MXPA97009813A - Method of regulation of quimi product synthesis procedures - Google Patents

Abstract

The present invention relates to a method for regulating a process for the synthesis of at least one chemical product in an installation comprising at least one reactor (R) which can be assimilated to a perfectly mixed reactor, in which control quantities ( GC) allow acting on the development of the procedure so that one or more magnitudes associated with the properties of the product and / or the development of the procedure, called regulated quantities (GR), are equal to the corresponding instructions (CGR), whose method comprises the following steps: (a) entry of instructions concerning the regulated quantities (CGR), (b) calculation, by means of a prediction organ (OP), of predictions of the regulated quantities (PGR), based on measurements of the procedure control quantities (MGC), (c) use of a control organ (OC) to calculate setpoints of the procedure control variables (CGC), based on the instructions (CGR) and of the predictions (PGR) of the regulated quantities, (d) transmission of the instructions of the procedure control variables (CGC) to actuators in order to act on the development of the procedure, in which the prediction organ (OP) ) is based on a mathematical model of the procedure, called a direct model (M), and is designed in such a way that the mass MXR of at least one constituent (X) in the reactor (R) is predicted by the equation: MXR = LAG ( FXRin. tX, tX,) where FXRin is the mass flow of constituent X entering the reactor R; tX is the residence time of x in the reactor, the function y = LAG (u, t) is the solution of the differential equation u = t. dy / dt + y, calculated with the instantaneous value of u and t, as well as with the last value of y computes

Description

Method of regulation of chemical synthesis procedures Field of the Invention The present invention relates to a method of regulating chemical synthesis procedures. It also refers to a regulation device for the implementation of this method, as well as to a synthesis process, in particular of a polymer, regulated by this method.

Background of the Invention In a conventional chemical synthesis process, regulators of the PID (proportional-integral-differential) type are used to regulate individually a more or less important number of magnitudes (temperatures, flow rates, pressures, ...) that exert influence on the development of synthesis. In other words, for each temperature, flow or pressure to be regulated, its effective value is measured continuously (or intermittently), and a PID regulator compares this effective value with a value of c nsigna and acts on the quantity to be sent to reduce , HI comes the case, the difference between the setpoint value and the measured value. Given the complexity of most industrial processes of chemical synthesis, the slogan values of the different regulators must still be adjusted empirically to obtain the desired properties of the synthesized product. For this purpose, recipes are used, which provide combinations of empirically determined parameters to obtain, in stable regime, the desired properties of the synthesized product. From these recipes, empirical relationships between the regulated quantities and the properties of the synthesized product could be deduced, with the help of more or less sophisticated statistical tools. However, it is evident that these empirical relationships can hardly take into account the multiple interdependencies that exist between the different magnitudes regulated separately, no more than unknown perturbations such as the contents of impurities of the raw materials. It is also evident that a classical closed-loop regulation, using as return corrections measures of essential properties of the synthesized product, is hardly applicable to most synthesis procedures. In fact, the dead times involved in either the procedure, or in the measurements or analyzes used as return corrections, are too high and the interdependencies between the different magnitudes that govern the procedure are too complex. The international patent application WO 93/24533 describes a method of regulating a process for the polymerization of an alpha-olefin in the gas phase in a horizontal reactor, in which control parameters allow to act on the development of the process so that the index flow rate (MFR) of the polymer is equal to the corresponding setpoint value, the method comprising the following steps: - determining the relationships between the melt flow rate of the polymer leaving the reactor and a first series of parameters, - controlling this first series of parameters, - calculate the MFR of the polymer, and - adapt at least one of the parameters to adjust the calculated MFR to a previously determined value. It has been known for a long time that synthesis procedures, especially the continuous polymer synthesis procedures (polymerization procedures), provided with regulations with empirical adjustment of setpoints, have important drawbacks, which can be summarized as follows: - the start-up of the synthesis procedures take up a lot of time and generate significant quantities of "out of standards" product; - Quality transitions are slow, which also entails the production of important quantities of "out of standards" transition products; - the step of the process, that is, the mass of synthesized product (s) per unit of time, can hardly be modified without altering the properties of this or these products; - the constancy of the essential properties - the synthesized product or products often leaves to be desired, even in a stable regime. In order to avoid the empirical adjustment of the reference values in the specialized literature, it has been proposed to use methods of regulating synthesis procedures that have resorted to characteristic equations modeling the synthesis procedure to associate certain properties of the Synthesized products to the conditions of the operation of the reactors during the synthesis. However, in order to limit the complexity of these characteristic equations, it was considered until now that it was forced in practice either to consider exclusively the static case (stable regime), or to limit itself to a very simplified empirical modeling of the dynamics of the process. The use of a static model is limited to the control of a fairly stable production regime. In the case of empirical modeling, the characteristic equations are only valid for a narrow validity interval (in the vicinity of the point where the modeling has been carried out). In both cases, the start-up phases and the transition phases are poorly regulated. Certainly, a wider range of operating conditions must be "covered" by performing such local modeling at several different points in the space of the operating parameters, but such an approach becomes prohibitive when trying to regulate several magnitudes by acting on multiple parameters. Therefore, it would be desirable to have a simple method and regulation device, better adapted to the specificities of the dynamics of chemical synthesis procedures.

Summary of the Invention To this end, the present invention relates to a method for regulating a synthesis process of at least one chemical product in a plant comprising at least one reactor (R) which can be assimilated to a perfectly mixed reactor, in which at least one or several control variables (GC) allow to act on the development of the procedure so that one or more magnitudes associated with the properties of the product and / or the development of the process, called regulated quantities GR), are equal to corresponding instructions (CGR) (or at least as close as possible to these), whose method includes the following steps: (a) entry of instructions that refer to the regulated quantities (CGR); (b) calculation, by means of a prediction organ (OP), of predictions of the regulated quantities (PQ), based on measurements of the procedure control quantities (MGC); (c) use of a control body (OC) to calculate setpoints of the procedure control quantities (CGG), based on the setpoints (C - y;.) and of the predictions (PR) of the regulated quantities; (d) transmission of the instructions of the control variables of the procedure (C G) to actuators, or to regulating organs that control actuators, in order to act on the development of the procedure; wherein the prediction organ (OP) is based on a mathematical model of the method, called direct model (M) and is designed in such a way that the mass M? R of at least one constituent X in the reactor is predicted (R ) by the equation: M? R = LAG (F? Rin. t?, t?) in which: - F? Rj_n is the mass flow rate of constituent X entering reactor R; - t? is the residence time of x in the reactor (time constant), which is worth: t? = M? R / (or Fxdis) in which: - M? R designates the last calculated value of the mass of the constituent X present in the reactor R; -? Fxdis designates the sum of all the mass flows Fxdis with which the constituent X disappears from reactor R, especially by reaction and / or by reactor outlet; - the function y = LAG (u, t) is the solution of the differential equation: u = t. dy + and dt calculated with the instantaneous value of u and t, as well as with the last value of y calculated. The advantage of this method is that the aforementioned differential equation is solved by a simple algebraic calculation, for example by the following formula (designating T the time interval, usually small with respect to t, separating the successive calculations) or by an equivalent formula to it: T y (tT) + u (t). t (t) and (t) = 1 + T t (t; In the case where the masses of several constituents are evaluated as discussed above, the method of the invention is particularly advantageous insofar as these masses can be calculated sequentially by such simple algebraic calculations, often recalculated (T < &t; t in general). On the contrary, traditional methods require the simultaneous resolution of a system of differential equations, which usually requires a large calculation power and sophisticated numerical (integration) calculation algorithms; as a result, the duration of each iteration of the calculation is high, and consequently such a regulation reacts poorly to rapid variations. The regulated synthesis process can be used for the synthesis of a: monomer or polymer compound; very good results have been obtained for the regulation of polymerization procedures. The method also extends to the case where several useful products are synthesized simultaneously in the same procedure. The procedure can be continuous or discontinuous (batch); The method of regulation of the invention provides excellent results in the case of continuous processes. The regulated synthesis procedure may eventually constitute only a part of a larger procedure, whose other parts are regulated in a different way, or are not regulated. For the regulation method of the invention to be applicable, it is necessary that at least one reactor can be assimilated to a perfectly mixed reactor, that is, to a reactor in which the different magnitudes (temperature, concentration of the constituents present, etc. ) are almost identical at each point. Other possible reactors can be of the plug-flow type: they are modeled mathematically by dead times. The method is also applicable to a process that is carried out in several reactors arranged in series and / or in parallel, which may give rise to products with identical or different properties. By "constituents", it is meant to designate the set of substances present in: the reactor and intended to participate in the synthesis or allow it; not only the salt starting reagents such as the synthesized product (s), but also the optional substance (s) that do not undergo any transformation, such as solvents, catalysts, etc. In addition to one or several reactors, the installation in which the process is carried out may possibly comprise other conventional devices such as decompressors, strippers, condensers, dryers, distillation columns, etc. In general, these auxiliary devices can also be considered as reactors (perfectly mixed or piston type), even if no chemical reaction takes place. In the case of a polymerization process, the "magnitudes associated with the properties of the product" can be chosen, for example, between the molecular mass, the melt flow index, the standard density, the comonomer content when it is found. present a comonomer, etc.

Examples of "quantities associated with the process" are, in particular, the temperature and the pressure prevailing in the reactor, the step of the procedure, the concentrations of the different reagents in the reactor, etc. The step of the march designates the mass of product synthesized per unit of time, which is not, however, by force equal to the flow rate of synthesized product leaving the reactor; thus, by way of example, in the start-up phases in particular, the mass flow rate of the synthesized product leaving the reactor is generally very small, even nil, although the synthesis has started, that is to say that the flow rate that exits is then less than step of the march. In stable regime, on the other hand, the step of the march can be assimilated to the mass of product synthesized per unit of time. Examples of "control quantities" are the flow rates of reagents entering the reactor, the power supplied to the heating devices, etc. It is these magnitudes that allow acting on the development of the procedure as well as, in general, on the properties of the synthesized product. The setpoint or setpoints of the control quantity or magnitudes are transmitted directly or indirectly to conventional actuators such as, in particular, valves, heating elements, -; c.

