MXPA97006666A - Method and apparatus for controlling a processomulatory - Google Patents

Abstract

The present invention relates to a process in which an input power is processed having at least one fluctuation parameter to obtain an output power, and in which at least one controlled variable in the output power varies In response to changes in the reference point of at least one manipulated variable, a method to vary the reference point of the manipulated variable to carry out an objective level of the variable independently of the controlled variable, the method is characterized because it includes the steps of: (a) recording and collecting a plurality of process parameters that affect the variable controlled and affected by the manipulated variable, (b) in response to the parameters of the process collected in stage (a) and the present value of the reference point of the manipulated variable determine the difference between the present value and the optimal value for the reference point of the vari able manipulated to reach the objective level of the controlled variable, (c) in response to the process parameters collected in stage (a) and the difference determined in stage (b), predict the level of the controlled variable that may result of applying the optimal manipulated reference point value to the process, (d) obtaining the present level of the controlled variable, (e) comparing the predicted result of step (c) with the present level obtained in stage (d) to provide a feedback signal; (f) in response to the feedback signal and the present value of the reference point of the manipulated variable modify the difference determined in step (b) in such a way that the value of the reference point of the optimal manipulated variable reach more approximately the target level of the controlled variable, and (g) apply a signal representing the value of the modified optimal reference point derived in step (f) to control the p reference variable of the variable manipulated in the process and therefore control the controlled variable in the output power, in which steps (a) to (g) constitute a cycle of optimization repeated regularly at a predetermined frequency; that step (d) includes the steps of: (d.1) directly measuring the current present level of the controlled variable, and (d.2) inferring the present level of the controlled variable from at least one parameter of the related process , collected in step (a), and in which step (e) comprises comparing the predetermined result of step (c) with both current measured levels obtained in step (d.1) and the inferred level in the stage (d.2) to provide the feedback signal, and further comprises the steps of: (h) storing, for each optimization cycle, the difference determined in step (b) and the current level of the controlled variable measured in the stage (d.1); and (i) during each optimization cycle n M-th, incorporate a tuning factor in the feedback signal to reduce the magnitude of the modification required in the difference determined in step (b) to reach the target value of the controlled variable, in which the tuning factor it is derived from trends in the difference and the current level stored in step (h), where M is a predetermined multiple of optimization cycles of such magnitude as to encompass a plurality of hours, wherein step (i) includes modifying the feedback signal by modifying the entire polynomial by a common integration factor, and also includes the stage of: (j) after each optimization cycle L, recalibrate the process to accommodate variations in the performance of the resulting process of the changes, cancellations or additions of process equipment, the variations in the execution that are observed from the differences and levels stored in stage (h), in which L is a multiple of optimization cycles that encompass a time structure in the order of a few days

Description

METHOD AND APPARATUS FOR CONTROLLING A MULTIVARIABLE PROCESS DESCRIPTION OF THE INVENTION The present invention relates to a method and apparatus for optimizing the process operations of a plant such as, but not limited to, alumina refining, fractionation, cryogenic expansion and processing and gas treatment. More particularly, the invention is directed to a method and apparatus for improving forward feeding, plural variable control techniques in the operations of the plant process. As indicated in U.S. Patent No. 4,349,869 (Prett et al), it is important to minimize the losses inherent in the processes performed in industrial plants, and this is similarly important for the simultaneous benefits in administration. Prett et al recognize that the control of forward feeding is important for the optimization of the process since this allows the user to initiate the action of the controller based on a prediction of the values of the controlled variables. The patent indicates that the previous forward feed controllers have had certain inherent problems based on the fact that the controllers have no "knowledge" as to what effect their controlled condition will have elsewhere in the total process, so it is required Operator intervention to alleviate problems. In addition, the above systems are described as allowing to handle the "large scale" power flow disturbances in the order of 10 to 15%, which are considered usual in oil fractionation processes, for example. As a solution to that problem, Prett et al provide a method to control and optimize the operation of a process that has plural input variables, independently manipulated plural variables and plural controlled variables that are dependent on input variables and manipulated variables. Input variables by themselves - for example, input flows, compositions, etc., may or may not be subject to manipulation but are classified as manipulated variables for the purposes of this discussion. The method involves introducing disturbance tests on the manipulated variables and then measuring the effect on the controlled variables, thereby allowing the response characteristics of the controlled variables (for a given change in one of the manipulated variables) to be easily calculated. The existing values of the manipulated variables and the controlled variables can then be measured, and the calculated response of the controlled variables can be used to calculate a new set of movements for the manipulated variables. The manipulated variables can then be adjusted according to the new set of movements to reach a new set of values. These movements, when implemented, have the effect of moving the controlled variables towards their optimal reference points. A characteristic of the Prett et al system is that it allows the formation of a projection for a future time of future controlled variable values. In a similar way, a number of future movements of each manipulated variable can be calculated to control the future values of the controlled variables at their desired operating points. In this way, the control of forward feeding is implemented by predicting, at one or more points in the future, the response of a process to changes in the manipulated variables. Based on the prediction trend of the process, a number of future movements for the manipulated variables can be calculated to minimize the error between the desired reference point and the predicted future response of the process. Importantly, feedback is used to predict - and therefore minimize - the benchmark error. The system described in the above has a number of limitations, at least none of which is incapable of responding to variations in a large scale of quality in forward feeding parameters such as production capacity and composition, and for volatile economic conditions that affect the system. In this context, "large scale of quality" means variations in the order of one hundred percent or more. For example, the system described in the Prett et al patent performs a fractionation process in an oil refinery. Raw material (for example, unpurified oil) is typically available in storage tanks containing two to three supplies per week for the system, allowing the system operator to control the speed of entry and composition within reasonably narrow limits . However, in gas processing fractionation systems, or in other systems where large supplies of raw material are not capable of being stored, such control of the operator is hardly feasible. Considering the gas fractionation situation, the incoming gases typically arrive directly through a pipeline from remote locations, usually without any control by the plant operator as to the time, speed of arrival, or composition of the total stream. The result is a large and random variation in the front end of the system, that is, in the raw material. Similarly, at the rear end of the plant there are no storage tanks typically for the output product as in the case of oil fractionation processes; either, the outflow of the gas fractionation process (typically clean gas and natural gas liquids) proceeds directly to the pipeline. Processes such as gas fractionation are also affected by economic parameters that are much more volatile than the economic parameters that affect the final product of the Prett et al system. Federal regulation has relegated the gas transportation industry to a common carrier status, effectively pre-including carriers from the contractor to the producers' gas vendor and then contracting to sell it downstream to users. As a consequence, most of the gas is sold in a free market characterized by significant and frequent variation of the price. Although the system described in the Prett et al patent can accommodate changes in economic conditions, it can only do so by being stopped and being reconfigured to function with the changed parameters. Such time of standstill and reconfiguration off-line is totally impractical for industries such as the natural gas industry where economic parameters are extremely volatile. In a fundamental sense, Prett et al's model is limited in that it is only mechanical rather than based on the physics and chemistry of the processes that are controlled.

