Enterprise pollution discharge behavior analysis method based on power data

Abstract

The application discloses an enterprise pollution discharge behavior analysis method based on power data, comprising the following steps: S1, acquiring original power data, constructing a power history database, classifying the power data and marking the power data with labels; S2, performing data cleaning on the power data; S3, based on original power data features, constructing new power features through feature combination; S4, mining the correlation between power data features and the correlation between power data features and equipment operation states; S5, fusing the new power features constructed in step S3 and the original feature set after removing redundant features in step S4, to obtain a power data key feature library; S6, building a production equipment and pollution control equipment operation state recognition model; and S7, training and parameter tuning are performed on the model built in step S6. The enterprise pollution discharge behavior analysis method based on power data can greatly improve pollution discharge supervision efficiency, reduce supervision cost, and realize intelligentized supervision of enterprise pollution discharge behaviors.

Description

Technical Field

[0001] This invention relates to the field of environmental monitoring technology, and in particular to a method for analyzing corporate pollution discharge behavior based on electricity data. Background Technology

[0002] With the increasing severity of global climate change and environmental problems, green, low-carbon, and circular economic development has become a new trend, making green transformation an urgent and inevitable choice for enterprises. However, the supervision of enterprise pollution generation and discharge still faces severe practical difficulties. On the one hand, the number and types of enterprises are vast, and their production processes are complex; some enterprises, while pursuing economic benefits, have neglected their responsibilities and obligations to environmental protection, leading to significant challenges in pollution discharge supervision. On the other hand, relevant environmental regulatory units generally adopt a crude approach of surprise inspections and periodic checks, resulting in an imperfect regulatory mechanism, low efficiency, and poor regulatory effectiveness. Furthermore, the number of regulatory personnel is far from meeting the actual regulatory needs, leading to heavy regulatory pressure and a heavy workload, making it difficult to cover all enterprises. Therefore, how to strengthen the supervision of enterprise pollution discharge and promote the green transformation of enterprises is a crucial issue that urgently needs to be addressed.

[0003] Enterprises' production relies on equipment, and the operation of equipment depends on electricity. Electricity data is characterized by its large volume, high dimensionality, and high value. Therefore, this invention starts with electricity data, mining the hidden patterns within it to construct a correlation analysis method between enterprise electricity consumption information and production and pollution control equipment. Furthermore, it utilizes artificial intelligence technology to build a status identification model for pollution-generating and pollution-control equipment, enabling the analysis and evaluation of enterprise pollution discharge behavior. This invention provides a new approach to enterprise pollution discharge supervision, which is beneficial for improving the efficiency of pollution discharge supervision, alleviating regulatory pressure, and reducing regulatory costs, thus possessing certain practical value. Summary of the Invention

[0004] The purpose of this invention is to provide a method for analyzing enterprise pollution discharge behavior based on electricity data. By cleaning and mining the electricity data from the enterprise's production equipment and pollution control equipment, the method accurately perceives the correlation between the enterprise's electricity consumption information and the operating status of the production and pollution control equipment. Through feature reconstruction and Pearson correlation analysis, key features in the electricity data are extracted. Based on artificial intelligence technology, a model for recognizing the operating status of production equipment and pollution control equipment based on the CNN-LSTM-Attention algorithm is built. This assists in intelligent decision-making for pollution discharge supervision and achieves a technological breakthrough in pollution discharge supervision from manual supervision to intelligent and automated supervision.

[0005] To achieve the above objectives, this invention provides a method for analyzing enterprise pollution discharge behavior based on electricity data, comprising the following steps:

[0006] S1. Obtain the raw power data of production equipment and pollution control equipment, build a historical power database, classify the power data and label it;

[0007] S2. Perform data cleaning on the power data;

[0008] S3. Based on the original power data characteristics, new power characteristics are constructed through feature combination;

[0009] S4. Mine the correlation between power data features and the correlation between power data features and equipment operating status. For highly correlated feature pairs, remove redundant features to obtain the original feature set.

[0010] S5. Merge the new power features constructed in step S3 with the original feature set after removing redundant features in step S4 to obtain the key feature library of power data.

[0011] S6. Build a model for recognizing the operating status of production equipment and pollution control equipment based on the CNN-LSTM-Attention algorithm;

[0012] S7. Train and optimize the operation status identification model of production equipment and pollution control equipment built in step S6.

