Method for evaluating uncertainty of greenhouse gas emission data

By combining data collection and deep learning model training with uncertainty assessment, the uncertainty problem in greenhouse gas emission assessment has been solved, enabling more accurate emission prediction and uncertainty assessment, and supporting intelligent decision-making and environmental protection.

CN119049591APending Publication Date: 2024-11-29TIANJIN RICHSOFT ELECTRIC POWER INFORMATION TECH +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410906457.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-08
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

Existing methods for assessing greenhouse gas emissions are subject to uncertainty. Traditional measurement techniques and model simulations have limitations in accuracy and reliability. While intelligent methods have improved upon these limitations, further development is still needed.

Method used

The system employs a data acquisition module to monitor greenhouse gas emission sources in real time. Through deep learning model training and uncertainty assessment modules, combined with probabilistic and statistical methods, it provides accurate emission predictions and uncertainty assessments.

Benefits of technology

It improves the accuracy and reliability of greenhouse gas emission assessment, supports real-time monitoring and intelligent decision-making, and promotes environmental protection and sustainable development.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure BDA0004932257370000031
    Figure BDA0004932257370000031
  • Figure BDA0004932257370000051
    Figure BDA0004932257370000051
  • Figure FDA0004932257350000021
    Figure FDA0004932257350000021
Patent Text Reader

Abstract

The invention discloses an uncertainty evaluation method for greenhouse gas emission data, which comprises the following steps: monitoring the greenhouse gas emission of a greenhouse gas emission source in real time and acquiring real-time data by using a data acquisition module; the data processing module is used for preprocessing, cleaning and correcting the collected real-time data and integrating the data from different sources; and utilizing an uncertainty evaluation module to evaluate the uncertainty of the greenhouse gas emission based on the prediction model, and the like. The intelligent greenhouse gas emission uncertainty evaluation system and method have multiple benefits and effects of improving the evaluation accuracy, evaluating the uncertainty, monitoring in real time, integrating data, supporting intelligent decision, protecting the environment, realizing sustainable development and the like.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of environmental assessment, and in particular to an uncertainty assessment method for greenhouse gas emission data. Background Art

[0002] The assessment of greenhouse gas emissions is crucial for climate change research and environmental protection. However, due to the complex characteristics of greenhouse gases and the diversity of emission sources, accurately assessing greenhouse gas emissions and their uncertainties has always been a challenging problem.

[0003] At present, the commonly used greenhouse gas emissions assessment methods mainly rely on traditional measurement techniques and model simulation methods. The measurement technology includes using sensors and instruments to directly measure the greenhouse gas content or flow of emission sources, and combining it with the activity data of the emission sources for calculation. However, due to the limitations of sensor accuracy, equipment maintenance and field conditions, the measurement technology has uncertainties and faces the challenges of high cost and difficult implementation in large-scale emission source monitoring.

[0004] The model simulation method models and simulates the greenhouse gas emission process based on mathematical models. These models usually use equations based on physical principles or statistical methods, combined with activity data of emission sources and environmental parameters, to predict greenhouse gas emissions. However, due to the selection and uncertainty of model parameters, as well as the model's simplified assumptions about the emission process, the model simulation method also has certain uncertainties.

[0005] In order to solve the uncertainty problem in the assessment of greenhouse gas emissions, some studies have attempted to introduce uncertainty analysis methods. These methods mainly consider the sources of uncertainty as parameter changes or data errors, and apply probabilistic and statistical methods to model and calculate the uncertainty. However, in practical applications, these methods still have certain limitations due to the complexity of data and models, as well as the technical challenges of uncertainty modeling.

[0006] With the rapid development of artificial intelligence and machine learning, intelligent methods have been widely used in the assessment of greenhouse gas emissions. Deep learning technology can learn and predict emissions from large-scale greenhouse gas data sets by constructing complex neural network models. In addition, deep learning can also provide an assessment of the uncertainty of greenhouse gas emissions by establishing a model uncertainty estimation method. The introduction and application of these technologies have effectively improved the accuracy and reliability of greenhouse gas emissions assessment.

