Missing data interpolation-based sparse attention atmospheric carbon concentration prediction method and system

Abstract

The invention provides a sparse attention atmospheric carbon concentration prediction method and system based on missing data interpolation. The method comprises the following steps: acquiring atmospheric carbon data; wherein the atmospheric carbon data comprises two meteorological variables, namely XCO2 and XCH4; adopting a multiple interpolation method based on historical backfill auxiliary data to carry out missing data interpolation on the atmospheric carbon data to obtain a complete data sequence; performing standardization and abnormal value processing on the complete data sequence to obtain an atmospheric carbon data set; and performing atmospheric concentration dual-mode multi-scale prediction based on the atmospheric carbon data set and the sparse attention mechanism atmospheric carbon concentration prediction model to obtain a concentration prediction result. According to the technical scheme, the prediction accuracy of the multi-element dual-carbon data time series can be effectively improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of machine learning, deep learning and time series prediction, and specifically relates to a sparse attention atmospheric carbon concentration prediction method and system based on missing data interpolation. Background Art

[0002] As the global carbon emission problem intensifies, under the goal of carbon neutrality, the study of changes in the concentration of carbon products in the atmosphere has become increasingly urgent. Carbon dioxide (CO2) and methane (CH4) are important greenhouse gases that have a significant impact on global climate change. XCO2 and XCH4 are parameters that describe the concentrations of XCO2 and XCH4 in the atmosphere, usually expressed as volume fractions, and are used to monitor and assess the impact of the global carbon cycle, climate change, and greenhouse gas emissions. TCCON (Total Carbon Column Observing Network) is currently the most widely used and recognized monitoring network, providing researchers with an important reference for understanding the spatiotemporal changes of XCO2 and XCH4 in the atmosphere, and is used to monitor greenhouse gas emissions and assess the global carbon cycle.

[0003] Transformer has a wide range of applications in time series forecasting, but Transformer has poor ability to capture underlying sequence representations and describe multivariate correlations, which limits its ability and generalization to different time series data. To address this issue, Tsinghua University recently proposed a new Transformer-like model called iTransformer, which combines a more variable-centric representation of global time series labels and a multivariate correlation attention mechanism to significantly improve performance. iTransformer solves the problem of Transformer's insufficient ability to predict long time series.

[0004] The present invention improves the multivariate attention component of the iTransformer model, obtains a sparse attention mechanism atmospheric carbon concentration prediction model with higher performance, and obtains more accurate atmospheric carbon concentration prediction results. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a sparse attention atmospheric carbon concentration prediction method and system based on missing data interpolation, which effectively improves the accuracy of multivariate dual-carbon data time series prediction.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A sparse attention atmospheric carbon concentration prediction method based on missing data interpolation includes the following steps:

[0008] Collect atmospheric carbon data; wherein the atmospheric carbon data includes two meteorological variables XCO2 and XCH4;

[0009] A multiple interpolation method based on historical backfill auxiliary data is used to interpolate missing data of the atmospheric carbon data to obtain a complete data series;

[0010] Standardizing the complete data series and processing outliers to obtain an atmospheric carbon data set;

[0011] Based on the atmospheric carbon dataset and the sparse attention mechanism atmospheric carbon concentration prediction model, a dual-mode multi-scale prediction of atmospheric concentration is performed to obtain concentration prediction results.

[0012] Preferably, the method for interpolating missing data of the atmospheric carbon data to obtain a complete data sequence is:

[0013] Identify missing data segments and adjacent complete data segments of the atmospheric carbon data;

[0014] Based on the time tags of the identified data segments, the corresponding historical complete data sequence is obtained;

[0015] Based on dynamic time warping, the similarity between the missing data segment and the corresponding historical complete data sequence is calculated;

[0016] Calculating interpolation results of the missing data segments based on a multiple interpolation method;

[0017] Based on the similarity and the interpolation results, the final interpolation results of the missing data segments are obtained;

[0018] Based on the final interpolation result, the complete data sequence is obtained.

