Aerobic composting intelligent control method based on artificial intelligence
By using AI-based sensors and machine learning models, the composting environment can be monitored and optimized in real time, solving the problem of lagging temperature and humidity control in traditional composting and improving decomposition efficiency and resource utilization.
Patent Information
- Application Number
- CN202511244173.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-02
- Publication Date
- 2025-10-31
AI Technical Summary
In the traditional agricultural waste composting process, fluctuations in temperature and humidity affect decomposition efficiency, are difficult to control precisely, and rely on manual monitoring, resulting in significant lag in adjustments and serious waste of resources.
An artificial intelligence-based control method is adopted, which uses sensors to monitor the composting environment in real time, combines machine learning models to predict and adjust composting parameters, establishes a dynamic microbial model, and automatically optimizes the composting environment.
It achieves precise temperature and humidity control during the composting process, improves decomposition efficiency, reduces resource waste, and ensures the stability and efficiency of the composting process.
Smart Images

Figure CN120864913A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of aerobic composting technology, specifically an intelligent control method for aerobic composting based on artificial intelligence. Background Technology
[0002] Aerobic composting is a process that uses aerobic microorganisms to transform organic waste into humus. It not only transforms waste into useful soil conditioners but also reduces waste volume and environmental pollution.
[0003] Chinese patent application number CN202210612548.6 discloses an intelligent control system and method for aerobic composting. The configuration method includes: a drive rod positioned on top of an aerobic composting turner to measure the temperature within the compost pile; an oxygen content probe positioned on top of the turner to measure the oxygen content of the compost pile; multiple proximity switches arranged at intervals along the x and y axes of a rectangular coordinate system corresponding to the fermentation tank, dividing the fermentation tank into multiple composting units; multiple aeration ducts laid along the fermentation tank; each aeration duct connected to an aeration fan; and a controller communicatively connected to each proximity switch, temperature probe, and oxygen content probe, and electrically connected to the turner and each aeration fan, capable of determining the current composting unit based on proximity switch signals and optimizing the control of the aeration fans in the current composting unit according to the temperature and oxygen content of the compost pile.
[0004] In the field of aerobic composting technology, although there are technical solutions that use temperature and oxygen content probes on the turner and multiple proximity switches to divide the fermentation tank into multiple composting units along the x and y axes, thereby achieving more precise control of the start and stop of the aeration blowers in each composting unit based on temperature and oxygen content, existing technologies still have the following problems in traditional agricultural waste composting scenarios:
[0005] 1. Fluctuations in temperature and humidity during the composting process directly affect decomposition efficiency. The humidity and temperature of traditional agricultural waste composting are difficult to control precisely, leading to incomplete or excessive fermentation.
[0006] 2. Traditional agricultural waste composting methods usually rely on manual monitoring and adjustment of composting conditions, making it difficult to obtain temperature and humidity data in a timely manner. The adjustments are often delayed and inaccurate, resulting in the waste of resources such as energy and water.
[0007] Therefore, an intelligent control method for aerobic composting based on artificial intelligence is needed, which can detect changes in temperature and humidity in real time and make precise adjustments. Summary of the Invention
[0008] This invention provides an intelligent control method for aerobic composting based on artificial intelligence, aiming to solve the problem that environmental parameters during the composting process are difficult to observe and adjust in real time in traditional agricultural waste composting scenarios, which affects fermentation and efficiency.
[0009] The technical solution adopted by this invention to solve the above-mentioned technical problems is as follows: An intelligent control method for aerobic composting based on artificial intelligence is provided, comprising:
[0010] Installing a sensor acquisition system to dynamically monitor the microbial community involves collecting environmental data and gas generation during the composting process in real time. The sensor data is then uploaded to a monitoring platform, where the activity of the microbial community is obtained through corresponding gas weight fusion.
[0011] Processing and analyzing the collected data to establish a microbial dynamic model involves preprocessing the collected data, filling in missing items and removing outliers, arranging the collected data in chronological order, and selecting representative features from each collected data point through correlation analysis to establish a microbial dynamic model.
