Method for controlling a chemical process

Description

30 METHOD FOR CONTROLLING A CHEMICAL PROCESS

[001] The present invention relates to the control of chemical processes by computational means. Particularly, but not exclusively, the methods that follow have particular application to processes with unpredictable inputs in the form of raw material flow rates and compositions or energy sources. A preferred example in which the disclosed methods can be employed is the process of producing ethanol.

[002] The e-methanol production process involves supplying a raw material from a renewable source, such as biomass or biogas, to a chemical plant. The process is typically powered, at least in part, by a renewable energy source (e.g., wind power). As such, the supply of raw material and electricity is significantly variable and unpredictable, as it depends on the composition of the raw material and the availability of energy. While this problem is present in numerous types of chemical plants, it is particularly exacerbated in e-methanol production, which involves many non-linear processes due, for example, to the presence of recycling cycles. In scenarios like this, processing the raw material in a chemical plant presents a more difficult control problem than previously considered.

[003] Prior art control systems use mechanistic models to identify setpoints for control variables (such as heat input, pump pressure, etc.) based on sets of detected parameters (such as temperature, pH, flow rate, etc.). That is, mechanistic models were formed from systems of deterministic physical equations, derived to represent the connection between process parameters and variables. Such models must be carefully tuned by process engineers using chemical plant data. An example of Petition 870250088545, dated 09 / 30 / 2025, page 13 / 56 / 30 such state-of-the-art method is described in “Optimal Control of Methanol Synthesis Fixed-Bed Reactor” by Flavio Manenti and Giulia Bozzano, Industrial & Engineering Chemistry Research 2013 52 (36), 13079-13091.

[004] There is therefore a need for a better approach to the problem of controlling chemical processes by computational means.

[005] According to the invention, a method is provided for controlling a chemical process, comprising: supplying a raw material to a chemical plant; processing the raw material in the chemical plant to implement a chemical process to produce a product; repeatedly obtaining a set of parameters of the chemical process using sensors; and operating the chemical plant according to control variables based on the parameter sets: discarding results outside the limits of the parameter sets; and simulating the chemical process using the retained parameter sets to estimate a set of control variables for the chemical plant that optimizes an objective function, wherein the step of discarding results outside the limits of the parameter sets comprises: projecting the parameter sets into a lower first-dimensional space to provide projected parameters;Identify out-of-bounds results by using projections of the parameters onto the lowest first-dimensional space; and discarding at least some of the parameter sets corresponding to the out-of-bounds results.

[006] As will be explained in more detail below, the use of a lower dimensional space for preprocessing parameter sets can enable more accurate detection of faulty data. Faulty data can result, for example, from an error in a detected parameter. This is particularly effective in the context of a highly variable, non-linear chemical process, in which the methods Petition 870250088545, dated 09 / 30 / 2025, p. 14 / 56 / 30 traditional preprocessing methods (such as low-pass filtering) are unable to correctly respond to variation in the data. In particular, the proposed method can avoid the unnecessary removal of data resulting from transitions between operating modes of a chemical plant, while data from such transitions would potentially be considered an out-of-bounds result by conventional approaches.

[007] According to the invention, a method is also provided for controlling a chemical process, comprising: providing a raw material to a chemical plant; processing the raw material in the chemical plant to implement a chemical process to produce a product; repeatedly obtaining a set of parameters of the chemical process using sensors; and operating the chemical plant according to control variables based on the parameter sets using a simulation of the chemical process to estimate a set of control variables for the chemical plant that optimizes an objective function, wherein the simulation is obtained by: providing a mechanistic model of the chemical process; generating first training data using the mechanistic model; obtaining second training data by detecting a set of parameters of a test process implemented by a chemical plant;and train a statistical model of the chemical process using the first and second training data to provide a trained statistical model.

[008] As will be explained in more detail below, the use of a mechanistic model of the chemical process can generate training data that provide a broad representation of the state space in which a process model should be applied. This is important when the real-world data available for a chemical process are limited to data surrounding the preferred operating points of a chemical plant. Such real-world data are more accurate than data Petition 870250088545, dated 09 / 30 / 2025, page 15 / 56 / 30 generated, but only in relation to the limited range related to the operating modes of the chemical plant, which do not allow a model to be extrapolated to new scenarios. Furthermore, real-world data suffer from process and instrument noise, which can negatively impact model predictions. Combining the two datasets allows a model to have broad applicability without losing the accuracy obtained using real-world data.

[009] For a better understanding of the invention and to show how it can be put into practice, reference will now be made, by way of example only, to the attached drawings, in which: Figure 1 shows a schematic representation of a chemical plant for the production of e-methanol; Figure 2 shows a flowchart of a control method for a chemical plant, such as the one shown in Figure 1; Figure 3A shows a flowchart of a model training method for use in the method of Figure 2; Figure 3B shows a two-dimensional representation of training data for use in the method of Figure 3A; Figure 4 shows a flowchart of a method for detecting defective data for use in the method of Figure 2; Figure 5 shows a flowchart of another method for detecting defective data for use in the method of Figure 2; and Figure 6 shows a flowchart of an additional method for detecting defective data for use in the method of Figure 2.

[0010] Figure 1 shows a chemical plant 100. An exemplary chemical plant 100 comprises: chemical processing apparatus 101; a power source 105; a raw material source 110; a catalyst 115; one or more sensors 120; one or more controllable components 125; and a controller 130. Petition 870250088545, dated 09 / 30 / 2025, page 16 / 56 / 30

[0011] The chemical processing apparatus 101 is configured to process the raw material supplied by the raw material source 110 to produce a product 135. Preferably, the chemical process is the production of methanol using electrolytic hydrogen. The precise apparatus shown in the figure is merely illustrative.