[0019] Preferably, the method for obtaining the atmospheric carbon data set is:

[0020] Converting the complete data sequence into a uniform daily granularity scale and performing averaging processing on the daily atmospheric carbon data to obtain an average value sequence;

[0021] Performing outlier detection and removal on the mean value sequence;

[0022] Based on the average value sequence with outliers removed, an atmospheric carbon data set is obtained, and the atmospheric carbon data set is divided into a training set, a validation set, and a test set.

[0023] Preferably, the probabilistic sparse self-attention method is used to improve the self-attention module of the iTransformer model to obtain the sparse attention mechanism atmospheric carbon concentration prediction model; the method for constructing the sparse attention mechanism atmospheric carbon concentration prediction model is:

[0024] The original time series of different meteorological variables in the training set are independently embedded into tokens;

[0025] Based on probabilistic sparse self-attention, the embedded original time series is uniformly sampled to select local context;

[0026] Based on the selection of the local context, compressing the original time series into a short time series of a preset length;

[0027] Based on the short time series, calculating a sparse attention weight matrix;

[0028] Based on the sparse attention weight matrix, sparse attention is performed on the local neighborhood in the short time series to obtain local neighborhood related information;

[0029] Based on the local neighborhood related information, the periodicity and trend information of the original time series of different meteorological variables and the dependency between multiple variables are captured to complete the construction of the atmospheric carbon concentration prediction model with a sparse attention mechanism.

[0030] Preferably, when performing dual-mode multi-scale prediction of atmospheric concentration, the dual mode includes multivariable prediction of single variable and multivariable prediction of multivariable.

[0031] The present invention also provides a sparse attention atmospheric carbon concentration prediction system based on missing data interpolation, which is used to implement the method, including:

[0032] A data acquisition module, used to collect atmospheric carbon data; wherein the atmospheric carbon data includes two meteorological variables XCO2 and XCH4;

[0033] An interpolation module, for interpolating missing data of the atmospheric carbon data using a multiple interpolation method based on historical backfill auxiliary data to obtain a complete data sequence;

[0034] A data set construction module, used for standardizing the complete data sequence and processing outliers to obtain an atmospheric carbon data set;

[0035] The multi-scale prediction module is used to perform dual-mode multi-scale prediction of atmospheric concentration based on the atmospheric carbon data set and the sparse attention mechanism atmospheric carbon concentration prediction model to obtain concentration prediction results.

[0036] Preferably, the interpolation module comprises:

[0037] A data segment identification unit, used to identify missing data segments and adjacent complete data segments of the atmospheric carbon data;

[0038] A historical data acquisition unit, used to obtain a corresponding historical complete data sequence based on the time tags of the identified data segments;

[0039] A similarity calculation unit, used for calculating the similarity between the missing data segment and the corresponding historical complete data sequence based on dynamic time warping;

[0040] An interpolation calculation unit, used for calculating the interpolation result of the missing data segment based on a multiple interpolation method;

[0041] The final interpolation acquisition unit is used to obtain the final interpolation result of the missing data segment based on the similarity and the interpolation result; and obtain the complete data sequence based on the final interpolation result.

[0042] Preferably, the data set construction module includes:

[0043] A data averaging unit, used to convert the complete data sequence into a unified daily granularity scale and perform averaging processing on the daily atmospheric carbon data to obtain an average value sequence;

[0044] An outlier detection unit, used for detecting and removing outliers from the mean value sequence;

[0045] The data set construction unit is used to obtain an atmospheric carbon data set based on a mean value sequence with outliers removed, and divide the atmospheric carbon data set into a training set, a validation set, and a test set.

[0046] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention proposes a multiple interpolation method based on historical backfill auxiliary data to solve the missing value characteristics inherent in carbon concentration time series data. The uncertainty of missing data can be retained to make the results more reliable. Compared with a single interpolation method, it can better reflect the randomness of missing data and is suitable for various types of missing patterns and statistical analysis methods. The sparse attention mechanism atmospheric carbon concentration prediction model constructed for accurate, multi-window carbon concentration prediction can capture long-term data dependencies. The complexity of the model is low, so that carbon time series data can be more effectively predicted in real time. The present invention achieves high-precision prediction of atmospheric carbon data and effectively solves the problems of computational complexity and real-time requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0048] Figure 1 The present invention is a flowchart of a sparse attention atmospheric carbon concentration prediction method based on missing data interpolation according to an embodiment of the present invention.

