Municipal sewage system odor concentration calculation method based on machine learning

Abstract

The invention provides a municipal sewage system odor concentration calculation method based on machine learning, and belongs to the technical field of data processing, and the method specifically comprises the steps: carrying out the data processing of the monitoring data of an odor monitoring target, building a machine learning model, and carrying out the calculation of the odor concentration of a municipal sewage system through the machine learning model and the monitoring data; determining a calculation result of the odor concentration of the odor concentration monitoring point, and determining an updating scheme of the odor monitoring target in the machine learning model on the basis of a matching condition of the odor monitoring target and the monitoring data of the concentration of the odor-causing substance of the odor concentration monitoring point, and the calculation result of the odor concentration of the odor concentration monitoring point is determined based on the machine learning model obtained by the updating scheme, so that the monitoring result of the odor concentration is more accurate.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, and in particular relates to a method for calculating odor concentration in municipal sewage systems based on machine learning. Background Technology

[0002] The odor generated by municipal wastewater systems is a mixture of gases containing various odor-causing substances. Its main components include hydrogen sulfide, ammonia, and various volatile organic compounds (VOCs). Currently, these substances can be accurately detected in laboratories and monitored online using electrochemical and optical methods, achieving good accuracy within a certain range. However, odor concentration is a direct indicator of the impact of ambient air odors on human senses, and this indicator can only be determined through manual olfactory identification. Many current online monitoring schemes attempt to calculate odor concentration online by using linear regression models to fit the detected odor component concentrations.

[0003] To solve the above-mentioned technical problems, the existing technical solution CN202310175765.8 "Method, device, electronic equipment and storage medium for determining the odor concentration of livestock and poultry houses" inputs the concentration values ​​of each target gas and environmental data obtained in real time into the odor concentration monitoring model, which can quickly obtain the target odor concentration in the target livestock and poultry house, and can quickly realize the detection of odor concentration in the target livestock and poultry house, which facilitates the real-time and continuous measurement of odor concentration.

[0004] However, unlike poultry houses, the odor components of municipal sewage systems are complex, and the mechanisms by which these components interact are not yet clear. Therefore, if a fixed odor concentration monitoring model is used, the accuracy of the overall odor concentration monitoring results may not meet the requirements.

[0005] To address the aforementioned technical problems, this application provides a machine learning-based method for calculating odor concentration in municipal wastewater systems. Summary of the Invention

[0006] To achieve the objectives of this invention, the following technical solution is adopted: Specifically, this application provides a machine learning-based method for calculating odor concentration in municipal wastewater systems, which includes: S1 uses the monitoring data of odor-causing substances at odor concentration monitoring points in the municipal system to determine the monitoring period when odor-causing substances are present at the odor concentration monitoring points. The number of odor-causing substances to be selected is determined by the changes in odor-causing substances at each odor concentration monitoring point during the monitoring period. The odor monitoring target is determined by the number of selected substances and the monitoring data at different odor concentration monitoring points. S2 processes the monitoring data of the odor monitoring target and constructs a machine learning model. Using the machine learning model and the monitoring data, the calculation result of the odor concentration at the odor concentration monitoring point is determined. S3 determines the update scheme for the odor monitoring target in the machine learning model based on the matching data of the odor monitoring target and the concentration of odor-causing substances at the odor concentration monitoring point, and determines the calculation result of the odor concentration at the odor concentration monitoring point based on the machine learning model obtained by the update scheme.

[0007] The beneficial effects of this invention are as follows: Using machine learning models and monitoring data, the calculation results of odor concentration at odor concentration monitoring points are determined, enabling real-time calculation of this sensory indicator. This is the core module for online odor concentration monitoring. Compared with multiple linear regression models and multinomial regression models for calculating odor concentration, machine learning has stronger resistance to data interference, higher accuracy, and the average absolute error can be reduced to 0.4, achieving a high accuracy rate in judging whether odor exceeds the standard. The overall accuracy rate for judging whether the odor exceeds 100 reaches 84%.

[0008] Based on the matching of odor monitoring targets and the concentrations of odor-causing substances at odor concentration monitoring points, an update scheme for the odor monitoring targets in the machine learning model is determined. This enables the determination of the update scheme for the odor monitoring targets in the machine learning model based on the monitoring matching of odor monitoring targets and the concentrations of odor-causing substances at odor concentration monitoring points. In other words, the update processing scheme is determined based on the severity of the monitoring deviation of odor-causing substances. This reduces the difficulty of updating the machine learning model while also meeting the needs of monitoring, analyzing, and processing the odor concentrations of odor-causing substances.

