A process parameter optimization method for a continuous production process of 2-pentylanthraquinone

Abstract

This invention provides a method for optimizing process parameters in the continuous production of 2-pentylanthraquinone, relating to the field of industrial process adaptive control technology. It constructs a coupling matrix reflecting the dynamic coupling relationship between parameters, calculates cross-correlation delay response coefficients to generate a parameter transfer effect spectrum, and then constructs a multi-objective constrained optimization function considering transfer delays, performing dynamic weight allocation and adaptive optimization. Finally, it outputs the coordinated process parameter setpoints for each reactor. This effectively characterizes the strong coupling and time-delay transfer characteristics among multiple parameters in continuous production, overcoming the shortcomings of traditional static parameter setting schemes that cannot adapt to dynamic disturbances, improving product quality stability and reducing energy consumption, and achieving adaptive global coordinated optimization of process parameters.

Description

Technical Field

[0001] This invention relates to the field of adaptive control technology for industrial processes, and more specifically, to a method for optimizing process parameters in a continuous production process of 2-pentylanthraquinone. Background Technology

[0002] In the field of process industry automation, process parameter optimization methods have evolved from manual experience-based control to the application of classical control theory, and then to data-driven intelligent optimization. Early process parameter settings relied primarily on the historical operational experience of engineers and discrete experimental data, maintaining production line operation through fixed parameter schemes or local single-variable adjustments. With the introduction of advanced control strategies such as Model Predictive Control (MPC) and Multivariable Statistical Process Control (MPC), process control systems have improved their ability to handle multivariable coupling problems to some extent. However, these methods are mostly based on assumptions of local linearization or fixed models identified offline, making it difficult to fully adapt to the nonlinear dynamic characteristics of industrial scenarios, such as operating condition drift, random disturbances, and time-varying constraints. In recent years, intelligent process control integrating artificial intelligence has become a research hotspot. Hybrid modeling methods combining data-driven modeling and mechanistic knowledge (such as the mechanistic-aware multimodal modeling framework based on state modulation graph networks) are dedicated to characterizing the nonlinear dynamic coupling relationships between variables in industrial processes. Simultaneously, rolling time-domain dynamic real-time optimization (D-RTO) and model predictive control frameworks have also made preliminary explorations in handling the inherent transmission delays caused by material recirculation. However, existing optimization methods generally employ fixed-weight allocation strategies to address the trade-offs between multiple objectives such as purity and energy consumption, lacking the ability to adaptively adjust to the dynamic changes in the importance of different optimization objectives as operating conditions change. Reports indicate that current AI-integrated intelligent control systems can reduce energy consumption in industrial processes by 15% to 25%, but in the field of continuous fine chemical production, process parameter optimization methods still have significant room for improvement in areas such as multivariate dynamic coupling characterization and prediction of time-delay transmission effects.

[0003] In the field of hydrogen peroxide (H2O2) production, over 95% of H2O2 internationally is currently produced via the anthraquinone process. 2-Pentylanthraquinone, due to its high solubility and resistance to degradation under high hydrogenation efficiency, has become a favored working medium in the anthraquinone process and is widely used by domestic and international producers such as Solvay, Evonik, and Sinopec. The continuous production process of 2-pentylanthraquinone typically involves multi-stage reactor operations in series, including alkylation and acylation reactions. This process includes multiple steps such as catalytic reactions and separation / purification. The process parameters within each reactor, such as temperature, pressure, material flow rate, and catalyst concentration, are interdependent and mutually restrictive. Furthermore, when material reflux exists in the system, adjusting a single parameter can trigger a chain reaction along the transmission path between multiple nodes, exhibiting typical characteristics of strong multivariable coupling and large time-delay dynamics. Currently, 2-pentylanthraquinone production generally suffers from problems such as lengthy production processes, low yields, numerous byproducts, and high energy consumption. The industry's demand for technologies to shorten production processes, improve product quality, and reduce energy consumption is increasingly urgent. Existing process parameter optimization methods typically rely on historical experience or discrete experimental data to formulate fixed parameter schemes. These methods cannot handle dynamic disturbances caused by changes in raw material composition or fluctuations in market demand during continuous production, and are even less effective in addressing system dynamic instability caused by strong coupling and time delay effects between parameters. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method for optimizing process parameters in the continuous production of 2-pentylanthraquinone. This method can, to some extent, solve the problem that existing process parameter optimization methods, which are based on historical experience or discrete experimental data, cannot handle the dynamic coupling effects between multiple parameters in continuous production. In particular, when there is material circulation and reflux between reactors, adjusting a single parameter can trigger a chain reaction, affecting the stability and economy of the entire production system.

[0005] According to one aspect of the present invention, a method for optimizing process parameters in a continuous production process of 2-pentylanthraquinone is provided, comprising:

[0006] Real-time process parameter data of each reactor in the 2-pentylanthraquinone continuous production line were obtained, and temperature, pressure and flow parameters were correlated and labeled to construct a dynamic coupling relationship matrix reflecting the degree of mutual influence between parameters.

[0007] Based on the dynamic coupling relationship matrix, the cross-correlation delay response coefficients of each process parameter are calculated, the transmission path characteristics of the influence of a single parameter change on other parameters are identified, and a parameter transmission effect spectrum is generated.

[0008] Based on the parameter transfer effect spectrum, a multi-objective constrained optimization function considering the transfer delay between parameters is constructed. The product purity objective and the energy consumption objective per unit output are dynamically weighted to generate a parameter optimization search domain with time-varying characteristics.