[0049] Figure 2A schematic diagram of a multiple interpolation method for backfilling auxiliary data of history according to an embodiment of the present invention;

[0050] Figure 3 This is a schematic diagram of an atmospheric carbon concentration prediction model using a sparse attention mechanism according to an embodiment of the present invention;

[0051] Figure 4 This is a schematic diagram of the structure of a sparse attention atmospheric carbon concentration prediction system based on missing data interpolation according to an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0054] Embodiment 1

[0055] like Figure 1 As shown, a sparse attention atmospheric carbon concentration prediction method based on missing data interpolation includes the following steps:

[0056] S1: Collect atmospheric carbon data. In this embodiment, the data of TCCON ground detection sites are collected: Lamont, Edwards, and Parkfalls. Atmospheric carbon data includes two meteorological variables, XCO2 and XCH4; in this embodiment, XCO2 and XCH4 (the number of molecules of carbon dioxide and methane per million air molecules) data are sampled and recorded at the same time interval and frequency. However, the amount of data is different every day, and there are abnormal data exceeding the empirical threshold and a few dates are missing. Therefore, the collected atmospheric carbon data needs to be preprocessed.

[0057] S2: Use multiple interpolation methods based on historical backfill auxiliary data to interpolate missing data of atmospheric carbon data and obtain a complete data series. Figure 2 shown.

[0058] During the data analysis process, it is observed that the same time period in different years presents a similar distribution trend. Therefore, the present invention believes that this characteristic can be used to obtain reference values ​​for missing data during the substitution process. Based on this observation, the present invention proposes a historical review method assisted by multiple interpolation.

[0059] A further implementation method is to interpolate missing data of atmospheric carbon data to obtain a complete data sequence by:

[0060] Identify missing data segments and adjacent complete data segments for atmospheric carbon data;

[0061] Based on the time tags of the identified data segments, the corresponding historical complete data sequence is obtained; specifically, the corresponding historical data is found from the same time period in previous years. The sequence with missing data is called a missing data sequence, and the sequence with complete historical data is called a historical complete data sequence.

[0062] Based on dynamic time warping (DTW), the similarity C between the missing data segment i and the corresponding historical complete data sequence j is calculated ij , and obtain the historical corresponding value of the missing position (time t) of the j sequence.

[0063] Calculate the interpolation result M of the missing data segment based on the multiple interpolation method i ; Specifically, when using spline interpolation, a cubic spline function can be constructed to fit the data and interpolate the missing points.

[0064] Based on the similarity and the interpolation results, the final interpolation results of the missing data segments are obtained;

[0065] Based on the final interpolation result X, a complete data sequence is obtained. Specifically, the interpolation result of the current data is combined with the interpolation result of the historical data by using the weighted average method. The calculation equation is as follows:

[0066] X=W1(i * C ij )+W2(M i ).

[0067] W1: represents the weight of the interpolation part based on similarity, where i * C ij It is the historical sequence His-Len i and His-Len j The similarity scores between them are calculated by dynamic time warping (DTW). W1 emphasizes the contribution of historical pattern similarity to the interpolation process.

[0068] W2: represents the interpolation value M generated from the multiple interpolation method i The weight of W2 controls the impact of the prediction interpolation technique on the result.

[0069] In this embodiment, a grid search method is used to determine the optimal values ​​of W1 and W2. The goal of this search is to balance the contributions of similarity-based interpolation and multiple interpolation methods.

[0070] Grid search parameters: define a range for W1 and W2, both weight values ​​should be between 0 and 1, and satisfy W1+W2=1 to ensure the balance of the two weights. For example, set W1={0.1, 0.2, 0.3, ... 0.9} and W2=1-W1.