[0013] Preferably, in step S1, the raw power data of the enterprise's production equipment and pollution control equipment collected by the contactless smart meter is first obtained. The raw power data includes voltage, current, and power power parameter data. A power history database is constructed, and the raw power data is classified and tagged according to the operating status of the production equipment and pollution control equipment.

[0014] Preferably, the power data is divided into four categories based on the operating status of the production equipment and the pollution control equipment: both production equipment and pollution control equipment are running, neither production equipment nor pollution control equipment is running, production equipment is running and pollution control equipment is not running, and production equipment is not running and pollution control equipment is running.

[0015] Preferably, in step S2, the data cleaning of the power data specifically includes:

[0016] S21. Use the Laida criterion to remove outliers from the power data;

[0017] S22. Use the mean imputation method to fill in the missing values ​​in the power data;

[0018] S23. A Savitzky-Golay filter is used to reduce noise in the power data.

[0019] Preferably, in step S3, three new electrical characteristics are constructed: the power difference between the production equipment and the pollution control equipment, the voltage ratio between the production equipment and the pollution control equipment, and the current ratio between the production equipment and the pollution control equipment.

[0020] Preferably, in step S4, for the original power data feature library, a Pearson correlation coefficient matrix is ​​constructed by calculating the Pearson correlation coefficients between power data features and between power data features and equipment operating status categories. The Pearson correlation coefficient is used to measure the linear correlation between two variables. Let the feature data X be X = {X1, X2, ..., X...} n The feature data of Y are Y = {Y1, Y2, ..., Y}. n The formula for calculating the Pearson correlation coefficient between features X and Y is as follows:

[0021]

[0022] in, Let X be the mean of the feature data. The mean of the Y feature data;

[0023] By setting a correlation coefficient threshold, for the constructed Pearson correlation coefficient matrix, features with correlation coefficient values ​​greater than the threshold are identified as highly correlated feature pairs, and redundant features in the highly correlated feature pairs are removed.

[0024] Preferably, in step S6, a production equipment and pollution control equipment operation status recognition model based on the CNN-LSTM-Attention algorithm is built. The key features obtained in step S5 are used as the input of the model, and the output results are used as the operation status category labels of the production equipment and pollution control equipment mentioned in step S1. The specific process is as follows:

[0025] CNN layers extract local features from the input data through convolution operations. For a convolution kernel W and input X, the formula for calculating the convolution output Y is:

[0026]

[0027] Where b is the bias term, M is the width of the convolution kernel, and N is the height of the convolution kernel;

[0028] LSTM layers achieve precise control over the data flow through forget gates, input gates, cell states, and output gates. The update equation for an LSTM cell within a time step is:

[0029] f t =σ(W f ·[h t-1 ,x t ]+b f )i t

[0030] Among them, f t It is the activation value of the forget gate, σ represents the sigmoid function, and W f It is the weight matrix, h t-1 Indicates the hidden state at the previous time step, x t b represents the input at the current time step. f For the bias term, i t It is the activation value of the input gate;

[0031] Applying the Attention mechanism to the output of an LSTM layer: Let the output sequence of the LSTM layer be H = [h1, h2, ..., h...]. T ],in Let d represent the hidden state of the LSTM at time step t, where T is the sequence length and d is the dimension of the hidden layer. Then, the attention score at time step t for the i-th position in the sequence is expressed as:

[0032]

[0033] Among them, w a W h W s b a These are model parameters, tanh is the hyperbolic tangent activation function, and s t-1 Indicates the hidden state of the previous time step;

[0034] The attention score is converted into attention weights using the softmax function, expressed as follows:

[0035]

[0036] Where, α ti e represents the attention weight at time step t for the i-th position in the sequence. ti This represents the attention score at time step t for the i-th position in the sequence;

[0037] The hidden states of the LSTM are weighted and summed using attention weights to obtain the context vector c. t The context vector c is used as input for subsequent layers. t The formula for expressing it is:

[0038]

[0039] Preferably, in step S7, the model training and parameter tuning specifically include the following steps:

[0040] S71. Normalize the power data from the production equipment end and the pollution control equipment end after data cleaning.