It is an object of the present invention to provide a method and apparatus for optimizing the process control of the plant where large-scale variations of quality in input conditions can be accommodated online, ie, without interrupting the process. It is another object of the present invention to provide a method and apparatus for optimizing the control of the plant process wherein the highly volatile economic parameters that affect the processed product can be accommodated online. Another object of the present invention is to provide an improved method and apparatus of employing control of the forward feed process in a process to direct the continuously adaptive control arrangement to wide fluctuations in uncontrolled parameters that affect the process, as well as changes in the efficiency of the process equipment. It is yet another object of the present invention to provide an advanced multivariable control system for optimal on-line control of continuous processes, the optimal control that is achieved by continuously simulating the fundamental physical-chemical processes that occur in the plant, and by means of a reduction in plant operation variability, adjust specification objectives and improved control over variable cost.

In accordance with the present invention, the feed control to adalant is periodically optimized at regular short-term intervals (eg, every thirty seconds) using a feedback arrangement based on the results of the current process against the predicted one to ensure that the Control of forward feeding does not produce an anomalous effect. The control is based on a rigorous simulation of the realization process in the current plant conducted online. The multiple field variables are recorded on a continuous basis to provide input signals to predict in a forward feed manner the reference points of the optimal manipulated variable, required to achieve the desired results in the controlled variables. The feedback arrangement is calculated before each control of the reference point based on the forward feed is applied to the system and serves as a "health check" of the output parameter of the predicted plant against the output parameter of the measured plant, the latter being subject to vary due to non-linearities between the loading of the unit and the efficiency of realization of the unit. The measured output parameters are filtered to accommodate the deviation factors, therefore, effectively providing on-line calibration of the measurement instruments. The filtering, the corrected feedback, the calculation of the reference point of the optimal manipulated variable, and its application to the process, occurs each control cycle (for example, typically in the order of every thirty seconds). The replacement, addition or cancellation of equipment (and concomitant changes in the execution of the process), such as the addition of exchangers, pipes, etc., are automatically accommodated through the self-tuning routines of the system. In this respect, the calculator equations used in the online system simulation are relinearized by means of a re-calculation of the coefficients of the calculator equations. The purpose of changing the coefficients more than the simulation model is that small corrections of the coefficient provide greater alignment of the tuning without invalidating the structure of the model. Therefore, the equations of the calculator remain similar for the operations of the given process and therefore minimize the need to modify the model while the basic function of the unit of the process remains the same. The multivariable control method of the present invention allows efficient and optimal control over the total range of a normal plant and economic plant operations (usually, but it is not limited to ± 25% of the loads of the unit of the process, and / or several times in the composition, and / or for the widest variation in the economic parameters). It is important in this aspect that the ultivariable equations, rigorous, robust, linear and non-linear polynomials of high order, solve the nonlinear behavior of the process and, therefore, the non-linear nature of the benefits of the manipulated process.; while mechanical methods are limited to fixed beneficial solutions (that is, they are confined to a narrow process range). Filtering is used for both active measurements and manual data inputs since variations of the individual point can have a significant effect on model output predictions. Essentially, it is the repeatability of the instrument's performance, rather than the accuracy of the absolute instrument, which is of key importance in data filtering and reconciliation. The present invention relieves the operator of the need to constantly adjust the reference points and overhead costs of over-utilization of resources (such as utilities) in order to remain secure within the specification. The multivariable control system of the present invention performs an optimal continuous operation of a process unit through on-line prediction and control of reference points for the key process variables in the unit. This is carried out under the constant change of the conditions found, with robust solutions that require a minimum of operator intervention, manufacturing effort or updating in response to the change of plant. The above and still further objects, aspects and advantages of the present invention will become apparent upon consideration of the following detailed description of a specific embodiment thereof, particularly when taken in conjunction with the accompanying drawings in which like reference numbers in the various figures are used to designate similar components. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a schematic diagram of a fractional distillation process illustrating the optimization control aspects of the present invention. Figure 2 is a block diagram of the components of the multivariable control system of the present invention. Figure 3 is a schematic diagram illustrating the procedure of the feedback arrangement used in the process control system of the present invention. The Figure is ur. flow chart of a program that can be used to implement the present invention.

