Waste gas emission remote management system in kitchen waste treatment process

Abstract

The invention provides a remote management system for waste gas emission in a kitchen waste treatment process, and relates to the technical field of intelligent control, and the method comprises a prediction module which is used for building a convolutional neural network model between waste gas parameters and the opening degree of a discharge port by using preprocessed historical data and preprocessed real-time data; setting a prediction time domain, and determining exhaust gas parameters in a future time domain by using the trained convolutional neural network model based on a current state and a control variable; the comparison module is used for determining the pollutant concentration change rate according to the exhaust gas parameters in the future time domain; comparing the pollutant concentration change rate with a dynamic threshold value to obtain a comparison result; the control module is used for generating a final control action at the current moment according to the comparison result; and according to the final control action, executing opening or closing operation of the waste gas exhaust port. The management level of waste gas emission in the kitchen waste treatment process is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control, and particularly to a remote management system for exhaust gas emissions during the treatment of kitchen waste. Background Art

[0002] When dealing with a large amount of kitchen waste in the city, the exhaust gas generated during the treatment process contains various harmful substances, such as hydrogen sulfide, ammonia, and volatile organic compounds (VOCs). In order to control exhaust gas emissions, exhaust gas monitoring equipment has been installed, which can collect data such as the concentration of pollutants and the emission flow rate in the exhaust gas in real time. However, in actual operation, these data have not been fully utilized, resulting in some problems in exhaust gas emission management.

[0003] For example, due to the lack of professional data analysis, the collected exhaust gas data cannot be deeply analyzed. Even if a large amount of data is collected, valuable information such as the exhaust gas emission pattern and the change trend of pollutant concentration cannot be extracted from it, resulting in the lack of intelligence in management decisions.

[0004] For another example, the exhaust gas emission data is mainly used to meet the supervision requirements of environmental protection departments, rather than being used to optimize the exhaust gas treatment process or adjust the emission strategy. The exhaust gas treatment equipment may operate in an inefficient state for a long time, resulting in an increase in treatment costs. At the same time, the exhaust gas emissions may still exceed the standards, causing pollution to the environment. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a remote management system for exhaust gas emissions during the treatment of kitchen waste, which improves the management level of exhaust gas emissions during the treatment of kitchen waste.

[0006] To solve the above technical problem, the technical solution of the present invention is as follows:

[0007] In a first aspect, a remote management system for exhaust gas emissions during the treatment of kitchen waste includes:

[0008] A sensor group, which is arranged at the exhaust gas emission port of the kitchen waste treatment facility and is used for real-time monitoring of exhaust gas parameters;

[0009] A processing module, which is used for preprocessing the exhaust gas parameters to obtain preprocessed real-time data; determining an adjustment weight according to the preprocessed historical data and the preprocessed real-time data; and determining a dynamic threshold according to the adjustment weight;

[0010] A prediction module, which is used for using the preprocessed historical data and the preprocessed real-time data to establish a convolutional neural network model between the exhaust gas parameters and the opening degree of the emission port; setting a prediction time domain, and based on the current state and control variables, using the trained convolutional neural network model to determine the exhaust gas parameters within the future time domain;

[0011] A comparison module, configured to determine the change rate of pollutant concentration according to the waste gas parameters within a future time domain; compare the change rate of pollutant concentration with a dynamic threshold to obtain a comparison result;

[0012] A control module, configured to generate a final control action at the current moment according to the comparison result; perform an opening or closing operation on the waste gas discharge port according to the final control action.

[0013] Furthermore, determine an adjustment weight according to the preprocessed historical data and the preprocessed real-time data, including:

[0014] Extract all pollutant concentration values from the real-time data window to determine the real-time average value of all pollutant concentration values; determine the real-time standard deviation based on the pollutant concentration values within the real-time data window;

[0015] Determine the historical concentration values of all pollutants after preprocessing within a past period of time, and determine the historical average value and historical standard deviation of all pollutants according to the historical concentration values;

[0016] Determine the absolute difference between the real-time average value and the historical average value; determine the standardized deviation value according to the absolute difference and the historical standard deviation;

[0017] Determine the real-time volatility according to the real-time standard deviation and the historical standard deviation;

[0018] Determine the mean deviation term according to the mean deviation item and the corresponding adjustment coefficient; determine the volatility term according to the real-time volatility and the corresponding adjustment coefficient;

[0019] Fuse the mean deviation term and the volatility term to obtain the adjustment weight.

[0020] Furthermore, determine a dynamic threshold according to the adjustment weight, including:

[0021] Obtain the concentration data of each pollutant after preprocessing within a past period of time;

[0022] Determine the historical average value and historical standard deviation of each pollutant according to the concentration data of each pollutant after preprocessing;

[0023] Fuse the product of the historical average value, the weight coefficient and the historical standard deviation to determine the basic threshold of each pollutant;

[0024] Determine the dynamic threshold of each pollutant according to the basic threshold.

[0025] Furthermore, use the preprocessed historical data and the preprocessed real-time data to establish a convolutional neural network model between the waste gas parameters and the opening degree of the discharge port, including:

[0026] Determine the Pearson correlation coefficient between the waste gas parameters and the opening degree of the emission port;

[0027] Determine the corresponding waste gas parameter as a feature according to the magnitude of the Pearson correlation coefficient;

[0028] Divide the features into a training set and a test set;

[0029] Use the training set data to train the convolutional neural network model, and optimize the parameters of the convolutional neural network model through the backpropagation algorithm; during the training process, evaluate the stability and generalization ability of the convolutional neural network model through the K-fold cross-validation method;

[0030] Use the grid search method to adjust the hyperparameters of the convolutional neural network model to obtain the final hyperparameter combination, and use the test set to evaluate the performance of the convolutional neural network model to obtain the accuracy rate;

[0031] Optimize the convolutional neural network model according to the accuracy rate to obtain the trained convolutional neural network model.