[0012] Energy source 105 can be a source of electricity. In particular embodiments, energy source 105 is a renewable source. For example, energy source 105 may include at least one of: a solar panel; a wind turbine; and / or a hydroelectric generator. The energy can be used for some or all of the components of the chemical plant 100.

[0013] The raw material source 110 can be a supply of raw materials or it can be an apparatus that processes raw materials to provide a refined raw material 110. In the example shown in Figure 1, the raw material source 110 can be an electrolysis system, powered by the energy source 105 (in this case renewable). Carbon dioxide (CO2) can be passed to the raw material 110 from any source, such as direct air capture, CO2 recovery from flue gases or from sequestered CO2 sources.

[0014] For example, the electrolysis system may use electricity supplied by a renewable source 105 to produce hydrogen as raw material 110 for the chemical plant 100. It is also possible to consider the electrolysis system 110 as a component of the chemical processing apparatus 101, in which case the fluid source for electrolysis would be considered the raw material source 110. In both cases, the intermittent nature of the energy from the energy source 105 may produce a significant and unpredictable variation in the hydrogen production by the electrolysis system 110.

[0015] Catalyst 115 can be of any type, but for a Petition 870250088545, dated 09 / 30 / 2025, page 17 / 56 / 30 methanol plant may be a component of, or a lining of, pellets forming at least part of a reactor bed.

[0016] Catalyst 115 may be of a type that fouls or degrades over time. The performance of a catalyst 115 can vary based on temperature, pressure, and the presence of contaminants in the feedstock. It is not usually possible to detect the catalyst's condition directly. Therefore, this must be inferred from the set of parameters obtained from sensor data using a model as discussed below. Catalysts are often temperature sensitive, and overheating can result in permanent damage that reduces catalyst activity and selectivity, both of which are used in chemical plant design. Therefore, maintaining catalyst activity and selectivity is fundamental for optimal and safe production.

[0017] The 120 sensors (there may be more than those shown) can be conventional sensors for detecting chemical process parameters. The chemical process parameters may include one or more of: raw material parameters, chemical plant parameters and / or product parameters.

[0018] The sensors 120 may be located, mounted, or integrated within components of the chemical processing apparatus 101. The sensors 120 communicate sensor data to the controller 130. Collectively, the sensors 120 provide a set of parameters that represent a state of the chemical plant 100 and the chemical process. Hereafter, the phrase “set of parameters” refers to a plurality of values ​​that represent the quantity measured by each sensor. For example, these may be presented as a vector. For example, a set of parameters may be a measurement each of temperature, inlet pressure, outlet pressure, and flow rate.

[0019] The 120 sensors can simultaneously provide data from Petition 870250088545, dated 09 / 30 / 2025, page 18 / 56 / 30 sensor or provide sensor data at different times. However, the controller 130 can form a set of parameters from the sensor data that represent a common time. This can be done, for example, by grouping detected data from similar times or by interpolating the time series of detected values ​​obtained from each sensor 120 to obtain a set of parameters for a specific time. Regardless of the particular method, the controller 130 is prepared to use the sensors 120 to repeatedly obtain a set of parameters from the chemical process. The sensors 120 may include one or more of: a temperature sensor; a flow sensor; a valve sensor to detect the open state of a valve; a pH sensor; a pressure sensor; a concentration sensor; etc.

[0020] The controllable components 125 (there may be more than those represented) may include one or more of: an actuator; a valve; a pump; a heater; a cooler; and / or an agitator. The controllable components 125 may be controlled by the controller 130 to influence the chemical process.

[0021] The controller 130 communicates with the sensors 120 and controllable components 125 to monitor and influence the chemical process. The controller 130 may include a local processor or a remote processor. For example, the controller 130 may include a computer. The controller 130 may include storage, such as a local historian, and / or store data on a remote server.

[0022] The controller 130 may include or implement one or more control systems arranged to maintain certain parts of the chemical process at setpoints. In this context, a setpoint is a desired parameter of the chemical process (whether directly measurable by the sensors 120 or not). For example, a setpoint may define a particular flow rate along a conduit of the chemical processing apparatus. Petition 870250088545, dated 09 / 30 / 2025, page 19 / 56 / 30 101, while another setpoint may define a particular temperature within a chamber or container of the chemical processing apparatus 101.

[0023] Controllable components 125 can be more or less direct in their influence on setpoints. For example, a heating jacket can be a controllable component 125 that directly heats a vessel, with the vessel temperature being desired to match a setpoint. Alternatively, the influence can be indirect, such as a downstream flow rate of the product being the result of a chemical reaction in the vessel, influenced by that heating jacket, with the flow rate being desired to match a setpoint. Furthermore, setpoints may not be directly measurable but can be inferred.

[0024] The controller 130 is arranged to control the controllable components 130 to influence the chemical process. For example, a controllable component 125 might be a heater, and the controller 130 might increase the heat output of the heater. As another example, a controllable component 125 might be a continuously variable valve, and the controller 130 might open or close the valve to modify a flow rate.

[0025] Controller 130 may include a processor and be able to access a model of the chemical process. The model can be used to simulate the chemical process using the parameter set as inputs. It may be preferable to preprocess the parameter set to remove noise. This can be done in the conventional way, for example, using a low-pass filter on each parameter individually to remove short-term effects, so that the data more closely represent an overall trend. Preferably, however, as will be described below, entire sets of parameters can be classified as out-of-bounds results using a faulty data detection and discarded method. Petition 870250088545, dated 09 / 30 / 2025, page 20 / 56 / 30 so that they are not used by controller 130 in the chemical process simulation. That is, the entire set of parameters for a given time can be discarded.

