Agricultural net carbon sink prediction method and related equipment

Abstract

The invention provides an agricultural net carbon sink prediction method and related equipment, and the method comprises the steps: carrying out the measurement and calculation of agricultural net carbon sink, and obtaining the amount of agricultural net carbon sink; important factors influencing the agricultural net carbon sink are screened by using a Lasso regression algorithm and an ISM algorithm, and the scientificity and reliability of a screening result are ensured; the method comprises the following steps of: constructing a data set comprising a training set and a test set by utilizing the agricultural net carbon sink quantity, important factors influencing the agricultural net carbon sink and province variables, inputting the training set into a random forest model, training a random forest, setting parameters in the random forest model according to an optimal parameter combination optimized by a grey wolf optimization algorithm, and establishing a random forest model; evaluating by using the test set to obtain a net carbon sink prediction model; and inputting the obtained important factors and province variables of the target area influencing the agricultural net carbon sink into the net carbon sink prediction model for prediction to obtain a net carbon sink prediction result of the target area, thereby improving the precision and stability of agricultural net carbon sink prediction.

Description

Technical Field

[0001] This invention relates to the field of net carbon sink prediction technology, and in particular to a method and related equipment for predicting net carbon sinks in agriculture. Background Technology

[0002] Currently, in the field of agricultural carbon sink and carbon source research, domestic scholars focus on two main aspects: macroscopic measurement and analysis of the current status of carbon sinks and carbon sources, and analysis of the driving forces and influencing factors of carbon sinks and carbon sources. Regarding macroscopic measurement and analysis of the current status of carbon sinks and carbon sources, there are many results based on provincial-level measurements. Net carbon sinks, which organically link carbon sinks and carbon sources, are an important prerequisite and cornerstone for in-depth research on the carbon effects of agricultural production.

[0003] In the prediction of net carbon sinks in agriculture, existing research mainly focuses on traditional statistical analysis and modeling. Most researchers use the STIRPAT model to select key influencing factors, then use ridge regression to screen important influencing factors for each province and establish their own prediction models. However, this prediction model lacks generalization ability. Furthermore, the ridge regression model assumes a linear relationship between the predictor and target variables, which is insufficient in capturing complex nonlinear relationships. Existing research shows that net carbon sinks in agriculture are affected by multiple nonlinear factors, such as climate change, policy intervention, and soil properties, thus limiting the application of the ridge regression model in net carbon sink prediction. Some researchers directly use the GM(1,1) grey prediction model for prediction. The GM(1,1) grey prediction model is only suitable for systems with relatively simple data variation patterns and is inadequate in handling complex multivariate relationships; moreover, this model is highly dependent on the initial data, and if the data quality is poor or highly volatile, the accuracy of the model's prediction decreases significantly. However, in the prediction of net carbon sinks in agriculture, carbon sink changes are often influenced by multiple complex factors, meaning the model lacks predictive accuracy, thus limiting its application in net carbon sink prediction.

[0004] In summary, existing technologies have significant shortcomings in screening and predictive modeling of factors influencing agricultural net carbon sinks. Factor screening relies too heavily on traditional methods and has significant limitations; in predictive modeling, the combined effects of multiple variables and nonlinearities are not considered, resulting in models lacking generalization and accuracy. Summary of the Invention

[0005] This invention provides a method and related equipment for predicting net carbon sinks in agriculture, with the aim of improving the accuracy and stability of net carbon sink prediction in agriculture.

[0006] To achieve the above objectives, the present invention provides a method for predicting net carbon sinks in agriculture, comprising: Step 1: Calculate the net carbon sink in agriculture using the obtained carbon emission data and crop data to obtain the net carbon sink in agriculture. Step 2: Use the Lasso regression algorithm and the ISM algorithm to screen the factors affecting agricultural net carbon sinks and obtain the important factors affecting agricultural net carbon sinks. Step 3: Using the net carbon sink in agriculture, the important factors affecting the net carbon sink in agriculture, and the provincial variables obtained by analyzing the factors affecting the net carbon sink in agriculture, construct a dataset including a training set and a test set, and input the training set into the random forest model to train the random forest. Step 4: During the training process of the random forest model, the optimal parameter combination of the random forest model is optimized by the gray wolf optimization algorithm. The parameters in the random forest model are set according to the optimal parameter combination, and the random forest model with the optimal parameter combination is evaluated using the test set to obtain the net carbon sink prediction model. Step 5: Input the key factors affecting agricultural net carbon sinks and provincial variables obtained from the target area into the net carbon sink prediction model for prediction, and obtain the net carbon sink prediction results for the target area.

[0007] Furthermore, step 1 includes: The total agricultural carbon emissions are calculated using the acquired carbon emission data. The calculation expression is as follows:

[0008] in, This represents the total carbon emissions from agriculture. Indicates the first Carbon emissions from carbon-like sources Indicates the first Carbon source data, Indicates the first Emission coefficients of carbon-like sources; The carbon uptake of crops is calculated using the acquired crop data to obtain the total carbon uptake by crops. The calculation expression is as follows:

[0009] in, This indicates the total amount of carbon absorbed by crops. Indicates the first Carbon absorption rate of crops Indicates the first Biomass yield of crop-like crops Indicates the first Economic yield of similar crops Indicates the first Economic coefficients of similar crops Indicates the first The water content of similar crops; The net carbon sink of agriculture is calculated based on the total agricultural carbon emissions and the total carbon absorption of crops. The calculation expression is as follows:

[0010] in, This indicates the net carbon sink in agriculture.

