Alcohol precipitation temperature prediction method and device and electronic equipment

Abstract

The application relates to the technical field of deep learning, and discloses an alcohol precipitation temperature prediction method and device and electronic equipment, the method comprising the following steps: obtaining historical process data of a target substance in an alcohol precipitation process; preprocessing the historical process data to generate standard data; constructing an initial alcohol precipitation temperature prediction model; training the initial alcohol precipitation temperature prediction model multiple times based on the standard data to obtain multiple basic models; based on the prediction error indicators of the multiple basic models, integrating and optimizing the multiple basic models to generate a target alcohol precipitation temperature prediction model, which is used for predicting the temperature of the target substance in the alcohol precipitation process. The application can capture the spatiotemporal correlation of global information and local features, effectively solves the problem that the existing model only focuses on a single time or space scale and cannot capture fine-grained and coarse-grained features at the same time, and effectively improves the accuracy and robustness of the prediction model.

Description

Technical Field

[0001] This invention relates to the field of deep learning technology, specifically to a method, apparatus, and electronic device for predicting alcohol precipitation temperature. Background Technology

[0002] Alcohol precipitation is a common process in the extraction of traditional Chinese medicine, encompassing key fields such as chemical engineering, oil refining, pharmaceuticals, and metallurgy. Its production process is characterized by strong continuity, multi-variable coupling, and dramatic dynamic changes in operating conditions. Temperature, as a core process parameter throughout the entire production process, directly determines reaction efficiency, product quality stability, and energy consumption levels. Therefore, achieving accurate temperature prediction, especially multi-step prediction (i.e., predicting temperature values ​​at multiple future time points), is of irreplaceable value for optimizing production control, reducing energy consumption, and ensuring safe production.

[0003] While numerous explorations have been conducted in related technologies, they all possess certain limitations. Many temperature prediction models suffer from unstable prediction accuracy and susceptibility to data noise. Especially with multivariate time-series data, existing single-model methods often struggle to effectively capture complex spatiotemporal relationships. Therefore, improving the accuracy and stability of alcohol precipitation temperature prediction has become a pressing technical challenge. Summary of the Invention

[0004] This invention provides a method, apparatus, and electronic device for predicting alcohol precipitation temperature, in order to solve the problem of unstable accuracy of single models in predicting alcohol precipitation temperature in the prior art.

[0005] In a first aspect, the present invention provides a method for predicting alcohol precipitation temperature, the method comprising: Obtain historical process data of the target substance during alcohol precipitation; Historical process data is preprocessed to generate standard data; An initial alcohol precipitation temperature prediction model was constructed, which included features for capturing the spatiotemporal correlation between multiple variables and key features. The initial alcohol precipitation temperature prediction model was trained multiple times based on standard data to obtain multiple basic models. Based on the prediction error index of multiple basic models, the multiple basic models are integrated and optimized to generate a target alcohol precipitation temperature prediction model, which is used to predict the temperature of the target substance during the alcohol precipitation process.

[0006] This invention effectively addresses the problem that existing models typically focus only on a single time or spatial scale and cannot simultaneously capture fine-grained and coarse-grained features by comprehensively considering features across multiple time and spatial scales. By constructing an initial alcohol precipitation temperature prediction model and training it multiple times based on standard data, it effectively captures the spatiotemporal correlation between global information and local features, thereby improving the accuracy and robustness of the prediction model. Furthermore, by integrating and optimizing multiple basic models, a final target alcohol precipitation temperature prediction model is obtained. This model not only fully utilizes features at different scales, solving the problem of traditional methods neglecting local features or global information, but also provides more accurate and stable results in temperature prediction during complex alcohol precipitation processes, thus providing more reliable decision support for industrial applications.

[0007] In one optional implementation, the constructed initial alcohol precipitation temperature prediction model includes: The multi-scale global average pooling module is used to perform multi-scale global average pooling on the input standard data to obtain multi-scale feature data. Multiple encoders of different scales are connected to the output of a multi-scale global average pooling layer; each encoder includes a multi-head attention layer, a dropout layer, a normalization layer, a BiLSTM module, and a convolutional layer connected in sequence. The spatial-channel fusion attention module is connected to the output of the encoder one by one. The spatial-channel fusion attention module includes a spatial attention mechanism layer and a channel attention mechanism layer. The spatial attention mechanism layer is used to assign weights to different spatial locations, and the channel attention mechanism layer is used to assign weights to each variable channel.

[0008] In one alternative implementation, the multi-scale global average pooling module includes: three parallel one-dimensional global average pooling layers; The formula for calculating the output features of the one-dimensional global average pooling layer is as follows: ; In the formula, s is the dimension of the pooling size. X is the input time series feature sequence, i is the index of the output feature, and k is the local index within the pooling window. One-dimensional global average pooling layers with different pooling sizes extract features of different granularities, and finally, multi-scale feature data is obtained by splicing.