Referring in more detail to Figure 1 of the accompanying drawings, a gas fractionation process employing the principles of the present invention is illustrated. As noted, gas fractionation is only one of several different types of processes for which the control techniques of the present invention are applied; therefore, the invention should not be construed as limited in its application to the particular process described herein. A fractional distillation column 11 receives the incoming unpurified product via the inlet conduit 12, the unpurified product in this case being a liquid and / or gas composed of hydrocarbon constituents. Column 11 includes a rectification section having trays 1 through g and a distillation section having trays g + l through m. A higher distillate vapor stream is withdrawn from column 11 via conduit 13 and is conducted to heat exchanger 14 where the thermal energy is exchanged between the Frp stream of the upper distillate product and a cooling medium provided to the exchanger of heat by means of an additional conduit 15. The condensed upper distillate product is then passed to an accumulator 16 by means of a conduit 17. The liquid is withdrawn from the accumulator 16 by means of the conduit 18 and divided between the outlet conduit 19 and the reflux conduit 20 of the distillate product. higher under the control of a reflux pump 21. The reflux flow FR is supplied back to column 11 via conduit 22. A reboiler 30, typically structured as a heat exchanger, receives the bottom product from column 11 via conduit 31 and, after adding heat, returns the product to the column through the conduit 32. The heat QRß is supplied to the reboiler 30 through the pipe 33. The reboiler also provides the output flux Fß of the bottom product. In accordance with the present invention, the process illustrated in Figure 1 is controlled by a multivariable controller (MVC) 40 in the manner described herein in the following. In particular, in the process illustrated in Figure 1, the feed line or conduit 12 supplies an output feed flux Fp comprising, in this example, a pseudobinary mixture of hydrocarbon constituents x and y, where x is a point constituent low boiling, (ie light) ey is a high boiling (heavy) constituent. For example, x can be ethane (and light) and can be propane (and weighed) in a typical process.The objective of the MVC controller is to separate constituents x and y at a maximum benefit within a predetermined purity specification for the products. Separated, within the process and contractual constraints, and in the reduced variability of the process The manipulated variables, controlled by the MVC 40 to achieve these objectives are the reflow FR, the heat service of the QRß boiler and the column pressure Pc. The target controlled variables affected by the manipulated variables are: the upper specification, namely the optimal concentration of the heavy key (y) in the upper product (X), and the background specification, namely the optimal concentration of the key light (x) in the background product (Y) The field variables periodically recorded by the controller 40 to effect control over the manipulated variables are Tethered and transmitted by a variety of transducer units appropriately located to detect the respective variable levels and transmit the levels detected as signals to the controller 40 when queried sequentially in a conventional telemetric manner. In this regard, it should be noted that, for proper clarity in the drawings, the feeder links between the MVC 40 and the various field variable transmitters are not shown, however, it being understood that the registration can be effected by means of wired or remote link transmission. The field variable transmitters include a pressure transmitter 41, a flow transmitter 42 and a temperature transmitter 43 for detecting and transmitting to the controller 40 the pressure, flow and temperature levels, respectively, of the input feed flow Fp. . An on-line analyzer 44 detects and transmits the composition (i.e., the percent of x and the percent of y) present in the input feed flow Fp. Similarly, the temperature, composition and flow rate in the outflow Fß of the bottom product from the reboiler 30 in the conduit 34 are detected and transmitted by the temperature transmitter 45, the analyzer 46 and the flow transmitter 47 , respectively. The temperature of the upper distillation product Ft in the conduit 20 upstream of the reflux pump 21 is measured and transmitted to the MVC 40 by the temperature transmitter 48. The flow and composition of the reflow FR downstream of the pump 21 are measured and transmitted by the flow transmitter 49 and the in-line analyzer 50, respectively. In-line analyzers 44 and 50 are typically gas chromatography units that are commercially available. The inference temperature near the top of the column 11, and the inference temperature near the bottom of the column, are detected and transmitted by the temperature transmitters 51 and 52, respec ivapv = r.te. The differential pressure between the rectification section ll through g) of the column and the differential pressure between the distillation section (g + 1 through m) of the column are measured and transmitted by the pressure transmitters 53 and 54 ^ respectively. The ambient temperature is transmitted to the MVC 40 from the ambient temperature transmitter 55. The manipulated variable of pressure P ^ in the column 11, is adjusted by a pressure controller 60 under the control of the MVC 40. The reflow flow FR, another manipulated variable, is adjusted by the flow controller 61, also under the control of the MVC 40. The other manipulated variable, the amount of heat and cooling per unit of time (ie, heat service QRB), is controlled by the MVC 40 through a flow controller 62 that controls the steam that passes through the conduit 33. Figure 2 illustrates the essential aspects of the modules of the MVC resident in computer and its interface with the controlled process of the plant. An MVC module defines a self-contained process unit that has a distinctive function of the process. For example, a fractionation column and its associated equipment (for example, condensers, boilers, other heat exchangers, pumps, compressors, etc.) constitute such a module. With respect to the control aspects of such a process module, it is typical for the equipment that such valves, pumps, etc. they are controlled in a predetermined pattern by a regulatory control system such as a DCS (distributed control system), a PLC (programmable logic control) or individual circuit controllers (either pneumatic or electronic). In the systems of the prior art, the regulatory control system is generally located in a control room where the operator manually adjusts the reference points of each manipulated variable or 'allows the regulatory control system to automatically adjust these reference points. . In the automatic mode of such systems, the adjustment of the reference point of a manipulated variable is made in the base of the feedback information specifically in relation to that variable. In the present invention, however, these reference points of the manipulated variables are instead optimally adjusted on the basis of both feedback information and forward feeding in relation not only to that variable but to all the variables (i.e. process, economy, contract variables and calculated) simultaneously. The software of the MVC modules resides on a computer which can be a personal computer, a work station or another type of computer hardware. In the preferred embodiment, the computer used is a personal computer with a high 32-bit speed. As illustrated in Figure 2, the multivariable control variable (MVC) of the present invention effects communication in two ways with the regulatory control system through its I / O driver. The actuator is a conventional type communication protocol translator for input and output signals between the modules and the regulatory control system. These 1/0 signals are handled by the data acquisition and sequencing control software (SCADA), a commercially available software that functions as a data acquisition system for the MVC of the present invention. As the SCADA acquires data from the process;, it sends the data to the different software blocks to be used in carrying out the operations of the corresponding system; and simultaneously empties the same data in the history block, serving as storage of files, so that the data may be available for comparisons as described herein. Both SCADA and history software allow the system to easily access both current data and historical processes. The SCADA registers all the variables of the process, both manipulated and others, through the 1/0 actuator, usually every one or two seconds. The MVC modules are the executive software of the system and are described in detail in the following. Its function is to generate the optimal reference points for all the variables manipulated in each optimization cycle.