[0009] Within the parameter optimization search domain, a gradient adaptive search algorithm is used to perform global optimization calculations and output the optimal process parameter coordination configuration scheme for each reactor.

[0010] Furthermore, the correlation marking is performed using physical causal chain tracing logic, specifically including:

[0011] Identify the thermal, mechanical, and physical property effects caused by parameter changes, mark the affected parameters along the heat transfer path, pressure propagation path, or physical property change path, and mark them in order of priority according to the speed of effect propagation, forming a correlation marking network that covers the physical causal relationship between parameters.

[0012] Furthermore, the dynamic coupling relationship matrix is ​​shown in the following equation:

[0013]

[0014] in, For a moment Time parameters For parameters The dynamic coupling strength, For a moment Time parameters With parameters The associated marker value, which takes the value 0 or 1. For a moment Time parameters With parameters The normalized correlation coefficient, For a moment Time parameters With parameters The intensity of physical effects For a moment Time parameters to parameters The cumulative path complexity For a moment Time parameters Influencing parameters Normalized response time Adjust parameters for response time sensitivity.

[0015] Furthermore, when calculating the cross-correlation delay response coefficient, for each parameter pair with non-zero coupling strength in the dynamic coupling relationship matrix, the time series data of one parameter in the parameter pair is used as the input signal and the time series data of the other parameter is used as the response signal. The cross-correlation function is calculated by time offset scanning, the time offset corresponding to the peak value of the cross-correlation function is used as the delay response time, and the peak value is used as the delay response coefficient.

[0016] When the same parameter affects multiple other parameters simultaneously, separate input / output analysis channels are established for calculation.

[0017] Furthermore, the parameter transfer effect spectrum is generated as follows:

[0018] Starting with the target parameter, the objects directly affected by the parameter are determined based on the delay response coefficient and designated as the first-level transmission nodes, recording the corresponding delay time and response intensity.

[0019] Then, starting from the first-level transmission node, the process expands step by step to identify multi-level transmission nodes, marks the paths that form loops, and removes paths whose accumulated delay time exceeds the maximum allowable delay of the process or whose product of response intensity is lower than the threshold. Finally, the influence transmission network of each parameter is output in the form of a directed graph, where nodes correspond to process parameters and directed edges are labeled with delay time and response intensity.

[0020] Furthermore, constructing the multi-objective constrained optimization function includes the following steps:

[0021] Based on the parameter transfer effect spectrum, the steady-state gain, pure delay time, and inertial time constant of each process parameter on product purity and energy consumption per unit output are extracted, and a first-order inertial prediction model with pure delay is established.

[0022] A rolling time-domain optimization framework is adopted, with the sum of squares of the deviations of product purity from the set value and the sum of squares of the ratios of energy consumption per unit output to the benchmark value as objective terms, and a penalty term for the magnitude of parameter changes is added.

[0023] The dynamic weight configuration is based on the deviation of the product purity and energy consumption values ​​from the target benchmark values ​​monitored in real time, the production plan requirements, and the improvement speed of each target, and adaptively adjusts the weight coefficients of the purity target and the energy consumption target.

[0024] Furthermore, the first-order inertial prediction model with pure delay is shown in the following equation:

[0025]

[0026]

[0027] in, For at any time The predicted change in product purity due to adjustments in various process parameters. The corresponding predicted change in energy consumption per unit of output. For parameters The steady-state gain on product purity, Let i be the pure delay time from the change in parameter i until the product purity begins to respond. The first-order inertial time constant, For the Laplace operator, Let i be the adjustment amount of parameter i at time t. , , These are the steady-state gain, pure delay time, and inertial time constant of parameter i per unit output energy consumption, respectively.

[0028] According to another aspect of the present invention, a process parameter optimization system for a continuous production process of 2-pentylanthraquinone is provided, comprising:

[0029] The data acquisition and coupling analysis module is used to acquire real-time process parameter data of each reactor in the 2-pentylanthraquinone continuous production line, to perform correlation marking on temperature, pressure and flow parameters, and to construct a dynamic coupling relationship matrix that reflects the degree of mutual influence between parameters.

[0030] The transfer path identification and effect spectrum generation module is used to calculate the cross-correlation delay response coefficient of each process parameter based on the dynamic coupling relationship matrix, identify the transfer path characteristics that affect other parameters by a change in a single parameter, and generate a parameter transfer effect spectrum.

[0031] The multi-objective optimization function construction and search domain generation module is used to construct a multi-objective constrained optimization function considering the transmission delay between parameters based on the parameter transfer effect spectrum, dynamically configure the product purity target and the unit output energy consumption target with dynamic weights, and generate a parameter optimization search domain with time-varying characteristics.

[0032] The global optimization and scheme output module is used to perform global optimization calculations using a gradient adaptive search algorithm within the parameter optimization search domain, and output the optimal process parameter coordination configuration scheme for each reactor.

[0033] Compared with existing technologies, the process parameter optimization method for the continuous production of 2-pentylanthraquinone provided by this invention constructs a coupling matrix reflecting the dynamic coupling relationship between parameters, calculates the cross-correlation delay response coefficient to generate a parameter transfer effect spectrum, and then constructs a multi-objective constrained optimization function considering the transfer delay, performs dynamic weight configuration and adaptive optimization, and finally outputs the coordinated process parameter setpoints for each reactor. This effectively characterizes the strong coupling and time-delay transfer characteristics between multiple parameters in continuous production, overcomes the shortcomings of traditional static parameter setting schemes that cannot adapt to dynamic disturbances, improves product quality stability and reduces energy consumption, and achieves adaptive global coordinated optimization of process parameters. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0035] Figure 1 This is a flowchart of a method for optimizing process parameters in a continuous production process of 2-pentylanthraquinone according to an embodiment of the present invention.