[0012] Using machine learning for prediction and automatic adjustment of the composting environment involves selecting a regression decision tree as the machine learning model. Based on features, the collected data is divided into subsets of different categories, the effectiveness of the features in the classification is calculated, the data is input into the regression decision tree to obtain the prediction results, and the prediction results are organized into instructions for execution by the monitoring platform.
[0013] The statistical composting environment feedback data is used to iteratively optimize the dynamic model. This involves statistically analyzing environmental parameters and microbial activity data after the execution of statistical instructions, calculating statistical indicators, compiling them into a feedback dataset, calculating errors, and using these errors to iterate the machine learning model.
[0014] As a preferred embodiment, the specific steps of installing the sensor acquisition system to dynamically monitor the microbial community are as follows: sensors are installed at the center and edge of the compost pile to monitor data from various areas within the compost pile, collecting environmental data and the generation of common gases during the composting process in real time, and obtaining information on the activity of the microbial community through changes in gas concentration; a server is installed in the central control area of the composting site, and a data acquisition system is deployed, connecting all sensors wirelessly to the data acquisition system, uploading the collected real-time data to the monitoring platform of the data acquisition system, which displays environmental parameters in real time, and is also responsible for real-time monitoring of various environmental parameters during the composting process, data storage, processing and analysis, and intelligent decision support, and is divided into a data acquisition layer, a data processing and analysis layer, and a decision support layer; the data acquisition layer transmits the real-time collected data to the platform via Wi-Fi, the data processing and analysis layer cleans the data, analyzes data change trends using machine learning algorithms and provides predictions, and the decision support layer generates corresponding decisions based on the analysis to adjust the composting environment to stabilize the process;
[0015] Monitoring the gases produced by microbial activity reveals several indicators. Increased carbon dioxide concentration indicates active organic matter decomposition, while decreased oxygen concentration suggests aerobic microorganisms are consuming oxygen for metabolism. The rate of oxygen concentration decrease allows calculation of the activity intensity of aerobic microorganisms in the compost. Increased ammonia concentration indicates microorganisms are decomposing nitrogenous organic matter; excessively high ammonia concentrations indicate excessive moisture in the compost pile. Increased methane concentrations indicate anaerobic decomposition in the compost pile, requiring increased ventilation or humidity adjustment. The activity rate of the microbial community is calculated based on the gas production indicators using the following formula:
[0016] rate v =ω1rate c +ω2rate o +ω3rate n +ω4rate ch ,
[0017] Where rate v This represents the activity rate of the microbial community, where ω1 represents the weight of carbon dioxide activity rate, ω2 represents the weight of oxygen activity rate, ω3 represents the weight of ammonia activity rate, and ω4 represents the weight of methane activity rate. c The rate represents the rate at which carbon dioxide is released. o The rate indicates the rate of oxygen consumption. n The rate indicates the release rate of ammonia. ch This indicates the rate of methane production.
[0018] As a preferred implementation, the specific steps for processing and analyzing the collected data to establish a microbial dynamic model are as follows: First, noise in the sensor data is removed using a smoothing algorithm; missing data is filled using interpolation; abnormal data is detected and processed using the Z-Score method; representative features are selected from each collected data set through correlation analysis; statistical relationships of the collected data are calculated; linear relationships of different features are calculated using the Pearson correlation coefficient; and features with linear relationships to the composting microbial activity range are selected. The formula is:
[0019]
[0020] Where relative represents the Pearson correlation coefficient, i represents the i-th observation of the data, and X i Let X represent the i-th value, and Y represent the ith value. i This represents the i-th value of data Y. This represents the mean of the data X. This represents the mean of the data Y;
[0021] The collected data were arranged chronologically, and their features were extracted. A Monod dynamic model was established to describe the relationship between microbial growth rate and substrate concentration. A curve equation was hypothesized between the microbial growth rate and substrate concentration. The growth rate of microorganisms at different substrate concentrations was measured using experimental data. The maximum growth rate at the highest substrate concentration and the substrate concentration at half the maximum growth rate were calculated. The model formula is as follows:
[0022]
[0023] Where v(d) represents the microbial growth rate of substrate concentration d, v max k represents the maximum growth rate at high substrate concentrations. d d represents the substrate concentration at half the maximum growth rate.