[0011] Furthermore, step 2 includes: The factors affecting agricultural net carbon sinks were standardized to obtain the standardized factors affecting agricultural net carbon sinks. The Lasso regression algorithm was used to perform Lasso regression on net carbon sink and standardized factors affecting agricultural net carbon sink to obtain the first screening results. The survey results of agricultural production areas were analyzed to identify the factors affecting net carbon sinks in agriculture and their relative importance. An artificial intelligence model was used to score these factors, resulting in an initial direct impact matrix. A reachability matrix consisting of 0s and 1s is generated based on the initial direct influence matrix. The reachability matrix includes multiple rows and columns. The reachability matrix is ​​hierarchically partitioned using the ISM algorithm to determine the second screening result; The union of the first and second screening results yields the key factors influencing net carbon sinks in agriculture.

[0012] Furthermore, based on the initial direct influence matrix, a reachability matrix consisting of 0s and 1s is generated. The reachability matrix includes multiple rows and columns, including: A standardized direct influence matrix is ​​constructed based on the initial direct influence matrix. The calculation expression is:

[0013] in, Represents the elements of the normalized matrix. This represents the result obtained by taking the average score of all artificial intelligence models. This indicates the maximum value that initially directly affects all elements in the matrix. This indicates the number of factors affecting net carbon sinks in agriculture; A comprehensive influence matrix is ​​constructed based on the standardized direct influence matrix. The calculation expression is:

[0014] Incorporating the overall influence matrix into the identity matrix The global influence matrix is ​​obtained. The calculation expression is:

[0015] Global influence matrix Set threshold The global influence matrix The value is greater than or equal to the threshold. Assigning the value 1 to the elements will change the global influence matrix. Medium less than the threshold The elements are assigned the value 0, and the reachability matrix is ​​obtained.

[0016] Furthermore, the ISM algorithm is used to hierarchically partition the reachability matrix to determine the second screening results, including: The expression for hierarchical partitioning of the reachability matrix using the ISM algorithm is as follows:

[0017]

[0018] in, Represents the th in the reachability matrix The sum of the elements in the row. Represents the th in the reachability matrix The sum of the column elements, Represents the th in the reachability matrix Line 1 Column elements; According to the reachability matrix, the first Sum of row elements With the reachability matrix of the th Sum of column elements Determine the results of the second screening.

[0019] Furthermore, the calculation expression for the provincial variable obtained by analyzing the factors affecting agricultural net carbon sink is as follows:

[0020] in, Indicates the first Province variables for each province, Indicates the first The average net agricultural carbon sink of each province Indicates the first The average value of factors affecting net agricultural carbon sinks in each province. This represents the least squares estimate of the fixed effects coefficient.

[0021] Furthermore, step 3 includes: After normalizing the key factors affecting net carbon sink in agriculture and the provincial variables, we added the net carbon sink in agriculture to obtain a dataset including training and test sets. The training set was input into decision tree model, random forest model, support vector machine model, linear regression model, fully connected neural network, boosting tree model and k-nearest neighbor algorithm for prediction. The prediction results were verified by five-fold cross-validation, R² evaluation criterion and error test. Based on the verification results, random forest model was determined as the target prediction model. The training set is input into the random forest model to train the random forest.

[0022] Furthermore, the parameters of the random forest model include the number of decision trees, the number of features, and the maximum depth.

[0023] The present invention also provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a method for predicting net carbon sinks in agriculture.

[0024] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements a method for predicting net carbon sinks in agriculture.

[0025] The above-described solution of the present invention has the following beneficial effects: This invention calculates the net agricultural carbon sink using acquired carbon emission and crop data. It then uses Lasso regression and ISM algorithms to screen factors influencing the net agricultural carbon sink, identifying key influencing factors. Using the net agricultural carbon sink, key influencing factors, and provincial variables obtained through analysis of these factors, a dataset including a training set and a test set is constructed. The training set is then input into a random forest model for training. During the training process, the optimal parameter combination for the random forest model is optimized using the Grey Wolf optimization algorithm. The parameters in the random forest model are then set according to this optimal parameter combination, and the results are analyzed using... The trial set evaluates the random forest model with optimal parameter combinations to obtain a net carbon sink prediction model. The important factors affecting agricultural net carbon sinks and provincial variables of the target region are input into the net carbon sink prediction model for prediction, yielding the net carbon sink prediction results for the target region. Compared with existing technologies, this invention uses two screening methods—Lasso regression and ISM algorithm—to screen factors affecting agricultural net carbon sinks, ensuring the scientific validity and reliability of the screening results. The important factors affecting agricultural net carbon sinks and provincial variables of the target region are input into the random forest model optimized by the Grey Wolf optimization algorithm for prediction, obtaining the net carbon sink prediction results for the target region, thus improving the accuracy and stability of agricultural net carbon sink prediction.

[0026] Furthermore, the ISM algorithm used in this invention is an optimized version of the ISM algorithm. By using the optimized ISM algorithm to screen factors affecting agricultural net carbon sinks, the screening results are combined with the results screened by the Lasso regression algorithm to obtain the important factors affecting agricultural net carbon sinks, which can effectively improve the prediction accuracy of net carbon sinks.

[0027] Other beneficial effects of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating an embodiment of the present invention; Figure 2 This is a Lasso regression coefficient path diagram in an embodiment of the present invention; Figure 3 This is a schematic diagram of the five-fold cross-validation results in an embodiment of the present invention; Figure 4 (a) shows the prediction results of the test set in the random forest model, (b) shows the prediction results of the test set in the boosting tree model, (c) shows the prediction results of the test set in the k-nearest neighbor algorithm, and (d) shows the error test results of the prediction results of the three models and the actual results. Figure 5 This is an optimization flowchart of the gray wolf optimization algorithm in an embodiment of the present invention; Figure 6 (a) is a schematic diagram of the prediction results of the test set in the parameter-optimized random forest model, and (b) is a comparison diagram of the error test results between the prediction results of the test set in the parameter-optimized random forest model and the actual results. Figure 7 This is a schematic diagram illustrating the verification results of the 10-fold cross-validation method and the learning curve method in an embodiment of the present invention; Figure 8 This is a schematic diagram of the structure of the terminal device in an embodiment of the present invention. Detailed Implementation

[0029] To make the technical problems, solutions, and advantages of this invention clearer, a detailed description will be provided below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0030] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0031] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a locking connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0032] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0033] This invention addresses existing problems by providing a method and related equipment for predicting net carbon sinks in agriculture.

