Industrial steam pipe network control method and system based on multi-source data fusion and adaptive learning

Abstract

The invention discloses an industrial steam pipe network control method and system based on multi-source data fusion and adaptive learning, and the method comprises the steps: obtaining multi-source real-time operation data of an industrial steam pipe network system, and carrying out the calculation based on a digital twin model to obtain a state variable calculation value of an industrial steam pipe network; comparing the state variable calculated value with a state variable measured value in real-time operation data to generate a residual error, taking minimization of the residual error as an optimization target, and automatically inverting and updating one or more time-varying physical parameters in the digital twinborn model on line through an optimization algorithm, so as to obtain a digital twinborn model. And performing system simulation, state monitoring or optimization control by using the updated digital twin model. The simulation error is controlled within an extremely low range allowed by engineering for a long time; a self-adaptive learning mechanism enables the model to be quickly self-adjusted and keep effective when facing working conditions such as pipe network topology change and equipment performance degradation, so that the adaptability of the digital twin system to a complex dynamic environment is greatly enhanced.

Description

Technical Field

[0001] This invention relates to an industrial steam pipeline network control method and system based on multi-source data fusion and adaptive learning, belonging to the interdisciplinary field of industrial Internet of Things and smart energy management. Background Technology

[0002] The closest existing technology is a digital twin technology for steam pipeline networks based on physical mechanisms and SCADA data. Its main technical features include:

[0003] Modeling method: A mechanistic model of the pipe network is established based on fluid mechanics and the first law of thermodynamics (conservation of mass and energy) (e.g., discretization using the finite element method or finite volume method), forming a system of equations. Model parameters (such as pipe friction coefficient and heat dissipation coefficient) are usually initially set based on design drawings and empirical values.

[0004] Data-driven: Using real-time data (such as pressure, temperature, and flow rate) collected by the SCADA system as boundary conditions for model calculations or for simple data verification.

[0005] Simulation function: Under given operating conditions (such as heat source outlet pressure and flow demand of each user), solve the model equations and calculate the state variables such as pressure, temperature and flow rate of each node in the pipeline network.

[0006] Application level: It is mainly used for "visual management" and "offline simulation", that is, to display the current status on the UI interface, or to perform scenario analysis after engineers manually change the parameters.

[0007] Insufficient model accuracy and reliability: The initial parameters of the mechanistic model (such as the actual roughness of the pipe and the thermal insulation performance) deviate from the physical entity, and these parameters change over time as the pipe ages, scales, and the performance of the thermal insulation material deteriorates. Traditional methods lack a mechanism for online automatic correction of these key parameters, leading to an accumulation of simulation errors and a loss of credibility after long-term model operation.

[0008] The data-model fusion is superficial: it only uses real-time data as input and does not deeply utilize it for model calibration and optimization. Data and model are separate entities, failing to form a closed-loop feedback loop.

[0009] Passive simulation rather than active optimization: The system can only reflect the current state or perform post-event analysis. It cannot make high-precision predictions of future operating conditions, nor does it have the ability to automatically generate optimal control commands based on model prediction results and send them to actuators (such as regulating valves and pumps). Its level of intelligence is limited.

[0010] Weak ability to handle complex operating conditions: For complex dynamic operating conditions such as changes in pipeline topology (e.g., branch line maintenance and closure) and multi-heat source network operation, traditional static models require manual intervention and adjustment, resulting in a delayed response and inability to achieve adaptive simulation. Summary of the Invention

[0011] This invention provides an industrial steam pipeline network control method and system based on multi-source data fusion and adaptive learning, which solves the problems disclosed in the background art.

[0012] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0013] A control method for industrial steam pipeline networks based on multi-source data fusion and adaptive learning:

[0014] Acquire real-time multi-source operational data of industrial steam pipeline networks;

[0015] The real-time operation of a pre-constructed digital twin model inputting multi-source data;

[0016] The digital twin model calculates the state variable values ​​of the real-time running multi-source data, and compares the calculated state variable values ​​with the measured state variable values ​​in the real-time running multi-source data to generate residuals;

[0017] With minimizing the residuals as the optimization objective, the parameters in the digital twin model are automatically inverted online and calibrated.

[0018] Update the calibrated parameters to the digital twin model;

[0019] The optimal output combination of each heat source in the industrial steam network is calculated based on the calibrated digital twin model under a given total load.

[0020] The industrial steam pipeline network is controlled based on the calculated optimal output combination.

