Ethylene concentration prediction method and system and storage medium

Abstract

The invention provides an ethylene concentration prediction method, an ethylene concentration prediction system and a computer readable storage medium. The ethylene concentration prediction method comprises the following steps: acquiring ethylene rectifying tower data, wherein the ethylene rectifying tower data comprises variables related to ethylene rectifying tower products; a principal component analysis method and a noise reduction auto-encoder are adopted to carry out dimension reduction extraction and noise reduction processing on the ethylene rectifying tower data so as to determine feature data; and inputting the feature data into a hybrid neural network model to predict the ethylene concentration, the hybrid neural network model being constructed based on a bidirectional gating circulation unit, and the preferred hyper-parameters of the hybrid neural network model being determined according to a whale optimization algorithm. According to the ethylene concentration prediction method provided by the invention, the ethylene concentration can be accurately predicted in real time, and the problems of low prediction precision and poor real-time performance caused by process state change and the like in an ethylene separation process are solved.

Description

Technical Field

[0001] The present invention relates to the field of ethylene production, and in particular to a method for predicting ethylene concentration, a system for predicting ethylene concentration, and a computer-readable storage medium. Background Art

[0002] Ethylene is an important basic raw material in the organic chemical industry. The raw materials of ethylene plants include refinery gas processed by the refinery and hydrogenated tail oil processed by the refinery, and its products include methane and ethylene. Ethylene can produce polyethylene, ethylene copolymers, etc. through polymerization reaction, ethylene oxide and acetaldehyde through oxidation reaction, ethylbenzene and alkyl aluminum through alkyl reaction, and ethylene dichloride through halogenation reaction. Therefore, ethylene is also known as the "mother of chemical industry".

[0003] The key to optimizing the ethylene separation production process is to monitor the ethylene concentration at the top of the ethylene distillation tower in real time. The real-time monitoring of the ethylene concentration can be used to control the parameters of the ethylene separation production process, thereby increasing product purity.

[0004] Traditional ethylene concentration monitoring methods rely on instruments such as online chromatographs to measure ethylene concentration. Although the measurement method based on instruments is accurate, it has a significant time lag and cannot meet the needs of real-time monitoring of ethylene concentration.

[0005] Existing soft sensing technology estimates the ethylene concentration that is difficult or impossible to measure directly in the ethylene production process by establishing a mathematical model between easily measurable process variables and target variables to predict the ethylene concentration at the top of the ethylene distillation tower, thereby optimizing production control. Neural networks are a type of soft sensing technology with powerful nonlinear mapping capabilities and are widely used in chemical process modeling. However, the complexity and variability of the ethylene production process pose challenges to the adaptability and prediction accuracy of the model. In particular, when production conditions change, existing neural network models cannot achieve accurate predictions and need to be retrained to adapt to new conditions, which is not only time-consuming but also costly.

[0006] In order to overcome the above-mentioned defects of the prior art, the art urgently needs an ethylene concentration prediction technology that can accurately predict ethylene concentration in real time and solve the problems of low prediction accuracy and poor real-time performance caused by process state changes in the ethylene separation process. Summary of the invention

[0007] A brief summary of one or more aspects is given below to provide a basic understanding of these aspects. This summary is not an exhaustive overview of all conceived aspects, and is neither intended to identify the key or critical elements of all aspects nor to define the scope of any or all aspects. Its only purpose is to give some concepts of one or more aspects in a simplified form as a prelude to a more detailed description that will be given later.

[0008] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method for predicting ethylene concentration, a system for predicting ethylene concentration, and a computer-readable storage medium, which can accurately predict ethylene concentration in real time and solve the problems of low prediction accuracy and poor real-time performance caused by changes in process conditions during ethylene separation.

[0009] Specifically, the above-mentioned ethylene concentration prediction method provided according to the first aspect of the present invention includes the steps of: acquiring ethylene distillation tower data, wherein the ethylene distillation tower data includes variables related to ethylene distillation tower products; performing dimensionality reduction extraction and denoising processing on the ethylene distillation tower data using principal component analysis and denoising autoencoder to determine feature data; and inputting the feature data into a hybrid neural network model to predict ethylene concentration, wherein the hybrid neural network model is constructed based on a bidirectional gated recurrent unit, and the preferred hyperparameters of the hybrid neural network model are determined according to a whale optimization algorithm.

[0010] Preferably, in one embodiment of the present invention, the denoising autoencoder includes an encoder and a decoder, the encoder and the decoder include two layers of fully connected neural networks, and the encoder and the decoder are symmetrical structures.

[0011] Preferably, in one embodiment of the present invention, the construction process of the encoder is:

[0012] X encoder =f(f(W 1 X+B 1 )W 2 +B 2 ),

[0013] Among them, f is the activation function, W 1 is the weight of the first layer of the fully connected neural network of the encoder, W 2 is the weight of the second layer of the fully connected neural network of the encoder, B 1 is the bias of the first layer of the fully connected neural network of the encoder, B 2 is the bias of the second fully connected neural network layer of the encoder, X encoder A first feature obtained by the encoder encoding the ethylene distillation tower data after dimension reduction;

[0014] The construction process of the decoder is:

[0015] X decoder =f(f(W 2 X encoder +B 2 )W 1 +B 1 ),

[0016] Among them, X decoder The feature data is output by the denoising autoencoder.

[0017] Preferably, in one embodiment of the present invention, the step of determining the preferred hyperparameters of the hybrid neural network model according to the whale optimization algorithm includes: in response to the prediction accuracy of the hybrid neural network model not meeting the requirements, determining the hyperparameters of the hybrid neural network model according to the whale optimization algorithm, the prediction accuracy of the hybrid neural network model being determined by an error discriminant function; training the hybrid neural network model according to the hyperparameters to update the parameters of the hybrid neural network model, and determining the prediction accuracy of the hybrid neural network model after the updated parameters; and repeating the above steps until the prediction accuracy of the hybrid neural network model meets the requirements, and determining the hyperparameters for training the hybrid neural network model as the preferred hyperparameters.