[0032] Furthermore, determine the pollutant concentration change rate according to the waste gas parameters in the future time domain, including:

[0033] Determine the size of the sliding window, that is, the number of data points included in the window;

[0034] Create an array to store the data within the sliding window, and the array is used to save the adjacent N data points, where N is the size of the sliding window;

[0035] When new preprocessed future time domain data arrives, perform the following operations:

[0036] Add the new data point to the end of the sliding window array; if the array is full, that is, it has reached the size N of the sliding window, remove the first data point in the array to keep the window size unchanged;

[0037] Whenever the sliding window is updated, perform a linear fit using the N data points within the window, specifically including:

[0038] Extract all the concentration data points within the sliding window, where the x value represents the time point and the y value represents the corresponding pollutant concentration; through the least squares formula, determine the slope and intercept of the fitting line, and the slope represents the change rate of the pollutant concentration.

[0039] Furthermore, the waste gas parameters include H2S concentration, NH3 concentration, VOCs concentration, waste gas temperature, waste gas humidity, and waste gas emission flow rate.

[0040] Furthermore, generate the final control action at the current moment according to the comparison result, including:

[0041] If the change rate of pollutant concentration is lower than the dynamic threshold, the current state of the emission port remains unchanged;

[0042] If the change rate of pollutant concentration is equal to the dynamic threshold, a warning is issued and preparations are made to adjust the opening of the emission port;

[0043] If the change rate of pollutant concentration is higher than the dynamic threshold, the opening of the emission port is adjusted or the emission port is completely closed.

[0044] In a second aspect, a method for remote management of waste gas emissions during the treatment of kitchen waste includes:

[0045] Real-time monitoring of the waste gas parameters at the waste gas emission port of the kitchen waste treatment facility;

[0046] Preprocessing the waste gas parameters to obtain preprocessed real-time data; determining an adjustment weight based on the preprocessed historical data and the preprocessed real-time data; determining a dynamic threshold based on the adjustment weight;

[0047] Using the preprocessed historical data and the preprocessed real-time data, establishing a convolutional neural network model between the waste gas parameters and the opening of the emission port; setting a prediction time domain, and based on the current state and control variables, using the trained convolutional neural network model to determine the waste gas parameters within the future time domain;

[0048] Determining the change rate of pollutant concentration based on the waste gas parameters within the future time domain; comparing the change rate of pollutant concentration with the dynamic threshold to obtain a comparison result;

[0049] Generating a final control action at the current moment based on the comparison result; performing an opening or closing operation of the waste gas emission port according to the final control action.

[0050] In a third aspect, a computing device includes:

[0051] One or more processors;

[0052] A storage device for storing one or more programs, which when executed by the one or more processors, cause the one or more processors to implement the method.

[0053] In a fourth aspect, a computer-readable storage medium stores a program that, when executed by a processor, implements the method.

[0054] The above solution of the present invention has at least the following beneficial effects:

[0055] Through the sensor group set at the exhaust gas outlet of the food waste treatment facility, the present invention can monitor the exhaust gas parameters in real time, ensuring the timeliness and accuracy of the data. Combining the preprocessing of the data by the processing module, the system can more accurately grasp the actual situation of the exhaust gas emissions.

[0056] The present invention adopts a dynamic threshold determination method, which adjusts the weight according to the preprocessed historical data and real-time data, so as to obtain a threshold that better fits the current actual situation. This dynamic adjustment method is more flexible and adaptable than the traditional fixed threshold, and can effectively cope with various changes in the exhaust gas emission process.

[0057] The prediction module uses a convolutional neural network model to predict the exhaust gas parameters in the future time domain based on historical data and real-time data. This prediction ability enables the system to take measures in advance to avoid excessive pollutant emissions, improving the predictability and initiative of the exhaust gas treatment.

[0058] By determining the change rate of the pollutant concentration and comparing it with the dynamic threshold, the comparison module can intelligently judge whether the exhaust gas emission meets the standard. The control module then generates the final control action according to the comparison result and automatically executes the opening or closing operation of the exhaust gas outlet, realizing the intelligent management of the exhaust gas emission. Description of the Drawings

[0059] Figure 1 It is a schematic diagram of a remote management system for exhaust gas emissions during the food waste treatment process provided by an embodiment of the present invention.

[0060] Figure 2 It is a schematic flowchart of a remote management method for exhaust gas emissions during the food waste treatment process provided by an embodiment of the present invention. Detailed Embodiments

[0061] Hereinafter, exemplary embodiments of the present disclosure will be described in more detail with reference to the drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0062] As Figure 1 shown, an embodiment of the present invention provides a remote management system for exhaust gas emissions during the food waste treatment process, including:

[0063] A sensor group, set at the exhaust gas outlet of the food waste treatment facility, for real-time monitoring of exhaust gas parameters, and the exhaust gas parameters include H2S concentration, NH3 concentration, VOCs concentration, exhaust gas temperature, exhaust gas humidity, and exhaust gas emission flow rate;

[0064] A processing module for preprocessing waste gas parameters to obtain preprocessed real-time data; determining an adjustment weight based on the preprocessed historical data and the preprocessed real-time data; and determining a dynamic threshold based on the adjustment weight.

[0065] A prediction module for using the preprocessed historical data and the preprocessed real-time data to establish a convolutional neural network model between the waste gas parameters and the opening degree of the emission port; setting a prediction time domain, and based on the current state and control variables, using the trained convolutional neural network model to determine the waste gas parameters in the future time domain.

