Convolutional neural network-based membrane distillation system performance prediction method and system

Abstract

The invention discloses a membrane distillation system performance prediction method and system based on a convolutional neural network, and belongs to the field of deep learning. The method comprises the following steps: collecting experimental data of the operation performance of the membrane distillation system; standardizing the collected data and then constructing a training data set sample and a test data set sample; constructing a convolutional neural network model, training the convolutional neural network model through a training data set sample to obtain an optimal model, and fixing model parameters of the optimal model; importing a test data set sample into the optimal model, and predicting the operation performance of the membrane distillation system; and evaluating the prediction result. According to the method, the multi-dimensional features of the system operation data are extracted and learned through the convolutional neural network, so that the performance of the membrane distillation system is efficiently and accurately predicted.

Description

Technical Field

[0001] The present invention belongs to the field of deep learning. Specifically, the present invention relates to a method and system for predicting the performance of a membrane distillation system based on a convolutional neural network. Background Art

[0002] Membrane distillation system is a technology that achieves liquid separation through vapor transmission through hydrophobic porous membranes. It is widely used in fields such as seawater desalination, wastewater treatment and resource recovery. Its performance usually depends on multiple operating parameters, such as temperature, pressure, flow rate and membrane fouling. Traditionally, the performance prediction of membrane distillation system relies on physical modeling or empirical formulas, but these methods often lead to inaccurate prediction results due to model simplification and complexity of actual operating conditions.

[0003] With the development of deep learning technology, especially convolutional neural network (CNN) and other technologies, it has shown strong ability in dealing with nonlinear problems and multidimensional complex data. At present, although some studies have used machine learning methods in the modeling of membrane distillation systems, the use of CNN for accurate prediction of system performance has not been widely used.

[0004] Existing membrane distillation system performance prediction is mainly based on physical modeling and statistical methods. These methods have the following defects:

[0005] 1. Nonlinear problems are difficult to handle accurately: The operating performance of the membrane distillation system is affected by a variety of complex nonlinear factors, and traditional mathematical models are difficult to effectively capture these complex interactions.

[0006] 2. Poor model robustness: When dealing with different working conditions and membrane materials, the generalization ability of traditional models is poor and it is difficult to adapt to changing working conditions.

[0007] 3. The calculation is complex and time-consuming: Physical modeling methods often require a large amount of experimental data and complex calculations, which restricts the real-time and accuracy of the predictions.

[0008] 4. Lack of explainability: Although existing deep learning models can improve prediction accuracy, they often lack the explainability of prediction results, which is a certain obstacle to the actual promotion of industrial applications. Summary of the invention

[0009] The present invention aims to overcome the shortcomings of the prior art and proposes a membrane distillation system performance prediction method and system based on convolutional neural network to achieve the following objectives: extracting and learning the multidimensional features of system operation data through convolutional neural network, and then efficiently and accurately predicting the performance of the membrane distillation system.

[0010] In order to achieve the above object, the technical solution adopted by the present invention is: a method for predicting the performance of a membrane distillation system based on a convolutional neural network, the method comprising:

[0011] Step S1, collecting experimental data of membrane distillation system operating performance;

[0012] Step S2: standardize the collected data to construct training data set samples and test data set samples;

[0013] Step S3, constructing a convolutional neural network model, training the convolutional neural network model with training data set samples to obtain an optimal model and fix its model parameters;

[0014] Step S4, importing the test data set sample into the optimal model of step S3 to predict the operating performance of the membrane distillation system;

[0015] Step S5: Evaluate the prediction results.

[0016] Preferably, in step S1, the conductive membrane in the membrane distillation system is subjected to a power-on experiment under different parameter conditions; in the power-on experiment under each parameter condition, the power-on process is repeated and monitored in real time by an electronic balance and a conductivity meter connected to a computer, and the flux and conductivity of the conductive membrane during the experiment are recorded at preset time intervals, thereby collecting experimental data on the operating performance of the membrane distillation system.