[0009] Furthermore, the monitoring data of the concentration of the odor-causing substance is determined based on the monitoring results of the sensors at the odor concentration monitoring points.

[0010] Furthermore, the monitoring period during which odor-causing substances are present at the odor concentration monitoring point is the period during which specific substances are present at the odor concentration monitoring point.

[0011] It should be noted that the specific substances mentioned include two inorganic substances, ammonia and hydrogen sulfide; eight sulfur-containing organic substances, such as methanethiol and dimethyl sulfide; four nitrogen-containing organic substances, such as methylamine; ten organic substances, including alcohols, ethers, aldehydes, ketones, and esters; eighteen organic substances, including benzene and halogenated benzenes; and thirty-six organic substances, including hydrocarbons and halogenated hydrocarbons.

[0012] Furthermore, the method for determining the selected quantity of the odor-causing substance is as follows: Based on the monitoring data of odor-causing substances at the odor concentration monitoring points during various monitoring periods, the odor-causing substances present at the odor concentration monitoring points are determined. Based on the changes in odor-causing substances at the odor monitoring points during various monitoring periods, monitoring periods with the same odor-causing substances are grouped into the same monitoring period combination; The number of odor-causing substances selected is determined based on the distribution data of odor-causing substances present at each odor concentration monitoring point and the combination of monitoring time periods.

[0013] Furthermore, the construction of machine learning models is carried out, specifically including: The original data is constructed using monitoring data including the odor monitoring indicators as input and odor concentration as output. For odor monitoring targets that are not present in the original data, replace them with 0. Divide the concentration of each component by the odor threshold of that component to obtain the dilution factor. Then take the common logarithm of the dilution factor (base 10). Replace illegal and negative values ​​with 0 to obtain the amplified original data. Using this method, components with a threshold dilution factor of less than 1 are screened out. For odor concentration indicators, the common logarithm is directly taken, and the range of the processed data distribution is reduced from tens to thousands to a range of 1 to 4.

[0014] Based on the amplified original data, the machine learning method is trained to obtain the trained machine learning model.

[0015] Furthermore, the method for determining the calculation result of the odor concentration at the odor concentration monitoring point is as follows: Based on the aforementioned machine learning model and monitoring data, the monitoring data of the odor monitoring target is used as the input to the machine learning model. The output of the machine learning model is used as the calculation result of the odor concentration.

[0016] It is understood that the machine learning model is constructed using one or more of the following: a BP neural network, an LSTM neural network, and an RNN neural network.

[0017] Furthermore, after obtaining the calculated odor concentration, the output of a binary classification algorithm, including the XGBoost algorithm, can be used to determine whether the odor concentration exceeds the standard.

[0018] Furthermore, the method for determining the update scheme of the odor monitoring target in the machine learning model is as follows: Based on the matching data of the odor monitoring target and the concentration of odor-causing substances at the odor concentration monitoring point, the time period in which the odor-causing substances at the odor concentration monitoring point do not exist is determined and is taken as the period of poor calculation. Based on the data from the poor calculation periods of each odor concentration monitoring point, odor-causing substances that are not odor monitoring targets in different poor calculation periods are identified and regarded as monitoring deviation substances. Based on the data from the poorly calculated periods at each odor concentration monitoring point and the data on the monitoring deviation substances during those periods, an update scheme for the odor monitoring target in the machine learning model is determined.

[0019] Other features and advantages will be set forth in the following description, and the objects and other advantages of the invention are realized and obtained through the structures particularly pointed out in the description and the drawings.

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0021] The above and other features and advantages of the present invention will become more apparent from a detailed description of exemplary embodiments thereof with reference to the accompanying drawings; Figure 1 This is a flowchart of a machine learning-based method for calculating odor concentration in municipal wastewater systems. Figure 2 This is a flowchart illustrating the method for determining the selected quantity of odor-causing substances; Figure 3 This is a flowchart illustrating the method for determining odor monitoring targets. Detailed Implementation

[0022] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0023] Example 1 like Figure 1 As shown, this application provides a machine learning-based method for calculating odor concentration in municipal wastewater systems, specifically including: S1 uses the monitoring data of odor-causing substances at odor concentration monitoring points in the municipal system to determine the monitoring period when odor-causing substances are present at the odor concentration monitoring points. The number of odor-causing substances to be selected is determined by the changes in odor-causing substances at each odor concentration monitoring point during the monitoring period. The odor monitoring target is determined by the number of selected substances and the monitoring data at different odor concentration monitoring points. Furthermore, the monitoring data of the concentration of the odor-causing substance is determined based on the monitoring results of the sensors at the odor concentration monitoring points.