[0071] Grid search execution: Iterate the preset W1 and W2 weight combinations. For each set of W1 and W2 combinations, use the following formula to calculate the final interpolation result: X = W1 (i * C ij )+W2(M i ).

[0073] In this embodiment, in order to ensure the accuracy and reliability of the interpolation result, the interpolation result needs to be verified and optimized.

[0074] Cross-validation: Divide the data set into a training set and a validation set, use the training set for interpolation calculations, and evaluate the interpolation effect on the validation set. Through cross-validation, the generalization ability of the interpolation method can be tested.

[0075] Error analysis: Calculate the error between the interpolation result and the real data, such as mean square error (MSE), mean absolute error (MAE), etc. According to the error analysis results, adjust the interpolation method or parameters to improve the interpolation accuracy.

[0076] Visualization of interpolation results: By drawing a comparison chart of the interpolation results and the real data, you can intuitively evaluate the interpolation effect. Draw a time series chart before and after interpolation to observe whether the interpolation results are reasonable and smooth.

[0077] S3: Standardize the complete data sequence and process outliers to obtain an atmospheric carbon data set; a further implementation method is that the method for obtaining the atmospheric carbon data set is:

[0078] The complete data sequence is converted into a unified daily granularity scale and the daily atmospheric carbon data is averaged to obtain an average value sequence; specifically, carbon data will be collected multiple times a day, and the daily data will be averaged for better data analysis, retaining the daily average as the only data at a given time point.

[0079] Detect and remove outliers from the mean value sequence; specifically, use statistical methods or machine learning algorithms (such as IQR method, Z-score, DBSCAN, etc.) to detect outliers to avoid negative impact of these outliers on model training.

[0080] Based on the average value sequence with outliers removed, the atmospheric carbon data set was obtained and divided into training set, validation set and test set in a ratio of 7:2:1. The training set is used to train the model, the validation set is used to adjust the model parameters, and the test set is used to evaluate the generalization ability of the model.

[0081] S4: Based on the atmospheric carbon dataset and the sparse attention mechanism atmospheric carbon concentration prediction model, dual-mode multi-scale prediction of atmospheric concentration is performed to obtain concentration prediction results.

[0082] The overall structure of the proposed C-TimeSeries (sparse attention mechanism atmospheric carbon concentration prediction model) has the same module layout as the Transformer encoder. In this model, the original sequences of different variables are embedded into tokens independently. This module reveals the multivariate correlations between variables by applying a multi-head self-attention mechanism to the embedded variable tokens.

[0083] like Figure 2 As shown in the figure, the constructed C-TimeSeries deep learning model (sparse attention mechanism atmospheric carbon concentration prediction model) model can capture the periodicity, trend information and complex dependencies between multiple variables in time series data. The C-TimeSeries model uses an inverted version of the feedforward network and applies it to channels of different tokens to extract complex features in time series. The observed time series is encoded by superimposing inverted blocks, and dense nonlinear connections are used to decode the features of future sequences. The sparse self-attention method is an improved self-attention mechanism that aims to reduce the computational complexity of long sequence processing and maintain efficient performance.

[0084] For independent time series, the same linear operation, i.e., the combination of linear predictor and channel-independent strategy, helps to extract sequence features. The temporal features extracted by multi-layer perceptron (MLP) should be shared across the entire time series. The neurons of MLP are trained to capture the intrinsic properties of any time series (such as amplitude periodicity). This prediction model has advantages over the self-attention mechanism that relies only on time points.

[0085] A further implementation method is to use a probabilistic sparse self-attention method to improve the self-attention module of the iTransformer model to obtain a sparse attention mechanism atmospheric carbon concentration prediction model; the method for constructing a sparse attention mechanism atmospheric carbon concentration prediction model is:

[0086] The original time series of different meteorological variables in the training set are embedded into tokens independently;

[0087] Based on probabilistic sparse self-attention, the embedded original time series is uniformly sampled to select local context;

[0088] Based on the selection of local context, the original time series is compressed into a short time series of preset length; In this way, the long input sequence is compressed into a shorter fixed-length representation while reducing the amount of computation.

