verified again.

[0037] The observation probability model is constructed using probabilistic statistical methods in the following steps: 1. Determine the model type: A Gaussian mixture model is selected as the observation probability model, and the core structure of the model is defined as including the observation layer, probability distribution layer, and parameter layer; 2. Input basic parameters: A subset of the observation data is used as the model input data, and the measurement accuracy parameters of each sensor are also input, including: opening sensor accuracy ±0.5°, torque sensor accuracy ±0.1 N·m, pressure sensor accuracy ±0.01 MPa, and temperature sensor accuracy ±0.5℃; 3. Set the error distribution type: The model error distribution is set to Gaussian distribution, and the initial value of the mean of the error distribution is defined as 0, while the initial value of the variance is set according to the measurement accuracy parameters of each sensor; 4. Determine the core parameters of the model: Substitute the subset of observation data into the model one by one, and iteratively calculate the model's mixing coefficients, mean, variance, and other core parameters using the maximum likelihood estimation method. The number of iterations is set to 50-100 times until the parameters converge; 5. Model validation: Select 30% of the observed data subset as validation data, substitute it into the constructed model, and calculate the deviation between the model prediction value and the actual value of the validation data.

[0038] The observation probability model is invoked, and a subset of the observation data is substituted into the model for calculation to obtain the model prediction value corresponding to each observation data. The difference between the actual value and the model prediction value of each observation data is calculated one by one to obtain the observation residual of each observation data. The observation residuals are arranged in the order of timestamps to form an observation residual sequence. The threshold range of the observation residuals is set as follows: opening observation residual threshold ±1.0°, torque observation residual threshold ±0.2N·m, inlet pressure observation residual threshold ±0.02MPa, outlet pressure observation residual threshold ±0.02MPa, and temperature observation residual threshold ±1.0℃. The observation residual sequence is iterated one by one, and each observation residual is compared with the preset threshold value of the corresponding parameter. For observation data with observation residuals within the preset threshold range, a weight value of 0.8-1.0 is assigned; for observation data with observation residuals exceeding the preset threshold but not exceeding twice the threshold, a weight value of 0.4-0.7 is assigned; for observation data with observation residuals exceeding twice or more the threshold, a weight value of 0.1-0.3 is assigned. Weight normalization: The weight values ​​assigned to all observation data are normalized to ensure that the sum of the weight values ​​of all observation data is 1. A robust weight factor sequence corresponding one-to-one with the observation data is generated and stored in association with the observation data according to the timestamp.

[0039] The existing observation probability model is invoked, the model error correction module is unlocked, and the robust weight factor sequence generated by S304 is embedded into the model error correction stage through the parameter adjustment interface to establish the correlation between the weight factor and the model error calculation. Combined with the observation residual sequence, a reasonable correction coefficient is set, the model error distribution parameters are dynamically adjusted, the impact of low-weight observation data on the model is reduced, and an observation model with robust correction is constructed.

[0040] Select 20% of the observed data subset as validation data, substitute it into the corrected model, calculate the deviation between the model's predicted value and the actual value of the validation data, and simultaneously simulate the addition of observed data with residuals exceeding three times the threshold; if the deviation exceeds the preset range or the abnormal data suppression effect is not good, adjust the weight allocation rules and correction coefficients, and repeat the correction-validation process.

[0041] For observation models with robust corrections, parameters are extracted, and the robust weight allocation rules, observation residual thresholds, and error correction logic in the model are extracted as the core content of robust observation constraints. The core content is organized in a standardized format to form standardized robust observation constraints, and then labeled and stored.