[0024] Furthermore, the monitoring period during which odor-causing substances are present at the odor concentration monitoring point is the period during which specific substances are present at the odor concentration monitoring point.

[0025] It should be noted that the specific substances mentioned include two inorganic substances, ammonia and hydrogen sulfide; eight sulfur-containing organic substances, such as methanethiol and dimethyl sulfide; four nitrogen-containing organic substances, such as methylamine; ten organic substances, including alcohols, ethers, aldehydes, ketones, and esters; eighteen organic substances, including benzene and halogenated benzenes; and thirty-six organic substances, including hydrocarbons and halogenated hydrocarbons.

[0026] Specifically, such as Figure 2 As shown, the method for determining the selected quantity of the odor-causing substance is as follows: Based on the monitoring data of odor-causing substances at the odor concentration monitoring points during various monitoring periods, the odor-causing substances present at the odor concentration monitoring points are determined. Based on the changes in odor-causing substances at the odor monitoring points during various monitoring periods, monitoring periods with the same odor-causing substances are grouped into the same monitoring period combination; The number of odor-causing substances selected is determined based on the distribution data of odor-causing substances present at each odor concentration monitoring point and the combination of monitoring time periods.

[0027] It is understandable that the selection quantity of the odor-causing substances is determined based on the distribution data of the odor-causing substances present at each odor concentration monitoring point and the combination of monitoring time periods, specifically including: Based on the odor-causing substances present at each odor concentration monitoring point, determine the total amount of odor-causing substances present at all odor concentration monitoring points; When the total number of odor-causing substances meets the requirement, i.e., it is not greater than the preset number, it is less difficult to construct a machine learning model based on all odor-causing substances. Therefore, the total number of odor-causing substances is used as the number of odor-causing substances selected.

[0028] In one possible embodiment, when the total number of odor-causing substances is no more than 10, the total number of odor-causing substances is used as the number of odor-causing substances selected.

[0029] It should also be noted that when the total amount of the odor-causing substances does not meet the requirements, the number of monitoring time period combinations for each odor concentration monitoring point is determined based on the distribution data of the monitoring time period combinations in each odor concentration monitoring point. When the average number of monitoring time period combinations at each odor concentration monitoring point exceeds the preset combination number threshold, it indicates that the odor-causing substances at different odor concentration monitoring points are changing significantly. Therefore, the machine learning model is likely to need to be updated later. Based on this, we should select as few odor monitoring targets as possible to ensure that the difficulty of retraining is sufficient.

[0030] In one possible embodiment, when the average number of monitoring time period combinations of each odor concentration monitoring point is greater than a preset combination number threshold, the number of odor monitoring targets is determined by the proportion of odor concentration monitoring points with a number of monitoring time period combinations greater than the preset combination number threshold. Specifically, the correction number is determined by multiplying the proportion of odor concentration monitoring points with a number of monitoring time period combinations greater than the preset combination number threshold among all odor concentration monitoring points by the baseline number. The difference between the baseline number and the correction number is used to determine the number of odor-causing substances selected.

[0031] In one possible specific embodiment, the product of the proportion of odor concentration monitoring points with more than 20 monitoring time period combinations and 8 is used as a correction quantity, and the difference between 8 and the correction quantity is rounded down to determine the number of odor-causing substances selected.

[0032] It should also be noted that when the average number of monitoring time period combinations at each odor concentration monitoring point is not greater than the preset combination number threshold, the number of odor-causing substances selected is determined based on the baseline number.