Before the system can be operated, it is necessary to load the equations of the multivariable control in the section of the control modules of the computer. These equations express the relationship between the registered variables, the economy of the system (that is, differential prices of the X and Y products, utility costs to provide energy loads to the condenser and boiler, and pumping costs), construction restrictions and specifications (ie, maximum levels of impurities) in the upper and bottom products), equipment restrictions (that is, for the column, the condenser, the accumulator, the boiler, the pumps, etc.) and the dynamics of the process and analyzer. The typical multivariate control equations used for the exemplary process described herein include: Manipulated Variable FR: Reflow Service FR = a * b. { W.) ° + cPí? C * * dKíti < • eKttJ * + fN ^ * q. { [irt) m - (i / r) rectr ...

Manipulated Variable QRE: Service of the QRB - a'F Boiler? • D '(l / r,) a' • c'Pf | * d'Kf | '- eftí | lf / * Manipulated Variable: Pressure e Column Pc. a% t3"• b". { p ^ / p and ". c" (Nrect, Nstrípr "+ ....

Where: a, b, c, d, e, f, g; a ', b', c ', d', e ', f', g '; a ", b", x: ", etc, are numerical coefficients characteristic of the particular distillation column used and are changeable when mechanical changes are made in the system, F is flow, T is absolute temperature, P is pressure, K is composition, N is the number of theoretical fractionation stages, p ° is saturation vapor pressure (of a component), and where the subscripts: f = power, c = column, i, j, k components (for example , light key, heavy key, etc.), rect = rectification section of a column, strip = distillation section, amb = environment, and where the exponents m, n, q, r, s, t, u, m ' , n ', q', r ', s', t1, u', r ", v", w "are constants. The function of the described equations is to allow the software routines of the control modules to simulate the operation of the process, based on the data of the process variable received from the SCADA, on a continuous basis. The simulations of this form follow the current plant or the realization of the process to allow the prediction of the optimal reference points of the manipulated variables, subject to the economy of the process, contractual and equipment constraints provided by the routines of the appropriate MVC module .

A functional block diagram representing the sequence of events during an optimization cycle typical of the operation of the system is illustrated in Figure 3 of the accompanying drawings. In this regard, optimization cycles occur at regular intervals, pre-selected by the operator, with a typical frequency in the range of half to five minutes. During each cycle the total process or the plant is optimized. The specific optimization cycle illustrated in Figure 3 is the cycle (n + l) th, the beginning of which occurs with the application of the signal level SPn to the software block 71 of the control algorithm. SP represents the reference point for any of the manipulated variables, and SPn represents the value of the reference point in the preceding cycle (ie, th). The values of all relevant forward feed variables (for example, the input feed flow rate, input feed composition, input feed temperature, etc.) are also applied to block 71 of the feed algorithm. control along with all the specifications of the applicable process (for example, product purity, economic requirements, constraints, etc.) All feed forward variables and process specifications are used in the control algorithm 71 to calculate the value SP -, of the optimal reference point of the variable manipulated during the current cycle (ie n + l) Before the value of the reference point SPR + 1 derived from the forward feed in the process is used , a connection is made in the feedback arrangement, Specifically, the result of the current process obtained from block 82 of the r is compared. The result of the analyzer with the predicted result derived from equation 81 of the result calculator. The result of the current process used for this comparison can be derived as the result of the analyzer of block 82 or the inference result of block 83. The difference between the predicted and actual results is sent to the algorithm 84 PID of the feedback array which derives the appropriate correction of the feedback to the value derived from the feed forward of the reference point of the manipulated variable. Both of the section 70 of specifications and feed forward and section 80 of the feedback arrangement are automatically tuned, linearized, derived and deferred to ensure the continuous adaptability of the optimization calculations for the execution of the change and dynamics of the process units. ? r. In particular, the 72 and 85 tuning factors of the equation represent the routines of the software * that automatically check for trends in differences in stored values of the predicted forward feed and feedback observed in history software. These trends, which signify possible changes in the execution of the system, are analyzed statistically and appropriate tuning factors are derived in order to keep the magnitude of the corrections of the feedback arrangement to a minimum. To manually reset the tuning procedure on the computer, the system operator-retrieves-and displays the equation tuning menu that lists the various tuning equations used by the system. In the command of pressing a key, the operator can then reset the operation of the feedback arrangement for each parameter of the main process. On an automatic basis, the tuning program highlights the differences between the values of the predicted and observed parameters over a pre-specified period of time (for example, a job change) and calculates the appropriate correction factor for each predictive equation. In this way, the procedure compensates for any systematic deviation in the execution of a component of the process, either due to gradual creep, changes in variables that do not exist. have been accounted for, or simply imperfect, of the system control modules by itself.