[0041] S72. Divide the dataset obtained in step S71 into a training set, a validation set, and a test set according to the proportions.

[0042] S73. Input the training set into the production equipment and pollution control equipment operation status recognition model, select four indicators: accuracy, precision, recall and F1-Score to evaluate the model performance, and use ten-fold cross-validation to obtain the final indicator score of the model.

[0043] S74. Set model evaluation index standards, verify model performance on the validation set, continuously optimize and adjust model parameters until the model output meets the index requirements, and obtain a trained production equipment and pollution control equipment operation status identification model.

[0044] S75. Input the test set into the trained model for testing, and output the operating status information of production equipment and pollution control equipment to assist in pollution discharge supervision and decision-making.

[0045] Therefore, the present invention employs the above-mentioned method for analyzing enterprise pollution discharge behavior based on electricity data, and the beneficial effects are as follows:

[0046] (1) This invention analyzes and processes the power data of the enterprise's production equipment and pollution control equipment, thereby realizing the automatic identification of the operating status of the production equipment and pollution control equipment, thus realizing data-driven analysis and evaluation of the enterprise's pollution discharge behavior, providing a new approach to pollution discharge supervision, which can greatly improve the efficiency of pollution discharge supervision, reduce supervision costs, and realize intelligent supervision of the enterprise's pollution discharge behavior.

[0047] (2) This invention uses data cleaning and mining techniques such as the Laida criterion, Savitzky-Golay filtering, and Pearson correlation analysis to effectively improve data quality, enhance data interpretability, mine implicit information from different dimensions of power data, construct the coupling relationship between enterprise power consumption information and the operating status of production equipment and pollution control equipment, solve the feature redundancy problem, and lay a solid foundation for subsequent model building.

[0048] (3) This invention establishes a production equipment and pollution control equipment operation status identification model based on CNN-LSTM-Attention algorithm. By combining CNN, LSTM and Attention mechanism, this model can efficiently extract local features in the data, accurately capture long-term dependencies in time series data, and enhance the ability to focus on key information during model training. This is beneficial to improving the accuracy and robustness of the classification model and has good practical application value.

[0049] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0050] Figure 1 This is an overall flowchart of an embodiment of the enterprise pollution discharge behavior analysis method based on power data of the present invention. Detailed Implementation

[0051] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0052] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0053] like Figure 1 As shown, a method for analyzing enterprise pollution discharge behavior based on electricity data includes the following steps:

[0054] S1. Obtain the raw power data of production equipment and pollution control equipment, build a historical power database, classify the power data and label it.

[0055] First, the raw power data of the enterprise's production equipment and pollution control equipment is collected by non-contact smart meters. The raw power data includes power parameters such as voltage, current, and power. A historical power database is then built, and the raw power data is classified and tagged according to the operating status of the production equipment and pollution control equipment.

[0056] Specifically, based on the operating status of production equipment and pollution control equipment, power data can be divided into four categories: both production equipment and pollution control equipment are running, neither production equipment nor pollution control equipment is running, production equipment is running and pollution control equipment is not running, and production equipment is not running and pollution control equipment is running. In actual production, the last category will basically not occur.

[0057] S2. Perform data cleaning on the power data to improve data quality, specifically including:

[0058] S21. Use the Laida criterion to remove outliers from the power data.

[0059] Specifically, for a certain power parameter data D={x1,x2,...,x n} Calculate the mean and standard deviation of this set of data using the following formulas:

[0060]

[0061] in, Let be the mean of the data set, and σ be the standard deviation of the data set. For the i-th data point x in D... i ,like Then it means x i This is abnormal data and should be removed.

[0062] S22. Use the mean imputation method to fill in the missing values ​​in the power data;

[0063] Specifically, the `pandas.isnan()` function in Python is used to detect missing values ​​in the data. If missing values ​​are found, imputation is performed using the mean of the data before and after the missing value to fill the gap. For a given power parameter dataset D = {x1, x2, ..., x...}, ... n The formula for calculating the fill value is as follows:

[0064]

[0065] Where, x' j To fill in the value, x j-5 x j-4 ....x j+5 Data before and after the missing point.

[0066] S23. A Savitzky-Golay filter is used to reduce noise in the power data.

