Medicament adding precision control method and system based on medicament injection timing prediction

Abstract

The application relates to a medicine adding precision control method and system based on medicine injection timing prediction, and relates to the technical field of industrial medicine control, which comprises the following steps: acquiring real-time operation parameters of a cracking furnace medicine adding system; calculating an actual comprehensive lag time length and a lag uncertainty, acquiring a nonlinear jump rate and a working condition disturbance increment; obtaining a front prediction medicine adding coefficient through linkage operation of the three; and generating a medicine control instruction based on dynamic prediction compensation of the medicine adding timing and injection amount based on the coefficient. The application solves the technical problem that, in traditional large time-lag medicine adding control, lag uncertainty, nonlinear response and external disturbance coupling influence cannot be considered, resulting in lagging adjustment of medicine adding timing and injection amount and poor precision.

Description

Technical Field

[0001] This application relates to the field of industrial dosing control technology, and in particular to a precise control method and system for dosing based on the prediction of dosing injection timing. Background Technology

[0002] In the high-pressure steam system of a cracking furnace, the phosphate concentration in the boiler water needs to be maintained within a set range to inhibit scaling and corrosion of the metal tube walls. Therefore, phosphate reagents must be continuously added to the tubes via a reagent injection actuator. The reagents are transported with the steam-water mixture and undergo a chemical reaction within the furnace. Changes in phosphate concentration are then detected at sampling points by online instruments. Traditional reagent addition control methods typically employ a feedback regulation strategy based on real-time concentration detection. The controller periodically calculates the dosage based on the deviation between the current phosphate concentration and the set value, driving the injection actuator to adjust accordingly.

[0003] However, significant time lags exist between reagent injection and concentration detection response, including tube-side transport lag, chemical reaction lag, and instrument detection lag, which combine to create a large time delay characteristic. Furthermore, external disturbances such as flow velocity fluctuations within the tube, unit load changes, and periodic blowdowns can cause nonlinear changes in the phosphate concentration response amplitude under the same dosage, further exacerbating the control difficulty. Therefore, traditional feedback control methods struggle to achieve precise adjustment, leading to incorrect dosing timing, large injection volume deviations, frequent concentration fluctuations, and a significant decrease in control accuracy and system robustness, failing to meet the stringent requirements of the high-pressure steam system for long-term boiler water quality stability in cracking furnaces. Summary of the Invention

[0004] This application provides a precise control method and system for drug addition based on drug injection timing prediction. It improves the technical problem in the prior art where traditional control methods are difficult to comprehensively evaluate the effects of hysteresis uncertainty, response nonlinearity and disturbance in drug injection scenarios with large time delay, nonlinearity and multiple disturbances. This makes it impossible to achieve dynamic and accurate prediction of injection timing and dosage, resulting in adjustment hysteresis and reduced robustness.

[0005] This application discloses the following technical solution:

[0006] In a first aspect, this application provides a method for precise control of drug addition based on prediction of drug injection timing, the method comprising:

[0007] Obtain real-time operating status parameters of the pyrolysis furnace dosing system;

[0008] Based on the time difference between the action time of the drug injection actuator, the real-time flow rate of the fluid in the tube, and the response time of the phosphate concentration, the actual comprehensive lag time is calculated, and the lag uncertainty is analyzed and obtained.

[0009] The nonlinear jump rate is calculated based on the actual response amplitude between the change in drug injection volume and the change in phosphate concentration during the current sampling period.

[0010] The external disturbance component is separated from the change in phosphate concentration in the current sampling period, and the operating condition disturbance increment is calculated by combining the phosphate concentration value.

[0011] The hysteresis uncertainty, the nonlinear jump rate, and the operating condition disturbance increment are linked together to calculate the pre-judgment dosing coefficient.

[0012] Based on the pre-judgment dosing coefficient, the injection timing and injection volume of the drug are dynamically predicted and compensated, and a prediction dosing control command is generated and output to the drug injection actuator for execution.

[0013] Secondly, this application provides a precise drug addition control system based on drug injection timing prediction, the system comprising:

[0014] The status acquisition module is used to acquire real-time operating status parameters of the pyrolysis furnace dosing system;

[0015] The hysteresis analysis module is used to calculate the actual overall hysteresis time and analyze and obtain the hysteresis uncertainty based on the time difference between the action time of the drug injection actuator, the real-time flow rate of the fluid in the tube, and the response time of the phosphate concentration.

[0016] The jump analysis module is used to calculate the nonlinear jump rate based on the actual response amplitude between the change in drug injection volume and the change in phosphate concentration during the current sampling period.

[0017] The disturbance separation module is used to separate and obtain the external disturbance component from the change in phosphate concentration in the current sampling period, and calculate the operating condition disturbance increment by combining the phosphate concentration value.

[0018] The coefficient generation module is used to perform a linkage calculation on the hysteresis uncertainty, the nonlinear jump rate and the operating condition disturbance increment to calculate the pre-judgment dosing coefficient.

[0019] The prediction and compensation module is used to dynamically predict and compensate the injection timing and injection volume of the agent based on the pre-prediction dosing coefficient, generate prediction dosing control instructions, and output them to the agent injection actuator for execution.

[0020] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0021] The technical solution of this application provides a precise control method for reagent addition based on reagent injection timing prediction. First, by acquiring the real-time operating status parameters of the cracking furnace dosing system, it synchronously collects multi-dimensional data such as the time of reagent injection, the real-time flow rate of the fluid in the tube, and the response time of the phosphate concentration in the boiler water, and establishes a unified time-aligned data benchmark. This provides complete and synchronous input information for subsequent timing prediction, and solves the problem of weak data foundation caused by incomplete acquisition of operating parameters and time asynchrony in traditional control.

[0022] Furthermore, by calculating the actual integrated lag time and comparing it with the average integrated lag time under the same historical operating conditions, the lag uncertainty is obtained. At the same time, the integrated lag is decomposed into pipe-pass transmission lag and chemical response lag, realizing the dynamic evaluation of time-varying lag characteristics and the quantification of its fluctuation degree, solving the problem that the fixed lag assumption cannot accurately reflect the large-scale variation of the actual lag time.

[0023] Furthermore, by calculating the actual response gain of the current sampling period and comparing it with the linear response gain of the same concentration range in history, the nonlinear jump rate is obtained, which quantifies the degree to which the dosing response deviates from the linear law online, and solves the problem that fixed control parameters cannot adapt to the dynamic changes in dosing response characteristics.

[0024] Furthermore, the pure dosing response component was separated from the change in phosphate concentration by the dosing injection response prediction model. After hysteresis confidence correction, the external disturbance component and the operating condition disturbance increment were obtained, realizing the quantitative separation and normalized evaluation of the influence of external disturbances. This solved the problem that the source of concentration change could not be distinguished and the degree of disturbance was difficult to quantify due to the coupling of dosing effect and external disturbance.

[0025] Furthermore, by performing a linked calculation of multiplication and summation of hysteresis uncertainty, nonlinear jump rate, and operating condition disturbance increment, a pre-judgment dosing coefficient is generated. This comprehensively characterizes the overall impact intensity of the mutual aggravation of hysteresis, nonlinearity, and disturbance on dosing control, solving the problem that single-factor assessment cannot reflect the deteriorating effect of multi-factor coupling and lacks a unified and comprehensive adjustment basis.

[0026] Finally, by dynamically adjusting the advance amount of dosing and the injection amount based on the pre-judgment dosing coefficient, and using a nonlinear jump rate to suppress the dosing amount, the dosing parameters are frozen when the risk is too high, and a pre-judgment dosing control command is generated. This achieves precise advance dosing and safe control under large time delay conditions, and solves the problems of traditional feedback control, such as lagging adjustment action, easy overshoot, and lack of extreme condition protection mechanism.