[0036] Figure 2 This diagram illustrates an application scenario of the process parameter optimization method for the continuous production process of 2-pentylanthraquinone according to an embodiment of the present invention. Detailed Implementation

[0037] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0038] Figure 1 This is a flowchart illustrating a method for optimizing process parameters in the continuous production process of 2-pentylanthraquinone according to an embodiment of the present invention. Figure 1 As shown, the process parameter optimization method for the continuous production of 2-pentylanthraquinone includes:

[0039] S1: Obtain real-time process parameter data of each reactor in the 2-pentylanthraquinone continuous production line, and perform correlation marking on temperature, pressure and flow parameters to construct a dynamic coupling relationship matrix that reflects the degree of mutual influence between parameters;

[0040] In the continuous production process of 2-pentylanthraquinone, real-time process parameter data are first collected from the sensor network of each reactor and its supporting equipment through a distributed control system. This data includes temperature and pressure parameters inside each reactor, as well as flow parameters in connecting pipelines. Temperature parameters cover data from multiple measuring points such as reactor wall temperature, catalyst bed temperature, and inlet / outlet material temperatures. Pressure parameters include monitored values ​​such as reactor internal pressure, pipeline differential pressure, and pumping pressure. Flow parameters include material flow data from various pathways, such as feed flow, product discharge flow, and recirculation flow.

[0041] Furthermore, when labeling temperature, pressure, and flow parameters, a unified physical causal chain tracing logic is used for systematic labeling. If any parameter changes, the type of direct physical effect of that parameter change is first identified, including three basic categories: thermal effect, mechanical effect, and physical property effect. When a parameter change produces a thermal effect, all affected temperature measurement points are labeled as heat transfer correlations along the heat transfer path, and the pressure and flow parameters corresponding to changes in density or viscosity due to temperature changes are labeled as thermophysical property correlations along the physical property change path. If a parameter change produces a mechanical effect, all upstream and downstream pressure measurement points are labeled as pressure transfer correlations along the pressure propagation path, and all flow measurement points affected by pressure are labeled as flow dynamic correlations along the flow resistance change path. When a parameter change produces a physical property effect, the relevant flow and pressure parameters are labeled as physical property flow correlations based on the impact of the physical property change on flow characteristics, and the relevant temperature parameters are labeled as physical property heat transfer correlations based on the impact of the physical property change on heat transfer characteristics. If a parameter produces multiple physical effects simultaneously, they are associated and labeled in order of priority according to the intensity of the effects, ultimately forming a complete association labeling network that covers the physical causal relationships between all parameters.

[0042] For example, when assigning correlation labels to temperature, pressure, and flow parameters, the specific correlation labeling logic is illustrated using the outlet pressure parameter of the first reactor as an example. If the outlet pressure of the first reactor increases, this pressure change first produces a mechanical effect, which is transmitted along the material flow direction to the connecting pipe and the inlet of the second reactor. Therefore, the outlet pressure of the first reactor is labeled as pressure transmission correlation with the pressure of the connecting pipe and the inlet pressure of the second reactor. When the outlet pressure of the first reactor increases, it increases the driving force for the material to flow through the pipe, resulting in an increase in the flow rate within the pipe. Therefore, the outlet pressure of the first reactor is labeled as flow dynamic correlation with the corresponding pipe flow rate parameter. If the pressure increase also causes the boiling point of the material in the reactor to rise, it will affect the temperature distribution and phase change behavior within the reactor. Therefore, the outlet pressure of the first reactor is labeled as thermodynamic correlation with the internal temperature parameter of the first reactor. When the pipe flow rate increases due to the pressure change, the increased flow velocity enhances the convective heat transfer at the pipe wall, affecting the temperature change of the material in the pipe. Therefore, the pipe flow rate parameter is labeled as convective heat transfer correlation with the pipe outlet temperature parameter. If the temperature at the pipe outlet changes and the material enters the second reactor, it will alter the feed temperature and reaction conditions of the second reactor. Therefore, the pipe outlet temperature and the reaction temperature of the second reactor are marked as heat transfer correlations. This correlation marking is completed by gradually tracing all physical effect chains starting from a single parameter.

[0043] Furthermore, a matrix framework is established using all process parameters as row and column indices, with each matrix element representing the degree of influence of the corresponding row parameter on the column parameter. If two parameters are identified as having a direct physical connection in the correlation marker, the degree of influence is quantified by analyzing the response time characteristics of parameter changes. If there is an intermediate transmission link between two parameters, the degree of influence is adjusted according to the complexity of the transmission path. When the transmission path includes multiple intermediate devices, the degree of influence is weakened accordingly; when the transmission path is a single direct-connection pipe, the degree of influence remains at a high level. When there is a bidirectional interaction between parameters, the degree of influence in both directions is calculated and filled into the corresponding positions in the matrix, forming an asymmetric coupling relationship representation. If there is no direct or indirect physical connection path between parameters, a zero value is filled into the corresponding position in the matrix to indicate no coupling relationship. After the degree of influence of all parameter pairs is evaluated, the matrix is ​​dynamically updated by introducing a time window sliding mechanism. The matrix element values ​​are adjusted in real time according to the changes in parameter response characteristics during actual production, ultimately forming a dynamic coupling relationship matrix that reflects the true coupling strength between parameters under the current process state. The dynamic coupling relationship matrix is ​​shown in the following formula:

