Intelligent process control system for butane separation tower

Abstract

The invention discloses an intelligent process control system for a butane separation tower, belongs to the field of petrochemical process control, and aims to solve the problems of high energy consumption and weak interference resistance caused by poor loop collaboration and dependence on univariate adjustment in the control of the existing butane separation tower. The system comprises a sensing layer, a control layer and an execution layer, the sensing layer collects full-process physical parameters and material component data, the control layer achieves data processing, trend prediction and multivariable optimization adjustment through data collection processing, multivariable prediction control (MPC) and an intelligent coordination module, and the execution layer executes a flow adjustment instruction. Through multi-variable cooperation, dynamic energy consumption optimization and feed-forward adjustment, the product quality stability and the system anti-interference capability are improved, the energy consumption is reduced, the manual intervention is reduced, and accurate, efficient and low-consumption control of the butane separation tower is realized.

Description

Technical Field

[0001] This invention belongs to the field of petrochemical production technology, specifically relating to an intelligent process control system for a butane separation tower. Background Technology

[0002] Butane separation towers are key equipment in petrochemical alkylation units. Their core function is to separate isobutane, lighter components than isobutane, and n-butane from alkylation oil and other materials. The control effect of this separation tower directly affects product quality and production energy consumption.

[0003] The existing butane separation tower control scheme is mainly based on single-loop independent control, which includes: the raw material is fed into the separation tower under the control of the liquid level in the storage tank; the outlet temperature of the reboiler at the bottom of the tower is controlled by setting the flow rate of the heating medium, and the liquid level in the bottom of the tower is maintained stable by relying on the set alkylation oil extraction rate; an online quality analyzer is installed on the n-butane extraction line to adjust the extraction rate according to the isobutane concentration in n-butane; the distillation rate of the overhead product is set by the pressure at the top of the tower, the reflux tank is operated when full, and the opening of the reflux flow regulating valve is controlled by a fixed reflux ratio signal.

[0004] However, existing control schemes have significant drawbacks: First, each control loop operates independently, resulting in insufficient coordination between variables. When faced with complex operating conditions such as fluctuations in the reaction feed, asynchronous adjustments can easily lead to fluctuations in product quality (such as alkylation oil vapor pressure and n-butane purity). Second, control parameter adjustments rely on only a single variable. For example, the reboiler outlet temperature is adjusted based solely on temperature, without considering comprehensive information such as the composition of the top distillate and the concentration gradient within the column, which can easily lead to energy waste. The n-butane extraction is based solely on the isobutane concentration, without being linked to the top isobutane distillate rate and the bottom alkylation oil quality, which may result in low isobutane recovery or insufficient n-butane purity. Third, the reflux ratio is fixed, lacking a dynamic optimization mechanism based on the real-time separation status within the column. When the feed composition changes, it is difficult to balance separation efficiency and energy consumption.

[0005] The root cause of the above problems lies in the lack of a globally optimized control strategy in the existing system and the absence of a multi-variable collaborative adjustment mechanism, resulting in significant room for improvement in ensuring product quality stability and reducing energy consumption. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent process control system for butane separation towers to solve existing problems.

[0007] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0008] An intelligent process control system for a butane separation tower includes a sensing layer, a control layer, and an execution layer. The sensing layer is used to collect physical parameters and material composition data of the butane separation tower feed, internal components, and products throughout the entire process. The control layer receives data from the sensing layer, processes, predicts, and optimizes it to generate multi-variable coordinated adjustment commands. The execution layer receives adjustment commands from the control layer and executes flow regulation operations to control the separation process within the tower.

[0009] The sensing layer includes detection elements and online analysis instruments;

[0010] The control layer includes a data acquisition and processing module, a multivariate predictive control (MPC) module, and an intelligent coordination module;

[0011] The execution layer includes multiple sets of flow control valves.

[0012] In a further embodiment, the sensing elements of the intelligent process control system for the butane separation tower include at least a bottom temperature sensor, a bottom pressure sensor, a top temperature sensor, a top pressure sensor, a bottom level gauge, a reflux tank level gauge, a feed flow transmitter, an alkylate oil extraction flow transmitter, a n-butane extraction flow transmitter, and an isobutane distillation flow transmitter, used to collect real-time data on temperature, pressure, level, and flow rate of each material within the tower.

[0013] In a further embodiment, the online analytical instruments in the sensing layer of the intelligent process control system for the butane separation tower include at least an isobutane purity analyzer at the top of the tower, an isobutane concentration analyzer in the n-butane extraction pipeline, an alkylation oil vapor pressure detector at the bottom of the tower, and an online analyzer for C4 components in the feed, for real-time analysis of the isobutane purity at the top of the tower, the isobutane concentration in n-butane, the vapor pressure of the alkylation oil at the bottom of the tower, and the C4 component content in the feed.