[0027] In summary, the technical solution of this application realizes real-time parameter acquisition, lag delay decomposition and uncertainty quantification, nonlinear response evaluation, external disturbance separation, multi-factor linkage coefficient generation, and dynamic prediction and compensation of dosing timing and injection volume for the large time-delay reagent injection process of the pyrolysis furnace dosing system. By unifying time reference acquisition, time delay decomposition and deviation comparison, gain deviation calculation, prediction model disturbance separation, product summation linkage operation, and dynamic correction and freeze protection of dosing volume, it effectively improves the technical problems in the prior art where dosing timing depends on lag feedback, injection volume lacks dynamic prediction and compensation, and safety protection mechanism is insufficient, resulting in adjustment lag, high overshoot risk, and decreased control accuracy and robustness. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 A schematic diagram of the overall process of the precise control method for drug addition based on drug injection timing prediction provided in the embodiments of this application;

[0030] Figure 2 A flowchart illustrating the predictive dosing control process of the precise dosing control method based on predictive dosing timing provided in this application embodiment;

[0031] Figure 3 This is a schematic diagram of the structure of a precise drug addition control system based on drug injection timing prediction provided in an embodiment of this application.

[0032] In the attached diagram, Figure 3 The components represented by each number are described as follows: Status acquisition module 11, hysteresis analysis module 12, sudden jump analysis module 13, disturbance separation module 14, coefficient generation module 15, and prediction compensation module 16. Detailed Implementation

[0033] This application provides a precise control method and system for drug addition based on drug injection timing prediction, which is used to solve the technical problem that in the existing technology, due to the time-varying lag, nonlinear response characteristics and the coupling effect of external disturbance events, traditional feedback adjustment mechanisms cannot provide timely and accurate prediction compensation, resulting in a decrease in control accuracy and system robustness in the drug addition process with large time delay.

[0034] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. It should be noted that the numerical values ​​in the embodiments are for illustrative purposes only and do not constitute a limitation on this application.

[0035] Example 1, as shown in the appendix Figure 1 As shown, this application provides a method for precise control of drug addition based on drug injection timing prediction, the method comprising the following steps:

[0036] S100: Obtain the real-time operating status parameters of the pyrolysis furnace dosing system;

[0037] In this embodiment of the application, in the drug injection scenario with large time delay, nonlinearity and multiple disturbances, in order to provide an accurate data foundation for subsequent timing prediction, this step needs to simultaneously acquire multi-dimensional real-time operating parameters such as the drug injection action time, tube flow rate and concentration response.

[0038] Step S100 in the method provided in this application embodiment includes:

[0039] Real-time operating data of the high-pressure steam system of the pyrolysis furnace is collected to obtain the actuation time and injection volume of the reagent injection actuator; the real-time flow velocity of the fluid in the tube is also collected. Detailed explanation follows:

[0040] In this embodiment, the pyrolysis furnace dosing system refers to a reagent dosing control device used in the high-pressure steam system of the pyrolysis furnace, including a reagent storage tank, an injection pump, and an actuator, used to inject phosphate reagents into the tube side to adjust water quality. The reagent injection actuator is a terminal device that receives control commands and executes the reagent injection action, typically an electrically controlled injection pump or regulating valve; its activation time refers to the point at which the actuator begins injecting the reagent. The fluid in the tube side refers to the steam-water mixture flowing in the tube side of the high-pressure steam system of the pyrolysis furnace; its flow velocity directly affects the transport time of the reagent from the injection point to the detection point. Real-time flow velocity refers to the instantaneous flow velocity of the fluid in the tube side measured by an online flow meter, in m / s.

[0041] In this step, firstly, to obtain the accurate time and dosage of the reagent injection, real-time operating data of the high-pressure steam system of the pyrolysis furnace needs to be collected. This involves recording the action time and injection volume of the reagent injection actuator, which solves the problem of time lag or omissions in traditional manual recording. For example, the distributed control system records the injection pump start time as 08:30:00.200 and the single injection volume as 0.5L. Further, to calculate the transport lag of the reagent within the tube side, the real-time flow velocity of the fluid within the tube side needs to be collected, providing dynamic input for the subsequent time delay model. For example, the real-time flow velocity is displayed as 2.5 m / s.

[0042] For example, taking the chemical dosing control of a cracking furnace as an example, operational data is collected in real time through a distributed control system. The chemical injection actuator starts at 09:15:00.050, with a single injection volume of 0.6L. Simultaneously, the flow meter at the tube inlet measures a real-time flow velocity of 2.8m / s. The data acquisition equipment also simultaneously obtains the current boiler water phosphate concentration of 2.0mg / L and the blowdown operation record, etc. All parameters are accompanied by precise timestamps, providing a unified benchmark for subsequent hysteresis analysis.

[0043] In summary, this step established a synchronous acquisition mechanism for multi-dimensional data by collecting real-time operating status parameters of the pyrolysis furnace dosing system. This provides necessary input information for subsequent lag time calculation, nonlinear analysis, and disturbance separation, ensuring the reliability and timeliness of predictive control data.

[0044] S200: Based on the time difference between the action time of the drug injection actuator, the real-time flow rate of the fluid in the tube, and the response time of the phosphate concentration, calculate the actual comprehensive lag time and analyze and obtain the lag uncertainty.

[0045] In this embodiment, under the agent injection scenario characterized by large time delays, nonlinearity, and multiple disturbances, the agent undergoes two stages from the injection point to the detection point: tube-side transport and chemical reaction. The duration of each stage dynamically changes with the operating conditions, and the delays of the two parts are difficult to observe independently, resulting in fluctuations in the actual overall delay duration. This step requires accurately calculating the actual overall delay duration and quantifying its uncertainty.

[0046] Step S200 in the method provided in this application embodiment includes:

[0047] The action time of the reagent injection actuator is obtained and recorded as the injection time; the time when the detection of the phosphate concentration in the boiler water begins to respond is obtained and recorded as the response time; the difference between the response time and the injection time is calculated as the actual comprehensive lag time; based on the real-time flow rate, the tube-side transport lag time is calculated; the difference between the actual comprehensive lag time and the tube-side transport lag time is calculated to obtain the chemical response lag time; the historical average comprehensive lag time of the pyrolysis furnace dosing system under historical normal operating conditions for the same flow rate range is obtained, and the deviation between the actual comprehensive lag time and the historical average comprehensive lag time is calculated as the lag uncertainty. Detailed explanation follows:

[0048] In this embodiment, the actual overall lag time refers to the total time elapsed from the action of the reagent injection actuator to the start of a identifiable change in the boiler water phosphate concentration detection value, measured in seconds (s). The tube-side transport lag time refers to the time required for the reagent to be transported from the injection point to the sampling detection point within the tube, calculated based on the tube length and real-time flow rate, measured in seconds. The chemical response lag time refers to the delay in chemical reaction and instrument response experienced after the reagent enters the boiler water until the phosphate concentration detection value begins to change, obtained by subtracting the tube-side transport lag time from the actual overall lag time, measured in seconds. The lag uncertainty refers to the normalized relative deviation between the current actual overall lag time and the historical average overall lag time under the same operating conditions, reflecting the degree to which the current lag deviates from the normal level, with a value ranging from 0 to 1; a larger value indicates higher uncertainty.