The linearization factors 73 and 86 represent software routines similar to the tuning factor routines but are invoked when significant mechanical changes are introduced into the system. Examples of such mechanical changes are changes or additions in the pipe, pumps, heat exchangers, etc. In this way, the function of the routines 73, 86 of the linearization factor is to recalibrate the plant or the system following the mechanical changes. The linearization factor program is evoked in a similar manner as described for the equation tuning program. The operator can retrieve and display the linearization factor menu when a mechanical change has been made to the system. In this case, however, the computer derives new numerical values for each coefficient of each term affected from the system equations rather than providing a new individual total correction factor for each equation. The routine 87 of the deviation factor allows the operator. automatically correct deviations in key instruments, for example a gas chromatograph. Once the operator calls the menu of the deviation factor, and for each sample of the process identified in the time on the screen, enter the results of the same analyzes carried out in laboratories identified in time. The system retrieves the corresponding online analyzer values, calculates the resulting correction or deviation factor for each selected component and, if approved by the operator, installs these factors in the system by key performance. The wait software 88 represents the routines for automatically adjusting the dynamic characteristics of the process and the measuring instruments. In the example illustrated in Figure 3, the wait 88 represents the sum of the process down time (i.e., sampling time from the power input to the analyzer) and dead time of the analyzer 82 itself. In this way, the value of equation 81 of the result calculator is calculated in a waiting time before the time of the current calculated result from the analyzer result block 82. Block 83 of the inference result represents a calculation of the composition of the resulting product as inferred from the temperature measurement, for example. The inference results are used due to the rapid dynamic response of the temperature measurement devices compared to the response times of the er analyzer. line. Refer to the arrangement of the resulting feedback odor-e-form a modifier to the reference point derived from the feed forward and the specification section 70 in order to generate, and send to the process, the value of the optimal validated reference point SPn, for the variable manipulated during the optimization cycle (n + l) th of the process. The software to perform an optimization cycle is represented by the flow chart illustrated in Figure 4 for which the specific reference will now be made. It is understood that the equations described in the above have been loaded into the control modules, and that the system has been initialized. The initialization process includes setting the system clock and activating the 1/0, SCADA and historical actuator. The various MVC modules, control modules and open / closed circuit breakers are also initialized. These switches allow to manually control the operation of the open circuit of the process in the selection of the operator or under certain circumstances such as calculated reference points that exceed established limit values in the range of the module routine. In addition, the initialization includes the selection by the operator of the frequency of the optimization cycle, and the numbers M and L determine the tuning frequencies and iinearization procedures. Come is illustrated in the flow chart, each optimization cycle begins by recording all the process variables shown in Figure 1 as being monitored. These include flows, temperatures, pressures, fluid compositions, etc. The recording is carried out over adjustable time intervals characteristic of the response times of the sensing medium in the process. Once the data has been collected, the change of the optimal reference point, for example, for the reflux flow FR is calculated. Before this change of the reference point can be sent by the MVC 40 (Figure 1) to the reflux flow controller 61, the predicted reflux composition that can result from such a change is compared to the inferred reflux composition (block 83). , Figure 3) based on the inference temperature Tp1 at the top of the column 11 as monitored by the temperature sensing means 51. In addition, the predicted reflux composition is compared to the current reflux composition (block 82, Figure 3) measured by the analyzer 50 (Figure 1). These comparisons are made with the appropriate process recesses (block 88, Figure 1) appropriately incorporated. The predicted differences, which correspond to the feedback arrangement, provide the feedback modification towards the change of the optimal reference point calculated for the reflux flow FR. The change of this modified form er. The reference wave is sent to the controller 61 of theor reflux. Still as part of the first cycle of eptimizaoiór. ie n = 1), a procedure similar to that described above is performed for the reflux flow FR for each of the other manipulated variables (ie, QRß and Pc). Once the first optimization cycle for all manipulated variables is finished, the same procedure is repeated automatically for the optimization cycles M-1, where M is pre-selected during the initialization of the system. The information received and derived in each of these cycles is directed to the history module. Among this information is the difference between the prediction of forward feeding and the currently observed results of product purity, etc. During the M-th optimization cycle, usually after several hours of operation, the process of tuning equations begins. That is, in the program illustrated in Figure 4, the tuning proceeds on a determination that n is equal to M. Without the interruption of the optimization procedure, the cumulative differences between the predicted and measured results are examined. (for example, product purities predicted against current superiors and inferiors) for systematic trends, if any. The tuning factors derived from such examinations are then applied as changes to the integration factor for the appropriate control and calculating equations. In other words, the change of the calculated reference point is integrated up or down in magnitude according to the derived tuning factor. In this way, the corrections of the required feedback arrangement become small at the base of the system experience, namely, cumulative differences between the predicted and actual results in the preceding optimization cycles. Similarly, during the M-th cycle, the deviation factors are applied as multipliers to observed values of in-line variables against measured calculations. For example, the laboratory calibrations printed in time of the analyzers 44, 50 and 46 in line for the feeding currents Fp, reflux FR and funds Fß, respectively, are used for statistical application for the deviation factors towards the readings in line . Likewise, the time-printed calibrations of the inference temperature monitors 51 and 52 of upper and bottom, respectively, against the readings of the corresponding on-line analyzers are employed for the automatic calculation of the deviation factors for the inferred in-line compositions. in block 83. The applications of the deviation and tuning factor of the equation are carried out during each M-th optimization cycle. As indicated in the above, M is set by the operator and is typically provided for several hours of operation between the successive tuning processes. The system also has the flexibility of adaptation for mechanical or plant changes. For example, if condenser surfaces are added, or the feed input position is changed, or the components or aggregates are changed, an automatic relinearization procedure is invoked without interrupting the process of the optimization cycle. In the illustrated example, the relinearization of each optimization cycle L is invoked, typically providing a few days between each relinearization. In contrast to the effect of the equation tuning process (in which the integration factor of the total equation is modified), the relinearization results in the modification of specific coefficients of terms that construct the calculator and control equations. These modifications are performed as demands for calibration requirements, typically requiring, as indicated, a few days of predicted versus observed results after a change to the plant is made. The system as described herein is an advanced multivariable control system designed for optimal online control of continuous processes. The computer of the system can be a personal computer or a work station and has software actuators that connect this to the system of regulatory control of the plant. Using modules of rigorous simulations of the process tuned to the execution of the current plant, the system continuously registers the many field variables and then predicts the optimal reference points for the control of the manipulated variables of the process, subject to the economy of the plant , contractual and equipment restrictions. The manipulated variables of the process are optimized approximately every thirty seconds (or in a few minutes, if desired). A characteristic of the main system is its adaptability in which, before each optimum reference point preprocessed, based on the economy, and based on forward feeding is sent to the process, a feedback arrangement is calculated based on the reactions of the current process against the predicted one. The recesses of the process, the critical equipment restrictions, the realization and deviation of the instrument are updated in the online routine through self-tuning. The equations of the reference point of the standardized manipulated variable are used and have coefficients adapted to the unit of process. To eliminate the data problems associated with mechanical versus fixed models, the equations are calibrated over the range of operation of the unit.