[0014] In a further embodiment, the data acquisition and processing module of the control layer in the intelligent process control system for the butane separation tower receives data from the sensing layer and uses a Kalman filter algorithm to filter the data. The Kalman filter algorithm utilizes state equations... and Kalman gain equation K k =P k||k-1 H T HP k||k-1 H T +R) -1 Remove high-frequency interference signals from the data; among which, This is the state estimate updated by observation at time k. K is the predicted value at time k based on the state at time k-1. k Let z be the Kalman gain at time k. k Here, H represents the raw observation data collected by the sensing layer at time k, where H is the observation matrix and P is the data from the observation layer at time k.k||k-1 Let R be the covariance matrix of the predicted state at time k, and R be the observation noise covariance matrix. Simultaneously, the data acquisition and processing module removes abnormal data by comparing historical normal operating data with preset process thresholds, and then processes the data according to the formula... Parameters in different units are converted into 0-1 standardized data to adapt to subsequent module calculations; where x norm Here are the standardized parameter values, and x is the original collected parameter value. min x is the minimum allowable value for this parameter in the process. max This is the maximum allowable value for this parameter in the process.

[0015] In a further embodiment, the multivariate predictive control (MPC) module of the control layer in the intelligent process control system for the butane separation tower is used to establish a nonlinear regression mathematical model of multiple variables such as temperature, pressure, liquid level, and component concentration within the tower. This model is expressed using the formula... Construct; where y is the key quality indicator, w0 is the model constant term, n is the number of input variables, and w i Let x be the coefficient of the linear term of the i-th input variable. i For the i-th input variable, w ij Here, ε represents the coefficient of the interaction term between the i-th and j-th input variables, and ε represents the model error term. The module combines real-time and historical operational data to predict the changing trends of key quality indicators based on this model, while simultaneously identifying coupling interferences and compensating for them using the deviation formula Δu = K. p (y set -y pred )+K i ∫(y set -y pred The interference compensation amount is calculated using )dt; where Δu is the adjustment compensation amount for coupled interference, and K p y is the proportionality constant. set y is the target value for the key quality indicator. pred K represents the predicted value of the key quality indicator obtained based on a nonlinear regression model. i For the integral coefficient, ∫(y set -y pred )dt is the integral term of the deviation between the target value and the predicted value of the key quality indicator.

[0016] In a further embodiment, the intelligent coordination module of the control layer of the intelligent process control system for the butane separation tower receives the output results of the multivariate predictive control (MPC) module, combines them with the built-in energy consumption model of process constraints, and generates optimized adjustment instructions for reboiler heating, reflux ratio, n-butane extraction, and isobutane distillation to balance product quality and energy consumption.

[0017] In a further embodiment, the flow control valves in the execution layer of the intelligent process control system for the butane separation tower include at least a reboiler heating medium flow control valve, a reflux flow control valve, a n-butane extraction control valve, an isobutane distillation control valve, and a feed flow control valve, which are used to execute control layer commands and adjust the corresponding material flow rates to control the separation process.

[0018] In a further embodiment, the data acquisition and processing module of the intelligent process control system for the butane separation tower, when performing data verification, preset process thresholds are determined based on the statistical range of various parameters during normal operation of the butane separation tower. When real-time data exceeds... Data within the specified range is identified as abnormal and removed; among them, σ is the historical average of the parameter, and σ is the standard deviation of the parameter; during standardization, x min x max These are the minimum and maximum allowable values ​​for the corresponding parameters, set based on the butane separation tower design parameters and production process requirements.

[0019] In a further embodiment, the nonlinear regression mathematical model coefficients w0 and w... of the multivariate predictive control (MPC) module in the intelligent process control system for the butane separation tower... i w ij The data was obtained by fitting historical operating data using the least squares method, with the fitting objective being to achieve the sum of squares of the error terms ε. Minimum; where m is the number of historical data samples, The square of the error term for the k-th historical sample, y k Let x be the actual observed value of the key quality indicator in the k-th historical sample. ik Let x be the actual observed value of the i-th input variable in the k-th historical sample. jk This represents the actual observed value of the j-th input variable in the k-th historical sample.

[0020] In a further proposed solution, when the multivariable predictive control (MPC) module of the intelligent process control system for the butane separation tower calculates the disturbance compensation amount, the proportional coefficient K... p and integral coefficient K i Based on the dynamic response characteristics of the butane separation tower, the system response curve was obtained through a step test. Then, the overshoot and settling time indices of the response curve were optimized to ensure rapid compensation and adjustment without excessive fluctuations. Among these, K... p K is used to adjust the sensitivity of the deviation response. i Used to eliminate the cumulative effect of deviation, the values ​​of both are matched with the separation efficiency, material residence time and other characteristics of the butane separator.

[0021] The present invention has the following beneficial effects:

[0022] By constructing a three-layer architecture of "perception layer - control layer - execution layer," with multivariate predictive control and intelligent coordination as the core, the traditional single-loop independent control mode is replaced, effectively improving the stability of product quality in the butane separation tower and solving the problems of insufficient coordination among control loops and weak ability to cope with operating condition fluctuations in existing technologies. At the same time, by using dynamic energy consumption optimization algorithms and multi-parameter linkage adjustment, system energy consumption is reduced and separation efficiency is improved while ensuring product quality. In addition, through feedforward adjustment and dynamic disturbance compensation mechanisms, the system's anti-interference ability to fluctuations in feed flow rate and composition is enhanced, the operating condition recovery time is shortened, the frequency of manual intervention is reduced, and the operational intensity and probability of human error are lowered. Overall, precise, efficient, and low-consumption intelligent process control of the butane separation tower is achieved, significantly improving the economic benefits and operational reliability of the unit. Attached Figure Description

[0023] Figure 1 This is a block diagram of the overall structure of the present invention.