[0042] S4: Unify the motion prior factors with the robust observation constraints, construct a factor graph, embed the robust weight factors into the factor graph, and construct a joint objective function;

[0043] The process involves calling upon the motion prior factors and robust observation constraints, extracting their parameter types, data formats, and numerical units, and establishing a parameter information lookup table. The parameter formats are standardized by converting the numerical units of the motion prior factors to the standard units consistent with those of the robust observation constraints, with torque units standardized to N·m and rotational acceleration units standardized to rad / s², ensuring no deviation in numerical units. The parameter relationships are clarified by associating the minimum and maximum allowable torque ranges in the motion prior factors with the torque observation residual thresholds in the robust observation constraints, and the minimum and maximum allowable rotational acceleration ranges with the acceleration error correction parameters in the error correction logic of the robust observation constraints, establishing a one-to-one correspondence mapping table. The consistency of the associated parameters is verified by comparing the constraint ranges of the motion prior factors with the error thresholds of the robust observation constraints, eliminating misaligned or conflicting parameters, and readjusting and matching conflicting parameters. Once the unification is complete, all consistent parameters are categorized into "motion prior constraint class" and "robust observation constraint class", and the correlation relationships of each type of parameter are marked to form a standardized joint constraint dataset, which is then stored in a designated storage unit and properly labeled.

[0044] A graphical modeling approach was adopted, and a factor graph construction tool was launched to build the basic framework of the factor graph. The framework was defined to include two core layers: a node layer and a factor layer. A communication interface between the two layers was established to ensure bidirectional parameter transmission. The node layer was constructed as follows: The node layer is divided into state nodes and observation nodes. State nodes are categorized according to the ball valve's operating parameters, specifically including opening state nodes, torque state nodes, inlet pressure state nodes, outlet pressure state nodes, and temperature state nodes. Each state node has a unique identifier, which corresponds one-to-one with the parameter identifiers of the observed data subset. Observation nodes are set according to the type of observed data. Each observation node corresponds to a set of observed data in the observed data subset, carrying the timestamp, parameter value, and data identifier of the corresponding observed data, and establishing a data association with the state nodes. The factor layer was constructed as follows: The factor layer is divided into prior factor nodes and observed factor nodes. Prior factor nodes are set according to the type of motion prior factors, namely torque constraint factor nodes and rotational acceleration constraint factor nodes. Factor nodes are constructed by embedding a prior factor from the joint constraint dataset into each prior factor, clearly defining the constraint threshold for each prior factor node. Observation factor nodes are set according to the observation parameter type, namely, opening observation factor nodes, torque observation factor nodes, pressure observation factor nodes, and temperature observation factor nodes. Each observation factor node reserves an interface for embedding robust weight factors. Hierarchical associations are established: prior factor nodes are associated with corresponding state nodes, where torque constraint factor nodes are associated with torque state nodes, and rotational acceleration constraint factor nodes are associated with torque state nodes and opening state nodes. Observation factor nodes are associated with corresponding state nodes and observation nodes, with each observation factor node associated with one state node and one observation node, ensuring that observation data can be transmitted to state nodes through observation factor nodes. After the factor graph is constructed, the node associations and parameter embedding are comprehensively verified to identify and correct issues such as misaligned node identifiers, broken associations, and missing parameters, and nodes with problems are adjusted and repaired.

[0045] The robust weighting factor sequence is retrieved, organized chronologically by timestamp, and the observation data identifier and parameter type corresponding to each robust weighting factor are extracted to establish a correlation table between weighting factors and observation data. The weighting factor embedding program is then initiated, embedding the robust weighting factor sequence one by one into the observation factor nodes through the embedding interface reserved by the observation factor nodes. The specific embedding rules are as follows: the opening observation factor node embeds the robust weighting factor corresponding to the opening observation data; the torque observation factor node embeds the robust weighting factor corresponding to the torque observation data; the pressure observation factor node embeds the robust weighting factor corresponding to the inlet and outlet pressure observation data respectively; and the temperature observation factor node embeds the robust weighting factor corresponding to the temperature observation data. During the embedding process, the embedding status of each observation factor node is verified in real time, comparing the embedded weighting factor values ​​with the observed data. Check if the data identifiers match the association lookup table. If there are numerical misalignments or identifier mismatches, immediately stop embedding and adjust to ensure that the robust weight factor embedded in each observation factor node completely matches the corresponding observation data. After all weight factors are embedded, start the global factor graph verification program, traverse all observation factor nodes, read the weight factor information of each node, and compare it one by one with the robust weight factor sequence to confirm that there are no omissions, no mis-embedding, and no duplicate embedding. After verification, mark the observation factor nodes after embedding weight factors to clearly mark that each node has completed weight factor embedding, and update the node parameter information of the factor graph.