The system automatically compensates for process dynamics, compositions or feed rate to vary a set of product specifications, prices and utility costs, which result in the economic factors used to set the optimal operating reference points. The advantage of adaptive forward feeding control is its ability to react to these continuous changes and predict anticipated results, with time, appropriate set by the process transaction times, before the system moves away from the specification constraints. The resulting capacity of the optimization control, without exceeding or minimizing the product specification, increases the utility of the system. The system of the present invention is a level of control over the regulatory control system of the plant. In the case of failures or out of range operation, the operator can interpose the control of the regulatory open circuit. However, with the present invention in line, the operator simply provides desired operation objectives, product specifications and equipment selections. The system optimally fixes the conditions of the process in operation for better achievements of the objective requirements. Economic and engineering information can be entered manually as required by the operator or received from remote locations. The operator has the option of manually entering the recovery values of optimal components from the economic aspect module or having these entries automatically made by the system. The economic aspect module provides the operator with a current optimal target for the recovery of the final product in any given set of process conditions. This data is displayed on a screen that can be displayed at the operator initialization. The recovery of the optimal product is typically shown as a function of the values of flow components, utility costs, and the like, changes in these economic parameters are introduced by means of the keyboard by the operator or other authorized person. Alternatively, these factors can be introduced from remote locations. The operator is also provided with the ability to activate or deactivate each of these MVC modules. In the active position, the system operates in the closed circuit and readjusts the reference points of the variables manipulated directly. In the deactivated position, the system receives and processes all the information online, including the calculation of the optimal reference points, but stops the current sending of such reference points to the process.

Another feature of the system is its ability to automatically adapt different ranges of operation under the control of the range module in Figure 2. Specifically, in certain processes, such as gas processing, the input power and the output power are subject to huge changes in the-flow rates. The function of the range module is to automatically change from one set of equations to another set of equations as the method of parameters at the end of one range and the beginning of another. In essence, these amounts are to reconfigure the system or simulated model to accommodate different ranges of operation. The equations reside in the RAM in the computer. Therefore, for example, a set of equations can be invoked when the system is operating in the range between twenty percent of normal capacity to seventy percent of normal capacity; another set of equations can be invoked for ranges of operation between seventy percent and one hundred and fifty percent of the normal capacity; etc. The equations for each parameter in the different ranges remain in the same form, only their coefficients differ. Another important feature of the present invention is the frequent optimization control. Specifically, the optimization cycles, which occur in the order of every thirty seconds, reduce the impact on the system of severe variability of conditions and inputs and outputs and the volatility of economic parameters. From the above description it will be appreciated that the invention makes available a method and apparatus of control of the novel multivariable process to optimize the control of the plant process in the presence of large-scale variations in the conditions of entry and exit and highly economic parameters. volatile that affect the processed product. Having described a preferred embodiment of the improved and novel multivariable process control method and apparatus of the present invention, it is believed that other modifications, variations and changes will be suggested by those skilled in the art in view of the teachings set forth herein. Accordingly, it is understood that all variations, modifications and changes may fall within the scope of the present invention as defined by the appended claims.

