System for determining individual product quantities of multiple chemical reactors

Abstract

The present invention relates to a system (100) for determining individual product amounts of multiple chemical reactors (11, 12, ..., 1K) that contribute to a combined product amount. The system includes a measurement providing unit (101) that provides measured individual reactant amounts for each reactor, and an artificial intelligence providing unit (102) that provides each reactor with artificial intelligence (21, 22, ..., 2K) that is trained upon receiving as input the individual reactant amounts for the reactor to provide as output the individual product amounts for the reactor that are combined into a combined product amount related to the individual reactant amounts received as input. The system further includes an individual product amount determining unit (103) that determines the individual product amounts for the reactor based on the measured individual reactant amounts and the trained artificial intelligence. The system allows for increased chemical production efficiency whenever multiple reactors contribute to a combined product amount.

Description

[Technical Field]

[0001] The present invention relates to a system, method and computer program for determining the individual product amounts of multiple chemical reactors that contribute to a combined product amount. [Background technology]

[0002] Chemical manufacturing processes often require reactions to occur simultaneously in two or more chemical reactors. Nevertheless, in many cases, only a single product output—i.e., a single combined product amount from the individual reactors—can be delivered. For example, conduits from the individual reactors may connect to a common supply conduit, and the product may only be accessible through the common supply conduit. In such cases, unless the individual product amounts of the multiple reactors are measurable, it can be difficult to control the individual reactors. However, suboptimal control of the reactors can lead to an unsatisfactory manufacturing process. Summary of the Invention [Problem to be solved by the invention]

[0003] It is an object of the present invention to enable increased efficiency in chemical manufacturing processes in which multiple chemical reactors contribute to a combined product load. [Means for solving the problem]

[0004] In a first aspect of the present invention, there is provided a system for determining individual product amounts of a plurality of chemical reactors that contribute to a combined product amount, the system comprising: - a measurement providing unit configured to provide a measured individual reactant quantity for each of the plurality of chemical reactors; an artificial intelligence providing unit configured to provide a trained artificial intelligence to each of the chemical reactors, the provided trained artificial intelligence being trained upon receiving as input individual reactant quantities for the chemical reactors to provide as output individual product quantities of the chemical reactors that are combined into a combined product quantity related to the individual reactant quantities received as input; an individual product amount determination unit configured to determine individual product amounts for the plurality of chemical reactors based on the measured individual reactant amounts and a trained artificial intelligence; A system is provided that includes:

[0005] Thus, multiple artificial intelligences are used to individually model the chemical reactors, and the artificial intelligences are trained so that the individual product quantities provided as outputs are combined into a combined product quantity related to the individual reactant quantities received as inputs. This has been found to allow for accurate determination of the individual product quantities of the multiple reactors based on their respective individual reactant quantities, which can be used for optimal control of the reactors, thereby improving the efficiency of each manufacturing process.

[0006] The determined individual product amounts may be replaced by corresponding measurements. Such individual product amount measurements may not be possible if only the combined product amount is provided for use. Also, if such measurements are possible, such measurements may no longer be necessary when using the presented system.

[0007] Because the reactant amounts provided to each reactor and the additional process parameters selected to control the reactions occurring in each reactor may differ from one another, it can be difficult to control a chemical production process in which multiple chemical reactors contribute to a combined product amount without being able to measure the individual product amounts of the reactors, even if all reactors are identical and the combined product amount provided is known. The reactant amounts provided to each reactor and the additional process parameters for multiple reactors may be intentionally selected to be different for practical reasons. However, this may only be achievable with a finite precision, even when identical control is desired, and even relatively small variations in the individual reactant amounts and additional process parameters may have a relatively large impact on the individual product amounts produced by each reactor. Accurate determination of the individual product amounts of multiple chemical reactors that contribute to a combined product amount provides richer information, and the production process can be controlled based on this information.

[0008] The amount of an individual reactant may refer to the amount of one reactant among a plurality of reactants or the amount of each of two or more reactants, where the plurality of reactants are chemically distinguishable from one another. It may be sufficient to measure the amount of only one reactant among the reactants for each of a plurality of chemical reactors. However, for particularly complex chemical reactions, the amounts of two or more reactants, particularly all of the reactants, may also be measured for each of the chemical reactors.

[0009] The amount of combined products may refer to the amount of one product among a plurality of products, where the plurality of products are chemically distinguishable from one another. The combined product whose amount is referred to herein may specifically be the product whose production is the primary purpose of the chemical reactor. Other products among the plurality of products may be produced only as side reactions. Nevertheless, these other products may still be useful, such as for further reactions in additional reactors.

[0010] The multiple chemical reactors can be acetylene reactors for producing acetylene. In particular, acetylene, or a raw form of acetylene that can later be processed into actual acetylene, can be produced in the multiple chemical reactors from oxygen and natural gas as reactants. In addition to acetylene, synthesis gas can be produced. In this particular case, for example, the measured individual reactant amounts can refer to the individual amounts of natural gas and oxygen delivered to each of the multiple reactors, and the combined product amount can refer collectively to the amount of acetylene produced by the multiple chemical reactors.

[0011] The term "artificial intelligence" is understood herein to include any type of machine learning model. Among the artificial intelligences that have proven particularly useful for the present purpose are artificial neural networks and regression trees, in particular gradient-boosted trees, such as the XGBoost model. However, it will be understood that these are only specific examples of the types of artificial intelligence that can be used.

[0012] Specifically, the artificial intelligence can be trained to provide as output the individual product quantities of the chemical reactors that sum to a combined product quantity when the artificial intelligence receives as input the individual reactant quantities for the chemical reactors. Thus, "combination" can specifically refer to summation. The amounts of individual and combined products can be expressed, for example, in terms of volume or weight, specifically in terms of volume or weight flow rates, i.e., volume or weight, respectively, fed or delivered per unit time.

[0013] The association of the individual reactant amounts received as input by the artificial intelligence with the combined individual product amounts provided as output by the artificial intelligence may correspond, in particular, to an assignment made to train the artificial intelligence. For example, the combined product amount may be associated with the individual reactant amounts received as input by the artificial intelligence in that it is used in the combined training of multiple artificial intelligences as a combined training output provided by the artificial intelligences that received the individual reactant amounts as input. Thus, training data for the combined training of the artificial intelligences may include, for example, a) measured individual reactant amounts as training input data and b) measured combined product amounts as combined training output data. The artificial intelligence may be trained such that the individual product amounts provided as output upon receiving the measured individual reactant amounts used as training input data sum to the measured combined product amount used as combined training output data.