[0009] In one optional implementation, based on the prediction error indices of multiple basic models, the multiple basic models are integrated and optimized to generate a target alcohol precipitation temperature prediction model, including: Clustering algorithms are used to classify multiple basic models into clusters that correspond one-to-one with the prediction error index. Each cluster obtains a corresponding predictive sub-model through an ensemble learning algorithm; All the prediction sub-models are then integrated again using an ensemble learning algorithm to obtain the target alcohol precipitation temperature prediction model.

[0010] In one alternative implementation, the prediction error metrics include: mean error, variance error, skewness error, and kurtosis error.

[0011] In one alternative implementation, the output of the prediction sub-model is: ; In the formula, Let be the predicted value of the k-th base model within the i-th cluster. Let be the weight of the k-th base model within the i-th cluster, and , Let $\frac{i}{k}$ be the regression error rate of the $k$-th base model within the $i$-th cluster. Let i be the number of basic models within the i-th cluster; The output after integrating all the predictive sub-models again using the ensemble learning algorithm is: ; In the formula, For prediction sub-model The weight, and , To predict the ensemble error rate of the sub-models, satisfying .

[0012] In one optional implementation, historical process data includes: The parameters for the target substance during the alcohol precipitation process include the temperature at the top of the industrial tank, the temperature at the bottom of the industrial tank, the gas pressure inside the tank, the main steam valve, the steam bypass regulating valve, the inlet valves for ethanol and drinking water in the industrial tank, and the reflux liquid temperature in the industrial tank.

[0013] This invention proposes a method for predicting alcohol precipitation temperature. It utilizes multi-scale global average pooling to process sparse Boolean variable data and extract multi-scale information. By combining the multi-scale feature extraction capabilities of the BTHE module and the spatial-channel fusion attention mechanism, it can more comprehensively capture the changing patterns of the production temperature of azurite alcohol precipitation, thus significantly improving prediction accuracy. Furthermore, a two-stage hierarchical AdaBoost ensemble framework enhances the model's adaptability and generalization ability in different scenarios. This not only meets the real-time monitoring requirements of azurite alcohol precipitation production temperature in intelligent factories but also improves process efficiency and product quality uniformity while reducing energy consumption. As an innovative time series prediction model, this invention can integrate multi-scale features, capture long-range dependencies, adapt to data sparsity, and possess strong robustness.

[0014] Secondly, the present invention provides a method for predicting alcohol precipitation temperature, the method comprising: Obtain the current process data of the target substance during the alcohol precipitation process; Preprocess the current process data to generate the current standard data; Input the current standard data into the target alcohol precipitation temperature prediction model established based on the method in any of the above embodiments to obtain the temperature prediction result of the target substance during the alcohol precipitation process.

[0015] This invention effectively improves the accuracy of predicting the alcohol precipitation temperature of substances such as azurite, solves the problem of traditional methods neglecting local features or global information, and can effectively guide industrial production control.

[0016] Thirdly, the present invention provides an alcohol precipitation temperature prediction device, the device comprising: The acquisition module is used to acquire historical process data of the target substance during the alcohol precipitation process. The preprocessing module is used to preprocess historical process data and generate standard data; The building module is used to construct an initial alcohol precipitation temperature prediction model, which includes features for capturing the spatiotemporal correlation between multiple variables and key features. The training module is used to train the initial alcohol precipitation temperature prediction model multiple times based on standard data to obtain multiple basic models. The optimization module is used to integrate and optimize multiple basic models based on the prediction error index of multiple basic models, and generate a target alcohol precipitation temperature prediction model. The target alcohol precipitation temperature prediction model is used to predict the temperature of the target substance during the alcohol precipitation process.

[0017] Fourthly, the present invention provides an electronic device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the alcohol precipitation temperature prediction method of the first aspect or any corresponding embodiment described above.

[0018] Fifthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the alcohol precipitation temperature prediction method of the first aspect or any corresponding embodiment described above.

[0019] In a sixth aspect, the present invention provides a computer program product, including computer instructions for causing a computer to execute the alcohol precipitation temperature prediction method of the first aspect or any corresponding embodiment described above. Attached Figure Description

[0020] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 This is a schematic flowchart of the alcohol precipitation temperature prediction method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the initial alcohol precipitation temperature prediction model according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the UWH-AdaBoost integrated framework structure according to an embodiment of the present invention; Figure 4 This is a schematic diagram comparing the first prediction results according to an embodiment of the present invention; Figure 5 This is a schematic diagram comparing the second prediction results according to an embodiment of the present invention; Figure 6 This is a structural block diagram of an alcohol precipitation temperature prediction device according to an embodiment of the present invention; Figure 7 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0024] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0025] For alcohol precipitation temperature prediction, traditional machine learning methods rely on manually generated feature equations, which are difficult to handle high-dimensional time-series data. Among traditional deep learning models, CNNs (Convolutional Neural Networks) excel at local feature extraction but neglect long-term dependencies; LSTMs (Long Short-Term Memory) and GRUs (Gated Recurrent Units) are inefficient when handling long-sequence predictions; and while Transformers (a model based on self-attention) and their variants achieve parallel processing through self-attention, their performance is limited when handling multi-scale sparse data and multivariate coupling relationships. Existing models often focus on a single temporal or spatial scale, failing to simultaneously capture the correlation between fine-grained and coarse-grained features to capture both global information and local feature relationships.