"Indirectly" means that the control variables can be transmitted by means of one or more control elements (which usually control only one variable, for example, PID controllers) that regulate the actuator (s) ("local" regulation). In the material plane, the prediction body and the control body are, in general, classic calculation devices that allow calculations to be made according to their wiring or their programming: in particular, they can be computers or numerical control systems ( SNCC). A single device can advantageously combine the prediction and control functions. The calculation device (s) used are preferably of the numerical type, and periodically (intermittently) provide the results of their calculations. The time intervals that separate the provision of these results may vary over time, and may also differ according to the result involved: it is clear that the quantities with rapid variation must be recalculated more frequently than the magnitudes of lenca variation. Phase shift registers can be used to materially simulate downtime. The prediction organ is based on a direct mathematical model of the procedure (M), in which the reactor (R) is assimilated to a perfectly mixed reactor; one or several net delays (dead times) can, if necessary, be taken into account to represent any piston-type reactors, possible transport delays or delays in obtaining measurement results, ... The control body is based on preference, in the inverse of the direct model used in the prediction body (inverse model). In general, the sum E Fxdis of all the basic flows (Fxdis) with which the constituent X disappears from the reactor R comprises two terms: * FR? which designates the mass flow rate according to which X is consumed by one or several eventual chemical reactions; * FXout < 5ue designates the eventual mass flow of X that leaves the reactor by extraction in the course of the reaction, in the (usual) case in which X is not totally consumed by reaction in this reactor; or even, for example, by evaporation in the case of an open reactor. The interest of the method is that the terms Fxdis are, in general, proportional to M? R; for example, you have in general FXout = MXR / CR (designating tR the residence time of reactor R) and FR? = R? . M? R (designating R? The reactivity of X in reactor R).

In this case, the expression that gives t? It simplifies and becomes: t? = 1 / (R? + L / tR) This expression is independently of M? R, which is a very interesting simplification. Another advantage of the method lies in the periodic calculation of the dwell time t? In effect, t? represents the dynamics of the constituent considered in the reactor. This allows in particular to follow the evolution of this parameter, which is important for understanding the dynamics of the procedure, and therefore for its regulation. On the contrary, empirical methods of type "black box" (boite noire) does not allow access to this parameter. Advantageously, the calculation of the predictions of the regulated quantities (PGp can, additionally, take into account one or more measures of regulated quantities (MQR), of control quantities (MGC) and / or other quantities associated with the development of the procedure (M ^ p). Likewise, advantageously, the calculation of the setpoints of the control variables of the procedure (CGG) can additionally take into account one or several measures of regulated quantities (MQR '' ^ e control quantities (MGG) and / or other quantities associated with the development of the rM procedure?). identical or different from those taken into account, possibly, for the calculation of the predictions of the regulated quantities (PQR) • All the measures in question in the present description are not necessarily direct measures, in the sense in which one or several of them can possibly be measures that are inferred, that is, values obtained by calculation from one or several other direct measures. Thus, for example, the passage of certain exoteric synthesis procedures can not be measured directly, but a measurement can be obtained by calculation, for example from (direct) measurements of the flow and of the inlet temperatures and of cooling fluid outlet. In the particular case of polymerization processes, the property or properties of the polymer involved in the regulation are preferably chosen between the specific density (MVS) of the polymer, the rheological properties of the polymer in the molten state, and its comonomer content. In particular, the property or rheological properties that are involved in the regulation method are advantageously the melt flow index (eit index) of the polymer and / or a viscosity measurement. Advantageously, one or more properties of the polymer is evaluated using a technique chosen from Near Infrared Spectroscopy (NIR), Fourier Transform Infrared Spectroscopy (FTIR) and Nuclear Magnetic Resonance (NMR). In particular, one or more properties of the polymer can be advantageously evaluated by applying a pre-established correlation relationship to the results of measurements made by near-infrared spectroscopy.

(NIR) at various predetermined wavelengths depending on the nature of the polymer, chosen between 0.8 and 2, 6 m. Disadvantage of details concerning the performance of such measures in the framework of the regulation of polymerization procedures can be found in the patent application EP 328826 (US 5155184). In order to take into account possible deviations between the measures and the regulated predictions, it may be useful to resort to a correction. A first type of correction consists in that the setpoint of at least one regulated quantity is corrected based on the deviation (advantageously filtered) between the measure (MQR ^ ^ ^ a prediction (pGR ^ ^ e esCa regulated quantity, so that the regulation is effective (Mr-, = CG) even in the presence of an error in the prediction of this regulated quantity.This technique is commonly referred to by the expression "internal model control" (IMC).

A second type of correction is that the model (M) of the procedure is adapted periodically on the basis of the deviation (advantageously filtered) between the predictions (PR) and the measurements (MQR1 of the regulated quantities, so that, here again , the model of the procedure provides predictions of the regulated quantities (PQR) as close as possible (ideally, the same) of the measurements of these quantities (MGR), which is essential for an effective regulation.The adaptation consists in recalibrating the model In other words, in recalculating one or several parameters, normally the number of recalculated parameters does not exceed the number of regulated quantities for which both a prediction and a measurement are available. time) of these measures is often desirable, especially when it comes to measures of properties of the synthesized product whose duration of collection is l This second type of correction is more advantageous in that it allows the model to be adapted equally on the plane of its dynamics. The adaptation concerns not only the direct model of the procedure (prediction organ) but also the inverse model (control organ). According to an advantageous variant, the measurements (MR) of the regulated quantities only intervene in the eventual adaptation of the procedure model, and do not intervene directly in the calculation of the instructions for the procedure control quantities (CGC). That is to say, that the measures of the regulated quantities do not intervene in the actual regulation: the advantage is that the quality of the regulation is not affected thus by the eventual slowness of the evaluation of the properties of the product. Another aspect of the invention relates to a method of regulation such as that described above, applied to a polymerization process. comprising one or more of the following supplementary steps: - calculation of a temperature setpoint in the reactor according to one or several slogans of product properties; and transmitting this temperature setpoint to one or more actuators that allow to modify the temperature in the reactor (possibly indirectly, that is to say through one or more regulating elements, for example, PID regulators, which control the actuator or actuators). ); - calculation of a thermal balance for the reactor, especially on the basis of temperature measurements; use of this thermal balance to determine the amount of polymer synthesized per unit time (step of the march) and / or the productivity of the catalyst and / or the concentration of at least one reagent in the reactor; - calculation of the amount of heat produced by the polymerization, thanks to a calculation of the amount of reagent or reagents that are polymerized; determination through it of the amount of heat that it is necessary to add or evacuate to maintain the temperature of the reactor; use of the result of said calculation (for example by feed-for ard) to improve the temperature regulation, in order to respect the temperature setpoint in the best possible way, especially in the case of changes in the pace of the march. These variants are based on the link between the amount of reagent or reagents involved in the reaction and the amount of heat produced or absorbed by the reaction. According to an advantageous variant, the property PxR of a constituent "x" in reactor R, assimilated to a perfectly mixed reactor, is calculated as follows: PxR = LAG (PxIN, MxR / FxIN) in which "Px" is a property of a constituent "x", which responds substantially to the law of the linear mixture 1 + 2 = l 'P? l + w' P: < ^ • where ^ and 2 are the mass proportions of two fractions 1 and 2 of property Px-j_ and Px2 that are mixed (1 + v¡2 = 1); P ?? + 2 is - * - a Property of x at its exit from the reactor, after mixing; PXIN is ^ to Pr of the constituent "x" at its entry into reactor R; MxR is the mass of constituent x in reactor R; FxjN is the mass flow rate of the constituent that enters the reactor R. A mathematical transformation sometimes allows to return linear (additive) certain quantities that are not, - for example, the melt flow index of a polymer (melt index) it does not respond to a law of linear mixing, but to its logarithm; the aforementioned calculation of P _ + 2 is therefore carried out on the logarithm of this parameter. According to another advantageous variant, the regulation method of the invention comprises the following steps: - entry of instructions relating to one or several properties of the product to be synthesized, in a main algorithm; - input of the step order of the procedure, in an accessory algorithm (slave); - Calculation of the instructions for concentration of the constituents in the reactor by means of the main algorithm, depending in particular on the instructions and measures of product properties, as well as measurements or predictions of the concentrations of the different constituents in the reactor. reactor; - transmission of the concentration slogans calculated by the main algorithm as input quantities to the accessory algorithm; calculation of flow rates of the constituents that enter the reactor, with the help of the accessory algorithm, depending in particular on the step order of the procedure, the concentration slogans and the flow measurements of the constituents that enter the reactor, and transmission of the flow instructions calculated by means of the accessory algorithm to one or several actuators (possibly indirectly, that is to say, by means of one or several regulating organs, for example, FID regulators, which control the actuator or actuators) in order to regulate the flow rates of constituents in the reactor, in which the main algorithm and / or the accessory algorithm are implemented as described above, ie, using the LAG function to calculate the mass of at least one constituent in the reactor.