[0044]

[0045] in,

[0046]

[0047]

[0048]

[0049] in, For a moment Time parameters For parameters The dynamic coupling strength, For a moment Time parameters With parameters The associated marker value, which takes the value 0 or 1. For a moment Time parameters With parameters The normalized correlation coefficient, For a moment Time parameters With parameters The intensity of physical effects For a moment Time parameters to parameters The cumulative path complexity For a moment Time parameters Influencing parameters Normalized response time For a moment Time parameters The measured value, For a moment Time parameters The measured value, Parameters within the time window The arithmetic mean, Parameters within the time window The arithmetic mean, This is the numerical stability protection constant. This is the weighting coefficient for the thermal effect. For mechanical effect weighting coefficients, This is the weighting coefficient for physical property effects. For parameters With parameters The thermal effect is marked with a value of 0 or 1. For parameters With parameters The mechanical effect is marked with a value of 0 or 1. For parameters With parameters The physical property effect label value takes the value of 0 or 1. For parameters to parameters The total number of intermediate transmission links, For the first The complex quantification value of each intermediate transmission link, For a moment Time parameters Influencing parameters The measured response delay time, This represents the maximum possible response delay within the system. The parameter is adjusted to respond to time sensitivity. The length of the sliding time window. For the current moment, For matrix row index, For matrix column index, Index of moments within a time window This is an index for the transmission process.

[0050] S2: Calculate the cross-correlation delay response coefficients of each process parameter based on the dynamic coupling relationship matrix, identify the transmission path characteristics of the influence of a single parameter change on other parameters, and generate a parameter transmission effect spectrum;

[0051] First, parameter pairs with non-zero coupling strength are selected from the matrix as the computational objects, and their corresponding historical time-series data sequences are extracted. If the matrix elements show that parameter A is coupled to parameter B, the time-series data of parameter A is used as the input signal, and the time-series data of parameter B is used as the response signal for delay analysis. Once the input and response signals are determined, the time position of the input signal is gradually adjusted using a time offset scanning method, and the correlation between the input and response signals is calculated at each time offset. If the correlation between the two signals reaches a peak at a specific time offset, then that offset is the delay response time, and the corresponding correlation peak is the delay response coefficient. After calculating the delay response coefficient for a single parameter pair, all parameter pairs corresponding to non-zero coupling relationships in the matrix are processed sequentially, and their respective delay response times and response strength coefficients are calculated. If a parameter simultaneously affects multiple other parameters, the independent delay response coefficients between the parameter and each affected parameter need to be calculated separately, forming a one-to-many response coefficient set, ultimately obtaining a complete delay response coefficient dataset covering all effective parameter coupling relationships.

[0052] In calculating the correlation between the input and response signals at each time offset, the input signal is first shifted along the time axis according to the current time offset, causing a corresponding misalignment between the time labels of the input signal and the response signal. If the time offset is positive, the input signal is shifted backward by the corresponding time step; if the time offset is negative, the input signal is shifted forward by the corresponding time step. After time alignment, corresponding data points of the two signals within the overlapping time period are extracted, and the numerical difference and consistency of the trend of each pair of corresponding data points are calculated one by one. If both signals show an upward trend or a downward trend at a certain time point, it is recorded as a positive correlation contribution; if one signal rises while the other falls, it is recorded as a negative correlation contribution. After completing the correlation assessment for all overlapping time points, all positive and negative correlation contributions are statistically combined. The overall correlation value at that time offset is obtained by calculating the weighted average of the ratio of the number of positively correlated points to the total number of comparison points and the correlation strength. This value reflects the degree of matching between the change patterns of the input signal and the change patterns of the response signal under the current time offset condition.

[0053] When calculating the independent delay response coefficients between the parameter and each affected parameter, the parameter is used as a unified input signal source, and independent input-output analysis channels are established for each affected parameter. If the parameter simultaneously affects three different affected parameters, three independent calculation channels need to be established, with each channel considering only the binary relationship between the parameter and its corresponding affected parameter. After establishing independent calculation channels, time offset scanning and correlation calculation are performed in each channel. The optimal matching delay time with the current affected parameter is found by adjusting the time position of the parameter's time series data. If the parameter's change pattern is found to have the highest correlation with the affected parameter at a specific time offset in a certain channel, the offset time is recorded as the delay response time of the corresponding affected parameter, and the highest correlation value is recorded as the corresponding response intensity coefficient. After completing the delay response analysis of all independent channels, the delay time and response coefficients corresponding to each affected parameter are stored separately, forming a multivariate delay response coefficient combination with the parameter as the core. Each coefficient independently reflects the individual effect of the parameter on a specific affected parameter, avoiding the influence of mutual interference between multiple parameters on the calculation results.

[0054] Furthermore, starting with the target parameter, the set of objects directly affected by that parameter is determined based on the calculation results of the delay response coefficient. If the delay response coefficient of a parameter shows that it has a direct impact on multiple parameters, these parameters are marked as first-level transmission nodes, and the corresponding delay time and response intensity are recorded as first-level transmission features. After determining the first-level transmission nodes, each first-level node is used as a new starting point to continue searching for its corresponding delay response coefficient, identifying the parameters further affected by these nodes as second-level transmission nodes. If a parameter identical to the original parameter appears in a second-level node, a transmission loop is formed, and loop marking is required in path identification to avoid infinite loops. After the identification of multi-level transmission nodes is completed, the effectiveness of the transmission path is evaluated by analyzing the cumulative delay time and response intensity decay law of each level of transmission. Paths with delay times exceeding the maximum allowable time delay of the process or response intensity products below a threshold are marked as negligible paths and removed from the analysis. If the transmission path has reasonable time characteristics and sufficient response strength, the starting point parameters, intermediate node parameters, ending point parameters, and corresponding transmission time and strength information of the path are integrated to form a complete transmission path description. Finally, all effective transmission paths are classified and organized according to the starting point parameters to generate a parameter transmission effect spectrum in the form of a graph.