[0034] like Figure 1 As shown, an embodiment of the present invention provides a method for predicting net carbon sinks in agriculture, comprising: Step 1: Calculate the net carbon sink in agriculture using the obtained carbon emission data and crop data to obtain the net carbon sink in agriculture. Step 2: Use the Lasso regression algorithm and the ISM algorithm to screen the factors affecting agricultural net carbon sinks and obtain the important factors affecting agricultural net carbon sinks. Step 3: Using the net carbon sink in agriculture, the important factors affecting the net carbon sink in agriculture, and the provincial variables obtained by analyzing the factors affecting the net carbon sink in agriculture, construct a dataset including a training set and a test set, and input the training set into the random forest model to train the random forest. Step 4: During the training process of the random forest model, the optimal parameter combination of the random forest model is optimized by the gray wolf optimization algorithm. The parameters in the random forest model are set according to the optimal parameter combination, and the random forest model with the optimal parameter combination is evaluated using the test set to obtain the net carbon sink prediction model. Step 5: Input the key factors affecting agricultural net carbon sinks and provincial variables obtained from the target area into the net carbon sink prediction model for prediction, and obtain the net carbon sink prediction results for the target area.

[0035] Specifically, step 1 includes: Agricultural carbon emissions were calculated using carbon emission data from 2000 to 2022 for each region. The total agricultural carbon emissions for each region from 2000 to 2022 were calculated using the following expression:

[0036] in, This represents the total carbon emissions from agriculture. Indicates the first Carbon emissions from carbon-like sources Indicates the first Carbon source data (also known as activity level data). Indicates the first Emission coefficients of carbon-like sources; In this embodiment of the invention, the carbon emission data for each region from 2000 to 2022, including the carbon source type and the corresponding carbon source data, emission coefficient, and carbon emission amount for each carbon source type, are shown in Table 1 below: Table 1 Carbon Emission Data

[0037] It should be noted that the carbon source types in this embodiment of the invention include, but are not limited to, agricultural machinery, crop irrigation, fertilizers, pesticides, agricultural films, and farmland tillage listed in Table 1 above. In practical applications, crop planting and livestock breeding can also be included. When calculating the carbon emissions from crop planting and livestock breeding, it is first necessary to calculate the emission coefficients and corresponding carbon source data for crop planting and livestock breeding based on the formula in the paper "Spatiotemporal Differentiation Characteristics and Driving Forces of Net Carbon Sinks from Conservation Tillage in China" to obtain the emissions of CH4 and N2O. The carbon source data for crop planting is the rice planting area at different growth stages, and the carbon source data for livestock breeding is the number of livestock slaughtered and the number of livestock in stock during the year. Then, according to the standard: the greenhouse effect produced by 1 ton of CH4 and N2O is equivalent to 6.8182 tons of standard carbon and 81.2727 tons of standard carbon, respectively, the emissions of CH4 and N2O are converted. Finally, the two conversion results are added together to obtain the carbon emissions.

[0038] Using crop data from 2000 to 2022, the carbon uptake of crops in each region was calculated, yielding the total carbon uptake of crops in each region from 2000 to 2022. The calculation expression is as follows:

[0039] in, This indicates the total amount of carbon absorbed by crops. Indicates the first Carbon absorption rate of crops Indicates the first Biomass yield of crop-like crops Indicates the first Economic yield of similar crops Indicates the first Economic coefficients of similar crops Indicates the first The water content of similar crops; In this embodiment of the invention, the total carbon absorption of crops is referred to as agricultural carbon sink. Agricultural carbon sink broadly includes forest carbon sink, grassland carbon sink, arable land carbon sink, and soil carbon sink. Existing technologies define agricultural carbon sink as the fixation of atmospheric carbon by crops through photosynthesis during their growth process, i.e., carbon sink measurement. In this embodiment of the invention, crop data includes crop types and their corresponding economic coefficients, carbon absorption rates, water content, and economic yields corresponding to crop types obtained from the *China Rural Statistical Yearbook* and *China Agricultural Statistical Data*. This embodiment of the invention only shows the specific numerical values ​​of crop data as shown in Table 2. Table 2. Crop Types and Their Corresponding Economic Coefficients, Carbon Absorption Rates, and Moisture Contents

[0040] Since net carbon sinks comprehensively reflect the balance between carbon emissions and carbon absorption, and are also an important indicator of regional emission reduction efficiency, this embodiment of the invention calculates the net agricultural carbon sinks of each region from 2000 to 2022 based on the total agricultural carbon emissions and total carbon absorption by crops. The calculation expression is as follows:

[0041] in, Indicates net carbon sequestration in agriculture, when A value greater than 0 indicates that the carbon sink is greater than the carbon emissions; when When the carbon sink is less than the carbon emissions; when At that time, carbon sinks and carbon emissions were equal.

