Method for multivariable predictive process control

Abstract

The invention is used in the fields of petroleum refining and petrochemistry and is particularly applicable for controlling the processes of fractionation, reforming, hydrocracking, hydrorefining, and the like. The technical result consists in faster prediction of the behaviour of technical systems and more accurate control of multiparameter technological processes. In the event that one parameter changes, the influence of a change of input parameters on a change of output parameters is predicted, for which purpose transient process models are constructed, where the dependencies of each output parameter on each input parameter are determined by functions of the transient processes in the form of lag elements. A set of models of the influence of each input parameter on each output parameter is obtained. For each combination of models, the root mean square deviation of the parameters of the models from reference data is calculated and the combination of output parameter values that has the least root mean square deviation is found.

Description

[0001] Method of multi-connected control of technological processes with forecasting

[0002] This invention relates to methods of multi-connected control of a technological process with forecasting. This technical solution is used in the field of oil refining, petrochemistry, in particular, it is used in controlling the processes of rectification, reforming, hydrocracking, hydrotreatment, etc., but can also be used when working with other technological systems and processes.

[0003] In the process of oil refining, as a rule, a mixture of various oil products is obtained. The cost of oil products differs significantly from each other depending on the possibility of its further use: ethane can be used as a raw material for the production of polyethylene, its cost is significantly higher than that of methane, which is almost always used as fuel. The requirements for each type of oil product are strictly regulated by regulatory documents, but for most oil products, the yield of an absolutely "pure" product is not required, the presence of impurities is allowed. Therefore, as a rule, the goal of the technological process is the maximum yield of the most valuable product with strict restrictions on the composition of each product.

[0004] Oil is collected and prepared at various fields and can be mixed in different proportions in oil pipelines, so the composition of raw materials at oil refining and petrochemical process facilities is constantly changing. Under the influence of external disturbances, such as, for example, changes in atmospheric air temperature during the day, constant correction of the process is required to achieve maximum efficiency. In general, the processes of oil refining and petrochemical production, as control objects, are multi-connected, characterized by significant inertia, constant change in the dynamic characteristics of control channels.

[0005] Let us give an example of a method for controlling a technological process using a rectification column without additional product outlets. Here, the main technological task is to separate the feedstock entering it into separate streams of substances with different molar masses. Separation is performed according to the boiling point parameter of the fractions. Such thermobaric conditions are created in the column so that one fraction remains predominantly in liquid form, and the other in gaseous form. The higher-boiling fraction is removed through the lower outlet of the column, the low-boiling fraction through the upper one. Separation is performed by heating and subsequent evaporation of the oil product located in the lower part of the column (still) and simultaneous cooling (condensation) of the evaporated oil product from the upper outlet and its subsequent supply as irrigation to the upper part of the column.The vapors of the oil product from the column cube rise upward, passing through a special device - plates, on which the condensed oil product flows (irrigation). When the gaseous and liquid phases come into contact, a multiple phase transition occurs, the end result of which is quite distinct fractions of the oil product.

[0006] To control the described technological process, it is necessary to simultaneously regulate various technological parameters - the heating power of the bottom product, the cooling power of the upper (gaseous) product, the degree of opening of the control valves that ensure the removal of the product from the unit, and this regulation is carried out manually by the operator based on data from the automated control system of the technological complex (ACS TP).

[0007] These APCS systems perform information and control functions. Information functions include collecting the values ​​of measured process variables (pressure, temperature), their subsequent filtering, checking their reliability, saving and displaying on the human-machine interface (on the control system screen in the control room). Control functions include issuing signals to actuators that change their state (switching on pumps, changing the degree of opening of control valves, the rotation frequency of air-cooling unit fans, opening shut-off valves, etc.). Control functions are used when a dangerous process mode is detected and it is necessary to transfer the controlled object to a safe state, when issuing commands by an operator (person) or when it is necessary to introduce actions to maintain the specified values ​​of process variables.

[0008] Structurally, modern control systems consist of sensors that measure physical parameters and transmit the measured values ​​to the "lower" level of the APCS, represented by controllers. Controllers provide control actions and ensure the transfer of information to the "upper" level of the APCS, represented, as a rule, by a set of servers and automated workstations (AWP) of process operators.