[0014] The data "pairs" of a) measured individual reactant amounts and b) measured combined product amounts can also be obtained at the time the manufacturing process is controlled, i.e., at the time the state of the manufacturing process is checked and / or changed. Therefore, the association between the individual reactant amounts received as input by the artificial intelligence and the combination of individual product amounts provided as output by the artificial intelligence can also correspond to the assignment of a) measured individual reactant amounts for multiple reactors at the time the manufacturing process is controlled and b) the combined product amounts that can be expected and / or measured at this time. In other words, the trained artificial intelligence can be considered a model of "real" data that can be measured during manufacturing and characterizes the ongoing manufacturing process.

[0015] It is understood that perfect training of an artificial intelligence is generally not possible in practice. Thus, the output provided by a trained artificial intelligence may simply be combined or summed to approximately the combined product amount, i.e., the combined product amount related to the individual reactant amounts received as input by the artificial intelligence. This is true not only for "real" data, but also for training data, one reason being that overtraining should be avoided.

[0016] The provided trained artificial intelligence can be trained to provide as an output the individual product quantities of a chemical reactor upon receiving additional input values ​​derived from the individual reactant quantities. It has been found that even if the additional input values ​​can be derived solely from the individual reactant quantities, i.e., without further information, such as a function dependent only on the individual reactant quantities, the use of such additional input values ​​allows for a more accurate determination of the individual product quantities. For example, the artificial intelligence can receive as input the individual reactant quantities and, additionally, the ratio between the individual reactant quantities for each chemical reactor. For example, in the case of acetylene production from natural gas and oxygen, it may be preferable to use an artificial intelligence that not only receives the quantities of natural gas and oxygen as inputs, but also additionally receives the ratio of the respective quantities of oxygen and natural gas for multiple chemical reactors. It is somewhat surprising that such additional inputs can improve system performance, since the ratio conveys less information than the measured individual quantities of oxygen and natural gas themselves, since it can be calculated by dividing one measured quantity by the other.

[0017] 1. A system for determining control parameters for chemical reactions in a plurality of chemical reactors that contribute to a combined product amount, comprising: - a system for determining the individual product quantities of a plurality of chemical reactors as defined above; a control parameter determination unit configured to determine control parameters for chemical reactions in the plurality of chemical reactors based on the determined individual product amounts; A system including:

[0018] The ability to accurately determine individual product quantities using multiple individual artificial intelligences allows for the determination of control parameters that allow optimal control of the chemical reactor and therefore more efficient production.

[0019] In particular, the control parameter determination unit may be configured to determine the amounts of individual reactants delivered to the plurality of chemical reactors and / or further process parameters for the plurality of chemical reactions based on the determined amounts of individual products. Thus, the determined control parameters may correspond in particular to the amounts of individual reactants delivered to the plurality of chemical reactors and / or further process parameters for the plurality of chemical reactions. It is understood that the plurality of reactions may be substantially chemically identical, the "plurality" arising simply from the fact that the reactions are carried out in different reactors, thereby introducing slight variations.

[0020] A human operator may control the multiple chemical reactors based on the determined control parameters. Alternatively, a control system for controlling the multiple chemical reactors may be provided, and the control system may be configured to control the multiple chemical reactors based on the determined control parameters. Controlling the reactors based on the determined control parameters may refer to adjusting the actual observed control parameters to the determined control parameters. For example, the flow of one or more reactants to the reactors may be increased or decreased. The determined control parameters may particularly refer to target control parameters.

[0021] The control parameter determination unit may be configured to determine the control parameters further based on the measured individual reactant amounts and / or the measured combined product amounts. In this way, the current state of each reactor can be taken into account. This may allow the reactor state to limit the control parameters achievable within a desired time frame, which may result in more efficient determination of the control parameters, since other control parameters do not need to be considered.

[0022] Additionally or alternatively, the control parameter determination unit may be configured to determine the control parameters based on trained artificial intelligence. For example, the trained artificial intelligence may be used to determine individual product amounts for target individual reactant amounts. Then, using the trained artificial intelligence, the target individual reactant amount that results in the most favorable individual product amount may be selected as the individual reactant amount actually delivered to the multiple chemical reactors. As will be understood, the term "most favorable" may refer to any measure for evaluating the performance of the multiple chemical reactors. For example, the "most favorable" individual product amount may not necessarily be the highest, but may be one that satisfies a predetermined relationship with respect to the target individual reactant amount.

[0023] The control parameter determination unit can be configured to determine the control parameters for the chemical reactions such that a combined amount of at least one of the reactants for the plurality of chemical reactors is minimized without reducing a combined product amount. In this way, resource-saving chemical production can be achieved. Furthermore, the cost for supplying the at least one reactant, and therefore the overall production cost, can also be minimized.

[0024] In particular, the control unit can be configured to control the chemical reactions such that the combined amount of at least one of the reactants for the plurality of chemical reactors is minimized without changing the combined product amount. Thus, in such an embodiment, the combined product amount is not only not decreased, but also not increased. In other words, the combined product amount is maintained.

[0025] In fact, the control unit may also be configured to minimize the combined production cost rather than minimizing the combined amounts of certain reactants, since the costs of supplying different reactants and of controlling chemical reactions according to certain process parameters, which may include certain energy consumption, may not be constant over time, so minimizing the combined production cost does not necessarily correspond to minimizing the amounts of certain reactants.

[0026] It has been found that the dependence of the combined product quantity on the multiple individual reactant quantities and / or further process parameters is generally relatively complex and typically includes several local minima when understood as a function of the space of possible reactant quantities and / or further process parameters for the multiple chemical reactors. Therefore, it is preferred that the respective minimization is performed globally, i.e., globally in the space of possible reactant quantities and / or further process parameters for the multiple chemical reactors. In particular, evolutionary algorithms, such as differential evolution or genetic algorithms, can be used for the respective minimization.

[0027] In one embodiment, the control parameter determination unit is configured to determine a function

number

number

number

number

number

number

[0028] Alternatively, the control parameters can be optimized using, for example, a genetic algorithm ("GA"). The control parameter determination unit then determines, in particular, the function

number

number

number

number

number

number

number

[0029] Both differential evolution and genetic algorithms can be implemented in their different variants, some of which may allow for a faster and / or more accurate determination of the optimal control parameters, so that an "optimization in optimization procedure" can be performed, i.e., optimization over an ensemble of methods considered to be used to optimize the control parameters.

[0030] As far as differential evolution is considered to find the optimal control parameters, the search for the optimal variant can be limited to the variants shown in Table 1(a) below, where N pop where σ denotes the size of each population considered, D denotes the number of control parameters considered, CR denotes the crossover probability, F denotes the differential weight, and "Strategy" denotes the evolution strategy, e.g., as used in SciPy v1.9.2. As far as a genetic algorithm is considered to find the optimal control parameters, the search for its optimal variant can be limited to the variants shown in Table 1(b) below, where N pop Here again, refers to the size of each population considered, CR again refers to the crossover probability, MR refers to the mutation probability, and "crossover" refers to the type of crossover considered.