[0007] In summary, the current background technology mainly involves traditional measurement technology, model simulation methods, and the introduction of uncertainty analysis and intelligent technology in the application of greenhouse gas emissions assessment. However, further research and development is still needed to improve the accuracy, reliability and intelligence level of greenhouse gas emissions assessment. Summary of the invention

[0008] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a method for uncertainty assessment of greenhouse gas emissions data, which has many benefits and effects such as improved assessment accuracy, assessment uncertainty, real-time monitoring and data integration, intelligent decision support, environmental protection and sustainable development.

[0009] A method for evaluating uncertainty of greenhouse gas emission data comprises the following steps:

[0010] Step 1: using a data acquisition module to monitor the greenhouse gas emissions of greenhouse gas emission sources in real time and obtain real-time data; the data acquisition module includes a sensor network and a data acquisition device, which is used to monitor multiple parameters in real time at the greenhouse gas emission source, including emissions P, soil temperature T, humidity PF, and organic matter content OSC;

[0011] Step 2: preprocessing, cleaning and correcting the collected real-time data using a data processing module, and integrating data from different sources; the data processing module further includes a data preprocessing step, a data cleaning step and a data correction step to improve the accuracy and consistency of the collected data;

[0012] Step 3: using a model training module to train a greenhouse gas emission prediction model using deep learning model training and a greenhouse gas emission dataset; wherein the model training module uses deep learning technology, including neural network structure design, training data preparation, and model parameter tuning, to improve the accuracy and generalization ability of the greenhouse gas emission prediction model;

[0013] Step 4: using the uncertainty assessment module to assess the uncertainty of greenhouse gas emissions based on the prediction model; the uncertainty assessment module calculates the uncertainty of greenhouse gas emissions using a probability statistics method by integrating factors such as data accuracy, sensor accuracy, and model reliability;

[0014] Step 5: Use the result display module to display the assessed greenhouse gas emissions and their uncertainty indicators to the user in an intuitive manner.

[0015] Preferably, in step 2, in the data processing module, preprocessing refers to preliminary processing of the collected real-time data, including data format conversion, data denoising, and data missing value filling; data cleaning refers to screening the data to remove outliers and erroneous data; data correction refers to correcting the collected data based on known standard data or reference data to improve the accuracy of the data; integration refers to merging and integrating data from different sources to form a unified data set;

[0016] Data format conversion, specifically: converting data from text format to digital format;

[0017] Data denoising, specifically: use the existing moving average method to filter the data, the formula is as follows:

[0018] Moving average = (data 1 + data 2 + ... + data n) / n, where n represents the number of data, data 1 to data n represent the data to be averaged, remove the noise points in the data, and consider the points in the data that differ from the average by more than 3% as noise data and remove them;

[0019] Filling missing values ​​in data, specifically: Determination of missing values: If three consecutive data are eliminated using the moving average method, it is considered that there are missing values, and interpolation methods are used to fill them, using linear interpolation, polynomial interpolation or KNN interpolation;

[0020] The correction method is as follows: the sensor output value adopts linear correction, nonlinear correction or zero point correction;

[0021] Data integration, specifically: merging and integrating the above data, merging and classifying the denoised and filled data according to soil temperature T, humidity PF, and organic matter OSC content, so as to facilitate subsequent model training.

[0022] Preferably, the formula for deep learning model training in step 3 is expressed in the following form:

[0023] Input features: X = [x1, x2, ..., x n ];

[0024] Output prediction: Y = f(X; θ);

[0025] Loss function: L(Y,Y_true);

[0026] Parameter tuning: θ_opt = argmin L(Y,Y_true);

[0027] Among them, X represents the input feature, Y represents the predicted output of the model, θ represents the parameters of the model, Y_true represents the true greenhouse gas emissions, L represents the loss function, and θ_opt represents the optimal parameters obtained by tuning; the goal of model training is to minimize the loss function so that the model can more accurately predict greenhouse gas emissions.

