A coal gas desulfurization parameter optimization regulation method and system based on causal identification and counterfactual deduction

Abstract

The application discloses a coal gas desulfurization parameter optimization regulation method and system based on causal identification and counterfactual deduction. It relates to the technical field of industrial process control. The method obtains desulfurization system process parameters, physical boundary limits and over-standard threshold values; a target causal diagram model is constructed based on a process flow chain; abnormality is identified, and the causal effect is calculated by reverse tracing the parent node in the causal diagram to lock the root cause variable; the adjustable parameters in the root cause variable are taken as intervention variables to perform counterfactual deduction, and the expected sulfur content is iteratively predicted under the physical boundary limits; based on process joint regulation constraints, the intervention strategy with the minimum adjustment cost and the expected sulfur content meeting the standard is selected, industrial physical adjustment instructions are generated and executed. The application realizes interpretable precise root cause tracing, reduces the on-site trial and error cost through pre-sand table deduction, and guarantees the physical safety and economy of regulation.

Description

Technical Field

[0001] This invention relates to the field of industrial process control technology, and in particular to a method and system for optimizing and controlling coal gas desulfurization parameters based on causal identification and counterfactual deduction. Background Technology

[0002] In iron and steel metallurgy and chemical production, coke oven gas desulfurization is a crucial environmental and technological step. Model- and data-based optimization methods have become the main means to improve the efficiency of desulfurization processes. However, existing process simulation software (such as Aspen Plus) and traditional regression models and machine learning methods are still insufficient in simulating dynamic disturbances and complex operating conditions in industrial plants, and cannot fully reflect the frequent operational adjustments and changes in conditions in actual production.

[0003] Existing technologies for addressing excessive sulfur content in desulfurization systems suffer from the following main drawbacks: First, multivariate nonlinear coupling makes root cause identification difficult. Numerous parameters in desulfurization systems are coupled, and traditional correlation analysis is susceptible to spurious correlations and confounding factors, making it difficult to distinguish the true causal drivers. Second, reliance on manual trial and error is inefficient and costly. Traditional control methods rely on repeated manual trials, resulting in slow response times and an inability to accurately predict outcomes before adjustments. Third, the evaluation of parameter adjustment effects is delayed. In actual production, the effectiveness of adjustments can only be evaluated after the system has physically stabilized, exhibiting significant lag and making it difficult to support dynamic collaborative optimization and precise control of the desulfurization system under full-cycle, variable-load operation. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a method and system for optimizing and controlling coal gas desulfurization parameters based on causal identification and counterfactual deduction.

[0005] To solve the above technical problems, the technical solution of the present invention is as follows: In a first aspect, the present invention provides a method for optimizing and controlling coal gas desulfurization parameters based on causal identification and counterfactual deduction, the method specifically including the following steps: Obtain multi-dimensional process parameter data of the coal gas desulfurization system; Based on the coal gas desulfurization production process chain and multi-dimensional process parameter data, a causal graph model is constructed to characterize the causal dependencies between desulfurization process parameters. Anomaly identification is performed based on multi-dimensional process parameter data to obtain an abnormal parameter set. The parent node of the abnormal parameter set is traced back in the causal graph model to calculate the causal effect and determine the root variable that causes the abnormal sulfur content of coal gas. Using the adjustable parameters in the root cause parameters as intervention variables, counterfactual inference is performed on the causal graph model, and the expected sulfur content of coal gas after adjusting the intervention variables is iteratively predicted under the condition that the physical boundary limits of the adjustable parameters are met. Based on the preset process joint control constraints, the intervention strategy that minimizes the expected sulfur content of coal gas and the excessive sulfur content threshold is selected, and the industrial physical control command is generated and sent to the field process equipment to perform physical control.

[0006] The aforementioned technical solution integrates the desulfurization process mechanism with multi-dimensional process parameter data to construct a causal graph model characterizing the true causal dependencies between desulfurization process parameters. Based on this model, it performs reverse tracing of the real-time identified abnormal parameter sets. This mechanism fundamentally overcomes the shortcomings of traditional data-driven models, which are easily misled by false data associations under strongly coupled multivariate conditions. Through the mechanistic constraints and reverse tracing of the causal graph, the system can accurately isolate interference caused by covariance of process parameters and precisely pinpoint the true physical source of abnormal sulfur content in coal gas, thereby significantly improving the accuracy and technical reliability of anomaly diagnosis in complex industrial systems.