[0031] [Table 1]

[0032] Table 1a, b: Parameter choices for a) Differential Evolution (DE, left) and b) Genetic Algorithm (GA, right) considered for control parameter optimization in one embodiment.

[0033] To perform the optimization of the control parameters in a meaningful way, the control parameter determination unit relies on an already trained artificial intelligence.

[0034] Preferably, the trained artificial intelligence is a training method comprising: - providing a measured individual reactant amount for each of the plurality of chemical reactors and a measured combined product amount of the plurality of chemical reactors associated with the measured individual reactant amounts; - providing an estimated individual product amount for each of a plurality of chemical reactors; - providing a trained artificial intelligence for each of a plurality of chemical reactors; - pre-training the provided artificial intelligence such that, upon receiving measured individual reactant amounts as input, the pre-trained artificial intelligence provides estimated individual product amounts as output; -

number

number

number

number

number

[0035] The above definitions of pre-training and supplemental training should not be misconstrued as requiring that the outputs provided by the trained artificial intelligences, respectively, exactly match the target outputs. Rather, the target outputs, i.e., the estimated individual product amounts during pre-training and the matched individual product amounts y' during supplemental training, serve as training output data for training input data, which are measured individual reactant amounts, and each of the pre-training and supplemental training, i.e., when considered independently, can follow known training protocols.

[0036] Artificial intelligence is a set of functions that collectively represent x1, x2, ..., x P The individual reactant amounts and optionally further input amounts are referred to as individual product amounts.

number

number

[0037] The quantity y' can be understood as the adapted combined product amount. However, since the combined product amount, as opposed to the individual product amounts, has been measured, and these measurements are preferably reliable, the quantity y' is preferably treated simply as a fictitious combined product amount.

[0038] With a given form of S and (shape of) P, the above equation (3) for the fitted individual product amounts becomes:

number

number

[0039] This means that the fitted individual product quantities sum up to a "fictitious" fitted combined product quantity. Thus, the fitted quantities y' and y' behave as expected from their "real" counterparts, i.e., the measured combined product quantity and the respective unmeasurable individual product quantity. On the other hand, this generally means that the individual product quantities y arising as output from pre-training cannot be expected to sum up to the real measured combined product quantity y already, and therefore the corresponding vector (y, y) T If the individual product quantities are expected to combine into the combined product quantity in a way other than by addition, then S, and in particular its first row, can be adapted accordingly.

[0040] Estimated individual product amounts for multiple chemical reactors can be provided based on the measured combined product amounts and / or the measured individual reactant amounts. In particular, the estimated individual product amounts can be provided in combination, i.e., summed to the measured combined product amount. For example, if multiple chemical reactors are measured to all receive the same individual reactant amounts, the individual product amounts can be estimated to be the same, i.e., the measured combined product amount divided by the number of chemical reactors. This particular estimate can also be referred to as an average. More generally, a breakdown of the measured combined product amount based on the measured individual reactant amounts can be used to estimate the individual product amounts. That is, if multiple chemical reactors are measured to receive different individual reactant amounts, the individual product amounts can be estimated to be related to the measured combined product amount just as the measured individual reactant amounts are related to the combination of the measured individual reactant amounts. A specific one of the reactants can be selected as the basis for this breakdown. In other words, the estimated individual product amounts may be such that they relate to the measured combined product amounts, just as the measured individual reactant amount for a particular one of the reactant amounts relates to the combination of the measured individual reactant amounts for this particular one of the reactants. Combination may particularly refer to a sum.

[0041] Therefore, the estimated amounts of each product were

number

number

number

number

number

number

[0042] The measured individual reactant amounts and measured combined product amounts provided in the training method may be measured over time, such that a plurality of individual reactant amounts and associated combined product amounts measured at several points in time are provided, and the steps of providing estimated individual product amounts, pre-training the artificial intelligence, determining adapted individual product amounts, and supplemental training the artificial intelligence are performed on a plurality of individual reactant amounts and associated combined product amounts measured at several points in time.

[0043] If the chemical reactions of interest, i.e., each of the chemical reactions in the multiple chemical reactors, occur sufficiently rapidly, and if the individual products transported from the multiple reactors combine along the way and the time required for the combined product amount to reach a location where it can be measured is sufficiently short, then time delays can be ignored to a good approximation, i.e., the combined product amount measured at a given time point can be assigned to the individual reactant amounts measured at that time. Otherwise, the assignment of the measured combined product amount to the measured individual reactant amounts can be applied based on a predetermined time delay.

[0044] If the measurements used for training are taken over time, some of the quantities used to describe the training method will acquire a time dependency. For practical reasons, measurements may only be taken at specific time points so that the time dependency can also be indicated by additional indicators. Nevertheless, for presentation purposes, the time dependency can be represented as if it were continuous, to distinguish it from indicators indicating the amounts of inputs received by the chemical reactor / AI and the AI. Thus, for example, the measured amounts of individual reactants and the combined product amount can be expressed as x 1,2 =x 1,2 (t) and y = y(t), where t indicates the respective measurement time. Thus, the output quantity y = y(t) is obtained by pre-training, and the determination of the estimated individual product quantities and the determination of the fitted individual product quantities defined in the above equation (3) are performed in the same way for all measurement times t, thereby obtaining the quantity

number

[0045] training input data pairs, i.e., a) measured individual reactant amounts x 1,2 and a possible further input quantity x 3,…,P , and b) the training output data, i.e., the respective quantities

number

[0046] The fitted individual product amounts y' are P=(S T W -1 S) -1 S T W -1 (8) can be determined using where W is of the form a) W ∝ I K+1 ,

number

number

number

number

number

number

[0047] Any of these choices for P have been found to allow good precision in the tailored individual product amounts.

[0048] Preferably, the determination of the adapted individual product amounts and the supplementary training of the artificial intelligence are repeated in a training method, and in each iteration: - upon receiving the measured individual reactant amounts as input, the individual product amounts provided as output by the already trained artificial intelligence are assumed as individual product amounts to be adapted for determining further adapted individual product amounts based on the measured individual reactant amounts, The artificial intelligence is additionally trained such that upon receiving the measured individual reactant amounts as input, the additionally trained artificial intelligence provides further adapted individual product amounts as output.

[0049] As mentioned above, the quantity y' is preferably considered to be a fictitious combined product quantity. Accordingly, the quantity y' can be discarded between iterations, which means that the value determined in one iteration in the process of determining the adapted individual product quantities is not used in the next iteration, i.e., to determine further adapted individual product quantities. Instead, each measured combined product quantity can be used again. Therefore, when repeating supplementary training, formula (3) can be used again, with y being replaced by the individual product quantities determined by the artificial intelligence obtained from the previous supplementary training, but without replacing y.