The main and accessory algorithms are also implemented by means of one or several classical calculation devices. According to an advantageous variant, the calculation set (prediction, control, etc.) of these two algorithms are carried out by the same calculation device. Advantageously, temperature measurements (for example, temperature in the reactor, and / or inlet and / or outlet temperature of a possible cooling fluid) intervene as supplementary input quantities in the prediction and / or control element. Preferably, the accessory algorithm also takes into account measurements of the composition of constituents present in the reactor or leaving it. Advantageously, the regulation method also comprises a calculation step by means of the accessory algorithm, based on the measurements of flow rates, of predictions of concentrations transmitted to the main algorithm to calculate properties predictions that serve as supplementary input quantities in the calculation of concentration slogans. The main and accessory algorithms form a cascade type regulation. It is par- ticularly advantageous that the main algorithm and / or the accessory algorithm are adaptive, ie that some of its parameters are recalculated periodically (at regular or irregular intervals). Such an adaptation makes it possible to guarantee, in particular, that the mathematical model describes as accurately as possible the procedure in its current state, even in the case of modification of certain operating conditions (temperature, pressure, gait, etc.) and in the case of disturbances (poisoning of the catalyst, ...). The main algorithm performs the control of the properties of the product by means of a model based on characteristic equations that associate the properties of the product with the concentrations of the different constituents in the reactor, as well as with the temperature prevailing in the reactor. The accessory algorithm regulates the concentrations of one or more constituents by acting on the feed rates of one or several optionally different constituents. The interest of this "principal-accessory" cascade resides in the fact that the main model accurately determines the concentrations of constituents necessary to obtain the properties desired for the synthesized product, and that the accessory model ensures respect for the values imposed by the principal. Guided by the principal, the attachment is able, therefore: - to quickly bring the concentrations to the values desired by the principal and 'maintain them; - Effectively monitor the progress of the procedure without disturbing concentrations. This "main-accessory" cascade is particularly effective because both the main and the accessory take into account the dynamics of the procedure thanks to the use of the LAG function in the calculus. The accessory algorithm can also be designed to provide predictions of reliable concentrations to the main algorithm. From these predictions or from measurements of concentrations, the main algorithm deduces predictions of reliable properties of the product in the act of being synthesized in the reactor. Comparing these predictions of properties with the property slogans, the main algorithm can, if necessary, intervene and correct the concentration slogans. This correction is possible even before there is a deviation between a variable and its slogan. The taking into consideration of predictions of properties obtained from predictions or from measurements of concentrations makes it possible to considerably reduce the temporal fluctuations of the properties of the synthesized product, resulting in a better constancy of the quality of the product. If the properties of the product to be synthesized are dependent on the temperature in the reactor or reactors, it is preferable to provide a temperature regulation by the accessory algorithm. The latter straightens the thermal balance of each reactor and determines, by means of the calculation of the step of the march, the amount of heat that must be added or evacuated in order to respect the temperature setpoints calculated by the main algorithm. From these results it deduces input slogans for the thermal regulation organs of the synthesis installation. It will be appreciated that this procedure allows intervention on the thermal regulation organs of the synthesis system before the temperature changes. Temperature measurements. they also advantageously intervene as supplementary input quantities in the accessory algorithm. The main algorithm advantageously comprises the following structure: - a prediction organ, based on a direct model of the method that allows to provide a prediction of the properties of the synthesized product as a function of measurements and / or predictions of the concentrations of the constituents; an adaptation organ that compares the predictions of properties calculated by the prediction organ with the values actually measured in the synthesized product, deriving from this comparison adaptation parameters, said adaptation parameters intervening as supplementary input quantities in said prediction body of the main algorithm; and - a control organ, based on an inverse model of the method, for calculating, according to the slogans and predictions of the properties of the product to be synthesized, the concentration slogans for the accessory algorithm, said adaptation parameters also acting as additional input magnitudes in said control body. The accessory algorithm advantageously comprises the following structure: - a prediction organ, based on a direct model of the method that allows to provide a prediction of the concentrations of one or more of the constituents on the basis of a 'material balance in the reactor; an adaptation organ, which compares the predictions of concentrations calculated by the direct model with concentration measurements, deriving from this comparison the adaptation parameters, said adaptation parameters intervening as supplementary input quantities in said prediction organ of the accessory algorithm; and - a control organ, based on an inverse model of the method, for calculating, based on the setpoint of the gait step, the concentration setpoints calculated by the control organ of the main algorithm and the calculated predictions of concentration by the prediction organ of the accessory algorithm, the instructions for the flows entering the reactor, said adaptation parameters intervening as supplementary input quantities in said accessory algorithm organ. The dynamics of the procedure is advantageously described and calculated by means of functions of the type y = LAG (u, z), this function being the solution of the differential equation u = t dv + and dt in which the arguments u and t vary with time. The use of this function in accordance with Theorems 1 and 2 discussed later in this report allows sequentially solving the material balances used by the slave algorithm and describing the kinetics of the procedure by simple characteristic equations in the main algorithm. The LAG function also makes it possible to considerably reduce the volume of the necessary calculations and therefore makes the use of fast and powerful computers useless. In addition, this function allows the direct and inverse models of the procedure or some of its parts to be established in a particularly simple way. The main qualities of the proposed regulation can be summarized as follows: - anticipation: the regulation begins to correct the disturbances measured even before its effect is manifested on the measures of the properties (use of predictions of concentrations, of predictions of properties and of predictions of temperatures in the algorithms), - accuracy even in the presence of disturbances: the direct model and the inverse model are permanently recalibrated using the property measurements (adaptation), - extended validity: the algorithm maintains its validity during the step transitions of the march and of the quality, as well as in the starts and stops (equation of the dynamics of the procedure, use of predictions for the magnitudes whose measurements imply important dead times), - simplicity: development and employment are facilitated thanks to an original method of putting into equation the dynamics of the procedure (LAG function). The procedure of synthesis to be regulated is modeled in this way under a form qu is generally qualified as a "model of knowledge" ("first principie model"), that is to say, that its model is elaborated from equations that reflect the detailed physicochemical development of the procedure. Such an approach makes it possible to obtain, by means of a relatively simple set of equations in the mathematical plane, results superior to those that could be obtained by means of a empirical black box model, especially providing parameters associated with real magnitudes and a better validity on the outside of the identification space (extrapolation). Most empirical models use complex equations, often of high order, if you want to obtain a correct simulation of the dynamics of the procedure, and whose parameters (especially time constants) must be identified for a precise operating point; the model is valid only in the immediate vicinity of this point of operation. Such an approach can hardly be generalized to a large number of operating points in the case of a real chemical synthesis procedure in which numerous quantities are involved. On the contrary, according to the regulation method of the invention, a set of simple, purely static equations is used; the dynamics of the procedure is simulated by simple functions (see, the previous LAG function). Advantageously, the dwell times (time constants of the equations) can be recalculated as frequently as desired, which does not impose any problem given the simplicity of the equations. In the end, we obtain a set of extremely simple equations that are easy to solve in real time, even with a high frequency. The proposed method of regulation is advantageously applicable to synthesis processes, especially continuous, of polymers (polymerization), and in particular to the continuous polymerization of olefins such as, for example, ethylene or propylene, both in the liquid phase as in gaseous phase. The present invention also relates to a process for the synthesis of one or more chemical products, regulated by means of a regulation method according to the invention. In particular, good results have been obtained for the regulation of a continuous synthesis process of polyethylene by polymerization of ethylene in at least one reactor, the reactants comprising ethylene, hydrogen and / or an optional comonomer, the reaction taking place. of polymerization in a solvent in the presence of a catalyst and removing a portion of the contents of the reactor permanently or intermittently. This procedure can also be developed both in the liquid phase and in the gas phase; it preferably develops in liquid phase (in a solvent). The method is applied analogously to the synthesis of polypropylene (the main starting material in this case being propylene instead of ethylene), propane can also be present if the process is carried out in the gas phase. For polypropylene, the melt flow index is often designated by MFI instead of MI. The invention also relates to a regulating device intended to implement the regulation method of the invention, as well as to a synthesis device for one or more chemical products comprising such a regulation device. More precisely, the invention also relates to a device for regulating a process for the synthesis of a chemical in a synthesis device comprising at least one reactor, the device comprising: - at least one calculation unit; - means to enter slogans of properties of the product to be synthesized in the calculation unit; means to enter the calculation unit slogans of the step of the march of the product to be synthesized; - bodies for measuring the flow of the currents entering the reactor; - organs for measuring the composition of the streams leaving the reactor; - flow control devices (actuators) for regulating the flow rates of the currents entering the reactor; - means of communication between said calculation unit, said flow measurement organs and said regulating organs; wherein: - the mass of at least one constituent is calculated by the LAG function as stated above: - the calculation unit is capable of calculating with reference to a main algorithm, according to the slogans of properties, slogans of concentration of reactants in the reactor; - the calculation unit is suitable for calculating with the aid of an accessory algorithm, according to the production slogans and the concentration slogans, flow rates for the currents entering the reactor, these flow rates being transmitted as entry slogans in the flow regulation bodies; the measurements carried out by the flow measurement bodies intervene as supplementary input quantities in said accessory algorithm, in order to allow the latter to calculate, based on these flow measurements, predictions of concentrations, these concentration predictions intervening in the main algorithm for calculate predictions of properties used as supplementary input quantities in the calculation of concentration slogans. The invention also relates to a regulating device such as that described above, in which: - the synthesis device further comprises: thermal regulation devices suitable for controlling the temperature in the reactor, and - temperature sensors; - the main algorithm is apt to calculate, according to the property setpoints, temperature setpoints for the reactor, - - the accessory algorithm is suitable for: - calculating a thermal balance for the reactor, - solving this or these thermal balances for determine the heat that needs to be added, respectively exclude from the synthesis to respect the temperature setpoints, and deduce from this or these thermal balances input slogans for the thermal regulators of the reactor, and receive as additional input quantities the measurements made by the temperature sensors. The invention also relates to a device as described above, in which measurements made by the temperature sensors act as supplementary input variables for the main algorithm. The invention also relates to a device as described above comprising: - at least one analyzer capable of providing measures of the properties involved in the main algorithm; and - means for entering these measures of properties in the calculation unit; said calculation unit comprising: - a first prediction organ, based on a first direct model of the procedure that allows the prediction of the properties of the synthesized product, based on the predictions of concentrations calculated by the accessory algorithm; - a first adaptation organ, which compares the predictions of properties calculated by the first prediction organ with the values actually measured on the synthesized product, deducting adaptation parameters from this comparison, which intervene as supplementary input quantities in said first evaluation organ. prediction; and a first control device, based on a first inverse model, for calculating, according to the setpoints and for the predictions of properties, concentration setpoints for the accessory algorithm, said adaptation parameters also intervening as supplementary input variables in said first control body,. The invention also relates to a device as described above, further comprising: - at least one analyzer capable of providing measures of concentration of the reagents; and means for entering these concentration measures into the calculation unit; said calculation unit comprising: a second prediction organ, based on a second direct model that allows the prediction of concentrations, as a function of the balance of matter in the reactor; - a second adaptation organ, which compares the predictions of concentrations calculated by the second prediction organ with the concentration measurements, deducting from this comparison other adaptation parameters, which intervene as supplementary input quantities in said second prediction organ; and - a second control organ, based on a second inverse model, for calculating, based on the production slogans, the concentration slogans calculated by the main algorithm and the predictions of the concentration of the second prediction organ, the slogans for the flows entering the reactor, said other adaptation parameters intervening as supplementary input quantities in said second control element.