[0018] Preferably, in one embodiment of the present invention, in response to the prediction accuracy of the hybrid neural network model not meeting the requirements, the step of determining the hyperparameters of the hybrid neural network model according to the whale optimization algorithm includes: S1: in response to the accuracy of the hybrid neural network model not meeting the requirements, initializing the number of whale individuals, whale positions and algorithm parameters, the whale positions representing the hyperparameters of the hybrid neural network model; S2: calculating the fitness of the whale individual according to the whale position of the whale individual to determine the initial optimal whale individual position as the preferred solution; S3: updating the algorithm parameters and updating the whale position of the whale individual according to the algorithm parameters by bubble net predation or encircling prey or searching for prey; S4: calculating the fitness of the whale individual according to the updated whale position of the whale individual to determine the optimal whale individual position; S5: in response to the optimal whale individual position being better than the preferred solution, taking the optimal whale individual position as the preferred solution; and S6: repeating steps S3 to S5 until the termination condition is met, outputting the whale position of the preferred solution to determine the hyperparameters of the hybrid neural network model.

[0019] Preferably, in one embodiment of the present invention, the step S3 includes: in response to the probability p being greater than or equal to 0.5, the whale position of the individual whale is updated by a bubble net predation method; in response to the probability p being less than 0.5 and the absolute value of the coefficient vector A being less than 1, the whale position of the individual whale is updated by a prey encirclement method; and in response to the probability p being less than 0.5 and the absolute value of the coefficient vector A being greater than or equal to 1, the whale position of the individual whale is updated by a search predation method.

[0020] Preferably, in an embodiment of the present invention, the error discrimination function includes root mean square error, mean absolute error, mean absolute percentage error and determination coefficient.

[0021] Preferably, in one embodiment of the present invention, the variables related to the ethylene distillation tower product include ethylene feed amount, ethane feed amount, feed temperature, feed pressure, reflux ratio, reboiler heat duty and top condenser heat duty.

[0022] In addition, the above-mentioned ethylene concentration prediction system provided according to the second aspect of the present invention includes a memory and a processor. The memory stores computer instructions. The processor is connected to the memory and is configured to execute the computer instructions stored in the memory to implement the ethylene concentration prediction method provided by any one of the above-mentioned embodiments.

[0023] In addition, the computer-readable storage medium provided according to the third aspect of the present invention stores computer instructions, which, when executed by a processor, implement the method for predicting ethylene concentration provided by any one of the above embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The above features and advantages of the present invention can be better understood after reading the detailed description of the embodiments of the present disclosure in conjunction with the following drawings. In the drawings, the components are not necessarily drawn to scale, and components with similar related properties or features may have the same or similar reference numerals.

[0025] Figure 1 A schematic diagram of a prediction system for ethylene concentration according to some embodiments of the present invention is shown;

[0026] Figure 2 A flow chart showing a method for predicting ethylene concentration according to some embodiments of the present invention is shown;

[0027] Figure 3 A schematic diagram of a hybrid neural network model provided according to some embodiments of the present invention is shown;

[0028] Figure 4A flowchart of updating parameters of a hybrid neural network model according to some embodiments of the present invention is shown;

[0029] Figure 5 A flowchart of a whale optimization algorithm provided according to some embodiments of the present invention is shown;

[0030] Figure 6 A comparison diagram showing the prediction results of the hybrid neural network model provided according to an embodiment of the present invention and the prediction results of other models is shown; and

[0031] Figure 7 A comparison chart of the prediction results of the hybrid neural network model provided according to an embodiment of the present invention and the prediction results of other models is shown.

[0032] Reference numerals:

[0033] 100: Ethylene concentration prediction system;

[0034] 110: memory;

[0035] 111: Computer readable storage medium;

[0036] 120: Processor;

[0037] 200: Prediction method of ethylene concentration;

[0038] 300: Hybrid neural network model;

[0039] 310: forward gated recurrent unit;

[0040] 320: Backward Gated Recurrent Unit;

[0041] S210-S230: steps; and

[0042] S410~S440: steps. DETAILED DESCRIPTION

[0043] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. Note that the aspects described below in conjunction with the accompanying drawings and specific embodiments are only exemplary and should not be construed as limiting the scope of protection of the present invention in any way.

[0044] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0045] In addition, the terms "upper", "lower", "left", "right", "top", "bottom", "horizontal" and "vertical" used in the following description should be understood as the directions shown in the paragraph and the related drawings. Such relative terms are only used for the convenience of description and do not mean that the device described therein must be manufactured or operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0046] It is understood that although the terms "first", "second", "third", etc. may be used herein to describe various components, regions, layers and / or parts, these components, regions, layers and / or parts should not be limited by these terms, and these terms are only used to distinguish different components, regions, layers and / or parts. Therefore, the first component, region, layer and / or part discussed below may be referred to as a second component, region, layer and / or part without departing from some embodiments of the present invention.

[0047] As mentioned above, existing soft sensing technology estimates the ethylene concentration that is difficult or impossible to measure directly in the ethylene production process by establishing a mathematical model between the easily measurable process variables and the target variables to predict the ethylene concentration at the top of the ethylene distillation tower, thereby optimizing production control. Neural networks are a type of soft sensing technology with powerful nonlinear mapping capabilities and are widely used in chemical process modeling. However, the complexity and variability of the ethylene production process pose challenges to the adaptability and prediction accuracy of the model. In particular, when production conditions change, the existing neural network model cannot achieve accurate predictions and needs to be retrained to adapt to the new operating conditions, which is not only time-consuming but also costly.

[0048] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method for predicting ethylene concentration, a system for predicting ethylene concentration, and a computer-readable storage medium, which can accurately predict ethylene concentration in real time and solve the problems of low prediction accuracy and poor real-time performance caused by changes in process conditions during ethylene separation.