[0007] Secondly, the present invention also provides a gas desulfurization parameter optimization and control system based on causal identification and counterfactual deduction, comprising: The data acquisition module is used to acquire multi-dimensional process parameter data of the gas desulfurization system; The causal graph construction module is used to construct a causal graph model to characterize the causal dependencies between desulfurization process parameters based on the coal gas desulfurization production process chain and multi-dimensional process parameter data. The root cause tracing module is used to identify abnormal parameters based on multi-dimensional process parameter data, obtain a set of abnormal parameters, and trace back the parent node of the abnormal parameter set in the cause-effect graph model to calculate the causal effect and determine the root cause variable that leads to the abnormal sulfur content of coal gas. The counterfactual inference module is used to perform counterfactual inference on the causal graph model with the adjustable parameters in the root cause parameters as intervention variables, and iteratively predict the expected sulfur content of coal gas after adjusting the intervention variables, under the condition that the physical boundary limits of the adjustable parameters are met. The instruction output module is used to select the intervention strategy that minimizes the expected sulfur content of coal gas and the excessive sulfur content threshold based on preset process joint control constraints, generate industrial physical control instructions, and send them to the field process equipment for physical control.

[0008] Thirdly, the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-mentioned method for optimizing and controlling coal gas desulfurization parameters based on causal identification and counterfactual deduction.

[0009] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-described method for optimizing and controlling coal gas desulfurization parameters based on causal identification and counterfactual deduction.

[0010] The beneficial effects of this invention are: (1) This invention uses the root cause adjustable parameters determined by the system as intervention variables, performs counterfactual deduction within the physical boundary limits of the parameters, and iteratively predicts the distribution of sulfur content changes under different intervention strategies in the algorithm model. This allows the control system to accurately and quantitatively predict the true impact of parameter adjustments on desulfurization indicators before applying actual physical actions to the field equipment. Compared with the traditional hysteresis-based adjustment based on sensor feedback, this scheme achieves "feedforward optimization" of the control strategy, avoids applying ineffective or reverse control disturbances to physical equipment, greatly reduces the fluctuation range of desulfurization process parameters, and improves the dynamic stability of the overall system operation.

[0011] (2) This invention uses a combination of point anomaly detection and subsequence anomaly detection to calculate the standardized anomaly comprehensive score, which can capture both sudden problems and identify potential chronic drift risks. Simultaneously, when determining intervention variables, the intersection of the root variable and the adjustable parameter set is calculated, and a collinearity test is performed based on the covariate set, eliminating variables whose collinearity test index exceeds the preset collinearity threshold. This series of data cleaning and screening mechanisms ensures that the variables input for counterfactual inference are both causally deterministic and physically adjustable, and are mutually independent, avoiding wasted computing power and inference distortion caused by multicollinearity. (3) This invention uses preset process joint control constraints to screen intervention strategies and ultimately generates industrial physical adjustment commands that are directly sent to the field process equipment for execution. Due to the strong physical and chemical coupling between various devices in the desulfurization system, the optimal parameter adjustment at a single point can easily lead to global process failures. This solution, by forcibly introducing joint control constraints at the control logic level, technically shields dangerous control commands that may cause equipment failures, effectively ensuring the physical operation safety of the desulfurization unit; at the same time, it directly converts the results of the deduction and optimization into standard underlying equipment drive messages for execution, opening up the fully automated control link and significantly improving the closed-loop control level of the industrial site. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A flowchart illustrating the method for optimizing and controlling coal gas desulfurization parameters based on causal identification and counterfactual deduction provided by this invention. Figure 2 This is a schematic diagram of the gas desulfurization parameter optimization and control device based on causal identification and counterfactual deduction provided by the present invention. Figure 3 A schematic diagram of the computer device provided by the present invention. Detailed Implementation

[0014] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0015] Example 1: This example uses the coke oven gas desulfurization system of a steel enterprise as an example to illustrate in detail the implementation process of the gas desulfurization parameter optimization and control method based on causal identification and counterfactual deduction provided by this invention. The enterprise's desulfurization system is divided into a first-phase production process and a second-phase production process. The gas sources for the two phases are different but they are carried out simultaneously, without any specific order. All processes in the first-phase production process are named after the equipment, in the following order: gas-liquid separation, primary cooling, electrostatic precipitator, blast furnace, precooling, ammonia stripping, No. 1 desulfurization, No. 2 desulfurization, No. 3 desulfurization, preheating, saturation, final cooling, and benzene washing. The second-phase production process adds a direct cooling process after the primary cooling in addition to the first-phase process. Furthermore, there is a No. 7 desulfurization tower as a backup tower for both phases, which can be switched and started / stopped at any time. According to environmental protection requirements and production needs, the total sulfur content of the desulfurization system is specified to be no more than 120 mg / m³ (i.e., the threshold for exceeding the sulfur content limit), otherwise it will affect the production of downstream users. The above-mentioned gas desulfurization parameter optimization and control method based on causal identification and counterfactual deduction specifically includes the following steps: Step S1: Obtain multi-dimensional process parameter data of the coal gas desulfurization system.