[0024] Figure 2 This is a logic diagram of each module of the present invention. Detailed Implementation

[0025] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0026] See Figures 1-2 As shown, an intelligent process control system for a butane separation tower includes a sensing layer, a control layer, and an execution layer. The sensing layer is used to collect physical parameters and material composition data of the butane separation tower feed, the tower interior, and the entire product process, providing data support for the control layer.

[0027] The control layer receives data from the sensing layer, processes, predicts, and optimizes it to generate multivariate collaborative regulation commands; the execution layer receives regulation commands from the control layer and executes flow regulation operations to control the separation process within the tower.

[0028] The sensing layer includes detection elements and online analysis instruments; the control layer includes a data acquisition and processing module, a multivariable predictive control (MPC) module, and an intelligent coordination module; and the execution layer includes multiple sets of flow control valves.

[0029] The detailed implementation of the system adopts a "distributed acquisition + centralized control" model. The sensing elements and online analytical instruments of the sensing layer are distributed and installed at key points in the butane separation tower, such as the feed pipeline, the bottom of the tower, the top of the tower, and the n-butane extraction pipeline. Data interaction is achieved with the control layer of the Zhejiang University Control System (ZJU Control) deployed in the central control room via industrial Ethernet (PROFINET protocol). The flow control valves of the execution layer receive commands from the control layer via 4-20mA analog signals, with a command transmission delay of no more than 50ms. In terms of hardware selection, the control layer server uses the ZJU Control JX-300XP system host, with an Intel Core i7-10700 CPU and 16GB of memory, ensuring sufficient computing power for parallel operation of multiple modules. Both the sensing layer and execution layer devices are explosion-proof (ExdIIBT4Ga), suitable for the explosion-proof environment of petrochemical workshops. The sensing layer is equipped with devices according to the principle of "full-process coverage." The number of detection elements is determined based on the tower size. For example, one temperature sensor and one pressure sensor are installed at the bottom and top of the tower, respectively, and one flow transmitter is installed on each of the key pipelines such as the feed and alkylate extraction lines. Online analysis instruments are installed in pipeline sections with stable material flow rates, such as the feed pipeline at a distance of three times the pump outlet diameter and the horizontal section of the n-butane extraction pipeline, to avoid flow field disturbances affecting analysis accuracy. The three modules of the control layer are integrated into the same DCS system platform, sharing data through an internal data bus without the need for additional communication links. The module operation priority is set as "MPC module > intelligent coordination module > data acquisition and processing module" to ensure that prediction and optimization tasks are executed first. The flow control valve in the execution layer is a pneumatic diaphragm control valve (model ZJHP) equipped with an intelligent valve positioner (accuracy ±0.1%), which can provide real-time feedback on the actual valve opening to the control layer, forming a local closed loop and avoiding adjustment deviations.

[0030] The sensing elements of the sensing layer include at least a bottom temperature sensor, a bottom pressure sensor, a top temperature sensor, a top pressure sensor, a bottom level gauge, a reflux tank level gauge, a feed flow transmitter, an alkylate oil extraction flow transmitter, a n-butane extraction flow transmitter, and an isobutane distillation flow transmitter, used to collect real-time data on temperature, pressure, level, and flow rate of each material within the column.

[0031] The bottom and top temperature sensors are model WZP-240, installed on the reboiler outlet line at the bottom and the condenser inlet line at the top, respectively. The insertion depth is 1 / 3 to 1 / 2 of the pipe's inner diameter, at a 45° angle to the pipe axis to avoid direct erosion of the sensor probe. After power-on, they are calibrated using a 0-200℃ standard temperature source, with an error ≤ ±0.1℃. The bottom and top pressure sensors are model PT201, installed on the side of the bottom end cap and the gas phase outlet line at the top, respectively. They are vertically installed with the sensor interface flush with the pipe to prevent liquid accumulation from affecting detection. They are calibrated using a 0-1MPa standard pressure source, with an error ≤ ±0.01MPa. The column bottom level gauge and reflux tank level gauge are UHZ-58 / C type, side-mounted on the side of the column bottom (level range 0-3m) and the reflux tank (level range 0-2m), respectively. The measurement range covers the normal level range (column bottom normal level 50% ± 10%, reflux tank normal level 60% ± 10%). Calibration is performed by adding water, and the error is ≤ ±1mm when compared with the reading of the manual glass plate level gauge. Each flow transmitter is an LDG-MIK type electromagnetic flow meter, installed on the horizontal sections of the feed, alkylate oil extraction, n-butane extraction, and isobutane distillation pipelines. The straight pipe sections before and after the installation are 10 times the pipe diameter (front) and 5 times the pipe diameter (rear), respectively, ensuring that the fluid flows through fully developed pipes, with a range of 0-100m. 3 The standard flow rate is calibrated to ≤±0.5% per hour. The output signal of the detection element is uniformly a 4-20mA analog signal. After common-mode interference is eliminated by an NPEXA-1300 signal isolator, it is connected to the DCS system AI module. Each AI module can connect to a maximum of 16 signals, with 20% reserved as spare channels for future expansion. Regarding the data acquisition cycle, temperature, pressure, and liquid level data are acquired once per second, and flow rate data is acquired once every 2 seconds. The acquired data is stored in the DCS system historical database in the format of "location number-parameter name-acquisition time-value-status" for one year, facilitating operational condition traceability and model optimization.