[0066] A comparison module for determining the pollutant concentration change rate according to the waste gas parameters in the future time domain; comparing the pollutant concentration change rate with the dynamic threshold to obtain a comparison result.

[0067] A control module for generating a final control action at the current moment according to the comparison result; and performing an opening or closing operation of the waste gas emission port according to the final control action.

[0068] In the embodiment of the present invention, the sensor group can monitor a variety of waste gas parameters in real time, including H2S concentration, NH3 concentration, VOCs concentration, waste gas temperature, waste gas humidity, and waste gas emission flow rate, ensuring a comprehensive grasp of the waste gas emission status, improving the monitoring accuracy and the pertinence of waste gas treatment. The processing module can remove noise and outliers by preprocessing the waste gas parameters, improving the data quality. At the same time, the dynamic threshold is determined by using the preprocessed historical data and real-time data, enabling the system to flexibly adjust the threshold according to different situations, enhancing the adaptability and accuracy of the system. The prediction module can accurately predict the waste gas parameters in the future time domain by using the convolutional neural network model. This prediction function enables the system to make a response in advance, effectively preventing the waste gas emission from exceeding the standard, and improving the forward-looking and control effect of the system. The comparison module can timely detect abnormal situations of waste gas emissions by determining the pollutant concentration change rate and comparing it with the dynamic threshold. The control module generates a final control action according to the comparison result and automatically performs the opening or closing operation of the waste gas emission port, realizing the intelligent and automatic management of waste gas emissions, and improving the processing efficiency and response speed.

[0069] In a preferred embodiment of the present invention, the sensor group is a key component deployed at the waste gas emission port of the food waste treatment facility, and it undertakes the important task of monitoring waste gas parameters in real time. These waste gas parameters are crucial for evaluating the quality of waste gas and its environmental impact. Specifically, the waste gas parameters monitored by the sensor group include:

[0070] The H2S concentration, hydrogen sulfide (H2S) is a toxic gas that is harmful to both humans and the environment. Monitoring the H2S concentration can help detect potential safety risks in a timely manner.

[0071] NH3 concentration: Ammonia (NH3) is also a toxic gas. Excessive concentration of it will pose a threat to human health and may have negative impacts on the environment.

[0072] VOCs concentration: Volatile organic compounds (VOCs) are a type of widespread air pollutants, which are harmful to both human health and the environment. Monitoring the VOCs concentration is an important measure for controlling air pollution.

[0073] Exhaust gas temperature: The exhaust gas temperature is an important factor affecting the exhaust gas treatment effect. Too high or too low temperature may affect the efficiency of the treatment facilities.

[0074] Exhaust gas humidity: Humidity will affect the treatment and emission effects of the exhaust gas. Especially in some treatment processes, humidity control is crucial.

[0075] Exhaust gas emission flow rate: The flow rate directly affects the amount of exhaust gas discharged from the treatment facilities per unit time, and is an important indicator for evaluating the treatment efficiency and emission control.

[0076] By real-time monitoring of these parameters, the sensor group can provide real-time data on the exhaust gas condition, providing key information for subsequent exhaust gas treatment and control.

[0077] The processing module receives the raw data from the sensor group and performs necessary preprocessing to obtain more accurate and reliable real-time data. The preprocessing steps may include:

[0078] Data cleaning: Removing outliers or noisy data caused by sensor failures, electromagnetic interference or other reasons. Data standardization: Converting the data into standard units or ranges for subsequent data analysis and processing. Data smoothing: Reducing the random fluctuations in the data and highlighting the main trends of the data; The preprocessed real-time data is not only more accurate, but also easier to be used by subsequent analysis and control modules.

[0079] In a preferred embodiment of the present invention, according to the preprocessed historical data and the preprocessed real-time data, an adjustment weight is determined, including:

[0080] Extract all pollutant concentration values from the real-time data window and determine the real-time average value of all pollutant concentration values; based on the pollutant concentration values within the real-time data window, determine the real-time standard deviation. For example: Set a real-time data window that contains pollutant concentration data for a recent period of time. Extract all relevant pollutant concentration values from this window, which are stored in a time series format. Sum up all the extracted pollutant concentration values, divide the sum by the total number of pollutant concentration values to obtain the real-time average value. This average value represents the average level of pollutant concentration within the current window. Square the difference between each pollutant concentration value and the real-time average value, sum up all the squared differences, then divide by the total number of pollutant concentration values minus one (to obtain the sample standard deviation), and take the square root of the sum to get the real-time standard deviation. This standard deviation reflects the fluctuation of pollutant concentration within the current window.

[0081] Determine the historical concentration values of all pollutants after pretreatment over a past period of time, and determine the historical average value and historical standard deviation of all pollutants based on the historical concentration values. For example: Select a historical time period that should be long enough to contain a sufficient number of historical data points. Extract all the pollutant concentration values after pretreatment within this time period from the historical database; the determination of the historical average value is similar to the above, except that historical data is used; the determination of the historical standard deviation is similar to the above steps and is also based on historical data.

[0082] Determine the absolute difference between the real-time average value and the historical average value; divide the absolute difference by the historical standard deviation to obtain the standardized deviation value. For example: Subtract the historical average value from the real-time average value to get the difference, take the absolute value of this difference to get the absolute difference. This absolute difference reflects the degree of deviation of the current pollutant concentration from the historical average level; divide the obtained absolute difference by the historical standard deviation, and the result is the standardized deviation value, which represents the standardized degree of deviation of the current pollutant concentration from the historical average level.