[0026] Controller 130 uses the remaining parameter sets to estimate a set of control variables for the chemical plant. It estimates the control variables using the model and parameter sets to simulate the chemical process and selecting the control variables that optimize the chemical process with respect to some objective function. The objective function can be any suitable function, but it can be a function of one or more of: product yield (e.g., methanol yield); energy efficiency of the chemical plant; catalyst performance; catalyst lifespan; financial profitability of the chemical plant; raw material usage; etc.

[0027] The 130A model can be trained. The 130A model can include a set of model coefficients. For example, the 130A model can include a set of model coefficients for the functions that enable the 130 controller to simulate the chemical process and thus establish the effect of the control variables on the objective function for a given set of parameters.

[0028] The model coefficients can be derived from training data in a known manner or by the method set out below.

[0029] The 130A model may be a mechanistic model. Such a model may comprise systems of deterministic equations describing underlying physical processes, with the coefficients of these equations being the model coefficients.

[0030] The 130A model may be a statistical model. Such a model may comprise a representation of the data used to train the model, with the model coefficients defining the representation.

[0031] A preferred statistical model is a Gaussian process or Petition 870250088545, dated 09 / 30 / 2025, page 21 / 56 / 30 a multi-objective Gaussian process, which can be derived from training data by Gaussian process regression. This technique is known in mathematics. An explanation of how Gaussian process regression can be used to model chemical processes can be found in “A Bayesian data modelling framework for chemical processes using adaptive sequential design with Gaussian process regression” by L. Fleming et al, Applied Stochastic Models in Business and Industry, vol. 38, no. 5, pp. 787-805, 2022. A preferred kernel function for this application is Matern.

[0032] An alternative statistical model is a neural network. This technique is known in mathematics. An explanation of how neural networks can be used to model chemical processes can be found in “The Rise of Neural Networks for Materials and Chemical Dynamics” by M Kulichenko et al, J. Phys. Chem. Lett., 12(26), 6227-6243, 2021.

[0033] Figure 2 is a flowchart of a method for controlling a chemical plant 100.

[0034] The method comprises the steps of: providing a raw material 210; processing the raw material to produce a product 220; repeatedly obtaining a set of parameters using sensors 230; deriving a set of control variables 240; and operating the chemical plant based on the sets of parameters 250.

[0035] The step of supplying a raw material 210 may involve supplying raw material to a chemical plant 100 at an unpredictable rate.

[0036] For example, the step of supplying a raw material 210 may involve supplying hydrogen to a chemical plant 100. This may involve using an electrolysis system 110 to obtain the hydrogen. The electrolysis system 110 may be powered by a renewable energy source 105. In this way, the rate of hydrogen supply depends on unpredictable wind conditions. Petition 870250088545, dated 09 / 30 / 2025, page 22 / 56 / 30

[0037] The raw material processing step 220 may involve processing the raw material at the chemical plant 100 to implement a chemical process to produce methanol (e.g., using hydrogen as a raw material).

[0038] The step of repeatedly obtaining a set of chemical process parameters using sensors 230 preferably involves repeatedly measuring at least one chemical process parameter using at least one sensor 120 and providing this to the controller 130. More preferably, this involves repeatedly measuring at least one chemical process parameter using each of a plurality of sensors 120. Optionally, this additionally or alternatively involves repeatedly estimating indirectly (i.e., not measuring) at least one chemical process parameter using the plurality of sensors 120.

[0039] It is not essential that the sensors 120 obtain sensor data simultaneously. Some parameters may vary more slowly than others and therefore it may be unnecessary to measure or estimate this parameter with the same frequency as a faster-changing parameter. Even so, the controller 130 is capable of obtaining a set of parameters from the sensor data.

[0040] A set of parameters can be, for example, a vector that represents the value of each parameter at a given time (for example, whether measured, interpolated, or estimated).

[0041] The step of deriving a set of control variables 240 may include estimating the set of control variables for the chemical plant that optimizes a particular objective function. To achieve this, the controller 130 simulates the chemical process using the parameter sets to estimate the set of control variables that optimize the objective function. The simulation is achieved using a model, such as a mechanistic or statistical model. Preferably, the model used in the step Petition 870250088545, dated 09 / 30 / 2025, page 23 / 56 / 30 240 is a statistical model, optionally trained using the method in Figure 3A. The parameter set is an input to the statistical model.

[0042] The objective function may include one or more terms (e.g., a weighted sum) related to different desired outcomes. An example objective function might include a term related to a methanol production rate. In step 240, controller 130 can simulate the chemical process using the parameter set (and optionally the previously obtained parameter time series and / or chemical process states derived from them) to derive the control variables that would produce the best value of the objective function (e.g., those that maximize the methanol production rate).

[0043] The step of operating the chemical plant 100 based on the parameter sets 250 may therefore involve operating the chemical plant 100 according to control variables derived by simulating the chemical process using the parameter sets.

[0044] The chemical plant 100 can be configured to operate in one of a plurality of operating modes. Each operating mode can be defined by a plurality of control variables being within their respective predefined ranges. The controller 130 can operate the chemical plant 100 according to an operating mode by automatically adjusting the control variables within their respective predefined ranges for that operating mode.

[0045] In general, the controller 130 can be programmed to operate the chemical plant 100 by adjusting the control variables within the respective predefined ranges of one or more operating modes.