[0021] Furthermore, the method for automatically inverting and calibrating the parameters in the digital twin model online, with minimizing the residual as the optimization objective, is as follows:

[0022]

[0023] in, These are the decision variables, i.e., the set of model parameters that need to be inverted;

[0024] The objective function is the weighted sum of squares of the residuals at all measurement points; the goal is to find a set of... Make The value is the smallest;

[0025] The total number of all sensor measuring points in the pipeline network;

[0026] : Parameter After substituting into the digital twin model, the calculated first... The pressure, temperature, and flow rate values ​​at each measuring point; these values ​​are The function;

[0027] : No. Measured values ​​of pressure, temperature, and flow rate at each measuring point.

[0028] Weighting coefficients are used to balance the different dimensions and orders of magnitude of pressure, temperature, and flow rate.

[0029] By gradually adjusting the parameters using gradient descent, the minimum value is eventually obtained, and the parameters that yield the minimum value are retained.

[0030] Furthermore, methods for updating the calibrated parameters to the digital twin model include:

[0031] In the database or configuration file of the digital twin model, the original parameter values ​​in the model are... Replace directly with the newly calculated optimal value ;

[0032] The latest real-time operational multi-source data of industrial steam pipeline networks is used as the initial state and boundary conditions for the new, updated parameter model.

[0033] Start or continue running this program with the newly injected parameters A digital twin model.

[0034] Furthermore, methods for updating the calibrated parameters to the digital twin model also include:

[0035] Continuously compare the new output data of the model with the real-time operating data of the industrial steam pipeline network;

[0036] Feedback: Continuously monitor residuals If the residual increases again and exceeds a certain threshold, a new round of parameter inversion calibration will be automatically triggered to find new optimal parameters. .

[0037] Furthermore, methods for calculating the optimal output combination of each heat source in an industrial steam network under a given total load, based on a calibrated digital twin model, include:

[0038] An optimization problem is constructed with the objective function of minimizing the total steam supply cost of the industrial steam pipeline network and optimizing the economic output of heat sources in multiple different geographical locations.

[0039] The calibrated digital twin model is used as a prediction model and embedded in model predictive control.

[0040] The model predictive controller performs rolling optimization of the operating variables in a future time domain and solves for the optimal control sequence.

[0041] Furthermore, the objective function is the total cost. Minimize:

[0042]

[0043] in, For heat source collection, Number the heat source. heat source The heat supply, This is the fuel cost coefficient. This is the operation and maintenance cost coefficient. This refers to the opportunity cost or power loss cost on the power generation side caused by steam extraction / heating. heat source Start-up and shutdown costs heat source Start-stop state variables, This is a subset of heat sources that can be started and stopped; under the conditions of satisfying the heat balance and output constraints of the industrial steam pipeline network, with... With the goal of minimizing the output, the economically optimal combination of multiple heat sources is obtained by solving the problem.

[0044] Furthermore, the constraints include:

[0045] Heat balance: , This represents the total heat load of the industrial steam pipeline network.

[0046] Upper and lower limits of heat source output: , It is in running state.

[0047] Furthermore, methods for obtaining real-time multi-source operational data of industrial steam pipeline networks include:

[0048] Regularly scan the outer surface of the pipeline network to obtain the temperature field distribution, which is used to quantitatively assess the condition of the insulation layer and heat loss;

[0049] Ambient temperature, wind speed, and solar radiation intensity are obtained to calculate the dynamic heat loss of the pipeline.

[0050] A second aspect of the present invention provides an industrial steam pipeline network control system based on multi-source data fusion and adaptive learning, comprising:

[0051] The multi-source data acquisition module is used to acquire real-time multi-source operational data of the industrial steam pipeline network;

[0052] A digital twin model calculates the state variable values ​​of real-time running multi-source data, compares the calculated state variable values ​​with the measured state variable values ​​in the real-time running multi-source data, and generates residuals.

[0053] With minimizing the residuals as the optimization objective, the parameters in the digital twin model are automatically inverted online and calibrated.

[0054] Update the calibrated parameters to the digital twin model;

[0055] The optimal output combination of each heat source in the industrial steam network is calculated based on the calibrated digital twin model under a given total load.

[0056] The control module is used to control the industrial steam pipeline network based on the calculated optimal output combination.

[0057] Furthermore, it also includes:

[0058] The digital twin-based diagnostic and decision support module is used to generate energy efficiency reports regularly, highlighting pipe sections with high heat loss and recommending priority for insulation upgrades. When the deviation between the simulated value and the measured value exceeds the threshold, an alarm is automatically triggered, and the knowledge base is used to help locate the fault point.