[0083] Divide the real-time standard deviation by the historical standard deviation to obtain the real-time volatility. For example: Divide the real-time standard deviation by the historical standard deviation, and the result is the real-time volatility, which reflects the multiple of the current pollutant concentration fluctuation relative to the historical average fluctuation.

[0084] Multiply the mean deviation term by an adjustment coefficient to obtain the mean deviation term; multiply the real-time volatility by (1 - adjustment coefficient) to obtain the volatility term. For example: Select an adjustment coefficient that lies between 0 and 1 and is used to trade off between mean deviation and volatility. Multiply the standardized deviation value by the adjustment coefficient to obtain the mean deviation term. This term represents the importance of mean deviation when considering adjusting the weights; multiply the real-time volatility by (1 minus the adjustment coefficient), and the result is the volatility term, which represents the importance of real-time volatility when considering adjusting the weights. Among them, the formula for determining the adjustment coefficient α is:

[0085]

[0086] where k is the adjustment sensitivity coefficient (e.g., k = 0.5); when the mean deviation is large, α → 1 (prioritize responding to mean changes); when the mean deviation is small, α → 0.5 (balance the mean and volatility); μ r represents the real-time average value; μ t represents the historical average value; e represents the base of the natural logarithm.

[0087] Add the mean deviation term and the volatility term to obtain the adjusted weight. For example: Add the obtained mean deviation term and the obtained volatility term, and the result is the adjusted weight, which comprehensively considers the mean deviation and volatility of the current pollutant concentration and can be used to guide the adjustment of the waste gas treatment and control strategy.

[0088] In the embodiments of the present invention, by determining the absolute difference between the real-time average value and the historical average value and normalizing it, this method can capture the real-time changes in the pollutant concentration in the waste gas; the comparison between the real-time standard deviation and the historical standard deviation provides a quantitative index for the volatility of waste gas emissions. If the real-time volatility increases significantly, it may mean that there are abnormalities or upcoming problems in the current treatment process, thus providing a risk warning for the operator and helping to take timely measures to prevent potential environmental pollution incidents. By introducing an adjustment coefficient, this method allows the operator to adjust between the mean deviation and volatility according to actual needs. This flexibility enables the system to optimize the weight allocation under different circumstances. For example, when strict control of the pollutant concentration mean is required, the weight of the mean deviation term can be increased. By comprehensively considering real-time data and historical data, this method can more accurately evaluate the efficiency of the current waste gas treatment, thereby guiding the operator to optimize the treatment process, such as adjusting the dosage of chemical agents, changing the temperature or pressure of waste gas treatment, etc., to improve the treatment efficiency. This method analyzes and adjusts weights based on data, providing objective decision-making support for waste gas treatment. This data-driven method reduces human intervention and subjective judgment, making the waste gas treatment more scientific and precise. Through continuous monitoring and analysis of real-time data and historical data, the operator can continuously discover problems and improvement points in the waste gas treatment process, thereby promoting the continuous optimization and improvement of the entire treatment system.

[0089] In a preferred embodiment of the present invention, determining a dynamic threshold according to the adjusted weight includes:

[0090] Obtain the concentration data of each pollutant after pretreatment in the past period of time. For example, determine the specific range of the "past period of time", which can be a fixed time period (such as the past 24 hours, the past week, etc.) or a dynamic time window (such as the most recent N data points); retrieve all the concentration data of the pollutants after pretreatment within this time range from the database or data storage system. These data have undergone pretreatment steps such as cleaning, standardization, and smoothing to ensure data quality and consistency.

[0091] According to the concentration data of each pollutant after pretreatment, determine the historical average value and historical standard deviation of each pollutant. For example, for each pollutant, add up all its concentration data within the selected time range and then divide by the total number of data points to obtain the historical average value of this pollutant, which reflects the average concentration level of this pollutant in the past period of time; the standard deviation is a statistic that measures the volatility of data. For each pollutant, first determine the difference (deviation) between each data point and the historical average value, then add up the squares of all deviations, divide by the total number of data points minus one (to obtain the sample standard deviation), and finally take the square root of the result to obtain the historical standard deviation of this pollutant. This value reflects the fluctuation of the concentration of this pollutant in the past period of time.

[0092] The basic threshold of each pollutant is determined by historical average + weight coefficient × historical standard deviation. For example: The weight coefficient is a parameter set according to actual needs and is used to adjust the dependence of the basic threshold on the historical average and historical standard deviation. This coefficient can be determined based on experience; add the historical average of each pollutant to the result of multiplying the weight coefficient by the historical standard deviation to obtain the basic threshold of the pollutant. This value is a statistical threshold based on historical data and is used to preliminarily judge whether the current pollutant concentration is abnormal. Among them, the determination formula for the weight coefficient k is:

[0093]

[0094] where k0 is the basic weight (preset according to experience, such as k0 = 0.5); σ hist is the historical standard deviation; μ hist is the historical average.

[0095] The dynamic threshold of each pollutant is determined by the basic threshold × (1 + adjustment weight). For example: The adjustment weight is determined based on the comparison result of real-time data and historical data, which reflects the deviation degree and volatility of the current pollutant concentration relative to the historical level. This weight should have been determined in the previous steps; multiply the basic threshold of each pollutant by the result of (1 plus the adjustment weight) to obtain the dynamic threshold of the pollutant. This value is a threshold dynamically adjusted according to real-time situations and is used to more precisely judge whether the current pollutant concentration exceeds the standard or is abnormal. By introducing the adjustment weight, the dynamic threshold can better adapt to changes in actual situations and improve the sensitivity and accuracy of the system.