[0046] Based on the node relationships in the factor graph, and combining the constraints of the prior motion factors with the error correction logic of the robust observation constraints, a joint objective function is constructed. The objective function focuses on minimizing the deviation between the estimated value and the actual observed value of the state node, incorporating the differentiated influence of the robust weight factor, and clarifying the variables, constraints and calculation logic of the objective function. After construction, the parameters of the objective function are initialized to ensure that subsequent iterations can be solved normally.

[0047] S5: Set a fixed-length sliding window, optimize the state sequence in the factor graph and update the window content, solve the joint objective function iteratively, determine the iteration termination condition based on the convergence criterion, and output the optimal state estimate.

[0048] A fixed-length sliding window is preset, with a window length of 50-100 data sampling points, corresponding to a window time span of 5-100 seconds; the window sliding step size is set to 1 data sampling point, and the start and end sampling point identifiers of the window are clearly defined. The start sampling point identifier corresponds to the first sampling point of the observed data subset, and the end sampling point identifier = start sampling point identifier + window length - 1; the window update rule is set, and when sliding, the earliest data with the timestamp and its corresponding state sequence in the window are removed, and the latest sampled data and its corresponding state sequence are included, keeping the window data length fixed and synchronized with the timestamp.

[0049] The factor graph is invoked to extract initial parameters for all state nodes, including: opening degree, torque, inlet pressure, outlet pressure, and temperature. These parameters are then organized by timestamp as initial values ​​for the state sequence. The initial state sequence values ​​are filled into the sliding window in ascending order of timestamp. The consistency between the timestamps and the observed data subset is compared, and mismatched timestamp identifiers are adjusted. The window initialization is completed, and relevant identifiers and parameters are recorded. The factor graph optimization program is started, the constructed joint objective function is loaded, and the state sequence within the window is used as the optimization object. The joint constraint dataset is called for optimization calculation, and the optimized state sequence and corresponding timestamps are recorded. The window is moved in steps of one sampling point, and nodes are removed. The earliest data is incorporated, and the latest data is included. The integrity of the window data and the continuity of timestamps are verified. The optimization-update-verification process is repeated until the window traverses all observation data. The state sequence after the last window optimization is selected as the initial value for iteration. Gradient descent is used to substitute the values ​​into the joint objective function for calculation, and the first function calculation value is recorded. The gradient value of the function with respect to the parameters of each state node is calculated, and the parameters are adjusted according to a preset step size of 0.001-0.01 to obtain the new iteration value and calculate the corresponding function value. The iteration process is repeated, and the number of iterations, function values, and state parameters are recorded for each iteration. A dual convergence criterion is set: the difference between two consecutive iteration function values ​​is less than... The iteration may terminate when the number of iterations reaches 200. After each iteration, the difference in function values ​​is calculated and the number is recorded. If the criterion is met, the iteration stops and the corresponding state parameters are recorded. If the criterion is not met, the iteration continues. When the maximum number of iterations is reached, the iteration stops and the last state parameters are recorded. After the iteration stops, the reason for termination is marked and all calculation records are retained.

[0050] After the iteration terminates, the state sequence parameters of the last iteration are extracted, including: opening degree, torque, inlet pressure, outlet pressure, and optimal temperature estimates. The parameters are then categorized and organized by type, marked with corresponding timestamps, and compared with the timestamps of the observed data subset. The optimal estimated parameters are verified to be within the constraint range of the joint constraint dataset. Parameters that are out of range are removed and replaced with the parameters that are closest to the constraint range during the iteration. After verification, standardized optimal state estimates are formed.

[0051] S6: Generate control commands based on the optimal state estimate and the preset control target, transmit the control commands to the ball valve actuator, collect the actuator feedback state and compare it with the control target to form a closed-loop control circuit, and continuously correct the control commands.