[0028] Preferably, in step 3, an improved OLS-RBF neural network algorithm is used to train a greenhouse gas emission prediction model; wherein soil temperature T∈{21,22,23,24,25,26,27,28,29,30}, humidity PF∈{20,25,30,35,40}, and soil organic matter content is divided into six levels according to national standards; a total of 300 sets of data are measured, of which 80% are used for model training and 20% are used for test data, and the mapping relationship between greenhouse gas emissions P and T, PF, and OSC is obtained as follows:

[0029]

[0030] Among them, λ i is the output weight, c i is the center point of the neuron in the i-th hidden layer, σ i is the center point width of the i-th hidden layer.

[0031] Preferably, in step 4, the types of uncertainty and specific calculation formulas are as follows:

[0032] Absolute error: measures the difference between the measured value and the true value. The absolute error is calculated using the following formula: \(\text{Absolute Error}=|\text{measured value}-\text{true value}|\);

[0033] Relative error: Relative error is the ratio of absolute error to true value; relative error is calculated using the following formula: \(\text{Relative Error}=\left(\frac{|\text{measured value}-\text{true value}|}{\text{true value}}\right)\times100\%\);

[0034] Standard Deviation: The standard deviation is found by calculating the variance of the data and taking the square root.

[0035] The advantages and technical effects of the present invention are:

[0036] 1. Improve assessment accuracy: By adopting deep learning technology and large-scale greenhouse gas data sets for model training, the system can provide more accurate greenhouse gas emissions prediction and assessment results.

[0037] 2. Evaluate uncertainty: The system uses probabilistic statistical methods to take into account factors such as data accuracy, sensor accuracy and model reliability, and provides an uncertainty assessment of greenhouse gas emissions. This helps provide more comprehensive information and better manage and respond to uncertainty in decision-making and policy-making.

[0038] 3. Real-time monitoring and data integration: The system's data acquisition module can monitor the data of greenhouse gas emission sources in real time, and pre-process, clean and correct the data from different sources to improve data quality and consistency. In this way, the latest emission information can be obtained in a timely manner, and greenhouse gas emissions can be effectively monitored and managed.

[0039] 4. Intelligent decision support: Through the charts, reports and statistical data generated by the result display module, users can intuitively understand the assessment results of greenhouse gas emissions and their uncertainty. This provides a scientific basis for decision makers and policy makers to make decisions on environmental management, climate change control and policy adjustments based on accurate information.

[0040] 5. Environmental protection and sustainable development: The application of this system is conducive to the realization of environmental protection and sustainable development goals. By accurately assessing greenhouse gas emissions, it helps to discover and solve emission source problems, optimize industrial production processes and energy utilization, reduce environmental load, and promote low-carbon development and climate change response. DETAILED DESCRIPTION

[0041] In order to further understand the content, characteristics and effects of the present invention, the following embodiments are given for detailed description. It should be noted that the embodiments are illustrative, not restrictive, and the protection scope of the present invention cannot be limited thereby.

[0042] A method for evaluating the uncertainty of greenhouse gas emission data includes a data acquisition module, a data processing module, a model training module, an uncertainty evaluation module, and a result display module.

[0043] ① Data acquisition module, used to monitor the greenhouse gas emissions of greenhouse gas emission sources in real time and obtain real-time data; the data acquisition module includes a sensor network and data acquisition equipment, which is used to monitor multiple parameters in real time at the greenhouse gas emission source, including emission P, soil temperature T, humidity PF, and organic matter content OSC.

[0044] ②Data processing module, including data preprocessing steps, data cleaning steps and data correction steps, is used to preprocess, clean and correct the collected real-time data, and integrate data from different sources to improve the accuracy and consistency of the collected data

[0045] In the data processing module, preprocessing refers to the preliminary processing of the collected real-time data, including data format conversion, data denoising, and data missing value filling. Cleaning refers to screening the data to remove outliers and erroneous data. Correction refers to correcting the collected data based on known standard data or reference data to improve the accuracy of the data. Integration refers to merging and integrating data from different sources to form a unified data set.