[0033] After sampling at odor concentration monitoring points, the concentrations of typical odor-causing substances were measured, including sulfur-containing organic compounds such as ammonia and hydrogen sulfide, methanethiol and dimethyl sulfide, nitrogen-containing organic compounds such as methylamine, alcohols, ethers, aldehydes, ketones, esters, benzene and halogenated benzenes, and hydrocarbons and halogenated hydrocarbons. Based on the odor thresholds of each substance, the substances contributing the most to odor were calculated and ranked. The concentrations of 78 typical odor-causing substances were measured, including 2 inorganic substances (hydrogen sulfide), 8 sulfur-containing organic compounds such as methanethiol and dimethyl sulfide, 4 nitrogen-containing organic compounds such as methylamine, 10 organic compounds (alcohols, ethers, aldehydes, ketones, esters), 18 organic compounds (benzene and halogenated benzenes), and 36 organic compounds (hydrocarbons and halogenated hydrocarbons). The concentrations of each substance (in μg / m³) were obtained, and the odor thresholds of each substance were then looked up. The odor thresholds of some substances (in ppm) are listed in the table below: Table 1. Odor threshold of different types of substances Substance Name Odor threshold (ppm) 3-Methylindole 0.0000056 Methanethiol 0.000067 hydrogen sulfide 0.0012 Trimethylamine 0.0009 Methyl sulfide 0.0020 Dimethyl disulfide 0.011 ammonia 0.3 Methylamine 0.035 After converting the concentration unit to ppm, calculate the threshold dilution factor for each substance using the following formula.

[0034] The results were then sorted from largest to smallest according to the threshold dilution factor, and some of the results are listed below: Table 2 Threshold dilution factors for different types of substances Substance Name Threshold dilution factor (dimensionless) Methanethiol 12904 hydrogen sulfide 9229 3-Methylindole 5328 Trimethylamine 21 ammonia 9 Methyl sulfide 8 Dimethyl disulfide 5 Therefore, the table above shows that methanethiol, hydrogen sulfide, and 3-methylindole can be the main components of odor monitoring targets.

[0035] Specifically, such as Figure 3 As shown, the method for determining the odor monitoring target is as follows: Odor concentration monitoring points with a number of monitoring time period combinations exceeding a preset combination number threshold are designated as fluctuation monitoring points. Based on the monitoring time period data of the odor-causing locations at different fluctuation monitoring points, determine the total number of monitoring time periods in which the odor-causing substances are present; The odor monitoring targets among the odor-causing substances are determined by the total number of monitoring periods in which the odor-causing substances are present and the number of selected periods.

[0036] In one possible embodiment, the odor-causing substance with the largest total number of monitoring periods in the fluctuation monitoring node is selected as the odor monitoring target. That is, if the number of odor-causing substances that is greater than the total number of monitoring periods in the fluctuation monitoring node is less than the selected number, then the odor-causing substance is determined as the odor monitoring target.

[0037] Based on the monitoring data from the odor concentration monitoring points, the odor concentration monitoring points where the odor-causing substances are present are determined; The ratio of the number of monitoring time period combinations of odor concentration monitoring points containing the odor-causing substance to a preset combination number threshold is used as the fluctuation weight value of the odor concentration monitoring points containing the odor-causing substance. The weighted sum of the odor-causing substances is determined by summing the fluctuation weight values ​​of the odor concentration monitoring points where the odor-causing substances are present. Based on the weighted sum and the selected quantity, the odor monitoring targets among the odor-causing substances are determined.

[0038] S2 processes the monitoring data of the odor monitoring target and constructs a machine learning model. Using the machine learning model and the monitoring data, the calculation result of the odor concentration at the odor concentration monitoring point is determined. Furthermore, the construction of machine learning models is carried out, specifically including: The original data is constructed using monitoring data including the odor monitoring indicators as input and odor concentration as output. For odor monitoring targets that are not present in the original data, replace them with 0. Divide the concentration of each component by the odor threshold of that component to obtain the dilution factor. Then take the common logarithm of the dilution factor (base 10). Replace illegal and negative values ​​with 0 to obtain the amplified original data. First, the concentration of undetected components was replaced with 0. The threshold dilution factor (dimensionless) for each component was calculated using the formula for threshold dilution factors in odor component analysis. Then, values ​​with threshold dilution factors less than or equal to 1 were replaced with 1 to prevent negative numbers from occurring during logarithmic calculations. Since theoretically, a human would not smell the substance when the threshold dilution factor is less than 1, this data simplification does not lose the information about the degree of odor represented by the data. Finally, the threshold dilution factors and odor concentrations were calculated using the common logarithm (base 10). The processed data narrowed the data distribution range from tens to thousands to 1 to 4. Partial data processing steps are shown below. The original data, i.e., the concentrations of each odor component (in μg / m³) and the measured values ​​of odor concentration, are shown in the table below: Table 3. Concentrations of each odor component (in μg / m3) and measured values ​​of odor concentration.