[0049] In some non-limiting embodiments, the above-mentioned ethylene concentration prediction method provided in the first aspect of the present invention can be implemented via the above-mentioned ethylene concentration prediction system provided in the second aspect of the present invention.

[0050] Please refer to Figure 1 , Figure 1 A schematic diagram of a system for predicting ethylene concentration according to some embodiments of the present invention is shown.

[0051] like Figure 1 As shown, the ethylene concentration prediction system 100 may be configured with a memory 110 and a processor 120. The memory 110 includes but is not limited to the computer-readable storage medium 111 provided in the third aspect of the present invention, on which computer instructions are stored. The processor 120 is connected to the memory 110 and is configured to execute the computer instructions stored in the memory 110 to implement the ethylene concentration prediction method provided in the first aspect of the present invention.

[0052] The following will first describe the working principle of the above-mentioned ethylene concentration prediction system in conjunction with some examples of ethylene concentration prediction methods. Those skilled in the art will understand that these examples of ethylene concentration prediction methods are only some non-limiting implementation methods provided by the present invention, and are intended to clearly demonstrate the main concept of the present invention and provide some specific solutions that are convenient for the public to implement, rather than to limit all functions or all working methods of the ethylene concentration prediction system. Similarly, the ethylene concentration prediction system is also only a non-limiting implementation method provided by the present invention, and does not constitute a limitation on the execution subject and execution order of each step in these ethylene concentration prediction methods.

[0053] Please refer to Figure 2 , Figure 2 A flow chart of a method for predicting ethylene concentration provided according to some embodiments of the present invention is shown.

[0054] like Figure 2 As shown, the method 200 for predicting ethylene concentration may include step S210: acquiring ethylene distillation tower data, where the ethylene distillation tower data includes variables related to ethylene distillation tower products.

[0055] In some embodiments, the above-mentioned ethylene concentration prediction method provided in the first aspect of the present invention can be implemented in two parts: an offline training stage and an online analysis stage.

[0056] In the online analysis stage, the prediction system of ethylene concentration can select variables related to the ethylene distillation tower as input for predicting ethylene concentration. In the offline training stage, the ethylene distillation tower data can also include actual ethylene concentration. The prediction system of ethylene concentration can obtain actual ethylene concentration as output to realize the offline training stage of the prediction method of ethylene concentration while obtaining ethylene distillation tower data. In some embodiments, the actual ethylene concentration can be the impurity concentration at the top of the tower measured by the analyzer. The variables related to the ethylene distillation tower product can include ethylene feed amount, ethane feed amount, feed temperature, feed pressure, reflux ratio, reboiler heat load and tower top condenser heat load.

[0057] In the offline training stage, the ethylene distillation tower data including variables related to the ethylene distillation tower products and the actual ethylene concentration can be divided into a training set, a test set and a validation set. In addition, the ethylene concentration prediction system can perform standardized preprocessing on the ethylene distillation tower data.

[0058] like Figure 2 As shown, the method 200 for predicting ethylene concentration may include step S220: using principal component analysis and denoising autoencoder to perform dimension reduction extraction and denoising on ethylene distillation tower data to determine characteristic data.

[0059] First, the prediction system of ethylene concentration can perform principal component analysis on the original ethylene distillation tower data to reduce the dimension to obtain the data set X. Specifically, the prediction system of ethylene concentration can calculate the characteristic covariance matrix based on the ethylene distillation tower data to determine the linear relationship between the characteristics of the ethylene distillation tower data. Then, the prediction system of ethylene concentration can calculate the eigenvalues ​​and corresponding eigenvectors of the covariance matrix. Then, the eigenvectors whose cumulative contribution reaches a preset threshold are selected to obtain the reduced-dimensional data set X. In some embodiments, the preset threshold can be 80%.

[0060] Afterwards, the ethylene concentration prediction system can perform denoising on the dimensionally reduced dataset X based on the constructed and trained denoising autoencoder.

[0061] The denoising autoencoder may include an encoder and a decoder, the encoder and the decoder may include two layers of fully connected neural networks, and the encoder and the decoder may be symmetrical structures.

[0062] Specifically, in a non-limiting embodiment, the construction of the denoising autoencoder includes the construction of an encoder and a decoder. The construction process of the encoder is:

[0063] X encoder =f(f(W 1 X+B 1 )W 2 +B 2 ),

[0064] Among them, f is the activation function, W 1 is the weight of the first layer of the fully connected neural network of the encoder, W 2 is the weight of the second layer of the fully connected neural network of the encoder, B 1 is the bias of the first layer of the fully connected neural network of the encoder, B 2 is the bias of the second fully connected neural network of the encoder, X encoder It is the first feature obtained by the encoder for encoding the ethylene distillation tower data after dimension reduction, that is, the first feature obtained by the encoder for encoding the data set X.

[0065] The construction process of the decoder is:

[0066] X decoder =f(f(W 2 X encoder +B 2 )W 1 +B 1 ),

[0067] Among them, X decoder is the feature data output by the denoising autoencoder. In this non-limiting embodiment, the decoder and the encoder are symmetrical structures and include two layers of fully connected neural networks. Preferably, the activation function f can adopt a RELU (Rectified Linear Unit) activation function.

[0068] Next, the constructed denoising autoencoder is trained to denoise the ethylene distillation tower data after dimension reduction. In some embodiments, the prediction system for ethylene concentration can use mean square error as an evaluation index for evaluating the denoising effect. The characteristic data obtained after dimension reduction and denoising of the ethylene distillation tower data can be used as the input of the hybrid neural network model provided by the prediction method for ethylene concentration.

[0069] In this way, the ethylene concentration prediction system can not only significantly reduce the computational complexity of the subsequent hybrid neural network model by implementing dimensionality reduction and noise reduction processing on the input ethylene distillation tower data, but also significantly enhance the generalization ability of the hybrid neural network model for new data, thereby improving the effectiveness and reliability of the hybrid neural network model in actual industrial applications.