[0032] The online analytical instruments in the sensing layer include at least an isobutane purity analyzer at the top of the tower, an isobutane concentration analyzer in the n-butane extraction pipeline, an alkylation oil vapor pressure detector at the bottom of the tower, and an online analyzer for C4 components in the feed, which are used to analyze the isobutane purity at the top of the tower, the isobutane concentration in n-butane, the vapor pressure of the alkylation oil at the bottom of the tower, and the content of C4 components in the feed in real time.

[0033] The detailed implementation details are as follows: The isobutane purity analyzer at the top of the column and the isobutane concentration analyzer in the n-butane extraction line are both GC-2030 gas chromatographs (FID detector), installed on the isobutane distillation line (near the reflux tank inlet) and the n-butane extraction line (5 times the pipe diameter upstream of the regulating valve), respectively. Standard gases (isobutane purity 90%, 95%, 100%, isobutane concentration 1%, 1.5%, 2%) are introduced for calibration, with analytical errors ≤ ±0.5% and ±0.1%, respectively. The alkylation oil vapor pressure analyzer at the bottom of the column is a VP-3000 model, using the static headspace method, vertically installed on the alkylation oil extraction line at the bottom of the column, calibrated using standard oil samples (vapor pressure 44 kPa, 46 kPa, 48 kPa), with an error ≤ ±0.5 kPa. The online analyzer for C4 components in the feed is also a GC-2030 gas chromatograph, installed at a distance of 10 times the pump outlet diameter from the feed line. Calibration is performed using a standard gas containing 30% isobutane, 50% n-butane, and 20% light components, with an analysis error of ≤±0.5% for each component. Regarding the analysis cycle, the isobutane purity analyzer at the top of the column and the isobutane concentration analyzer in the n-butane extraction line operate at 2 minutes per analysis, with each analysis including 30 seconds of sampling, 20 seconds of vaporization, 40 seconds of separation, and 30 seconds of detection and data output. The alkylation oil vapor pressure detector at the bottom of the column and the online analyzer for C4 components in the feed operate at 1 minute per analysis, with a built-in sample cell simplifying the sampling and pretreatment process. When the online analysis instrument shows abnormalities such as "no signal" or "analysis result out of range", it automatically switches to the "historical average replacement" mode, using the average of the normal analysis results of the past 10 minutes as temporary data. At the same time, it triggers the audible and visual alarm in the central control room (red light flashing + buzzer). After the fault is cleared, real-time analysis is restored, and the replacement data during the fault period is marked for later verification.

[0034] The data acquisition and processing module of the control layer receives data from the perception layer and uses the Kalman filter algorithm to filter the data. The Kalman filter algorithm uses state equations... and Kalman gain equation K k =P k||k-1 H T HP k||k-1 H T +R) -1 Remove high-frequency interference signals from the data; among which, This is the state estimate updated by observation at time k. K is the predicted value at time k based on the state at time k-1. k Let z be the Kalman gain at time k. k Here, H represents the raw observation data collected by the sensing layer at time k, where H is the observation matrix and P is the data from the observation layer at time k. k||k-1Let R be the covariance matrix of the predicted state at time k, and R be the observation noise covariance matrix. Simultaneously, the data acquisition and processing module removes abnormal data by comparing historical normal operating data with preset process thresholds, and then processes the data according to the formula... Parameters in different units are converted into 0-1 standardized data to adapt to subsequent module calculations; where x norm Here are the standardized parameter values, and x is the original collected parameter value. min x is the minimum allowable value for this parameter in the process. max This is the maximum allowable value for this parameter in the process.