[0046] The data format conversion: converts data from text format to digital format.

[0047] The data denoising: using the existing moving average method to filter the data, the formula is as follows:

[0048] Moving average = (data 1 + data 2 + ... + data n) / n, where n represents the number of data, data 1 to data n represent the data to be averaged, remove the noise points in the data, and consider the points in the data that differ from the average by more than 3% as noise data and eliminate them.

[0049] The data missing value filling: For missing values ​​(characteristics of missing values: if three consecutive data are eliminated using the moving average method, it is considered that there are missing values), existing interpolation methods can be used to fill them. Linear interpolation, polynomial interpolation, and KN N interpolation are all known and mature technologies that can be used.

[0050] The correction methods include linear correction, nonlinear correction and zero point correction of sensor output values.

[0051] The data integration: the above data are merged and integrated, and the denoised and filled data are merged and classified according to soil temperature T, humidity PF, and organic matter OSC content to facilitate subsequent model training.

[0052] ③ Model training module, used to train the greenhouse gas emissions prediction model using deep learning model training (consisting of improved neural network and parameter tuning) and greenhouse gas emission data set (providing the model formula: in the back); the model training module adopts deep learning technology, including neural network structure design, training data preparation (obtained from actual measurements) and model parameter tuning to improve the accuracy and generalization ability of the greenhouse gas emissions prediction model.

[0053] The formula for deep learning model training is expressed in the following form:

[0054] Input features: X = [x1, x2, ..., xn]

[0055] Output prediction: Y = f(X; θ) (obtained by training with the improved OLS-RBF neural network algorithm)

[0056] Loss function: L(Y,Y_true)

[0057] Parameter tuning: θ_opt = argmin L(Y,Y_true)

[0058] Among them, X represents the input features, Y represents the predicted output of the model, θ represents the parameters of the model, Y_true represents the true greenhouse gas emissions, L represents the loss function, and θ_opt represents the optimal parameters obtained through tuning. The goal of model training is to minimize the loss function so that the model can more accurately predict greenhouse gas emissions.

[0059] The improved OLS-RBF neural network algorithm is used to train the greenhouse gas emission prediction model. Among them, soil temperature T∈{21,22,23,24,25,26,27,28,29,30}, humidity PF∈{20,25,30,35,40}, and soil organic matter content is divided into six levels according to national standards. A total of 300 sets of data were measured, of which 80% were used for model training and 20% were used for test data. The mapping relationship between greenhouse gas emissions P and T, PF, and OSC is as follows:

[0060]

[0061] Among them, λ i is the output weight, c i is the center point of the neuron in the i-th hidden layer, σ i is the center point width of the i-th hidden layer.

[0062] ④ Uncertainty assessment module, which is used to assess the uncertainty of greenhouse gas emissions based on the prediction model; the uncertainty assessment module calculates the uncertainty of greenhouse gas emissions using probabilistic statistical methods by integrating factors such as data accuracy, sensor accuracy and model reliability.

[0063] Calculation uncertainty:

[0064] Absolute Error: A measure of the difference between a measured value and the true value. Absolute error can be calculated using the following formula: Absolute Error = |Measured Value - True Value|.

[0065] The true value is derived from a standard instrument or reference device: A more accurate instrument or device is used as a reference to determine the true value. These devices are usually calibrated and verified to provide highly accurate measurements.

[0066] Standard data set: If there is a verified and recognized greenhouse gas emissions data set, the values ​​in the data set can be used as the true value. This data set is usually provided by professional institutions or organizations and is widely recognized and used.

[0067] Certification or inspection agency: The true value is obtained by entrusting professional certification or inspection agencies to conduct measurements and evaluations. These agencies are authoritative and credible, and they use independent verification procedures to ensure the accuracy of the results they provide.