[0024] As a preferred implementation, the specific steps of using machine learning to predict and automatically adjust the composting environment are as follows: select a regression decision tree as the machine learning model according to the characteristics of the data, integrate historical data as the training set and validation set and input them into the regression decision tree, construct the decision tree by selecting data features with discriminative power, divide the dataset into subsets according to the different values of each feature, and input the trained regression decision tree into the current collected data to obtain the prediction results.
[0025] The Gini index is used to replace the effectiveness of each feature in classification, and the formula is:
[0026]
[0027] Where data represents the dataset, Gini(data) represents the Gini index of the dataset data, j represents the class, c represents the number of classes in the dataset, and p j Represents the probability of category j;
[0028] The performance of the regression decision tree is evaluated by recursively splitting the dataset until the depth of the tree is limited to 4 levels, with each leaf node of the decision tree representing a predicted value.
[0029] The dataset is evenly divided into multiple subsets, and then training and validation are performed a number of times for each subset. Each time, a different subset is selected as the validation set, and the remaining subset is used as the training set. After each training, the performance indicators, including mean squared error and root mean square error, are calculated and averaged. The average error under different parameters is compared and used as the overall error evaluation index of the regression decision tree model.
[0030] Based on the prediction results of the regression decision tree, instructions are compiled and issued to adjust the environmental parameters of the compost pile. When the decision tree predicts a low microbial growth rate, the monitoring platform issues instructions to increase temperature and humidity to promote microbial activity. By predicting changes in gas factors, the regression decision tree guides the monitoring platform to adjust the aeration system of the compost pile.
[0031] As a preferred implementation, the specific steps for iterative optimization of the dynamic model using statistical composting environmental feedback data are as follows: Analyze the environmental parameters and microbial community activity data of the compost after executing the command, calculate the mean, standard deviation, and variation range of parameters and data at different times, construct a dataset arranged in chronological order, and input it into the regression decision tree as a feedback dataset. Simultaneously, input the environmental feedback data into the Monod dynamic model to obtain the actual observed microbial community activity data at the corresponding time, calculate the error between the predicted value given by the regression decision tree and the actual observed value of the Monod dynamic model, use the mean absolute error as an evaluation index to iterate the parameters of the regression decision tree model, and repeat the above iterative process. In each iteration, update the parameters of the regression decision tree model, re-predict and compare the error until the regression decision tree model reaches the set number of iterations.
[0032] The difference between the predicted values of the regression decision tree at different time points and the actual observed values is calculated. The mean squared error is obtained by squaring all the differences and averaging them. The key parameters in the regression decision tree model are iteratively adjusted based on the mean squared error index so that the predictions given by the regression decision tree are closer to the actual observed data.
[0033] The beneficial effects of this invention are as follows:
[0034] 1. This invention uses a sensor acquisition system to monitor temperature and humidity data in real time during the composting process, enabling precise control of the composting environment. By combining dynamically monitored microbial community data with machine learning predictions, the temperature and humidity of the compost pile are automatically adjusted, avoiding the problem of excessive temperature and humidity fluctuations in traditional composting methods and improving decomposition efficiency.
[0035] 2. This invention continuously optimizes the dynamic model by statistically analyzing the feedback data of the composting environment, thereby achieving continuous improvement and precise control of the composting environment. The data collected during each composting process is fed back into the machine learning model for continuous iteration and optimization, ensuring that the model can more accurately predict and adjust the composting environment as the composting process progresses.
[0036] Legend
[0037] Figure 1 This is a flowchart of an intelligent control method for aerobic composting based on artificial intelligence.