[0035] In the initial parameter configuration of the Kalman filter algorithm, the observation matrix H is set according to the type of detection element. For temperature sensors, H = 1 (the state variable directly corresponds to the observation variable), and for flow transmitters, H is calculated based on fluid density and pipeline cross-sectional area (H = ρ × A). The observation noise covariance matrix R is determined based on the sensor accuracy, such as R = (0.1℃) for temperature sensors. 2 Pressure sensor R = (0.01 MPa) 2 Predicted state covariance matrix P 0||0 The initial matrix is ​​set as an identity matrix, and it is automatically updated as data is acquired. The filtering effect is verified using bottom temperature data (1-second acquisition cycle). Before filtering, there were high-frequency fluctuations of ±0.3℃ due to equipment vibration; after filtering, the fluctuations decreased to ±0.05℃, and the lag time for tracking actual temperature changes was ≤1 second, meeting the model's requirements for data stability. During data verification, preset process thresholds were generated statistically from the DCS system's historical database. Normal operating data from the past 3 months (sample size ≥10,000 sets) were selected, and the historical average values ​​of each parameter were calculated. With standard deviation σ, the threshold range is set as follows: For example, the temperature at the bottom of the tower σ = 1℃, threshold range 92-98℃. Data with a real-time temperature of 99℃ is considered outlier and removed, replaced by filtered data from the previous time step. (x is the standardization factor for the data.) min x max Based on the butane separation tower design documents and production process specifications, such as the bottom temperature in °C, the value is determined. min =90℃, ℃x max =100℃, feed flow rate 3 x min =50m 3 / h、 3 x max =150m 3 The standardization process is automatically executed in the DCS system's computing module, and the results are stored in a real-time database for use by the MPC module, with a call latency of ≤100ms.

[0036] The multivariate predictive control (MPC) module in the control layer is used to establish a nonlinear regression mathematical model for multiple variables such as temperature, pressure, liquid level, and component concentration within the tower. This model is expressed using the formula... Construct; where y is the key quality indicator, w0 is the model constant term, n is the number of input variables, and w i Let x be the coefficient of the linear term of the i-th input variable. i For the i-th input variable, w ij Here, ε represents the coefficient of the interaction term between the i-th and j-th input variables, and ε represents the model error term. The module combines real-time and historical operational data to predict the changing trends of key quality indicators based on this model, while simultaneously identifying coupling interferences and compensating for them using the deviation formula Δu = K. p (y set -y pred )+K i ∫(y set -y pred The interference compensation amount is calculated using )dt; where Δu is the adjustment compensation amount for coupled interference, and K p y is the proportionality constant. set y is the target value for the key quality indicator. pred K represents the predicted value of the key quality indicator obtained based on a nonlinear regression model. i For the integral coefficient, ∫(y set -y pred )dt is the integral term of the deviation between the target value and the predicted value of the key quality indicator.

[0037] A nonlinear regression model was constructed using six input variables (feed flow rate, feed isobutane content, reboiler heating amount, reflux ratio, n-butane extraction rate, and top pressure) and two output variables (top isobutane purity and bottom alkylation oil vapor pressure). The least squares method was used to fit the coefficients w0 and w2 based on normal operating data (sample size m = 5000 sets) in MATLAB software. i w ij The calculation of the sum of squared error terms is achieved by writing a script, and the coefficient of determination of the fitted model is obtained. 2 R 2 ≥0.95 (Isobutane purity model at the top of the column) 2 R 2 =0.96, vapor pressure model 2R of alkylation oil at the bottom of the tower 2 =0.97). The model refits coefficients every 7 days based on the latest 1000 sets of normal operating condition data. The update process is executed automatically. During the update, the MPC module temporarily uses the old model for prediction, and seamlessly switches over after the update is complete. The proportional coefficient K for interference compensation. p With integral coefficient K iThe parameters, such as the isobutane purity adjustment K, were determined by step testing. This was achieved by applying a +5% step disturbance to the reboiler heating rate and recording the response process (overshoot and settling time) of the isobutane purity at the top of the column. The Ziegler-Nichols tuning method was then used to obtain these parameters. p =2.5, K i =0.1, K is used to regulate the vapor pressure of the alkylation oil at the bottom of the tower. p =3.0, K i =0.08. Simulating the disturbance of a step increase in feed isobutane content from 30% to 35%, the predicted isobutane purity at the top of the column dropped from 90.5% to 88.2% without compensation. The MPC module calculated the compensation amount (increasing the reflux ratio by 0.1 and the reboiler heating amount by 2%). After execution, the actual purity stabilized at 89.8%-90.2%, with a deviation from the target value (90%) of ≤0.2%, demonstrating a significant compensation effect.

[0038] The intelligent coordination module of the control layer receives the output results of the multivariate predictive control (MPC) module, combines them with the built-in energy consumption model of process constraints, and generates optimized adjustment instructions for reboiler heating, reflux ratio, n-butane extraction, and isobutane distillation to balance product quality and energy consumption.