[0050] Furthermore, to adapt the individual product quantities between supplementary training, a formula corresponding to equation (8) can again be used, in particular with one of the above options a) to c) for W. If option b) is used, then the quantity e t is redefined to refer to the error, i.e., deviation, of i) the individual product amounts provided as outputs by each pre-trained supplementary artificial intelligence upon receiving the measured individual reactant amounts as inputs, relative to ii) the corresponding individual product amounts, which are used as training outputs for each previous supplementary training. Since the estimated individual product amounts serve as training outputs during pre-training, it can also be said that the training outputs used in each previous training can be used for each adaptation. Meanwhile, in this embodiment, the training outputs used for supplementary training are the adapted training results of each previous training, so that in the adaptation preparing for each next supplementary training, the quantity e t It can be said that σ denotes the error, i.e., deviation, between the current training result and the adapted training result of each previous training.

[0051] The pre-training, supplemental training, and any iterations of supplemental training can each be performed in many ways, for example, possibly following known training protocols depending on the type of artificial intelligence used, and in particular using loss functions of essentially unlimited type.

[0052] As outlined above, it may be preferable to train multiple artificial intelligences over several training epochs rather than in a single training, however, it has been found that equally preferred or even more preferred embodiments can be achieved using only a single training, as long as an appropriate loss function is selected for the single training.

[0053] As outlined above, according to one possible embodiment, without the need to train over several training epochs, the (single) training of the artificial intelligence is performed by using a loss function

number

number

[0054] It will be appreciated that in principle, the loss function given by equation (9) can be used at different epochs of the step-wise training procedure outlined further above, i.e., pre-training, supplemental training, and any iteration of supplemental training. For pre-training, equation (9) can be used in the same way, but for one or more supplemental trainings,

number

number

number

[0055] As can be seen from equation (9), the loss function

number

number

number

number

number

[0056] Another possible loss function that can be used specifically with single training techniques, but which generalizes equally well to multi-stage training as shown above for the loss function from equation (9), is:

number

[0057] This alternative loss function, which is similar to the previous loss function L1, is:

number

number

[0058] With trained artificial intelligence at hand, for a set of newly measured individual reactant quantities for multiple chemical reactors, it is possible to determine the corresponding individual product quantities, even if it is not possible to measure the corresponding individual product quantities. As already indicated above, this can enable improved control of multiple chemical reactors, since more sophisticated optimization of control parameters can be performed. For example, as already outlined above, specific evolutionary algorithms can be applied, such as differential evolution or genetic algorithms.

[0059] It should be noted that optimization can be performed on an ensemble of considered AIs to find the one that should actually be implemented. When only a given type of AI is considered, this process can be called hyperparameter optimization. When two or more types of AI are considered, this optimization can be extended to different types of AIs. Typically, optimization performed for AI involves pre-selecting one or more types of AIs with different hyperparameters, training them, and then comparing the trained AIs according to a predetermined performance metric. Thus, for example, to find an AI to be used to optimize a control parameter, artificial neural networks and XGBoost models with different hyperparameters can be pre-selected, trained, and then compared, and the best-performing one among them can then be selected for actual use. The term "best-performing" can refer, for example, to how well each AI can reproduce measured combined product amounts from corresponding measured reactant amounts, i.e., measured validation data.

[0060] When using an artificial neural network or XGBoost model as the artificial intelligence, the hyperparameters may be limited to the values ​​shown in Tables 2a and 2b below, respectively.

[0061] [Table 2]

[0062] Table 2a, b: Selection of hyperparameters of the artificial intelligence considered in the embodiment using a) Artificial Neural Network (ANN, left) and b) XGBoost model (XGB, right).

[0063] Furthermore, the training of the artificial intelligence and the determination of optimized control parameters can be repeated over time, such as at predetermined intervals. In this way, changes in different chemical reactors and / or their environments can be taken into account, and the control parameters for the reactors can be (re)set accordingly. The periodic intervals for training the artificial intelligence, optimizing the control parameters, and (re)setting the control parameters to each "new" optimal control parameter may be referred to as a control cycle. A typical control cycle may have a duration of, for example, one hour.

[0064] In a further aspect, the present invention provides a training system for training a plurality of artificial intelligences used to determine individual product amounts of a plurality of chemical reactors that contribute to a combined product amount, the training system comprising: - a training data providing unit configured to provide training data, the training data including: a) training input data corresponding to individual reactant amounts received by the plurality of chemical reactors; and b) training output data corresponding to combined product amounts associated with the individual reactant amounts provided as input data; an artificial intelligence providing unit configured to provide, for each of the chemical reactors, an artificial intelligence to be trained; a training unit configured to train an artificial intelligence with the combined training using the training data, such that upon receiving as input individual reactant quantities for the chemical reactor, the trained artificial intelligence provides as output individual product quantities for the chemical reactor that are combined into a combined product quantity related to the individual reactant quantities received as input; and The training performed by the training unit may be of any of the types described above.

[0065] The present invention also provides a method for determining individual product amounts of a plurality of chemical reactors that contribute to a combined product amount, the method comprising: - providing measured individual reactant amounts for each of a plurality of chemical reactors; - providing a trained artificial intelligence for each of the chemical reactors, the trained artificial intelligence being trained to, upon receiving as inputs individual reactant quantities for the chemical reactors, provide as output individual product quantities for the chemical reactors that are combined into a combined product quantity related to the individual reactant quantities received as inputs; determining individual product quantities for the plurality of chemical reactors based on the measured individual reactant quantities and trained artificial intelligence; This method, in any of its embodiments, can be performed by a corresponding system, as further described above.

[0066] Another aspect of the present invention is a training method for training a plurality of artificial intelligences used to determine individual product amounts of a plurality of chemical reactors that contribute to a combined product amount, the training method comprising: - providing training data, the training data including: a) training input data corresponding to individual reactant amounts received by the plurality of chemical reactors; and b) training output data corresponding to combined product amounts associated with the individual reactant amounts provided as input data; - providing for each of the chemical reactors an artificial intelligence to be trained; training an artificial intelligence with combined training such that, upon receiving as input individual reactant quantities for a chemical reactor, the trained artificial intelligence provides as output individual product quantities for the chemical reactor that are combined into a combined product quantity related to the individual reactant quantities received as input; This method can be performed by the training system described above in any of its embodiments.

[0067] The present invention also relates in one aspect to a computer program for determining individual product amounts of a plurality of chemical reactors that contribute to a combined product amount, the program comprising program code means for causing the above-described system for determining individual product amounts to execute a corresponding method for determining individual product amounts.