[0046] Figure 3A shows a flowchart of a method for training a statistical model 300 for use in the method of Figure 2, in which the chemical process is simulated using a statistical model in the step Petition 870250088545, dated 09 / 30 / 2025, page 24 / 56 / 30 240 in order to derive the control variables.

[0047] In this mode, the statistical model training method is a two-stage process (the process uses what is known in mathematics as transfer learning). Two training datasets are obtained, a first training dataset and a second training dataset (despite the terms “first” and “second”, the order in which the data are obtained is not important). The statistical model is then trained (i.e., its model coefficients are determined) on both the first and second training datasets.

[0048] Although it is possible to simply group the first and second training data sets, it is preferable to use the transfer learning approach, as is well known in the field. In this context, the first training data set would be considered the source task data and the second training data set would be considered the target task data.

[0049] The method comprises the steps of: providing a mechanistic model 310; generating first training data from the mechanistic model 320A; obtaining second training data from a real process 320B; and training a statistical model of the chemical process by means of transfer learning using the first training and second training data 340. It is important to note that the order of the first and second training data is not important and not limiting (e.g., step 320B can be the first). Preferably, the faulty data detection methods 400, 500, 600 discussed below are applied to the second training data in step 330.

[0050] The step of providing a mechanistic model of the chemical process 310 may involve providing a set of deterministic equations that describe underlying physical processes within the product plant. Petition 870250088545, dated 09 / 30 / 2025, page 25 / 56 / 30 chemicals 100. As is known in the field, the coefficients of these equations can be derived from real-world historical data or can be derived analytically from first principles and / or estimates.

[0051] The step of generating initial training data using the mechanistic model 320A may include simulating the use of the chemical plant 100 through an initial range of control variables and parameter values. The initial range is not a one-dimensional range, but the range for all possible control variables and parameter values.

[0052] For example, initial training data can be generated by analyzing the mechanistic model's response to a range of inlet flow rates (e.g., from the lowest possible to the highest possible) in combination with a range of inlet compositions (e.g., from the richest to the poorest) in combination with a range of reactor temperatures (e.g., from the highest heat input to the lowest heat input). Furthermore, the model's uncontrolled variable can be perturbed to provide extra data not obtainable from a real chemical plant.

[0053] The first range may be a range that encompasses extreme values ​​of the control variables and parameter values ​​in a plurality of permutations.

[0054] The first range may extend beyond the collective ranges of control variables and parameter values ​​defined for the operating modes of the chemical plant 100 that the controller 300 is programmed to implement.

[0055] In other words, generating initial training data using the mechanistic model may involve generating training data covering an initial range of at least one of the parameter sets using control variables outside the range used in Petition 870250088545, dated 09 / 30 / 2025, page 26 / 56 / 30 one or more modes of operation of the chemical plant.

[0056] Step 320B of obtaining second training data involves using a chemical plant (preferably chemical plant 100 that will use the statistical model) to implement the chemical process. The sensors 120 of chemical plant 100 are used to detect sets of parameters as the chemical plant is used to perform a test process.

[0057] In some cases, it may be appropriate to obtain secondary training data from a chemical plant different from the chemical plant 100 in which the statistical model will be employed.

[0058] The test process can be designed to mimic the expected operating range of the control variables within the operating modes of the chemical plant 100 for the chemical process.

[0059] In other words, obtaining second training data using a chemical plant may involve obtaining training data covering a second range of at least one parameter from the parameter set using control variables within the range used in one or more operating modes of the chemical plant.

[0060] In practice, it may be appropriate to sample the data obtained from the chemical plant to obtain the second training data. This can be done using known methods, preferably using Quasi-Monte Carlo or Monte Carlo sampling.

[0061] Consequently, the first range of the parameter set is wider than, and encompasses, the second range of the parameter set. In other words, for each parameter in the parameter set, the range of values ​​for that parameter in the first training data is greater than for the second training data.

[0062] Figure 3B shows a representation in representation Petition 870250088545, dated 09 / 30 / 2025, page 27 / 56 / 30 two-dimensional training data for use in the method of Figure 3A. Although two dimensions are shown by a simple representation, the training data would naturally have a much larger number of dimensions. As can be seen in Figure 3B, the first training data covers the state space, while the second training data provides a grouping representing the range of values ​​that would result from a mode of operation (in the example shown, only one mode of operation was used, while multiple groupings may be present for multiple modes of operation).The real data, the second training data, can enable the model to be more accurate within the expected ranges for chemical plant 100, while the data generated from the mechanistic model, the first training data, can enable the model to be more accurate in extrapolating to points not represented in the state space.

[0063] The first training data, therefore, represent a broader theoretical extension of the state space for the chemical plant 100 and the chemical process than the second training data. The second training data are representative of the preferred sub-region of this state space in which the chemical process is expected to produce the best results (e.g., minimizing the objective function).

[0064] Preferably, the number of data points in the first training data is greater than the number of data points in the second training data.

[0065] The step of training a statistical model of the chemical process via transfer learning using both the first training data and the second training data provides a trained statistical model 340. This step can use any known training methodology for the particular model being used.

[0066] For example, when a neural network is used as a model Petition 870250088545, dated 09 / 30 / 2025, page 28 / 56 / 30 statistical, for example, this can simply be trained sequentially using backpropagation on the first training data. The model can then be refined on the second training data.

[0067] As another example, when a Gaussian process (or multi-objective Gaussian process) is used as the statistical model, for example, a suitable method for training the statistical model using the first and second training data is known in the technique as coregionalization (sometimes known as co-kriging).