[0059] The beneficial effects achieved by this invention are as follows:

[0060] Significantly improves model accuracy and reliability: By introducing multi-source data (especially non-traditional data such as infrared and acoustic data) and an online automatic parameter calibration mechanism, the digital twin can dynamically track the aging and changes of the physical entity, keeping simulation errors within an extremely low range permissible by engineering standards (e.g., <2%) over the long term. This is a direct result of the "parameter estimator" and "multi-source data fusion" in the technical solution.

[0061] Enhanced system resilience and adaptability: The adaptive learning mechanism enables the model to quickly self-adjust and remain effective when faced with operating conditions such as changes in pipeline topology and equipment performance degradation, greatly enhancing the adaptability of the digital twin system to complex dynamic environments.

[0062] Uncovering deeper value and expanding application boundaries: Based on a more reliable model, new advanced applications such as accurate carbon accounting and predictive maintenance are derived, providing enterprises with digital value that goes beyond the traditional scope of energy conservation. This is a derivative benefit directly generated by the high precision and intelligent features of this solution. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0064] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0065] Example 1

[0066] This embodiment provides an industrial steam pipeline network control method based on multi-source data fusion and adaptive learning, which includes the following parts:

[0067] Temporal alignment and fusion processing of multi-source heterogeneous data:

[0068] This involves an algorithm to align and fuse SCADA data, acoustic data, and infrared thermal imaging data with different sampling frequencies and transmission delays along a unified time axis. This is not simply a data stacking process, but a prerequisite for providing high-quality, highly correlated input data for subsequent parameter estimation.

[0069] Online automatic inversion and calibration of key model parameters based on real-time data:

[0070] Identify which model parameters (such as pipe equivalent roughness ε and comprehensive heat dissipation coefficient K) are critical, time-varying, and can be inverted from existing data. Optimize the model construction by transforming the objective of "minimizing the residual between calculated and measured values" into a mathematical optimization problem solvable in real-time / near real-time. Select and adapt algorithms for this optimization problem (which may be non-convex or nonlinear), employing or improving specific optimization algorithms (such as gradient descent, genetic algorithms, particle swarm optimization, etc.) to ensure computation can be completed within an engineering-acceptable timeframe, meeting the requirements of online calibration.

[0071] Embed the corrected high-precision digital twin model into the Model Predictive Control (MPC) framework:

[0072] How to use a dynamically updated digital twin model as a predictive model for MPC, and design an efficient rolling optimization strategy to transform simulation results into the optimal control command sequence for actuators (regulating valves, pumps) to achieve closed-loop control? This involves techniques such as system identification and control model simplification to ensure real-time performance.

[0073] Construction of a Fault Diagnosis and Localization Knowledge Base Based on Digital Twin Deviation Analysis:

[0074] Establish a rule base or machine learning model to map the relationship between "specific deviation patterns" (such as abnormal pressure drop or abnormal temperature distribution in a branch) and "specific fault types" (such as blocked steam traps, damaged insulation layers, or leaks).

[0075] Example 2

[0076] This embodiment proposes a more specific industrial steam pipeline network control method based on multi-source data fusion and adaptive learning, building upon Embodiment 1. The method includes the following steps:

[0077] Step 1: Construct a deeply integrated multi-source data sensing system;

[0078] In addition to traditional SCADA data (pressure P, temperature T, flow rate F), the following deployment is added:

[0079] Unmanned automatic inspection infrared thermal imager: periodically scans the outer surface of the pipeline network to obtain the temperature field distribution, which is used to quantitatively assess the condition of the insulation layer and heat loss.

[0080] Weather station: Acquires ambient temperature, wind speed, and solar radiation intensity to accurately calculate the dynamic heat loss of pipelines.

[0081] All data is preprocessed by the edge gateway, including time alignment, filtering, and noise reduction, before being uploaded to the platform system.

[0082] Step 2: Establish an adaptive mechanism model for parameter self-calibration, i.e., a digital twin model;

[0083] Core innovation: Based on the traditional mechanistic model, a "parameter estimator" module is introduced.

[0084] Operating steps:

[0085] 2-1. Input the real-time collected multi-source data ({P, T, F} measured values) into the model.

[0086] 2-2. Compare the calculated state variables ({P, T, F} calculated values) with the measured values ​​to generate a residual sequence.