[0026] According to an embodiment of the present invention, an embodiment of a method for predicting alcohol precipitation temperature is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0027] This embodiment provides a method for predicting alcohol precipitation temperature, which can be used on servers, terminals, and mobile terminals, such as mobile phones and tablets. Figure 1 This is a flowchart of the alcohol precipitation temperature prediction method according to an embodiment of the present invention, as follows: Figure 1 As shown, the process includes the following steps: Step S101: Obtain historical process data of the target substance during the alcohol precipitation process.

[0028] In this embodiment, the target substance is *Aquilaria sinensis*, specifically a mixture of honeysuckle and artemisia annua. *Aquilaria sinensis* precipitation is a crucial step in the production process; however, the precipitation temperature is affected by various factors, such as complex measurement environments, multivariate coupling, and data sparsity. These factors exhibit strong randomness and uncertainty, leading to complex temperature variations in the precipitation process and posing a significant challenge to accurate prediction. This embodiment can collect historical process data related to *Aquilaria sinensis* precipitation over a time period T prior to the prediction time, including precipitation parameters and time-series temperature data during the production process.

[0029] In some optional implementations, historical process data includes parameters of the target substance during the alcohol precipitation process. These parameters include various data such as the temperature at the top and bottom of the industrial tank, the internal gas pressure, the main steam valve, the steam bypass regulating valve, the ethanol and drinking water inlet valves of the industrial tank, and the reflux liquid temperature of the industrial tank. This provides rich samples for subsequent model training, enabling the discovery of patterns and features within the data. All gold-acid alcohol precipitation parameters can be collected and transmitted to the control center via sensors. Various data are acquired in real time through sensor monitoring equipment installed on the gold-acid alcohol precipitation production line. Then, the real-time acquired and preprocessed gold-acid alcohol precipitation data is input into the optimized power prediction model for calculation, resulting in a high-precision prediction of the gold-acid alcohol precipitation production temperature for future periods.

[0030] Step S102: Preprocess the historical process data to generate standard data.

[0031] In this embodiment, preprocessing may include data cleaning and normalization.

[0032] Missing values ​​can be handled using moving window average interpolation, which smoothly fills in the missing values ​​with valid data before and after them. Let the input time series be... middle, If the value is missing, the expression is: ; in, The length of the moving window is determined. If other missing values ​​exist within the window, the window is automatically reduced to the valid data range to ensure interpolation reliability. Statistical methods using 3... The criteria identify outliers in the data and correct them using a moving window average interpolation method, with the same processing method as above.

[0033] The collected data undergoes cleaning. Outliers can be identified using the three-standard-deviation method. For various data points, including industrial tank top temperature, industrial tank bottom temperature, tank internal pressure, main steam angle valve, steam bypass regulating valve, industrial tank ethanol and drinking water inlet valves, and industrial tank reflux liquid temperature, their means and standard deviations are calculated separately. If a data point's difference from the mean is greater than three standard deviations, it is considered an outlier and removed. Missing values ​​are imputed using moving window average interpolation. This method can reasonably estimate missing values ​​based on data trends, ensuring data continuity and integrity. Multivariate data, due to differences in units, can affect the convergence speed of the model. This invention uses min-max standardization to normalize the input data. All features are mapped to the [0,1] interval, as shown in the following expression: ; in, For input data values, , and These represent the normalized value, the minimum value, and the maximum value of the data column, respectively. Normalization is calculated only for the training set to avoid data leakage on the test set. For Boolean variables, the original value remains unchanged after standardization. The min-max normalization method is used to map the cleaned data to the [0,1] interval. This method unifies data of different magnitudes to the same scale range, helping to improve the efficiency and stability of model training and making the model converge more easily.

[0034] Finally, the preprocessed standard data is divided into training set, validation set and test set according to a ratio of 7:2:1.

[0035] Step S103: Construct an initial alcohol precipitation temperature prediction model. The initial alcohol precipitation temperature prediction model includes features for capturing the spatiotemporal correlation between multiple variables and key features.

[0036] In this embodiment, the constructed initial alcohol precipitation temperature prediction model can be configured with average pooling layers of different scales to extract features of different scales, such as fine-grained features and coarse-grained features; it can also be configured with an encoder to capture the spatiotemporal correlation features between multiple variables; and it can also be configured with an attention module to extract key features of different variables.

[0037] Step S104: The initial alcohol precipitation temperature prediction model is trained multiple times based on standard data to obtain multiple basic models.