[0068] An example of how this can be accomplished is described in “Multi-task Gaussian Process Prediction” by Edwin V. Bonilla, Kian Ming A. Chai and Christopher KI Williams. In Advances in Neural Information Processing Systems 20: NIPS'08.

[0069] As another available alternative, well-known hierarchical Bayesian methods can be employed.

[0070] Figure 4 shows a flowchart of a faulty data detection method 400 for use in the method of Figure 2. The 400 method can be particularly beneficial for removing anomalous data from parameter sets. Anomalous data removal is important for chemical processes that include significant non-linearities, such as e-methanol production.

[0071] The faulty data detection method 400 comprises: projecting the parameter sets into a lower dimensional space 410; identifying out-of-bounds results in the lower dimensional space 415; discarding the parameter sets corresponding to the out-of-bounds results 420; and then using the remaining parameter sets in the simulation 425.

[0072] The step of projecting the parameter sets into a lower dimensional space 410 provides a projected parameter set.

[0073] For example, using training data (for example, the Petition 870250088545, dated 09 / 30 / 2025, page 29 / 56 / 30 (the first or second training data discussed above, or some combination of both), a dimensionality reduction technique known in mathematics, such as principal component analysis (PCA) or canonical correlation analysis (CCA), can be used to generate a projection function that projects the data into a smaller number of dimensions. PCA, for example, provides an ordered set of orthogonal vectors that sequentially describe the directions of greatest variance in the training data. By removing the lowest order of the vectors, the lowest-dimensional projection provides a representation of the parameter datasets that best explains the potential variability in the data. CCA provides a similar approach, identifying an ordered set of orthogonal vectors that sequentially describe the directions of greatest correlation between the parameters in the training data.

[0074] The step of identifying out-of-bounds results in the lower dimensional space 415 involves projecting the parameters into the lower dimensional space to provide a lower dimensional representation of the parameter set. The lower dimensional representations are then classified to establish whether they are out-of-bounds results. The classification of a lower dimensional representation of a parameter set as an out-of-bounds result, in turn, indicates that the parameter set from which it was derived is also an out-of-bounds result.

[0075] A preferable way to classify lower-dimensional representations as out-of-bounds or non-bounds results is to use a metric and compare it to a threshold. For example, a metric can be calculated from the lowest-dimensional representation of the parameter set, and the parameter sets discarded as out-of-bounds results or retained by comparing the metric to a threshold.

[0076] Preferred metrics include the t-squared statistic of Petition 870250088545, dated 09 / 30 / 2025, page 30 / 56 / 30 Hotelling distance or mean forecast error squared. Alternatively, Euclidean distance or Mahalanobis distance can be used.

[0077] The step of discarding the parameter sets corresponding to results outside the 420 limits is thus performed for each of the parameter sets based on an analysis of the lowest dimensional representation of that parameter set.

[0078] Only the non-discarded parameter sets are used in the simulation in step 425. The chemical plant 100 is thus operated according to control variables based on the retained parameter sets (i.e., the parameter sets retained after discarding those considered as out-of-bounds results) using a chemical process simulation to estimate a set of control variables for the chemical plant that optimizes an objective function.

[0079] In other words, in step 425 (which may represent an implementation of step 240), the chemical process is simulated using only the retained parameters (those that are not out-of-bounds results) to estimate a set of control variables for the chemical plant that optimizes an objective function.

[0080] Advantageous results were achieved by performing this data cleaning in the lowest dimensional space. Furthermore, although PCA and CCA (along with other such methods) can provide suitable lower dimensional spaces, combining multiple lower dimensional spaces may provide additional advantages.

[0081] The use of a plurality of different lower-dimensional spaces can reduce errors in identifying out-of-bounds results in the parameter sets.

[0082] As shown in Figure 5, the plurality of lower-dimensional spaces provides a plurality of projections of the sets. Petition 870250088545, dated 09 / 30 / 2025, page 31 / 56 / 30 of parameters in the respective lower dimensional spaces. The application of metrics established above can be repeated for each lower dimensional representation. A combined metric can be obtained (for example, a weighted sum of the metric calculated in each lower dimensional space) and compared to a threshold.

[0083] In an alternative approach, shown in Figure 6, the plurality of lower-dimensional spaces provides a plurality of projections of the parameter sets onto the respective lower-dimensional spaces. The application of metrics established above can be repeated for each lower-dimensional representation, and a respective limit can be applied to the metrics in each lower-dimensional space separately. A voting system can then be employed across the plurality of lower-dimensional spaces, or logic can be employed. For example, the parameter set can be discarded if it is identified as an out-of-bounds result in each lower-dimensional space (i.e., an “AND” test).

[0084] Figure 5 shows a flowchart of a method for detecting defective data 500 for use in the method of Figure 2.

[0085] The faulty data detection method 500 comprises the following steps: projecting the parameter sets into a first lower dimensional space 510A; projecting the parameter sets into a second lower dimensional space 510B; calculating a first metric in the first lower dimensional space 515a; calculating a second metric in the second lower dimensional space 515b; calculating a combined metric 518; discarding the parameter sets corresponding to out-of-bounds results 520; and then using the remaining parameter sets in the simulation 525.

[0086] The combined metric 518 may be a weighted sum of the first and second metrics. Petition 870250088545, dated 09 / 30 / 2025, page 32 / 56 / 30

[0087] The step of discarding the parameter sets corresponding to the results outside the 520 limits is thus performed for each of the parameter sets based on an analysis of the metrics of both lower-dimensional representations of that parameter set.

[0088] Figure 6 shows a flowchart of an additional method for detecting defective data 600 for use in the method of Figure 2.