[0016] Specifically, the system collects multi-dimensional process parameter data in real time through DCS and LIMS systems, with a collection frequency on the order of minutes. The data includes, but is not limited to: gas temperature, desulfurization liquid temperature, ammonia temperature, lean oil temperature, gas flow rate, desulfurization liquid flow rate, air flow rate, rich oil flow rate, residual ammonia water flow rate, desulfurization liquid level, foam tank level, gas-liquid separator suction, resistance and pressure of each piece of equipment, voltage, and effective components detected by testing. Simultaneously, it acquires the physical boundary limits (adjustable upper and lower limits) of each adjustable parameter and the threshold for exceeding the sulfur content limit in the gas (120 mg / m³).

[0017] After acquiring the data, data checks and preprocessing are performed: the data structure integrity, format consistency and content validity are checked; invalid data (including duplicate records, redundant information and empty or meaningless values) are cleaned up, missing values ​​are filled by linear interpolation, high-frequency data is denoised, and outliers are truncated based on the 3σ principle.

[0018] The system synchronously acquires the physical boundary limits (constraint information) of adjustable parameters. The number of desulfurization process parameters for this enterprise is as follows: 32 processes before desulfurization, 57 processes after desulfurization, and 30 processes after desulfurization. Some core adjustable parameters and their adjustable ranges are shown in Table 1. Table 1. Some adjustable parameters and their adjustable ranges

[0019] Step S2: Based on the coal gas desulfurization production process chain and multi-dimensional process parameter data, construct a causal graph model to characterize the causal dependencies between desulfurization process parameters.

[0020] Specifically, a cause-effect graph is established based on the processes and procedures in the production workshop. In the formula, V is the vertex set, containing all vertices in the graph, E is the edge set, containing all directed edges, and each edge connects two vertices, forming a tuple. Indicates from vertex Pointing to the vertex A directed edge, where, The specific steps to set it up are as follows: Step S2-1: Based on the production process flow, the entire production system is divided into three core process stages, denoted as V1, V2, and V3 respectively. In the formula, V1 is the pre-desulfurization process, V2 is the desulfurization process, and V3 is the post-desulfurization process. Directed edges are defined. edge set .

[0021] Step S2-2: Based on the major process categories, refine the process nodes into specific process steps. The pre-desulfurization process includes raw material testing (V). 11 Gas-liquid separation V 12 Primary gas cooling V 13 Electrostatic capture V 14 And the blower V 15 Five processes; the desulfurization process includes gas precooling V 21 Desulfurization V 22 Chemical product testing V 23 Three processes; the post-desulfurization process includes gas preheating V 31 Deamination V 32 Final cooling V 33 and benzene removal V 34Four processes. These processes form a complete process chain according to the actual production sequence, with the output of the preceding process serving as the input of the following process. The node set is updated to... .

[0022] Step S2-3: Define directed edges between processes and add them to edge set E: In the formula, All are positive integers.

[0023] Step S2-4: Each process node is further refined into monitoring and control parameters (including but not limited to gas temperature, gas flow rate, liquid temperature, liquid flow rate, liquid level, equipment resistance, equipment pressure, voltage, and detection of effective components). Based on the process mechanism, define the following core causal relationships and set path coefficients: ① The sulfur content of coal gas is directly affected by the temperature of the coal gas before desulfurization, the temperature and flow rate of the desulfurization liquid, and the concentration of effective components, as well as indirectly affected by other processes. ② The gas temperature is affected by the combined effects of the preceding gas temperature and the current liquid flow rate and temperature through direct gas-liquid heat exchange; ③ The equipment resistance is determined by both the pressure of the preceding equipment and the pressure of the current process; ④ The sulfur content in the system follows the law of conservation of materials, and its form is transformed between processes. Ultimately, the sulfur content of the coal gas is regulated by the oxidation-reduction reaction between the raw material sulfur and chemical products.

[0024] Based on the above relationships, directed edges are added to the causal graph to form a directed acyclic graph structure.