[0042] Specifically, step 2 includes: The factors affecting agricultural net carbon sinks were standardized to obtain the standardized factors affecting agricultural net carbon sinks. The Lasso regression algorithm was used to perform Lasso regression on net carbon sink and standardized factors affecting agricultural net carbon sink to obtain the first screening results. The survey results of agricultural production areas were analyzed to identify the factors affecting net carbon sinks in agriculture and their relative importance. An artificial intelligence model was used to score these factors, resulting in an initial direct impact matrix. A reachability matrix consisting of 0s and 1s is generated based on the initial direct influence matrix. The reachability matrix includes multiple rows and columns. The reachability matrix is ​​hierarchically partitioned using the ISM algorithm to determine the second screening result; The union of the first and second screening results yields the key factors influencing net carbon sinks in agriculture.

[0043] In the embodiments of the present invention, as shown in Table 3, the factors affecting net carbon sinks in agriculture include social factors, environmental factors, and agricultural factors. Among them, social factors include per capita education level, regional economic development level, urbanization level, per capita agricultural output value, and industrial structure; environmental factors include temperature, precipitation, and sunshine; and agricultural factors include agricultural mechanization level, cultivated land area, fertilizer application intensity, and effective irrigation area.

[0044] Table 3 Factors Affecting Net Carbon Sink in Agriculture

[0045] In this embodiment of the invention, the factors affecting agricultural net carbon sinks in Table 3 above can be obtained by compiling and extracting publicly available data from the Rural Statistical Yearbook, Agricultural Machinery Industry Yearbook, local statistical yearbooks and meteorological data center websites from 2000 to 2022.

[0046] In this embodiment of the invention, the factors affecting agricultural net carbon sinks are standardized using the Min-Max normalization method to obtain the standardized factors affecting agricultural net carbon sinks.

[0047] In this embodiment of the invention, the Lasso regression algorithm is an extension of a linear regression model. When the regularization parameter exceeds a certain threshold, the regression coefficients of some redundant or irrelevant features will be compressed to zero, thereby automatically filtering out key variables. This method can not only reduce the complexity of the model, but also improve the accuracy of prediction to a certain extent.

[0048] Specifically, the Lasso regression algorithm was used to perform Lasso regression on the factors affecting net carbon sink in agriculture and the standardization of these factors. This yielded a first screening result, which identified key factors with non-zero coefficients affecting net carbon sink in agriculture. By observing the coefficients of these factors under different regularization parameters, the value of the regularization parameter was determined. In the case of selection, such as Figure 2As shown, key factors influencing net carbon sinks in agriculture include effective irrigated area of ​​farmland, level of agricultural mechanization, arable land area, fertilizer application intensity, and urbanization level.

[0049] In this embodiment of the invention, to enable the Lasso regression algorithm model to automatically select key features and control complexity, the Lasso regression algorithm model can be represented as minimizing the objective function. :

[0050] in, Indicates the first The response variable for each observation is the net carbon sink in agriculture. Indicates the first The first observation value There are several predictor variables, which are the factors that affect net carbon sinks in agriculture. Indicates the first The coefficients of the predictor variables, Indicates the number of samples. Indicates the number of predictor variables. This represents the regularization parameter, used to control the strength of the penalty term.

[0051] In this embodiment of the invention, the survey results of agricultural production areas are analyzed to obtain the factors affecting net carbon sinks in agriculture and their relative importance. An artificial intelligence model is then used to score these factors, resulting in an initial direct impact matrix, which specifically includes: The survey results of agricultural production areas were analyzed, and the following table shows the factors affecting net carbon sinks in agriculture, as shown in Table 4 below, and the relative importance of the factors affecting net carbon sinks in agriculture, as shown in Table 5 below. Artificial intelligence models, such as Kimi, Chatgpt4o, and Doubao, were used to score the factors influencing net agricultural carbon sinks based on the interaction evaluation criteria (the evaluation criteria are: the interaction relationship between factors influencing net agricultural carbon sinks is scored on a scale of 4, 3, 2, 1, and 0 respectively, indicating strong, relatively strong, moderate, weak, and none) and the relative importance of the factors influencing net agricultural carbon sinks as shown in Table 5 below. This process was used to construct an initial direct influence matrix. ,in Indicates the first The first artificial intelligence model evaluation Factors affecting agricultural net carbon sinks For the Factors affecting agricultural net carbon sinks The direct impact intensity, i.e., the evaluation of the first Factors affecting agricultural net carbon sinks With the Factors affecting agricultural net carbon sinks The relative importance, and These represent the number of artificial intelligence models and the number of factors affecting net carbon sinks in agriculture, respectively. hour, .

[0052] It should be noted that, in this embodiment of the invention, the survey results of agricultural production areas were first sorted out to obtain multiple factors affecting agricultural net carbon sinks and their relative importance as shown in Tables 4 and 5 below. Then, based on Tables 4 and 5 below, the factors affecting agricultural net carbon sinks as shown in Table 3 were selected, including social factors, environmental factors, and agricultural factors.

[0053] Table 4. Summary of Factors Affecting Net Carbon Sequestration in Agriculture

[0054] Table 5. Comparison of the relative importance of factors affecting net carbon sinks in agriculture.