[0038] In this embodiment, the initial alcohol precipitation temperature prediction model can be repeatedly trained using standard data, resulting in a base model after each training iteration. This can be achieved by augmenting the standard dataset (e.g., perturbing, flipping, or randomly pruning time-series data) or by training the initial alcohol precipitation temperature prediction model based on different subsets of data.

[0039] Step S105: Based on the prediction error index of multiple basic models, integrate and optimize multiple basic models to generate a target alcohol precipitation temperature prediction model. The target alcohol precipitation temperature prediction model is used to predict the temperature of the target substance during the alcohol precipitation process.

[0040] Specifically, gradient descent-based ensemble optimization can be used to dynamically adjust the weights of multiple base models in the ensemble, minimizing the prediction error of the final target model. In this embodiment, the UWH-AdaBoost ensemble framework is preferred. This framework integrates and outputs the optimal model, which then outputs the final prediction result for the gold-acid alcohol precipitation production temperature. The UWH-AdaBoost ensemble framework will be described in detail in the next embodiment.

[0041] This invention effectively addresses the problem that existing models typically focus only on a single time or spatial scale and cannot simultaneously capture fine-grained and coarse-grained features by comprehensively considering features across multiple time and spatial scales. By constructing an initial alcohol precipitation temperature prediction model and training it multiple times based on standard data, it effectively captures the spatiotemporal correlation between global information and local features, thereby improving the accuracy and robustness of the prediction model. Furthermore, by integrating and optimizing multiple basic models, a final target alcohol precipitation temperature prediction model is obtained. This model not only fully utilizes features at different scales, solving the problem of traditional methods neglecting local features or global information, but also provides more accurate and stable results in temperature prediction during complex alcohol precipitation processes, thus providing more reliable decision support for industrial applications.

[0042] This embodiment also provides a method for predicting alcohol precipitation temperature, which can be used on servers, terminals, and mobile terminals, such as mobile phones and tablets. The process includes the following steps: Step S201: Obtain historical process data of the target substance during the alcohol precipitation process. For details, please refer to [link to relevant documentation]. Figure 1 Step S101 of the illustrated embodiment will not be described again here.

[0043] Step S202 involves preprocessing historical process data to generate standard data. For details, please refer to [link to relevant documentation]. Figure 1 Step S102 of the illustrated embodiment will not be described again here.

[0044] Step S203: Construct an initial alcohol precipitation temperature prediction model. The initial alcohol precipitation temperature prediction model includes features for capturing the spatiotemporal correlation between multiple variables and key features.

[0045] Specifically, refer to Figure 2 As shown, the constructed initial alcohol precipitation temperature prediction model includes: a multi-scale global average pooling module, multiple encoders of different scales, a spatial-channel fusion attention module, and a feedforward neural network layer. A feature extraction mechanism is used to model the interrelationships of time points in the industrial dataset.

[0046] The multi-scale global average pooling module is used to perform multi-scale global average pooling on the input standard data to obtain multi-scale feature data. In some optional implementations, the multi-scale global average pooling module includes three parallel one-dimensional global average pooling layers. In this embodiment, the pooling sizes are 1×1, 2×2, and 3×3, and the stride is consistent with the pooling size. For pooling sizes of... The formula for calculating the output features of a one-dimensional global average pooling layer is as follows: ; In the formula, s is the dimension of the pooling size. X is the input time series feature sequence, i is the index of the output feature, and k is the local index within the pooling window.

[0047] One-dimensional global average pooling layers with different pooling sizes extract features of different granularities, i.e., they extract fine-grained features. and coarse grain The features are ultimately combined to obtain the fused multi-scale feature data through a concatenation operation. .

[0048] In this embodiment, by setting three one-dimensional global average pooling operations with different pooling sizes in parallel, fine-grained features and coarse-grained features are extracted simultaneously from the sparse standard dataset obtained after preprocessing, thereby achieving collaborative extraction of multi-scale features. Finally, the outputs of the three branches are zero-filled, concatenated, and linearly transformed to fuse them into a multi-size feature matrix, which effectively solves the sparsity problem of alcohol precipitation data of target substances (such as azurite).

[0049] The encoder is connected to the output of a multi-scale global average pooling layer; each encoder includes a multi-head attention layer, a dropout layer, a normalization layer, a BiLSTM module, and a convolutional layer connected in sequence.

[0050] The encoder in this embodiment is a hybrid encoder, employing the BTHE (BiLSTM-Transformer-Hybrid-Encoder) module, and positional encoding is removed to improve computational efficiency. The BTHE module is responsible for processing the input data through the Transformer multi-head attention layer before feeding it into the BiLSTM module, thereby enhancing the ability to capture key spatiotemporal features.

[0051] The input feature matrix of the multi-head attention layer is ,in, For time step, The number of channels; a query matrix is ​​generated through linear transformation. Key matrix Sum matrix ,in , For the number of attention heads.