[0096] In the embodiments of the present invention, by introducing the adjustment weight, the dynamic threshold can be flexibly adjusted according to actual situations. This dynamic nature enables the system to better adapt to the exhaust gas emission characteristics in different treatment stages or environmental conditions, improving the adaptability and practicality of the system. The setting of the dynamic threshold is not only based on the historical average but also takes into account the historical standard deviation, which means that the system not only focuses on the average concentration of pollutants but also on their fluctuations. When the concentration or its volatility of pollutants in the exhaust gas increases, the dynamic threshold will increase accordingly, so as to be able to trigger early warning or control measures earlier and effectively prevent potential situations of exceeding the standard emissions. By updating the dynamic threshold in real time, the system can continuously monitor the exhaust gas emission status and respond in a timely manner when necessary. A reasonable setting of the dynamic threshold can avoid overly conservative or overly lenient control strategies, which helps to optimize the utilization of treatment resources while ensuring the exhaust gas treatment effect.

[0097] In a preferred embodiment of the present invention, a convolutional neural network model between exhaust gas parameters and the opening degree of the emission port is established using preprocessed historical data and preprocessed real-time data, including:

[0098] Determine the Pearson correlation coefficient between the exhaust gas parameters and the opening degree of the emission port. For example, collect a data set containing exhaust gas parameters (such as temperature, pressure, concentration, etc.) and the corresponding emission port opening degree data, and determine the Pearson correlation coefficient between each exhaust gas parameter and the emission port opening degree. This coefficient measures the linear correlation between the two variables.

[0099] Determine the corresponding exhaust gas parameters as features according to the magnitude of the Pearson correlation coefficient. For example, according to the magnitude of the correlation coefficient determined in the previous step, select those exhaust gas parameters with a relatively high correlation with the emission port opening degree as features. A threshold can be set to only select the exhaust gas parameters whose correlation coefficient exceeds this threshold as features to ensure the representativeness of the features.

[0100] Divide the features into a training set and a test set. For example, divide the selected feature data into a training set and a test set; usually, the training set accounts for 70%-80% of the data, and the test set accounts for 20%-30%. This ratio can be adjusted according to the data volume and actual situation.

[0101] Use the training set data to train the convolutional neural network model, and optimize the parameters of the convolutional neural network model through the backpropagation algorithm; during the training process, evaluate the stability and generalization ability of the convolutional neural network model through the K-fold cross-validation method. For example, use a deep learning framework to build a convolutional neural network (CNN) model, input the training set data into the CNN model for training, and adjust the weights and parameters of the model according to the prediction error of the model through the backpropagation algorithm. Iterate the training process multiple times until the model reaches the preset training accuracy or the number of iterations.

[0102] The hyperparameters of the convolutional neural network model are adjusted using the grid search method to obtain the final hyperparameter combination. The performance of the convolutional neural network model is evaluated using the test set to obtain the accuracy. For example: the training set data is further divided into K subsets, each subset is used as the validation set in turn, and the remaining subsets are used as the training set. The model is trained using K - 1 subsets in turn, and the performance of the model is verified using the remaining one subset. The performance metrics (such as accuracy, recall, etc.) of each verification are recorded, and the average performance metric of K verifications is determined. By analyzing the average performance metric and the performance fluctuations of each verification, the stability and generalization ability of the model are evaluated; a search range is set for the hyperparameters of the model (such as learning rate, batch size, number of iterations, etc.), and the grid search algorithm is used to try different hyperparameter combinations within the set hyperparameter range. For each set of hyperparameters, the performance of the model is evaluated using the cross - validation method, and according to the evaluation results, the hyperparameter combination with the best performance is selected as the final hyperparameters of the model.

[0103] The convolutional neural network model is optimized according to the accuracy to obtain the trained convolutional neural network model. For example: the test set data is input into the trained CNN model to obtain the prediction results of the model. By comparing the prediction results of the model with the actual results, performance metrics such as the accuracy, recall, and F1 - score of the model are determined; according to the performance evaluation results of the test set, the model is adjusted and optimized, such as adjusting the model architecture, adding data augmentation, etc. The optimized model is used to evaluate on the test set again to ensure that the model performance is improved.

[0104] In the embodiment of the present invention, by training the CNN model to learn the complex relationship between the exhaust gas parameters and the opening degree of the emission port, the automatic control of the opening degree of the emission port can be realized. This automatic control can quickly respond to the changes in the exhaust gas parameters, thereby more effectively managing the exhaust gas emissions. The CNN model can capture the non - linear relationship between the exhaust gas parameters and the opening degree of the emission port, which makes the adjustment of the emission port more accurate. Precise control of the opening degree of the emission port can ensure that the exhaust gas is emitted in an optimal way, improving the overall emission efficiency. By optimizing the opening degree of the emission port, unnecessary exhaust gas emissions can be reduced, thereby reducing environmental pollution. The CNN model is trained with a large amount of data and can better adapt to various complex working conditions and environmental changes. This stability enables the exhaust gas treatment system to maintain a robust performance in the face of emergencies. Precise emission port control can reduce energy waste, thereby reducing the operating cost. In addition, by reducing the excessive wear and repair frequency of the equipment, the maintenance cost can be further reduced. Using the CNN model, the future opening degree requirement of the emission port can be predicted based on the current exhaust gas parameters. This prediction ability helps to make adjustments in advance to ensure the stable operation of the exhaust gas treatment system.