[0039] The energy consumption model built into the intelligent coordination module uses "unit product energy consumption (kJ / t)" as the objective function, expressed as E = a × Q + b × L, where E is the unit product energy consumption, Q is the reboiler heating capacity (kJ / h), and L is the return flow rate (m³ / h). 3 / h), a is the energy consumption coefficient of the heating medium (steam a=0.86kJ / (kJ·t)), b is the energy consumption coefficient of the reflux pump ( 3 b = 2.5 kJ / (m 3 Parameters a and b are calibrated monthly based on actual on-site energy consumption data (steam consumption, reflux pump power consumption, product output). Multi-objective optimization uses "product quality qualification (top isobutane purity ≥90%, bottom alkylation oil vapor pressure 44-48 kPa, isobutane concentration in n-butane ≤1.5%)" as constraints and "minimum E" as the objective. The optimal operating parameters are calculated using a particle swarm optimization (PSO) algorithm. The algorithm has 50 particles, 30 iterations, an inertia weight of 0.7, and acceleration coefficients c1 = c2 = 2.0. The optimization calculation time is ≤10 seconds. Generated adjustment commands are prioritized according to their "affected area" (reboiler heating command > reflux ratio command > n-butane extraction command > isobutane distillation command) to avoid simultaneous execution of multiple commands leading to operational chaos. For example, when adjusting heating and reflux ratio simultaneously, the heating command is executed first, followed by the reflux ratio command 5 seconds later. In addition, the instruction is set with limit protection: the single adjustment range of reboiler heating amount is ≤ ±5%, reflux ratio is ≤ ±0.1, and n-butane extraction amount is ≤ ±10%. When the limit is exceeded, the instruction is issued according to the limit value and an alarm is triggered.

[0040] The flow control valves in the execution layer include at least a reboiler heating medium flow control valve, a reflux flow control valve, a n-butane extraction control valve, an isobutane distillation control valve, and a feed flow control valve, which are used to execute control layer commands and adjust the corresponding material flow rates to control the separation process.

[0041] The reboiler heating medium flow regulating valve (DN50, PN1.6MPa) is installed on the reboiler heating steam inlet pipeline; the reflux flow regulating valve (DN80, PN1.0MPa) is installed on the reflux inlet pipeline from the reflux tank to the top of the column; the n-butane extraction regulating valve (DN65, PN1.0MPa) is installed on the n-butane extraction pipeline; the isobutane distillation regulating valve (DN80, PN1.0MPa) is installed on the top isobutane distillation pipeline; and the feed flow regulating valve (DN100, PN1.6MPa) is installed on the feed pipeline. All are ZJHP type pneumatic diaphragm regulating valves, equipped with intelligent valve positioners. During commissioning, with a flow rate of 0-100% of the rated flow, the valve opening and flow linearity error should be ≤±1%; the leakage when the valve is fully closed should be ≤0.01% of the rated flow; the response time should be ≤1 second (0% to 100% opening); the positioner accuracy should be ≤±0.1%; and the valve pressure drop loss at rated flow should be ≤0.1MPa. The valve control adopts a closed-loop opening mode. The 0-100% opening command issued by the control layer is converted into a 4-20mA current signal by the intelligent valve positioner, which drives the actuator. The positioner collects the actual valve opening in real time (built-in displacement sensor) and feeds it back to the control layer to ensure that the deviation between the actual opening and the commanded opening is ≤±0.5%. When the valve is "stuck" (deviation >5% and lasts for 10 seconds), the control layer automatically outputs a "valve purging" command (quickly opening and closing 3 times) to attempt to clear the obstruction; if purging is ineffective, an emergency alarm is triggered, and the system switches to manual control mode, where the operator can adjust the opening remotely or on-site.

[0042] When the data acquisition and processing module performs data verification, the preset process threshold is determined based on the statistical range of each parameter during normal operation of the butane separator. If the real-time data exceeds this range... Data within the specified range is identified as abnormal and removed; among them, σ is the historical average of the parameter, and σ is the standard deviation of the parameter; during standardization, x min x max These are the minimum and maximum allowable values ​​for the corresponding parameters, set based on the butane separation tower design parameters and production process requirements.

[0043] The process threshold is automatically calculated on the 1st of each month by a dedicated statistical script in the DCS system, which reads historical data from the past month (excluding fault and downtime periods). σ represents the arithmetic mean of the valid data, and σ is the sum of the valid data and the arithmetic mean of the valid data. The square root of the sum of squared deviations and the mean. When making temporary adjustments to the process (such as changes to product quality indicators), operators can manually modify the threshold. The system records the person, time, and reason for the modification, and automatically restores the statistical threshold after the adjustment is completed. Standardized parameter x min x max Determined by combining tower design documents and process specifications, such as feed flow rate x min 50% of the design flow rate (50m³) 3 / h), x max The design flow rate is 110% (110m³). 3 ( / h), ensuring parameters are within a safe and effective range. Use a feed flow rate of (50-110 m³ / h). 3 / h) Verify the standardization effect. The standardized values ​​are distributed in the 0-1 range and have a strict linear relationship with the original values ​​(correlation coefficient r=1). There is no data distortion, which meets the input requirements of the MPC module.

[0044] The coefficients w0 and w1 of the nonlinear regression mathematical model in the multivariate predictive control (MPC) module. i w ij The data was obtained by fitting historical operating data using the least squares method, with the fitting objective being to achieve the sum of squares of the error terms ε. Minimum; where m is the number of historical data samples, The square of the error term for the k-th historical sample, y k Let x be the actual observed value of the key quality indicator in the k-th historical sample. ik Let x be the actual observed value of the i-th input variable in the k-th historical sample. jk This represents the actual observed value of the j-th input variable in the k-th historical sample.