[0070] Please continue to refer to Figure 2 The method 200 for predicting ethylene concentration may include step S230: inputting feature data into a hybrid neural network model to predict ethylene concentration, wherein the hybrid neural network model is constructed based on a bidirectional gated recurrent unit, and the preferred hyperparameters of the hybrid neural network model are determined according to a whale optimization algorithm.

[0071] LSTM (Long Short-Term Memory) can reduce the vanishing gradient by introducing a special gate structure, allowing the error to propagate over a long distance, thus having the ability of long-term memory. GRU (Gated Recurrent Unit) is an improved LSTM algorithm. GRU merges the forget gate and the input gate into an update gate, and merges the data state and the hidden state. Compared with LSTM, GRU has a simpler structure and faster training speed.

[0072] The prediction system of ethylene concentration can construct a hybrid neural network model based on a bidirectional gated recurrent unit. In some embodiments, the hybrid neural network model can be a WOA-BiGRU (Whale Optimization Algorithm-Bidirectional Gated Recurrent Unit) model.

[0073] Please refer to Figure 3 , Figure 3 A schematic diagram of a hybrid neural network model provided according to some embodiments of the present invention is shown.

[0074] like Figure 3 As shown, the hybrid neural network model 300 may be composed of bidirectional gated recurrent units, which include a forward gated recurrent unit 310 and a backward gated recurrent unit 320. The forward gated recurrent unit 310 and the backward gated recurrent unit 320 may each include a plurality of gated recurrent units GRU. The input of the hybrid neural network model 300 may include inputs of different time steps, x t is the input of the current time step, x t-1 is the input of the previous time step, x t+1 The output of the hybrid neural network model 300 can be the hidden state of different time steps, h t is the hidden state of the current time step, h t-1 is the hidden state of the previous time step, h t+1 is the hidden state at the next time step.

[0075] The gated recurrent unit GRU merges the forget gate and input gate of LSTM into an update gate, which can be used to control which information in the state information of the current time step needs to flow into the candidate state of the next time step.

[0076] like Figure 3 As shown, taking the forward gated recurrent unit 310 as an example, the input in the GRU of the forward gated recurrent unit 310 includes the hidden state h output by the GRU in the previous time stept-1 and the input x at the current time step t The specific formula for updating the gate is as follows:

[0077] z i =sigmod(W z [h t-1 , x t ]+b z ),

[0078] Among them, z t is the output of the update gate, h t-1 is the hidden state of the previous time step, x t is the input of the current time step, W z is the weight matrix of the update gate, b z is the bias vector for the update gate.

[0079] Candidate states for GRU As shown below:

[0080]

[0081] Among them, W z , W h is the weight parameter of the gated recurrent unit, b z , b h is the bias of the gated recurrent unit, r t is the output of the reset gate.

[0082] The reset gate determines the output h of the hidden layer at the previous time step. t-1 , that is, the hidden state h of the previous time step t-1 Candidate status The specific formula is as follows:

[0083] r t =sigmod(W r [h t-1 , x t ]+b r ),

[0084] Among them, W r is the weight parameter of the reset gate; b r To reset the gate bias.

[0085] Thus, the output of the GRU of the forward gated recurrent unit 310 can be determined by the output of the hidden layer of the previous time step, the output of the update gate, and the candidate state of the current time step. The output of the GRU of the forward gated recurrent unit 310 is the forward hidden state

[0086]

[0087] Accordingly, in the backward gated recurrent unit 320, the input to the GRU of the backward gated recurrent unit 320 includes the hidden state h of the GRU output in the next time step. t+1 and the input x at the current time step t Afterwards, according to the calculation of the update gate, the candidate state and the reset gate, the prediction system of ethylene concentration can determine the output of the backward gated recurrent unit 320, i.e., the reverse hidden state, based on the output of the hidden layer at the next time step, the output of the update gate and the candidate state at the current time step.

[0088] Please continue to refer to Figure 3 In the bidirectional gated recurrent unit of the hybrid neural network model 300, the feature data can be respectively transmitted to the forward gated recurrent unit 310 and the backward gated recurrent unit 320, and the forward hidden state at the current time step t and the reverse hidden state The combination of is the output of the hybrid neural network model 300, that is,

[0089] In some embodiments, the preferred hyperparameters of the constructed hybrid neural network model can be determined according to the whale optimization algorithm. Through the whale optimization algorithm, the prediction system of ethylene concentration can optimize and improve the performance of the bidirectional gated recurrent unit constituting the hybrid neural network model, obtain the preferred hyperparameter weights for training the hybrid neural network model, and then determine the parameters of the hybrid neural network model through training and verification. Hyperparameters may include the number of hidden layer nodes, the number of training rounds, and the learning rate.

[0090] Please refer to Figure 4 , Figure 4 A flowchart for updating parameters of a hybrid neural network model provided according to some embodiments of the present invention is shown.

[0091] like Figure 4 As shown, the method for predicting ethylene concentration may further include a step S410 of determining preferred hyperparameters of the hybrid neural network model: in response to the prediction accuracy of the hybrid neural network model not meeting the requirements, determining the hyperparameters of the hybrid neural network model according to the whale optimization algorithm, and the prediction accuracy of the hybrid neural network model is determined by the error discriminant function. Afterwards, executing step S420: training the hybrid neural network model according to the hyperparameters to update the parameters of the hybrid neural network model, and determining the prediction accuracy of the hybrid neural network model after the updated parameters. Repeating the above steps S410 and S420 until the prediction accuracy of the hybrid neural network model meets the requirements, determining the hyperparameters of the hybrid neural network model trained to meet the prediction accuracy requirements as the preferred hyperparameters.