[0017] Preferably, the step S2 comprises:

[0018] Step S21, screening the collected experimental data to remove experimental data under parameter conditions that are invalid for the operating performance of the membrane distillation system;

[0019] Step S22, normalizing the screened data;

[0020] Step S23: Divide the normalized data into training data set samples and test data set samples according to a preset ratio.

[0021] Preferably, in step S21, a correlation analysis method is used for data screening, that is, the correlation coefficient between each parameter condition and the target variable is calculated, wherein if the correlation coefficient value is less than a preset correlation coefficient threshold, the corresponding parameter condition is eliminated; the target variable includes the flux and conductivity of the conductive film.

[0022] Preferably, in step S22, the normalization adopts maximum-minimum normalization, that is:

[0023]

[0024] Among them, X normrepresents the normalized value; X represents the characteristic value of an experimental data after screening; X max , X min They represent the maximum and minimum values ​​of the feature in the entire dataset, respectively.

[0025] Preferably, step S3 comprises:

[0026] Step S31, constructing a convolutional neural network model; the convolutional neural network consists of a convolutional layer, a pooling layer, a fully connected layer and an output layer; wherein the convolutional layer is used to extract local features in the input data, and the pooling layer is used to downsample the feature map output by the convolutional layer; the fully connected layer is used to flatten the feature map extracted by the convolution and pooling layers and then output the prediction result through the output layer;

[0027] Step S32, determining a loss function;

[0028] Step S33, after randomly initializing the model parameters, use the training data set samples as the input of the model, use the gradient descent algorithm to train the model iteratively, and when the loss function value no longer decreases within a preset period, retain the model and fix its parameters as the optimal model parameters.

[0029] Preferably, in step S32, the loss function is the mean square error, that is:

[0030]

[0031] Among them, MSE represents mean square error; represents the predicted value of the i-th sample; y i represents the actual value of the i-th sample; n represents the total number of samples.

[0032] Preferably, step S5 comprises:

[0033] Step S51, evaluating the difference between the model predicted value and the true value in step S4 by using the mean square error;

[0034] Step S52: Determine the coefficient R 2 Assess the explanatory power of the optimal model;

[0035] Step S53: Visually compare the model prediction value in step S4 with the true value, and the visualization includes drawing line graphs of the model prediction value and the true value, and drawing a residual graph between the model prediction value and the true value.

[0036] At the same time, the present application also proposes a membrane distillation system performance prediction system based on convolutional neural network, including a computer, wherein the computer is used to execute a computer program constructed according to the above-mentioned membrane distillation system performance prediction method based on convolutional neural network.

[0037] The technical effects of the present invention are:

[0038] (1) Through the introduction of convolutional neural networks, the model can capture the complex nonlinear relationship of multi-dimensional input parameters in the membrane distillation system, thereby significantly improving the prediction accuracy of flux and conductivity.

[0039] (2) The convolutional neural network model of the present invention has strong robustness through training and optimization of a large amount of data and can adapt to the prediction needs under different working conditions.

[0040] (3) The convolutional neural network model has high computational efficiency and can predict system performance in real time in industrial production, helping to adjust operating parameters in a timely manner to optimize system efficiency.

[0041] (4) This method is applicable to various types of membrane distillation systems and can be extended to the performance prediction and optimization of other membrane separation processes. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a flow chart of a method for predicting membrane distillation system performance based on a convolutional neural network according to an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The following is a further detailed description of the specific implementation of the present invention by describing the embodiments with reference to the accompanying drawings, with the purpose of helping those skilled in the art to have a more complete, accurate and in-depth understanding of the inventive concept and technical solution of the present invention and to facilitate its implementation. In order to make the technical solution of the present invention clearer, the present invention is explained by the following embodiments.