[0089] Based on short time series, a sparse attention weight matrix is ​​calculated; specifically, most of the elements of the matrix are zero. This sparse matrix can effectively reduce computational costs and memory, thereby improving the efficiency of the model in long sequence tasks.

[0090] Based on the sparse attention weight matrix, sparse attention is performed on the local neighborhood in the short time series to obtain local neighborhood related information and reduce the computational complexity.

[0091] Based on the local neighborhood related information, the periodicity and trend information of the original time series of different meteorological variables and the dependency between multiple variables are captured to complete the construction of the atmospheric carbon concentration prediction model with sparse attention mechanism. The present invention realizes the high-precision prediction of atmospheric carbon data and effectively solves the problems of computational complexity and real-time requirements.

[0092] In order to meet the characteristics of carbon data, we also formulated the following requirements:

[0093] Focus on local information: The model should focus on the local information of carbon time series data, such as irregular changes in carbon data records caused by emergencies.

[0094] Capturing long-term dependencies: The model should capture long-term data dependencies, such as seasonal characteristics.

[0095] Reduced complexity: The complexity of the model should be low so that the carbon time series data can be more effectively predicted in real time.

[0096] A further implementation method is that when performing dual-mode multi-scale prediction of atmospheric concentration, the dual mode includes multivariable prediction of single variable and multivariable prediction of multivariable.

[0097] Specifically,

[0098] Multivariate Prediction of Univariate: A predictive model uses multiple input variables to predict a single output variable.

[0099] Univariate Prediction Univariate: A predictive model uses a single input variable to predict a single output variable.

[0100] In both multivariate and univariate modes, the input prediction window is set to 96 steps. After modeling XCO2 and XCH4, the output prediction window is 12, 24, 36, 48, 96, and 192, a total of six prediction scales.

[0101] The experiment selected mean square error (MSE) and mean absolute error (MAE) as the evaluation indicators of the model to quantitatively analyze and evaluate the experimental performance of the model. MSE and MAE are two commonly used prediction error evaluation indicators, which calculate the difference between the predicted value and the true value respectively as the final prediction evaluation indicators of the model.

[0102] In order to better process input and output data, this embodiment adopts three different types of sliding window strategies, namely M, MS and S modes:

[0103] M-mode (multiple variables in, multiple variables out): The forecasting model predicts multiple output variables based on multiple input variables. For example, for multivariable XCO2 and XCH4, the forecasting model predicts the future multivariable XCO2 and XCH4.

[0104] MS mode (multiple variables in, single variable out): The forecasting model uses multiple input variables to predict a single output variable. For example, predicting the future values ​​of XCO2 and XCH4 is based on the past values ​​of multiple variables such as XCO2 and XCH4.

[0105] S-mode (single variable in, single variable out): The forecasting model uses a single input variable to predict a single output variable. For example, forecasting XCO2 and XCH4 for the next few days is based only on the previous XCO2 and XCH4 data respectively.

[0106] Most of the time series prediction studies in recent years use M or MS mode. In this study, TCCON, XCO2 and XCH4 data of three sites are used. There are three types of data in the data set: date, XCO2 and XCH4 data. Therefore, the present invention also selects two processing modes, M and MS, to compare and predict XCO2 and XCH4 of three sites respectively.

[0107] In order to conduct ablation experiments and quantitatively test the effectiveness of the improved model, two models were tested separately. The specific experimental steps are as follows:

[0108] M-mode experiment:

[0109] Input variables: XCO2 and XCH4.

[0110] Output: Joint prediction results of XCO2 and XCH4.

[0111] Experimental set: data from three sites.

[0112] MS mode experiment:

[0113] Input variables: XCO2 and XCH4.

[0114] Output results: predict XCO2 and XCH4 respectively.

[0115] Experimental set: data from three sites.

[0116] Next, the same six experiments were conducted on the proposed C-TimeSeries model based on the XCO2 and XCH4 time series data. In the experiment, the input prediction window (SeqLen) was set to a length of 96, and the output prediction windows of the six scales were 12, 24, 36, 48, 96, and 192, respectively. Corresponding training and testing were performed, and the MSE and MAE values ​​of each group of experiments were calculated to quantitatively evaluate the performance of the model.