[0067] Specifically, a Savitzky-Golay filter is used to process data noise, achieving data denoising and smoothing effects. The Savitzky-Golay filter is a special type of low-pass filter that fits the value of each data point by constructing a local polynomial and then uses the least squares method to optimize the polynomial coefficients, achieving the best fitting effect and thus denoising the data.

[0068] For each data point x in the power data i x after being processed by the Savitzky-Golay filter i It can be calculated using the following formula:

[0069]

[0070] Where m is the order of the polynomial, c k x represents the polynomial coefficients calculated using the least squares method. i+k x is in the original data sequence i Data values ​​at a distance k around the perimeter.

[0071] S3. Based on the original power data characteristics, new power characteristics are constructed through feature combination.

[0072] Specifically, since the raw power data contains information on voltage, current, power, etc. of both production equipment and pollution control equipment, these parameters do not intuitively reflect the relationship between the enterprise's power consumption information and the operating status of production and pollution control equipment. In order to explore the deeper meaning of the data, three new power features that are more interpretable are constructed by combining the features of the raw parameters: the power difference between production equipment and pollution control equipment, the voltage ratio between production equipment and pollution control equipment, and the current ratio between production equipment and pollution control equipment.

[0073] S4. By constructing a Pearson correlation coefficient matrix, we can mine the correlations between power data features and between power data features and equipment operating status. Here, power data features refer to the parameter features in the power data, such as current, voltage, and power as mentioned in the previous steps. Specifically, we can explore the correlations between power data features, such as the correlation between voltage and power; and the correlations between power data features and equipment operating status, such as the relationship between current and equipment switching status.

[0074] For highly correlated feature pairs, redundant features are removed to obtain the original feature set, thus achieving feature dimensionality reduction.

[0075] Specifically, for the original power data feature library, a Pearson correlation coefficient matrix is ​​constructed by calculating the Pearson correlation coefficients between power data features and between power data features and equipment operating status categories. The Pearson correlation coefficient is used to measure the linear correlation between two variables. Let the feature data X be X = {X1, X2, ..., X...} n The feature data of Y are Y = {Y1, Y2, ..., Y}. n The formula for calculating the Pearson correlation coefficient between features X and Y is as follows:

[0076]

[0077] in, Let X be the mean of the feature data. The mean of the Y feature data.

[0078] By setting a certain correlation coefficient threshold, features with correlation coefficient values ​​greater than the threshold are identified as highly correlated feature pairs for the constructed Pearson correlation coefficient matrix. Highly correlated feature pairs often exhibit strong interdependencies and may carry a large amount of overlapping information. To reduce model complexity and computational cost, and improve model generalization ability, redundant features in highly correlated feature pairs need to be removed.

[0079] Specifically, for highly correlated feature pairs, only the feature with the largest absolute value of the correlation coefficient with the equipment operating status category is retained, and the remaining redundant features are eliminated.

[0080] S5. The new power features constructed in step S3 are recombined and fused with the original feature set after removing redundant features in step S4 to obtain a power data key feature library with minimized dimensionality that encompasses almost all power data information. These key features are used as input parameters for subsequent models.

[0081] S6. Build a production equipment and pollution control equipment operation status recognition model based on CNN-LSTM-Attention algorithm; take the key features obtained in step S5 as the input of the model, and output the operation status category labels of production equipment and pollution control equipment in step S1.

[0082] Specifically, CNN layers extract local features from the input data through convolution operations. The core process involves multiplying and summing the convolution kernel and the input features. For a convolution kernel W and an input X, the formula for calculating the convolution output Y is:

[0083]

[0084] Where b is the bias term, M is the width of the convolution kernel, and N is the height of the convolution kernel.

[0085] An LSTM layer is a special type of recurrent neural network structure that achieves precise control over the flow of data information through forget gates, input gates, cell states, and output gates. The update equation for an LSTM unit within one time step is:

[0086] f t =σ(W f ·[h t-1 ,x t ]+b f )i t ,

[0087] Among them, f t It is the activation value of the forget gate, σ represents the sigmoid function, and W f It is the weight matrix, h t-1 Indicates the hidden state at the previous time step, x t b represents the input at the current time step. f For the bias term, i t It is the activation value of the input gate.