[0049] In this step, firstly, in order to obtain a complete lag time reference, it is necessary to record the action time of the reagent injection actuator and the start time of the response of the boiler water phosphate concentration, and calculate the difference between the two as the actual comprehensive lag time. The total delay of the reagent action within the current sampling period can be obtained. For example, if the injection time is 09:15:00.050 and the response time is 09:15:42.350, the actual comprehensive lag time is 42.3s.

[0050] Furthermore, in order to separate the physical lag of the tube path transmission, the tube path transmission lag time needs to be calculated based on the tube path length and the real-time flow velocity. For example, if the tube path length is 100m and the real-time flow velocity is 2.8m / s, the calculated tube path transmission lag time is 35.7s.

[0051] Furthermore, in order to obtain the chemical response lag time, the tube-side transmission lag time needs to be subtracted from the actual comprehensive lag time. For example, 42.3s minus 35.7s gives 6.6s, which is the total time of chemical reaction and instrument delay after the reagent enters the boiler water.

[0052] Finally, to quantify the uncertainty of lag fluctuations, it is necessary to look up the average comprehensive lag time of the same flow velocity range under historical normal operating conditions, and calculate the deviation between the actual comprehensive lag time and this average value as the lag uncertainty. For example, if the historical average comprehensive lag time of the same flow velocity range is 40.0s, the deviation is |42.3-40.0| / 40.0=0.058, that is, the lag uncertainty is 0.058, indicating that the current lag has a small deviation from the historical benchmark and is highly reliable.

[0053] For example, taking the chemical dosing control of a cracking furnace as an example, the chemical injection actuator was activated at 09:15:00.050 during the sampling period. The detection system recorded that the phosphate concentration in the boiler water first increased by more than 0.1 mg / L at 09:15:42.350, resulting in an actual overall lag time of 42.3 s. The real-time flow velocity in the tube was 2.8 m / s, the tube length was 100 m, and the tube-side transport lag time was calculated to be 35.7 s, thus yielding a chemical response lag time of 6.6 s. The average overall lag time under historical normal operating conditions for this flow velocity range was 40.0 s, and the calculated deviation yielded a lag uncertainty of 0.058.

[0054] In summary, this step, by calculating the pipe-side transport lag and chemical response lag separately, and then using the average lag under historical normal operating conditions as a benchmark to obtain the lag uncertainty, achieves the decomposition of the actual comprehensive lag duration and the quantitative assessment of the degree of fluctuation, providing dynamic delay and its reliability information for subsequent analysis.

[0055] S300: Calculate the nonlinear jump rate based on the actual response amplitude between the change in drug injection volume and the change in phosphate concentration during the current sampling period;

[0056] In this embodiment of the application, under the drug injection scenario characterized by large time delays, nonlinearity, and multiple perturbations, the relationship between changes in drug dosage and the response of phosphate concentration is not a constant ratio. Fluctuations in operating conditions cause the actual response amplitude to deviate from the historical linear pattern. It is necessary to quantify the degree to which the actual response between changes in drug dosage and concentration deviates from the linear pattern within the current sampling period.

[0057] Step S300 in the method provided in this application embodiment includes:

[0058] The system acquires the change in reagent injection rate and the change in phosphate concentration for the current sampling period, and obtains the actual response gain based on the changes in phosphate concentration and reagent injection rate. It also acquires the linear response gain of the pyrolysis furnace dosing system for the current phosphate concentration range in historical data, where the linear response gain is the arithmetic mean of the ratios of the changes in phosphate concentration to the changes in reagent injection rate within the same concentration range in the historical data. Finally, it calculates the deviation between the actual response gain and the linear response gain to obtain the nonlinear jump rate. A detailed explanation follows:

[0059] In this embodiment, the change in reagent injection volume refers to the difference between the cumulative injection volume of the reagent injection actuator in the current sampling period and the previous sampling period, in units of L. The change in phosphate concentration refers to the change in the detected value of boiler water phosphate concentration at the end of the current sampling period relative to the beginning of the period, in units of mg / L. The actual response gain refers to the ratio of the change in phosphate concentration to the change in reagent injection volume in the current sampling period, reflecting the actual change in concentration caused by a unit dosage. The linear response gain refers to the arithmetic mean of the ratios of the change in phosphate concentration to the change in reagent injection volume in all sampling periods within the current phosphate concentration range under historical normal operating conditions, representing the average linear response level of that concentration range. The nonlinear jump rate refers to the normalized deviation between the actual response gain and the linear response gain, ranging from 0 to 1; a larger value indicates a higher degree of deviation of the current dosing response from historical patterns.

[0060] In this step, firstly, in order to obtain the actual dosing response intensity of the current cycle, it is necessary to record the difference between the total amount of dosing injected in the current cycle and the previous cycle as the change in the amount of dosing injected, and simultaneously record the change in phosphate concentration at the beginning and end of the cycle. Calculate the ratio of the two to obtain the actual response gain, which can reflect the concentration response efficiency per unit dosing under the current operating conditions. For example, if the dosing amount in the current cycle increases by 0.1L compared to the previous cycle, the phosphate concentration increases by 0.3mg / L, and the actual response gain is 3.0(mg / L) / L.

[0061] Furthermore, in order to establish a reference benchmark for nonlinear deviation, it is necessary to retrieve all normal operating condition samples corresponding to the current phosphate concentration range from the historical database, and calculate the arithmetic mean of the ratio of concentration change to injection volume change in each sample as the linear response gain. For example, if the current concentration is in the range of 2.0 to 2.5 mg / L, the historical average ratio is 2.5 (mg / L) / L.

[0062] Finally, to quantify the nonlinearity of the current response, the absolute deviation between the actual response gain and the linear response gain needs to be calculated and divided by the linear response gain to obtain the normalized deviation as the nonlinear jump rate. For example, |3.0-2.5| / 2.5=0.2, which indicates that the current response deviates from the linear reference by 20%.

[0063] For example, taking the chemical dosing control of a cracking furnace as an example, the current sampling cycle is 45 seconds. At the beginning of the cycle, the phosphate concentration in the boiler water is 2.0 mg / L. During the cycle, the chemical injection actuator injects a cumulative 0.6 L, compared to 0.5 L in the previous cycle, resulting in a change of 0.1 L in the injected amount. At the end of the cycle, the phosphate concentration rises to 2.3 mg / L, with a change of 0.3 mg / L. The calculated actual response gain is 0.3 / 0.1 = 3.0 (mg / L) / L. There are 120 samples within the same concentration range of 2.0 to 2.5 mg / L in the historical data. The arithmetic mean of the gain ratio of each group is 2.5 (mg / L) / L. Therefore, the nonlinear jump rate is |3.0 - 2.5| / 2.5 = 0.2.

[0064] In summary, this step, by calculating the deviation between the actual response gain and the linear response gain of the historical concentration range, obtains the nonlinear jump rate, thereby achieving an online quantitative assessment of the nonlinearity of the current dosing response and providing a basis for adjusting the dosing control strategy.

[0065] S400: Separate and obtain the external disturbance component from the change in phosphate concentration in the current sampling period, and calculate the operating condition disturbance increment by combining the phosphate concentration value;

[0066] In this embodiment, based on the obtained hysteresis uncertainty and nonlinear jump rate, the change in phosphate concentration is not only caused by the injection of the reagent, but also affected by external events such as sewage discharge and load fluctuations, resulting in uncontrollable disturbance components in the concentration change. This step requires separating the pure dosing response component from the change in phosphate concentration to quantify the degree of influence of external disturbances on the concentration.