[0052] Based on the operational requirements of the industrial pipeline network, preset ball valve control targets are established. The core target parameter is the target opening value, which ranges from 0 to π / 2 radians. It can be remotely set by the IoT cloud platform or automatically adapted according to the pipeline network pressure and flow requirements. Auxiliary control targets include pipeline network pressure targets and flow targets, with the pipeline network pressure target set at 0.8MPa and the flow target set at 50m³ / h.

[0053] Based on the core parameters in the optimal state estimate, a comparison is made with the preset control target parameters to calculate the control deviation. The calculation method is to subtract the optimal state estimate parameters from the control target parameters. A PID control algorithm is used to generate actuator control commands, i.e., actuator output torque control values, based on the corrected control deviation. The calculation process of the PID control algorithm includes three parts: proportional term, integral term, and derivative term. The proportional term is the product of the proportional coefficient and the corrected control deviation, with a proportional coefficient of 5.0. The integral term is the product of the integral coefficient and the accumulated value of the corrected control deviation from the first sampling period to the current sampling period, and then the sampling period, with an integral coefficient of 0.1. The derivative term is the result of the derivative coefficient and the difference between the current corrected control deviation and the corrected control deviation of the previous sampling period, divided by the sampling period, with a derivative coefficient of 0.5. The generated actuator control commands are then compared with the theoretical torque model. The absolute difference between the two values ​​is calculated. If the absolute difference is greater than 2 N·m, the control command is calibrated by multiplying the generated control command by 0.8 and adding the theoretical torque model calculation value multiplied by 0.2 to obtain the calibrated control command. The calibrated control command is transmitted to the ball valve electric actuator via NB-IoT or LoRa network. Cyclic redundancy check is used during transmission to ensure error-free command transmission. At the same time, the control transmission delay does not exceed 100ms to meet real-time control requirements. After receiving the control command, the actuator performs the opening adjustment operation in real time. Simultaneously, the actual opening and actual output torque after execution are collected by the torque sensor and opening sensor integrated on the actuator, respectively. The collected actual opening and actual output torque are transmitted to the edge computing node through the IoT terminal module to form a feedback data sequence, which includes the actual opening and actual output torque of the current sampling period.

[0054] Calculate the absolute difference between the actual opening degree and the preset target opening degree, and set the allowable range of this difference to be no more than 0.01 radians; at the same time, determine whether the actual output torque collected is within a reasonable range of 0 to 50 N·m, and the specific judgment criteria are as follows:

[0055] If the absolute difference between the actual opening degree and the target opening degree does not exceed 0.01 radians, and the actual output torque is within a reasonable range of 0 to 50 N·m, it indicates that the control command has been executed in place. The current control command is maintained, and the next sampling cycle is entered.

[0056] If the absolute difference between the actual opening degree and the target opening degree exceeds 0.01 radians, or the actual output torque exceeds the reasonable range of 0 to 50 N·m, it indicates that there is a deviation in the execution of the control command. At this time, the collected feedback data sequence is input into the preprocessing process of step S1, and the complete process from S1 to S6 is re-triggered to update the optimal state estimate and control command, thereby achieving closed-loop optimization.

[0057] If the deviation exceeds the allowable range for three consecutive sampling cycles, it is determined that the actuator is faulty or the sensor is abnormal. A fault alarm signal is immediately generated and transmitted to the IoT cloud platform. At the same time, the control command is adjusted to a safe value (i.e., the actuator output torque control value is set to 0 N·m, and the actuator operation is stopped to avoid equipment damage). Through closed-loop control testing for 10 consecutive sampling cycles, it is ensured that the deviation between the actual opening degree and the target opening degree is stable within 0.01 radians, and the actual output torque is maintained within a reasonable range of 5 to 35 N·m. The effectiveness of the closed-loop control is verified, and high-precision and robust IoT remote intelligent control of the ball valve is achieved.