[0038] Figure 2 An illustration of the effect of an artificial intelligence-based intelligent control method for aerobic composting. Detailed Implementation
[0039] To make the technical means, creative features, objectives, and effects of this invention easier to understand, the invention is further described below with reference to specific embodiments. However, the following embodiments are merely preferred embodiments of this invention and not all of them. Other embodiments obtained by those skilled in the art based on the embodiments described herein without creative effort are all within the protection scope of this invention.
[0040] Example 1, such as Figure 1 This is an intelligent control method for aerobic composting based on artificial intelligence, including the following steps:
[0041] Install a sensor data acquisition system to dynamically monitor the microbial community;
[0042] Process and analyze the collected data to establish a dynamic model of microorganisms;
[0043] Using machine learning to make predictions and automatically adjust the composting environment;
[0044] Statistical data on composting environmental feedback is used to iteratively optimize the dynamic model.
[0045] The specific implementation steps are as follows: An intelligent control method for aerobic composting based on artificial intelligence, wherein the specific steps for installing a sensor acquisition system to dynamically monitor the microbial community are as follows: Sensors are installed at the center and edge of the compost pile to monitor data from various areas within the compost pile, such as temperature and humidity sensors, gas sensors, and microbial activity sensors, to collect environmental data and the generation of common gases during the composting process in real time, and to obtain the activity of the microbial community through changes in gas concentration; A server is installed in the central control area of the composting site, and a data acquisition system is deployed, connecting all sensors wirelessly to the data acquisition system, and uploading the collected real-time data to the monitoring platform of the data acquisition system, which displays environmental parameters such as temperature, humidity, gas concentration, and microbial activity in real time;
[0046] Specifically, obtaining information about microbial community activity through changes in gas concentration involves monitoring gases produced by microbial activity, such as carbon dioxide, oxygen, ammonia, and methane. An increase in carbon dioxide concentration indicates active organic matter decomposition, while a decrease in oxygen concentration indicates that aerobic microorganisms are consuming oxygen for metabolism. By monitoring the rate of decrease in oxygen concentration, the activity intensity of aerobic microorganisms in the compost can be calculated. An increase in ammonia concentration indicates that microorganisms are decomposing nitrogenous organic matter; excessively high ammonia concentrations indicate excessive moisture in the compost. An increase in methane concentration indicates anaerobic decomposition in the compost pile, requiring increased ventilation or humidity adjustment. The activity rate of the microbial community is calculated based on the indicators of gas production using the following formula:
[0047] rate v =ω1rate c +ω2rate o +ω3rate n +ω4rate ch ,
[0048] Where rate v This represents the activity rate of the microbial community, where ω1 represents the weight of carbon dioxide activity rate, ω2 represents the weight of oxygen activity rate, ω3 represents the weight of ammonia activity rate, and ω4 represents the weight of methane activity rate. c The rate represents the rate at which carbon dioxide is released. o The rate indicates the rate of oxygen consumption. n The rate indicates the release rate of ammonia. ch Indicates the rate of methane production;
[0049] Specifically, the monitoring platform of the data acquisition system is responsible for real-time monitoring of various environmental parameters, data storage, processing and analysis, and intelligent decision support during the composting process. It is divided into a data acquisition layer, a data processing and analysis layer, and a decision support layer. The data acquisition layer transmits the real-time collected data to the platform via Wi-Fi. The data processing and analysis layer cleans the data, analyzes the data change trends through machine learning algorithms, and provides predictions. The decision support layer generates corresponding decisions based on the analysis of the situation to adjust the composting environment to stabilize the process.
[0050] Based on the above steps, the specific steps for processing and analyzing the collected data and establishing a microbial dynamic model are as follows: First, noise in the sensor data is removed using a smoothing algorithm; then, missing data items are filled using interpolation; and finally, outlier data is detected and processed using the Z-Score method. The smoothing algorithm applies a weighted average to the data sequence to smooth data changes and ensure data stability and usability. The interpolation method calculates the value of missing data items based on the relationships between known data items to fill in the gaps in the data. The Z-Score method identifies and marks outliers in the data by calculating the standard deviation distance between the data points and the mean, ensuring the normal distribution and validity of the data.