[0092] In some embodiments, the ethylene concentration prediction system can train and verify the constructed hybrid neural network model in the offline training stage. When the prediction accuracy of the established hybrid neural network model does not meet the requirements, the hyperparameters of the hybrid neural network model can be determined according to the whale optimization algorithm, and the hybrid neural network model can be retrained and verified to update the parameters until the prediction accuracy meets the requirements. In the online analysis stage, the actual ethylene concentration value measured can be verified with the predicted ethylene concentration. When the prediction accuracy of the established hybrid neural network model does not meet the requirements, the hyperparameters of the hybrid neural network model can be determined according to the whale optimization algorithm, and the hybrid neural network model can be retrained and verified to update the parameters. Until the hybrid neural network model meets the prediction accuracy requirements, the ethylene concentration prediction system can determine that the hyperparameters of training the hybrid neural network model are preferred hyperparameters, and then determine the updated parameters based on the hybrid neural network model, thereby realizing dynamic adjustment of the parameters of the hybrid neural network model.

[0093] Preferably, the ethylene concentration prediction system can update the connection weights of the hybrid neural network model through the back propagation algorithm during the training of the hybrid neural network model, and measure the difference between the predicted ethylene concentration and the actual ethylene concentration through an appropriate loss function until the maximum number of training rounds is met or the error reaches the requirement, and then use historical data or simulation data to evaluate the performance of the optimized hybrid neural network model. In some embodiments, the simulation data can be obtained by simulating the operation process of the ethylene distillation tower by using the aspenplus software.

[0094] Please refer to Figure 5 , Figure 5 A flowchart of a whale optimization algorithm provided according to some embodiments of the present invention is shown.

[0095] like Figure 5 As shown, in response to the prediction accuracy of the hybrid neural network model not meeting the requirements, the step of determining the hyperparameters of the hybrid neural network model according to the whale optimization algorithm may include step S1: in response to the accuracy of the hybrid neural network model not meeting the requirements, the number of whale individuals and the whale positions and algorithm parameters are initialized, and the whale positions represent the hyperparameters of the hybrid neural network model. The algorithm parameters of the whale optimization algorithm may include the number of whale individuals, the maximum number of iterations, the parameter a, the coefficient vector A, the coefficient vector C, the probability p, and the random number l.

[0096] Afterwards, the ethylene concentration prediction system may execute step S2: calculating the fitness of the individual whales according to the individual whale positions of the individual whales to determine the position of the initial optimal individual whale as the preferred solution.

[0097] The prediction system of ethylene concentration can calculate the fitness of individual whales according to the whale positions of the individual whales based on the determined objective function.

[0098] In some embodiments, the objective function may be:

[0099]

[0100] Among them, N is the sample size, that is, the number of whale individuals, is the predicted i-th output of the hybrid neural network model, yi is the actual output value of the i-th sample, W is the set of all weights in the hybrid neural network model, b is the set of all biases in the hybrid neural network model, η is the learning rate of the hybrid neural network model, η min is the lower limit of the learning rate, η max is the upper limit of the learning rate, E is the number of training rounds of the hybrid neural network model, and E min is the lower limit of the number of training rounds, E max The upper limit of the number of training rounds.

[0101] Next, step S3 is executed: updating the algorithm parameters and updating the whale position of the individual whale by using bubble net hunting, encircling prey, or searching for prey according to the algorithm parameters.

[0102] The prediction system of ethylene concentration can update the parameters a, coefficient vector A, coefficient vector C, probability p, and random number l of the whale optimization algorithm. And the whale position of the individual whale is updated in a corresponding manner according to the updated probability p and coefficient vector A. In this process, the current individual whale determines whether to swim towards the prey in a spiral curve or in a shrinking encirclement through the probability p between (0, 1). When the probability p is greater than or equal to 0.5, it can swim towards the prey in a spiral curve, that is, the bubble net predation method is adopted to update the whale position of the individual whale; otherwise, when the probability p is less than 0.5, it swims towards the prey in a shrinking encirclement, that is, the method of encircling the prey or searching for prey is determined according to the absolute value of the coefficient vector A as the method of updating the whale position of the individual whale.

[0103] Depending on the value of probability p, the way the whale optimization algorithm updates the whale position of an individual whale can switch between spiral motion and circular motion. In response to the probability p being greater than or equal to 0.5, the whale position of an individual whale is updated by bubble net predation. Bubble net predation is a way to update the whale position of an individual whale based on the hunting behavior of humpback whales, where the bubble net predation method swims toward the prey in a spiral motion. After the spiral motion update, the whale position X(n+1) of the individual whale is as follows:

[0104] D k =|X * (n)-X(n)|,

[0105] X(n+1)=X * (n)+D k ·e bl ·cos 2πl,

[0106] Among them, D k Represents the whale position X(n) of the current whale individual and the whale position X(n) of the current optimal whale individual * (n), b is the logarithmic spiral shape constant, and l is a random number between [-1, 1].

[0107] In response to the probability p being less than 0.5 and the absolute value of the coefficient vector A being less than 1, the prediction system for ethylene concentration can update the whale position of the individual whale in a way of encircling prey.

[0108] The search scope of the whale optimization algorithm is the global solution space. The range of the prey needs to be determined first for encirclement, but the optimal position in the search space is unknown. Therefore, the whale optimization algorithm can assume that the position of the current optimal whale individual is the target prey or close to the optimal solution. After defining the position of the target prey, other whale individuals can try to encircle the position of the target prey, that is, to capture the prey. Encircling the position of the prey to update the whale position X(n+1) of the whale individual can be shown as follows:

[0109] D=|C×X * (n)-X(n)|,

[0110] X(n+1)=X * (n)-A·D,

[0111] Among them, n represents the current iteration number, A and C are the coefficient vectors of the whale optimization algorithm, and D represents the distance between the whale position of the current whale individual and the whale position of the current optimal whale individual.