[0045] When filtering historical data, data from faulty or start-up / shutdown conditions are excluded, retaining only data from normal operating conditions (product quality qualified, parameter fluctuation ≤ ±1%). Anomaly periods are automatically identified and removed using a Python script. The filtered data then undergoes "missing value imputation" (linear interpolation) and "outlier removal" (removing values ​​exceeding the limit). The data was preprocessed and stored as a CSV file. Fitting was performed in MATLAB R2023a, using a least squares fitting script. The sum of squares of the error terms was defined, and the coefficients were solved using the fmincon function. Examples of coefficients include the isobutane purity model at the top of the column (w0 = 85.2) and the isobutane content in the feed (w). i =0.3, coefficient w of the interaction term between feed flow rate and reflux ratio ij =0.02. The accuracy was verified by leaving one out cross-validation. The 5000 samples were divided into 4999 training sets and 1 test set, and repeated 5000 times. The average prediction error of the isobutane purity model at the top of the tower was ≤±0.3%, and the vapor pressure model of the alkylation oil at the bottom of the tower was ≤±0.5kPa, which met the control requirements.

[0046] When the multivariable predictive control (MPC) module calculates the disturbance compensation, the proportional coefficient K... p and integral coefficient K i Based on the dynamic response characteristics of the butane separation tower, the system response curve was obtained through a step test. Then, the overshoot and settling time indices of the response curve were optimized to ensure rapid compensation and adjustment without excessive fluctuations. Among these, K... p K is used to adjust the sensitivity of the deviation response. i Used to eliminate the cumulative effect of deviation, the values ​​of both are matched with the separation efficiency, material residence time and other characteristics of the butane separator.

[0047] The step test was conducted during a stable operating period of the equipment (product quality qualified, parameter fluctuation ≤ ±1%). The system was switched to "manual control mode," and data was recorded in real time using LabVIEW software (sampling period 0.5 seconds). ±5% step disturbances were applied to the reboiler heating rate, reflux ratio, and other adjustment parameters. For example, a +5% disturbance in heating rate was recorded, and the response curves for the bottom temperature and top isobutane purity were recorded. Each disturbance was repeated three times, and the average value was used for analysis. The response curves were analyzed using MATLAB Control System Toolbox, and the overshoot (e.g., top isobutane purity overshoot = 5%) and settling time (e.g., bottom temperature settling time = 150 seconds) were calculated. The Cohen-Coon tuning method was used to optimize the coefficients. Considering the column characteristics, a longer material residence time (approximately 2 hours) resulted in a smaller K value. i (such as regulating the vapor pressure at the bottom of the tower K) i =0.08), to avoid excessive accumulation of integral terms; if the separation efficiency is high, increase K. p (such as adjusting the purity at the top of the column K) p =2.5), improving sensitivity. Verification shows that after tuning, the system disturbance overshoot is ≤8%, and the settling time is ≤200 seconds, meeting the stability control requirements.

[0048] The system operates on a closed-loop process of "sensing-processing-prediction-optimization-execution-feedback" to achieve intelligent control: Sensing layer detection elements and online analytical instruments collect physical parameters, composition, and quality index data, which are transmitted to the control layer via industrial Ethernet; the data acquisition and processing module uses Kalman filtering to remove interference, threshold comparison to eliminate anomalies, and standardized formulas to unify the data format; the MPC module calls standardized data and a nonlinear regression model to predict the trend of key quality indicators within 15 minutes, identify coupled interference, calculate compensation amounts, and output to the intelligent coordination module; the intelligent coordination module combines process constraints and energy consumption models, uses a particle swarm optimization algorithm to generate optimal adjustment commands, and issues them to the execution layer control valves according to priority; the control valves execute the commands and provide feedback on the opening degree; the sensing layer collects the adjusted data and sends it back to the control layer to update model coefficients and optimization strategies, forming a closed loop and continuously achieving precise, efficient, and low-consumption control of the butane separation tower.

[0049] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart process control system for a butane separation tower, characterized in that: It includes a sensing layer, a control layer, and an execution layer; the sensing layer is used to collect physical parameters and material composition data of the butane separation tower feed, the tower interior, and the entire product process; the control layer is used to receive data from the sensing layer, process, predict, and optimize it to generate multi-variable coordinated adjustment instructions; the execution layer is used to receive adjustment instructions from the control layer and execute flow regulation operations to control the separation process inside the tower. The sensing layer includes detection elements and online analysis instruments; The control layer includes a data acquisition and processing module, a multivariate prediction and control module, and an intelligent coordination module. The execution layer includes multiple sets of flow control valves.

2. The intelligent process control system for the butane separation tower according to claim 1, characterized in that, The sensing elements of the sensing layer include at least a bottom temperature sensor, a bottom pressure sensor, a top temperature sensor, a top pressure sensor, a bottom level gauge, a reflux tank level gauge, a feed flow transmitter, an alkylate oil extraction flow transmitter, a n-butane extraction flow transmitter, and an isobutane distillation flow transmitter, used to collect real-time data on temperature, pressure, level, and flow rate of each material within the tower.