[0055] It should be noted that the maximum allowable time delay for the process is determined based on the control cycle and response requirements of the production process. If the standard control adjustment cycle of the production line is one time unit, the maximum allowable time delay is typically set to two to three times that control cycle to ensure that parameter adjustments produce observable effects within a reasonable control time range. When the process requires high response speed, the maximum time delay is set to one to two times the control cycle. When the process is relatively stable and allows for a longer response time, the maximum time delay can be set to three to five times the control cycle.

[0056] The response intensity product threshold is set based on the measurement accuracy and normal fluctuation range of the endpoint parameter. If the measurement system accuracy of a parameter is a specific proportion of its normal value, the response intensity product threshold is set to two to three times that measurement accuracy to ensure that parameter changes generated by the transmission path can be reliably detected and distinguished. When the normal process fluctuation range of the endpoint parameter is known, the threshold is set to one-tenth to one-fifth of that fluctuation range, so that only transmission paths that can produce significant effects beyond normal fluctuations are retained. If the process system requires extremely high control accuracy for certain key parameters, the response intensity threshold of the transmission path with that parameter as the endpoint is increased accordingly to ensure that the retained paths can produce significant effects in an engineering sense.

[0057] For example, the reactor temperature parameter in a continuous 2-pentylanthraquinone production line illustrates the composition of the parameter transfer effect spectrum. When the reaction temperature of the first reactor, used as a starting parameter, changes, the transfer effect spectrum first shows its direct impact on the discharge temperature of the first reactor. The delay time is the mixing time of the materials in the reactor, and the response intensity is relatively high, forming the first-level transfer node. If the discharge temperature of the first reactor increases, it affects the feed temperature of the second reactor through pipeline transmission. The delay time is the transmission time of the materials in the connecting pipeline, and the response intensity is slightly reduced due to heat dissipation from the pipeline, forming the second-level transfer node. When the feed temperature of the second reactor changes, it further affects the reaction temperature and catalyst activity in the second reactor, while also changing the pressure distribution and discharge flow rate of the second reactor. The delay time is the heat equilibrium time inside the reactor, and the response intensity is amplified or reduced according to the heat effect of the reaction, forming the third-level transfer node. If the change in the flow rate of the second reactor, in turn, affects the back pressure of the first reactor, a feedback transfer path is formed from the third-level node back to the first-level node. The parameter transfer effect spectrum represents these transfer paths in the form of a network diagram. Nodes represent the affected parameters, lines represent the transfer paths, and delay time and response intensity information are marked on the lines. Overall, it forms a complete effect spectrum that reflects the influence range and transfer mechanism of the temperature parameter of the first reactor in the entire production system.

[0058] It should be noted that in step S1 The coupling strength was calculated using the response time factor. The reason for recalculating in step S2 is that S1 and S2 address two different levels of problems in parameter coupling analysis. The dynamic coupling matrix constructed in step S1 primarily addresses the identification of whether and how strong the influence relationship exists between parameters. Its calculated coupling strength reflects the potential and theoretical impact of the interaction between parameters, evaluated based on engineering principles such as equipment physical connections and heat and mass transfer mechanisms. After calculating the coupling strength in S1, although the existence of an influence relationship between parameters is known, the specific time-series manifestation of this relationship remains undetermined. Specifically, it cannot answer how long it takes for a change in parameter A to affect parameter B, or what response pattern this effect exhibits over time. Based on the coupling relationship determined in S1, the temporal dynamic characteristics of the influence relationship between parameters are further explored. If S1 shows a strong coupling relationship between parameter A and parameter B, S2 identifies the optimal time-series match between the change pattern of parameter A and the response pattern of parameter B by analyzing historical time-series data of parameters A and B in actual production processes. This recalculation is not a negation or repetition of the S1 results, but a dynamic extension based on the static coupling analysis of S1. It verifies and quantifies the actual performance of the coupling relationships identified in S1 in the time dimension through actual data, and finally forms a complete parametric relationship description that includes both the intensity of influence and the temporal characteristics.

[0059] S3: Construct a multi-objective constrained optimization function that considers the transmission delay between parameters based on the parameter transfer effect spectrum, dynamically configure the product purity target and the energy consumption target per unit output, and generate a parameter optimization search domain with time-varying characteristics.

[0060] First, extract the delay time of each propagation path from the parameter transfer effect spectrum. and response strength (Regarding purity) and (Regarding energy consumption). For each adjustable process parameter, accumulate the total gain and total delay based on its transmission path to product purity P; similarly, obtain the total gain for energy consumption per unit output E. Total delay When multiple parallel paths exist, the dominant path or the superimposed responses are selected.