[0052] The output of single-head attention is: ; Multi-head attention is obtained by concatenating the outputs of all heads and performing a linear transformation: ; Multi-head attention output via residual connection and After normalization, the data is fed into the BiLSTM module: ; in, This is a random deactivation operation to prevent overfitting.

[0053] The BiLSTM module includes forward LSTM and backward LSTM to process the input sequence. The forward and reverse hidden states are: ; The output of the BiLSTM module is obtained by concatenating the bidirectional hidden states: ; Finally, local features are extracted and the dimensions are adjusted using a one-dimensional convolutional layer. ,in The kernel size is [size]. The number of output channels is used as the dimension, and the final output should be consistent with the number of input channels to facilitate processing by subsequent modules.

[0054] The spatial-channel fusion attention module is connected to the output of the encoder in a one-to-one correspondence. The spatial-channel fusion attention module includes a spatial attention mechanism layer and a channel attention mechanism layer. The spatial attention mechanism layer is used to assign weights to different spatial locations, and the channel attention mechanism layer is used to assign weights to each variable channel.

[0055] Channel attention assigns different weights to each variable channel, thus highlighting the more important channel features; spatial attention assigns weights to specific spatial locations to highlight the spatial regions that contribute the most to the task. The spatial-channel fusion attention module outputs weighted spatial and channel branches to obtain attention-enhanced features that combine global and local characteristics; the feedforward neural network layer further performs nonlinear transformations and integration on the extracted features. The channel-spatial fusion attention module can enhance key features through weighted enhancement.

[0056] Channel attention mechanism uses one-dimensional convolution, temporal compression, and Normalization is performed to generate channel weights. The calculation process is as follows: ; Wherein, the input feature matrix , For time step, For the number of channels, the input features Temporal dimension compression (global average pooling) is performed to obtain After one-dimensional convolution and Activate and generate channel weights .

[0057] The spatial attention mechanism performs one-dimensional convolution and normalization on the input features to generate a spatial weight matrix. The specific calculation process is as follows: ; Wherein, the input feature matrix , For time step, Given the number of channels, a spatial feature matrix is ​​obtained through a one-dimensional convolutional layer. After Normalization and Activation Generate Spatial Weights .

[0058] The channel-weighted features and spatial-weighted features can then be expressed as: ; in, This indicates element-wise multiplication.

[0059] By fusing the weighted features of the two branches, the enhanced feature is output. :and ; in, The fusion coefficient is a learnable coefficient.

[0060] Step S204 involves training the initial alcohol precipitation temperature prediction model multiple times based on standard data to obtain several basic models. For details, please refer to [link to relevant documentation]. Figure 1 Step S103 of the illustrated embodiment will not be described again here.

[0061] Step S205: Based on the prediction error index of multiple basic models, integrate and optimize multiple basic models to generate a target alcohol precipitation temperature prediction model. The target alcohol precipitation temperature prediction model is used to predict the temperature of the target substance during the alcohol precipitation process.

[0062] Specifically, refer to Figure 3 The diagram shows the UWH-AdaBoost integration framework structure. The UWH-AdaBoost integration framework is used for integration optimization in this embodiment. Step S205 includes: Step a1: Using a clustering algorithm, multiple basic models are classified into clusters that correspond one-to-one with the prediction error index.

[0063] In some optional implementations, prediction error metrics include: mean error, variance error, skewness error, and kurtosis error. That is, the first-stage ensemble employs a clustering algorithm. The basic model was clustered into 4 clusters based on the prediction error index.

[0064] In step a2, each cluster obtains a corresponding predictive sub-model through an ensemble learning algorithm.

[0065] Each cluster is ensembled using the AdaBoost ensemble learning algorithm to obtain a sub-model. For the first... Clusters ( The output of the prediction sub-model is: ; in, Let be the predicted value of the k-th base model within the i-th cluster. Let be the weight of the k-th base model within the i-th cluster, and , Let $\frac{i}{k}$ be the regression error rate of the $k$-th base model within the $i$-th cluster. Let i be the number of basic models within the i-th cluster; The output after integrating all the predictive sub-models again using the ensemble learning algorithm is: ; in, For prediction sub-model The weight, and , To predict the ensemble error rate of the sub-models, satisfying .

[0066] Step a3: Integrate all the prediction sub-models again using an ensemble learning algorithm to obtain the target alcohol precipitation temperature prediction model.

[0067] In other words, the second-stage ensemble applies unweighted AdaBoost ensemble to the four prediction sub-models again, outputting the final predicted value. : ; in, For prediction sub-model The weight, and , To predict the ensemble error rate of the sub-models, satisfying .

[0068] In this embodiment, data related to the gold-aluminol precipitation process are first collected. The data is preprocessed using multi-scale global average pooling. A temperature prediction model for gold-aluminol precipitation is then constructed, and optimized using a two-stage hierarchical AdaBoost ensemble to enhance the model's generalization ability, ultimately yielding an optimized prediction model. In practical applications, various parameters from the production process are input into the prediction model to achieve the final prediction of the production temperature.