[0068] The present invention further relates in one aspect to a computer program for training a plurality of artificial intelligences used to determine the individual product amounts of a plurality of chemical reactors that contribute to a combined product amount, the program comprising program code means for causing the training system described above to perform the training method described above.

[0069] It is to be understood that the above-described aspects, in particular the system of claim 1, the method of claim 13 and the computer program of claim 14, as well as the training system, training method and corresponding computer program, have similar and / or identical preferred embodiments, in particular as defined in the dependent claims.

[0070] It is to be understood that a preferred embodiment of the invention can also be any combination of the dependent claims or the above embodiments with the respective independent claim.

[0071] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiment(s) described hereinafter. [Brief explanation of the drawings]

[0072] [Figure 1] 1 shows a schematic and exemplary installation for the production of acetylene. [Figure 2] 1 shows, in a schematic and exemplary manner, a system for determining individual product amounts. [Figure 3] 1 shows a schematic and exemplary representation of several artificial intelligences; [Figure 4a-4b] 10 illustrates exemplary manufacturing efficiency improvements achievable in one embodiment. [Figure 5a-5b] 10 illustrates, by way of example, the manufacturing efficiency improvements achievable in further embodiments. [Figure 6] 1 shows a schematic and exemplary method for determining the amount of each product. DETAILED DESCRIPTION OF THE INVENTION

[0073] FIG. 1 shows a schematic and exemplary diagram of a chemical production facility 10 for carrying out a production process. The facility 10 includes ten chemical reactors 11, 12, ..., 1K, each of which is supplied with oxygen (O) and natural gas (NG) as reactants. The chemical reactors 11, 12, ..., 1K carry out chemically corresponding reactions, thereby producing chemically corresponding products. In this case, raw acetylene (Ac) and additionally synthesis gas (SG) are produced from the reactants O and NG. For example, due to variations in the amounts of reactants supplied to the different reactors and in the reactor behavior, each of the reactors 11, 12, ..., 1K produces a different product amount. The outputs of the individual reactors, i.e., the individual product amounts, are combined via conduits and directed into a fractionation column 12′. In the illustrated embodiment, three groups of reactors are formed, and a separate fractionation column 12′ is provided for each of the three groups. The product, i.e., the already partially combined product amount, is further conducted from the fractionation column 12' to a compressor 13', from which it is conducted to a dedicated device 14' for separating the product into its chemical components, i.e., raw AC and SG. Since the chemical reactions carried out in the reactors 11, 12, ..., 1K involve gas decomposition, the separation process carried out by the device 14' is sometimes called cracked gas separation. Although a separate compressor 13' is still provided for each of the three reactor groups, after passing through the compressor 13', the products are combined, and the total product amount produced by the ten reactors 11, 12, ..., 1K enters the device 14'. Although not shown in FIG. 1, after the raw AC and SG are separated from each other in the device 14', the raw AC is compressed and then treated in an acid scrubber to become its final form, AC, while the SG is conducted to a lean gas scrubber. Because the primary purpose of facility 10 is the production of acetylene, the amount of acetylene output by unit 14' is referred to as the "combined product amount," while the amount of synthesis gas output by unit 14' can be considered a by-product for the present purposes.Both the acetylene and synthesis gas produced are generally used in further chemical production processes, and it will nevertheless be understood that the embodiments described herein may also be used when the amount of synthesis gas produced is considered, rather, as the combined product amount for which production is optimized.

[0074] 2 schematically and exemplarily illustrates a system 100 for determining individual product amounts of multiple chemical reactors that contribute to a combined product amount. The system includes a measurement providing unit 101 configured to provide measured individual reactant amounts for each of the multiple chemical reactors. Furthermore, the system 100 includes an artificial intelligence providing unit 102 configured to provide trained artificial intelligence for each of the chemical reactors, the trained artificial intelligence being trained, upon receiving individual reactant amounts for the chemical reactors as inputs, to provide as outputs the individual product amounts of the chemical reactors that are combined into a combined product amount related to the individual reactant amounts received as inputs. The system 100 also includes an individual product amount determining unit 103 configured to determine individual product amounts for the multiple chemical reactors based on the measured individual reactant amounts and the trained artificial intelligence.

[0075] System 100 can be used to model the manufacturing facility 10 shown in FIG. 1. In this manner, information about the reactions occurring in each of the chemical reactors 11, 12, ..., 1K can be obtained even when it is not possible to measure the amount of each individual product produced by each reactor 11, 12, ..., 1K or the individual product streams routed from each reactor 11, 12, ..., 1K to the fractionation column 12′. Often, in manufacturing processes such as the one shown in FIG. 1, it is not possible to measure the amount of each individual product produced by each of the reactors 11, 12, ..., 1K, as opposed to measuring the amount of each reactant supplied to each of the reactors 11, 12, ..., 1K, i.e., the amount of oxygen and natural gas supplied to each of the reactors 11, 12, ..., 1K in the example of FIG. 1, and measuring the combined amount of products, i.e., specifically, the total amount of acetylene in the example of FIG. 1. Knowledge of the individual product amounts allows optimal control of the multiple reactors 11, 12, ..., 1K, thereby increasing the efficiency of the overall manufacturing process. For example, reductions in natural resources consumed for production can be made for a given amount of combined products. Thus, in the acetylene production example above, the amount of natural gas required can be reduced. This can lead to cheaper production and therefore a competitive advantage, as well as increased economic and political independence.

[0076] FIG. 3 shows, in a schematic and exemplary manner, the structure of the artificial intelligences that can be provided by the artificial intelligence providing unit, i.e., that can be used to model the individual reactors 11, 12, ..., 1K. In the illustrated case, the artificial intelligences are structurally identical and take the form of artificial neural networks 21, 22, ..., 2K. Since an individual artificial intelligence is provided for each of the chemical reactors 11, 12, ..., 1K, there are K artificial intelligences, which can be considered to form a larger shared artificial intelligence 20. In the example shown in FIG. 1, K=10. "Structurally identical" artificial neural networks 21, 22, ..., 2K may refer in particular to sharing the same set of hyperparameters, although their training may naturally lead to different internal parameters of the artificial neural networks, thereby leading to individual models for each chemical reactor 11, 12, ..., 1K.

[0077] In the embodiment of Figure 3, the artificial neural network is chosen to include three layers, namely a single hidden layer. Through the input layer, the individual reactant quantities x 1,2 is received and output through the output layer.

number

number

[0078] In the illustrated embodiment, the artificial neural networks 21, 22, ..., 2K calculate the individual reactant amounts x 1,2 One or more further input values ​​x derived from (t) 3,..,P Thus, the number of inputs (P) received by each of the K artificial neural networks can be greater than the number of reactants involved in the chemical reaction. For example, in the context of acetylene production, it has been found useful to use, apart from the amounts of oxygen (O2) and natural gas (NG) provided to the individual reactors, also the ratio between the amount of O2 and the amount of NG (e.g., the amount of O2 received divided by the amount of NG received) as a control parameter for controlling acetylene production.