[0044] This embodiment proposes a membrane distillation system performance prediction method based on a convolutional neural network, which aims to extract and learn the multi-dimensional features of system operation data through a convolutional neural network, and then efficiently and accurately predict the performance of the membrane distillation system. Figure 1 As shown, the method comprises the following steps:

[0045] Step S1, collecting experimental data of membrane distillation system operating performance;

[0046] Step S2: standardize the collected data to construct training data set samples and test data set samples;

[0047] Step S3, constructing a convolutional neural network model, training the convolutional neural network model with training data set samples to obtain an optimal model and fix its model parameters;

[0048] Step S4, importing the test data set sample into the optimal model of step S3 to predict the operating performance of the membrane distillation system;

[0049] Step S5: Evaluate the prediction results.

[0050] Specifically, in step S1 of the present embodiment, the conductive membrane in the membrane distillation system is subjected to an energization experiment under different parameter conditions. Common parameter conditions include temperature, pressure, flow rate, humidity, inlet water pH value and membrane contamination degree. In the energization experiment under each parameter condition, the voltage setting is increased from 0V to 3V, the energization process is repeated and monitored in real time by an electronic balance and a conductivity meter connected to a computer, and the flux and conductivity of the conductive membrane during the experiment are recorded at preset time intervals, wherein the flux of the mixture passing through the conductive membrane after distillation can be measured by the amount of liquid falling on the electronic balance, and the conductivity is directly measured using a conductivity meter, and finally the experimental data of the operating performance of the membrane distillation system are obtained after collection and aggregation.

[0051] In step S2 of this embodiment, standardization includes data screening, normalization, etc. Specifically, step S2 includes:

[0052] Step S21, screen the collected experimental data, and remove the experimental data under the parameter conditions that are invalid for the operating performance of the membrane distillation system; in this embodiment, the data screening is performed using the correlation analysis method, that is, calculating the correlation coefficient between each parameter condition and the target variable (including the flux and conductivity of the conductive membrane, etc.). In this embodiment, the expression of the correlation coefficient is:

[0053]

[0054] Where X represents the parameter condition; Y represents the target variable; Cov(X,Y) represents the covariance of X and Y; Var[] represents the variance. The larger the absolute value of the correlation coefficient, the higher the correlation between the parameter condition and the target variable. Therefore, the user can preset a correlation coefficient threshold according to actual needs. When the correlation coefficient value is less than the preset correlation coefficient threshold, the corresponding parameter condition (such as ambient temperature, ambient operating humidity, inlet water pH value, etc.) is eliminated.

[0055] Step S22: normalize the filtered data. The normalization in this embodiment adopts maximum and minimum normalization, that is:

[0056]

[0057] Among them, X norm represents the normalized value; X represents the characteristic value of an experimental data after screening; X max , X minRespectively represent the maximum and minimum values ​​of the feature in the entire data set. This method scales the feature value to a specified range, usually [0,1], thereby eliminating the impact of different feature scale differences. In specific implementation, other normalization methods can also be used as required.

[0058] Step S23, divide the normalized data into training data set samples and test data set samples according to a preset ratio. In this embodiment, the preset ratio is to use 80% of the entire data set as training data to train the model, and the remaining 20% ​​as test data to evaluate model performance.

[0059] Step S3 of this embodiment includes:

[0060] Step S31, construct a convolutional neural network model; the convolutional neural network consists of a convolutional layer, a pooling layer, a fully connected layer and an output layer; wherein the convolutional layer is used to extract local features in the input data, and the pooling layer is used to downsample the feature map output by the convolutional layer; the fully connected layer is used to flatten the feature map extracted by the convolution and pooling layers and output the prediction result through the output layer; in this embodiment, the flux and conductivity of the membrane are used as predicted performance indicators.

[0061] Convolution operation is the core of convolutional neural network. The calculation formula for the convolution operation of input data is:

[0062] Y(i,j)=∑ M-1 ∑ N-1 X(i+m,j+n)*K(m,n);

[0063] Among them, Y(i,j) is the value of the output feature map obtained after the convolution operation; X(i+m,j+n) is the value of position (i,j) in the input data, m and n are the sizes of the convolution kernel, K(m,n) is the value of the convolution kernel; M and N are the sizes of the convolution kernel.