Claims (20)

1. CLAIMS 1. In a process in which an input power is processed that has at least one fluctuation parameter to obtain an output power, and in which at least one controlled variable in the output power varies in response to the changes in the "reference point of at least one manipulated variable, a method for varying the reference point of the manipulated variable to carry out an objective level of the variable independently of the controlled variable, the method is characterized in that it comprises the stages of: (a) recording and collecting a plurality of process parameters that affect the variable controlled and affected by the manipulated variable, (b) in response to the parameters of the process collected in stage (a) and the present value of the reference point of the manipulated variable, determine the difference between the present value and the optimal value for the reference point of the manipulated variable p to reach the target level of the controlled variable; (c) in response to the process parameters collected in step (a) and the difference determined in step (b), predict the level of the controlled variable that may result from applying the optimal manipulated reference point value to the process; (d) obtain the present level of the controlled variable; (e) comparing the predicted result of step (c) with the present level obtained in step (d) to provide a feedback signal; (f) in response to the feedback signal and the present value of the manipulated variable's reference point, modify the difference determined in the stage (b) in such a way that the value of the reference point of the optimal manipulated variable reaches more approximately the target level of the controlled variable; and (g) apply a signal representing the value of the modified optimal reference point derived in the stage (f) to control the reference point of the variable manipulated in the process and therefore control the controlled variable in the output power; wherein steps (a) to (g) constitute an optimization cycle repeated regularly at a predetermined frequency; wherein step (d) includes the steps of: (d.l) directly measuring the current present level of the controlled variable; and (d.2) inferring the present level of the controlled variable from at least one parameter of the related process, collected in step (a); and wherein step (e) comprises comparing the _ predetermined result of step (c) with both current measured levels obtained in step (d.l) and the inferred level in step (d.2) to provide the feedback signal; and further comprises the steps of: (h) storing, for each optimization cycle, the difference determined in step (b) and the current level of the controlled variable measured in step (d.l); and (i) during each M-th optimization cycle, incorporate a tuning factor in the feedback signal to reduce the magnitude of the modification required in the difference determined in step (b) to achieve the target value of the controlled variable , wherein the tuning factor is derived from trends in the difference and the current level stored in step (h); wherein M is a predetermined multiple of optimization cycles of such magnitude as to encompass a plurality of hours; wherein step (i) includes modifying the feedback signal by modifying the entire polynomial by a common integration factor; and further comprises the stage of: (j) after each optimization cycle L, recalibrate the process to accommodate variations in the performance of the process resulting from changes, cancellations or additions of process equipment, variations in the execution that they are observed from the differences and levels stored in step (h); where L is a multiple of optimization cycles that encompass a time structure in the order of a few days.
2. 2. The method according to claim 1, further characterized in that it comprises the step of adjusting the level of the feedback signal to accommodate inherent delay times in the process.
3. The method according to claim 1, characterized in that step (b) includes calculating the value of the reference point of the optimal manipulated variable from a stored polynomial, in which each term includes a coefficient and at least one variable , each variable at least corresponds to a parameter of the respective process collected in step (a).
4. The method according to claim 1, further characterized in that it comprises the step of adjusting the level of the feedback signal to accommodate the delay times inherent in the process.
5. 5. The method according to claim 1, characterized in that there are a plurality of controlled variables that respond to changes of reference points of three manipulated variables, and in which the stages in each optimization cycle are performed separately and in respective sequences for each of the three manipulated variables.
6. 6. The method of compliance with the claim 14, characterized in that the process is a gas fractionation process, in which the input feed is a gas having at least first and second constituents, in which the process uses a fractionation column to separate the constituents of the gas in the flow of superior products leaving the top of the column and the flow of bottom products leaving the bottom of the column, in which the controlled variables are the concentration of the first constituent in the flow of the superior product and the concentration of the second constituent in the bottom product stream, and in which the manipulated variables are the pressure in the fractionation column P ^, the reflux flow rate FR of the upper product back to the column, and the velocity of heating of the QRß flow of the fund product back to the column.
7. The method according to claim 6, characterized in that step (a) includes: determining the value of the optimal reference point of FR by a calculation from a first polynomial, namely: FR - aF, "4 b. {L / Tty * cP, ^ * dKt? Go 'eKttJ * + fWrßct £ + g ((i / r)« * - d / r) r? Ct) u < • • • determine the value of the optimal reference point for QRB by calculating from a second polynomial, namely: QRB - a'F? * b '. { l / Tt) n '- c'Pt * d'K, * e'Kt? l? * and determine the optimal reference point for Pc by a calculation from a third polynomial, namely: Pc «a% tij" "* cw (Nrect, Wstrip)" "+ .... where a, b, c, d, e, f, g; a ', b', c ', d', e ', f', g '; a ", b", c ", etc, are characteristic numerical coefficients of the particular distillation column used and are changeable when the mechanical changes are made in the system, F is flow, T is absolute temperature, P is pressure, K is composition, N is the number of theoretical fractionation steps, p ^ is saturation vapor pressure (of one component), and where the subscripts: f = food, c = column, i, j, k = components (for example, light key, heavy key, etc.), rect = rectification section of a column, strip = distillation section, amb = environment, and where the exponents m, n, q, r, s, t, u, m1 , n ', q1, r', s1, t1, u ', r ", v", w "are constants.
8. 8. The method according to claim 1, characterized in that a fluctuation parameter is subject to unpredictable changes in level in the order of at least one hundred percent.