[0025] Step S2-5: Define parameters that simultaneously affect multiple indirectly related processes as confounding variables and add them to the covariate set X. cov Furthermore, directed edges originating from sulfur content in coal gas are removed from the initial causal graph structure to block causal inversion paths.

[0026] Step S2-6: Perform conditional independence tests on the initial causal graph structure based on actual production data, remove pseudo-causal edges and add data-driven causal edges to obtain the causal graph model. .

[0027] Step S3: Based on multi-dimensional process parameter data, anomaly identification is performed to obtain an abnormal parameter set. In the causal graph model, the parent node of the abnormal parameter set is traced back to calculate the causal effect and determine the root cause variable that leads to the abnormal sulfur content of coal gas.

[0028] Specifically, firstly, a combination of point anomaly detection and subsequence anomaly detection is used to calculate a standardized anomaly comprehensive score for each process parameter. The specific calculation formula is as follows: In the formula, For point anomaly scores, Scoring for subsequence anomalies and Let be the weighting coefficient, satisfying , and They are respectively The mean and standard deviation, and They are respectively The mean and standard deviation of the parameters are calculated. Z-score standardization is used to eliminate dimensional differences between heterogeneous physical quantities. Parameters are sorted in descending order of score, and the top 5 parameters in the sorted sequence are selected to form an outlier parameter set. .

[0029] In this embodiment, some of the anomaly detection results are shown in Table 2: Table 2 Partial Results of Anomaly Detection

[0030] Next, initialize the candidate set. For each node In a causal graph Traverse its parent node If a certain parent node Also belongs to Then it is believed The abnormality may be caused by [the following], Remove from R. Use machine learning methods (such as causal forest models) and regression methods, adjusting the set of covariates X using the backdoor and frontdoor criteria. cov To eliminate confounding factors and block backdoor paths (non-causal paths), the average causal effect (including direct and indirect effects) of each candidate root cause node on the target gas sulfur content is calculated. Parameters are sorted in descending order of the absolute value of the average causal effect, and the top predetermined number of parameters in the sorted sequence are selected as the root cause variable set. .

[0031] In this embodiment, the results of root cause analysis are shown in Table 3: Table 3 Results of Root Cause Analysis

[0032] Step S4: Using the adjustable parameters in the root cause parameters as intervention variables, perform counterfactual inference on the causal graph model, and iteratively predict the expected sulfur content of coal gas after adjusting the intervention variables, provided that the physical boundary limits of the adjustable parameters are met.

[0033] Specifically, this step includes the following sub-steps: Step S4-1: Intervention variable screening: Obtain the set of adjustable parameters T of the system and calculate the set of root dependent variables. The intersection of this intersection with the adjustable parameter set T yields candidate intervention variables. This is based on the covariate set X. cov The candidate intervention variables are subjected to collinearity testing (e.g., using a LinearRegression model). Variables with collinearity test indicators (e.g., Pearson correlation coefficient absolute value greater than 0.8, or variance inflation factor (VIF) greater than 10) exceeding a preset collinearity threshold are removed, resulting in the final set of intervention variables Y. When the intersection is an empty set, a process design anomaly warning is generated indicating the absence of effective intervention variables.

[0034] Step S4-2: Two-way counterfactual inference: For the current value of each intervention variable in the intervention variable set Y Adjust the step size ratio according to the preset settings. Adjustments (e.g., 5%) are made upwards and downwards, and the calculations are performed. (In the formula, To adjust upwards, (To be adjusted downwards). The variable values ​​after intervention... Substitute the values ​​into a causal machine learning model (such as a causal forest model) that is constrained by the topology of the causal graph model to predict the corresponding expected sulfur content of the coal gas and its 95% confidence interval.

[0035] Step S4-3: Iteration Termination Condition: Continue iterative bidirectional counterfactual deduction until the expected sulfur content of the coal gas is lower than the threshold for exceeding the sulfur content limit. (120mg / m³), or The physical boundary limit is reached, or the maximum number of iterations is reached. For each intervention variable, the upward and downward adjustments are independently extrapolated, and the average effect is calculated separately. The direction that results in a lower predicted sulfur content is selected, and the complete information is stored in the candidate strategy dictionary for that variable.

[0036] In this embodiment, some results of counterfactual reasoning are shown in Table 4: Table 4 Counterfactual Reasoning Results

[0037] Step S5: Based on the preset process joint control constraints, select the intervention strategy that minimizes the expected sulfur content of coal gas below the threshold for exceeding the sulfur content limit and minimizes the adjustment cost, generate industrial physical control instructions, and send them to the field process equipment for physical control.