[0105] In a preferred embodiment of the present invention, a prediction time domain is set, and based on the current state and control variables, a trained convolutional neural network model is used to determine exhaust gas parameters within the future time domain, including:

[0106] Determine the prediction duration, such as the next 1 hour, 4 hours, 1 day, etc.; divide the entire prediction time domain into smaller time steps, such as one time step every 5 minutes, 10 minutes, or half an hour. These time steps will serve as the basic units for model prediction. Obtain the exhaust gas parameters and the opening degree of the emission port at the current moment; the control variables are parameters that can be artificially adjusted within the prediction time domain, such as the opening degree of the emission port, the fuel supply amount, etc. These variables will be part of the model input; combine the current state data and the control variables into an input sequence, and this sequence should match the input format used during model training; send the input sequence into the trained CNN model for forward propagation determination to obtain the output of the exhaust gas parameters predicted by the model. The output of the model may be a sequence representing the predicted values of the exhaust gas parameters at each future time step. Parse these outputs into a practical and usable data format.

[0107] In a preferred embodiment of the present invention, based on the exhaust gas parameters within the future time domain, determine the pollutant concentration change rate, including:

[0108] Determine the size of the sliding window, that is, the number of data points included in the window. For example: determine the purpose of the analysis. For example, if it is to capture rapid changes in the short term, a smaller window is required; if it is to analyze long-term trends, a larger window may be needed; based on the requirements analysis, set a specific size of the sliding window, that is, determine the number of data points N included in the window. This N value should be a positive integer, which determines the number of data points used for linear fitting.

[0109] Create an array to store the data within the sliding window. The array is used to save the adjacent N data points, where N is the size of the sliding window. For example: create an empty array or list to store the data within the sliding window. The length of this array can be dynamically adjusted to adapt to the update of the sliding window. Although the length of the array will change dynamically, the maximum length should be set to the size N of the sliding window to ensure that too much data is not stored.

[0110] When new preprocessed future-time domain data arrives, perform the following operations: Add the new data point to the end of the sliding window array; If the array is full, i.e., it has reached the size N of the sliding window, remove the first data point in the array to keep the window size unchanged; Whenever the sliding window is updated, perform a linear fit using the N data points within the window. Specifically, it includes: Extract all the concentration data points within the sliding window, where the x-value represents the time point and the y-value represents the corresponding pollutant concentration; Determine the slope and intercept of the fitting line through the least squares formula. The slope represents the rate of change of the pollutant concentration. For example:

[0111] When new preprocessed future-time domain data arrives, receive this data; Add the newly received data point to the end of the sliding window array. This step ensures that the sliding window always contains the latest data; If the sliding window array is full (i.e., it has reached the set maximum length N), then before adding a new data point, the first data point in the array needs to be removed. This step ensures that the size of the sliding window always remains unchanged. Whenever the sliding window is updated, extract all the data points saved in the window. These data points will be used for the subsequent linear fitting process; Divide the extracted data points into two parts: the x-values (representing the time points) and the y-values (representing the corresponding pollutant concentrations), ensuring that the lengths of these two parts of data are the same and are arranged in chronological order; Perform a linear fit on the x-values and y-values using the least squares formula. The purpose of this step is to find a straight line such that the sum of the perpendicular distances from all data points to this line is minimized. The fitting result will obtain a slope and an intercept, where the slope represents the rate of change of the pollutant concentration. Store the determined slope and intercept.

[0112] Among them, in order to determine the rate of change of the pollutant concentration, a linear fit can be performed on the data within the sliding window using the least squares method. The goal of the least squares method is to find a straight line such that the sum of the perpendicular distances from all data points to this line is minimized. The slope and intercept of this line can be determined through the following formula:

[0113] Given a set of data points (x1, y1), (x2, y2), …, (x N , y N ), it is desired to find a straight line y = kx + b, where k is the slope and b is the intercept. Among them,

[0114]

[0115] Among them, N is the number of data points in the sliding window; x i is the i-th time point; y iis the pollutant concentration corresponding to the i-th time point; i represents the index value. The slope k represents the change rate of the pollutant concentration. If k > 0, it indicates that the pollutant concentration is increasing; if k < 0, it indicates that the pollutant concentration is decreasing.

[0116] In the embodiment of the present invention, by continuously adding new data points to the sliding window and removing old data points, this method can reflect the latest change trend of the pollutant concentration in real time; the size of the sliding window (N value) can be adjusted according to actual needs. A smaller window size can more sensitively capture short-term changes in concentration, while a larger window can provide a more stable long-term trend analysis. By linearly fitting the data points within the sliding window, this method can provide a quantitative index of the change rate of the pollutant concentration (i.e., the slope of the fitted straight line), which can more accurately reflect the change trend of the concentration than a simple concentration difference because it takes into account the time factor and is based on the statistical results of multiple data points.

[0117] In a preferred embodiment of the present invention, the change rate of the pollutant concentration is compared with a dynamic threshold to obtain a comparison result, including:

[0118] The change rate of the pollutant concentration is determined by using the sliding window and the least squares method in the previous steps, representing the average change speed of the pollutant concentration within a specific time window. This value can be positive (indicating an increase in concentration), negative (indicating a decrease in concentration), or zero (indicating that the concentration remains basically unchanged).

[0119] The dynamic threshold is used to determine whether the change rate of the pollutant concentration has reached a level that requires action, and the determined change rate of the pollutant concentration is compared with the set dynamic threshold. Specifically, if the change rate exceeds the threshold (whether it exceeds positively or negatively, depending on the setting method of the threshold), it means that the concentration of the pollutant is changing at an unacceptable speed, and actions may need to be taken to prevent potential environmental problems or compliance risks.