[0058] Based on the above method, this application also discloses a remote intelligent control system for ball valves based on the Internet of Things (IoT), see reference. Figure 3 The IoT-based remote intelligent control system for ball valves shown includes:

[0059] Multi-source sensor acquisition module: Deployed in the ball valve body and corresponding locations in the pipeline network, it includes multi-source sensors and IoT terminals to collect operating data such as ball valve opening degree, torque, and pressure, and outputs standardized observation sequence data after preprocessing;

[0060] Model building and factor extraction module: Receives observation sequence data, establishes data mapping relationships, constructs a theoretical torque model, extracts motion prior factors, and generates robust observation constraints;

[0061] Robust optimization module: Constructs an observation probability model based on observation sequence data, calculates observation residuals and generates robust weighting factors, constructs an observation model with robust correction, and outputs robust observation constraints;

[0062] Factor graph optimization module: Receives motion prior factors and robust observation constraints, constructs a factor graph, embeds robust weights into the observation factors, and outputs the optimal state estimate through iterative solution;

[0063] Closed-loop control module: Receives the optimal state estimate, generates control commands and transmits them to the actuator, and simultaneously collects execution feedback status, compares it with the preset target, and continuously corrects the control commands;

[0064] IoT communication module: Responsible for the two-way transmission of data transmission, control commands, and feedback information between modules, ensuring the coordinated operation of the entire system.

[0065] The electronic device 1000 may include: at least one processor 1001, at least one communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005.

[0066] The communication bus 1002 is used to realize the connection and communication between the above components, ensuring the stable transmission of timing data, state estimation results, control commands and feedback information, and avoiding data packet loss and command errors.

[0067] The user interface 1003 may include a display screen and a touch interaction unit, and optionally also includes a standard wired debugging interface and a wireless Bluetooth interaction interface. The user interface 1003 is used to realize the operation interaction of on-site maintenance personnel, visualize the operation data such as ball valve opening degree / torque / pressure, configure control parameters on-site, provide early warning pop-up prompts for deviation exceeding limits, and view fault information. Optionally, it can also be connected to an audible and visual alarm unit to realize on-site early warning of actuator abnormalities and sensor failures.

[0068] The network interface 1004 may include a standard wired Ethernet interface, an industrial fieldbus interface, or a wireless interface. The network interface 1004 is used to realize bidirectional communication between electronic devices and multi-source sensor acquisition modules, ball valve electric actuators, edge computing nodes, and IoT cloud platforms, and to complete core data interactions such as raw time-series data acquisition, uploading of optimal state estimation results, issuance of control commands, feedback data transmission, and fault alarm push.

[0069] The processor 1001 may include one or more processing cores. The processor 1001 connects to various functional units within the electronic device through various interfaces and lines. It executes all functions and data processing steps of the IoT-based remote intelligent control method for ball valves described in this invention by running or executing instructions, programs, code sets, or instruction sets stored in the memory 1005, and by calling data stored in the memory 1005. These functions include, but are not limited to: preprocessing of raw time-series data, construction of theoretical torque models, extraction of motion prior factors, calculation of observation residuals and generation of robust weight factors, construction of factor graphs and iterative solution of joint objective functions, sliding window optimization and optimal state estimation output, generation of PID control instructions and calibration of torque models, calculation of closed-loop control deviations and dynamic correction of instructions, anomaly detection and fault safety protection, etc. Optionally, the processor 1001 may be implemented using at least one hardware form selected from digital signal processing (DSP), field-programmable gate array (FPGA), and programmable logic array (PLA); the processor 1001 may integrate one or more combinations of a central processing unit (CPU) and an embedded communication modem. The CPU primarily handles the operating system, control business logic, user interaction programs, factor graph optimization, and iterative solutions; the modem is used to handle the parsing and data transmission of various IoT communication protocols.

[0070] The memory 1005 may include random access memory (RAM) or read-only memory (ROM); optionally, the memory 1005 may include a non-transitory computer-readable medium. The memory 1005 can be used to store instructions, programs, code, code sets, or instruction sets, and includes a program storage area and a data storage area: the program storage area can store instructions for implementing an operating system, instructions for implementing at least one function, and instruction programs for implementing all steps of the above-described method embodiments of the present invention; the data storage area can store various types of data involved in the implementation of the method of the present invention, including but not limited to: original time-series data of opening / torque / pressure, preprocessed observation sequence data, theoretical torque model parameters, motion prior factors, robust weighting factors, factor graph nodes and constraint parameters, sliding window optimization data, optimal state estimates, PID control coefficients, control target thresholds, allowable deviation ranges, fault triggering conditions, historical execution records, and closed-loop optimization logs, etc.