[0038] Specifically, this step includes the following sub-steps: Step S5-1: Joint Constraints: Identify the parameter groups that need to be jointly controlled based on the preset joint process control constraints (business rules). For adjustable parameters within the same parameter group, force them to maintain a consistent direction of change (i.e., increase or decrease simultaneously) when performing counterfactual deduction, in order to avoid violating process constraints such as the conservation of chemical materials or introducing new abnormal operating conditions.

[0039] Step S5-2: Cost optimization: reducing the total sulfur content of coal gas to the threshold exceeding the standard. The objective is to calculate the total adjustment cost (comprehensive adjustment cost) of each intervention strategy using a adjustment cost model based on machine learning or statistical methods, while satisfying the physical boundary limits of each adjustable parameter and the joint unidirectional adjustment constraint. The strategy with the minimum total adjustment cost is selected as the optimal intervention strategy.

[0040] Step S5-3: Instruction generation and issuance: Generate structured control instructions containing intervention methods, adjustment ratios, absolute adjustment amounts and their confidence intervals, encapsulate them as industrial physical adjustment instructions and issue them to the field PLC / DCS equipment for physical control.

[0041] In this embodiment, the final generated operation strategy text is shown in Table 5: Table 5 Operation Strategy Text Example 2: This example provides a gas desulfurization parameter optimization and control system based on causal identification and counterfactual inference. See [link to example]. Figure 2 The system includes: a data acquisition module, a cause-effect graph construction module, a root cause tracing module, a counterfactual deduction module, and an instruction output module.

[0042] Specifically, the data acquisition module is used to acquire multi-dimensional process parameter data of the coal gas desulfurization system. This module includes the following sub-units: Multi-source heterogeneous data acquisition unit: Real-time acquisition of multi-dimensional process parameters of Phase I and Phase II chemical production at a frequency of minutes (such as gas temperature of each tower, desulfurization liquid flow rate, equipment resistance, and effective components of the test and analysis, etc.), and simultaneous retrieval of historical time series data.

[0043] Data cleaning and preprocessing unit: performs structural integrity and format consistency checks on the collected data; performs linear interpolation to fill in missing values; filters and denoises high-frequency noise; and truncates outliers based on the 3σ principle, outputting a standardized two-dimensional process parameter list.

[0044] Boundary and threshold configuration unit: Dynamically reads environmental assessment indicators (such as the threshold of 120mg / m³ for sulfur content exceeding the standard in coal gas) from the MES system, and loads the physical boundary limits of adjustable parameters for each process (such as the lower limit of 16℃ and the upper limit of 25℃ for coal gas temperature in the primary cooler of Phase I; the lower limit of 500m³ / h and the upper limit of 1000m³ for desulfurization liquid flow rate in Phase I No. 1 desulfurization tower, etc.), forming an adjustable parameter set T in nested object format.

[0045] The causal graph construction module is used to construct a causal graph model to characterize the causal dependencies between desulfurization process parameters based on the coal gas desulfurization production process chain and multi-dimensional process parameter data. This module specifically includes the following sub-units: Process topology initialization unit: Based on the desulfurization production process flow, the system is divided into three core stages: pre-desulfurization process, desulfurization process, and post-desulfurization process, and further subdivided into 15 specific process nodes, from raw material testing to benzene removal. According to the material flow direction and heat exchange mechanism, directed edges between stages and between processes are automatically generated, constructing an initial directed acyclic graph structure.

[0046] Hybrid Identification and Mechanism Constraint Unit: Parameters that simultaneously affect multiple indirectly related processes are identified using graph algorithms, defined as hybrid variables, and added to the covariate set X. cov At the same time, hard rule constraints are enforced to forcibly remove all directed edges in the initial structure that start with "coal gas sulfur content", thus completely blocking the causal inversion path from a physical and logical perspective.

[0047] Data-driven validation unit: Loads historical production data and uses conditional independence testing algorithms (such as PC or FCI) to perform secondary validation on the initial graph structure, removing pseudo-causal edges and supplementing data-driven hidden causal edges, finally outputting a high-confidence target causal graph model. .

[0048] The root cause analysis module is used to identify anomalies based on multi-dimensional process parameter data, obtain a set of anomalous parameters, and then trace back the parent nodes of the anomalous parameter set in a causal graph model to calculate the causal effect and determine the root cause variable leading to the abnormal sulfur content in the coal gas. This module specifically includes the following sub-units: Dual anomaly detection units: a parallel point anomaly detection algorithm (capturing instantaneous limit violations) and a subsequence anomaly detection algorithm (capturing trend drift and morphological distortion within the historical half-hour). After extracting time-series features, Z-score normalization is used to eliminate the dimensional barriers of heterogeneous physical quantities, and then the formula is applied. Perform weighted fusion and output a set of anomaly parameters arranged in descending order of standardized anomaly comprehensive scores.