[0079] While Figure 3 illustrates an embodiment in which an artificial neural network is used to model the reactors 11, 12, ..., 1K, in other embodiments, other types of artificial neural networks can be used. In particular, gradient-boosted trees, particularly trees from the XGBoost library, can be used instead. When alternative artificial intelligences are used to model the individual reactors 11, 12, ..., 1K, the internal structure of the artificial intelligences will be different, but they can function using the same input and output data as described above with respect to Figure 3. Furthermore, it may still be preferable to use multiple artificial intelligences to create a larger shared artificial intelligence whose output is formed by combining, in particular adding, the outputs provided by the individual artificial intelligences.

[0080] To train the artificial intelligence, a training system is used, the training system including a training data providing unit configured to provide training data, the training data including: a) training input data corresponding to individual reactant amounts received by a plurality of chemical reactors; and b) training output data corresponding to combined product amounts associated with the individual reactant amounts provided as input data. The training system may further include an artificial intelligence providing unit configured to provide, for each of the chemical reactors, an artificial intelligence to be trained, and a training unit configured to train the artificial intelligence with the combined training using the training data, such that, upon receiving the individual reactant amounts for the chemical reactor as inputs, the trained artificial intelligence provides as outputs the individual product amounts of the chemical reactors that are combined into a combined product amount associated with the individual reactant amounts received as inputs.

[0081] Training data can be collected by measurements during ongoing production, such as while the facility 10 shown in FIG. 1 is in operation. Thus, for example, pairs can be constructed from such measurements: a) the measured amounts of oxygen and natural gas supplied to the K reactors 11, 12, ..., 1K at a given time, and additionally the ratio of these amounts, and b) the amount of acetylene recovered from the apparatus 14′ at that same time. These pairs, collected over time, can be used as training data. Similarly, validation data can be obtained and used in conjunction with the training data to train each artificial intelligence. The collected training (and validation) data can be considered combined training (and validation) data. While the combined training (and validation) data can be used in a single combined training of multiple artificial intelligences, it may also be preferable to divide the combined training into several stages, in each of which multiple artificial intelligences are trained using individual training data. Each individual training data can be derived from the combined training data and / or from the training results of each preceding training stage. In particular, the training unit of the training system can be configured to perform the following training method:

[0082] In a first step of the training method, measured individual reactant amounts for each of a plurality of chemical reactors and measured combined product amounts for the plurality of chemical reactors associated with the measured individual reactant amounts are provided, i.e., combined training data is provided.

[0083] In a second step of the training method, estimated individual product quantities are provided for each of the multiple chemical reactors. The estimated individual product quantities may correspond to average values ​​of the combined product quantities provided in the first step of the training method, and the average values ​​may be weighted according to the amount of one of the reactants supplied to each reactor 11, 12, ..., 1K. For example, the average values ​​may be weighted according to the amount of natural gas supplied to each reactor 11, 12, ..., 1K. Such weighting may lead to more accurate estimates, since the more natural gas supplied to a reactor, the more acetylene is expected to contribute to the total amount of acetylene produced by that reactor.

[0084] In a third step of the training method, an artificial intelligence to be trained is provided for each of the plurality of chemical reactors 11, 12, ..., 1K. For example, artificial neural networks 21, 22, ..., 2K as shown in Figure 3 can be provided.

[0085] In the fourth step of the training method, the AI ​​is pre-trained so that the pre-trained AI receives measured amounts of each reactant as input and provides estimated amounts of each product as output. Apart from the selection of this training data, the pre-training of each AI can be performed using a known method for each type of AI. For example, a known loss function can be used.

[0086] In a fifth step of the training method, the adapted individual product quantities for each of the plurality of chemical reactors 11, 12, ..., 1K are determined according to the approach expressible by equations (3) and (4) above. In particular, to determine the matrix P used in equation (3), equation (9) can be used with any of the options a) to c) for the matrix W used therein.

[0087] Next, in the sixth step of the training method, the pre-trained AI is supplementarily trained so that the supplementarily trained AI receives the measured individual reactant amounts as input and provides the respective adapted individual product amounts as output. Again, apart from the selection of training data, the supplementary training of the individual AIs can be performed in a known manner for each type of AI. For example, a known loss function can also be used for the supplementary training.

[0088] To collect training (and validation) data, the measured individual reactant amounts and measured combined product amounts provided in the first step of the training method can be measured over time, resulting in a plurality of individual reactant amounts and associated combined product amounts measured at several time points. The steps of providing estimated individual product amounts (second step), pre-training the artificial intelligence (fourth step), determining adapted individual product amounts (fifth step), and supplemental training the artificial intelligence (sixth step) can then be performed on the plurality of individual reactant amounts and associated combined product amounts measured at several time points.

[0089] Preferably, the fifth and sixth steps of the training method are repeated, and in each iteration, the individual product amounts provided as output by the trained artificial intelligence upon receiving the measured individual reactant amounts as input are assumed to be the individual product amounts adapted to determine further adapted individual product amounts based on the measured individual reactant amounts, and the artificial intelligence is additionally trained to provide further adapted individual final product amounts as output upon receiving each individual reactant amount as input. To avoid overtraining the artificial intelligence, a stopping criterion may be applied, and if the stopping criterion is met, the repeated supplemental training is terminated. For example, the stopping criterion may be selected so that the stopping criterion is met whenever at least one of the following conditions is met: a) a predetermined number of iterations has elapsed; or b) the performance of the artificial intelligence has not improved over the predetermined number of iterations, where the performance may be measured in terms of the mean absolute error of the sum of the individual outputs of the artificial intelligence relative to the measured combined product amounts, and the mean absolute error is evaluated on the training data set.

[0090] In an alternative training method, the combined training is based on the same training data but is not divided into several stages. Instead, the known training protocol can be followed, and one of the functions given in equation (10) above is used as the loss function for training (11).

[0091] Regardless of the training protocol followed, the hyperparameters of each used AI can be optimized. Hyperparameter optimization can, in principle, be performed by training an AI with different hyperparameter settings, and then comparing the different trained AIs in terms of their performance. However, it may be more efficient to make the final selection of hyperparameters before the actual training begins. Therefore, for example, hyperparameter optimization can be performed based on estimated individual product quantities assumed as training output data. In particular, in the case of the multi-stage training described above, which involves pre-training and one or more supplemental training rounds, hyperparameter optimization can be performed only on the pre-trained AI.

[0092] Once the artificial intelligence has been trained on several choices of hyperparameters, it is possible to accurately model reactors 11, 12, ..., 1K.