[0089] The faulty data detection method 600 comprises the following steps: projecting the parameter sets into a first lower dimensional space 610a; projecting the parameter sets into a second lower dimensional space 610b; identifying out-of-bounds results in the first lower dimensional space 615a; identifying out-of-bounds results in the second lower dimensional space 615B; discarding the parameter sets corresponding to the out-of-bounds results 620; and then using the remaining parameter sets in the simulation 625.

[0090] Identifying out-of-bounds results in the first lowest dimensional space in step 615A and identifying out-of-bounds results in the second lowest dimensional space in step 615B can be performed as described above for step 420.

[0091] The step of discarding the parameter sets corresponding to the results outside the limits 620 can be performed when the results of steps 615A and 615B indicate that the respective lowest dimensional representation of the parameter set is an outlier and thus agree that the corresponding parameter set is an outlier.

[0092] The statistical method of step 240 may include a dimensionality reduction step as a preprocessing step to provide model input data in a lower dimensional space. Petition 870250088545, dated 09 / 30 / 2025, page 33 / 56 / 30 It is important to emphasize that it is preferable that the lower dimensional spaces used in steps 410, 510A, 510B, 610a, and 610b be different from any lower dimensional space forming preprocessing for the statistical model. In other words, the dimensional reduction techniques ideal for fault detection methods are not necessarily suitable for the statistical model.

[0093] On the other hand, it is preferable that the second training data used in step 340 to train the statistical model be cleaned in step 330 using one of the defective data detection methods 400, 500, 600.

[0094] Similarly, it is preferable that step 420 of method 400 simulate the chemical process using the model trained according to method 300.

[0095] The following establishes preferred arrangements in the form of clauses.

[0096] Clauses: Clause 1. A method for controlling a chemical process that includes: to supply a raw material for a chemical products factory; Process the raw material in the chemical plant to implement a chemical process to produce a product; To repeatedly obtain a set of parameters from the chemical process using sensors; and to operate the chemical plant according to control variables based on the parameter sets by: Discard results outside the limits of the parameter sets; and simulate the chemical process using the parameter sets. Petition 870250088545, dated 09 / 30 / 2025, page 34 / 56 / 30 retained to estimate a set of control variables for the chemical plant that optimizes an objective function, where the step of discarding results outside the limits of the parameter sets comprises: To project the parameter sets into a lower first-dimensional space, provide projected parameters; Identify out-of-bounds results using projections of the parameters in the lowest first-dimensional space; and discard at least some of the parameter sets corresponding to the out-of-bounds results.

[0097] Clause 2. The method of clause 1, wherein the discarding of at least some of the parameter sets corresponding to out-of-bounds results, comprises: To project the parameter sets into a second, lower-dimensional space to provide projected parameters, where the second lower-dimensional space is different from the first lower-dimensional space; Identify out-of-bounds results from the projected parameters in the second lower dimensional space; and discard the parameter sets corresponding to the out-of-bounds results in both the first and second lower dimensional spaces.

[0098] Clause 3. The method of clause 1 or clause 2, with the out-of-bounds results being identified in the first lowest dimensional space by: Obtain training data by detecting a set of parameters from a test process; Apply dimensionality reduction to the training data to identify a lower first-dimensional space; Petition 870250088545, dated 09 / 30 / 2025, p. 35 / 56 / 30 calculate a metric from the projection of the training data onto the first lowest dimensional space; Derive a limit from the projection of the training data in the first lowest dimensional space; and apply the limit to the metric calculated from the projected parameters.

[0099] Clause 4. The method of clause 2 or clause 3, where the metric is one or more of: Hotelling's t-squared statistic; or mean squared forecast error.

[00100] Clause 5. The method of any preceding clause, which additionally comprises deriving the first lowest dimensional space of the parameter sets using at least one of: principal component analysis; canonical component analysis.

[00101] Clause 6. The method of any preceding clause, wherein the simulation of the chemical process using the retained parameter sets comprises projecting the parameter sets into a third lower dimensional space to provide input parameters for the simulation, wherein the third lower dimensional space is different from the first lower dimensional space.

[00102] Clause 7. The method of any preceding clause, wherein the simulation of the chemical process using the retained parameter sets comprises the use of a Gaussian process model.

[00103] Clause 8. A method for controlling a chemical process that includes: to supply a raw material for a chemical products factory; Process the raw material in the chemical plant to implement a chemical process to produce a product; repeatedly obtain a set of process parameters Petition 870250088545, dated 09 / 30 / 2025, page 36 / 56 / 30 chemical using sensors; and operate the chemical plant according to control variables based on parameter sets using a chemical process simulation to estimate a set of control variables for the chemical plant that optimizes an objective function, where the simulation is obtained by: to provide a mechanistic model of the chemical process; Generate initial training data using the mechanistic model; Obtain secondary training data by detecting a set of parameters from a test process implemented by a chemical plant; and train a statistical model of the chemical process using the first and second training data to provide a trained statistical model.

[00104] Clause 9. The method of clause 8, wherein the step of training a statistical model of the chemical process using the first and second training data to provide a trained statistical model uses transfer learning.

[00105] Clause 10. The method of clause 8, wherein the step of training a statistical model of the chemical process using the first and second training data to provide a trained statistical model uses co-regionalization.

[00106] Clause 11. Method, in accordance with any of clauses 8 to 10, wherein: Generating initial training data using the mechanistic model involves generating training data covering an initial range of at least one of the parameter sets; Obtaining secondary training data involves Petition 870250088545, dated 09 / 30 / 2025, page 37 / 56 / 30 obtaining training data that covers a second range of at least one of the parameter sets; and the first range is wider and covers the second range.