[0049] Graph Traversal and Effect Calculation Unit: In the Cause-and-Effect Graph Model In China, Recursively trace upwards through the parent node. Apply the Causal Forest algorithm, strictly using the backdoor and frontdoor criteria to adjust X. cov The confounding factor is used to block non-causal paths, and the average causal effect (including direct and indirect effects) of each candidate parent node on the sulfur content of the target coal gas is calculated.

[0050] Root Cause Locating and Ranking Unit: Sort the variables in descending order of the absolute value of the average causal effect, extract the top 5 parameters from the ranking sequence, and output a set of root cause variables with clear physical orientation. .

[0051] The counterfactual inference module is used to perform counterfactual inference on a causal graphical model using adjustable parameters from the root cause parameters as intervention variables. Under the condition that the physical boundary limits of the adjustable parameters are met, iteratively predicts the expected sulfur content of the coal gas after adjusting the intervention variables. This module specifically includes the following sub-units: Intervention variable selection and collinearity circuit breaker unit: calculation The intersection of this intersection with the adjustable parameter set T yields candidate intervention variables. Based on X cov Perform collinearity tests on candidate variables (e.g., calculate the variance inflation factor (VIF) or Pearson correlation coefficient), and eliminate variables whose collinearity test indicators exceed a preset threshold. If the intersection is determined to be an empty set, immediately trigger the circuit breaker mechanism, generate an abnormal prompt "No effective intervention variables, process design needs to be checked," and stop the simulation.

[0052] Topologically constrained bidirectional inference unit: using the current abnormal operating condition as the factual benchmark, it applies the selected intervention variables according to a preset step size ratio. Perform bidirectional iterative calculations for both up and down adjustments. .Will Substitute the values ​​into a causal machine learning model constrained by the topological structure of the target causal graph to predict the expected sulfur content of the coal gas and its 95% confidence interval. The iterative process is strictly limited by physical boundary values ​​until the predicted value is below 120 mg / m³ or the maximum number of iterations is reached. .

[0053] Joint Intervention Verification Unit: This unit incorporates a process business rule engine to identify parameter groups requiring joint control (such as gas flow rate and desulfurization liquid flow rate). During simulations, it forces adjustable parameters within the same parameter group to maintain consistent change directions (both increasing or decreasing) to prevent counterfactual interventions from violating the conservation of chemical materials or introducing new system oscillations.

[0054] The instruction output module is used to select the intervention strategy that minimizes the expected sulfur content of the coal gas and reduces the regulation cost, based on preset process joint control constraints. It then generates industrial physical regulation instructions and sends them to the field process equipment for execution. This module specifically includes the following sub-units: Cost assessment and optimization unit: Invokes a preset adjustment cost model to comprehensively calculate the weighted sum of the energy consumption increment and material consumption increment caused by each target-achieving intervention strategy (i.e., the total adjustment cost). Under the premise of satisfying physical boundaries and joint unidirectional constraints, the strategy with the minimum total adjustment cost is selected as the optimal intervention strategy.

[0055] Structured instruction encapsulation unit: Transforms the optimal intervention strategy into structured control instruction text that is both human-readable and machine-parseable. For example, it generates: "Phase II gas-liquid separation gas temperature decreases by 5.0% (approximately 4.242 units) → Expected total sulfur content -9.765 (95% CI: [-15.681, -3.849])".

[0056] The secure command delivery and closed-loop unit verifies the packaged instructions (e.g., rate of change limits, extreme value protection), and after confirming safety, delivers them to the field PLC / DCS via the industrial gateway for physical control. Simultaneously, it records the actual sulfur content change after execution and feeds it back to the data acquisition module for online incremental learning and closed-loop correction of the causal graph model and the inference model.

[0057] Example 3: See Figure 3 This embodiment provides a computer device, the components of which may include, but are not limited to: one or more processors or processing units, system memory, and buses connecting different system components (including system memory and processing units).

[0058] A bus refers to one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. Examples of these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0059] Computer systems / servers typically include a variety of computer system-readable media. These media can be any available media that can be accessed by the computer system / server, including volatile and non-volatile media, and removable and non-removable media.