[0071] In the electronic device disclosed in this embodiment, the processor 1001 can call the executable program of the IoT-based ball valve remote intelligent control method stored in the memory 1005. When the executable program is executed by one or more processors 1001, the electronic device performs all the steps of the IoT-based ball valve remote intelligent control method as described in any of the above embodiments, realizing the integrated intelligent operation of ball valve data acquisition, state estimation, instruction generation, closed-loop control and fault protection.

[0072] An electronic device readable storage medium stores instructions that, when executed by one or more processors, cause the electronic device to perform one or more of the methods described in the above embodiments.

[0073] It should be noted that the computer-readable storage medium may include, but is not limited to, USB flash drives, portable hard drives, magnetic disks, optical disks, solid-state drives (SSDs), embedded read-only memory, random access memory, NB-IoT terminal built-in flash memory, edge computing node local storage, and various non-volatile memories, etc., any medium capable of storing program code. When the computer program runs on hardware carriers such as electronic devices, edge computing terminals, IoT gateways, cloud servers, and ball valve electric actuator controllers, it can completely execute the entire process of data acquisition, preprocessing, factor graph optimization, state estimation, instruction generation, closed-loop calibration, and fault protection of the method embodiments of the present invention. This enables IoT remote intelligent control of industrial ball valves with full coverage, low-latency transmission, high-precision adjustment, and highly robust operation, adapting to the automation control needs of long-term unattended operation in industrial sites.

[0074] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0075] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A remote intelligent control method for ball valves based on the Internet of Things, characterized in that, include: S1: Deploy multi-source sensors at corresponding locations on the ball valve body and pipeline network, collect multi-source sensor output data, form raw time series data, and perform preprocessing to obtain observation sequence data; S2: Based on the observed sequence data, establish the mapping relationship between opening degree and torque, and construct a theoretical torque model by combining pressure influence parameters; based on the theoretical torque model, integrate the mechanical characteristics of the ball valve to directly construct the dynamic prior constraint equation, and extract the motion prior factor from the dynamic prior constraint equation; S3: Based on the observed sequence data, construct an observation probability model, calculate the observation residuals and generate robust weighting factors, construct an observation model with robust correction based on the robust weighting factors, and output robust observation constraints; S4: Unify the motion prior factors with the robust observation constraints, construct a factor graph, embed the robust weight factors into the factor graph, and construct a joint objective function; S5: Set a fixed-length sliding window, optimize the state sequence in the factor graph and update the window content, solve the joint objective function iteratively, determine the iteration termination condition based on the convergence criterion, and output the optimal state estimate. S6: Generate control commands based on the optimal state estimate and the preset control target, transmit the control commands to the ball valve actuator, collect the actuator feedback state and compare it with the control target to form a closed-loop control circuit, and continuously correct the control commands.

2. The remote intelligent control method for ball valves based on the Internet of Things according to claim 1, characterized in that, The preprocessing includes: Outlier detection is performed on the collected raw time series data to identify and remove abnormal data points containing sensor drift and transient interference, while retaining valid time series segments; The effective data after removing outliers is denoised by smoothing filtering to eliminate high-frequency noise interference and restore the true data characteristics. The denoised data is normalized to unify the data units and numerical ranges, generating standardized observation sequence data.

3. The remote intelligent control method for ball valves based on the Internet of Things according to claim 1, characterized in that, The establishment of the mapping relationship between opening degree and torque includes: Extract aperture and torque-related data from the observation sequence data, filter data samples to obtain valid data, and remove abnormal data to ensure data authenticity; Based on the aforementioned valid data, a fitting algorithm was used to establish a correlation model between opening degree and torque, clarifying the correspondence between the two. Verify the accuracy of the correlation model, ensure that the mapping relationship conforms to the actual operating conditions, and complete the mapping between opening degree and torque.