[0120] In a preferred embodiment of the present invention, according to the comparison result, the final control action at the current moment is generated, including:

[0121] If the change rate of the pollutant concentration is lower than the dynamic threshold, the current emission port state remains unchanged;

[0122] If the change rate of the pollutant concentration is equal to the dynamic threshold, a warning is issued and the emission port opening is prepared to be adjusted;

[0123] If the change rate of the pollutant concentration is higher than the dynamic threshold, the emission port opening is adjusted or the emission port is completely closed.

[0124] In an embodiment of the present invention, when the change rate of the pollutant concentration is lower than the set dynamic threshold, it means that the change trend of the current pollutant concentration is within an acceptable range and does not exceed the predetermined safety or environmental protection standards; in this case, the system will decide to maintain the current state of the emission port unchanged. That is to say, if the emission port is currently open, it will remain open; if it is closed, it will remain closed. This is because the current state has been considered safe and no additional adjustment is required; when the change rate of the pollutant concentration exactly equals the dynamic threshold, this can be regarded as a critical point, indicating that the change of the pollutant concentration is on the verge of being controllable and uncontrollable. At this time, the system will issue a warning to remind the operator to pay attention to the current environmental change and prepare to adjust the opening degree of the emission port. This warning is a preventive measure designed to let the operator make preparations in advance in case the pollutant concentration further increases and exceeds the safe range.

[0125] If the change rate of the pollutant concentration exceeds the dynamic threshold, this indicates that the concentration of the pollutant is rising at too fast a rate and may have already or will soon have an adverse impact on the environment. In this case, the system will take more active control measures. Specifically, it may automatically adjust the opening degree of the emission port to reduce the pollutant emission amount in order to reduce the growth rate of the pollutant concentration. In extreme cases, if adjusting the opening degree still cannot effectively control the increase of the pollutant concentration, the system may even choose to completely close the emission port to cut off the pollutant emission source. This step is the last link in the entire pollutant control process, which ensures that when the pollutant concentration shows an adverse change, the system can respond quickly and accurately, thereby minimizing the impact on the environment and human health. Through this dynamic adjustment mechanism, the system can meet the environmental protection and safety requirements while maintaining normal operation.

[0126] When specifically applied, a control instruction generated based on the comparison between the change rate of the pollutant concentration and the dynamic threshold is received. This instruction clearly indicates whether the exhaust gas emission port should be opened, closed or the opening degree adjusted. Before performing any action, the system will verify whether the received instruction is valid and accurate to prevent incorrect actions caused by misoperation or system failures. If the control instruction requires opening the exhaust gas emission port, the system will send a signal to the emission port control device to trigger the opening mechanism, which usually involves activating a motor or a pneumatic device to change the emission port from the closed state to the open state.

[0127] If the control instruction requires closing the exhaust gas emission port, the system will close the emission port through the corresponding control signal, which also involves the operation of a motor or a pneumatic device to ensure that the emission port is completely sealed to prevent exhaust gas leakage; when the control instruction requires adjusting the opening degree of the emission port, the system will accurately adjust the opening degree of the emission port according to the opening degree value specified in the instruction through the control device.

[0128] As shown Figure 2 in the figure, an embodiment of the present invention further provides a remote management method for exhaust gas emissions during the treatment of kitchen waste, including:

[0129] Real-time monitoring of the exhaust gas parameters at the exhaust gas outlet of the kitchen waste treatment facility;

[0130] Preprocessing the exhaust gas parameters to obtain preprocessed real-time data; determining an adjustment weight based on the preprocessed historical data and the preprocessed real-time data; determining a dynamic threshold based on the adjustment weight;

[0131] Using the preprocessed historical data and the preprocessed real-time data to establish a convolutional neural network model between the exhaust gas parameters and the opening degree of the exhaust gas outlet; setting a prediction time domain, and based on the current state and control variables, using the trained convolutional neural network model to determine the exhaust gas parameters within the future time domain;

[0132] Determining the pollutant concentration change rate based on the exhaust gas parameters within the future time domain; comparing the pollutant concentration change rate with the dynamic threshold to obtain a comparison result;

[0133] Generating a final control action at the current moment according to the comparison result; performing an opening or closing operation of the exhaust gas outlet according to the final control action.

[0134] It should be noted that this system corresponds to the above method. All implementation manners in the above method embodiments are applicable to this embodiment and can also achieve the same technical effects.

[0135] An embodiment of the present invention further provides a computing device, including: a processor and a memory storing a computer program. When the computer program is run by the processor, it executes the method described above. All implementation manners in the above method embodiments are applicable to this embodiment and can also achieve the same technical effects.

[0136] An embodiment of the present invention further provides a computer-readable storage medium storing instructions. When the instructions are run on a computer, the computer is caused to execute the method described above. All implementation manners in the above method embodiments are applicable to this embodiment and can also achieve the same technical effects.

Claims

1. A remote management system for waste gas emissions during the treatment of kitchen waste, characterized in that, Including: A sensor group, which is set at the exhaust gas outlet of the kitchen waste treatment facility and is used to monitor exhaust gas parameters in real time; A processing module, which is used to preprocess the exhaust gas parameters to obtain preprocessed real-time data; Determine the adjustment weight according to the preprocessed historical data and preprocessed real-time data; Determine the dynamic threshold according to the adjustment weight; A prediction module, which is used to establish a convolutional neural network model between the exhaust gas parameters and the opening degree of the exhaust outlet by using the preprocessed historical data and preprocessed real-time data; set the prediction time domain, and based on the current state and control variables, use the trained convolutional neural network model to determine the exhaust gas parameters in the future time domain; A comparison module, which is used to determine the pollutant concentration change rate according to the exhaust gas parameters in the future time domain; Compare the pollutant concentration change rate with the dynamic threshold to obtain a comparison result; A control module, which is used to generate the final control action at the current moment according to the comparison result; and perform the opening or closing operation of the exhaust gas outlet according to the final control action.