[0061] Furthermore, a first-order inertial model with pure delay is constructed to approximate the effect of each parameter on the two targets:

[0062]

[0063]

[0064] in, Indicates at time The predicted change in product purity due to adjustments in various process parameters. The corresponding predicted change in energy consumption per unit of output. For parameters The steady-state gain on product purity (i.e., the magnitude of purity change caused by a unit change in parameter). Let i be the pure delay time from the change in parameter i until the product purity begins to respond. It is the first-order inertial time constant (characterizing the rate of response rise). For the Laplace operator, Let this be the adjustment amount of parameter i at time t. Similarly, , , These are the steady-state gain, pure delay time, and inertial time constant of parameter i per unit output, respectively. These parameters are extracted from the parameter transfer effect spectrum generated by S2 through path gain accumulation and delay time.

[0065] Based on the above prediction model, a rolling time-domain multi-objective optimization function is constructed:

[0066]

[0067]

[0068]

[0069] in, This indicates finding the minimum value, where the decision variable is the value from the current time t to the control time domain. In-step parameter adjustment sequence , ,…, , where t after the vertical bar indicates that the sequence is the planned value set at time t; This represents the total cost function value; The time-varying weighting coefficient (ranging from 0 to 1) is used to dynamically balance the importance of product purity targets and energy consumption targets. Its value is determined by real-time monitoring deviations and production plans through a dynamic weighting configuration mechanism. For prediction of the time domain (number of steps). For the prediction step index; Set a target value for product purity. This represents the purity value predicted by the model at time t+h. The allowable fluctuation range of purity (used for dimensionless conversion). This represents the energy consumption per unit output at the predicted time t+h. This serves as a benchmark reference value for energy consumption (also used for dimensionless conversion). This is the regularization coefficient (non-negative), used to penalize the magnitude of parameter adjustment to suppress drastic fluctuations. To control the time domain (number of steps). Let ii be the adjustment amount for parameter ii in step hh. This represents the maximum allowable variation of this parameter per step (dimensionless denominator). and These are the baseline values ​​for product purity and energy consumption at the current moment, respectively. For continuous delay time, For sampling and control cycle, This is an indicator function. It takes the value 1 when the current prediction step h is greater than or equal to the delay step, and 0 otherwise, indicating that the parameter adjustment has not yet had an effect during the delay period.

[0070] The constraints include:

[0071]

[0072]

[0073]

[0074]

[0075] in, and These are the minimum and maximum allowable operating values ​​for parameter i (e.g., minimum / maximum reactor temperature); Limit the rate of change per step; The normalized coupling strength is derived from the S1 dynamic coupling matrix. The delay time at which the change in parameter i affects parameter j (from S2). This serves as the upper bound of the i-th transitive delay constraint, used to limit the strongly coupled parameters from changing too rapidly during the delay time. The final safety constraint is... For example, it means that the predicted value of the reactor temperature must not exceed the safety limit, which can be extended to other safety constraints such as pressure and flow rate in practice.

[0076] The dynamic weight configuration mechanism in the multi-objective optimization function is as follows: The current product purity and energy consumption per unit output are obtained through a real-time monitoring system, and the deviation from their respective benchmark values ​​is calculated. If the product purity is lower than the set benchmark for three consecutive sampling periods, the weight is adjusted according to the step size. Increase (Purity weight), while reducing If the energy consumption per unit of output continuously exceeds the preset control line, then reduce... When the deviations between the two targets are close, manual settings are made based on the production plan (such as high-purity orders or cost constraints). The baseline is established. Furthermore, the descent gradient of the two objective function values ​​is monitored within a continuous optimization cycle, and weights are shifted towards the objective that improves more slowly to achieve balanced optimization. When external conditions change (such as decreased raw material purity or fluctuations in market demand), weight adjustments are automatically triggered through preset rules to ensure that the optimization direction remains consistent with production and operation objectives.

[0077] Furthermore, a time-varying parameter optimization search domain is generated: the initial search domain consists of the safe operating range of each parameter. Confirmed. If the current weight Higher purity levels will prioritize high-temperature, high-pressure regions that improve reaction selectivity, while regions that may lead to side reactions will be narrowed down or eliminated. If the parameter transfer effect spectrum shows that increasing a certain parameter (such as temperature) will result in a delay time... If this leads to downstream pressure exceeding limits, the potentially dangerous range of that parameter will be removed from the search domain at the current moment based on the prediction model. When production scheduling requires increased output, the search range of the flow parameter will shift towards the high flow range, and the temperature and pressure ranges will be adjusted accordingly to match the high flow conditions. When the purity of the raw materials decreases, the search range of the reaction temperature will shift towards the high temperature range, and the flow range will be adjusted towards the low flow range (extending the residence time). Through the above mechanisms, the search domain can dynamically change according to weights, delay effects, production plans, and raw material conditions.

[0078] S4: Within the parameter optimization search domain, a gradient adaptive search algorithm is used to perform global optimization calculations and output the optimal process parameter coordination configuration scheme for each reactor.

[0079] First, the time-varying parameter optimization search domain generated by S3 is used as the solution space, and the multi-objective optimization function JJ is used as the cost function. Since the objective function may be non-convex with respect to the control variables and has multiple local minima, a strategy combining gradient descent with an adaptive momentum term and random restart is adopted.

[0080] Initialization: Set the actual parameter values ​​at the current moment. As initial points, randomly generated initial points with minor disturbances are used to cover the search domain. For each initial point, the Adam optimizer is used to iteratively update the parameter sequence. In each iteration, the gradient of the cost function with respect to the control sequence is calculated (which can be obtained through analytical differentiation or finite difference), and then the control variables are updated. The process terminates when the cost function change is less than a threshold for several consecutive steps or when the maximum number of iterations is reached, and the optimal solution at that initial point is recorded. The solution with the minimum cost function among all initial points is selected as the global approximate optimal solution.