[0068] 2. Relative Error: Relative error is the ratio of absolute error to true value. Relative error can be calculated using the following formula: \(\text{Relative Error}=\left(\frac{|\text{measured value}-\text{true value}|}{\text{true value}}\right)\times 100\%\).

[0069] 3. Standard Deviation: Standard Deviation is used to measure the degree of dispersion of a set of data. The standard deviation can be obtained by calculating the variance of the data and taking the square root.

[0070] Standard deviation calculation formula: S = Sqrt[(∑(measured value - average value)^2) / (N-1)].

[0071] 4. Confidence Interval: The confidence interval indicates the potential error range of the measured value (the confidence interval is the range of an unknown population parameter estimated at a certain confidence level. The confidence level is usually expressed as 95%, 99%, etc. Mathematically, for a set of sample data, assuming that the sample mean is x, the sample standard deviation is s, and the sample size is n, the sample confidence interval formula can be expressed as:

[0072] Confidence interval = x±t(n-1,α / 2)*s / √n.

[0073] Among them, t(n-1,α / 2) represents the value of the t distribution with n-1 degrees of freedom and α / 2 confidence level, s represents the sample standard deviation, and n represents the sample size. This formula can give the upper and lower bounds of the confidence interval, so that the true population parameter (such as the population mean) has a certain degree of confidence (such as 95% or 99%) that it falls within the confidence interval. Confidence intervals for measured values ​​can be calculated using statistical methods, such as t distribution or normal distribution.

[0074] ⑤Result display module, used to display the assessed greenhouse gas emissions and their uncertainty indicators to the user in an intuitive way. Uncertainty calculation module that considers data accuracy, sensor accuracy and model reliability; uncertainty calculation module that uses probabilistic statistical methods, wherein the result display module further displays the assessment results in the form of charts, reports, etc. The result display module can generate charts, reports or statistical data with visual effects to intuitively and clearly display the assessment results of greenhouse gas emissions and their uncertainties (the interface is designed using the Labview software platform, and automatic drawing and report output are performed based on the result data obtained from the prediction model).

[0075] Better manage and respond to uncertainty in decision-making and policy-making; Real-time monitoring and data integration: The system's data acquisition module can monitor data from greenhouse gas emission sources in real time, and pre-process, clean and correct data from different sources to improve data quality and consistency. In this way, the latest emission information can be obtained in a timely manner, and greenhouse gas emissions can be effectively monitored and managed; Intelligent decision support: Through the charts, reports and statistical data generated by the result display module, users can intuitively understand the assessment results of greenhouse gas emissions and their uncertainty. This provides a scientific basis for decision makers and policy makers to make decisions such as environmental management, climate change control and policy adjustments based on accurate information; Environmental protection and sustainable development: The application of this system contributes to the realization of environmental protection and sustainable development goals. By accurately assessing greenhouse gas emissions, it helps to discover and solve problems with emission sources, optimize industrial production processes and energy utilization, reduce environmental loads, and promote low-carbon development and climate change response.

[0076] Finally, all the parts not described in the present invention adopt mature products and mature technical means in the prior art.

[0077] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all these improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims

1. A method for evaluating uncertainty of greenhouse gas emission data, characterized in that: The following steps are involved: Step 1: using a data acquisition module to monitor the greenhouse gas emissions of greenhouse gas emission sources in real time and obtain real-time data; the data acquisition module includes a sensor network and a data acquisition device, which is used to monitor multiple parameters in real time at the greenhouse gas emission source, including emissions P, soil temperature T, humidity PF, and organic matter content OSC; Step 2, using a data processing module to integrate data from different sources, the data processing module further includes a data preprocessing step, a data cleaning step, and a data correction step to improve the accuracy and consistency of the collected data; Step 3: using a model training module to train a greenhouse gas emissions prediction model using a deep learning model and a greenhouse gas emissions dataset; wherein the model training module uses deep learning technology, including neural network structure design, training data preparation, and model parameter tuning, to improve the accuracy and generalization ability of the greenhouse gas emissions prediction model; Step 4: using the uncertainty assessment module to assess the uncertainty of greenhouse gas emissions based on the prediction model; the uncertainty assessment module calculates the uncertainty of greenhouse gas emissions using a probability statistics method by integrating factors such as data accuracy, sensor accuracy, and model reliability; Step 5: Use the result display module to display the assessed greenhouse gas emissions and their uncertainty indicators to the user in an intuitive manner.