[0064] Step S32, determine the loss function; in this embodiment, the loss function uses the mean square error, that is:

[0065]

[0066] Among them, MSE represents mean square error; represents the predicted value of the i-th sample; y i represents the actual value of the ith sample; n represents the total number of samples. The calculation of the mean square error is relatively simple and can be implemented quickly, especially on large data sets, with high efficiency. Since the error is squared, the mean square error gives a higher penalty to larger errors, which encourages the model to minimize the occurrence of larger errors during model training, thereby improving the model's predictive performance.

[0067] Step S33, after randomly initializing the model parameters, use the training data set samples as the input of the model, use the gradient descent algorithm as the back propagation algorithm to train the model iteratively, and when the loss function value no longer decreases within a preset period (this embodiment is set to 10 periods, which can be flexibly selected according to actual conditions), retain the model and fix its parameters as the optimal model parameters.

[0068] The convolutional neural network model of the present invention has strong robustness through training and optimization of a large amount of data, and can adapt to the prediction requirements under different working conditions. The convolutional neural network model has high computational efficiency and can predict system performance in real time in industrial production, helping to adjust operating parameters in time to optimize system efficiency.

[0069] In step S5 of this embodiment, the flux and conductivity of the membrane predicted by the model are compared with the actual values ​​in the measurement data set sample to evaluate the model performance, which specifically includes the following steps:

[0070] Step S51, evaluate the difference between the model prediction value and the true value in step S4 by using the mean square error. In this embodiment, by continuing to use the loss function in the training process as a part of the model prediction performance evaluation, the reliability of the model prediction performance evaluation result can be guaranteed.

[0071] S52, through the coefficient of determination R 2 Evaluate the explanatory power of the optimal model. Based on step S51, this embodiment proposes a determination coefficient R 2 As an evaluation criterion to further evaluate the prediction performance of the model, the setting of multiple evaluation criteria can further ensure the reliability of the evaluation results. 2 The expression is:

[0072]

[0073] in, is the predicted value; y i is the actual value; is the average of the actual values. 2 The closer the value is to 1, the better the prediction ability of the model is. 2 The closer the value is to 0, the poorer the prediction ability of the model.

[0074] Step S53: Visually compare the model prediction value and the true value in step S4 to intuitively display the prediction performance evaluation result of the model. The visualization includes drawing a line graph of the model prediction value and the true value, and drawing a residual graph between the model prediction value and the true value.

[0075] After experimental measurement, the method of this embodiment shows significant performance improvement over traditional machine learning methods, reaching 95.9% and 92.5% in the prediction accuracy of flux and conductivity, respectively, and an increase of 13.2% and 14.1% respectively compared to the traditional XGBoost model (an ensemble learning model based on a gradient boosting framework that iteratively generates multiple decision trees to improve the prediction accuracy of the model).

[0076] At the same time, the present application also proposes a membrane distillation system performance prediction system based on convolutional neural network, including a computer, wherein the computer is used to execute a computer program constructed according to the above-mentioned membrane distillation system performance prediction method based on convolutional neural network.

[0077] In this embodiment, by introducing convolutional neural networks, the model can capture the complex nonlinear relationship of multi-dimensional input parameters in the membrane distillation system, thereby significantly improving the prediction accuracy of flux and conductivity. This method is applicable to various types of membrane distillation systems and can be extended to the performance prediction and optimization of other membrane separation processes.

[0078] The present invention is described above by way of example in conjunction with the accompanying drawings. Obviously, the specific implementation of the present invention is not limited to the above-mentioned methods. As long as various non-substantial improvements are made using the method concept and technical solution of the present invention; or the above concept and technical solution of the present invention are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.