[0117] C-TimeSeries helps to understand carbon trends and provide information for management strategies, as shown in Table 1 (Lamont test results), Table 2 (Parkfalls test results), and Table 3 (Edwards test results).

[0118] Table 1

[0119]

[0120] Table 2

[0121]

[0122] Table 3

[0123]

[0124] In practical applications, the present invention proposes a complete set of solutions. For those skilled in the art, they can recognize that the technical solution of the present invention can be implemented by a computer program and its corresponding general hardware platform. Based on this understanding, the technical solution of the present invention, or its contribution to the prior art, can be embodied as a computer program software product, which is stored in the above-mentioned storage medium and contains several instructions for instructing a device including a data processing unit to execute the methods described in various embodiments of the present invention or specific parts of its embodiments.

[0125] Embodiment 2

[0126] like Figure 4 As shown, the present invention also provides a sparse attention atmospheric carbon concentration prediction system based on missing data interpolation, which is used to implement the method, including:

[0127] The data acquisition module is used to collect atmospheric carbon data, wherein the atmospheric carbon data includes two meteorological variables, XCO2 and XCH4;

[0128] The interpolation module is used to interpolate missing data of atmospheric carbon data using multiple interpolation methods based on historical backfill auxiliary data to obtain a complete data series;

[0129] The dataset construction module is used to standardize the complete data series and process outliers to obtain the atmospheric carbon dataset;

[0130] The multi-scale prediction module is used to perform dual-mode multi-scale prediction of atmospheric concentration based on the atmospheric carbon dataset and the sparse attention mechanism atmospheric carbon concentration prediction model to obtain concentration prediction results.

[0131] A further implementation is that the interpolation module comprises:

[0132] A data segment identification unit, used to identify missing data segments and adjacent complete data segments of atmospheric carbon data;

[0133] A historical data acquisition unit, used to obtain a corresponding historical complete data sequence based on the time tags of the identified data segments;

[0134] A similarity calculation unit, used for calculating the similarity between the missing data segment and the corresponding historical complete data sequence based on dynamic time warping;

[0135] An interpolation calculation unit, used for calculating the interpolation result of the missing data segment based on a multiple interpolation method;

[0136] A final interpolation acquisition unit, used to obtain a final interpolation result of the missing data segment based on the similarity and the interpolation result;

[0137] Based on the final interpolation result, a complete data sequence is obtained.

[0138] A further implementation is that the data set construction module includes:

[0139] The data averaging unit is used to convert the complete data series into a unified daily granularity scale and average the daily atmospheric carbon data to obtain a mean value series;

[0140] An outlier detection unit, used to detect and remove outliers from the mean value sequence;

[0141] The data set construction unit is used to obtain an atmospheric carbon data set based on an average value sequence with outliers removed, and divide the atmospheric carbon data set into a training set, a validation set, and a test set.

[0142] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A sparse attention atmospheric carbon concentration prediction method based on missing data interpolation, characterized in that: The following steps are involved: Collect atmospheric carbon data; wherein the atmospheric carbon data includes two meteorological variables XCO2 and XCH4; A multiple interpolation method based on historical backfill auxiliary data is used to interpolate missing data of the atmospheric carbon data to obtain a complete data series; Standardizing the complete data series and processing outliers to obtain an atmospheric carbon data set; Based on the atmospheric carbon dataset and the sparse attention mechanism atmospheric carbon concentration prediction model, a dual-mode multi-scale prediction of atmospheric concentration is performed to obtain concentration prediction results.

2. The sparse attention atmospheric carbon concentration prediction method based on missing data interpolation according to claim 1 is characterized in that: The method for interpolating missing data of the atmospheric carbon data to obtain a complete data series is as follows: Identify missing data segments and adjacent complete data segments of the atmospheric carbon data; Based on the time tags of the identified data segments, the corresponding historical complete data sequence is obtained; Based on dynamic time warping, the similarity between the missing data segment and the corresponding historical complete data sequence is calculated; Calculating interpolation results of the missing data segments based on a multiple interpolation method; Based on the similarity and the interpolation results, the final interpolation results of the missing data segments are obtained; Based on the final interpolation result, the complete data sequence is obtained.