[0051] Representative features were selected from each collected data point through correlation analysis. Specifically, statistical relationships such as the mean and standard deviation of the collected data were calculated, and the linear relationships of different features were calculated using the Pearson correlation coefficient. Features with linear relationships to the activity range of composting microorganisms were selected. The formula is as follows:
[0052]
[0053] Where relative represents the Pearson correlation coefficient, i represents the i-th observation of the data, and X i Let X represent the i-th value, and Y represent the ith value. i This represents the i-th value of data Y. This represents the mean of the data X. This represents the mean of the data Y;
[0054] The collected data were arranged in chronological order, and the features of the collected data were extracted to establish a Monod dynamic model to describe the relationship between the growth rate of microorganisms and the substrate concentration.
[0055] Specifically, the aforementioned Monod dynamic model assumes a curve equation relating microbial growth rate and substrate concentration. By measuring the microbial growth rate at different substrate concentrations using experimental data, the maximum growth rate at the highest substrate concentration and the substrate concentration at half the maximum growth rate are calculated. The model formula is as follows:
[0056]
[0057] Where v(d) represents the microbial growth rate of substrate concentration d, v max k represents the maximum growth rate at high substrate concentrations. d d represents the substrate concentration at half the maximum growth rate;
[0058] Based on the above steps, the specific steps for using machine learning to make predictions and automatically adjust the composting environment are as follows: Select a regression decision tree as the machine learning model according to the characteristics of the data, integrate historical data as the training set and validation set and input them into the regression decision tree, construct the decision tree by selecting data features with discriminative power, divide the dataset into subsets according to the different values of each feature, and input the trained regression decision tree into the currently collected data to obtain the prediction results.
[0059] Specifically, the process of dividing the dataset into subsets based on feature values and using the Gini index to represent the effectiveness of each feature in classification is described by the following formula:
[0060]
[0061] Where data represents the dataset, Gini(data) represents the Gini index of the dataset data, j represents the class, c represents the number of classes in the dataset, and p j Represents the probability of category j;
[0062] The performance of the regression decision tree is evaluated by recursively splitting the dataset until the depth of the tree is limited to 4 levels, with each leaf node of the decision tree representing a predicted value.
[0063] Specifically, the cross-validation method for evaluating the performance of the regression decision tree involves dividing the dataset evenly into multiple subsets, then performing training and validation multiple times for each subset. Each time, a different subset is selected as the validation set, and the remaining subset is used as the training set. After each training iteration, performance metrics, including mean squared error and root mean square error, are calculated and averaged. The average error under different parameters is compared and used as the overall error evaluation metric for the regression decision tree model.
[0064] Based on the prediction results of the regression decision tree, instructions are compiled and issued to adjust the environmental parameters of the compost pile. When the decision tree predicts a low microbial growth rate, the monitoring platform issues instructions to increase temperature and humidity to promote microbial activity. By predicting changes in gas factors, the regression decision tree guides the monitoring platform to adjust the aeration system of the compost pile.
[0065] Based on the above steps, the specific steps for iterative optimization of the dynamic model are as follows: Analyze the environmental parameters and microbial community activity data of the compost after the execution of the command, calculate the mean, standard deviation, and variation of parameters and data at different times, construct a dataset arranged in chronological order, and input it into the regression decision tree as the feedback dataset. At the same time, input the environmental feedback data into the Monod dynamic model to obtain the actual observed microbial community activity data at the corresponding time, calculate the error between the predicted value given by the regression decision tree and the actual observed value of the Monod dynamic model, use the mean absolute error as the evaluation index to iterate the parameters of the regression decision tree model, and repeat the above iterative process. In each iteration, update the regression decision tree model parameters, re-predict and compare the error until the regression decision tree model reaches the maximum number of iterations, such as 200 times.