[0067] Step S400 in the method provided in this application embodiment includes:

[0068] A pre-constructed reagent injection response prediction model is obtained. The change in reagent injection volume during the current sampling period, the real-time flow rate, and the phosphate concentration value of the boiler water at the start of the current sampling period are input into the reagent injection response prediction model to obtain the reagent injection response component. The hysteresis uncertainty is obtained, and the hysteresis confidence is obtained based on the hysteresis uncertainty, wherein the hysteresis confidence is inversely proportional to the hysteresis uncertainty. The hysteresis confidence is used to correct the reagent injection response component to obtain the corrected injection component. Based on the corrected injection component, the external disturbance component is calculated, and the operating condition disturbance increment is obtained based on the ratio of the external disturbance component to the phosphate concentration value. Detailed explanation is as follows:

[0069] In this embodiment, the reagent injection response prediction model refers to a mapping model trained based on historical normal operating data. The inputs are the change in reagent injection rate, real-time flow rate, and initial boiler water phosphate concentration. The output is the phosphate concentration response component caused solely by reagent injection, used to eliminate the influence of external disturbances. The reagent injection response component refers to the theoretical change in phosphate concentration caused solely by the change in reagent injection rate within the current sampling period, calculated by the prediction model based on current operating parameters, and is expressed in mg / L.

[0070] The hysteresis confidence factor, derived from the transformation of hysteresis uncertainty, is used to correct the weights of the prediction model output. It ranges from 0 to 1, with lower hysteresis uncertainty indicating higher confidence. The corrected injection component, calculated by multiplying the drug injection response component by the hysteresis confidence factor, is used to reduce the weights of the prediction model output when hysteresis uncertainty is high. Its unit is mg / L. The operating condition disturbance increment is the ratio of the absolute value of the external disturbance component to the current phosphate concentration value, normalized after upper limit limiting. It ranges from 0 to 1 and is used to quantify the degree of interference of external events on the concentration.

[0071] In this step, firstly, in order to obtain the theoretical pure dosing response, it is necessary to call the pre-built dosing injection response prediction model, input the change in dosing amount, real-time flow rate and initial concentration of the current cycle, and obtain the dosing injection response component. For example, if the dosing amount is 0.1L, the flow rate is 2.8m / s and the initial concentration is 2.0mg / L, the model outputs a predicted response component of 0.28mg / L.

[0072] Furthermore, in order to assess the impact of lag on prediction reliability, lag confidence needs to be calculated. Lag confidence = 1 - lag uncertainty. For example, if the lag uncertainty is 0.058, then the lag confidence is 0.942.

[0073] Furthermore, in order to introduce hysteresis reliability correction into the prediction, the drug injection response component needs to be multiplied by the hysteresis reliability to obtain the corrected injection component, for example, 0.28 mg / L × 0.942 = 0.264 mg / L.

[0074] Finally, in order to separate external disturbances and calculate their impact, the external disturbance component is obtained by subtracting the correction injection component from the current periodic phosphate concentration change. The absolute value of the external disturbance component is then divided by the current concentration value to obtain the operating condition disturbance increment, with an upper limit of 1. For example, if the concentration change is 0.3 mg / L, the external disturbance component is 0.3 - 0.264 = 0.036 mg / L, and the operating condition disturbance increment is min(|0.036| / 2.3,1) = 0.016.

[0075] Step S400 in the method provided in this application embodiment further includes:

[0076] Multiple sets of operational data for the pyrolysis furnace dosing system under historical normal operating conditions are collected. Each set of data includes the change in reagent injection rate, the real-time flow rate of the fluid in the tube, the phosphate concentration of the boiler water at the start of the sampling period, and the phosphate concentration response value when no external disturbance occurs during the sampling period. These are used as sample data, where the phosphate concentration response value when no external disturbance occurs serves as the label for the reagent injection response component of that sampling period. An initial prediction model is constructed, taking the change in reagent injection rate, the real-time flow rate of the fluid in the tube, and the phosphate concentration of the boiler water at the start of the sampling period as inputs, and the reagent injection response component as output. The initial prediction model is trained using the sample data until the deviation between the output reagent injection response component and the reagent injection response component label meets a preset convergence condition, thus obtaining the completed reagent injection response prediction model. Detailed explanation follows:

[0077] In this embodiment, "historical normal operating conditions" refers to a period of time during which the pyrolysis furnace dosing system has not experienced external disturbances such as blowdown operations or load fluctuations, and the dosing response follows a linear pattern. "No external disturbances" means that no disturbances such as blowdowns, sudden load changes, or abnormal feedwater flow occurred within the corresponding sampling period, ensuring that changes in phosphate concentration are solely caused by reagent injection. The reagent injection response component label refers to the measured change in phosphate concentration within the sampling period without external disturbances, serving as the target output value for model training, expressed in mg / L. The initial prediction model refers to the reagent injection response prediction model constructed before training begins, without parameter optimization, typically employing a feedforward neural network structure. The preset convergence condition refers to the stopping criterion when the loss function value decreases by less than a set threshold or reaches the maximum number of iterations during training.

[0078] In this step, firstly, to construct the training dataset, sampling periods without external disturbances need to be selected from historical operating data. Since absolutely undisturbed operating conditions are not easily obtained directly in actual operation, periods where the impact of disturbances is deemed negligible can be assessed by experts, using operating logs such as sewage discharge records and load curves as undisturbed samples; alternatively, during the stable trial operation phase after system maintenance, undisturbed data can be collected in an experimental environment free from interference such as sewage discharge and load adjustments. The changes in reagent injection volume, real-time flow rate, and initial concentration for each period are extracted as inputs, and the changes in phosphate concentration are used as output labels to obtain the training sample set, for example, selecting 1500 sets of normal operating condition data.

[0079] Furthermore, in order to construct a prediction model, the model structure needs to be determined. The initial prediction model can adopt a multi-layer feedforward neural network, with 3 neurons in the input layer corresponding to 3 input parameters and 1 neuron in the output layer corresponding to the response component.

[0080] Finally, to complete model training, the sample set needs to be divided into a training set and a validation set. The model parameters are iteratively updated using an optimization algorithm until the deviation between the output and the label meets the convergence condition. For example, with 1200 training sets and 300 validation sets, the root mean square error of the model on the validation set after training is 0.02 mg / L.

[0081] For example, the method for constructing a drug injection response prediction model is as follows: A feedforward neural network is constructed as the initial prediction model. This network contains three fully connected hidden layers with 64, 32, and 16 neurons respectively. The activation function of the hidden layers is ReLU, and the activation function of the output layer is a linear function. The model input is a 3-dimensional feature vector, including the change in drug injection volume in L, the real-time flow velocity of the fluid in the pipe in m / s, and the phosphate concentration of the boiler water at the beginning of the sampling period in mg / L. The model output is a 1-dimensional predicted value, i.e., the drug injection response component in mg / L. The training sample set consists of 1500 sets of event samples without external disturbances selected from historical normal operating conditions, randomly divided into a training set of 1200 sets and a validation set of 300 sets in an 8:2 ratio. During model training, the mean squared error is used as the loss function, the optimizer is Adam, the learning rate is set to 0.001, the batch size is 32, the maximum number of iterations is 500, and an early stopping mechanism is set: training is terminated when the validation set loss does not decrease for 20 consecutive iterations. After 328 iterations, the validation set loss converged to 0.0004, and the model training was complete. The trained model parameters were then fixed and deployed into the online prediction process for the real-time output of the drug injection response component.