[0081] Furthermore, to meet real-time requirements, a rolling optimization strategy of Model Predictive Control (MPC) is adopted: the control increment is executed only for the first time step. It outputs the set value, remeasures the state at the next sampling time, and solves the optimization problem again. At the same time, it uses historical optimization results to identify and correct model parameters (gain, delay, time constant) online, improving model accuracy.

[0082] If, during the optimization process, a feasible solution satisfying all constraints cannot be found within the current search domain (e.g., constraint conflict), an incremental expansion of the search domain boundary is automatically triggered: first, secondary constraints (such as rate of change limits) are relaxed; if no solution is found, the parameter boundaries are gradually relaxed step by step (not exceeding the equipment's safety limit) until a feasible solution is obtained. This condition is recorded for subsequent adjustment of the constraint boundaries.

[0083] The final output of coordinated process parameter settings includes: temperature setpoint, pressure setpoint, feed flow rate setpoint, circulation reflux flow rate setpoint, catalyst addition rate, etc. for each reactor. These are distributed to each actuator (heater, regulating valve, pump, etc.) through a distributed control system to achieve adaptive optimization control of the continuous production process.

[0084] Example: For a continuous production line of 2-pentylanthraquinone, assume the current product purity is 2% lower than the target value, while energy consumption is within the normal range. S3 increases the dynamic weight α(t) to 0.7, shifting the temperature range of the first reactor upwards by 5°C and the pressure range of the second reactor upwards by 0.1 MPa in the search domain. After 5 iterations of optimization, S4 obtains a coordinated solution that increases the temperature of the first reactor by 3°C, the pressure of the second reactor by 0.05 MPa, and reduces the feed flow rate by 2%. After output, the production line operates under the new settings, and the product purity recovers to the target range after 15 minutes.

[0085] In summary, the method 1 for optimizing process parameters in the continuous production process of 2-pentylanthraquinone based on the embodiments of the present invention is explained. It constructs a coupling matrix reflecting the dynamic coupling relationship between parameters, calculates the cross-correlation delay response coefficient to generate a parameter transfer effect spectrum, then constructs a multi-objective constrained optimization function considering the transfer delay, performs dynamic weight configuration and adaptive optimization, and finally outputs the coordinated process parameter settings for each reactor. This effectively characterizes the strong coupling and time-delay transfer characteristics among multiple parameters in continuous production, overcomes the shortcomings of traditional static parameter setting schemes that cannot adapt to dynamic disturbances, improves product quality stability and reduces energy consumption, and achieves adaptive global coordinated optimization of process parameters.

[0086] According to another aspect of the present invention, a process parameter optimization system for a continuous production process of 2-pentylanthraquinone is provided, comprising:

[0087] The data acquisition and coupling analysis module is used to acquire real-time process parameter data of each reactor in the 2-pentylanthraquinone continuous production line, to perform correlation marking on temperature, pressure and flow parameters, and to construct a dynamic coupling relationship matrix that reflects the degree of mutual influence between parameters.

[0088] The transfer path identification and effect spectrum generation module is used to calculate the cross-correlation delay response coefficient of each process parameter based on the dynamic coupling relationship matrix, identify the transfer path characteristics that affect other parameters by a change in a single parameter, and generate a parameter transfer effect spectrum.

[0089] The multi-objective optimization function construction and search domain generation module is used to construct a multi-objective constrained optimization function considering the transmission delay between parameters based on the parameter transfer effect spectrum, dynamically configure the product purity target and the unit output energy consumption target with dynamic weights, and generate a parameter optimization search domain with time-varying characteristics.

[0090] The global optimization and scheme output module is used to perform global optimization calculations using a gradient adaptive search algorithm within the parameter optimization search domain, and output the optimal process parameter coordination configuration scheme for each reactor.

[0091] Here, those skilled in the art will understand that the specific operations of each step in the process parameter optimization system for the continuous production of 2-pentylanthraquinone have been referenced above. Figure 1 and Figure 2 The method for optimizing process parameters in the continuous production process of 2-pentylanthraquinone has been described in detail, and therefore, its repeated description will be omitted.

[0092] In summary, the process parameter optimization system for the continuous production process of 2-pentylanthraquinone based on the embodiments of the present invention has been clarified. It constructs a coupling matrix reflecting the dynamic coupling relationship between parameters, calculates the cross-correlation delay response coefficient to generate a parameter transfer effect spectrum, and then constructs a multi-objective constrained optimization function considering the transfer delay, performing dynamic weight configuration and adaptive optimization, ultimately outputting the coordinated process parameter setpoints for each reactor. This effectively characterizes the strong coupling and time-delay transfer characteristics among multiple parameters in continuous production, overcomes the shortcomings of traditional static parameter setting schemes that cannot adapt to dynamic disturbances, improves product quality stability, reduces energy consumption, and achieves adaptive global coordinated optimization of process parameters.

Claims

1. A process parameter optimization method for a continuous production process of 2-pentylanthraquinone, characterized by, include: Real-time process parameter data of each reactor in the 2-pentylanthraquinone continuous production line were obtained, and temperature, pressure and flow parameters were correlated and labeled to construct a dynamic coupling relationship matrix reflecting the degree of mutual influence between parameters. Based on the dynamic coupling relationship matrix, the cross-correlation delay response coefficients of each process parameter are calculated, the transmission path characteristics of the influence of a single parameter change on other parameters are identified, and a parameter transmission effect spectrum is generated. Based on the parameter transfer effect spectrum, a multi-objective constrained optimization function considering the transfer delay between parameters is constructed. The product purity objective and the energy consumption objective per unit output are dynamically weighted to generate a parameter optimization search domain with time-varying characteristics. Within the parameter optimization search domain, a gradient adaptive search algorithm is used to perform global optimization calculations and output the optimal process parameter coordination configuration scheme for each reactor.