[00107] Clause 12. Method, in accordance with any of clauses 8 to 11, wherein: Generating initial training data using the mechanistic model involves generating training data from an initial set of data points; Obtaining a second set of training data involves obtaining training data with a second set of data points; and the first set of data points is larger than the second set of data points.

[00108] Clause 13. Method, in accordance with any of clauses 8 to 12, wherein: The chemical plant is configured to operate in one of a plurality of operating modes; Each operating mode is defined by a set of control variables within their respective predefined ranges; and generating initial training data using the mechanistic model involves generating training data covering an initial range of at least one of the parameter sets using control variables outside the ranges used in the plurality of operating modes.

[00109] Clause 14. Method, in accordance with any of clauses 8 to 13, wherein: Obtaining secondary training data involves implementing the chemical process in the chemical plant; and detecting a set of parameters of the chemical process using the sensors. Petition 870250088545, dated 09 / 30 / 2025, page 38 / 56 / 30

[00110] Clause 15. Method, in accordance with any of clauses 8 to 14, wherein: Obtaining second training data involves implementing the chemical process in an additional chemical plant; and detecting a set of parameters of the chemical process using a set of sensors from the additional chemical plant.

[00111] Clause 16. The method of any of clauses 8 to 15, wherein operating the chemical plant according to control variables based on the parameter sets using a simulation of the chemical process involves discarding results outside the limits of the parameter sets.

[00112] Clause 17. Method for controlling a chemical process, characterized by operating the chemical plant according to control variables based on parameter sets using a simulation of the chemical process to estimate a set of control variables for the chemical plant that optimizes an objective function comprises: Discard results outside the limits of the parameter sets; and simulate the chemical process using the retained parameter sets to estimate a set of control variables for the chemical plant that optimizes an objective function, where the step of discarding results outside the limits of the parameter sets comprises: To project the parameter sets into a lower first-dimensional space, provide projected parameters; identify out-of-bounds results using projections of the parameters in the first lowest dimensional space; and Petition 870250088545, dated 09 / 30 / 2025, page 39 / 56 / 30 discard at least some of the parameter sets corresponding to the out-of-limits results.

[00113] Clause 18. Method for controlling a chemical process, characterized by: Obtaining secondary training data by detecting a set of parameters from a test process implemented by a chemical plant involves discarding results outside the limits of the secondary training data parameter sets; and the step of discarding results outside the limits of the parameter sets comprises: To project the parameter sets into a lower first-dimensional space, provide projected parameters; Identify out-of-bounds results using projections of the parameters in the lowest first-dimensional space; and discard at least some of the parameter sets corresponding to the out-of-bounds results.

[00114] Clause 19. The method of clause 17 or clause 18, in which discarding at least some of the parameter sets corresponding to out-of-bounds results comprises: To project the parameter sets into a second, lower-dimensional space to provide projected parameters, where the second lower-dimensional space is different from the first lower-dimensional space; Identify out-of-bounds results from the projected parameters in the second lower dimensional space; and discard the parameter sets corresponding to the out-of-bounds results in both the first and second lower dimensional spaces.

[00115] Clause 20. The method of any of clauses 17 to 19, Petition 870250088545, dated 09 / 30 / 2025, page 40 / 56 / 30, where results outside the limits are identified in the first lowest dimensional space by: Obtain training data by detecting a set of parameters from a test process; Apply dimensionality reduction to the training data to identify a lower first-dimensional space; Calculate a metric from the projection of training data onto the first, lowest dimensional space; Derive a limit from the projection of the training data in the first lowest dimensional space; and apply the limit to the metric calculated from the projected parameters.

[00116] Clause 21. The method of any of clauses 18 to 19, where the metric is one or more of the following: Hotelling's t-squared statistic; or mean squared forecast error.

[00117] Clause 22. The method of any of clauses 17 to 21, which additionally comprises deriving the first lowest dimensional space of the parameter sets using at least one of: principal component analysis; canonical component analysis.

[00118] Clause 23. The method of any of clauses 17 to 22, wherein the simulation of the chemical process using the retained parameter sets comprises projecting the parameter sets into a lower third-dimensional space to provide input parameters for the simulation, wherein the lower third-dimensional space is different from the lower first-dimensional space.

[00119] Clause 24. The method of any of clauses 17 to 23, wherein the simulation of the chemical process using the retained parameter sets comprises the use of a Gaussian process model.

[00120] Clause 25. The method of any of clauses 8 to 24, Petition 870250088545, dated 09 / 30 / 2025, page 41 / 56 / 30, where the statistical model is obtained by Gaussian process regression.

[00121] Clause 26. The method of any previous clause, where: The chemical plant is configured to operate in one of a plurality of operating modes; and each operating mode is defined by a set of control variables.

[00122] Clause 27. The method of any previous clause, provided that the chemical process is powered at least in part by renewable energy.

[00123] Clause 28. The method of any previous clause, wherein the product is methanol.

[00124] Clause 29. The method of any previous clause, provided that the raw material is a renewable source or that the raw material is produced using renewable energy.

[00125] Clause 30. The method of any previous clause, provided that the chemical plant includes a catalyst.