[0066] Specifically, the iterative parameter adjustment of the model based on the evaluation index involves calculating the difference between the predicted values of the regression decision tree at different time points and the actual observed values, averaging all the differences to obtain the mean square error, and iteratively adjusting the key parameters in the regression decision tree model based on the mean square error index, such as the microbial specific growth rate, half-saturation constant, and environmental response coefficient, so that the predictions given by the regression decision tree are closer to the actual observed data.
[0067] Example 2, based on Example 1 above, describes the application of an AI-based intelligent control method for aerobic composting in a traditional agricultural waste composting scenario, specifically as follows:
[0068] Step 1 involves deploying microbial DNA sensors to periodically collect DNA samples from microbial communities in the compost pile. Real-time polymerase chain reaction (PCR) technology is used to monitor the types and quantities of microbial communities, calculate the relative abundance of different microbial populations at various time points, and combine this data with environmental data such as temperature, humidity, and gas concentration monitored in real time during the composting process to analyze the relationship between environmental factors and microbial community dynamics. By establishing a microbial community dynamic model, microbial activity can be monitored in real time, and its growth trend can be predicted.
[0069] Step 2: Combining multi-source data such as temperature, humidity, gas concentration, and microbial growth rate, calculate the mean and standard deviation of the collected data, extract key features affecting microbial community activity such as gas concentration, and dynamically establish an autoregressive model. First, set an initial autoregressive order to determine the linear relationship between the current microbial community activity and the activity at previous orders. Select a suitable order through autocorrelation plots and partial autocorrelation plots, calculate the difference between the actual observed value and the current autoregressive model output value, sum the squares of the errors to obtain the sum of squares, and adjust the autoregressive coefficients to minimize the sum of squares to fit the model and form a dynamic model of the microbial community.
[0070] Step 3: Select the random forest model as the machine learning model. Input multi-source feature values such as temperature, humidity, oxygen, carbon dioxide and ammonia concentration collected by sensors to predict the environmental parameters of the compost pile. Specifically, construct multiple regression decision trees, train each tree using a random training subset and a random feature subset, independently fit the relationship between the features and the prediction target, form a set of multiple prediction trees, and average the prediction results of all trees to obtain the overall prediction value of the random forest. Based on the overall prediction value, generate corresponding instructions to regulate the compost environment, such as adjusting the ventilation system and maintaining temperature and humidity.
[0071] Step four: Divide the compost pile into layers according to height. Determine the height of each layer based on the total height of the pile, such as 20 cm. Deploy sensors independently in each layer to collect and statistically analyze the time series data of temperature, humidity, and gas concentration in each layer. Build multiple regression decision tree models for each layer's data and compare the prediction error with the actual observation value to iterate the parameters of the multiple regression decision tree models for each layer until the difference between the prediction error and the actual observation value is within 5%. Average and merge the parameters of the multiple regression decision tree models from multiple layers to obtain the overall parameters of the multiple regression decision tree model.
[0072] The embodiments of the present invention described above are subject to modification and change of method by those skilled in the art without departing from the embodiments and broader aspects of the present invention. The appended claims are intended to include all such modifications and changes of method that do not depart from the present invention.
Claims
1. An intelligent control method for aerobic composting based on artificial intelligence, characterized in that, include: Installing a sensor acquisition system to dynamically monitor the microbial community involves collecting environmental data and gas generation during the composting process in real time. The sensor data is then uploaded to a monitoring platform, where the activity of the microbial community is obtained through corresponding gas weight fusion. Processing and analyzing the collected data to establish a microbial dynamic model involves preprocessing the collected data, filling in missing items and removing outliers, arranging the collected data in chronological order, and selecting representative features from each collected data point through correlation analysis to establish a microbial dynamic model. Using machine learning for prediction and automatic adjustment of the composting environment involves selecting a regression decision tree as the machine learning model. Based on features, the collected data is divided into subsets of different categories, the effectiveness of the features in the classification is calculated, the data is input into the regression decision tree to obtain the prediction results, and the prediction results are organized into instructions for execution by the monitoring platform. The statistical composting environment feedback data is used to iteratively optimize the dynamic model. This involves statistically analyzing environmental parameters and microbial activity data after the execution of statistical instructions, calculating statistical indicators, compiling them into a feedback dataset, calculating errors, and using these errors to iterate the machine learning model.