[0055] In this embodiment of the invention, the above-mentioned agricultural production area survey results are as follows: For example, the analysis of the spatiotemporal evolution characteristics and driving factors of arable land carbon source and sink in a major grain-producing area can be "Analysis of the spatiotemporal evolution characteristics and driving factors of arable land carbon source and sink in the major grain-producing area of ​​the middle reaches of the Yangtze River"; the analysis of the spatiotemporal differentiation and evolution of influencing factors of net agricultural carbon sink in a certain area can be "Analysis of the spatiotemporal differentiation and evolution of influencing factors of net agricultural carbon sink in China"; the spatiotemporal distribution and driving factors of net agricultural carbon sink in a certain agricultural area can be "The spatiotemporal distribution and driving factors of net agricultural carbon sink in Northeast China"; the spatiotemporal differentiation characteristics and driving force of net carbon sink of conservation tillage in a certain area can be "The spatiotemporal differentiation characteristics and driving force of net carbon sink of conservation tillage in China"; and the agricultural carbon emission of a certain province... Research on the spatiotemporal evolution characteristics and influencing factors of agricultural carbon emissions could be titled "Research on the Spatiotemporal Evolution Characteristics and Influencing Factors of Agricultural Carbon Emissions in Sichuan Province". Analysis of the spatiotemporal evolution and driving factors of net carbon sinks in rice production in a certain province from the perspective of externalities could be titled "Analysis of the Spatiotemporal Evolution and Driving Factors of Net Carbon Sinks in Rice Production in Hunan Province from the Perspective of Externalities". The spatiotemporal pattern of net carbon sinks in conservation tillage in a certain region could be titled "The Spatiotemporal Pattern of Net Carbon Sinks in Conservation Tillage in China". The spatiotemporal pattern and driving factors of net carbon sinks in crop production in a certain region could be titled "The Spatiotemporal Pattern and Driving Factors of Net Carbon Sinks in Crop Production in a Major Development Area in Western China". Inter-provincial carbon equity in crop production and its driving factors could be titled "Inter-provincial Carbon Equity in Crop Production in China and its Driving Factors".

[0056] This invention achieves more accurate data results by averaging the initial direct influence matrices provided by multiple artificial intelligence models. The calculation expression is as follows:

[0057] Specifically, a reachability matrix consisting of 0s and 1s is generated based on the initial direct influence matrix. The reachability matrix includes multiple rows and columns, including: To ensure the initial direct influence matrix The elements in the range fall within the interval Internally, a standardized direct influence matrix is ​​constructed based on the initial direct influence matrix. The calculation expression is:

[0058] in, Represents the elements of the normalized matrix. This represents the result obtained by taking the average score of all artificial intelligence models. This indicates the maximum value that initially directly affects all elements in the matrix. This indicates the number of factors affecting net carbon sinks in agriculture; To eliminate the potential chain reactions caused by changes in various factors, it is necessary to consider the synergistic effects between factors, including both direct and indirect influences. This requires constructing a comprehensive influence matrix to reflect the cumulative effect of direct and indirect influences among factors. The purpose is to determine the overall impact of each factor on the most influential factor in the system. Therefore, a comprehensive influence matrix is ​​constructed based on a standardized direct influence matrix. The calculation expression is:

[0059] To account for the individual influence of each factor as well as the interactions between factors, the overall influence matrix is ​​incorporated into the identity matrix. The global influence matrix is ​​obtained. The calculation expression is:

[0060] Global influence matrix Set threshold =0.3, global influence matrix The value is greater than or equal to the threshold. Assigning the value 1 to the elements will change the global influence matrix. Medium less than the threshold The elements are assigned the value 0, resulting in a reachability matrix consisting of 0s and 1s. reachable matrix Including multiple rows and columns, for example The reachability matrix can be represented as , This indicates the number of factors affecting net carbon sinks in agriculture.

[0061] As the number of influencing factors increases, the computational complexity also increases, which is inconvenient in practical applications. To solve this problem, this embodiment of the invention uses an ISM algorithm (i.e., the classical interpretive structural model) for hierarchical partitioning. The basic idea is to use the reachability matrix... The sum of the elements in each row represents the influence of that row's elements on the other elements; this is represented by a reachability matrix. The sum of each column element in the table represents the degree to which that column element is affected by other elements.

[0062] Specifically, the ISM algorithm is used to hierarchically partition the reachability matrix to determine the second screening results, including: The expression for hierarchical partitioning of the reachability matrix using the ISM algorithm is as follows:

[0063]

[0064] in, Represents the th in the reachability matrix The sum of the elements in the row, the first row The row element corresponds to the first row. Factors affecting agricultural net carbon sinks , Represents the th in the reachability matrix The sum of the elements in the column, the first The column element corresponds to the first Factors affecting agricultural net carbon sinks , Represents the th in the reachability matrix Line 1 Column elements; According to the reachability matrix, the first Sum of row elements With the reachability matrix of the th Sum of column elements Determine the results of the second screening.

[0065] In the embodiments of the present invention, in obtaining and After obtaining the numerical value, the matrix will be accessible. Sum of factors in each row Sort the data from largest to smallest and from top to bottom, then arrange the column factors according to the row factor order, thus obtaining a reordered reachable matrix. As shown in Table 6 below. By treating factors whose sum is equal in each row as factors of the same structural level, we can obtain the hierarchical structure of factors, and thus the reachable matrix can be represented by the first factor. The sum of the elements in a row is larger (larger means the sum of the elements in the row of the reachable matrix is ​​the largest). The sum of the row elements is greater than or equal to a preset threshold. The preset threshold is set according to the number of factors that need to influence the net carbon sink in agriculture. In this embodiment of the invention, the number of factors that need to influence the net carbon sink in agriculture is 6, that is, the reordered reachability matrix is ​​taken. middle Factors affecting net agricultural carbon sinks with a value greater than or equal to 5, and the corresponding factors in the reachability matrix. The sum of the column elements is smaller (smaller means the first column in the reachable matrix has the smallest sum of its elements). The sum of the column elements is less than a preset threshold. The preset threshold is determined based on the number of factors that need to influence the net carbon sink in agriculture. In this embodiment of the invention, the number of factors that need to influence the net carbon sink in agriculture is 6, that is, the reordered reachability matrix is ​​taken. middle Factors affecting agricultural net carbon sinks whose values ​​are less than 4 are used as the second screening result. The second screening result includes effective irrigated area of ​​farmland, level of agricultural mechanization, cultivated land area, fertilizer application intensity, regional education level, and urbanization level.