[00126] Clause 31. The method of any previous clause, provided that the raw material is produced by electrolysis powered by at least one renewable source. Petition 870250088545, dated 09 / 30 / 2025, pp. 42-56

Claims

1 / 6 CLAIMS 1. A method for controlling a chemical process, characterized in that it comprises: supplying a raw material to a chemical plant; processing the raw material in the chemical plant to implement a chemical process to produce a product; repeatedly obtaining a set of parameters from the chemical process using sensors; and operating the chemical plant according to control variables based on the parameter sets by: discarding results outside the limits of the parameter sets; and simulating the chemical process using the retained parameter sets to estimate a set of control variables for the chemical plant that optimizes an objective function, wherein the step of discarding results outside the limits of the parameter sets comprises: projecting the parameter sets into a lower first-dimensional space to provide projected parameters;Identify out-of-bounds results using projections of the parameters in the lowest first-dimensional space; and discard at least some of the parameter sets corresponding to the out-of-bounds results.

2. Method according to claim 1, characterized in that discarding at least some of the parameter sets corresponding to out-of-bounds results comprises: projecting the parameter sets into a second lower dimensional space to provide projected parameters, wherein the second lower dimensional space is different from the first lower dimensional space; identifying out-of-bounds results from the projected parameters in the second lower dimensional space; and discarding the parameter sets corresponding to out-of-bounds results in both the first and second lower dimensional spaces.

3. A method according to claim 1 or claim 2, characterized in that out-of-bounds results are identified in the first lowest dimensional space by: obtaining training data by detecting a set of parameters from a test process; applying dimensionality reduction to the training data to identify a first lowest dimensional space; calculating a metric from the projection of the training data onto the first lowest dimensional space; deriving a threshold from the projection of the training data onto the first lowest dimensional space; and applying the threshold to the metric calculated from the projected parameters.

4. Method according to claim 2 or claim 3, characterized in that the metric is one or more of: Hotelling's t-square statistic; or mean squared forecast error.

5. A method according to any of the preceding claims, characterized in that it further comprises deriving the first lowest dimensional space of the parameter sets using at least one of: principal component analysis; canonical component analysis.

6. Method according to any of the preceding claims Petition 870250088545, dated 09 / 30 / 2025, page 44 / 56 3 / 6, characterized in that the simulation of the chemical process using the retained parameter sets comprises projecting the parameter sets into a lower third-dimensional space to provide input parameters for the simulation, wherein the lower third-dimensional space is different from the lower first-dimensional space.

7. A method according to any of the preceding claims, characterized in that the simulation of the chemical process using the retained parameter sets comprises the use of a Gaussian process model.

8. A method for controlling a chemical process, characterized in that it comprises: supplying a raw material to a chemical plant; processing the raw material in the chemical plant to implement a chemical process to produce a product; repeatedly obtaining a set of parameters of the chemical process using sensors; and operating the chemical plant according to control variables based on the parameter sets using a simulation of the chemical process to estimate a set of control variables for the chemical plant that optimizes an objective function, wherein the simulation is obtained by: supplying a mechanistic model of the chemical process; generating first training data using the mechanistic model; obtaining second training data by detecting a set of parameters of a test process implemented by a chemical plant; and Petition 870250088545, dated 09 / 30 / 2025, p.45 / 56 4 / 6 train a statistical model of the chemical process using the first and second training data to provide a trained statistical model.

9. Method according to claim 8, characterized in that the step of training a statistical model of the chemical process using the first and second training data to provide a trained statistical model uses transfer learning.

10. Method according to claim 8, characterized in that the step of training a statistical model of the chemical process using the first and second training data to provide a trained statistical model uses co-regionalization.

11. A method according to any one of claims 8 to 10, characterized in that: generating initial training data using the mechanistic model comprises generating training data covering a first range of at least one of the parameter set; obtaining second training data comprises obtaining training data covering a second range of at least one of the parameter set; and the first range being wider and encompassing the second range.

12. Method according to any one of claims 8 to 11, characterized in that: generating first training data using the mechanistic model comprises generating training data having a first number of data points; obtaining second training data comprises obtaining training data with a second number of data points; and the first number of data points being greater than the second number of data points.

13. A method according to any one of claims 8 to 12, characterized in that: the chemical plant is configured to operate in one of a plurality of operating modes; each operating mode is defined by a set of control variables being within their respective predefined ranges; and generating initial training data using the mechanistic model comprises generating training data covering an initial range of at least one of the parameter sets using control variables outside the ranges used in the plurality of operating modes.

14. A method according to any one of claims 8 to 13, characterized in that: obtaining secondary training data comprises implementing the chemical process in the chemical plant; and detecting a set of parameters of the chemical process using sensors.

15. A method according to any one of claims 8 to 14, characterized in that: obtaining secondary training data comprises implementing the chemical process in an additional chemical plant; and detecting a set of parameters of the chemical process using a set of sensors from the additional chemical plant.

16. Method according to any one of claims 8 to 15, characterized in that operating the chemical plant according to control variables based on parameter sets using a chemical process simulation comprises discarding results outside the limits of the parameter sets. Petition 870250088545, dated 09 / 30 / 2025, pp. 47 / 56 6 / 6 17. Method according to any one of claims 8 to 16, characterized in that the statistical model is obtained by Gaussian process regression.

18. A method according to any one of the preceding claims, characterized in that: the chemical plant is configured to operate in one of a plurality of operating modes; and each operating mode is defined by a set of control variables.

19. A method according to any of the preceding claims, characterized in that the chemical process is powered at least in part by renewable energy.

20. A method according to any of the preceding claims, characterized in that the product is methanol.

21. A method according to any of the preceding claims, characterized in that the raw material is a renewable source or that the raw material is produced using renewable energy.

22. A method according to any of the preceding claims, characterized in that the chemical plant includes a catalyst.

23. A method according to any of the preceding claims, characterized in that the raw material is produced by electrolysis powered by at least one renewable source. Petition 870250088545, dated 09 / 30 / 2025, pp. 48 / 56

Images