[0009] To maintain the set values ​​of process variables in the APCS, PID controllers are used, to the input of which the current value and the set limit for maintaining the value of the process variable are fed. The PID controller provides control action through the corresponding controller. When the current value deviates from the set value (as well as the integral and differential of the deviation), the signal output to the actuator controlling this process variable is adjusted.

[0010] The disadvantage of this method of controlling technological processes is that the control system reacts only to the fact of deviation of technological variables from the task, controls each control loop separately, without taking into account various factors and reasons for the occurrence of deviations.

[0011] For example, in the technological process of a rectification column, such regulators are used to maintain a given pressure, a given temperature in the cube and the upper part of the column, to maintain the flow rates of products removed from the unit.

[0012] The main problems with using a set of PID controllers are their single-connectivity and lack of forecasting. In addition to the problems of control using PID controllers, one can also include the lack of the ability to involve output product quality indicators in control due to the complexity of their measurement and the lack of means for their indirect calculation.

[0013] Methods for single-loop control of technological processes using PID controllers are described, for example, in patents US 6424873 "System and method for limiting integral calculated components in PID controllers" (dated 2002-07-23, patented by HONEYWELL INC. (Us), IPC class G05B11 / 42, G05B 5 / 01, G05B 13 / 02) and KR 20180028305 (A), "Method for controlling a secondary valve using the integral action of a PID as a controlled variable required by a primary valve" (dated 2018-03-16, patented by KOREA SOUTH EAST POWER CO LTD [KR], IPC class G05B 11 / 40; G05B 11 / 42).

[0014] Patent US 6424873 of 2002.07.23 describes a control device that includes a primary proportional, integral, differential ("PID") controller capable of receiving a first setpoint and a first process variable and generating a second setpoint based on them, and a secondary controller capable of receiving the second setpoint and a second process variable and generating an output control signal based on them. In this case, the primary PID controller is capable of receiving from the secondary controller a feedback signal 1) that indicates that the previous value of the second setpoint exceeds a limit associated with the output control signal of the secondary controller, and 2) that transmits the value of the signal from the secondary controller. Then the main PID controller is capable of limiting the contribution of the integral calculation component to the PID calculation that generates a new current value of the second setpoint.The integral component of the calculation can be excluded, included or partially included in the PID controller calculation to effectively minimize the influence of unwanted erroneous output signals.

[0015] According to the invention of patent KR 20180028305 dated 2018.03.16, a method for controlling an auxiliary valve in accordance with a required work volume of a main valve using the integral action of a PID controller in a power plant includes: 1) a step: further adjusting the auxiliary PID controller to control the auxiliary valve in accordance with the output signal of the main PID controller controlling the main valve, which makes it possible to input a predetermined positive (+) error into the auxiliary PID controller when the output signal of the main PID controller is equal to or higher than the upper limit value, which makes it possible to input a predetermined negative (-) error when the output signal of the main PID controller is equal to or less than the lower limit value, and makes it possible to input an error of 0 when the output signal of the main PID controller is between the upper limit value and the lower limit value,for calculating the error of the auxiliary PID controller according to the output of the main PID controller; 2) a step of determining the integration time (Ti) of the auxiliary PID controller according to the opening degree of the auxiliary valve; and 3) a step of controlling the auxiliary valve using the output signal of the auxiliary PID controller, which is calculated by a formula using the error and integration time (Ti) calculated in steps 1) and 2) as input data.

[0016] These technical solutions illustrate the principle of single-loop control based on a PID controller. The prototype chosen is a method of single-loop control of technological processes at technological objects based on PID controllers, in which PID controllers are used for each technological parameter of the technological process. The PID controller system is controlled by a multidimensional predictive controller containing a model of the technological process describing the influence of the controlled (input) variables on the monitored (output) variables. The current values ​​of the measured technological and other variables, as well as the boundary values ​​of each input and output variable specified by the operator, are fed to the controller input. The controller calculates the values ​​of the technological variables on the forecast horizon based on the process model embedded in it.The controller output variables are fed to the input of existing PID controllers as settings for the corresponding actuators.

[0017] When changing one input technological parameter - the controlled parameter, during the operation of the system, some other parameters of the system may go beyond the limits set by the operator. This is displayed by a mismatch in the corresponding PID controllers. When a mismatch appears in the corresponding PID controllers of technological parameters, that is, when a specified limit is reached in any TTID controller, the parameters are adjusted to return the parameters within the specified limits. This control method is adopted as a prototype.