3. The sparse attention atmospheric carbon concentration prediction method based on missing data interpolation according to claim 1 is characterized in that: The method to obtain the atmospheric carbon dataset is: Converting the complete data sequence into a uniform daily granularity scale and performing averaging processing on the daily atmospheric carbon data to obtain an average value sequence; Performing outlier detection and removal on the mean value sequence; Based on the average value sequence with outliers removed, an atmospheric carbon data set is obtained, and the atmospheric carbon data set is divided into a training set, a validation set, and a test set.

4. The sparse attention atmospheric carbon concentration prediction method based on missing data interpolation according to claim 3 is characterized in that: The probabilistic sparse self-attention method is used to improve the self-attention module of the iTransformer model, and the sparse attention mechanism atmospheric carbon concentration prediction model is obtained; the method for constructing the sparse attention mechanism atmospheric carbon concentration prediction model is: The original time series of different meteorological variables in the training set are independently embedded into tokens; Based on probabilistic sparse self-attention, the embedded original time series is uniformly sampled to select local context; Based on the selection of the local context, compressing the original time series into a short time series of a preset length; Based on the short time series, calculating a sparse attention weight matrix; Based on the sparse attention weight matrix, sparse attention is performed on the local neighborhood in the short time series to obtain local neighborhood related information; Based on the local neighborhood related information, the periodicity and trend information of the original time series of different meteorological variables and the dependency between multiple variables are captured to complete the construction of the atmospheric carbon concentration prediction model with a sparse attention mechanism.

5. The sparse attention atmospheric carbon concentration prediction method based on missing data interpolation according to claim 1 is characterized in that: When conducting dual-mode multi-scale prediction of atmospheric concentration, the dual mode includes multivariable prediction of single variable and multivariable prediction of multivariable.

6. A sparse attention atmospheric carbon concentration prediction system based on missing data interpolation, used to implement the method according to any one of claims 1 to 5, characterized in that: include: A data acquisition module, used for collecting atmospheric carbon data; wherein the atmospheric carbon data includes two meteorological variables XCO2 and XCH4; An interpolation module, for interpolating missing data of the atmospheric carbon data using a multiple interpolation method based on historical backfill auxiliary data to obtain a complete data sequence; A data set construction module, used for standardizing the complete data sequence and processing outliers to obtain an atmospheric carbon data set; The multi-scale prediction module is used to perform dual-mode multi-scale prediction of atmospheric concentration based on the atmospheric carbon data set and the sparse attention mechanism atmospheric carbon concentration prediction model to obtain concentration prediction results.

7. The sparse attention atmospheric carbon concentration prediction system based on missing data interpolation according to claim 6 is characterized in that: The interpolation module comprises: A data segment identification unit, used to identify missing data segments and adjacent complete data segments of the atmospheric carbon data; A historical data acquisition unit, used to obtain a corresponding historical complete data sequence based on the time tags of the identified data segments; A similarity calculation unit, used for calculating the similarity between the missing data segment and the corresponding historical complete data sequence based on dynamic time warping; An interpolation calculation unit, used for calculating the interpolation result of the missing data segment based on a multiple interpolation method; The final interpolation acquisition unit is used to obtain the final interpolation result of the missing data segment based on the similarity and the interpolation result; and obtain the complete data sequence based on the final interpolation result.

8. The sparse attention atmospheric carbon concentration prediction system based on missing data interpolation according to claim 6 is characterized in that: The data set construction module includes: A data averaging unit, used to convert the complete data sequence into a unified daily granularity scale and perform averaging processing on the daily atmospheric carbon data to obtain an average value sequence; An outlier detection unit, used for detecting and removing outliers from the mean value sequence; The data set construction unit is used to obtain an atmospheric carbon data set based on a mean value sequence with outliers removed, and divide the atmospheric carbon data set into a training set, a validation set, and a test set.

Images