[0060] System memory may include computer system readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. The computer device may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system may be used to read and write non-removable, non-volatile magnetic media. Disk drives for reading and writing to removable non-volatile disks (e.g., "floppy disks") and optical disc drives for reading and writing to removable non-volatile optical discs (e.g., CD-ROMs, DVD-ROMs, or other optical media) may be provided. In these cases, each drive may be connected to a bus via one or more data media interfaces. The memory may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0061] A program / utility having a set (at least one) of program modules can be stored, for example, in memory. Such program modules include—but are not limited to—an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. The program modules typically perform the functions and / or methods described in the embodiments of this invention.

[0062] Computer devices can also communicate with one or more external devices (such as keyboards, pointing devices, monitors, etc.). This communication can be done through input / output (I / O) interfaces. Furthermore, computer devices can communicate with one or more networks (such as local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via network adapters.

[0063] The processing unit executes the functions and / or methods described in the embodiments of the present invention by running programs stored in the system memory.

[0064] The aforementioned computer program can be stored in a computer storage medium, that is, the computer storage medium is encoded with a computer program, which, when executed by one or more computers, causes one or more computers to perform the method flow and / or device operation shown in the above embodiments of the present invention.

[0065] With the development of time and technology, the meaning of "medium" has become increasingly broad. The dissemination of computer programs is no longer limited to tangible media; they can also be downloaded directly from the network. Any combination of one or more computer-readable media can be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. Computer-readable storage media can be, for example,—but not limited to—electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0066] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including—but not limited to—electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0067] The program code contained on a computer-readable medium may be transmitted using any suitable medium, including—but not limited to—wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0068] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0069] It should be noted that the technological achievements of this application were primarily supported by the Jiangsu Provincial Key Laboratory of Industrial Artificial Intelligence.

[0070] In addition to the above embodiments, the present invention may have other implementation methods; all technical solutions formed by equivalent substitution or equivalent transformation fall within the protection scope claimed by the present invention.

Claims

1. A method for optimizing and controlling coal gas desulfurization parameters based on causal identification and counterfactual inference, characterized in that, include: Obtain multi-dimensional process parameter data of the coal gas desulfurization system; Based on the coal gas desulfurization production process chain and the multi-dimensional process parameter data, a causal graph model is constructed to characterize the causal dependencies between desulfurization process parameters. Anomaly identification is performed based on the multi-dimensional process parameter data to obtain an abnormal parameter set. The parent node of the abnormal parameter set is traced back in the causal graph model to calculate the causal effect and determine the root variable that causes the abnormal sulfur content of coal gas. Using the adjustable parameters in the root cause parameters as intervention variables, counterfactual inference is performed on the causal graph model, and the expected sulfur content of coal gas after adjusting the intervention variables is iteratively predicted under the condition that the physical boundary limit of the adjustable parameters is met. Based on preset process joint control constraints, an intervention strategy is selected that minimizes the expected sulfur content of coal gas below the threshold for exceeding the sulfur content limit while minimizing adjustment costs. An industrial physical control command is then generated and sent to the field process equipment for physical control.

2. The method for optimizing and controlling coal gas desulfurization parameters based on causal identification and counterfactual inference according to claim 1, characterized in that, The causal graph model, constructed based on the coal gas desulfurization production process chain and the multi-dimensional process parameter data, is used to characterize the causal dependencies among desulfurization process parameters, including: Based on the desulfurization production process flow, the process stages are divided, and the initial cause-effect diagram structure is built layer by layer by combining the equipment monitoring parameters and control parameters corresponding to each process. Parameters that simultaneously affect multiple indirect processes are defined as hybrid variables and added to the covariate set, and directed edges starting from the sulfur content of coal gas are removed. Based on actual production data, the initial causal graph structure is subjected to conditional independence testing, pseudo-causal edges are removed, and data-driven causal edges are added to obtain the causal graph model.

3. The method for optimizing and controlling coal gas desulfurization parameters based on causal identification and counterfactual inference according to claim 1, characterized in that, The abnormal parameter set obtained by anomaly identification based on the multi-dimensional process parameter data includes: A combination of point anomaly detection and subsequence anomaly detection is used to calculate a standardized anomaly comprehensive score for each process parameter; The abnormal parameters are sorted in descending order based on the standardized anomaly comprehensive score, and the first preset number of parameters in the sorted sequence are selected to form the anomaly parameter set.