[0018] The disadvantage of this approach is that each PID controller strives to maintain a given value of the corresponding process variable without taking into account the others. Due to the fact that technological processes at process facilities are multi-connected, almost every control action of any PID controller causes multiple gradual damping disturbances in other control loops. Using the example of a rectification column, if the operator increased the task for maintaining the temperature in the column cube by 5°C, a mismatch appears in the PID controller for controlling the cube temperature and it issues a command to gradually open the valve regulating the flow rate of the coolant for heating the bottom product. After some time (transport delay), the temperature in the column cube begins to increase and the controller gradually brings it to the set value.But, as the overall energy balance in the column changes, this leads to a change in pressure and temperature across the entire column profile.

[0019] This, in turn, causes a mismatch to occur in the corresponding pressure and temperature PID controllers, and they begin to issue control signals to return the values ​​of their supported process variables to the setpoints.

[0020] In order to minimize mutual disturbances, and also due to the presence of significant transport delays, all changes in the technological process by PID controllers are made extremely slowly, since a rapid change is fraught with overshoot, undamped oscillatory processes, and in general to a violation of the technological regulations in terms of the quality of the output products.

[0021] The next problem is the lack of forecasting. Changes are made to the control signal only after a mismatch occurs. In the case of a rectification column, this can be shown using the example of increasing the flow rate of raw materials into the unit. With an increase in the flow rate of raw materials and an unchanged heating power of the bottoms product, the temperature of the bottoms product will decrease. The PID controller, which maintains this temperature, will begin to make control actions only after the temperature drops below the set one. Given the fairly large transport delay, compensation for this disturbance takes a long time. The PID controller sends a command to the valve regulating the steam flow in the heat exchanger, then a larger amount of steam begins to heat the circulating bottoms oil product more strongly. The heated oil product must get into the bottom of the column, mix with the entire volume of the substance in the column, and after that the readings on the temperature sensor will gradually increase to the required values.

[0022] The present invention is aimed at solving the above-mentioned problems. The objective of the present invention is to create a method for multi-connected control of technological processes, reducing or preventing disturbances in various control loops due to forecasting technological processes.

[0023] The technical result consists in preventing and reducing mutual disturbances in various control loops of technological objects during control action; increasing the speed of forecasting the behavior of technical systems (operability) and increasing the accuracy of control of multi-parameter technological processes; increasing the efficiency of the technological system by maintaining an optimal technological mode, for example, reducing energy consumption, optimizing the operation of the catalyst, increasing the yield of the most valuable products; reducing the time of transient processes and reducing the loss of product quality when changing production tasks; ensuring the possibility of operation of the technological object in an operating mode near the specified technological and economic limitations (reducing the quality margin).

[0024] The technical result is achieved by means of a method of multi-connected control of technological processes with forecasting, which consists in the fact that a system of PID controllers is used, which is controlled using a multidimensional predictive controller, the current values ​​of technological variables and specified limitations are fed to the controller input, the controller calculates the values ​​of technological variables on the forecast horizon according to the technological process model embedded in it, differs in that if the parameters do not fit into the limitations specified by the operator, the necessary values ​​of the input variables are calculated to ensure that all variables are in the specified ranges of values ​​on the forecast horizon, for this purpose the technological process model is identified, for which the dependencies of the output parameters on the input parameters are determined, where the dependency model is represented by a function of the type aperiodic dynamic link,a matrix of models of the influence of each input parameter on each output parameter is obtained, for each combination of influence models the standard deviation of the parameters of the influence models from the reference data is calculated, such a combination of values ​​of the output parameters is found for which the standard deviation from the reference data is minimal, due to this such a set of values ​​of the input parameters is found for which the values ​​of the output parameters of the system are maximally close to the specified indicators, these values ​​of the parameters are fed to the input of the multiparameter controller to implement control actions and maintain the values ​​of the process parameters in the specified ranges.

[0025] The essence of the technical solution is as follows. When changing the input - controlled parameter or external disturbance, the dependence of the output process parameters on the input parameters is modeled. For each output parameter, a model of the dependence on the input parameter is built using a transient process model - a function of the type of aperiodic dynamic link. The use of aperiodic dynamic links for modeling transient processes is known from the prior art and is used to create regulators (see, for example, Besekersky V.A., Popov E.P. Theory of automatic control systems. Moscow: Science, 1972, Chapter 14.1, https: / / scask.ru / g_book_b_tau. php?id=86). In our case, an aperiodic link of the 1st or 2nd order or an ideal or real integrating link is used.