[0066] The ISM algorithm described in this embodiment of the invention is an optimized version of the ISM algorithm. By using the optimized ISM algorithm to screen factors affecting net carbon sinks in agriculture, the screening results are combined with the results screened by the Lasso regression algorithm to obtain the important factors affecting net carbon sinks in agriculture, which can effectively improve the prediction accuracy of net carbon sinks.

[0067] In this embodiment of the invention, the union of the first screening result and the second screening result is used to obtain the important factors affecting agricultural net carbon sink as effective irrigated area of ​​farmland, level of agricultural mechanization, cultivated land area, fertilizer application intensity, regional education level, and urbanization level.

[0068] Table 6 Reordered Reachability Matrix ´

[0069] Specifically, step 3 includes: After normalizing the key factors affecting net carbon sink in agriculture and the provincial variables, we added the net carbon sink in agriculture to obtain a dataset including training and test sets. The training set was input into decision tree model, random forest model, support vector machine model, linear regression model, fully connected neural network, boosting tree model and k-nearest neighbor algorithm for prediction. The prediction results were verified by five-fold cross-validation, R² evaluation criterion and error test. Based on the verification results, random forest model was determined as the target prediction model. The training set is input into the random forest model to train the random forest.

[0070] It should be noted that the provincial variable was obtained by analyzing the factors affecting net carbon sink in agriculture using a fixed effects model. The fixed effects model is a powerful data analysis tool that takes into account the unique characteristics of each province. These characteristics do not change within a certain time range, such as natural resources and geographical location. These fixed characteristics are collectively referred to as "provincial differences." They exist among provinces but remain constant over time. By adding a fixed effects variable to each province, individual provincial differences can be considered, thereby improving the accuracy of predictions.

[0071] Specifically, by analyzing the factors affecting net carbon sinks in agriculture using a fixed effects model, the calculation expression for the province variable is obtained as follows:

[0072] in, Indicates the first Province variables for each province, Indicates the first The average net agricultural carbon sink of each province Indicates the first The average value of factors affecting net agricultural carbon sinks in each province. This represents the least squares estimate of the fixed effects coefficient.

[0073] This embodiment of the invention uses data from various provinces from 2000 to 2022 as an example. Indicates the first The average net agricultural carbon sink of each province from 2000 to 2022. Indicates the first The average values ​​of various factors affecting net agricultural carbon sinks in 12 provinces from 2000 to 2022. Assuming there are 12 factors affecting net agricultural carbon sinks, the... The average value needs to be calculated for each factor affecting the net carbon sink in agriculture in each province.

[0074] Subsequently, the key factors affecting net carbon sink in agriculture and the provincial variables were normalized and then added to the net carbon sink in agriculture to obtain a dataset including training and testing sets.

[0075] Specifically, the prediction results are validated using the five-fold cross-validation method, the R² evaluation criterion, and the error test method. Based on the validation results, the target prediction model is determined. The validation results are as follows: Figure 3 , Figure 4 As shown, by Figure 3 It can be seen that the random forest model, boosting tree model and k-nearest neighbor algorithm have the highest R² values ​​and the shortest box lengths, which indicates that these three algorithms are the most suitable for completing the predictions in the embodiments of the present invention. The training and test sets in the dataset are split at a ratio of 70% and 30%, respectively, and R² is used as the validation metric to evaluate the results. The computational results and errors of the random forest model, boosting tree model, and k-nearest neighbor algorithm on the test set are as follows: Figure 4 As shown, its parameters are set to default. Figure 4 (a) to (c) show the fitting performance of the three models on the test set; due to the normalization of some data in the dataset, the predicted and true values ​​range from [0,1]. Figure 4 As shown in (a) to (c), the predicted values ​​of the three models are very close to the actual values, and the R² on the test set is greater than 0.96, indicating that the overall model fit is good. Then, the error is tested by calculating the difference between the actual value and the predicted value. The test results are as follows: Figure 4 As shown in (d), the error range on the test set is [ The range is between 0.2 and 0.2, but most data points are concentrated in [...]. The prediction range is [0.1, 0.1]. Generally, a prediction error within 10% is considered excellent, indicating that all three models passed the error test. On the test set, the error range of the random forest model is smaller than that of the k-nearest neighbor algorithm and the boosting tree model, indicating that the random forest model performs better than these algorithms on the test set. According to the indicators shown in Table 7, the MSE and R² of the random forest model are better than those of the k-nearest neighbor algorithm and the boosting tree model, and the MAPE of the random forest model is not significantly different from those of the k-nearest neighbor algorithm and the boosting tree model. In summary, the random forest model shows the best performance on the test set, therefore, the random forest model is chosen as the target prediction model.

[0076] Table 7 Evaluation results of each model

[0077] Specifically, step 4 uses the Grey Wolf optimization algorithm to optimize the optimal parameter combination of the random forest model, including: The parameters of the gray wolf optimization algorithm are set as follows: the total number of wolves is 20, the maximum number of iterations is 10, and the initial coordinates of gray wolf α, β and δ are all [0,0,0]. Set the key parameter ranges for the random forest model: number of trees (n_estimators) [10, 200], maximum tree depth (max_depth) in [5, 50], and number of randomly selected features (max_features) [0.1, 1]. During the training process of the random forest model, the parameters of the random forest algorithm are optimized using the gray wolf optimization algorithm. The optimization process... Figure 5As shown, the optimal parameter combination is determined to be: n_estimators=186, max_depth=20, max_features=0.86.