Claims

1. A method for predicting membrane distillation system performance based on convolutional neural network, characterized in that: The method comprises: Step S1, collecting experimental data of membrane distillation system operating performance; Step S2: standardize the collected data to construct training data set samples and test data set samples; Step S3, constructing a convolutional neural network model, training the convolutional neural network model with training data set samples to obtain an optimal model and fix its model parameters; Step S4, importing the test data set sample into the optimal model of step S3 to predict the operating performance of the membrane distillation system; Step S5: Evaluate the prediction results.

2. The method for predicting membrane distillation system performance based on convolutional neural network according to claim 1, characterized in that: In step S1, the conductive membrane in the membrane distillation system is subjected to a power-on experiment under different parameter conditions; in the power-on experiment under each parameter condition, the power-on process is repeated and monitored in real time by an electronic balance and a conductivity meter connected to a computer, and the flux and conductivity of the conductive membrane during the experiment are recorded at preset time intervals, thereby collecting experimental data on the operating performance of the membrane distillation system.

3. The method for predicting membrane distillation system performance based on convolutional neural network according to claim 1, characterized in that: The step S2 comprises: Step S21, screening the collected experimental data to remove experimental data under parameter conditions that are invalid for the operating performance of the membrane distillation system; Step S22, normalizing the screened data; Step S23: Divide the normalized data into training data set samples and test data set samples according to a preset ratio.

4. The method for predicting membrane distillation system performance based on convolutional neural network according to claim 3, characterized in that: In step S21, data screening is performed using a correlation analysis method, that is, the correlation coefficient between each parameter condition and the target variable is calculated, wherein if the correlation coefficient value is less than a preset correlation coefficient threshold, the corresponding parameter condition is eliminated; the target variables include the flux and conductivity of the conductive film.

5. A membrane distillation system performance prediction method based on convolutional neural network according to claim 3 or 4, characterized in that: In step S22, the normalization adopts the maximum and minimum normalization, that is: Among them, X norm represents the normalized value; X represents the characteristic value of an experimental data after screening; X max , X min They represent the maximum and minimum values ​​of the feature in the entire dataset, respectively.

6. The method for predicting membrane distillation system performance based on convolutional neural network according to claim 1, characterized in that: The step S3 comprises: Step S31, constructing a convolutional neural network model; the convolutional neural network consists of a convolutional layer, a pooling layer, a fully connected layer and an output layer; wherein the convolutional layer is used to extract local features in the input data, and the pooling layer is used to downsample the feature map output by the convolutional layer; the fully connected layer is used to flatten the feature map extracted by the convolution and pooling layers and then output the prediction result through the output layer; Step S32, determining a loss function; Step S33, after randomly initializing the model parameters, use the training data set samples as the input of the model, use the gradient descent algorithm to train the model iteratively, and when the loss function value no longer decreases within a preset period, retain the model and fix its parameters as the optimal model parameters.

7. The method for predicting membrane distillation system performance based on convolutional neural network according to claim 6, characterized in that: In step S32, the loss function is the mean square error, that is: Among them, MSE represents mean square error; represents the predicted value of the i-th sample; y i represents the actual value of the i-th sample; n represents the total number of samples.

8. The method for predicting membrane distillation system performance based on convolutional neural network according to claim 1, characterized in that: The step S5 comprises: Step S51, evaluating the difference between the model predicted value and the true value in step S4 by using the mean square error; Step S52: Determine the coefficient R 2 Assess the explanatory power of the optimal model; Step S53: Visually compare the model prediction value in step S4 with the true value, and the visualization includes drawing line graphs of the model prediction value and the true value, and drawing a residual graph between the model prediction value and the true value.

9. A membrane distillation system performance prediction system based on convolutional neural network, characterized in that: It comprises a computer, which is used to execute a computer program constructed according to a method for predicting membrane distillation system performance based on a convolutional neural network according to any one of claims 1 to 8.

Images