2. The remote management system for waste gas emissions during the treatment of kitchen waste according to claim 1, characterized in that, Determine the adjustment weight according to the preprocessed historical data and preprocessed real-time data, including: Extract all pollutant concentration values from the real-time data window, and determine the real-time average value of all pollutant concentration values; based on the pollutant concentration values within the real-time data window, determine the real-time standard deviation; Determine the historical concentration values of all pollutants after preprocessing in the past period of time, and determine the historical average value and historical standard deviation of all pollutants according to the historical concentration values; Determine the absolute difference between the real-time average value and the historical average value; determine the standardized deviation value according to the absolute difference and the historical standard deviation; Determine the real-time volatility according to the real-time standard deviation and the historical standard deviation; Determine the mean deviation term according to the mean deviation term and the corresponding adjustment coefficient; determine the volatility term according to the real-time volatility and the corresponding adjustment coefficient; Fuse the mean deviation term and the volatility term to obtain the adjustment weight.

3. The remote management system for exhaust gas emission during the treatment of kitchen waste according to claim 2, characterized in that, Determine the dynamic threshold according to the adjustment weight, including: Obtain the concentration data of each pollutant after preprocessing in the past period of time; Determine the historical average value and historical standard deviation of each pollutant according to the concentration data of each pollutant after preprocessing; Fuse the product of the historical average value, the weight coefficient and the historical standard deviation to determine the basic threshold of each pollutant; Determine the dynamic threshold of each pollutant according to the basic threshold.

4. A remote management system for exhaust gas emissions during the treatment of kitchen waste according to claim 3, characterized in that, Establish a convolutional neural network model between the exhaust gas parameters and the opening degree of the exhaust outlet by using the preprocessed historical data and preprocessed real-time data, including: Determine the Pearson correlation coefficient between the exhaust gas parameters and the opening degree of the exhaust outlet; Determine the corresponding exhaust gas parameter as a feature according to the magnitude of the Pearson correlation coefficient; Divide the features into a training set and a test set; Use the training set data to train the convolutional neural network model, and optimize the parameters of the convolutional neural network model through the backpropagation algorithm; during the training process, evaluate the stability and generalization ability of the convolutional neural network model through the K-fold cross-validation method; Use the grid search method to adjust the hyperparameters of the convolutional neural network model to obtain the final hyperparameter combination, and use the test set to evaluate the performance of the convolutional neural network model to obtain the accuracy rate; Optimize the convolutional neural network model according to the accuracy rate to obtain the trained convolutional neural network model.

5. A remote management system for exhaust gas emissions during the treatment of kitchen waste according to claim 4, characterized in that Determine the pollutant concentration change rate according to the waste gas parameters in the future time domain, including: Determine the size of the sliding window, that is, the number of data points included in the window; Create an array to store the data within the sliding window. The array is used to save the adjacent N data points, where N is the size of the sliding window; When new preprocessed future time domain data arrives, perform the following operations: Add the new data point to the end of the sliding window array; if the array is full, that is, it has reached the size N of the sliding window, remove the first data point in the array to keep the window size unchanged; Whenever the sliding window is updated, perform linear fitting using the N data points within the window, specifically including: Extract all concentration data points within the sliding window. The x value represents the time point, and the y value represents the corresponding pollutant concentration; determine the slope and intercept of the fitting line through the least squares formula, and the slope represents the change rate of the pollutant concentration.

6. The remote management system for waste gas emission during the treatment of kitchen waste according to claim 5, characterized in that, The waste gas parameters include H2S concentration, NH3 concentration, VOCs concentration, waste gas temperature, waste gas humidity, and waste gas emission flow rate.

7. A remote management system for exhaust gas emissions during the treatment of kitchen waste according to claim 6, characterized in that, Generate the final control action at the current moment according to the comparison result, including: If the pollutant concentration change rate is lower than the dynamic threshold, keep the current emission port state unchanged; If the pollutant concentration change rate is equal to the dynamic threshold, issue a warning and prepare to adjust the opening of the emission port; If the pollutant concentration change rate is higher than the dynamic threshold, adjust the opening of the emission port or completely close the emission port.

8. A remote management method for exhaust gas emissions during the treatment of kitchen waste, characterized in that, The method is used to execute the system described in any one of claims 1 to 7, including: Real-time monitor the waste gas parameters of the waste gas emission port of the food waste treatment facility; Preprocess the waste gas parameters to obtain preprocessed real-time data; determine the adjustment weight according to the preprocessed historical data and preprocessed real-time data; determine the dynamic threshold according to the adjustment weight; Use the preprocessed historical data and preprocessed real-time data to establish a convolutional neural network model between the waste gas parameters and the opening of the emission port; set the prediction time domain, and based on the current state and control variables, use the trained convolutional neural network model to determine the waste gas parameters in the future time domain; Determine the pollutant concentration change rate according to the waste gas parameters in the future time domain; compare the pollutant concentration change rate with the dynamic threshold to obtain the comparison result; Generate the final control action at the current moment according to the comparison result; perform the opening or closing operation of the waste gas emission port according to the final control action.

9. A computing device, characterized in that, Including: One or more processors; A storage device for storing one or more programs, which when executed by the one or more processors, enable the one or more processors to implement the method described in claim 8.

10. A computer-readable storage medium, characterized in that, A program is stored in the computer-readable storage medium, and when the program is executed by the processor, it implements the method described in claim 8.

Images