[0069] Because noise in the production data of substances such as azurite and chlorite precipitates can easily lead to model overfitting of local features, hierarchical AdaBoost uses a two-stage ensemble to offset the overfitting of individual models while improving the model's generalization ability. However, due to the large dataset size, traditional AdaBoost strategies assign relatively small weights to each data point, resulting in minimal improvement when training with weak learners. Therefore, this invention proposes the UWH-AdaBoost ensemble strategy, which evaluates N models using four prediction error metrics: mean error, variance error, skewness error, and kurtosis error. In the first stage, the trained base models are clustered into four clusters using constrained K-means based on the prediction error metrics. Models within each cluster share similarity in specific error patterns. Furthermore, unweighted AdaBoost ensembles the models within the same cluster to specifically weaken common errors within that cluster, generating more stable prediction sub-models G1-G4. In the second stage, AdaBoost ensemble is applied again to the four prediction sub-models, combining the advantages of different clusters to further offset the biases of individual sub-models, ultimately generating the optimal prediction model. The prediction model provided by this invention can integrate multi-scale features, capture long-range dependencies, adapt to data sparsity, and has strong robustness, and has broad application prospects.

[0070] Temperature prediction results reference Figure 4 As shown, this is a schematic diagram comparing the predicted results of the gold cyanide precipitation temperature prediction model with the actual temperature; the predicted temperature curve matches the fluctuation trend of the actual temperature curve, the predicted temperature can reflect the actual temperature, and the prediction effect is good.

[0071] To verify the effectiveness of the method proposed in this invention, the PatchTST model was used for prediction on the same dataset, and the temperature prediction results are as follows: Figure 5 The diagram shown is a comparison between the predicted results of the PatchTST model and the actual temperature. The predicted power curve of the model is similar to the actual temperature change trend, but there is also a significant deviation. For a detailed introduction to the prediction model based on PatchTST, please refer to the literature "A Time Series is Worth 64 Words: Long-term Forecasting with Transformers".

[0072] To measure the accuracy of model predictions, this invention uses the mean error (MAE) and mean squared error (MSE) to detect the model's predictive ability. The calculation formulas are as follows: ; ; in, For the sample size, For the true value, These are predicted values.

[0073] The proposed method yields a mean squared error (MAE) of 0.0971 and a mean squared error (MSE) of 0.1563. In contrast, the test results using the PatchTST model show a MAE of 0.1697 and an MSE of 0.5342. This demonstrates that the feature extraction and attention enhancement modules in the proposed model effectively filter noise from the input data and capture its periodicity and trends. To further improve the model's generalization ability and prediction accuracy, a two-stage distributed AdaBoost ensemble is used to optimize the model, making its predictions more realistic. Comparative analysis shows that the model's prediction capabilities meet expectations.

[0074] Furthermore, the performance of the temperature prediction model can be tested using mean absolute error (MAE) and mean squared error (MSE). Based on the test results, the model's parameter settings and structure can be optimized accordingly. Parameter settings may include: input sequence length, output sequence length, learning rate, optimizer selection, patience value, dropout ratio, and time feature encoding type, etc.

[0075] This invention proposes a method for predicting the temperature of ethanol precipitation. It utilizes multi-scale global average pooling to process sparse Boolean variable data and extract multi-scale information. By combining the multi-scale feature extraction capabilities of the BTHE module and the spatial-channel fusion attention mechanism, it can more comprehensively capture the changing patterns of the temperature during the production of ethanol precipitation, thus significantly improving prediction accuracy. Furthermore, a two-stage hierarchical AdaBoost ensemble framework enhances the model's adaptability and generalization ability in different scenarios. This not only meets the real-time monitoring requirements of intelligent factories for the production temperature of ethanol precipitation but also improves process efficiency and product quality uniformity while reducing energy consumption. As an innovative time series prediction model, this invention integrates multi-scale features, captures long-range dependencies, adapts to data sparsity, and possesses strong robustness. This model uses multivariate industrial process datasets for multi-step temperature prediction, providing enterprises with reasonable industrial adjustments and representing a new trend in the intelligent upgrading of ethanol precipitation.

[0076] This embodiment also provides a method for predicting alcohol precipitation temperature, which can be used on servers, terminals, and mobile terminals, such as mobile phones and tablets. The process includes the following steps: Step S301: Obtain the current process data of the target substance in the alcohol precipitation process; Step S302: Preprocess the current process data to generate the current standard data; Step S303: Input the current standard data into the target alcohol precipitation temperature prediction model established based on the method of any of the above embodiments to obtain the temperature prediction result of the target substance during the alcohol precipitation process.