[0093] Next, a system for determining control parameters for chemical reactions in the plurality of chemical reactors 11, 12, ..., 1K may be provided, including a system 100 for determining individual product amounts for the plurality of chemical reactors and a control parameter determination unit configured to determine control parameters for the chemical reactions in the plurality of chemical reactors 11, 12, ..., 1K based on the determined individual product amounts. The control parameter determination unit is preferably configured to determine control parameters further based on the measured individual reactant amounts and / or the measured combined product amount, more specifically, such that the combined amount of at least one of the reactants for the plurality of chemical reactors is minimized without reducing the combined product amount. For example, in the case of acetylene production, the amount of natural gas used for production can be minimized for a given desired overall amount of acetylene produced. Particularly preferred schemes that can be followed to find optimal control parameters include differential evolution and genetic algorithms, as described above with respect to equations (1a)-(1d) and (2a)-(2c), respectively.

[0094] 4a and 4b illustrate, in a schematic and exemplary manner, efficiency improvements achievable in accordance with the above-described embodiments for an actual manufacturing facility 10. FIG. 4a is a scatter plot in which each dot represents the state of the manufacturing facility at a given time in the past. The horizontal axis shows the combined amount of acetylene produced in kilograms per hour, and the vertical axis shows the amount spent on consumed reactants in euros (EUR) per hour. As can be seen, the further down the scatter plot point is positioned relative to a fixed horizontal position, the higher the production efficiency. Conversely, the further to the right of a fixed vertical position, the higher the production efficiency. Due to the relative dispersion of the scatter plot points, it can be seen from FIG. 4a that the production efficiency of the considered manufacturing facility varies over the considered time interval. The two small clouds of points highlighted in Figure 4a correspond to the state of the considered production facility on that day; the upper cloud of points represents the production state without control parameter optimization, while the lower cloud of points represents the production state achievable through the aforementioned optimization of the control parameters. Thus, it can be observed that a significant increase in production efficiency was achievable on that day by implementing the control parameter optimization described above. This is further illustrated by Figure 4b, which plots the amount of money, again in euros per hour, spent on reactants consumed over the course of a single day. The upper line of the plot corresponds to the unoptimized production state, while the lower line corresponds to the optimized production state. The gap between the two lines has an approximate width of more than 200 euros per hour throughout the day, revealing considerable potential cost savings. It can be seen that this potential cost savings is closely linked to the potential for resource savings, particularly natural gas savings in the case of acetylene production.

[0095] Figures 5a and 5b differ from Figures 4a and 4b only with respect to the database, i.e., the manufacturing facility 10 from which the data was collected. Even for this different manufacturing facility, the potential for significant increases in manufacturing efficiency by employing control parameter optimization as described above becomes apparent.

[0096] 6 schematically and exemplarily illustrates a method 200 for determining individual product amounts of a plurality of chemical reactors that contribute to a combined product amount. The method includes step 201 of providing measured individual reactant amounts for each of a plurality of chemical reactors and step 202 of providing a trained artificial intelligence for each of the chemical reactors, the trained artificial intelligence being trained to, upon receiving as input the individual reactant amounts for the chemical reactors, provide as output the individual product amounts of the chemical reactors that are combined into a combined product amount related to the individual reactant amounts received as input. Furthermore, method 200 includes step 203 of determining individual product amounts for the plurality of chemical reactors based on the measured individual reactant amounts and the trained artificial intelligence. The method can be performed, for example, by system 100.

[0097] Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention from a study of the drawings, the disclosure, and the appended claims.

[0098] In the claims, the word "comprise" does not exclude other elements or steps and the indefinite article "a" or "an" does not exclude a plurality.

[0099] A single unit or device may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

[0100] The steps performed by one or more units or devices, such as providing reactant quantities or other data, providing artificial intelligence, determining individual product quantities or control parameters, any training of the artificial intelligence, etc., can be performed by any other number of units or devices. These steps can be implemented as program code means of a computer program and / or as dedicated hardware.

[0101] The computer program product may be distributed on a suitable medium, such as an optical storage medium or a solid-state medium, supplied together with or as part of other hardware, or may be distributed in other forms, for example via the Internet or other wired or wireless telecommunications systems.

[0102] Any reference signs in the claims should not be construed as limiting the scope of the invention.

[0103] The present invention relates to a system for determining individual product amounts of multiple chemical reactors that contribute to a combined product amount. The system includes a measurement providing unit that provides measured individual reactant amounts for each reactor, and an artificial intelligence providing unit that provides each reactor with artificial intelligence that, upon receiving as input the individual reactant amounts for the reactor, is trained to provide as output the individual product amounts of the reactor that are combined to result in a combined product amount related to the individual reactant amounts received as input. The system further includes an individual product amount determining unit that determines the individual product amounts of the reactor based on the measured individual reactant amounts and the trained artificial intelligence. The system allows for increased chemical production efficiency whenever multiple reactors contribute to a combined product amount.

Claims

1. A system (100) for determining the individual product amounts of a plurality of chemical reactors (11, 12, ..., 1K) that contribute to a combined product amount, said system comprising: a measurement providing unit (101) configured to provide measured individual reactant quantities for each of the plurality of chemical reactors; an artificial intelligence providing unit (102) configured to provide a trained artificial intelligence (21, 22, ..., 2K) to each of the chemical reactors (11, 12, ..., 1K), wherein the provided trained artificial intelligence (21, 22, ..., 2K) is trained, upon receiving as input individual reactant amounts for the chemical reactors (11, 12, ..., 1K), to provide as output individual product amounts of the chemical reactors (11, 12, ..., 1K) that are combined into a combined product amount related to the individual reactant amounts received as input; an individual product amount determination unit (103) configured to determine individual product amounts for the plurality of chemical reactors (11, 12, ..., 1K) based on the measured individual reactant amounts and the trained artificial intelligence (21, 22, ..., 2K); A system (100) comprising:

2. 2. The system of claim 1, wherein the plurality of chemical reactors (11, 12, . . . , 1K) are acetylene reactors for producing acetylene.

3. 3. The system of claim 1, wherein the provided trained artificial intelligence (21, 22, ..., 2K) is trained to provide as output the individual product quantities of the chemical reactors (11, 12, ..., 1K) upon additionally receiving input values ​​derived from the individual reactant quantities.

4. 1. A system for determining control parameters for chemical reactions in a plurality of chemical reactors (11, 12, . . . , 1K) that contribute to a combined product amount, said system comprising: A system (100) according to any one of claims 1 to 3 for determining the individual product amounts of the plurality of chemical reactors (11, 12, ..., 1K); a control parameter determination unit configured to determine control parameters for the chemical reactions in the plurality of chemical reactors (11, 12, ..., 1K) based on the determined individual product amounts; Including, the system.