[0112] In response to the probability p being less than 0.5 and the absolute value of the coefficient vector A being greater than or equal to 1, the prediction system for ethylene concentration can adopt a search and prey method to update the whale position of the individual whale.

[0113] To ensure sufficient search, a whale can randomly select the position of another whale as a reference to update its own whale position, rather than using the position of the current best whale as a reference. The whale position X(n+1) of a whale can be updated by searching and preying as follows:

[0114] X(n+1)=X rand (n)-A·D,

[0115] D=|C×X * (n)-X(n)|,

[0116] Among them, X rand (n) is the whale position of a randomly selected individual whale.

[0117] Please continue to refer to Figure 5 , the ethylene concentration prediction system can execute step S4: calculating the fitness of the whale individual according to the updated whale position of the whale individual to determine the position of the optimal whale individual.

[0118] Then, step S5 is executed: in response to the position of the optimal whale individual being better than the preferred solution, the position of the optimal whale individual is used as the preferred solution.

[0119] After the position update is completed, the ethylene concentration prediction system can recalculate the fitness of each individual whale. And determine the whale position of the optimal individual whale after the update iteration according to the fitness. After that, the whale position of the optimal individual whale after the update iteration is compared with the position of the optimal individual whale in the previous iteration round (i.e., the current optimal solution). If the whale position of the optimal individual whale after the update iteration is better than the optimal solution, the whale position of the optimal individual whale after the update iteration is used to replace the current optimal solution as the new optimal solution.

[0120] Afterwards, the ethylene concentration prediction system may execute step S6: repeating steps S3 to S5 until the termination condition is met, and outputting the whale position of the optimal solution to determine the hyperparameters of the hybrid neural network model.

[0121] The prediction system of ethylene concentration can determine whether the current number of iterations reaches the maximum number of iterations as a termination condition. In some embodiments, the termination condition can be that the fitness of the optimal whale individual no longer decreases as the number of iterations increases. If the termination condition is reached, the prediction system of ethylene concentration can output the preferred solution as the optimal solution through the whale optimization algorithm; if the termination condition is not reached, the next iteration is entered.

[0122] In this way, by adopting the whale optimization algorithm as a hyperparameter optimization strategy, the performance of the hybrid neural network model is further optimized by accurately adjusting the key non-learning parameters of the model, such as the learning rate and the number of nodes in the hidden layer. This optimization not only enhances the generalization ability of the hybrid neural network model and avoids overfitting, but also speeds up the convergence speed of the hybrid neural network model and improves the training efficiency.

[0123] like Figure 4As shown in the figure, in order to comprehensively evaluate the prediction performance of the hybrid neural network model, the prediction system of ethylene concentration uses an error discriminant function to determine the prediction accuracy of the hybrid neural network model. By minimizing the error discriminant function, the prediction system of ethylene concentration can dynamically adjust the hyperparameters of the hybrid neural network model through the whale optimization algorithm to train the hybrid neural network model and update the parameters, thereby achieving accurate prediction of ethylene concentration.

[0124] The error discrimination function uses a variety of evaluation indicators, including but not limited to root mean square error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and determination coefficient (R 2 ). Through multi-dimensional evaluation, the ethylene concentration prediction system can integrate the advantages of different metrics to ensure that the established hybrid neural network model achieves the best prediction accuracy and robustness.

[0125] In the specific calculation process of the error discriminant function, the root mean square error is the square root of the ratio of the square of the predicted ethylene concentration value to the actual ethylene concentration value and the number of observations (i.e., the number of samples) N, that is, the predicted value and the actual value t i The square root of the ratio of the square of the deviation to the number of observations N:

[0126]

[0127] When the predicted value Close to the actual value t i When the predicted value is Far from the actual value t i When , the RMSE value will become larger. Therefore, in the prediction of ethylene concentration, the smaller the RMSE, the better.

[0128] The mean absolute error is the average of the absolute errors, that is, the predicted value and the actual value t i The average error between:

[0129]

[0130] When the predicted value Close to the actual value t i When the predicted value With the actual value t i As the error between the two increases, the MAE value will increase linearly. Therefore, in the prediction of ethylene concentration, the smaller the MAE, the better.

[0131] The mean absolute percentage error is the predicted value With the actual value t i The average percentage of relative error between:

[0132]

[0133] Mean absolute percentage error is expressed as a percentage of the predicted value and the actual value t i The relative errors between them are more comparable for prediction problems of different magnitudes.

[0134] The coefficient of determination is the proportion of the sum of squares explained by the hybrid neural network model:

[0135]

[0136] in, is the average value of the actual value of ethylene concentration.

[0137] The determination coefficient can reflect the degree of fit of the hybrid neural network model and measure the degree to which the hybrid neural network model explains the changes in the dependent variable. Its value is usually between 0 and 1. The larger the value of the determination coefficient, the better the hybrid neural network model fits the data.

[0138] The whale optimization algorithm is used to optimize the hyperparameters of the hybrid neural network model, and the model parameters are dynamically adjusted based on the error discriminant function, which can improve the error caused by the weakening correlation between the training data and the actual prediction data.

[0139] In this way, the prediction system of ethylene concentration can update the weights of the hybrid neural network model through the back propagation algorithm, and use the root mean square error, mean absolute error, mean absolute percentage error and determination coefficient in the error discriminant function as evaluation indicators to evaluate the performance of the hybrid neural network model. In addition, through the error discriminant function, the prediction system of ethylene concentration can dynamically adjust the parameters of the hybrid neural network model to effectively predict the ethylene concentration, for example, effectively predict the impurity concentration at the top of the ethylene distillation tower.

[0140] The following are specific non-limiting preferred embodiments, based on which the ethylene concentration prediction system proposed by the present invention is further described.

[0141] Please refer to Figure 6 , Figure 6 A flowchart of constructing a hybrid neural network model provided according to an embodiment of the present invention is shown.