[0039] Next, the concentrations of undetected components were replaced with 0, the threshold dilution factor for each component was calculated, and values ​​with a threshold dilution factor less than or equal to 1 were replaced with 1. The results are shown in the table below: Table 4 Component Concentration Table

[0040] Finally, the threshold dilution factor and odor concentration were calculated using the commonly used logarithm (base 10), and the results are shown in the table below: Table 5 shows the commonly used logarithms for threshold dilution factors and odor concentrations.

[0041] The original data ranged from tens to thousands, but after data processing, the data ranged from 0 to 4, which is beneficial for improving the accuracy of machine learning.

[0042] Using this method, components with a threshold dilution factor of less than 1 are screened out. For odor concentration indicators, the common logarithm is directly taken, and the range of the processed data distribution is reduced from tens to thousands to a range of 1 to 4.

[0043] Based on the amplified original data, the machine learning method is trained to obtain the trained machine learning model.

[0044] Furthermore, the method for determining the calculation result of the odor concentration at the odor concentration monitoring point is as follows: Based on the aforementioned machine learning model and monitoring data, the monitoring data of the odor monitoring target is used as the input to the machine learning model. The output of the machine learning model is used as the calculation result of the odor concentration.

[0045] It is understood that the machine learning model is constructed using one or more of the following: a BP neural network, an LSTM neural network, and an RNN neural network.

[0046] Furthermore, after obtaining the calculated odor concentration, the output of a binary classification algorithm, including the XGBoost algorithm, can be used to determine whether the odor concentration exceeds the standard.

[0047] S3 determines the update scheme for the odor monitoring target in the machine learning model based on the matching data of the odor monitoring target and the concentration of odor-causing substances at the odor concentration monitoring point, and determines the calculation result of the odor concentration at the odor concentration monitoring point based on the machine learning model obtained by the update scheme.

[0048] Furthermore, the method for determining the update scheme of the odor monitoring target in the machine learning model is as follows: Based on the matching data of the odor monitoring target and the concentration of odor-causing substances at the odor concentration monitoring point, the time period in which the odor-causing substances at the odor concentration monitoring point do not exist is determined and is taken as the period of poor calculation. Based on the data from the poor calculation periods of each odor concentration monitoring point, odor-causing substances that are not odor monitoring targets in different poor calculation periods are identified and regarded as monitoring deviation substances. Based on the data from the poorly calculated periods at each odor concentration monitoring point and the data on the monitoring deviation substances during those periods, an update scheme for the odor monitoring target in the machine learning model is determined.

[0049] It is understandable that when the number of poorly calculated periods at each odor concentration monitoring point meets the requirements, since the number of poorly calculated periods is small, it can be directly determined that there is no need to update the odor monitoring target in the machine learning model.

[0050] In one possible embodiment, if the number of poorly calculated periods at each odor concentration monitoring point accounts for less than 0.05 of the total duration of the periods when odorous substances are present at the odor concentration monitoring point, then it is determined that no update processing of the odor monitoring target in the machine learning model is required.

[0051] Furthermore, when the number of poorly calculated periods at odor concentration monitoring points does not meet the requirements, it is necessary to further determine the monitoring deviation substance data in the poorly calculated periods. When there are no monitoring deviation substances in the poorly calculated periods that meet the requirements, that is, the distribution of monitoring deviation substances in different poorly calculated periods is relatively discrete, it is determined that there is no need to update the odor monitoring target in the machine learning model.

[0052] In one possible specific embodiment, if the proportion of the poorly calculated period of the monitored deviation substance in all poorly calculated periods is not less than 0.1, then the number of poorly calculated periods of the monitored deviation substance is determined to meet the requirements.

[0053] Additionally, it can be understood that when there are monitoring deviation substances in the period of poor calculation that meet the requirements, the number of monitoring deviation substances in the period of poor calculation that meet the requirements is obtained. When the number of monitoring deviation substances in the period of poor calculation that meet the requirements is greater than the preset threshold for the number of odor-causing substances, it is determined that the machine learning model should be updated, that is, all odor monitoring targets should be updated. Specifically, the odor monitoring targets are determined based on the odor-causing substance data in different monitoring periods.