[0082] For example, taking the chemical dosing control of a cracking furnace as an example, the current sampling cycle is 45s, the initial phosphate concentration is 2.0mg / L, the change in reagent injection rate is 0.1L, and the real-time flow rate is 2.8m / s. Inputting these parameters into a pre-trained reagent injection response prediction model, the model outputs a reagent injection response component of 0.28mg / L. The lag uncertainty for this cycle is 0.058, the lag confidence is 0.942, and the corrected injection component is 0.264mg / L. At the end of the cycle, the measured phosphate concentration is 2.3mg / L, the concentration change is 0.3mg / L, and the external disturbance component is 0.3-0.264=0.036mg / L. The current concentration is 2.3mg / L, and the operating condition disturbance increment = 0.036 / 2.3≈0.016, which is less than 1, so no limiting is required. This small value indicates that the external disturbance has a minor impact on the concentration during the current cycle.

[0083] In summary, this step estimates the pure dosing response using a predictive model and subtracts it from the total change after hysteresis confidence correction to obtain the external disturbance component. Then, it normalizes the current concentration value to obtain the operating condition disturbance increment, thus achieving a quantitative assessment of the impact of external disturbances and providing disturbance dimension input information for subsequent multi-factor coupling analysis.

[0084] S500: Perform a linkage calculation on the hysteresis uncertainty, the nonlinear jump rate and the operating condition disturbance increment to calculate the pre-judgment dosing coefficient;

[0085] In this embodiment of the application, based on the obtained hysteresis uncertainty, nonlinear jump rate and operating condition disturbance increment, the quantitative indicators of the above three dimensions need to be linked and processed into a unified pre-judgment dosing coefficient to characterize the comprehensive intensity of the dosing prediction adjustment required under the current operating condition.

[0086] Step S500 in the method provided in this application embodiment includes:

[0087] The hysteresis uncertainty is multiplied by the nonlinear jump rate, and the resulting product is denoted as the hysteresis-nonlinear coupling term. The disturbance amplification factor is calculated based on the operating condition disturbance increment. The hysteresis-nonlinear coupling term is multiplied by the disturbance amplification factor to obtain the pre-judgment dosing coefficient. A detailed explanation follows:

[0088] In this embodiment, the lag-nonlinear coupling term refers to the product of lag uncertainty and nonlinear jump rate, used to quantify the amplification effect of dosing control risk caused by the superposition of large lag and nonlinear response. The larger the value, the higher the degree of dual deterioration of lag and nonlinearity. The disturbance amplification factor refers to the expansion coefficient after adding the operating condition disturbance increment to a base of 1, used to reflect the further amplification of the lag-nonlinear coupling effect by external disturbance events, and the value is not less than 1. The pre-judgment dosing coefficient refers to the comprehensive adjustment coefficient obtained by amplifying the lag-nonlinear coupling term by the disturbance amplification factor, used to uniformly adjust the dosing sequence and injection volume in subsequent steps. The larger the value, the stronger the pre-judgment compensation force required. The linkage operation refers to the operation process of coupling lag uncertainty, nonlinear jump rate, and operating condition disturbance increment into a single coefficient through a combination of product and summation, so as to reflect the overall effect of the intertwined influence of multiple factors.

[0089] In this step, firstly, to capture the mutually reinforcing effect between large lag and nonlinear response, the lag uncertainty is multiplied by the nonlinear jump rate to obtain the lag-nonlinear coupling term. This addresses the problem that single-factor assessment cannot reflect the superposition of dual deterioration. For example, when the lag uncertainty is 0.058 and the nonlinear jump rate is 0.2, the lag-nonlinear coupling term is 0.058 × 0.2 = 0.0116. Furthermore, to incorporate the impact of external disturbances into the comprehensive judgment, the disturbance amplification factor needs to be calculated based on the operating condition disturbance increment, using the form 1 + operating condition disturbance increment. This ensures that the larger the external disturbance, the stronger the amplification effect. For example, when the operating condition disturbance increment is 0.016, the disturbance amplification factor is 1 + 0.016 = 1.016. Finally, in order to obtain the final prediction adjustment coefficient, the lag-nonlinear coupling term needs to be multiplied by the disturbance amplification factor to obtain the pre-prediction dosing coefficient. This coefficient integrates information from three aspects: lag uncertainty, nonlinear response degree and external disturbance intensity. For example, 0.0116×1.016≈0.0118 indicates that the overall risk is low under the current operating conditions and the dosing prediction does not require significant intervention.

[0090] For example, taking the chemical dosing control of a cracking furnace as an example, in the current sampling period, the calculated hysteresis uncertainty is 0.058, the nonlinear jump rate is 0.2, and the operating condition disturbance increment is 0.016. Multiplying the hysteresis uncertainty by the nonlinear jump rate yields the hysteresis-nonlinear coupling term 0.0116; the disturbance amplification factor of 1.016 is calculated based on the operating condition disturbance increment; and the two are then multiplied to obtain the pre-judgment dosing coefficient 0.0118.

[0091] In summary, this step, through the linked calculation of multiplication and summation of hysteresis uncertainty, nonlinear jump rate, and operating condition disturbance increment, obtains the pre-judgment dosing coefficient, realizing the quantitative integration of the coupled influence of large hysteresis, nonlinear response, and external disturbance, and providing a unified adjustment basis for the dynamic prediction and compensation of subsequent dosing timing and injection volume.

[0092] S600: Based on the pre-judgment dosing coefficient, dynamically predict and compensate the injection timing and injection volume of the agent, generate a pre-judgment dosing control command, and output it to the agent injection actuator for execution.

[0093] In this embodiment of the application, based on the pre-judgment dosing coefficient that comprehensively reflects lag, nonlinearity and disturbance, the coefficient needs to be specifically converted into a pre-judgment compensation amount for dosing timing and injection volume, and dosing safety is ensured under abnormal operating conditions, generating an executable pre-judgment dosing control command.

[0094] Step S600 in the method provided in this application embodiment includes:

[0095] The preset standard dosing timing advance amount is multiplied by the pre-judged dosing coefficient, and the product is taken as the actual dosing timing advance amount. The actual dosing timing advance amount is the advance of the execution time of the current chemical injection command relative to the execution time that should have been based on the current boiler water phosphate concentration. Based on the pre-judged dosing coefficient and the preset standard dosing amount, the initial actual dosing amount is calculated. The initial actual dosing amount is corrected using the nonlinear jump rate to obtain the actual dosing amount. The actual dosing timing advance amount and the actual dosing amount are combined into a pre-judged dosing control command, which is output to the chemical injection actuator for execution. Detailed explanation follows:

[0096] In this embodiment, the standard dosing timing advance refers to the baseline advance time of the dosing command execution relative to the detection concentration trigger time, pre-set under ideal operating conditions with no lag, no nonlinearity, and no external disturbances, to prevent concentration overshoot; the unit is seconds (s). The actual dosing timing advance refers to the actual advance execution time of the dosing command after adjustment based on the comprehensive risk of the current operating conditions; it is obtained by multiplying the standard advance by the pre-judged dosing coefficient, with a larger advance for higher risks. The standard dosing amount refers to the theoretical dosage calculated based on the phosphate concentration deviation under ideal operating conditions; the unit is liters (L). The initial actual dosing amount refers to the dosing amount after preliminary adjustment by the pre-judged dosing coefficient; it is calculated by multiplying the standard dosing amount by 1 and subtracting the pre-judged dosing coefficient. The actual dosing amount refers to the final dosing amount after secondary suppression of the initial actual dosing amount using a nonlinear jump rate; this is achieved by dividing the initial value by 1 and adding the nonlinear jump rate to avoid excessive dosing during nonlinear jumps.