2. The intelligent control method for aerobic composting based on artificial intelligence according to claim 1, characterized in that: The specific steps for installing the sensor acquisition system to dynamically monitor the microbial community are as follows: Sensors are installed at the center and edges of the compost pile to monitor data from various areas within the pile, collecting environmental data and the generation of common gases during the composting process in real time, and obtaining information on microbial community activity through changes in gas concentration. A server is installed in the central control area of the composting site, and a data acquisition system is deployed. All sensors are wirelessly connected to the data acquisition system, and the collected real-time data is uploaded to the monitoring platform of the data acquisition system. The platform displays environmental parameters in real time and is also responsible for real-time monitoring of various environmental parameters during the composting process, data storage, processing and analysis, and intelligent decision support. It is divided into a data acquisition layer, a data processing and analysis layer, and a decision support layer. The data acquisition layer transmits real-time collected data to the platform via Wi-Fi. The data processing and analysis layer cleans the data, analyzes data trends using machine learning algorithms, and provides predictions. The decision support layer generates corresponding decisions based on the analysis to adjust the composting environment to stabilize the process.
3. The intelligent control method for aerobic composting based on artificial intelligence according to claim 2, characterized in that: The specific steps for installing the sensor acquisition system to dynamically monitor the microbial community also include: Monitoring the gases produced by microbial activity reveals several indicators. Increased carbon dioxide concentration indicates active organic matter decomposition, while decreased oxygen concentration suggests aerobic microorganisms are consuming oxygen for metabolism. The rate of oxygen concentration decrease allows calculation of the activity intensity of aerobic microorganisms in the compost. Increased ammonia concentration indicates microorganisms are decomposing nitrogenous organic matter; excessively high ammonia concentrations indicate excessive moisture in the compost pile. Increased methane concentrations indicate anaerobic decomposition in the compost pile, requiring increased ventilation or humidity adjustment. The activity rate of the microbial community is calculated based on the gas production indicators using the following formula: rate v =ω1rate c +ω2rate o +ω3rate n +ω4rate ch , Where rate v This represents the activity rate of the microbial community, where ω1 represents the weight of carbon dioxide activity rate, ω2 represents the weight of oxygen activity rate, ω3 represents the weight of ammonia activity rate, and ω4 represents the weight of methane activity rate. c The rate represents the rate at which carbon dioxide is released. o The rate indicates the rate of oxygen consumption. n The rate indicates the release rate of ammonia. ch This indicates the rate of methane production.
4. The intelligent control method for aerobic composting based on artificial intelligence according to claim 1, characterized in that: The specific steps for processing and analyzing the collected data to establish a microbial dynamic model are as follows: First, noise in the sensor data is removed using a smoothing algorithm. Then, interpolation is used to fill in missing data. The Z-Score method is used to detect and process outlier data. Representative features are selected from each collected data set through correlation analysis. Statistical relationships between the collected data are calculated, and the Pearson correlation coefficient is used to calculate linear relationships between different features. Features exhibiting linear relationships with the composting microbial activity range are selected. The formula is as follows: Where relative represents the Pearson correlation coefficient, i represents the i-th observation of the data, and X i Let X represent the i-th value, and Y represent the ith value. i This represents the i-th value of data Y. This represents the mean of the data X. This represents the mean of the data Y.
5. The intelligent control method for aerobic composting based on artificial intelligence according to claim 4, characterized in that: The specific steps for processing and analyzing the collected data to establish a microbial dynamic model also include: The collected data were arranged chronologically, and their features were extracted. A Monod dynamic model was established to describe the relationship between microbial growth rate and substrate concentration. A curve equation was hypothesized between the microbial growth rate and substrate concentration. The growth rate of microorganisms at different substrate concentrations was measured using experimental data. The maximum growth rate at the highest substrate concentration and the substrate concentration at half the maximum growth rate were calculated. The model formula is as follows: Where v(d) represents the microbial growth rate of substrate concentration d, v max k represents the maximum growth rate at high substrate concentrations. d d represents the substrate concentration at half the maximum growth rate.