[0054] In one possible specific embodiment, if the number of monitoring deviation substances that meet the requirements during the poorly calculated time period is more than three, then it is determined that the number of monitoring deviation substances that meet the requirements during the poorly calculated time period is greater than a preset threshold for the number of odor-causing substances. At this time, if there is a monitoring time period of the original odor monitoring target that does not meet the requirements, then the original odor monitoring target is deleted. The deleted odor monitoring target and the number of monitoring deviation substances that meet the requirements during the poorly calculated time period are used as input quantities to retrain the machine learning model and then update it.

[0055] In one possible specific embodiment, if the number of odor monitoring points for the original odor monitoring target during the monitoring period is less than 5, then it is determined that the monitoring period of the original odor monitoring target does not meet the requirements.

[0056] Furthermore, when the number of monitoring deviation substances that meet the requirements during periods of poor calculation is not greater than the preset threshold for the number of odor-causing substances, the monitoring deviation substances that meet the requirements during periods of poor calculation will be used as new odor monitoring targets added to the machine learning model. The newly added odor monitoring targets and the original odor monitoring targets will be used as inputs to retrain the machine learning model and then update it.

[0057] Example 2 Furthermore, the method for determining the odor monitoring target is as follows: Based on the monitoring data from the odor concentration monitoring points, the odor concentration monitoring points where the odor-causing substances are present are determined; The ratio of the number of monitoring time period combinations of odor concentration monitoring points containing the odor-causing substance to a preset combination number threshold is used as the fluctuation weight value of the odor concentration monitoring points containing the odor-causing substance. The weighted sum of the odor-causing substances is determined by summing the fluctuation weight values ​​of the odor concentration monitoring points where the odor-causing substances are present. Based on the weighted sum and the selected quantity, the odor monitoring targets among the odor-causing substances are determined.

[0058] Specifically, the odor-causing substances with the largest weight and the largest number of selected substances are used as odor monitoring targets. That is, if the number of odor-causing substances with a greater weight than the sum of the weights of the odor-causing substances is less than the selected number, then the odor-causing substances are determined as odor monitoring targets.

[0059] Example 3 Optionally, the method for determining the selected quantity of the odor-causing substance is as follows: Based on the monitoring data of odor-causing substances at the odor concentration monitoring points during various monitoring periods, the odor-causing substances present at the odor concentration monitoring points are determined. Based on the changes in odor-causing substances at the odor monitoring points during various monitoring periods, monitoring periods with the same odor-causing substances are grouped into the same monitoring period combination; Based on the number of monitoring time period combinations for different odor concentration monitoring points, the average number of monitoring time period combinations for different odor concentration monitoring points is determined. Based on the average number of monitoring time period combinations for different odor concentration monitoring points, the number of odor-causing substances selected is determined.

[0060] In one possible specific embodiment, a correction ratio is determined by the ratio of the average number of monitoring time period combinations of different odor concentration monitoring points to a preset combination number threshold. The product of the correction ratio and a preset ratio factor is used as a correction factor. The value obtained by subtracting the product of the benchmark value and the correction factor from the benchmark value is used as the number of odor-causing substances selected.

[0061] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0062] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0063] The above description is merely one or more embodiments of this specification and is not intended to limit this specification. Various modifications and variations can be made to the one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of this specification.

Claims

1. A method for calculating odor concentration in a municipal wastewater system based on machine learning, characterized in that, Specifically, it includes: Based on the monitoring data of odor-causing substances at odor concentration monitoring points in the municipal system, the monitoring period containing odor-causing substances at the odor concentration monitoring points is determined. The number of odor-causing substances to be selected is determined based on the changes in odor-causing substances at each odor concentration monitoring point during the monitoring period. The odor monitoring target is determined based on the number of selected substances and the monitoring data at different odor concentration monitoring points. After processing the monitoring data of the odor monitoring target, a machine learning model is constructed, and the calculation result of the odor concentration at the odor concentration monitoring point is determined using the machine learning model and the monitoring data. Based on the matching data of the odor monitoring target and the concentration of odor-causing substances at the odor concentration monitoring point, an update scheme for the odor monitoring target in the machine learning model is determined, and the calculation result of the odor concentration at the odor concentration monitoring point is determined based on the machine learning model obtained from the update scheme.