[0097] In this step, firstly, to reflect the comprehensive risk coefficient in the dosing timing, the standard dosing timing advance amount needs to be multiplied by the pre-judgment dosing coefficient to obtain the actual dosing timing advance amount. The greater the risk, the earlier the command is advanced. For example, if the standard advance amount is 5.0s and the pre-judgment dosing coefficient is 0.0118, the calculated actual advance amount is 0.059s. Considering the minimum action cycle of the actuator and the timing control accuracy, when the calculated actual advance amount is less than 0.1s, the advance amount can be rounded to 0s, which means that no advance injection is needed for the current operating condition, avoiding unnecessary jitter caused by an excessively small advance amount. Therefore, the actual dosing timing advance amount is taken as 0s at this time.

[0098] Furthermore, in order to initially determine the dosage, the standard dosage needs to be reduced based on the pre-judged dosage coefficient to calculate the initial actual dosage. The coefficient is reduced by 1 to make the dosage more conservative. For example, if the standard dosage is 0.6L, the initial actual dosage = 0.6 × (1 - 0.0118) = 0.593L.

[0099] Furthermore, in order to suppress the overshoot risk caused by the nonlinear response, the initial actual dosage needs to be corrected twice by the nonlinear jump rate. The dosage is divided by 1 and then the nonlinear jump rate is added to suppress it. For example, when the nonlinear jump rate is 0.2, the actual dosage = 0.593 / (1+0.2) = 0.494L.

[0100] Finally, in order to generate an executable instruction, the actual dosing timing advance amount and the actual dosing amount need to be combined into a predictive dosing control instruction, which is then output to the drug injection actuator. For example, the generated instruction is "No need for advance injection, inject 0.494L", and is then sent to the injection pump for execution through the control system.

[0101] Step S600 in the method provided in this application embodiment further includes: when the pre-judgment dosing coefficient is greater than a preset dosing amount freezing threshold, or the nonlinear jump rate is greater than a preset nonlinear jump warning threshold, freezing both the actual dosing timing advance amount and the actual dosing amount in the current sampling period to the values ​​used in the previous sampling period; when the pre-judgment dosing coefficient is less than or equal to a preset reset threshold for M consecutive sampling periods, and the nonlinear jump rate is less than or equal to the nonlinear jump warning threshold, restoring the use of the actual dosing timing advance amount and the actual dosing amount, where M is a positive integer greater than 0. Detailed explanation follows:

[0102] In this embodiment, the dosing freeze threshold refers to the pre-judgment dosing coefficient threshold that triggers the freezing of dosing parameters. When the coefficient exceeds this threshold, it indicates that the overall risk is too high, and dynamic adjustment needs to be suspended to ensure safety. It is usually set to an empirical value between 0.3 and 0.5. The nonlinear jump warning threshold refers to the nonlinear jump rate threshold that triggers the freezing of dosing parameters. When the jump rate exceeds this threshold, it indicates that the system has entered a strong nonlinear region, and the dosing response is unpredictable, requiring a freeze command. It is usually set between 0.4 and 0.6. The reset threshold refers to the upper limit of the pre-judgment dosing coefficient that triggers the freeze recovery. Dynamic adjustment can only be resumed when the coefficient of the system is lower than this threshold for M consecutive cycles and the nonlinear jump rate is also lower than the warning threshold. It is usually set to a value lower than the freeze threshold to form hysteresis. M refers to the number of consecutive stable cycles required to trigger recovery. The recommended value is 3 to 5 to prevent frequent switching of control state due to random fluctuations.

[0103] In this step, firstly, to ensure safe dosing under extreme conditions, it is necessary to monitor the pre-judgment dosing coefficient and nonlinear jump rate in real time. When either indicator exceeds the corresponding threshold, the dosing timing shift and dosing amount for the current cycle are frozen, and the safety value of the previous cycle is used. For example, if the dosing amount freeze threshold is set to 0.3, and the current coefficient of 0.0118 is far below the threshold, no freezing is required. When the nonlinear jump rate exceeds the warning threshold of 0.5, freezing is also triggered.

[0104] Furthermore, in order to promptly unfreeze the system after the operating conditions stabilize, it is necessary to continuously monitor the changing trend of the coefficient. When the coefficient is below the reset threshold for M consecutive cycles and the jump rate is also below the warning threshold, the real-time calculated dosing parameters are restored. For example, if the reset threshold is set to 0.2 and M is 3, dynamic adjustment is restored after the coefficient is below 0.2 for 3 consecutive cycles and the jump rate is below 0.5.

[0105] The procedure for this step is as follows: Figure 2 As shown.

[0106] For example, taking the chemical dosing control of a cracking furnace as an example, the current sampling cycle pre-judgment dosing coefficient is 0.0118, the nonlinear jump rate is 0.2, the standard dosing timing advance is 5.0s, and the standard dosing quantity is 0.6L. The actual dosing timing advance = 5.0 × 0.0118 = 0.059s. Since the calculated advance of 0.059s is less than 0.1s, to avoid introducing invalid action jitter due to an excessively small advance, the actual dosing timing advance is rounded to 0s. The initial actual dosing quantity = 0.6 × (1 - 0.0118) = 0.593L, and the actual dosing quantity = 0.593 / (1 + 0.2) = 0.494L. The freeze threshold is set to 0.3, and the warning threshold is set to 0.5. The current coefficient and jump rate are both within the safe range, and the normal output pre-judgment dosing control command is given. The command content is: no need for advance injection, inject 0.494L. The output is sent to the drug injection actuator for execution.

[0107] In summary, this step dynamically adjusts the advance of dosing timing and injection volume by predicting the dosing coefficient in advance, and uses a nonlinear jump rate to suppress the dosing volume a second time. At the same time, a parameter freezing and reset mechanism is introduced under extreme conditions to generate safety instructions, realizing closed-loop control from comprehensive evaluation to specific execution instructions. This ensures the timeliness of dosing prediction and avoids the risk of concentration overshoot caused by excessive dosing.

[0108] The embodiments of this application, through the specific implementation methods described above, achieve the following technical effects:

[0109] This application proposes a precise control method for pesticide dosing based on predictive injection timing. First, operating parameters such as the injection time, real-time flow rate in the tubing, and phosphate concentration response time are acquired. Then, the actual combined lag time and lag uncertainty are calculated, and the nonlinear jump rate is obtained based on changes in injection volume and concentration. Next, external disturbance components are separated using a predictive model, and the increment of the operating condition disturbance is calculated. These three parameters are then linked together to generate a pre-predictive dosing coefficient. Finally, this coefficient is used to dynamically predict and compensate for the preset dosing timing and injection volume, generating actual dosing control commands, and freezing the dosing parameters when the risk is high. This method achieves precise proactive control under large time delays through predictive injection timing and multi-factor linkage compensation, eliminating the regulation lag problem caused by multi-stage lags and nonlinear disturbances.

[0110] Example 2, as shown in the appendix Figure 3 As shown, based on the inventive concept of the precise drug addition control method based on drug injection timing prediction provided in Embodiment 1, this application also provides a precise drug addition control system based on drug injection timing prediction, specifically including:

[0111] Status acquisition module 11 is used to acquire real-time operating status parameters of the pyrolysis furnace dosing system;

[0112] The hysteresis analysis module 12 is used to calculate the actual overall hysteresis time based on the time difference between the action time of the drug injection actuator, the real-time flow rate of the fluid in the tube, and the response time of the phosphate concentration, and to analyze and obtain the hysteresis uncertainty.

[0113] The jump analysis module 13 is used to calculate the nonlinear jump rate based on the actual response amplitude between the change in drug injection amount and the change in phosphate concentration in the current sampling period.