BRIEF DESCRIPTION OF THE FIGURES A specific embodiment of the invention is illustrated on the basis of a continuous synthesis process of polyethylene (PE), referring to figures 1 to 10. These show: Figure 1: a schematic of a polyethylene manufacturing circuit; Figure 2: a simplified diagram of the structure of an advanced regulation according to the invention; Figure 3: a schematic diagram of the advanced regulation applied to the manufacturing circuit of Figure 1; Figure 4: a schematic diagram of an adaptive control algorithm, as used in the advanced regulation system according to Figure 2. Figure 5: a schematic of the structure of the main algorithm in the advanced regulation system, according to the Figure 2; Figure 6: a diagram of the structure of the accessory algorithm in the advanced regulation system according to Figure 2, - Figure 7: the general scheme of a regulation method according to the invention, - 33 Fig.8-10: the scheme of particular variants of the method of the invention.

DETAILED DESCRIPTION OF THE INVENTION On figure 7, the synthesis procedure itself (Pr) is distinguished in the first place, which can be controlled by imposing at least one command value (CGC) command (for example one or several flow rates of constituents entering the control unit). reactor, a temperature, etc.) to a suitable actuator (valve, heating or cooling device, etc.). The regulation is carried out by means of a control organ (OC) based on the inverse mathematical model of the procedure, and whose main role is to compare the setpoint or slogans of the regulated quantities (CGR) (for example one or several properties of the product to synthesize and / or one or several magnitudes associated with the development of the procedure) with the prediction or predictions of these magnitudes (PGR). The prediction or predictions of the regulated magnitudes (PGR) are calculated by a prediction organ (OP) based on the direct mathematical model of the procedure, based on measurements of the control magnitudes (MGC). It will be appreciated that no measure of ownership (s) of the synthesized product is involved in regulation. Figure 8 represents a variant of the method of Figure 7, in which the mathematical model of the procedure is periodically adapted by an adaptation organ (OA), on the basis of the deviation (advantageously filtered or numerically treated) between the predictions ( PGR) and the measurements (MGR) of the regulated quantities. A resynchronization (reengagement in time) of these measurements and of these predictions is frequently necessary, for example when it comes to measures of properties of the synthesized product whose duration of obtaining is long. The adaptation organ (OA) transmits the results of its calculations, that is, its adaptation instructions, to the prediction organ (to adapt the direct model of the procedure) and to the control body (to adapt the inverse model of the procedure). It will be appreciated that the property measure (s) of the synthesized product are not taken into account but in the adaptation processes, which generally proceeds with a frequency much smaller than the normal control process. The eventual slowness of these measures therefore has no direct effect on the quality of regulation.

Figure 9 represents another variant of the invention, in which one or several measures of regulated magnitudes (MGR) are taken into account by the control body (OC), and one or several measures of regulated quantities (MGR,) (possibly different) are taken into account by the prediction body (OP). Also, one or several measures of control quantities (GG,) can also be taken into account by the control body (OC). It goes without saying that another variant of the method of the invention could be created by combining the variants of figures 8 and 9, that is to say, using at the same time an adaptation element and taking into account one or several measures of control variables in the control element, and / or one or several control variables in the prediction element and / or in the control element. In figure 10, the mathematical model of the procedure is not adapted, in the proper sense of the word, but the deviation (advantageously filtered) between the measurements and the predictions of the regulated quantities is used to correct the setpoints of the regulated quantities (CGR) In this circumstance, this correction has been represented here as a simple difference: a corrective term calculated by the adaptation body OA is subtracted from each of the slogans of regulated magnitudes (which is, in effect, not more than an organ of correction), which provides corrected setpoints CGR ', transmitted to the control body OC. Needless to say that in certain cases, the correction may involve more complex operations than a subtraction, for example a division (in this case, it can, however, be reduced to a subtraction considering the logarithms of the magnitudes considered). This method is commonly called Internal Model Control (IMC). If reference is made to Figure 1, which schematically represents a polyethylene (PE) continuous synthesis circuit, the polymerization of the ethylene is carried out in a loop reactor 10, in suspension in a suitable solvent, such as, for example, Hexane The process is continuous, that is, the reagents are continuously injected, and a portion of the contents of the reactor 10 is continuously extracted. A circulating pump (not shown) ensures the homogeneity of the contents of the reactor 10. The reagents introduced into the reactor are etiieno "Et", hydrogen "Hy" and butene "Bt" (see reference 11). A catalyst is also injected continuously. It is important to control the concentrations of reactants in the reactor, because the properties of the PE resin are determined mainly by the concentrations of Hy / Et and Bt / Et concentrations. The polymerization temperature in the reactor is an additional parameter that influences the properties of the PE resin. Since the polymerization reaction is strongly exothermic, the temperature of the reactor must be regulated using a cooling circuit 12. The reactor 10 in operation therefore contains solvent, polymer and reagents that have not yet reacted, and catalyst. Its content is permanently extracted through the extraction duct 14. This extracted content enters a STP 16 stripper, which separates the PE polymer and the fluids (solvent and reactants). These fluids are vaporized by injection of water vapor and evacuated to a CD 18 condenser. In the latter, the solvent condenses again before being recycled. Reagents that are lighter are separated from the solvent and are also recycled. A GC gas chromatograph (20) placed at the outlet of the condenser 18 allows to determine the Hy /? T, Bt / et concentrations of the reactants. The polymer expelled from desorber 16 is concentrated in a CFG 22 centrifuge and then dried in a SHLF 24 fluid bed dryer, before being sent to the termination to be granulated. Samples are taken at the outlet of the dryer 24 to measure the properties of the resin: crystallinity (measured by the specific density "MVS") and rheological properties (melt flow index ("melt index (MI)" or "melt flow"). index (MFI) ") and melt viscosity" μ2", measured in a capillary viscometer under a shear stress of 100 s" 1) The dynamics of this P? synthesis process are slow and complex: - The reactor loop 10 behaves like a perfectly mixed reactor, therefore, any change in the feed rate of one of the reactants will only translate progressively on the concentration of this reagent in the reactor., the new flow rate must be mixed in the total volume of the reactor 10 to bring it to the new equilibrium concentration. The measurement of the concentrations of the reagents is carried out by means of a gas chromatograph 20; it is a discontinuous device, which proceeds in successive stages: removal of a gas sample, analysis, and then elaboration of the results. There is, therefore, a dead time (5 to 15 minutes) between the changes in concentration and its measurement. - The properties of the polymer manufactured at each instant depend: pi incipalmente of the concentrations of the reagents. Any modification of these concentrations, therefore, instantaneously affects the properties of the polymer manufactured. On the other hand, the average properties in the reactor are modified only gradually, because the newly produced polymer must be mixed with the polymer already present in the reactor 10 (residence time: ± 2 h). - When the polymer is extracted from the reactor 10, it undergoes again a series of mixtures in the different apparatuses (STP, CFG and SHLF) destined to dry it (dwell time: + 2 h). Then samples of the polymer are taken and analyzed by the factory laboratory. The results of these analyzes will not be communicated, therefore, to the manufacturing process but after a new dead time, which may be important (± 2 h).

Modeling of the procedure with the aid of a LAG function According to the method of the invention, the dynamic modeling of a continuous synthesis process is carried out by resorting to hypotheses of perfect mixtures and net delays. The perfect mixes are put into equation by means of a well-known function of the engineers, the "LAG" function, or low-pass filter (passe-bas) (of the first order), - this function is linear and easily programmable.

It is defined as follows: y = LAG (u, t) (it is said "LAG of u during t") as being the solution of the differential equation dy u = t. + and dt whose arguments u and C vary with time. This equation can be solved numerically (even in real time) by means of a first order algebraic equation, which has as arguments the following variables: - the preparation period of the samples "T" (or time elapsed since the last iteration) - the time of permanence (or "time constant)" t "at the instant" t "the state variable" and "at the previous instant" tT "- the command variable" u "at the instant instant" t "(representing uyt , in effect, the values measured or calculated at time "t" on the magnitudes "u" and "t" that are assumed to have been checked over the entire preceding interval "T".) Preferably, T is small with respect to (for example at least 10 times smaller) to increase the accuracy of the calculation.

The solution of the aforementioned equation can be calculated, for example, by the following formula: y (t) = y (t-T). e ~ T / t: (t :) + u (t). (1 - e ~ t / t {)) Or even more simply (roughly): T y (t-T) + u (t) t (t) y (t) = 1 + t (t) The modeling of the procedure with the help of the LAG function is based on the following theorems: Be a perfectly mixed reactor (CSTR) of volume VR. Several constituents (reactive or inert) feed it, among which is the reagent "x" (input flow FxrN) that has the property "P JN" a ^ a input. The output flow "Fout" (extraction) is also measured.