6. The intelligent control method for aerobic composting based on artificial intelligence according to claim 1, characterized in that: The specific steps for using machine learning to predict and automatically adjust the composting environment are as follows: Based on the characteristics of the data, a regression decision tree is selected as the machine learning model. Historical data is integrated as the training set and validation set and input into the regression decision tree. The decision tree is constructed by selecting data features with discriminative power. For each feature, the dataset is divided into subsets according to the different values of the feature. The trained regression decision tree is then input into the currently collected data to obtain the prediction results.
7. The intelligent control method for aerobic composting based on artificial intelligence according to claim 6, characterized in that: The specific steps for using machine learning to predict and automatically adjust the composting environment also include: The Gini index is used to replace the effectiveness of each feature in classification, and the formula is: Where data represents the dataset, Gini(data) represents the Gini index of the dataset data, j represents the class, c represents the number of classes in the dataset, and p j This represents the probability of category j; The performance of the regression decision tree is evaluated by recursively splitting the dataset until the tree depth is limited to level 4, with each leaf node of the decision tree representing a predicted value.
8. The intelligent control method for aerobic composting based on artificial intelligence according to claim 6, characterized in that: The specific steps for using machine learning to predict and automatically adjust the composting environment also include: The dataset is evenly divided into multiple subsets, and then training and validation are performed a number of times for each subset. Each time, a different subset is selected as the validation set, and the remaining subset is used as the training set. After each training, the performance indicators, including mean squared error and root mean square error, are calculated and averaged. The average error under different parameters is compared and used as the overall error evaluation index of the regression decision tree model. Based on the prediction results of the regression decision tree, instructions are compiled and issued to adjust the environmental parameters of the compost pile. When the decision tree predicts a low microbial growth rate, the monitoring platform issues instructions to increase temperature and humidity to promote microbial activity. By predicting changes in gas factors, the regression decision tree guides the monitoring platform to adjust the aeration system of the compost pile.
9. The intelligent control method for aerobic composting based on artificial intelligence according to claim 1, characterized in that: The specific steps for iterative optimization of the dynamic model based on the statistical composting environment feedback data are as follows: After executing the statistical instructions, the environmental parameters and microbial community activity data of the compost were analyzed. The mean, standard deviation, and variation of parameters and data at different times were calculated, and a dataset arranged in chronological order was constructed. This dataset was then used as a feedback dataset and input into the regression decision tree. Simultaneously, the environmental feedback data was input into the Monod dynamic model to obtain the actual observed microbial community activity data at the corresponding time. The error between the predicted value given by the regression decision tree and the actual observed value of the Monod dynamic model was calculated. The mean absolute error was used as an evaluation index to iterate the parameters of the regression decision tree model. The above iterative process was repeated, updating the regression decision tree model parameters in each iteration, re-predicting, and comparing the error until the regression decision tree model reached the set number of iterations.
10. The intelligent control method for aerobic composting based on artificial intelligence according to claim 9, characterized in that: The specific steps for iterative optimization of the dynamic model using the statistical composting environmental feedback data also include: The difference between the predicted values of the regression decision tree at different time points and the actual observed values is calculated. The mean squared error is obtained by squaring all the differences and averaging them. The key parameters in the regression decision tree model are iteratively adjusted based on the mean squared error index so that the predictions given by the regression decision tree are closer to the actual observed data.
Citation Information
Patent Citations
Intelligent control system and method for aerobic composting
CN114956883A
Intelligent control method and system for sludge aerobic composting
CN115146776A
Intelligent control method and system for aerobic composting
CN115959933A
Preparation method of traditional Chinese medicine residue composting inoculant based on artificial intelligence
CN118800337A
Intelligent agricultural compost microbial flora optimization management system
CN118908760A