[0087] 2-3. For optimization problems with the goal of minimizing residuals, use optimization algorithms such as gradient descent or genetic algorithms to automatically invert and calibrate key time-varying parameters in the model online (such as pipe equivalent roughness ε and comprehensive heat dissipation coefficient K).

[0088]

[0089] in These are the decision variables, i.e., the set of model parameters that need to be inverted. In this invention, Main representatives:

[0090] (epsilon): Equivalent roughness of the pipe;

[0091] Overall heat dissipation coefficient of the pipeline.

[0092] The objective function is the weighted sum of squares of the residuals at all measurement points. The goal is to find a set of... Make The value is the smallest.

[0093] The total number of sensor measuring points in the pipeline network.

[0094] : Parameter After substituting into the digital twin mechanism model, the calculated first... The pressure, temperature, and flow rate values ​​at each measuring point. These values ​​are... The function.

[0095] : No. Measured values ​​of pressure, temperature, and flow rate at each measuring point.

[0096] Weighting coefficients. Used to balance the different dimensions and orders of magnitude of pressure, temperature, and flow rate, as well as the different accuracy requirements for them. For example, if pressure data is more reliable and more important to the system, then... It can be set to a larger size.

[0097] Gradient descent is an iterative algorithm based on the first derivative. It finds the minimum value by gradually adjusting the parameters "along the opposite direction of the gradient (the steepest descent direction)".

[0098] 1. The parameters are updated to the model after the standardization, realizing the "synchronization" between the model and the physical entity of the industrial steam pipeline network, enabling the model to have adaptive learning ability and maintain high accuracy over a long period of time.

[0099] In step 2-3, the time-varying physical parameters are retrieved by minimizing the residuals. Specifically, this mainly involves:

[0100] Pipe equivalent roughness (ε): a key parameter affecting hydraulic friction loss in pipelines.

[0101] Pipeline overall heat dissipation coefficient (K): a key parameter affecting pipeline heat loss.

[0102] In practical applications, parameters such as pump efficiency degradation coefficient and valve flow coefficient may also be included. These parameters together constitute a parameter set. (i.e., the optimal parameter set) .

[0103] Example:

[0104] (1) Model update - parameter replacement

[0105] Operation: In the database or configuration file of the digital twin model, change the original parameter values ​​in the model. Replace directly with the newly calculated optimal value .

[0106] Example: In the model calculation section pressure drop hour,

[0107] formula In the formula Updated from the old 0.0002 to the calibrated 0.00025.

[0108] In the calculation section temperature drop hour,

[0109] formula In the formula The old value of 1.5 W / m²·K has been updated to the calibrated value of 1.8 W / m²·K.

[0110] (2) Model initialization

[0111] Operation: Use the latest physical entity measured data (pressure, temperature, and flow rate at the heat source outlet, and pressure, temperature, and flow rate at each user terminal) as the initial state and boundary conditions for the new, updated parameter model.

[0112] Objective: To ensure that the corrected model starts running from a starting point that is completely synchronized with the physical world, avoiding transient errors introduced by inconsistent initial states.

[0113] (3) Perform a new round of simulation calculations

[0114] Action: Start or continue running this "new parameters injected" program. "Digital twin model".

[0115] Process: The model solver will use the new parameters. "Based on the current boundary conditions, we recalculate the new state variables such as pressure, temperature, and flow rate for each node in the entire pipeline network."

[0116] Results: Model's computational output It is derived based on parameters that are closer to the real world, and theoretically should be consistent with the physical entity of industrial steam pipeline networks. The measured values ​​are highly consistent. At this point, the model has completed a "synchronization" with the physical entity.

[0117] (4) Verification and closed-loop (continuous process)

[0118] Operation: The system continuously compares the new outputs of the model. With real-time data collection .

[0119] Feedback: Residual It will be continuously monitored. If the residual increases again and exceeds a certain threshold, a new round of parameter inversion calibration process will be automatically triggered (i.e., returning to the optimization algorithm step) to find new optimal parameters. Then repeat the update steps described above.

[0120] Step 3: Prediction and Active Optimization Based on the Corrected Model

[0121] Short-term load forecasting: Using historical data and LSTM neural networks, predict the steam demand load of each user for the next day.

[0122] Model prediction:

[0123] An optimization problem is constructed with the objective function of minimizing total steam supply cost and optimizing the economic output of heat sources in multiple different geographical locations.

[0124] The corrected high-precision digital twin model is used as a prediction model and embedded in model predictive control.