[0114] The disturbance separation module 14 is used to separate and obtain the external disturbance component from the change in phosphate concentration in the current sampling period, and calculate the operating condition disturbance increment in combination with the phosphate concentration value.

[0115] The coefficient generation module 15 is used to perform a linkage calculation on the hysteresis uncertainty, the nonlinear jump rate and the operating condition disturbance increment to calculate the pre-judgment dosing coefficient.

[0116] The prediction and compensation module 16 is used to dynamically predict and compensate the injection timing and injection amount of the agent based on the pre-prediction dosing coefficient, generate a prediction dosing control command, and output it to the agent injection actuator for execution.

[0117] In one embodiment, the status acquisition module 11 is also used to: acquire real-time operating data of the high-pressure steam system of the pyrolysis furnace, obtain the action time and injection volume of the reagent injection actuator, and acquire the real-time flow rate of the fluid in the tube.

[0118] In one embodiment, the hysteresis analysis module 12 is further configured to: obtain the action time of the reagent injection actuator, denoted as the injection time, and obtain the time when the phosphate concentration in the boiler water is detected to begin responding, denoted as the response time; calculate the difference between the response time and the injection time as the actual comprehensive hysteresis duration; calculate the tube-side transmission hysteresis duration based on the real-time flow rate; calculate the difference between the actual comprehensive hysteresis duration and the tube-side transmission hysteresis duration to obtain the chemical response hysteresis duration; obtain the historical average comprehensive hysteresis duration corresponding to the same flow rate range under historical normal operating conditions of the pyrolysis furnace dosing system, and calculate the deviation between the actual comprehensive hysteresis duration and the historical average comprehensive hysteresis duration as the hysteresis uncertainty.

[0119] In one embodiment, the jump analysis module 13 is further configured to: obtain the change in reagent injection amount and the change in phosphate concentration during the current sampling period, and obtain the actual response gain based on the change in phosphate concentration and the change in reagent injection amount; obtain the linear response gain of the pyrolysis furnace dosing system in the historical data corresponding to the current phosphate concentration range, wherein the linear response gain is the arithmetic mean of the ratios of the change in phosphate concentration to the change in reagent injection amount within the same concentration range in the historical data; calculate the deviation between the actual response gain and the linear response gain, and obtain the nonlinear jump rate.

[0120] In one embodiment, the disturbance separation module 14 is further configured to: acquire a pre-built reagent injection response prediction model; input the change in reagent injection amount, the real-time flow rate, and the phosphate concentration value of the boiler water at the start of the current sampling period into the reagent injection response prediction model to acquire a reagent injection response component; acquire the hysteresis uncertainty and acquire the hysteresis confidence based on the hysteresis uncertainty, wherein the hysteresis confidence is inversely proportional to the hysteresis uncertainty; correct the reagent injection response component using the hysteresis confidence to acquire a corrected injection component; calculate and acquire an external disturbance component based on the corrected injection component, and acquire the operating condition disturbance increment based on the ratio of the external disturbance component to the phosphate concentration value.

[0121] Furthermore, the disturbance separation module 14 is also used to: collect multiple sets of operating data of the pyrolysis furnace dosing system under historical normal operating conditions, wherein each set of data includes the change in reagent injection amount within a sampling period, the real-time flow velocity of the fluid in the tube, the phosphate concentration value of the boiler water at the beginning of the sampling period, and the phosphate concentration response value corresponding to the absence of external disturbance events within the sampling period, as sample data, wherein the phosphate concentration response value corresponding to the absence of external disturbance events is used as the reagent injection response component label for the sampling period; construct a reagent injection response prediction model, wherein the reagent injection response prediction model takes the change in reagent injection amount, the real-time flow velocity of the fluid in the tube, and the phosphate concentration value of the boiler water at the beginning of the sampling period as input, and the reagent injection response component as output; train the reagent injection response prediction model using the sample data until the deviation between the output reagent injection response component and the reagent injection response component label meets the preset convergence condition, thereby obtaining the constructed reagent injection response prediction model.

[0122] In one embodiment, the coefficient generation module 15 is further configured to: multiply the hysteresis uncertainty by the nonlinear jump rate, and record the resulting product as a hysteresis-nonlinear coupling term; calculate and obtain the disturbance amplification factor based on the operating condition disturbance increment; and multiply the hysteresis-nonlinear coupling term by the disturbance amplification factor to obtain the pre-judgment dosing coefficient.

[0123] In one embodiment, the prediction compensation module 16 is further configured to: multiply the preset standard dosing timing advance amount by the pre-prediction dosing coefficient, and use the resulting product as the actual dosing timing advance amount, wherein the actual dosing timing advance amount is the advance amount of the execution time of the current reagent injection command relative to the execution time that should have been based on the current boiler water phosphate concentration value; calculate and obtain the initial actual dosing amount based on the pre-prediction dosing coefficient and the preset standard dosing amount; correct the initial actual dosing amount using the nonlinear jump rate to obtain the actual dosing amount; combine the actual dosing timing advance amount and the actual dosing amount into a prediction dosing control command, and output it to the reagent injection actuator for execution.

[0124] Furthermore, the prediction compensation module 16 is also used to: when the pre-prediction dosing coefficient is greater than a preset dosing amount freezing threshold, or the nonlinear jump rate is greater than a preset nonlinear jump warning threshold, freeze the actual dosing timing advance amount and the actual dosing amount of the current sampling period to the values ​​used in the previous sampling period; when the pre-prediction dosing coefficient is less than or equal to a preset reset threshold for M consecutive sampling periods, and the nonlinear jump rate is less than or equal to the nonlinear jump warning threshold, restore the use of the actual dosing timing advance amount and the actual dosing amount, where M is a positive integer greater than 0.

[0125] The precise dosing control system based on predictive dosing injection timing provided in this application can achieve precise and proactive control management in scenarios with large time delays, such as high-pressure steam systems in pyrolysis furnaces and water quality regulation in industrial boilers. This management involves the acquisition of real-time operating parameters, the calculation of hysteresis uncertainty, nonlinear jump rate, and operating disturbance increments, the generation of pre-predicted dosing coefficients, and the dynamic prediction and compensation of dosing timing and injection volume, as well as the output of commands. It can be integrated into distributed control systems or dosing injection control workstations, effectively improving the dosing prediction accuracy under the influence of large time delays, nonlinearity, and multiple disturbances. It eliminates the regulation lag problem caused by lag in multiple stages, reduces the risk of overdosing, and provides structured data support including hysteresis characteristics, response nonlinearity, and disturbance effects for dosing process optimization and traceability. The specific control flow and identification details of this system are described in Example 1.

[0126] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

Claims

1. A precise control method for drug addition based on drug injection timing prediction, characterized in that, The method includes: Obtain real-time operating status parameters of the pyrolysis furnace dosing system; Based on the time difference between the action time of the drug injection actuator, the real-time flow rate of the fluid in the tube, and the response time of the phosphate concentration, the actual comprehensive lag time is calculated, and the lag uncertainty is analyzed and obtained. The nonlinear jump rate is calculated based on the actual response amplitude between the change in drug injection volume and the change in phosphate concentration during the current sampling period. The external disturbance component is separated from the change in phosphate concentration in the current sampling period, and the operating condition disturbance increment is calculated by combining the phosphate concentration value. The hysteresis uncertainty, the nonlinear jump rate, and the operating condition disturbance increment are linked together to calculate the pre-judgment dosing coefficient. Based on the pre-judgment dosing coefficient, the injection timing and injection volume of the drug are dynamically predicted and compensated, and a prediction dosing control command is generated and output to the drug injection actuator for execution.