Theorem 1: Application of the LAG method to the calculation of a mass balance: - At each instant, the mass "MxR" of a constituent "x" in a perfectly mixed reactor (CSTR) is equal to LAG i of the mass flow product entering "FxIN" for a time "t?" during this same time cx • MxR = LAG (FxIN.t ?, t?) (in kg) Time "t?" is the "dwell time of x"; it is worth the mass of the constituent MxR, divided by the sum of the "outgoing" mass flows (amount consumed by reaction "FRx", flow leaving the reactor "Fxout", etc.). t? = MXR / (FX0IJT + FRx + ...) (in h) This theorem provides an exact method for the dynamic calculation (even in real time) of the concentrations in a perfectly mixed reactor. In effect, the "CxR" concentration of constituent "x" expressed in kg / pr is valid, if "VR" is the reactor volume: CxR = MXR / VR (in kg / m3) Knowing the total volumetric flow rate "FVOUT" leaving the reactor, the residence time of the "tR" reactor is defined: ^ "VR / FV0UT Therefore, the mass flow" FxQrjrp "of the constituent" x "at the reactor outlet (extraction) is valid: FX0UT = MXR / R It will be appreciated that if "x" is an inert (not subjected to reaction and that does not leave the reactor except by extraction), we have: t? = tR Furthermore, in the frequent case in which the reaction speed of "x" is proportional to its concentration CxR, with a proportionality factor R ?, we have: FR? = R ?. MxR and therefore t? = 1 / (R? + L / tR) Theorem 2: Application of the LAG method to the calculation of a property of a mixture: Let a property "Px" of a constituent that responds to the law of linear mixing: PXl + 2 = wl- Pxl + w2 • Px2 where - ^ and w2 are the mass fraction of property Px- | _ and Px2 (where w-_ + w2 = 1). At each instant the property PxR in a perfectly mixed reactor (CSTR) is equal to the LAG of the property at the input PXIN 'during a dwell time that is worth the ratio of the mass MxR of the constituent in the reactor divided by the mass flow rate. enters (and / or appears by reaction) FxjN: PxR = LAG (PxIN, MxR / FxJN) It is thus possible to take into account the dynamics of the procedure, and permanently recalculate your time constants. As stated above, the "properties" in question here may, in certain cases, have undergone a mathematical transformation by making them linear (for example, the logarithm of the melt flow index of a polymer can be considered to respond to a law of linear mix) '.

Principle of regulation When the model of the synthesis procedure is established, an algorithm is needed to calculate the parameters necessary for the regulation of this procedure. Figure 2 shows a simplified general scheme of the advanced regulation type (Advanced Process Control or "APC") adopted for the polymerization process described above. It is appreciated that this regulation system comprises a cascade of two algorithms, guiding this cascade, in particular, the PID regulators of the reagent feed rates. The two cascade algorithms, called the main algorithm and the accessory algorithm (slave), are the two adaptive dynamic algorithms based on models derived from process knowledge (as opposed to empirical models), based in particular on the material balances and the kinetics of the regulated procedure. They use the LAG function defined above. In Figure 3, which shows the principle of the regulation system in the context of the polymerization process described above, it is seen that: the main algorithm is based on the equations characteristic of the catalysts, ie equations that provide the properties of the PE in function of polymerization temperature and reagent concentrations in the reactor, - this algorithm provides the accessory algorithm with the setpoints of reagent concentrations to satisfy the setpoints for PE properties; the accessory algorithm is based on a balance of matter and the chemical kinetics of the reactions; This algorithm provides the flow rate regulators of the reactants needed to satisfy the concentration orders imposed by the main algorithm and the step slogan of the procedure. Preferably, it also calculates an anticipative thermal ("feed-forward") for the temperature setpoint, improving the temperature regulation during changes in the pace of the march.

This type of regulation is not of perfect precision more than if the model is perfect and takes into account all possible disturbances. This is not the case, in general. This is why, in general (see Figure 4), the direct model (and the inverse model) is continuously adapted by comparing the predictions with the measurements of the properties. This "adaptation" of the model allows to maintain its accuracy in the presence of non-modeled disturbances, and therefore obtain a more precise regulation in all cases.

The accessory algorithm Figure 6 illustrates the principle of the accessory algorithm: 1. a prediction organ that uses a direct model of the procedure predicts, from the measurements of the flows that feed the reactor, the concentrations of reagents and polymer; 2. An adaptation organ compares the concentrations of ethylene (Et), hydrogen (Hy) and butene (Bt) measured by an analyzer (gas chromatograph) with the values predicted by the direct model, to determine three adaptation parameters: - The specific activity of the catalyst for the "AEt" etiien, in kg / h of polyethylene per kg of catalyst and per kg / m3 of etiieno, - the gain error on the measurement of the flow rate of the hydrogen "KfHy" - the purity of the butene feed "KfBt" 3. the control organ calculates, from the concentration slogans calculated by the main algorithm and the setpoint of the gait step, the setpoints for the reactive feed rates, - these flow rates are composed of a feed-forward based on the inverse model, and a feed-back proportional to the deviation between the direct model and the concentration slogans. To understand the calculations made by the accessory algorithm, in the case of a polyethylene synthesis procedure, it is first necessary to know that it is generally accepted that the polymerization rate "VitPEt-" is proportional: - to the concentration of unpolymerized Et ? tR (in kg / m3), - at the concentration of active catalyst in the cCAR reactor (in kg / m3), and at a proportionality factor, a (hardly quantifiable) function of the temperature, the concentrations of Hy, Bt and co-catalyst, the presence of poisons, etc. This term is called "catalytic activity" for the AEt etiieno. Except for major dysfunction (poison, etc.), it varies reasonably little during the course of the operation. I saw Pt? T = AEt _ cCAR. cEtR (kg / m3.h) The amount of ethylene "FpEt" that is polymerized in the reactor in each unit of time ("polymerization flow") is, therefore, VR being the volume of the reactor: FpEt = Vit.PEt . VR (kg / h) = AEt • cCAR • MEtR in which MEtR is the mass of ethylene in solution in the reactor (in kg). It is also known that the rate of incorporation of Hy is approximately 100 times slower than that of Et, and 10 times slower for Bt. It follows from this: FpHy = AEt. cCAR MHYR / 100 FpBt = AEt._ cCAR. MBtR / 10 in which MHyR is the mass of hydrogen in solution in the reactor (in kg) and MBtR is the mass of butene in solution in the reactor (in kg) The accessory model uses the following measurements in the meantime: FEtt «= flow ethylene feed (monomer) (kg / h) FSvIN = solvent feed rate (hexane) (kg / h) FCAJ J = catalyst feed rate (kg / h) FHyi = hydrogen feed rate (transfer agent i g / h) FBtIN = butene feed rate (comonomer) (kg / h) It also uses the following adaptation parameters: AEt = "catalytic acti-vi" for the etiieno KfHy = gain error on the measurement of the hydrogen feed KfBt = purity of the butene feed. The following calculations are carried out sequentially, in the following order, with a high frequency (the interval of time that separates each iteration from the smallest of the residence times t being small).

Since the volume of the reactor is constant, the volume flow, which leaves is equivalent to the flow, in volume, that enters incompressible fluids). The flow, in volume, that comes out "V0TjT" can be calculated, therefore, as the sum of the mass flows that enter divided by the specific density they have in the reactor: • FV0UT = FSvIN 650 + FEtIN / 950 + FBt N / 600 (m3 / h) (in which the densities are the following: 650 kg / mJ for the solvent, 950 kg / m for the polyethylene, 600 kg / m3 for the butene). The hypothesis is made here that all ethylene is instantly transformed into polyethylene, and the flow of hydrogen and catalyst (some kg) is discarded. The solvent is chemically inert and does not leave the reactor except by extraction from the reactor. The mass "MSvR" in the reactor is calculated using Theorem 1: ^ R = VR / FVOUT ^) (residence time in the reactor) MSvR = LAG (FSvIN _ tR, tR) (kg) The catalyst is deactivated with a time constant "kd"; the "MCAR" mass of active catalyst in the reactor is calculated as follows: tCA = 1/1 / tR + kd) (h) MCAR = LAG (FCAIN, CA CA) (kg) bC and therefore the concentration "cCAR" of active catalyst in the reactor is worth: cCAR = MCA / VR (kg / m3) The ethylene "leaves" the reactor in the extraction flow as well as in the polymerization reaction. Its mass in the "MEtR" reactor is calculated as follows: tEt = 1 /! L / cR + AEt. cCAR) (h) MEtR = LAG (FEtIN, tEt, tEt) (kg) In a similar way, the "gross" mass is calculated (uncalibrated) of hydrogen in the "MHyRAw" reactor: tHy = 1 / (l / tR + AEt cCAR / 100) (h) MHyRA = AG (FHyiN • CHy 'Hy] (k9 > The mass "MHyR" corrected to take into account the gain error on the measurement of the hydrogen supply, is valid: MHyR = KfHy. MHY ^ (kg) The ratio Hy: Et in the reactor is therefore worth: HyEtR = MHyR / MEtR The butene mass is calculated in a similar way "gross" "MBTRA" tBt = 1 / (l / tR + AEt cCAR / 10) (h) MBtRAW = AG (FBCIN • CBt 'CBt.}. (K9> The mass "MütR", corrected to take into account the purity of the butene feed, is worth: MBtR = KfBt. MBtRAW (kg) The relationship. Bt: Et in the reactor is therefore worth: BtEtR = MB- / MEtR It has been seen that the polymerization flow rate "FpEt" (step of the instantaneous gear) is worth: FpEt = AEt. cCAR MEtR (kg / h) Since the polymer is inert and not subject to reaction, its mass in the "MPER" reactor is: MPER = LAG (FpEt. TR, tR) (kg) The flow rate of polymer leaving the reactor "FPE0UT" is therefore: FPE0UT = MP? R / R (kg / h) Adaptation of the accessory model: The adaptation block uses an analyzer (for example a "GC" gas chromatograph) to obtain the measurements of the concentrations in the "cEtGC" reactor., of hydrogen "cHyGC" and of butene "cBtGC" (expressed, for example, in kg / m). These measurements are compared with the values predicted for the direct model, in order to determine the following three adaptation parameters: - the specific activity of the catalyst for the "AEt" etiien, in kg h of polyethylene per kg of catalyst and per kg / m3 of etiler.o, - the gain error on the measurement of the hydrogen flow rate "KfHy", - the purity of the butene feed "KfBt". Gas chromatography provides measurements of the samples taken with a delay of about 6 minutes. The calculation of the specific activity for the ethylene "AEt" is deduced from the following equations: cETGC = cEtR (kg / m3) = ME R / VR = LAG (FEtIN, tEt) / VR = LAG (FEtIN / (l / tR + A? t cCAR), tEt / VR -AEt "= (F? TIN / LEAD (cEtQC .VR .tEt) -l / tR) / cCAR The measure cEtGC (referred to the sample and with" noise ") is not to be found in a LEAD, therefore it is used rather: Aet '= LAG (AET ", tE¿) = LAG (FEtIN / cCAR, Et) / (cEtGC-Vp - 1 / The count of the 6 minute delay over the measure of cEtGG is entered by entering two LAGs in series of > 3 minutes each on the values of the model, obtaining the final formula: Aec = LAG (FEtIN / cCAR, t £ t, 3/60, 3/60) / (cEtGC.VR) 1 / LAG (tR.CCAR, tEc , 3/60, 3/60) The calculation of the hydrogen flow gain "KfHy" is deduced from the following equalities: cHyGC = cHYR (kg / m3) = MHYR VR = KfHy. MHYRAW / VR é > KfHy = cHyGC. VR / MHYRAW The delay of 6 minutes on the measure of cHyrG is taken into account by introducing two LAGs in series of 3 minutes each on the value of the model, obtaining the final formula: é > KfHy * cHyQC. VR / LAG 3 / 60,3 / 60) In a similar way to KfHy, the correction parameter for the purity of the butene "KfBt" is calculated: KfHy = cBtQC. VR / LAG (MBtRAW, 3/60, 3/60) The regulation algorithm has as entries: - the concentration instructions calculated by the main algorithm; more precisely the slogans for concentration ratios cHyR / cEtR '? eEtop "and cBtR / cEtR" BtEtSp "(in kg / kg) OR the setpoint for the passage of the procedure FpEtgp, set by the operator - the setpoint for the concentration of etiieno cEtSp set by the operator; - the concentrations calculated by the model. It calculates the setpoints for the feed rates of the reagents FEtgp, FCASp, FHy-SP and FBtSp. Different algorithms can be used, among which is the MBPC (Model Based Predictive Control). In general, they can be decomposed into a feed-forward based on the inverse model, and a feed-back proportional to the deviation between the direct model and the concentration slogans.