[0078] This invention demonstrates the testing and evaluation of a parameter-optimized random forest model, yielding the following evaluation results: Figure 6 As shown, the evaluation results on the test set are as follows: Figure 6 As shown in (a), the mean absolute percentage error (MAPE) of the random forest model optimized using the Grey Wolf optimization algorithm is reduced to 7%, which is an improvement compared to the MAPE (11.29%) of the unoptimized random forest model. Furthermore, the R² is increased to 0.98 while maintaining an MSE of 0. Figure 6 (b) shows that the error on the test set is between -0.1 and 0.1, with most points between -0.025 and 0.025, which is better than the random forest model with unoptimized parameters.

[0079] To verify the stability of the model, this invention employs a 10-fold cross-validation method and a learning curve method, yielding the following verification results: Figure 7 As shown, the results of the 10-fold cross-validation are R 2 Using the gray wolf optimization algorithm as an indicator, the learning curves of the optimized random forest model on the training and test sets are plotted, such as... Figure 7 As shown, from Figure 7 As can be seen, with the increase of the number of samples, the R-values ​​of the training and test sets of the three models increase. 2 The difference gradually decreased and then converged to a single value, without underfitting or overfitting. The convergence value of the random forest model optimized by the Grey Wolf optimization algorithm was 0.95. This indicates that the model can maintain a high degree of fit under different data partitions, and has excellent robustness and stability, further verifying the reliability and superiority of the model in repeated experiments. This result shows that the Grey Wolf optimization algorithm is effective in optimizing the parameters of the random forest model and can further improve the prediction accuracy of the model.

[0080] It should be noted that 10-fold cross-validation is a commonly used model evaluation method. The basic idea is to divide the original dataset into 10 subsets of similar size and then conduct 10 experiments. In each experiment, one subset is selected as the test set, and the remaining 9 subsets are used as the training set. By splitting the dataset 10 times, the performance of the model on different data subsets can be better estimated, thereby more accurately evaluating the model's generalization ability. The learning curve shows the relationship between a model's performance on the training and test sets and the number of training samples. By observing the learning curve, one can determine whether the model has overfitting or underfitting issues, and decide whether more data is needed or the model's complexity needs to be adjusted.

[0081] Specifically, step 5 involves inputting the key factors and provincial variables affecting the net carbon sink of agriculture in the target region for the year to be predicted into the net carbon sink prediction model for prediction, and obtaining the net carbon sink prediction result of the target region as the net carbon sink of agriculture in the target region in this year.

[0082] This invention utilizes acquired carbon emission data and crop data to calculate the net agricultural carbon sink, obtaining the net agricultural carbon sink amount. Lasso regression and ISM algorithms are used to screen factors influencing the net agricultural carbon sink, identifying key factors. A dataset including a training set and a test set is constructed using the net agricultural carbon sink amount, key factors influencing the net agricultural carbon sink, and provincial variables obtained through analysis of these factors. The training set is then input into a random forest model for training. During the training process, the optimal parameter combination for the random forest model is optimized using the Grey Wolf optimization algorithm, and the parameters in the random forest model are set according to this optimal parameter combination. The invention utilizes a test set to evaluate a random forest model with optimal parameter combinations, resulting in a net carbon sink prediction model. The important factors influencing agricultural net carbon sinks in the target region and provincial variables are input into the net carbon sink prediction model for prediction, yielding the net carbon sink prediction results for the target region. Compared to existing technologies, this embodiment uses both Lasso regression and ISM algorithms to screen factors influencing agricultural net carbon sinks, ensuring the scientific validity and reliability of the screening results. The input of the important factors influencing agricultural net carbon sinks in the target region and provincial variables into the net carbon sink prediction model for prediction improves the accuracy and stability of agricultural net carbon sink prediction.

[0083] This invention also provides a terminal device, such as... Figure 8 As shown, the terminal device D10 of this embodiment includes: at least one processor D100 ( Figure 8 The diagram shows only one processor, a memory D101, and a computer program D102 stored in the memory D101 and executable on the at least one processor D100. When the processor D100 executes the computer program D102, it implements the above-described method for predicting net carbon sinks in agriculture.

[0084] The terminal device D10 can be a desktop computer, laptop, handheld computer, server, server cluster, or cloud server, etc. This terminal device may include, but is not limited to, a processor D100 and a memory D101. Those skilled in the art will understand that... Figure 8 This is merely an example of terminal device D10 and does not constitute a limitation on terminal device D10. It may include more or fewer components than shown in the figure, or combine certain components, or different components, such as input / output devices, network access devices, etc.

[0085] The processor D100 can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0086] In some embodiments, the memory D101 may be an internal storage unit of the terminal device D10, such as a hard disk or memory of the terminal device D10. In other embodiments, the memory D101 may be an external storage device of the terminal device D10, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the terminal device D10. Furthermore, the memory D101 may include both internal and external storage units of the terminal device D10. The memory D101 is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory D101 can also be used to temporarily store data that has been output or will be output.

[0087] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0088] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0089] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements a method for predicting net carbon sinks in agriculture.

[0090] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to a building device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0091] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An agricultural carbon sink prediction method, characterized by, The method comprises the following steps: Step 1, calculating the agricultural net carbon sink by using the obtained carbon emission data and crop data to obtain the agricultural net carbon sink amount; Step 2, screening the factors affecting the agricultural net carbon sink by using the Lasso regression algorithm and the ISM algorithm to obtain the important factors affecting the agricultural net carbon sink; Step 3, constructing a data set comprising a training set and a test set by using the agricultural net carbon sink amount, the important factors affecting the agricultural net carbon sink, and the province variables obtained by analyzing the factors affecting the agricultural net carbon sink, inputting the training set into a random forest model, and training the random forest; Step 4, in the training process of the random forest model, the optimal parameter combination of the random forest model is optimized by the grey wolf optimization algorithm, the parameters in the random forest model are set according to the optimal parameter combination, and the random forest model with the optimal parameter combination is evaluated by using the test set to obtain a net carbon sink prediction model; Step 5, inputting the obtained important factors affecting the agricultural net carbon sink and the province variables of a target region into the net carbon sink prediction model for prediction to obtain the net carbon sink prediction result of the target region.