2. The precise control method for drug addition based on drug injection timing prediction according to claim 1, characterized in that, The acquisition of real-time operating status parameters of the pyrolysis furnace dosing system includes: Collect real-time operating data of the high-pressure steam system of the pyrolysis furnace to obtain the action time and injection volume of the reagent injection actuator; The real-time flow rate of the fluid inside the tube is collected.

3. The precise control method for drug addition based on drug injection timing prediction according to claim 1, characterized in that, The actual overall lag time is calculated based on the time difference between the action time of the drug injection actuator, the real-time flow rate of the fluid in the tubing, and the response time of the phosphate concentration, and the lag uncertainty is analyzed and obtained, including: The moment when the agent injection actuator operates is recorded as the injection moment, and the moment when the detection of the phosphate concentration in the boiler water begins to occur is recorded as the response moment. The difference between the response time and the injection time is calculated as the actual overall lag time; Based on the real-time flow rate, calculate the pipe transfer lag time; Calculate the difference between the actual overall lag time and the tube-pass transmission lag time to obtain the chemical response lag time; The historical average comprehensive lag time of the pyrolysis furnace dosing system under the same flow rate range under historical normal operating conditions is obtained, and the deviation between the actual comprehensive lag time and the historical average comprehensive lag time is calculated as the lag uncertainty.

4. The precise control method for drug addition based on drug injection timing prediction according to claim 1, characterized in that, The nonlinear jump rate is calculated based on the actual response amplitude between the change in drug injection volume and the change in phosphate concentration during the current sampling period, including: Obtain the change in drug injection volume and the change in phosphate concentration during the current sampling period, and obtain the actual response gain based on the change in phosphate concentration and the change in drug injection volume; Obtain the linear response gain of the pyrolysis furnace dosing system in the historical data corresponding to the current phosphate concentration range, wherein the linear response gain is the arithmetic mean of the ratio of the change in phosphate concentration to the change in the amount of reagent injected within the same concentration range in the historical data; Calculate the deviation between the actual response gain and the linear response gain to obtain the nonlinear jump rate.

5. The precise control method for drug addition based on drug injection timing prediction according to claim 1, characterized in that, The external disturbance component is separated from the change in phosphate concentration during the current sampling period, and the operating condition disturbance increment is calculated by combining the phosphate concentration value, including: Obtain a pre-built reagent injection response prediction model, input the change in reagent injection amount in the current sampling period, the real-time flow rate, and the phosphate concentration value of the boiler water at the beginning of the current sampling period into the reagent injection response prediction model to obtain the reagent injection response components; The hysteresis uncertainty is obtained, and the hysteresis confidence is obtained based on the hysteresis uncertainty, wherein the hysteresis confidence is inversely proportional to the hysteresis uncertainty; The hysteresis confidence level is used to correct the drug injection response component to obtain the corrected injection component; Based on the corrected injection component, the external disturbance component is calculated and obtained, and based on the ratio of the external disturbance component to the phosphate concentration value, the operating condition disturbance increment is obtained.

6. The precise control method for drug addition based on drug injection timing prediction according to claim 5, characterized in that, The construction of the drug injection response prediction model includes: Multiple sets of operating data of the pyrolysis furnace dosing system under historical normal operating conditions are collected. Each set of data includes the change in reagent injection amount within a sampling period, the real-time flow rate of the fluid in the tube, the phosphate concentration value of the boiler water at the beginning of the sampling period, and the phosphate concentration response value corresponding to the absence of external disturbance events within the sampling period. These are used as sample data. The phosphate concentration response value corresponding to the absence of external disturbance events is used as the reagent injection response component label for the sampling period. A reagent injection response prediction model is constructed. The reagent injection response prediction model takes the change in reagent injection amount, the real-time flow rate of the fluid in the tube and the phosphate concentration of the boiler water at the beginning of the sampling period as inputs and the reagent injection response component as output. The sample data is used to train the drug injection response prediction model until the deviation between the output drug injection response component and the drug injection response component label meets the preset convergence condition, thus obtaining the completed drug injection response prediction model.

7. The precise control method for drug addition based on drug injection timing prediction according to claim 1, characterized in that, The step of performing a linked calculation on the hysteresis uncertainty, the nonlinear jump rate, and the operating condition disturbance increment to calculate the pre-judgment dosing coefficient includes: Multiplying the hysteresis uncertainty by the nonlinear jump rate, the resulting product is denoted as the hysteresis-nonlinear coupling term. The disturbance amplification factor is calculated based on the disturbance increment under the aforementioned operating conditions. Multiply the hysteresis-nonlinear coupling term by the perturbation amplification factor to obtain the pre-judgment dosing coefficient.

8. The precise control method for drug addition based on drug injection timing prediction according to claim 1, characterized in that, The process of dynamically predicting and compensating for the injection timing and volume of the drug based on the pre-predicted dosing coefficient, generating a predictive dosing control command, and outputting it to the drug injection actuator for execution includes: Multiply the preset standard dosing timing advance amount by the pre-judged dosing coefficient, and the resulting product is the actual dosing timing advance amount, wherein the actual dosing timing advance amount is the advance amount of the execution time of the current agent injection command relative to the execution time that should have been based on the current boiler water phosphate concentration value; Based on the pre-judged dosing coefficient and the preset standard dosing amount, the initial actual dosing amount is calculated and obtained. The initial actual dosage is corrected using the nonlinear jump rate to obtain the actual dosage. The actual drug dosing timing advance amount and the actual drug dosing amount are combined into a predictive drug dosing control command, which is then output to the drug injection actuator for execution.

9. The precise control method for drug addition based on drug injection timing prediction according to claim 8, characterized in that, When the pre-judgment dosing coefficient is greater than the preset dosing amount freezing threshold, or the nonlinear jump rate is greater than the preset nonlinear jump warning threshold, the actual dosing timing advance amount and the actual dosing amount in the current sampling period are both frozen to the values ​​used in the previous sampling period. When the pre-judgment dosing coefficient is less than or equal to the preset reset threshold for M consecutive sampling periods, and the nonlinear jump rate is less than or equal to the nonlinear jump warning threshold, the actual dosing timing advance amount and the actual dosing amount are restored, where M is a positive integer greater than 0.

10. A precise drug addition control system based on drug injection timing prediction, characterized in that, The system is used to execute the precise drug addition control method based on drug injection timing prediction as described in any one of claims 1-9, and the system includes: The status acquisition module is used to acquire real-time operating status parameters of the pyrolysis furnace dosing system; The hysteresis analysis module is used to calculate the actual overall hysteresis time and analyze and obtain the hysteresis uncertainty based on the time difference between the action time of the drug injection actuator, the real-time flow rate of the fluid in the tube, and the response time of the phosphate concentration. The jump analysis module is used to calculate the nonlinear jump rate based on the actual response amplitude between the change in drug injection volume and the change in phosphate concentration during the current sampling period. The disturbance separation module is used to separate and obtain the external disturbance component from the change in phosphate concentration in the current sampling period, and calculate the operating condition disturbance increment by combining the phosphate concentration value. The coefficient generation module is used to perform a linkage calculation on the hysteresis uncertainty, the nonlinear jump rate and the operating condition disturbance increment to calculate the pre-judgment dosing coefficient. The prediction and compensation module is used to dynamically predict and compensate the injection timing and injection volume of the agent based on the pre-prediction dosing coefficient, generate prediction dosing control instructions, and output them to the agent injection actuator for execution.

Images