2. The method for calculating odor concentration in municipal wastewater systems based on machine learning as described in claim 1, characterized in that, The concentration of the odor-causing substance is determined based on the monitoring results of the sensors at the odor concentration monitoring points.

3. The method for calculating odor concentration in municipal wastewater systems based on machine learning as described in claim 1, characterized in that, The monitoring period during which odor-causing substances are present at the odor concentration monitoring point is the period during which specific substances are present at the odor concentration monitoring point.

4. The method for calculating odor concentration in municipal wastewater systems based on machine learning as described in claim 1, characterized in that, The method for determining the quantity of the odor-causing substance is as follows: Based on the monitoring data of odor-causing substances at the odor concentration monitoring points during various monitoring periods, the odor-causing substances present at the odor concentration monitoring points are determined. Based on the changes in odor-causing substances at the odor monitoring points during various monitoring periods, monitoring periods with the same odor-causing substances are grouped into the same monitoring period combination; The number of odor-causing substances selected is determined based on the distribution data of odor-causing substances present at each odor concentration monitoring point and the combination of monitoring time periods.

5. The method for calculating odor concentration in municipal wastewater systems based on machine learning as described in claim 4, characterized in that, Based on the distribution data of odor-causing substances at each odor concentration monitoring point and the combination of monitoring time periods, the number of odor-causing substances selected is determined, specifically including: Based on the odor-causing substances present at each odor concentration monitoring point, determine the total amount of odor-causing substances present at all odor concentration monitoring points; When the total number of odor-causing substances meets the requirement, i.e., it is not greater than the preset number, it is less difficult to construct a machine learning model based on all odor-causing substances. Therefore, the total number of odor-causing substances is used as the number of odor-causing substances selected.

6. The method for calculating odor concentration in municipal wastewater systems based on machine learning as described in claim 1, characterized in that, The method for determining the odor monitoring target is as follows: Odor concentration monitoring points with a number of monitoring time period combinations exceeding a preset combination number threshold are designated as fluctuation monitoring points. Based on the monitoring time period data of the odor-causing locations at different fluctuation monitoring points, determine the total number of monitoring time periods in which the odor-causing substances are present; The odor monitoring targets among the odor-causing substances are determined by the total number of monitoring periods in which the odor-causing substances are present and the number of selected periods.

7. The method for calculating odor concentration in municipal wastewater systems based on machine learning as described in claim 1, characterized in that, Building machine learning models specifically includes: The original data is constructed using monitoring data including the odor monitoring indicators as input and odor concentration as output. For odor monitoring targets that are not present in the original data, replace them with 0. Divide the concentration of each component by the odor threshold of that component to obtain the dilution factor. Then take the common logarithm of the dilution factor. Replace illegal and negative values ​​with 0 to obtain the amplified original data. Based on the amplified original data, the machine learning method is trained to obtain the trained machine learning model.

8. The method for calculating odor concentration in municipal wastewater systems based on machine learning as described in claim 1, characterized in that, The method for determining the calculation results of the odor concentration at the odor concentration monitoring point is as follows: Based on the aforementioned machine learning model and monitoring data, the monitoring data of the odor monitoring target is used as the input to the machine learning model. The output of the machine learning model is used as the calculation result of the odor concentration.

9. The method for calculating odor concentration in municipal wastewater systems based on machine learning as described in claim 1, characterized in that, After obtaining the calculated odor concentration, the output of a binary classification algorithm, including the XGBoost algorithm, can be used to determine whether the odor concentration exceeds the standard.

10. The method for calculating odor concentration in municipal wastewater systems based on machine learning as described in claim 1, characterized in that, The method for determining the update scheme of the odor monitoring target in the machine learning model is as follows: Based on the matching data of the odor monitoring target and the concentration of odor-causing substances at the odor concentration monitoring point, the time period in which the odor-causing substances at the odor concentration monitoring point do not exist is determined and is taken as the period of poor calculation. Based on the data from the poor calculation periods of each odor concentration monitoring point, odor-causing substances that are not odor monitoring targets in different poor calculation periods are identified and regarded as monitoring deviation substances. Based on the data from the poorly calculated periods at each odor concentration monitoring point and the data on the monitoring deviation substances during those periods, an update scheme for the odor monitoring target in the machine learning model is determined.

Images