2. The agricultural carbon sink prediction method of claim 1, wherein, The step 1 comprises: calculating the agricultural carbon emission amount by using the obtained carbon emission data to obtain the total agricultural carbon emission amount, and the calculation expression is: E = ∑E i = ∑A i x δ i Wherein, E represents the total amount of agricultural carbon emissions, E i represents the carbon emissions of the i-th carbon source, A i represents the i-th carbon source data, δ i represents the emission factor of the i-th carbon source; calculating the crop carbon absorption amount by using the obtained crop data to obtain the total crop carbon absorption amount, and the calculation expression is: C = ∑C j × D j = ∑c j × (1 - r j ) Y j / H j wherein C represents the total carbon absorption amount of the crop, C j represents the carbon absorption rate of the jth crop, D j represents the biological yield of the jth crop, Y j represents the economic yield of the jth crop, H j represents the economic coefficient of the jth crop, r j represents the water content of the jth crop; calculating the agricultural net carbon sink by using the total agricultural carbon emission amount and the total crop carbon absorption amount to obtain the agricultural net carbon sink amount, and the calculation expression is: Z = C - E = ∑C j x (1 - r j ) Y j / H j - ∑A i x δ i wherein Z represents the agricultural net carbon sink amount.

3. The agricultural carbon sink prediction method of claim 1, wherein, The step 2 comprises: standardizing the factors affecting the agricultural net carbon sink to obtain the standardized factors affecting the agricultural net carbon sink; performing Lasso regression on the net carbon sink amount and the standardized factors affecting the agricultural net carbon sink by using the Lasso regression algorithm to obtain a first screening result; combing the agricultural production area investigation results to obtain the factors affecting the agricultural net carbon sink and the relative importance of the factors affecting the agricultural net carbon sink and scoring by using an artificial intelligence model to obtain an initial direct influence matrix; generating a reachable matrix composed of 0 and 1 based on the initial direct influence matrix, wherein the reachable matrix comprises multiple rows and multiple columns; performing hierarchical division on the reachable matrix by using the ISM algorithm to determine a second screening result; taking the union of the first screening result and the second screening result to obtain the important factors affecting the agricultural net carbon sink.

4. The agricultural carbon sink prediction method of claim 3, wherein, The reachable matrix composed of 0 and 1 is generated based on the initial direct influence matrix, and the reachable matrix comprises multiple rows and multiple columns, and comprises: a standardized direct influence matrix D is constructed based on the initial direct influence matrix, and the calculation expression of the standardized direct influence matrix D is: D = (d pq = β pq / max(β pq )) n×n where d pq represents the normalized matrix element, β pq represents the result of taking the mean of all artificial intelligence model scores, max(β pq ) represents the maximum value of all elements in the initial direct influence matrix, and n represents the number of factors affecting the agricultural net carbon sink. a comprehensive influence matrix T is constructed based on the standardized direct influence matrix, and the calculation expression of the comprehensive influence matrix T is: a global influence matrix H is obtained by incorporating a unit matrix I into the comprehensive influence matrix, and the calculation expression of the global influence matrix H is: H = I + T A threshold value η is set for the global influence matrix H, elements greater than or equal to the threshold value η in the global influence matrix H are assigned a value of 1, and elements less than the threshold value η in the global influence matrix H are assigned a value of 0, to obtain a reachable matrix.

5. The agricultural carbon sink prediction method of claim 4, wherein, The reachable matrix is hierarchically divided using the ISM algorithm to determine a second screening result, including: The expression for hierarchically dividing the reachable matrix using the ISM algorithm is: where Q p represents the sum of the elements in the pth row of the reachability matrix, Y q represents the sum of the elements in the qth column of the reachability matrix, k pq represents the element in the pth row and qth column of the reachability matrix; According to the sum of the elements in the pth row of the reachability matrix Q p and the sum of the elements in the qth column of the reachability matrix Y q determining a second screening result.

6. The agricultural carbon sink prediction method of claim 5, wherein, The calculation expression of the province variable obtained by analyzing the factors affecting the agricultural net carbon sink is: wherein, denotes the province variable of the kth province, denotes the average value of the amount of net carbon sink of agriculture in the kth province, denotes the average value of the factors affecting the net carbon sink of agriculture in the kth province, denotes the least square estimate of the fixed effect coefficient.

7. The agricultural carbon sink prediction method of claim 6, wherein, The step 3 includes: The important factors affecting the agricultural net carbon sink and the province variable are normalized and added to the agricultural net carbon sink to obtain a data set including a training set and a test set; The training set is respectively input into a decision tree model, a random forest model, a support vector machine model, a linear regression model, a full connection neural network, a boosting tree model and a k-nearest neighbor algorithm for prediction, and the prediction results are verified by five-fold cross-validation, R 2 The evaluation criteria and error testing method are used to verify the prediction results, and based on the verification results, the random forest model is determined as the target prediction model; The training set is input into the random forest model to train the random forest.

8. The agricultural carbon sink prediction method of claim 7, wherein, The parameters of the random forest model include the number of decision trees, the number of features, and the maximum depth.

9. A terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the agricultural net carbon sink prediction method of any one of claims 1-8.

10. A computer-readable storage medium storing a computer program, the computer-readable storage medium comprising: The computer program is executed by the processor to implement the agricultural net carbon sink prediction method of any one of claims 1-8.

Images