5. The system of claim 4 , wherein the control parameter determination unit is configured to determine the control parameters further based on the measured individual reactant amounts and / or the measured combined product amount.

6. 6. The system according to claim 4 or 5, wherein the control parameter determination unit is configured to determine the control parameters for the chemical reaction such that a combined amount of at least one of the reactants for the plurality of chemical reactors (11, 12, ..., 1K) is minimized without reducing the combined product amount.

7. The system according to any one of claims 1 to 6, The trained artificial intelligence (21, 22, ..., 2K) is trained by a training method, providing measured individual reactant amounts for each of the plurality of chemical reactors (11, 12, ..., 1K) and measured combined product amounts of the plurality of chemical reactors (11, 12, ..., 1K) associated with the measured individual reactant amounts; providing an estimated individual product amount for each of said plurality of chemical reactors (11, 12, ..., 1K); providing a trained artificial intelligence (21, 22, ..., 2K) for each of said plurality of chemical reactors (11, 12, ..., 1K); pre-training the provided artificial intelligences (21, 22, ..., 2K) so that the pre-trained artificial intelligences (21, 22, ..., 2K) provide the estimated individual product amounts as an output upon receiving the measured individual reactant amounts as an input; [Equation 1] determining adapted individual product amounts for each of said plurality of chemical reactors (11, 12, ..., 1K) according to a formula expressible by: [Equation 2] (K is the number of chemical reactors (11, 12, ..., 1K)) refers to the individual product amounts provided as output by said pre-trained artificial intelligence (21, 22, ..., 2K) upon receiving said measured individual reactant amounts as input, and y refers to said measured combined product amount; [Equation 3] is the adapted individual product amount, [Equation 4] S stands for format [Equation 5] and Here, I Ｋ is a K-dimensional identity matrix, and each entry in the first row of S is equal to 1; and supplementally training the pre-trained artificial intelligences (21, 22, ..., 2K) so that the supplementally trained artificial intelligences (21, 22, ..., 2K) provide the respective adapted individual product amounts as outputs upon receiving the measured individual reactant amounts as inputs; The system according to any one of claims 1 to 6, obtained by the training method comprising:

8. 8. The system of claim 7, wherein the measured individual reactant amounts and the measured combined product amounts provided in the training method are measured over time, thereby providing a plurality of individual reactant amounts and associated combined product amounts measured at several points in time, and the steps of providing estimated individual product amounts, pre-training the artificial intelligence (21, 22, ..., 2K), determining adapted individual product amounts, and supplementary training the artificial intelligence (21, 22, ..., 2K) are performed on the plurality of individual reactant amounts and associated combined product amounts measured at several points in time.

9. The adapted individual product amounts are P=(S Ｔ W －１ S) －１ S Ｔ W －１ is determined using Here, W ｈ is of one of the following forms: a)W∝I Ｋ＋１ 、 [Equation 6] (In the formula, [Equation 7] is t=1...T ｔｏｔ (where y is the error in y relative to the estimated individual product amounts determined relative to the measured individual reactant amounts and combined product amounts indicated by c)W ｈ ∝diag(S １Ｋ ) (In the formula, 1 Ｋ is a K-dimensional vector 1 with only 1 as an entry Ｋ = (1,...,1) Ｔ (Refers to The system of claim 8.

10. The determination of the adapted individual product amounts and the supplementary training of the artificial intelligence (21, 22, . . . , 2K) are repeated in the training method, and in each repetition: Upon receiving the measured individual reactant amounts as input, the individual product amounts provided as output by the already trained artificial intelligence (12) are assumed as individual product amounts to be adapted for determining further adapted individual product amounts based on the measured individual reactant amounts, 10. The system of claim 8 or 9, wherein the artificial intelligences (21, 22, ..., 2K) are supplementarily trained so that upon receiving the respective individual reactant amounts as input, the supplementarily trained artificial intelligences (21, 22, ..., 2K) provide the further adapted individual final product amounts as output.

11. The control parameter determination unit is a function [Equation 8] Constrain the expressible quantities by y(t)·b≧y, [Equation 9] configured to determine the control parameters by minimizing under [Equation 10] comprises the respective amounts of the P reactants for the K reactors at time t; [0011] contains the costs of the P reactants, and b∈{0, 1} Ｋ indicates, for the K reactors (11, 12, ..., 1K), whether the reactor is in operation, y(t) refers to the individual product amounts of the K reactors determined by the trained artificial intelligence (21, 22, ..., 2K) for the individual reactant amounts X(t), y refers to the desired minimum combined product amount, and s １ and s ２ is a predetermined constant, [0012] contains the measured amounts of individual reactants, [0013] includes predetermined limits on the amounts of the individual reactants, or the control parameter determination unit determines the function [0014] Constrain the expressible quantities by [Equation 15] and configured to determine the control parameters by minimizing under [0016] comprises the respective amounts of the P reactants for the K reactors at time t; [Equation 17] contains the costs of the P reactants, and b∈{0, 1} Ｋ indicates, for the K reactors, whether the reactor is in operation, y(t) refers to the individual product amounts of the K reactors determined by the trained artificial intelligence for the individual reactant amounts X(t), y refers to the desired minimum combined product amount, and s １ and s ２ is a predetermined constant, [Equation 18] contains the measured amounts of individual reactants, [Equation 19] includes predetermined limits on the amounts of the individual reactants, [Equation 20] The system of claim 4, alone or in combination with any one of claims 5 to 10, wherein: is a predetermined weight.

12. The training of the artificial intelligence (21, 22, . . . , 2K) is performed by using a loss function [Equation 21] , or loss function [Equation 22] It is performed using During the ceremony, [Equation 23] 12. The system according to claim 1, wherein y(t) denotes estimated individual product amounts for the K reactors (11, 12, ..., 1K) at a given time t, y(t) denotes individual product amounts at the time t determined by the K artificial intelligences (21, 22, ..., 2K), and α is a predetermined training parameter.

13. A method (200) for determining the individual product amounts of a plurality of chemical reactors (11, 12, ..., 1K) that contribute to a combined product amount, said method comprising: providing (201) measured individual reactant quantities for each of the plurality of chemical reactors; providing (202) a trained artificial intelligence for each of the chemical reactors, the trained artificial intelligence being trained to, upon receiving as input individual reactant quantities for the chemical reactors, provide as output individual product quantities for the chemical reactors that are combined into a combined product quantity related to the individual reactant quantities received as input; determining (203) individual product amounts for the plurality of chemical reactors based on the measured individual reactant amounts and the trained artificial intelligence; A method comprising:

14. 14. A computer program for determining individual product amounts of a plurality of chemical reactors that contribute to a combined product amount, the program comprising program code means for causing the system of claim 1 to perform the method of claim 13.

Images