[0125] The model predictive controller performs rolling optimization of the operating variables (such as valve opening) in the next time domain and solves for the optimal control sequence.

[0126] Considering the operational constraints of each heat source in the industrial steam pipeline network, a mixed integer nonlinear programming (MINLP) model is established.

[0127] The objective function is the total cost. Minimize:

[0128]

[0129] in, For heat source collection, Number the heat source. heat source The heat supply, This is the fuel cost coefficient. This is the operation and maintenance cost coefficient. This refers to the opportunity cost or power loss cost on the power generation side caused by steam extraction / heating. heat source Start-up and shutdown costs heat source The start / stop state variables (1 / 0). This is a subset of heat sources that can be started and stopped. The model, under the conditions of satisfying system heat balance and output constraints, uses... With the goal of minimizing output, the economically optimal combination of multiple heat sources can be obtained by using an improved genetic algorithm.

[0130] The constraints include:

[0131] Heat balance: , This represents the total heat load of the industrial steam pipeline network.

[0132] Upper and lower limits of heat source output: , The running status is (0 / 1).

[0133] Climbing rate constraints, etc.

[0134] The MINLP model is solved using an improved genetic algorithm (GA), which can effectively handle nonlinearity and integer variables and obtain the optimal output combination of each heat source under a given total load.

[0135] Step 4: Digital Twin-Based Diagnosis and Decision Support

[0136] The system regularly generates energy efficiency reports, highlighting pipe sections with high heat loss and recommending them for priority insulation upgrades.

[0137] When the deviation between the simulated value and the measured value exceeds the threshold, an alarm is automatically triggered, and the knowledge base is used to help locate the fault point (such as "the XX steam trap may be malfunctioning").

[0138] Example 3

[0139] This embodiment proposes a more specific industrial steam pipeline control system based on multi-source data fusion and adaptive learning, building upon Embodiment 2. The system includes:

[0140] The multi-source data acquisition module is used to acquire real-time multi-source operational data of the industrial steam pipeline network;

[0141] A digital twin model calculates the state variable values ​​of real-time running multi-source data, compares the calculated state variable values ​​with the measured state variable values ​​in the real-time running multi-source data, and generates residuals.

[0142] With minimizing the residuals as the optimization objective, the parameters in the digital twin model are automatically inverted online and calibrated.

[0143] Update the calibrated parameters to the digital twin model;

[0144] The optimal output combination of each heat source in the industrial steam network is calculated based on the calibrated digital twin model under a given total load.

[0145] The control module is used to control the industrial steam pipeline network based on the calculated optimal output combination.

[0146] The digital twin-based diagnostic and decision support module is used to generate energy efficiency reports regularly, highlighting pipe sections with high heat loss and recommending priority for insulation upgrades. When the deviation between the simulated value and the measured value exceeds the threshold, an alarm is automatically triggered, and the knowledge base is used to help locate the fault point.

[0147] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

[0148] A computer-readable storage medium storing one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform an industrial steam pipeline control method based on multi-source data fusion and adaptive learning.

[0149] A computing device includes one or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for executing an industrial steam pipeline control method based on multi-source data fusion and adaptive learning.

[0150] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0151] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0152] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0153] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0154] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. An industrial steam pipeline network control method based on multi-source data fusion and adaptive learning, characterized in that: Acquire real-time multi-source operational data of industrial steam pipeline networks; The real-time operation of a pre-constructed digital twin model inputting multi-source data; The digital twin model calculates the state variable values ​​of the real-time running multi-source data, and compares the calculated state variable values ​​with the measured state variable values ​​in the real-time running multi-source data to generate residuals; With minimizing the residuals as the optimization objective, the parameters in the digital twin model are automatically inverted online and calibrated. Update the calibrated parameters to the digital twin model; The optimal output combination of each heat source in the industrial steam network is calculated based on the calibrated digital twin model under a given total load. The industrial steam pipeline network is controlled based on the calculated optimal output combination.

2. The industrial steam pipeline control method based on multi-source data fusion and adaptive learning according to claim 1, characterized in that: The method for automatically inverting and calibrating the parameters in the digital twin model online, with the goal of minimizing the residuals, is as follows: ； in, These are the decision variables, i.e., the set of model parameters that need to be inverted; The objective function is the weighted sum of squares of the residuals at all measurement points; the goal is to find a set of... Make The value is the smallest; The total number of all sensor measuring points in the pipeline network; : Parameter After substituting into the digital twin model, the calculated first... The pressure, temperature, and flow rate values ​​at each measuring point; these values ​​are The function; : No. Measured values ​​of pressure, temperature, and flow rate at each measuring point; Weighting coefficients are used to balance the different dimensions and orders of magnitude of pressure, temperature, and flow rate. By gradually adjusting the parameters using gradient descent, the minimum value is eventually obtained, and the parameters that yield the minimum value are retained.