2. The method for optimizing process parameters in the continuous production process of 2-pentylanthraquinone according to claim 1, characterized in that, The correlation marking is performed using physical causal chain tracing logic, specifically including: Identify the thermal, mechanical, and physical property effects caused by parameter changes, mark the affected parameters along the heat transfer path, pressure propagation path, or physical property change path, and mark them in order of priority according to the speed of effect propagation, forming a correlation marking network that covers the physical causal relationship between parameters.

3. The process parameter optimization method of the continuous 2-pentylanthraquinone production process according to claim 2, characterized in that, The dynamic coupling relationship matrix is ​​shown in the following equation: in, For a moment Time parameters For parameters The dynamic coupling strength, For a moment Time parameters With parameters The associated marker value, which takes the value 0 or 1. For a moment Time parameters With parameters The normalized correlation coefficient, For a moment Time parameters With parameters The intensity of physical effects For a moment Time parameters to parameters The cumulative path complexity For a moment Time parameters Influencing parameters Normalized response time Adjust parameters for response time sensitivity.

4. The method for optimizing process parameters in the continuous production process of 2-pentylanthraquinone according to claim 3, characterized in that, When calculating the cross-correlation delay response coefficient, for each parameter pair with non-zero coupling strength in the dynamic coupling relationship matrix, the time series data of one parameter in the parameter pair is used as the input signal and the time series data of the other parameter is used as the response signal. The cross-correlation function is calculated by time offset scanning, the time offset corresponding to the peak value of the cross-correlation function is used as the delay response time, and the peak value is used as the delay response coefficient. When the same parameter affects multiple other parameters simultaneously, separate input / output analysis channels are established for calculation.

5. The process parameter optimization method of the continuous production process of 2-pentylanthraquinone according to claim 4, characterized in that, The parameter transfer effect spectrum is generated as follows: Starting with the target parameter, the objects directly affected by the parameter are determined based on the delay response coefficient and designated as the first-level transmission nodes, recording the corresponding delay time and response intensity. Then, starting from the first-level transmission node, the process expands step by step to identify multi-level transmission nodes, marks the paths that form loops, and removes paths whose accumulated delay time exceeds the maximum allowable delay of the process or whose product of response intensity is lower than the threshold. Finally, the influence transmission network of each parameter is output in the form of a directed graph, where nodes correspond to process parameters and directed edges are labeled with delay time and response intensity.

6. The process parameter optimization method of the continuous 2-pentylanthraquinone production process according to claim 5, characterized in that, Constructing the multi-objective constrained optimization function includes the following steps: Based on the parameter transfer effect spectrum, the steady-state gain, pure delay time, and inertial time constant of each process parameter on product purity and energy consumption per unit output are extracted, and a first-order inertial prediction model with pure delay is established. A rolling time-domain optimization framework is adopted, with the sum of squares of the deviations between product purity and set values ​​and the sum of squares of the ratios of energy consumption per unit output to benchmark values ​​as objective terms, and a penalty term for the magnitude of parameter changes is added. The dynamic weight configuration is based on the deviation of the product purity and energy consumption values ​​from the target benchmark values ​​monitored in real time, the production plan requirements, and the improvement speed of each target, and adaptively adjusts the weight coefficients of the purity target and the energy consumption target.

7. The process parameter optimization method of the continuous 2-pentylanthraquinone production process according to claim 6, characterized in that, The first-order inertial prediction model with pure delay is shown in the following equation: in, For at any time The predicted change in product purity due to adjustments in various process parameters. The corresponding predicted change in energy consumption per unit of output. For parameters The steady-state gain on product purity, Let i be the pure delay time from the change in parameter i until the product purity begins to respond. The first-order inertial time constant, For the Laplace operator, Let i be the adjustment amount of parameter i at time t. , , These are the steady-state gain, pure delay time, and inertial time constant of parameter i per unit output energy consumption, respectively.

8. A system for optimizing process parameters of a continuous production process of 2-pentylanthraquinone, characterized in that include: The data acquisition and coupling analysis module is used to acquire real-time process parameter data of each reactor in the 2-pentylanthraquinone continuous production line, to perform correlation marking on temperature, pressure and flow parameters, and to construct a dynamic coupling relationship matrix that reflects the degree of mutual influence between parameters. The transfer path identification and effect spectrum generation module is used to calculate the cross-correlation delay response coefficient of each process parameter based on the dynamic coupling relationship matrix, identify the transfer path characteristics that affect other parameters by a change in a single parameter, and generate a parameter transfer effect spectrum. The multi-objective optimization function construction and search domain generation module is used to construct a multi-objective constrained optimization function considering the transmission delay between parameters based on the parameter transfer effect spectrum, dynamically configure the product purity target and the unit output energy consumption target with dynamic weights, and generate a parameter optimization search domain with time-varying characteristics. The global optimization and scheme output module is used to perform global optimization calculations using a gradient adaptive search algorithm within the parameter optimization search domain, and output the optimal process parameter coordination configuration scheme for each reactor. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. When the processor executes the computer program, it implements the steps of the process parameter optimization method for the continuous production process of 2-pentylanthraquinone according to any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the process parameter optimization method for the continuous production process of 2-pentylanthraquinone according to any one of claims 1 to 7.

Images