[0142] like Figure 6As shown, the prediction system of ethylene concentration can first obtain a historical data set, which can be simulation data or actual scene data. Then, the prediction system of ethylene concentration can process the historical data set through principal component analysis and denoising autoencoder to obtain feature data. The feature data and bidirectional gated recurrent unit (BIGRU) are then used to construct a hybrid neural network model.

[0143] The prediction system of ethylene concentration initializes the constructed hybrid neural network model and determines the accuracy of the constructed hybrid neural network model. When the accuracy of the constructed hybrid neural network model meets the requirements, the prediction result is directly output. When the accuracy of the constructed hybrid neural network model does not meet the requirements, the hyperparameters of the constructed hybrid neural network model are determined by the whale optimization algorithm, and the initialized parameters of the constructed hybrid neural network model are updated according to the hyperparameters.

[0144] Furthermore, if Figure 6 As shown in the figure, the whale optimization algorithm can first initialize the population, calculate the fitness of each individual whale and update the algorithm parameters. The method of updating the whale position of each individual whale is determined according to the updated algorithm parameters. When the probability p is greater than or equal to 0.5, the whale position is updated by bubble net predation; when the probability p is less than 0.5, the coefficient vector A is judged again. When the probability p is less than 0.5 and the absolute value of the coefficient vector A is less than 1, the whale position is updated by encircling the prey; when the probability p is less than 0.5 and the absolute value of the coefficient vector A is greater than or equal to 1, the whale position is updated by searching and preying.

[0145] After that, the whale position is continuously updated until the termination condition is reached, and the whale position as the optimal solution for the hyperparameters of the hybrid neural network model is output.

[0146] The parameters of the constructed hybrid neural network model are updated based on the hyperparameters determined by the whale optimization algorithm, and then the accuracy of the constructed hybrid neural network model is determined. When the accuracy of the constructed hybrid neural network model does not meet the requirements, the whale optimization algorithm is continued to be used to determine the hyperparameters of the constructed hybrid neural network model. When the accuracy of the constructed hybrid neural network model meets the requirements, the constructed hybrid neural network model is used to output the prediction result of ethylene concentration.

[0147] Please refer to Figure 7 , Figure 7 A comparison chart of the prediction results of the hybrid neural network model provided according to an embodiment of the present invention and the prediction results of other models is shown.

[0148] like Figure 7As shown, ethylene distillation tower data is obtained under the stable operation state of a petrochemical plant. In order to compare fairly with other models, the feature data input to the hybrid neural network model and other models are obtained based on the same data processing method, that is, the feature data obtained after processing based on principal component analysis and denoising autoencoder. Other models can be LSTM models and BiGRU (Bidirectional Gated Recurrent Unit) models. Based on the training data set in the acquired ethylene distillation tower data, the hybrid neural network model, LSTM model and BiGRU model are trained. In the verification process, the verification data set in the ethylene distillation tower data is input to respectively predict and calculate the concentration of ethylene at the top of the ethylene distillation tower under the conditions of the current ethylene feed flow rate, ethane feed flow rate, feed temperature and feed pressure, and the predicted ethylene concentration is compared with the actual ethylene concentration in the verification data set to evaluate the hybrid neural network model, LSTM model and BiGRU model.

[0149] In the indicators for evaluating the prediction performance of the hybrid neural network model, namely the WOA-BiGRU model, the root mean square error (RMSE) is 0.045, the mean absolute error (MAE) is 0.036, the mean absolute percentage error (MAPE) is 2.38%, and the coefficient of determination (R 2 ) is 0.94; in the indicators for evaluating the prediction performance of the LSTM model, the root mean square error (RMSE) is 0.057, the mean absolute error (MAE) is 0.044, the mean absolute percentage error (MAPE) is 2.87%, and the coefficient of determination (R 2 ) is 0.89; in the indicators for evaluating the prediction performance of the BiGRU model, the root mean square error (RMSE) is 0.052, the mean absolute error (MAE) is 0.041, the mean absolute percentage error (MAPE) is 2.78%, and the coefficient of determination (R 2 ) is 0.90. It can be seen that the WOA-BiGRU model outperforms the other two models in all evaluation indicators.

[0150] In summary, the ethylene concentration prediction method provided by the present invention can construct a hybrid neural network model based on a bidirectional gated recurrent unit, and optimize the parameters of the hybrid neural network model based on the whale optimization algorithm to improve the prediction accuracy of the model and improve the training efficiency of the model, so as to achieve the adjustment of the hybrid neural network model when the process state changes, thereby solving the problem of low prediction accuracy and poor real-time performance caused by process state changes in the ethylene separation process. In addition, through the data preprocessing technology combined with the principal component analysis method and the denoising autoencoder, the ethylene concentration prediction method provided by the present invention can effectively remove the noise in the data and enhance the stability of subsequent predictions.

[0151] Although the above methods are illustrated and described as a series of actions for simplicity of explanation, it should be understood and appreciated that these methods are not limited by the order of the actions, because according to one or more embodiments, some actions may occur in a different order and / or concurrently with other actions from those illustrated and described herein or not illustrated and described herein but understandable to those skilled in the art.

[0152] Those skilled in the art will appreciate that information, signals, and data may be represented using any of a variety of different technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols, and chips cited throughout the above description may be represented by voltage, current, electromagnetic waves, magnetic fields or magnetic particles, optical fields or optical particles, or any combination thereof.

[0153] Those skilled in the art will further appreciate that the various illustrative logic blocks, modules, circuits, and algorithm steps described in conjunction with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or a combination of the two. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps are generally described above in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system. The technician may implement the described functionality in different ways for each specific application, but such implementation decisions should not be interpreted as resulting in a departure from the scope of the present invention.

[0154] The various illustrative logic modules and circuits described in conjunction with the embodiments disclosed herein may be implemented or executed with a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in cooperation with a DSP core, or any other such configuration.