[0077] In this embodiment, azurite is used as the target substance. Specifically, multivariate time-series data from the azurite ethanol precipitation process are collected and preprocessed. Three parallel one-dimensional global average pooling layers (pooling sizes of 1×1, 2×2, and 3×3) are used to extract multi-scale features to address the data sparsity problem. A hybrid encoder is used to enhance the ability to capture spatiotemporal correlation features. A channel-space fusion attention mechanism is designed, which strengthens and highlights key features through weighted fusion of spatial weight matrices and channel weight matrices. Finally, the N trained sub-models are integrated through a two-stage clustering strategy framework using UWH-AdaBoost to output multi-step temperature prediction results. This invention effectively improves the accuracy of ethanol precipitation temperature prediction for substances such as azurite, solves the problem of traditional methods neglecting local features or global information, and can effectively guide industrial production control.

[0078] This embodiment also provides an alcohol precipitation temperature prediction device, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0079] This embodiment provides an alcohol precipitation temperature prediction device, such as... Figure 6 As shown, it includes: The acquisition module 601 is used to acquire historical process data of the target substance during the alcohol precipitation process; the parameters of the target substance during the alcohol precipitation process include the temperature of the industrial tank, the temperature of the industrial tank, the gas pressure inside the tank, the steam angle valve of the main pipe, the steam bypass regulating valve, the ethanol and drinking water inlet valves of the industrial tank, and the reflux liquid temperature of the industrial tank.

[0080] Preprocessing module 602 is used to preprocess historical process data to generate standard data; Module 603 is used to construct an initial alcohol precipitation temperature prediction model, which includes features for capturing spatiotemporal correlations and key features among multiple variables; a multi-scale global average pooling module, used to perform multi-scale global average pooling on the input standard data to obtain multi-scale feature data; multiple encoders of different scales, connected to the output of the multi-scale global average pooling layer; each encoder includes a multi-head attention layer, a dropout layer, a normalization layer, a BiLSTM module, and a convolutional layer connected in sequence; a spatial-channel fusion attention module, connected one-to-one with the output of the encoder; the spatial-channel fusion attention module includes a spatial attention mechanism layer and a channel attention mechanism layer, the spatial attention mechanism layer is used to assign weights to different spatial locations, and the channel attention mechanism layer is used to assign weights to each variable channel; the multi-scale global average pooling module includes three parallel one-dimensional global average pooling layers; the calculation formula for the output features of the one-dimensional global average pooling layer is: In the formula, s is the dimension of the pooling size. X is the input temporal feature sequence, i is the index of the output feature, and k is the local index within the pooling window. One-dimensional global average pooling layers of different pooling sizes extract features of different granularities, and finally, the fused multi-scale feature data is obtained through a concatenation operation.

[0081] Training module 604 is used to train the initial alcohol precipitation temperature prediction model multiple times based on standard data to obtain multiple basic models. The optimization module 605 is used to integrate and optimize multiple basic models based on the prediction error index of multiple basic models to generate a target alcohol precipitation temperature prediction model. The target alcohol precipitation temperature prediction model is used to predict the temperature of the target substance during the alcohol precipitation process.

[0082] In some alternative implementations, the optimization module 605 is specifically used for: Clustering algorithms are used to classify multiple basic models into clusters that correspond one-to-one with the prediction error indicators. The prediction error indicators include: mean error, variance error, skewness error, and kurtosis error.

[0083] Each cluster obtains a corresponding predictive sub-model through an ensemble learning algorithm; All the prediction sub-models are then integrated again using an ensemble learning algorithm to obtain the target alcohol precipitation temperature prediction model.

[0084] The output of the prediction sub-model is as follows: ; In the formula, Let be the predicted value of the k-th base model within the i-th cluster. Let be the weight of the k-th base model within the i-th cluster, and , Let $\frac{i}{k}$ be the regression error rate of the $k$-th base model within the $i$-th cluster. Let i be the number of basic models within the i-th cluster; The output after integrating all the predictive sub-models again using the ensemble learning algorithm is: ; In the formula, For prediction sub-model The weight, and , To predict the ensemble error rate of the sub-models, satisfying .

[0085] The alcohol precipitation temperature prediction device provided in this embodiment of the invention can execute the alcohol precipitation temperature prediction method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the various modules and units described above are the same as in the corresponding embodiments described above, and will not be repeated here.

[0086] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0087] The following is a detailed reference. Figure 7 This diagram illustrates a suitable structural schematic for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 701, which can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) 702 or a program loaded from memory 708 into random access memory (RAM) 703. The RAM 703 also stores various programs and data required for the operation of the electronic device. The processor 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0088] Typically, the following devices can be connected to I / O interface 705: input devices 706 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 707 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 708 including, for example, magnetic tapes, hard disks, etc.; and communication devices 709. Communication device 709 allows electronic devices to exchange data via wireless or wired communication with other devices. Although Figure 7 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0089] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 709, or installed from a memory 708, or installed from a ROM 702. When the computer program is executed by the processor 701, it performs the functions defined in the alcohol precipitation temperature prediction method of the embodiments of the present invention.