[0026] Functions of the type aperiodic dynamic links f (T, k, c) are identified by the coefficients of transient processes - transport delay, inertia coefficient and gain. Transport delay T shows how much time after the input variable changes the output variable will begin to change. Gain coefficient k shows how much the value of the output variable will change when the input variable changes by the value X. Inertia coefficient c - how much time the value of the output variable will change by approximately 2 / 3 of its full change k*X. The coefficients of transient processes T, k, c for various parameters are determined by retrospective system data or from preliminary experiments.The result is a matrix of models of the influence of controlled variables and measured disturbances on the controlled variables, where each model of influence is a model of the transient process with corresponding coefficients obtained from retrospective data or from an experiment.

[0027] For each combination of models, the standard deviation of the model parameters from the reference data is calculated. The standard deviation is the sum of the squares of the differences between the parameter values ​​obtained from the models and the values ​​obtained from the reference data. Reference data are calibration data that relate changes in input parameters and the system's response to these changes - real indicators of the system's state, recorded by sensors or obtained in advance in laboratory studies.

[0028] A combination of output parameter values ​​is found for which the standard deviation is minimal. Thus, a set of input parameters is determined for which the output parameter values ​​will be closest to the reference data. That is, a set of control actions is found that brings the system parameters as close as possible to the specified indicators for which the system is maximally stable. This minimizes the effects of mutual disturbances and eliminates the possibility of undamped oscillations in the system.

[0029] The difference between the model parameter and the reference data determines the range of values ​​that the output parameter value must fall within. These values ​​are fed to the input of the multiparameter controller, which calculates the settings for the PID controllers based on these values. The PID controllers adjust the signals sent to the actuators to maintain the values ​​of the process parameters within the specified ranges.

[0030] The difference from the prototype is that in our case, when changing one parameter, a forecast is built on how the change in input parameters will affect the change in output parameters. For this, transient process models are built, where the dependencies of each output parameter Pout on each input parameter Pin are specified by transient process functions. The coefficients of these functions T, k, c are determined in preliminary experiments or according to retrospective data.

[0031] Pout = £(T,k,s)(Pin) (1)

[0032] A set of models of the influence of each input parameter on each output parameter is obtained. Thus, the influence of a combination of input parameters on an output parameter is determined as the sum of the models of the influence of each input parameter on a given output parameter.

[0033] Рвых! = Efj(T,k,c)(PBxj) (2) where i is the number of the output parameter, j is the number of the input parameter.

[0034] For each combination of output parameters, the standard deviation of the output parameters of the models from the reference data is calculated.

[0035] Find such a combination of values ​​of the output parameters Pouthk for which the standard deviation is minimal, provided that the parameters fall within the specified constraints P: [Pmin, Pmax]

[0036] {Рвыхк | S(P) = minS, Рвыхтт <= Рвых <= Рвыхmax, Рвхтт <= Рвх <= Рвхmax} (4)

[0037] Using this combination, a set of control actions (input parameter values) is found, at which the values ​​of the system’s output parameters are as close as possible to the specified indicators.

[0038] These values ​​are set in the control system to implement control action on the actuators. The difference between the parameter value according to the model and the reference data determines the minimum permissible ranges of values ​​in which the output parameters must fall.

[0039] Unlike the prototype, in our case a set of system parameters is found, at which the technological object will reach the specified operating mode without additional sequential adjustments. These parameters are set as acceptable ranges of values ​​for the corresponding system parameters immediately at the moment of changing the system task. This can happen in an automated mode. In this case, such effects as self-oscillations are excluded, mutual disturbances are minimized, the speed and accuracy of technological process control are increased.By maintaining an optimal process mode, many beneficial effects are achieved - reduced energy consumption, optimized catalyst operation, increased yield of the most valuable products, reduced time of transient processes and reduced loss of product quality when production tasks change, ensuring the possibility of operating a process facility in a mode of operation close to specified technological and economic limitations (reduction of quality margin).