Feed regulation: Feed-forward: value to maintain the current concentration, based on the inversion of the stationary value of the following equation: MEtR = LAG ((FEtIN Et, t £ t) (kg) é> FEtFF = MEtR / tEt Feed-back: proportional to the deviation between the setpoint cEtESR and the model FEt FB = 5 (cEt SP Vt MEt R 'Setpoint FECSP = FEtFF + FEtFB Regulation of the feed of catalyst: Feed-forward: value to maintain the current concentration, based on the inversion of the stationary value of the following equation: MCAR = LAG (FCAIN, tCA, tCA) (kg) é> FCApp = MCAR / tCA Feed-back: proportional to the deviation between the setpoint FpEtsp and The model, according to the following formula: FpEt = A £ t MEtR MCAR / VR é> FCAFB = 5 • < FpEtsp / (AEt. MEtR / VR) - MCAR Slogan: FCAsp = FCAFF + FCApB Regulation of the hydrogen supply: Feed-forward: value to maintain the current concentration, based on the inversion of the stationary value of the following equation: MHyR = LAG (FHYJN. THy, tHy) (kg) é > FHyFF = MHYRAW / tHy Feed-back: proportional to the deviation between the HyEtsp relation setpoint and the model. FHYFB = 5. (HyEtgp. MEtR - MHyR Setpoint: FHysp = FHyFF + FHypB Regulation of the butene feed (similar to hydrogen): FBtpp = MBtRAW / tßt FBtF3 = 5. (BtEtSp. MEtR. 'MBtR) Btgp = FBtpp + FBtp The above equations summarize the equations of the accessory algorithm, executed every 10 seconds by the numerical control and command em (SNCC).

The main algorithm Figure 5 illustrates the principle of the main algorithm: 1. its predictive organ (based on a direct model) predicts the main properties of the polymer (MI and MVS), - uses for this the measurement of the polymerization temperature, the predictions for the concentrations in the reactor, provided by the accessory model, and the residence times of the PE in the different apparatuses; 2. its adaptation organ compares the measurements of MI and MVS made (at the exit of the dryer as well), either by the laboratory of measurement every 2 hours, or by an analyzer continuously, with the values predicted by the direct model , to determine the 2 adaptation parameters that are corrective parameters, multiplicative for the MI and additive for the MVS; 3. its control body (based on an inverse model) calculates, based on the slogans of MI and MVS provided by the operator, the instructions for the concentrations in the reactor (ratios Hy: Et and Bt: Et). As for the accessory algorithm, this calculation is composed of a feed-forward based on the direct model, and a feed-back proportional to the deviation between the direct model and the operator's slogans. For a given catalyst, the properties of the resins at steady state are functions of the polymerization temperature and the concentrations of the reactants. Among the different static equations outlined in the scientific literature, the following equations have been chosen: log (MI) = aQ + a1. T ° + a2. log (Hy / Et) + a3.Bt/Et MVS = bQ + b. T ° + b2. (Bt / Et) b3 + b4. log (MI) The parameters aQ a a ^ and bQ a b4 are obtained by identification in steady state, for various resins manufactured with the same catalyst. In addition, the different devices found by the polyethylene until the moment in which its properties are measured (reactor, desorber, centrifuge and then 0 dryer), they can all be assimilated in a first approximation to perfectly mixed reactors. The main algorithm uses the following measurements as inputs: TR = temperature in the reactor (° C) Vstp = volume of liquid in the desorber (obtained by level measurement) (m3) MIMES = measure of MI (melt-index) MVSMES = measure of MVS (Specific Density) as well as the following calculations made by the accessory algorithm: FpEt = instantaneous polymer production (step of the march) (kg / h) FPE0UT = flow of PE leaving the reactor (kg / h) MPER = mass of PE in the reactor (kg) HyEtR = ratio of Hy to Et in the reactor (kg / kg) BtEtR = ratio of Bt to Et in the reactor (kg / kg) The gross instantaneous values (before adaptation) of the MVS and the logarithm of the MI ( "IMI" are calculated by: MVSINS = b0 + b1, TR + b2. (BtEtR) b3 + b4.IMIINS 1MIINS = aQ + a1. TR + a2. log (HyEtR) + a3.BtEtR Using theorem 2, the gross average properties at the output of the reactor are calculated: lMIr = LAG (1MIINS, MP £ R / FpEt) MVSr = LAG (MVSINS, MPER / FpEt) In effect: - the 1MI and MVS properties respond quite well to a linear mixing law the loop reactor can be assimilated to a perfectly mixed reactor - the mass flow of PE "incoming" (which appears) in the reactor is ef fi ciently FpEt, the amount of PE that polymerizes at every moment (step of the march). Gross properties for the measurement: Knowing that the desorber has around 500 kg of PE per m3, and if the hypothesis that the desorber is a perfectly mixed reactor is formulated, the gross properties at the output of the desorber are calculated as follows : lMIstp = LAG (lMIr, 500. Vstp / FPE0UT) MVSstp = LAG (MVSr, 500. Vstp / FPEQUT) Since the residence time in the centrifuge is very short, it can be ignored. The dryer is a fluid bed; permanently contains around 1,400 kg of PE. The hypothesis can be formulated that the level in the desorbedor changes little, and that the flow that leaves is equal to the one that enters. Therefore, the flow of PE entering the dryer is worth FPEQTJT- Then, at the dryer exit, with respect to where the sample is taken for the measurement of property, the following raw values: lMIsh = LAG (lMIstp, 1400 / FPE0TJT) MVSsh = LAG (MVSstp, 1400 / FPE0TJT) The properties after adaptation are obtained by intervening the kMI adaptation parameters (multiplicative parameter) and kMVS (additive parameter), - the properties after adapting the output of the reactor, desorber and dryer are valid, therefore: MIrc = kMI. 10lMIr MVSrc = kMVS + MVSr MIstpc = kMI. 10lMIstP MVSstpc = kMVS + MVSstp MIshc = kMI. l0lMIsh MVSshc = kMVS + MVSsh Adaptation of the main algorithm: Property measurements take a certain time to be completed (± 5 min if there is an on-line analyzer, ± 1 h if performed by the laboratory). In order to calculate the adaptation parameters, it is necessary, therefore, to resynchronize (re-lock in time) the gross predictions of the model with the measurements. This can G7 be made for example by means of a phase shift register called here "DELAY function": 1MIDEL = D? LAY (lMIsh, tMI) MVSDEL = DELAY (MVSsh, tMVS) where tMI and tMVS = i 5 min or ± 1 h, according to whether the measurement is carried out by an analyzer continuously or by the laboratory. Once the new measurement of MI or MVS is reached, the raw adaptation parameter kMI 'or kMVS1 is recalculated, comparing the value of the resynchronized gross model with the measured value: kMI' = log (MIMES). 1 IDEL kMVS '= MVSMES - MVSDEL These values are filtered in order to attenuate the precipitated reactions that the possible perturbations (noise) of the measures impose in the procedure: kMI = LAG (kMI', ± 1 h) kMVS = LAG ( kMVS1, ± 1 h) Control organ The control body has the values MIsp and MVSgp entered by the operator. This organ calculates the slogans for the ratios of the concentrations in the HyEtgp and BtEtgp reactor necessary to quickly obtain the desired properties MIsp and MVSgp. This calculation is carried out in 2 stages: 1. the control body calculates, from the slogans MIsp and MVS3p provided by the operator and from the values after adaptation of the MI and the MVS in the different devices, the slogans MIisp and MVSigp for the instantaneous production. These instant slogans are composed of a feed-forward and a feed-back proportional to the deviation between the direct model and the operator's slogans. 2. the instructions for the concentration ratios HyEtgp and BtEtgp are then calculated by inverting the static equation previously used for the calculation of the instantaneous value of the MI and the MVS. Instructions for instantaneous properties: The properties at the output of the dryer are compared with the property instructions, to determine the desired setpoints for the properties at the output of the desorber. (the centrifuge is discarded): MIstp = l0 (log (MISP) + 0, l (log (MISP) -log (MIshC)) j MVSstpgp = MVSgp + 0.1. (MVSgp - MVSshc) Also, from the deviation between these setpoints to the output of the desorber and the values calibrated in the desorber, the desired setpoints are calculated at the output of the reactor: MIr _ 10 (log (MIstpSP) +0, 5. (log (MIstpSP) -? Jr log (MIstpC)) j MVSrSp = MVSstpgp + 0.5. (MVSstpSp - MVSstpc) Finally, from the deviation between these instructions to the output of the reactor and the corresponding calibrated values, the desired slogans for the instantaneous production are calculated: Mli = 10rlo9 (MIrSP> + 2- dog (MIrSP ) -log (MIrC))} tr MVSigp = MVSrSp + 2. (MVSrSp - MVSrc) Instructions for merger relationships: The slogans for the concentration ratios HyEtsp and BtEtgp are obtained by inverting the static equation previously used for the calculation of the instantaneous value of the MI and the MVS, substituting in the terms MI and MVS the desired slogans for the instantaneous production and applying the adaptation parameter. Starting from: Log (MIiSp / kMI) log (HyEtgp) + a3. BtEtR MVSigp - kMVS = bQ + b1.TR + b2. (BtEtgp) b3 + b4.IMIINS you get: a2. log (HyEtgp) = log (MIisp / kMI). (aQ + a-L -TR + a3. BtEtR) which gives: HyEt =? o < (lo9 (MIisp / kMI) -aO-al.TR-a3.BtEtR) / a2 y = b2 (BtEtgp) b3 = MVSiSp-kMVS- (b0 + b1, TR + b4, 1MIINS, which gives: BtEtgp = < (MVSigp - kMVS - bQ, b, TR - b4) The equations above summarize the equations of the main algorithm These equations are executed every 30 seconds by the SNCC With this procedure, it is possible to direct the polymerization with great precision, in particular: the controlled properties (MI and MVS) are maintained, at most, very close to the desired values, with a minimum dispersion - the quality changes (and therefore of the MI and MVS properties) are carried out quickly and precisely - the polymerization starts and stops, as well as the step changes of the process step are carried out in an accelerated manner, maintaining the MI and the MVS at all times very close to the desired values, even though the regulation method according to the invention has been presented with the aid of a polyethylene synthesis process by polymerization of ethylene. in a continuous manner, it is understood that this method of regulation will be effective, generally, for other synthesis procedures, and in particular for the procedures that have one or more of the following characteristics: a multivariable regulation is necessary because they influence the whole of properties to regulate several variables; - the dynamics of the process is slow: serial mixtures, important dead times; - the measurements of the properties are subjected to sampling with a small frequency and / or have noises; the regulation must be dynamic, that is to say valid whatever the step of the procedure, as well as during the transitions of the gait and quality (properties) of the product to be synthesized; - It is interesting to estimate certain variables not measured directly. In order to be easily implemented with the presented techniques, it suffices that: - the static equations of the process are known (frequently they are at least to a certain extent, otherwise the procedure could not be regulated); - the dynamics of the process can be approximated by perfect mixtures and dead times; - the necessary measures are available and of sufficient quality (in particular, the flow rates of the reactants and the flows through the deposits involved) The particular use of the LAG function described above, especially in Theorems 1 and 2, may to be extended, of course, to regulation methods based on a structure different from the one that has been exposed, which comprises a different main algorithm and an accessory algorithm. It can be applied, for example, in a regulation method that involves no more than a single algorithm.