[0088] In this model, the attention mechanism is applied to the output of the LSTM layer. Let the output sequence of the LSTM layer be H = [h1, h2, ..., h...]. T ],in Let d represent the hidden state of the LSTM at time step t, where T is the sequence length and d is the dimension of the hidden layer. Then, the attention score at time step t for the i-th position in the sequence can be expressed as:

[0089]

[0090] Among them, w a W h W s b a These are model parameters, tanh is the hyperbolic tangent activation function, and s t-1 This indicates the hidden state of the previous time step.

[0091] Next, the attention score is converted into attention weights using the softmax function, expressed as follows:

[0092]

[0093] Where, α ti e represents the attention weight at time step t for the i-th position in the sequence. ti Let represent the attention score at time step t for the i-th position in the sequence.

[0094] Next, the hidden states of the LSTM are weighted and summed using attention weights to obtain the context vector c. t The context vector c is used as input for subsequent layers. t The formula for expressing it is:

[0095]

[0096] S7. Train and optimize the operation status recognition model for production equipment and pollution control equipment built in step S6 to achieve the best recognition effect. When the model identifies that the pollution control equipment is off while the production equipment is running, it means that the enterprise has not taken pollution control measures, which constitutes illegal production. The enterprise needs to be warned and punished, thereby achieving effective supervision of the enterprise's production pollution discharge. The model training and parameter optimization specifically include the following steps:

[0097] S71. Normalize the power data from the production equipment end and the pollution control equipment end after data cleaning.

[0098] S72. Divide the dataset obtained in step S71 into a training set, a validation set, and a test set in a ratio of 8:1:1.

[0099] S73. Input the training set into the production equipment and pollution control equipment operation status recognition model, select four indicators—accuracy, precision, recall, and F1-Score—to evaluate the model performance, and use ten-fold cross-validation to obtain the final indicator score of the model.

[0100] The calculation formulas for the four indicators are as follows:

[0101]

[0102] In this system, TP represents true positives, TN represents true negatives, FP represents false positives, FN represents false negatives, and n represents the total number of classes. Precision, recall, and F1-Score are all calculated for specific classes.

[0103] S74. Set model evaluation index standards, verify model performance on the validation set, continuously optimize and adjust model parameters until the model output results meet the index requirements, and obtain a trained production equipment and pollution control equipment operation status identification model.

[0104] S75. Input the test set into the trained model for testing, and output the operating status information of production equipment and pollution control equipment to assist in pollution discharge supervision and decision-making.

[0105] Therefore, this invention employs the aforementioned method for analyzing enterprise pollution discharge behavior based on electricity data. Starting with electricity data, it mines the hidden patterns within the data to construct a correlation analysis method between enterprise electricity consumption information and production and pollution control equipment. Furthermore, it leverages artificial intelligence technology to build a status identification model for pollution-generating and pollution-control equipment, thereby enabling the analysis and evaluation of enterprise pollution discharge behavior. This provides a new approach to enterprise pollution discharge supervision, which is beneficial for improving the efficiency of pollution discharge supervision, alleviating regulatory pressure, and reducing regulatory costs, thus possessing certain practical value.