4. The method for optimizing and controlling coal gas desulfurization parameters based on causal identification and counterfactual deduction according to claim 2, characterized in that, The step of tracing back the parent nodes of the abnormal parameter set in the causal graph model to calculate the causal effect and determine the root cause variable leading to the abnormal sulfur content in the coal gas includes: In the target cause-effect graph model, for each abnormal parameter in the abnormal parameter set, its parent node is recursively traced upwards. A causal forest model was adopted, and the confounding factors in the set of covariates were adjusted using the backdoor and frontdoor criteria to block non-causal paths. The average causal effect of each candidate parent node parameter on the target sulfur content of coal gas was calculated. Sort the parameters in descending order according to the absolute value of the average causal effect, and select the first preset number of parameters in the sorted sequence as the root dependent variable.

5. The method for optimizing and controlling coal gas desulfurization parameters based on causal identification and counterfactual inference according to claim 4, characterized in that, The use of adjustable parameters from the root cause parameters as intervention variables includes: Obtain the adjustable parameter set of the gas desulfurization system; Calculate the intersection of the root dependent variable and the set of adjustable parameters to obtain candidate intervention variables; Based on the set of covariates, a collinearity test is performed on the candidate intervention variables, and candidate intervention variables with collinearity test indices greater than a preset collinearity threshold are eliminated to obtain the final intervention variables; When the intersection is an empty set, a process design anomaly alert is generated indicating that there are no effective intervention variables.

6. The method for optimizing and controlling coal gas desulfurization parameters based on causal identification and counterfactual inference according to claim 1, characterized in that, The step of performing counterfactual inference on the causal graphical model and iteratively predicting the expected sulfur content of the coal gas after adjusting the intervention variables, under the condition that the physical boundary limits of the adjustable parameters are met, includes: For the current value of the intervention variable, a two-way counterfactual inference is performed, adjusting upwards and downwards according to a preset adjustment step size ratio; Substitute the intervened variable values ​​into a pre-set causal machine learning model to predict the corresponding expected sulfur content of coal gas and its confidence interval. The bidirectional counterfactual deduction is continuously iterated until the expected sulfur content of the coal gas is lower than the sulfur content exceeding the standard threshold, or the physical boundary limit is reached, or the maximum number of iterations is reached.

7. The method for optimizing and controlling coal gas desulfurization parameters based on causal identification and counterfactual deduction according to claim 6, characterized in that, The aforementioned intervention strategy, based on preset process joint control constraints, selects the intervention strategy that minimizes the expected sulfur content in the coal gas and reduces the sulfur content exceeding the standard threshold, thereby generating industrial physical control instructions, including: Identify the parameter groups that need to be jointly controlled based on the preset process joint control constraints, and force the adjustable parameters within the same parameter group to maintain the same direction of change when performing counterfactual inference; With the goal of reducing the total sulfur content of coal gas to below the threshold for exceeding the sulfur content limit, and under the premise of satisfying the physical boundary limits of each adjustable parameter and the joint unidirectional regulation constraint, the comprehensive regulation cost of each intervention strategy is calculated. The strategy with the lowest total adjustment cost is selected as the optimal intervention strategy, and a structured control instruction containing the intervention method, adjustment ratio, absolute adjustment amount and its confidence interval is generated as the industrial physical adjustment instruction.

8. A gas desulfurization parameter optimization and control system based on causal identification and counterfactual inference, characterized in that, include: The data acquisition module is used to acquire multi-dimensional process parameter data of the gas desulfurization system; The causal graph construction module is used to construct a causal graph model to characterize the causal dependencies between desulfurization process parameters based on the coal gas desulfurization production process chain and the multi-dimensional process parameter data. The root cause tracing module is used to identify anomalies based on the multi-dimensional process parameter data to obtain an abnormal parameter set, and to trace back the parent node of the abnormal parameter set in the causal graph model to calculate the causal effect and determine the root cause variable that leads to the abnormal sulfur content of coal gas. The counterfactual deduction module is used to perform counterfactual deduction on the causal graph model using the adjustable parameters in the root cause parameters as intervention variables, and iteratively predict the expected sulfur content of the coal gas after adjusting the intervention variables, under the condition that the physical boundary limit of the adjustable parameters is met. The instruction output module is used to select the intervention strategy that minimizes the expected sulfur content of coal gas and the adjustment cost based on preset process joint control constraints, generate industrial physical control instructions, and send them to the field process equipment for physical control.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the gas desulfurization parameter optimization and control method based on causal identification and counterfactual deduction as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the gas desulfurization parameter optimization and control method based on causal identification and counterfactual deduction as described in any one of claims 1 to 7.

Images