[0155] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The software module may reside in a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor so that the processor can read and write information from / to the storage medium. In an alternative, a storage medium may be integrated into a processor. The processor and the storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In an alternative, the processor and the storage medium may reside in a user terminal as discrete components.

[0156] In one or more exemplary embodiments, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented as a computer program product in software, each function may be stored on or transmitted by a computer-readable medium as one or more instructions or codes. Computer-readable media include both computer storage media and communication media, including any medium that facilitates the transfer of a computer program from one place to another. Storage media may be any available medium that can be accessed by a computer. As an example and not limitation, such a computer-readable medium may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, disk storage or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of an instruction or data structure and can be accessed by a computer. Any connection is also properly referred to as a computer-readable medium. For example, if the software is transmitted from a website, a server, or other remote source using a coaxial cable, a fiber optic cable, a twisted pair, a digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwaves, the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwaves are included in the definition of the medium. Disk and disc as used herein include compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk and Blu-ray disc, wherein disk often reproduces data magnetically, while disc reproduces data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0157] The previous description of the disclosure is provided to enable any person skilled in the art to make or use the disclosure. Various modifications to the disclosure will be apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the spirit or scope of the disclosure. Thus, the disclosure is not intended to be limited to the examples and designs described herein, but should be granted the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting ethylene concentration, characterized in that: Includes steps: Obtaining ethylene distillation column data, the ethylene distillation column data including variables related to ethylene distillation column products; The principal component analysis method and the noise reduction autoencoder are used to perform dimension reduction extraction and noise reduction processing on the ethylene distillation tower data to determine the characteristic data; as well as The characteristic data are input into a hybrid neural network model to predict ethylene concentration, wherein the hybrid neural network model is constructed based on a bidirectional gated recurrent unit, and the preferred hyperparameters of the hybrid neural network model are determined according to a whale optimization algorithm.

2. The method for predicting ethylene concentration according to claim 1, characterized in that: The denoising autoencoder includes an encoder and a decoder, the encoder and the decoder include two layers of fully connected neural networks, and the encoder and the decoder are symmetrical structures.

3. The method for predicting ethylene concentration according to claim 2, characterized in that: The construction process of the encoder is: X encoder =f(f(W1X+B1)W2+B2), Wherein, f is the activation function, W1 is the weight of the first layer of the fully connected neural network of the encoder, W2 is the weight of the second layer of the fully connected neural network of the encoder, B1 is the bias of the first layer of the fully connected neural network of the encoder, B2 is the bias of the second layer of the fully connected neural network of the encoder, and X encoder A first feature obtained by the encoder encoding the ethylene distillation tower data after dimension reduction; The construction process of the decoder is: X decoder =f(f(W2X encoder +B2)W1+B1), Among them, Xd ecoder The feature data is output by the denoising autoencoder.

4. The method for predicting ethylene concentration according to claim 1, characterized in that: The steps of determining the optimal hyperparameters of the hybrid neural network model according to the whale optimization algorithm include: In response to the prediction accuracy of the hybrid neural network model not meeting the requirement, determining the hyperparameters of the hybrid neural network model according to the whale optimization algorithm, wherein the prediction accuracy of the hybrid neural network model is determined by an error discriminant function; Training the hybrid neural network model according to the hyperparameters to update the parameters of the hybrid neural network model, and determining the prediction accuracy of the hybrid neural network model after the updated parameters; and Repeat the above steps until the prediction accuracy of the hybrid neural network model meets the requirements, and determine that the hyperparameters for training the hybrid neural network model are the preferred hyperparameters.

5. The method for predicting ethylene concentration according to claim 4, characterized in that: In response to the prediction accuracy of the hybrid neural network model not meeting the requirement, the step of determining the hyperparameters of the hybrid neural network model according to the whale optimization algorithm comprises: S1: In response to the accuracy of the hybrid neural network model not meeting the requirement, the number of whale individuals, whale positions and algorithm parameters are initialized, wherein the whale positions represent hyperparameters of the hybrid neural network model; S2: Calculating the fitness of the individual whale according to the whale position of the individual whale to determine the position of the initial optimal individual whale as the optimal solution; S3: updating the algorithm parameters and updating the whale position of the individual whale by using a bubble net to prey, or to surround prey, or to search for prey according to the algorithm parameters; S4: Calculating the fitness of the individual whale according to the updated whale position of the individual whale to determine the position of the optimal individual whale; S5: In response to the position of the optimal whale individual being better than the preferred solution, taking the position of the optimal whale individual as the preferred solution; and S6: Repeat steps S3 to S5 until the termination condition is met, and output the whale position of the preferred solution to determine the hyperparameters of the hybrid neural network model.

6. The method for predicting ethylene concentration according to claim 5, characterized in that: The step S3 comprises: In response to the probability p being greater than or equal to 0.5, updating the whale position of the individual whale by means of bubble net predation; In response to the probability p being less than 0.5 and the absolute value of the coefficient vector A being less than 1, updating the whale position of the individual whale in a way of encircling prey; and In response to the probability p being less than 0.5 and the absolute value of the coefficient vector A being greater than or equal to 1, the whale position of the individual whale is updated in a search and prey manner.

7. The method for predicting ethylene concentration according to claim 4, characterized in that: The error discrimination function includes root mean square error, mean absolute error, mean absolute percentage error and determination coefficient.

8. The method for predicting ethylene concentration according to claim 1, characterized in that: The variables related to the product of the ethylene fractionation tower include ethylene feed amount, ethane feed amount, feed temperature, feed pressure, reflux ratio, reboiler heat duty and tower top condenser heat duty.

9. A prediction system for ethylene concentration, characterized in that: include: a memory having computer instructions stored thereon; and A processor is connected to the memory and configured to execute computer instructions stored in the memory to implement the method for predicting ethylene concentration according to any one of claims 1 to 8.

10. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed by a processor, the method for predicting ethylene concentration according to any one of claims 1 to 8 is implemented.

Images