3. The intelligent process control system for the butane separation tower according to claim 1, characterized in that, The online analytical instruments in the sensing layer include at least an isobutane purity analyzer at the top of the tower, an isobutane concentration analyzer in the n-butane extraction pipeline, an alkylation oil vapor pressure detector at the bottom of the tower, and an online analyzer for C4 components in the feed, used to analyze in real time the purity of isobutane at the top of the tower, the concentration of isobutane in n-butane, the vapor pressure of alkylation oil at the bottom of the tower, and the content of C4 components in the feed.

4. The intelligent process control system for the butane separation tower according to claim 1, characterized in that, The data acquisition and processing module of the control layer is used to receive data from the perception layer and to filter the data using a Kalman filter algorithm. The Kalman filter algorithm uses state equations... and Kalman gain equation K k =P k||k-1 H T HP k||k-1 H T +R) -1 Remove high-frequency interference signals from the data; in, This is the state estimate updated by observation at time k. K is the predicted value at time k based on the state at time k-1. k Let z be the Kalman gain at time k. k Here, H represents the raw observation data collected by the sensing layer at time k, where H is the observation matrix and P is the data from the observation layer at time k. k||k-1 Let R be the covariance matrix of the predicted state at time k, and R be the observation noise covariance matrix. Simultaneously, the data acquisition and processing module removes abnormal data by comparing historical normal operating data with preset process thresholds, and then processes the data according to the formula... Parameters in different units are converted into 0-1 standardized data to adapt to subsequent module calculations; where x norm Here are the standardized parameter values, and x is the original collected parameter value. min x is the minimum allowable value for this parameter in the process. max This is the maximum allowable value for this parameter in the process.

5. The intelligent process control system for the butane separation tower according to claim 1, characterized in that, The multivariate predictive control module of the control layer is used to establish a nonlinear regression mathematical model of multiple variables such as temperature, pressure, liquid level, and component concentration within the tower. This model is expressed using the formula... Build; Where y is the key quality indicator, w0 is the model constant term, n is the number of input variables, and w i Let x be the coefficient of the linear term of the i-th input variable. i For the i-th input variable, w ij ε represents the interaction term coefficient between the i-th and j-th input variables, and ε represents the model error term. The module combines real-time and historical operational data to predict the changing trends of key quality indicators based on the model, while also identifying coupling interferences and compensating for them using the deviation formula Δu=K. p (y set -y pred )+K i ∫(y set -y pred )dt calculates the interference compensation amount; Where Δu is the adjustment compensation amount for coupling interference, and K p y is the proportionality constant. set y is the target value for the key quality indicator. pred K represents the predicted value of the key quality indicator obtained based on a nonlinear regression model. i For the integral coefficient, ∫(y set -y pred )dt is the integral term of the deviation between the target value and the predicted value of the key quality indicator.

6. The intelligent process control system for the butane separation tower according to claim 1, characterized in that, The intelligent coordination module of the control layer is used to receive the output results of the multivariate predictive control module, and combine them with the built-in energy consumption model of process constraints to generate optimized adjustment instructions for reboiler heating, reflux ratio, n-butane extraction, and isobutane distillation, so as to balance product quality and energy consumption.

7. The intelligent process control system for the butane separation tower according to claim 1, characterized in that, The flow control valves of the execution layer include at least a reboiler heating medium flow control valve, a reflux flow control valve, a n-butane extraction control valve, an isobutane distillation control valve, and a feed flow control valve, which are used to execute control layer commands and adjust the corresponding material flow rates to control the separation process.

8. The intelligent process control system for the butane separation tower according to claim 4, characterized in that, When performing data verification, the data acquisition and processing module presets process thresholds based on the statistical range of various parameters during normal operation of the butane separation tower. If the real-time data exceeds... Data within the specified range is considered abnormal and removed; among them, σ is the historical average of the parameter, and σ is the standard deviation of the parameter; during standardization, x min x max These are the minimum and maximum allowable values ​​for the corresponding parameters, set based on the butane separation tower design parameters and production process requirements.

9. The intelligent process control system for the butane separation tower according to claim 5, characterized in that, The nonlinear regression mathematical model coefficients w0, w of the multivariate prediction and control module i w ij The data was obtained by fitting historical operating data using the least squares method, with the fitting objective being to achieve the sum of squares of the error terms ε. Minimum; Where m is the number of historical data samples. The square of the error term for the k-th historical sample, y k Let x be the actual observed value of the key quality indicator in the k-th historical sample. ik Let x be the actual observed value of the i-th input variable in the k-th historical sample. jk This represents the actual observed value of the j-th input variable in the k-th historical sample.

10. The intelligent process control system for the butane separation tower according to claim 5, characterized in that, When the multivariable predictive control module calculates the disturbance compensation amount, the proportional coefficient K... p and integral coefficient K i Based on the dynamic response characteristics of the butane separation tower, the system response curve is obtained through step test, and then optimized according to the overshoot and settling time indicators of the response curve to ensure that the compensation adjustment is fast and without excessive fluctuations. Among them, K p K is used to adjust the sensitivity of the deviation response. i The values ​​of both are used to eliminate the cumulative effect of deviations, and are matched with the separation efficiency, material residence time and other characteristics of the butane separation tower.

Images