Examples: 8 polyethylene (PE) synthesis tests of 4 different types (defined by their MI, MVS, ...) were carried out, respectively using a classical regulation method and using the method of the invention. The table below summarizes the checks that have been made on the basis of numerous measurements of melt flow index (melt index) of the 8 polymers obtained. Cpk designates the centered capacity index of the procedure.

It is verified that the capacity index Cpk is more than double thanks to the use of the method of the invention, which indicates that the properties are about twice less dispersed and / or better centered with respect to the imposed slogans.

Table of the abbreviations used to parameters for the static equation of MI (I = 0 to 3) bt parameters for the static equation of the MVS (I = 0 to 4) * Et catalytic activity for the etiieno (m3. Kg "1. .! "1) ex GC concentration of" x "obtained by the measure of the analyzer (kg / m'3) cxR concentration of" x "in the reactor (kg / m3) cxsp slogan for the concentration of" x "in the reactor (kg / m3) FDX mass flow polymerization of "x" (step of the march) (kg / h) FV, OUT volumetric flow leaving the reactor (ra3 / h) FxIN entering mass flow rate of "x" (kg / h) Fx OUT mass flow rate of "x" leaving (kg / h) kd catalyst deactivation constant (l / h) KfBt corrective parameter (adaptation) for butene KfHy corrective parameter (adaptation) for hydrogen corrective parameter kMI ( adaptation) for the MI kMVS corrective (adaptive) parameter for the MVS LAG (,) first-order low-pass filter function MY MONTH measured from the MI (Melt Index) (melt flow index) MI and MI (Melt Index ) gross (uncalibrated) in "y" MI c MI setpoint (calibrated) for "y" MiySP MI setpoint (calibrated) for "y" MVS MES measured from the KVS (Standard Density) MVSy MVSy (Standard Density) gross (uncalibrated) in "y" MVSyc MVS calibrated (with adaptation) in "y" MVSysp setpoint of MVS (calibrated) for "y" Mxjy ^ mass of "x" gross (uncalibrated) in the reactor (kg) Mx? mass of "x" calibrated (with adaptation) in "y" (kg) R? reactivity of X in reactor V? Volume of "y" (m3) tfl time spent in the reactor (h) t? dwell time for "x" in the reactor (h) "x" can represent the following constituents: Bt Buteno CA Catalyst Et Etiieno Hy Hydrogen Sv Solvent "y" can represent the following devices: r Polymerization reactor stp Desorber (stripper) sh Dryer (in fluid bed) Legend of figures 10 polymerization reactor 11 reagent feed (raw materials), catalyst, solvent 12 cooling circuit 14 extraction duct 16 desorber 18 condenser 20 gas chromatograph 22 centrifuge 24 fluid bed dryer 25 solvent and reagents to be recycled 26 polyethylene 27 reagents to be recycled 28 solvent to recycle 30 slogans of ownership of the polymer 31 setpoint of step of the procedure 32 main algorithm 33 accessory algorithm 34 concentration slogans 35 slogans of flows that enter 36 regulation (PID) of the flows 37 measures 38 temperature regulation 39 feed-forward temperature 40 polymerization dynamics: chemical kinetics and material balance 41 simulations of the gait step and Hy / Et and Bt / Et ratios 42 measurements of temperature, flow rates and concentrations 43 regulated procedure 44 analysis of a sample of the product if synthesized by the method 51 measure of magnitudes associated with the development of the procedure 52 measure of properties of the direct model polymer 53; prediction of the properties in the measure 54 comparison: correction of the model (adaptation) 55 regulation algorithm based on the inverse model (feed-forward + feed-back) 56 slogans for the quantities associated with the procedure 57 measurements and flow rates that enter 58 temperature measurement and prediction of concentrations 59 direct model: equations of properties as a function of concentrations 60 prediction of properties 61 slogans for concentrations in the reactor 62 measure of the flows that enter 63 measure of the concentrations in the direct model reactor 64; prediction of concentrations on the basis of the material balance 65 predictions of concentration 66 comparison; calculation of adaptation parameters.

Claims (18)

1. A method for regulating a synthesis process of at least one chemical in a plant comprising at least one reactor (R) which can be assimilated to a perfectly mixed reactor, in which one or more control variables (GC) allow act on the development of the procedure so that one or several magnitudes associated to the properties of the product and / or to the development of the procedure, called regulated quantities (GR), are equal to the corresponding instructions (CGR), whose method comprises the following stages : (a) entry of instructions concerning the regulated quantities (CGR); (b) calculation, by means of a prediction organ (OP), of predictions of the regulated quantities (PQJS) on the basis of measurements of the procedure control quantities (MQC); (c) use of a control body (OC) to calculate the setpoints of the procedure control quantities (CGC), based on the setpoints (CGR) and the predictions (P) of the regulated quantities; (d) transmission of the commands of the procedure control quantities (CGG) to actuators or to regulating organs that control the actuators, in order to act on the development of the procedure; wherein the prediction organ (OP) is based on a mathematical model of the procedure, called direct model (M), characterized in that the prediction organ (OP) is designed in such a way that the mass M? R of minus one constituent (X) in the reactor (R) by the equation: M? R = LAG (F? Rin. t? t?) in which: - FXRin is the mass flow rate of constituent X entering the reactor R; - t? is the residence time of x in the reactor, which is worth t? = M? R / (S Fxdis) in which: - M? R designates the last calculated value of the mass of constituent X present in reactor R; - E Fxdis designates the sum of all the mass flows Fxdis with which the constituent X disappears from the reactor R, in particular by reaction and / or by exit from the reactor; - the function y = LAG (u, t) is the solution of the differential equation dy. u = t dt + y ao calculated with the instantaneous value of u and t, as well as with the last value of y calculated.