[0106] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for analyzing enterprise pollution discharge behavior based on electricity data, characterized in that, Includes the following steps: S1. Obtain the raw power data of production equipment and pollution control equipment, build a historical power database, classify the power data and label it; S2. Perform data cleaning on the power data; S3. Based on the original power data characteristics, new power characteristics are constructed through feature combination; S4. Mine the correlation between power data features and the correlation between power data features and equipment operating status. For highly correlated feature pairs, remove redundant features to obtain the original feature set. S5. Merge the new power features constructed in step S3 with the original feature set after removing redundant features in step S4 to obtain the key feature library of power data. S6. Build a model for recognizing the operating status of production equipment and pollution control equipment based on the CNN-LSTM-Attention algorithm; S7. Train and optimize the operation status identification model of production equipment and pollution control equipment built in step S6. In step S6, a model for recognizing the operating status of production equipment and pollution control equipment based on the CNN-LSTM-Attention algorithm is built. The key features obtained in step S5 are used as the input to the model, and the output results are used as the category labels for the operating status of the production equipment and pollution control equipment mentioned in step S1. The specific process is as follows: CNN layers extract local features from the input data through convolution operations. For a single convolution kernel... W and input X Convolution output Y The calculation formula is: ； in, b For bias terms, M The width of the convolution kernel. N The height of the convolution kernel; LSTM layers achieve precise control over the data flow through forget gates, input gates, cell states, and output gates. The update equation for an LSTM cell within a time step is: ； in, It is the activation value of the Forgotten Gate. This represents the sigmoid function. It is a weight matrix. This indicates the hidden state of the previous time step. This represents the input at the current time step. For bias terms, It is the activation value of the input gate; Applying the Attention mechanism to the output of the LSTM layer: Let the output sequence of the LSTM layer be... ,in , represents the hidden state of the LSTM at time step t. T It is the sequence length. d If it is the dimension of the hidden layer, then the time step t For the first in the sequence i The attention score for each position is represented as follows: ； in, , , , These are the model parameters, and tanh is the hyperbolic tangent activation function. Indicates the hidden state of the previous time step; The attention score is converted into attention weights using the softmax function, expressed as follows: ； in, Indicates time step For the first in the sequence Attention weights for each position, Indicates time step For the first in the sequence Attention score for each position; The hidden states of the LSTM are weighted and summed using attention weights to obtain the context vector. Context vector used as input for subsequent layers The formula for expressing it is: 。 2. The method for analyzing enterprise pollution discharge behavior based on electricity data according to claim 1, characterized in that, In step S1, the raw power data of the enterprise's production equipment and pollution control equipment collected by the non-contact smart meter is first obtained. The raw power data includes voltage, current and power power parameter data. A power history database is constructed, and the raw power data is classified and tagged according to the operating status of the production equipment and pollution control equipment.

3. The method for analyzing enterprise pollution discharge behavior based on electricity data according to claim 2, characterized in that, Based on the operating status of the production equipment and pollution control equipment, the power data is divided into four categories: both production equipment and pollution control equipment are running, neither production equipment nor pollution control equipment is running, production equipment is running and pollution control equipment is not running, and production equipment is not running and pollution control equipment is running.

4. The method for analyzing enterprise pollution discharge behavior based on electricity data according to claim 3, characterized in that, In step S2, the data cleaning of the power data specifically includes: S21. Use the Laida criterion to remove outliers from the power data; S22. Use the mean imputation method to fill in the missing values ​​in the power data; S23. A Savitzky-Golay filter is used to reduce noise in the power data.

5. The method for analyzing enterprise pollution discharge behavior based on electricity data according to claim 4, characterized in that, In step S3, three new electrical characteristics are constructed: the power difference between the production equipment and the pollution control equipment, the voltage ratio between the production equipment and the pollution control equipment, and the current ratio between the production equipment and the pollution control equipment.

6. The method for analyzing enterprise pollution discharge behavior based on electricity data according to claim 5, characterized in that, In step S4, for the original power data feature library, a Pearson correlation coefficient matrix is ​​constructed by calculating the Pearson correlation coefficients between power data features and between power data features and equipment operating status categories. The Pearson correlation coefficient is used to measure the linear correlation between two variables. Let... X Feature data is , Y Feature data is ,but X and Y The formula for calculating the Pearson correlation coefficient between features is as follows: in, Let X be the mean of the feature data. The mean of the Y feature data; By setting a correlation coefficient threshold, for the constructed Pearson correlation coefficient matrix, features with correlation coefficient values ​​greater than the threshold are identified as highly correlated feature pairs, and redundant features in the highly correlated feature pairs are removed.

7. The method for analyzing enterprise pollution discharge behavior based on electricity data according to claim 1, characterized in that, Step S7, specifically the model training and parameter tuning, includes the following steps: S71. Normalize the power data from the production equipment end and the pollution control equipment end after data cleaning. S72. Divide the dataset obtained in step S71 into a training set, a validation set, and a test set according to the proportions. S73. Input the training set into the production equipment and pollution control equipment operation status recognition model, select four indicators: accuracy, precision, recall and F1-Score to evaluate the model performance, and use ten-fold cross-validation to obtain the final indicator score of the model. S74. Set model evaluation index standards, verify model performance on the validation set, continuously optimize and adjust model parameters until the model output meets the index requirements, and obtain a trained production equipment and pollution control equipment operation status identification model. S75. Input the test set into the trained model for testing, and output the operating status information of production equipment and pollution control equipment to assist in pollution discharge supervision and decision-making.

Images