9. 9. In a process in which an input power is processed having at least one jitter parameter to obtain an output power, and in which at least one controlled variable in the output power varies in response to changes in the reference point of at least one manipulated variable, a method for varying the reference point of the manipulated variable to carry out an objective level of the variable independently of the controlled variable, the method is characterized in that it comprises the steps from: (a) recording and collecting a plurality of process parameters that affect the variable controlled and affected by the manipulated variable; (b) in response to the parameters of the process collected in step (a) and the present value of the manipulated variable's reference point, determine the difference between the present value and the optimal value for the reference point of the manipulated variable to reach the objective level of the controlled variable; (c) in response to the process parameters collected in step (a) and the difference determined in step (b), predict the level of the controlled variable that may result from applying the optimal manipulated reference point value to the process; (d) obtain the present level of the controlled variable; (e) comparing the predicted result of step (c) with the present level obtained in step (d) to provide a feedback signal, - (f) in response to the feedback signal and the present value of the reference point of the manipulated variable, modify the difference determined in step (b) in such a way that the value of the reference point of the optimal manipulated variable reaches more approximately the target level of the controlled variable; and (g) apply a signal representing the value of the modified optimal reference point derived in the stage (f) to control the reference point of the variable manipulated in the process and therefore control the controlled variable- in the output power; wherein steps (a) to (g) constitute an optimization cycle repeated regularly at a predetermined frequency; and (h) in response to a command, change dynamically during the online operation between different modes of processes in which the manipulated variable and the controlled variable are different in each mode.
10. The method according to claim 2, further characterized in that it comprises the step of adjusting the level of the feedback signal to accommodate inherent delay times in the process.
11. The method according to claim 9, characterized in that step (b) includes calculating the value of the reference point of the optimal manipulated variable from a stored polynomial, in which each term includes a coefficient and at least one variable , each variable at least corresponds to a parameter of the respective process collected in step (a).
12. The method according to claim 9, further characterized in that it comprises the step of adjusting the level of the feedback signal to accommodate the delay times inherent in the process.
13. The method according to claim 9, characterized in that there are a plurality of controlled variables that respond to changes of reference points of three manipulated variables, and in which the stages in each optimization cycle are performed separately and in respective sequences for each of the three manipulated variables.
14. 14. The method according to claim 13, characterized in that the process is a gas fractionation process, in which the input feed is a gas having at least first and second constituents, in which the process uses a column of fractionation to separate the constituents of the gas in the flow of upper products leaving the top of the column and the flow of bottom products leaving the bottom of the column, in which the controlled variables are the concentration of the first constituent in the flow of the superior product and the concentration of the second constituent in the bottom product stream, and in which the manipulated variables are the pressure in the fractionation column Pc, the reflux flow rate FR of the superior product back to the column, and the rate of heating of the QRß flow of the fund product back to the column.
15. 15. In a process in which an input power is processed having at least one jitter parameter to obtain an output power, and in which at least one controlled variable in the output power varies in response to changes in the reference point of at least one manipulated variable, a method to vary the reference point of the manipulated variable to carry out an objective level of the variable independently of the controlled variable, the method is characterized in that it comprises stages of: (a) recording and collecting a plurality of process parameters that affect the variable controlled and affected by the manipulated variable; (b) in response to the parameters of the process collected in step (a) and the present value of the manipulated variable's reference point, determine the difference between the present value and the optimal value for the reference point of the manipulated variable to reach the objective level of the controlled variable; (c) in response to the process parameters collected in step (a) and the difference determined in step (b), predict the level of the controlled variable that may result from applying the optimal manipulated reference point value to the process; (d) obtain the present level of the controlled variable; (e) comparing the predicted result of step (c) with the present level obtained in step (d) to provide a feedback signal; (f) in response to the feedback signal and the present value of the manipulated variable reference point, modify the difference determined in step (b) in such a way that the value of the reference point of the optimal manipulated variable reaches more approximately the objective level of the variable -controlled; and (g) applying a signal representing the value of the modified optimal reference point derived in step (f) to control the reference point of the variable manipulated in the process and therefore controlling the controlled variable in the output power; wherein steps (a) to (g) constitute an optimization cycle repeated regularly at a predetermined frequency; in which the process uses multiple process units in a plant and has a total economic objective, and also includes the stage of (h) determining the total economic objective by providing real-time integration of the economic aspect and control of all units of process.
16. 16. The method according to claim 15, further characterized in that it comprises the step of adjusting the level of the feedback signal to accommodate inherent delay times in the process.
17. The method according to claim 15, characterized in that step (b) includes calculating the value of the reference point of the optimal manipulated variable from a stored polynomial, in which each term includes a coefficient and at least one variable , each variable at least corresponds to a parameter of the respective process collected in step (a).
18. 18. The method according to claim 15, further characterized in that it comprises the step of adjusting the level of the feedback signal to accommodate the delay times inherent in the process.
19. 19. The method according to claim 15, characterized in that there are a plurality of controlled variables that respond to changes of reference points of three manipulated variables, and in which the stages in each optimization cycle are performed separately and in respective sequences for each of the three manipulated variables.
20. 20. The method according to claim 19, characterized in that the process is a gas fractionation process, in which the input feed is a gas having at least first and second constituents, in which the process uses a fractionation column to separate the constituents of the gas in the flow of upper products leaving the top of the column and the flow of bottom products leaving the bottom of the column, in which the controlled variables are the concentration of the first constituent in the flow of the superior product and the concentration of the second constituent in the bottom product stream, and in which the manipulated variables are the pressure in the fractionation column Pc, the reflux flow rate FR of the superior product back to the column, and the rate of heating of the QRß flow of the fund product back to the column.