4. The remote intelligent control method for ball valves based on the Internet of Things according to claim 1, characterized in that, The robustness weighting factor includes: Based on the observation sequence data, an observation probability model is constructed and the observation residuals corresponding to each observation data point are calculated. Based on the magnitude of the observation residuals, a residual threshold is set to distinguish between normal and abnormal observation data; Based on the correspondence between the residual threshold and the observed residual, different robust weight factors are assigned. When the residual exceeds the threshold, a smaller weight is assigned, and when the residual is within the threshold range, a larger weight is assigned. The assigned robustness weighting factor is calibrated to ensure it is suitable for the operating conditions of the ball valve.

5. The remote intelligent control method for ball valves based on the Internet of Things according to claim 1, characterized in that, The output robust observation constraint includes: Based on the robust weighting factor and the observation residual, and by constructing an observation model with robust correction, the core parameters of the robust observation constraint are determined. Based on the mechanical characteristics of the ball valve operation and the preset control accuracy requirements, the threshold range of robust observation constraints is set, the constraint boundary corresponding to the normal observation data is clarified, and the robust weight factor is embedded into the observation constraint conditions to form observation constraint rules with robust characteristics. The robust observation constraint rules are validated to ensure they are compatible with the observation probability model and robust weighting factor, and finally standardized robust observation constraints are output.

6. The method for remote intelligent control of a ball valve based on the Internet of Things according to claim 1, characterized in that, The construction of the factor graph includes: By integrating the aforementioned motion prior factors and robust observation constraints, the core input parameters of the factor graph are determined. With ball valve opening, torque, and pressure as core nodes and motion priors and robust observations as constraint edges, the basic framework of the factor graph is constructed. The robust weighting factor and observation residuals are embedded as key parameters into the corresponding nodes of the factor graph to clarify the relationship logic between each node. Verify the compatibility of nodes and constraint edges in the factor graph to ensure that motion priors and robust observation constraints can be effectively integrated into the graph structure, thus completing the factor graph construction.

7. The remote intelligent control method for ball valves based on the Internet of Things according to claim 1, characterized in that, The formation of the closed-loop control circuit includes: Based on the optimal state estimate and the preset control target, control commands adapted to the ball valve's operating conditions are generated. The control command is transmitted to the ball valve actuator, allowing the ball valve actuator to perform the corresponding opening adjustment and load control operations. Real-time acquisition of feedback status data from the ball valve actuator; accurate acquisition of actual operating parameters after ball valve execution; comparison of the feedback status data with a preset control target; calculation of the deviation between the two. The control command is continuously corrected based on the deviation value, and the above control command process is repeated to form a complete closed-loop control circuit.

8. A remote intelligent control system for ball valves based on the Internet of Things, characterized in that, include: Multi-source sensor acquisition module: Deployed in the ball valve body and corresponding locations in the pipeline network, it includes multi-source sensors and IoT terminals to collect ball valve opening degree, torque, and pressure as operating data, and outputs standardized observation sequence data after preprocessing; Model building and factor extraction module: Receives observation sequence data, establishes data mapping relationships, constructs a theoretical torque model, extracts motion prior factors, and generates robust observation constraints; Robust optimization module: Constructs an observation probability model based on observation sequence data, calculates observation residuals and generates robust weighting factors, constructs an observation model with robust correction, and outputs robust observation constraints; Factor graph optimization module: Receives motion prior factors and robust observation constraints, constructs a factor graph, embeds robust weights into the observation factors, and outputs the optimal state estimate through iterative solution; Closed-loop control module: Receives the optimal state estimate, generates control commands and transmits them to the actuator, and simultaneously collects execution feedback status, compares it with the preset target, and continuously corrects the control commands; IoT communication module: Responsible for the two-way transmission of data transmission, control commands, and feedback information between modules, ensuring the coordinated operation of the entire system.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing a computer program that, when executed by the processor, implements the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 7.

Images