[0040] At the same time, the system constantly monitors whether the parameters go beyond the permissible values ​​or not. When external disturbances change, the control system reacts to parameter changes in real time and adjusts the effects on the actuators. For example, during the day there is a significant drop in air temperature, due to which the temperature inside the rectification column changes, this affects the thermodynamic processes inside the column. Using our method, in real time, the system automatically records the temperature change and exerts control effects on the actuators in order to level out unauthorized changes in the column temperature.

[0041] The technical solution can be explained by the following example. There is a production task to change the quality indicators of the bottom product. The task is usually formulated in terms of shifting the boiling temperatures of the fractions, or in the percentage of the upper fraction in the bottom product or the lower fraction in the top product. To do this, the process operator increases the temperature settings of the bottom product, due to which the distribution of fractions between the top and bottom products of the column changes (an increase in the end boiling temperature of the bottom product and the beginning of boiling of the top product, or a decrease in the concentration of the top product in the bottom).

[0042] When working with a standard method of controlling a technological object (prototype), the operator is forced to gradually change the settings to maintain the technological variables that affect the parameters specified in the production task. The operator makes these changes based on his own experience. As a rule, the technological parameters are brought to the specified values ​​within one day (or even several days). The long time of bringing them to the specified values ​​is due to the fact that product quality indicators are usually not measured in real time, but are determined in the laboratory with a delay of several hours from the process. That is, the operator learns about the actual indicators of the technological object's products only several hours after receiving the analysis results from the laboratory.All this time, the operator iteratively very slowly brings the process variables closer to the set values ​​as the results of laboratory tests are received, carried out every 4, 8, 12 hours.

[0043] In our case, when using the claimed method, the operator directly enters into the system the production task of quality indicators in natural units (boiling point, concentration, etc.). These indicators are entered in the form of an acceptable range of values ​​of product quality indicators - output parameters and controlled variables - input parameters. After entering the values, if the current or predicted values ​​of quality indicators go beyond the range of values ​​according to the built-in standard dynamic model of the technological process, using our method, a set of ranges of parameter values ​​is found so that when one parameter increases - for example, the boiling point of fractions, the remaining parameters remain in the specified ranges of values. These values ​​​​are fed to the input of the multi-parameter controller. In this case, the operator no longer needs to iteratively gradually adjust the process parameters so that they reach the specified values.Using our method, we find a set of values ​​for technological parameters, by setting which the system itself brings the technological process to a given operating mode with given indicators.

[0044] There are situations when there are pairs of multi-connected variables between which there is a conflict. When one parameter is in the given range, another condition cannot be met - the other parameter being in the given range. A distinctive feature of the claimed method is the ability to resolve such conflict situations when it is impossible to ensure that all variables are in the given ranges.

[0045] For example, let us consider a situation where the qualities of the upper and lower products, as well as the flow rate of the lower product, are specified for a binary rectification column (depropanizer). If it is impossible to simultaneously maintain the specified quality of both products, and priority is given to the quality of one product, while the quality of the second product is outside the range of specified values, then using our method it is possible to bring the quality of the second product as close as possible to the initially specified one, provided that the permissible range of the quality of the first product is not exceeded. Thus, a reduction in the quality reserve is achieved.

[0046] This method achieves a technical result in preventing and reducing mutual disturbances in various control loops of technological objects, increasing the speed of predicting the behavior of technical systems (operability), increasing the accuracy of control of multi-parameter technological processes, etc.

Claims

Method of multi-connected control of technological processes with forecasting Formula A method of multi-connected control of technological processes with forecasting, consisting in the fact that a system of PID controllers is used, which is controlled using a multidimensional predictive controller, the current values of the technological variables and specified limitations are fed to the controller input, the controller calculates the values of the technological variables on the forecast horizon according to the technological process model embedded in it, characterized in that if the parameters do not fit into the limitations specified by the operator, the necessary values of the input variables are calculated to ensure that all variables are in the specified ranges of values on the forecast horizon, for this purpose the technological process model is identified, for which the dependencies of the output parameters on the input parameters are determined, where the dependency model is represented by a function of the type aperiodic dynamic link, a matrix of models of the influence of each input parameter on each output parameter is obtained,for each combination of influence models, the standard deviation of the parameters of the influence models from the reference data is calculated, a combination of output parameter values is found for which the standard deviation from the reference data is minimal, due to this, a set of input parameter values is found for which the values of the system output parameters are as close as possible to the specified indicators, these parameter values are fed to the input of the multiparameter controller to implement control actions and maintain the values of the process parameters in the specified ranges.

Images