[0090] Figure 7 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0091] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the alcohol precipitation temperature prediction method shown in the above embodiments is implemented.

[0092] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0093] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for predicting alcohol precipitation temperature, characterized in that, The method includes: Obtain historical process data of the target substance during alcohol precipitation; The historical process data is preprocessed to generate standard data; An initial alcohol precipitation temperature prediction model is constructed, which includes features for capturing the spatiotemporal correlation between multiple variables and key features. The initial alcohol precipitation temperature prediction model was trained multiple times based on the standard data to obtain multiple basic models. Based on the prediction error index of the multiple basic models, the multiple basic models are integrated and optimized to generate a target alcohol precipitation temperature prediction model, which is used to predict the temperature of the target substance during the alcohol precipitation process.

2. The method according to claim 1, characterized in that, The constructed initial alcohol precipitation temperature prediction model includes: The multi-scale global average pooling module is used to perform multi-scale global average pooling on the input standard data to obtain multi-scale feature data. Multiple encoders of different scales are connected to the output of the multi-scale global average pooling layer; each encoder includes a multi-head attention layer, a dropout layer, a normalization layer, a BiLSTM module, and a convolutional layer connected in sequence. The spatial-channel fusion attention module is connected to the output of the encoder in a one-to-one correspondence. The spatial-channel fusion attention module includes a spatial attention mechanism layer and a channel attention mechanism layer. The spatial attention mechanism layer is used to assign weights to different spatial locations, and the channel attention mechanism layer is used to assign weights to each variable channel.

3. The method according to claim 2, characterized in that, The multi-scale global average pooling module includes: three parallel one-dimensional global average pooling layers; The formula for calculating the output features of the one-dimensional global average pooling layer is as follows: ; In the formula, s is the dimension of the pooling size. X is the input time series feature sequence, i is the index of the output feature, and k is the local index within the pooling window. The one-dimensional global average pooling layers with different pooling sizes extract features of different granularities, and finally obtain the fused multi-scale feature data through a splicing operation.

4. The method according to claim 1, characterized in that, The prediction error index based on the multiple basic models is used to integrate and optimize the multiple basic models to generate a target alcohol precipitation temperature prediction model, including: A clustering algorithm is used to classify the multiple basic models into clusters that correspond one-to-one with the prediction error index; Each cluster obtains a corresponding predictive sub-model through an ensemble learning algorithm; All the predicted sub-models are then integrated again using the ensemble learning algorithm to obtain the target alcohol precipitation temperature prediction model.

5. The method according to claim 4, characterized in that, The prediction error indicators include: mean error, variance error, skewness error, and kurtosis error.

6. The method according to claim 4, characterized in that, The output of the prediction sub-model is: ； In the formula, Let be the predicted value of the k-th base model within the i-th cluster. Let be the weight of the k-th basic model within the i-th cluster, and , Let $\frac{i}{k}$ be the regression error rate of the $k$-th base model within the $i$-th cluster. The number of the basic models within the i-th cluster; The output of integrating all the aforementioned prediction sub-models again using the ensemble learning algorithm is: ； In the formula, For the prediction sub-model The weight, and , The ensemble error rate of the prediction sub-model satisfies .

7. The method according to claim 1, characterized in that, The historical process data includes: The parameters of the target substance in the alcohol precipitation process include the temperature at the top of the industrial tank, the temperature at the bottom of the industrial tank, the gas pressure inside the tank, the main steam angle valve, the steam bypass regulating valve, the ethanol and drinking water inlet valves of the industrial tank, and the reflux liquid temperature of the industrial tank.

8. A method for predicting alcohol precipitation temperature, characterized in that, The method includes: Obtain the current process data of the target substance during the alcohol precipitation process; The current process data is preprocessed to generate current standard data; The current standard data is input into the target alcohol precipitation temperature prediction model established based on any one of claims 1-7 to obtain the temperature prediction result of the target substance during the alcohol precipitation process.

9. A device for predicting alcohol precipitation temperature, characterized in that, The device includes: The acquisition module is used to acquire historical process data of the target substance during the alcohol precipitation process. The preprocessing module is used to preprocess the historical process data to generate standard data; A construction module is used to build an initial alcohol precipitation temperature prediction model, which includes features for capturing the spatiotemporal correlation between multiple variables and key features. The training module is used to train the initial alcohol precipitation temperature prediction model multiple times based on the standard data to obtain multiple basic models; An optimization module is used to integrate and optimize the multiple basic models based on the prediction error index of the multiple basic models to generate a target alcohol precipitation temperature prediction model, which is used to predict the temperature of the target substance during the alcohol precipitation process.

10. An electronic device, characterized in that, include: A memory and a processor are communicatively connected, the memory storing computer instructions, and the processor executing the computer instructions to perform the alcohol precipitation temperature prediction method according to any one of claims 1 to 7.

Images