3. The industrial steam pipeline control method based on multi-source data fusion and adaptive learning according to claim 1, characterized in that: Methods for updating the calibrated parameters to the digital twin model include: In the database or configuration file of the digital twin model, the original parameter values ​​in the model are... Replace directly with the newly calculated optimal value ; The latest real-time operational multi-source data of industrial steam pipeline networks is used as the initial state and boundary conditions for the new, updated parameter model. Start or continue running this program with the newly injected parameters A digital twin model.

4. The industrial steam pipeline control method based on multi-source data fusion and adaptive learning according to claim 3, characterized in that: Methods for updating the calibrated parameters to the digital twin model also include: Continuously compare the new output data of the model with the real-time operating data of the industrial steam pipeline network; Feedback: Continuously monitor residuals If the residual increases again and exceeds a certain threshold, a new round of parameter inversion calibration will be automatically triggered to find new optimal parameters. .

5. The industrial steam pipeline control method based on multi-source data fusion and adaptive learning according to claim 1, characterized in that: Methods for calculating the optimal output combination of heat sources in an industrial steam network under a given total load based on a calibrated digital twin model include: An optimization problem is constructed with the objective function of minimizing the total steam supply cost of the industrial steam pipeline network and optimizing the economic output of heat sources in multiple different geographical locations. The calibrated digital twin model is used as a prediction model and embedded in model predictive control. The model predictive controller performs rolling optimization of the operating variables in a future time domain and solves for the optimal control sequence.

6. The industrial steam pipeline control method based on multi-source data fusion and adaptive learning according to claim 5, characterized in that: The objective function is the total cost. Minimize: ； in, For heat source collection, Number the heat source. heat source The heat supply, This is the fuel cost coefficient. This is the operation and maintenance cost coefficient. This refers to the opportunity cost or power loss cost on the power generation side caused by steam extraction / heating. heat source Start-up and shutdown costs heat source Start-stop state variables, This is a subset of heat sources that can be started and stopped; under the conditions of satisfying the heat balance and output constraints of the industrial steam pipeline network, with... With the goal of minimizing the output, the economically optimal combination of multiple heat sources is obtained by solving the problem.

7. The industrial steam pipeline control method based on multi-source data fusion and adaptive learning according to claim 6, characterized in that: The constraints include: Heat balance: , This represents the total heat load of the industrial steam pipeline network. Upper and lower limits of heat source output: , It is in running state.

8. The industrial steam pipeline control method based on multi-source data fusion and adaptive learning according to claim 1, characterized in that: Methods for obtaining real-time multi-source operational data of industrial steam pipeline networks include: Regularly scan the outer surface of the pipeline network to obtain the temperature field distribution, which is used to quantitatively assess the condition of the insulation layer and heat loss; Ambient temperature, wind speed, and solar radiation intensity are obtained to calculate the dynamic heat loss of the pipeline.

9. An industrial steam pipeline control system based on multi-source data fusion and adaptive learning, characterized in that, include: The multi-source data acquisition module is used to acquire real-time multi-source operational data of the industrial steam pipeline network; A digital twin model calculates the state variable values ​​of real-time running multi-source data, compares the calculated state variable values ​​with the measured state variable values ​​in the real-time running multi-source data, and generates residuals. With minimizing the residuals as the optimization objective, the parameters in the digital twin model are automatically inverted online and calibrated. Update the calibrated parameters to the digital twin model; The optimal output combination of each heat source in the industrial steam network is calculated based on the calibrated digital twin model under a given total load. The control module is used to control the industrial steam pipeline network based on the calculated optimal output combination.

10. The industrial steam pipeline control system based on multi-source data fusion and adaptive learning according to claim 9, characterized in that, Also includes: A digital twin-based diagnostic and decision support module is used to generate energy efficiency reports regularly, highlighting pipe sections with high heat loss and recommending priority for insulation retrofitting. When the deviation between the simulated value and the measured value exceeds the threshold, an alarm is automatically triggered, and the knowledge base is used to help locate the fault point.

Images