2. The uncertainty assessment method for greenhouse gas emission data according to claim 1, characterized in that: In the step 2, in the data processing module, preprocessing refers to preliminary processing of the collected real-time data, including data format conversion, data denoising, and data missing value filling; data cleaning refers to screening the data to remove abnormal values ​​and erroneous data; data correction refers to correcting the collected data according to known standard data or reference data to improve the accuracy of the data; integration refers to merging and integrating data from different sources to form a unified data set; The data format conversion specifically includes: converting the data from text format to digital format; The data denoising is specifically: using the existing moving average method to filter the data, the formula is as follows: Moving average = (data 1 + data 2 + ... + data n) / n, where n represents the number of data, data 1 to data n represent the data to be averaged, remove the noise points in the data, and consider the points in the data that differ from the average by more than 3% as noise data and remove them; The missing value filling of the data is specifically as follows: missing value determination: if three consecutive data are eliminated by the moving average method, it is considered that there are missing values, and interpolation methods are used to fill them, using linear interpolation, polynomial interpolation or KNN interpolation; The correction method is specifically: the sensor output value adopts linear correction, nonlinear correction or zero point correction; The data integration is specifically: merging and integrating the above data, merging and classifying the denoised and filled data according to soil temperature T, humidity PF, and organic matter OSC content, so as to facilitate subsequent model training.

3. The uncertainty assessment method for greenhouse gas emission data according to claim 1, characterized in that: The formula for deep learning model training in step 3 is expressed in the following form: Input features: X = [x1, x2, ..., x n ]; Output prediction: Y = f(X; θ); Loss function: L(Y,Y_true); Parameter tuning: θ_opt = argmin L(Y,Y_true); Among them, X represents the input feature, Y represents the predicted output of the model, θ represents the parameters of the model, Y_true represents the true greenhouse gas emissions, L represents the loss function, and θ_opt represents the optimal parameters obtained by tuning; the goal of model training is to minimize the loss function so that the model can more accurately predict greenhouse gas emissions.

4. The uncertainty assessment method for greenhouse gas emission data according to claim 1, characterized in that: In step 3, an improved OLS-RBF neural network algorithm is used to train a greenhouse gas emission prediction model; wherein soil temperature T∈{21,22,23,24,25,26,27,28,29,30}, humidity PF∈{20,25,30,35,40}, and soil organic matter content is divided into six levels according to national standards; a total of 300 sets of data are measured, of which 80% are used for model training and 20% are used for test data, and the mapping relationship between greenhouse gas emissions P and T, PF, and OSC is obtained as follows: Among them, λ i is the output weight, c i is the center point of the neuron in the i-th hidden layer, σ i is the center point width of the i-th hidden layer.

5. The uncertainty assessment method for greenhouse gas emission data according to claim 1, characterized in that: In step 4, the types of uncertainty and specific calculation formulas are as follows: Absolute error: measures the difference between the measured value and the true value. The absolute error is calculated using the following formula: \(\text{Absolute Error}=|\text{measured value}-\text{true value}|\); Relative error: Relative error is the ratio of absolute error to true value; The relative error is calculated using the following formula: \(\text{Relative Error}=\left(\frac{|\text{measured value}-\text{true value}|}{\text{true value}}\right)\times100\%\); Standard Deviation: The standard deviation is found by calculating the variance of the data and taking the square root.