2. A regulation method according to claim 1, wherein the setpoint of at least one regulated variable (CGR) is corrected on the basis of the deviation between the measure (MGR) and the prediction (PR) of this regulated quantity, so that regulation is effective even in the presence of an error in the prediction of this regulated quantity (PQR) •

3. A method of regulation according to claim 1, wherein the model (M) of the method is adapted periodically on the basis of the deviation between the predictions (PQR) and the measurements (MGR) of the regulated quantities, so that the model of the procedure provide predictions of the regulated quantities (Pgp) as close as possible to the measurements of these quantities (MGR) -

4. A method of regulation according to claim 3, wherein the measurements (MGR) of the regulated quantities intervene only in the eventual adaptation of the procedure model, and not they intervene directly in the calculation of the slogans of the procedure control quantities (CGG).

5. A regulating method according to one of claims 1 to 4, applied to a polymerization process, comprising one or more of the following supplementary steps: calculation of a temperature setpoint in the reactor as a function of one or several setpoints of properties of the product; and transmitting this temperature setpoint to one or more actuators that allow to modify the temperature in the reactor; - calculation of a thermal balance for the reactor, based, in particular, on temperature measurements; use of this thermal balance to determine the amount of polymer synthesized per unit of time and / or the productivity of the catalyst and / or the concentration of at least one reagent in the reactor; - calculation of the amount of heat produced by the polymerization, thanks to a calculation of the quantity of the reactants that are polymerized; determining by this means the amount of heat that needs to be added or evacuated to maintain the temperature of the reactor; use of the result of said calculation to improve the temperature regulation, in order to respect in the best possible way the temperature setpoint, especially in the case of changes in the pace of the march.

6. A method of regulation according to one of the preceding claims, in which the property PxR of a constituent "x" in the reactor R, assimilated to a perfectly mixed reactor, is calculated as follows: PxR = LAG (PxIN, MxR / FXIN) where "Px" is a property of a constituent "x", which responds substantially to the linear mixing law Pxl + 2 = wl - P? l + w2"Px2 'where 1 and w2 are the mass proportions of two fractions 1 and 2 of property Px, and Px-2 that are mixed; is the property of x at its exit from the reactor, after mixing; PXIN is the property of the constituent "x" upon its entry into reactor R; MxR is the mass of constituent x in reactor R; and Fx-N is the mass flow rate of the constituent x entering the reactor R.

7. A method of regulation according to one of the preceding claims, comprising the following steps: - input of instructions relating to one or more properties of the product to be synthesized, in a main algorithm; - input of the setpoint of the step of the procedure, in an accessory algorithm (slave), - - calculation of setpoints of concentration of the constituents in the reactor with the help of the main algorithm, depending in particular on the instructions and measurements of product properties as well as measurements or predictions of the concentrations of the different constituents in the reactor; - transmission of the concentration slogans calculated by the main algorithm as input quantities to the accessory algorithm; - Calculation of the flow instructions of the constituents entering the reactor, with the help of the accessory algorithm, depending in particular on the instructions for the passage of the procedure, the concentration slogans and the flow measurements of the constituents enter the reactor, and transmission of the flow instructions calculated with the aid of the accessory algorithm to one or more actuators in order to regulate the flow rates of the constituents entering the reactor, in which the main algorithm and / or the algorithm accessory are employed according to one of the preceding claims.

8. A method of regulation according to claim 7, characterized in that the main algorithm comprises: - a prediction organ, based on a direct model of the method that allows to provide a prediction of the properties of the synthesized product based on measurements and / or predictions of the concentrations of the constituents; an adaptation organ that compares the predictions of properties calculated by the prediction organ with values actually measured in the synthesized product and deriving adaptation parameters from said comparison, said adaptation parameters intervening as supplementary input magnitudes in said algorithm prediction organ principal; and a control organ, based on an inverse model of the procedure, to calculate, according to the slogans and the predictions of properties of the product to be synthesized, concentration slogans for the accessory algorithm, said adaptation parameters also intervening as magnitudes of supplementary entry into said control body.

9. A method of regulation according to any one of claims 7 or 8, wherein the accessory algorithm comprises: - a prediction organ, based on a direct model of the method that allows to provide a prediction of the concentrations of one or more of the constituents on the basis of a balance of matter in the reactor; an adaptation organ, which compares the predictions of concentrations calculated by the direct model with concentration measurements, deriving from this comparison the adaptation parameters, said adaptation parameters intervening as supplementary input quantities in said prediction organ of the accessory algorithm; and - a control organ, based on an inverse model of the method, for calculating, based on the setpoint of the gait step, the concentration setpoints calculated by the control organ of the main algorithm and the calculated predictions of concentration by the prediction organ of the accessory algorithm, the instructions for the flows entering the reactor, said adaptation parameters intervening as supplementary input quantities in said accessory algorithm control organ.

10. A method of regulation according to one of the preceding claims, applied to the regulation of the continuous synthesis of polyethylene by polymerization of ethylene in at least one reactor, the reactants comprising ethylene, hydrogen and / or an optional comonomer, the polymerization reaction in the presence of a catalyst and removing a part of the contents of the reactor permanently or intermittently.

11. A method of regulation according to claim 9 and 10, wherein the adaptation organ of the accessory algorithm compares the measurements of the concentrations of ethylene (Et), hydrogen (Hy) and / or optional comonomer (Bt) with the values predicted by the prediction organ of the accessory algorithm, to determine at least one of the following adaptation parameters: a) the specific activity of the catalyst for the ethylene "AEt", in kg / h of polyethylene per kg of catalyst and per kg / m3 of etiieno; b) the gain error on the measurement of the hydrogen flow rate "KfHy"; c) the purity of the "KfBt" comonomer feed.

12. A method of regulation according to one of claims 1 to 9, applied to the regulation of the continuous synthesis of polypropylene by polymerization of propylene in at least one reactor, the reactants comprising propylene, hydrogen and / or an optional comonomer, having The polymerization reaction takes place in the presence of a catalyst and a part of the content of the reactor is removed permanently or intermittently.

13. A method of regulation according to claims 9 and 12, wherein the adaptation organ of the accessory algorithm compares the measurements of the concentrations of propylene (Pe), hydrogen (Hy) and / or optional comonomer (Et) with the predicted values by the prediction organ of the accessory algorithm, to determine at least one of the following adaptation parameters: a) the specific activity of the catalyst for propylene "APe", in kg / h of polypropylene per kg of catalyst and per kg / mj of propylene; b) the gain error on the measurement of the hydrogen flow rate "KfHy"; c) the purity of the "KfEt" comonomer feed.

14. A method of regulation according to claim 8, applied to a polymerization process, in which: the melt flow index (MI) and / or the standard density (MVS) of the polymer and / or its content are periodically measured in comonomer; - the prediction organ of the main algorithm calculates gross predictions of MI and MVS as a function of the temperature in the reactor, of the concentrations in the reactor and of the residence times in the different devices of the polymerization circuit, - - periodically, the organ of adaptation of the main algorithm: - resynchronizes the gross predictions of MI and MVS taking into account the time elapsed between the taking of the measurements of MI and MVS and obtaining the result of the measurements, and compares the gross predictions of MI and MVS resynchronized with the measurements of MI and MVS, calculates a multiplicative adaptation parameter kMI applied to the raw prediction of MI to obtain a calibrated prediction of MI, and -calculates an additive adaptation parameter kMVS applied to the gross prediction of MVS to obtain a calibrated MVS prediction.

15. A method of regulation according to one of the preceding claims, applied to a polymerization process, in which one or several properties of the polymer are evaluated using a technique chosen from near infrared spectroscopy (NIR), infrared spectroscopy by Fourier transform (FTIR) and nuclear magnetic resonance (NMR).

16. A method of regulation according to one of the preceding claims, applied to a polymerization process, in which one or more polymer properties are evaluated by applying a pre-established correlation relationship to the results of measurements made by near infrared (NIR) spectroscopy. several wavelengths previously determined depending on the nature of the polymer, chosen between 0.8 and 2.6 mm.

17. A process for the synthesis of a chemical in an installation comprising at least one reactor that can be assimilated to a reactor perfectly i * o mixed, regulated by means of the regulation method according to one of the preceding claims.

18. A device for regulating a process for synthesizing a chemical in a synthesis plant comprising: - at least one reactor that can be assimilated to a perfectly mixed reactor; - at least one means to enter a property setpoint (CGR) of the product to be synthesized in the calculation unit; - at least one means for entering a setpoint for passing the product to be synthesized (CGC) in the calculation unit; - at least one control organ (OC), - - at least one prediction organ (OP); - at least one means for imposing a command of a command variable (CGC) on a. suitable actuator; wherein the method of regulation is according to one of claims 1 to 16. SUMMARY Method of regulation of chemical synthesis procedures A method for regulating a synthesis process of at least one chemical in a plant comprising at least one reactor (R) which can be assimilated to a perfectly mixed reactor, in which control quantities (GC) allow to act on the development of the procedure so that one or several magnitudes associated to the properties of the product and / or to the development of the procedure, called regulated quantities (GR) are equal to the corresponding instructions (CGR), whose method comprises the following steps: (a) entry of instructions concerning the regulated quantities (CGR); (b) calculation, by means of a prediction organ (OP), of predictions of the regulated quantities (PQR) > on the basis of measurements of the procedure control quantities (MGC); (c) use of a control body (OC) to calculate the setpoints of the procedure control variables (CGC), based on the setpoints (CGR) and the predictions (Pgp) of the regulated quantities; (d) transmission of the instructions of the procedure control quantities (CGC) to actuators in order to act on the development of the procedure; wherein the prediction organ '(OP) is based on a mathematical model of the method, called a direct model (M) and is designed in such a way that the mass M? R of at least one constituent (X) is predicted in the reactor (R) by the equation .- MXR = LAG (F? Rin.t?, t?) in which: - F Rin is the mass flow of constituent X entering reactor R; - t? is the residence time of x in the reactor; - the function y = LAG (u, t) is the solution of the differential equation dv u = t. dt + y calculated with the instantaneous value of u and t, as well as with the last value of y calculated. Figure 7