Bayesian optimization algorithm-based carbon sequestration amount accurate prediction system and method

Through the combination of Bayesian optimization algorithm and XGBoost model, combined with sensors to collect data in real-time, the real-time and accuracy problems in carbon sequestration prediction technology are solved, and efficient carbon sequestration monitoring and prediction are achieved, suitable for forest resource management and environment monitor.

CN118898028APending Publication Date: 2024-11-05NORTHEAST FORESTRY UNIV
View PDF 0 Cites -1 Cited by

Patent Information

Application Number
CN202410972177.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-19
Publication Date
2024-11-05

Smart Images

  • Figure CN118898028A_ABST
    Figure CN118898028A_ABST
Patent Text Reader

Abstract

The invention provides a Bayesian optimization algorithm-based carbon sequestration amount accurate prediction system and method. The system structurally comprises a sensor module, a data processing unit, a display and interaction unit and a prediction model unit, the sensor module is electrically connected with the data processing unit, the data processing unit is electrically connected with the prediction model unit and the display and interaction unit, and the prediction model unit is electrically connected with the display and interaction unit. The advanced sensor technology and the XGBoost machine learning model are integrated, real-time monitoring and processing of environment and growth data are achieved, and compared with the prior art, the system can collect and analyze data in real time, and high-accuracy carbon sequestration amount prediction is provided through the XGBoost model. The strong nonlinear modeling capability of XGBoost is utilized to reflect the influence of environmental change on the carbon sequestration amount in time, so that more accurate data is provided to support forest management and carbon credit evaluation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a system and method for accurately predicting carbon fixation based on a Bayesian optimization algorithm, and belongs to the technical field of vegetation carbon fixation calculation. Background Art

[0002] In recent years, with the increasing awareness of environmental protection and the increasingly prominent impact of climate change on ecosystems, the accurate measurement and prediction of carbon sequestration has become one of the research focuses. As a key indicator for measuring the carbon cycle process of ecosystems, carbon sequestration is of great significance to the health and stability of the ecological environment. At present, the carbon sequestration prediction technology based on climate factors and plant growth data has gradually matured. However, the existing technology still has some shortcomings in real-time data collection and prediction model application.

[0003] In the research of related fields, some carbon sequestration prediction models have been proposed, some of which are based on statistical methods, such as random forest regression model, and others use machine learning algorithms, such as XGBoost algorithm. These technical solutions usually require offline data processing and model training, and have certain limitations in real-time data collection and prediction. The following defects exist in the existing technology: 1. Lack of real-time performance: The existing technology usually relies on offline data processing and model training in carbon sequestration prediction, and cannot realize real-time monitoring and prediction of environmental factors and growth data. 2. Complex data processing: There are problems of complex data processing and high consumption of computing resources in the process of data collection and prediction model application, which limits the feasibility and efficiency of real-time prediction. 3. Low accuracy: Some existing models have certain limitations in the prediction accuracy of carbon sequestration of individual trees. Affected by data quality and model parameter adjustment, it is difficult to achieve high prediction accuracy and stability.

[0004] In summary, in order to overcome the above technical defects, the present invention proposes a system and method for accurate prediction of carbon fixation based on the Bayesian optimization algorithm. By encapsulating an accurate carbon fixation prediction model in the Raspberry Pi and combining sensors to collect environmental factors and plant growth data in real time, real-time monitoring and prediction of carbon fixation can be achieved. Summary of the invention

[0005] The present invention aims to solve the technical problems in the prior art that carbon fixation prediction lacks real-time performance, data processing is complex and prediction accuracy is low, and further proposes a carbon fixation accurate prediction system and method based on a Bayesian optimization algorithm.

[0006] The technical solution adopted by the present invention to solve the above problems is: the present invention proposes a carbon fixation accurate prediction system based on Bayesian optimization algorithm, comprising:

[0007] Sensor module, data processing unit, display and interaction unit and prediction model unit;

[0008] The sensor module is electrically connected to the data processing unit, the data processing unit is electrically connected to the prediction model unit and the display and interaction unit, and the prediction model unit is electrically connected to the display and interaction unit.

[0009] Optionally, the sensor module includes a temperature and humidity sensor, a photoresistor and a linear displacement sensor, and the temperature and humidity sensor, the photoresistor and the linear displacement sensor are all electrically connected to the data processing unit;

[0010] Temperature and humidity sensors are used to measure and record ambient temperature and humidity;

[0011] Photoresistors are used to monitor ambient light intensity;

[0012] Linear displacement sensors are used to measure the radial growth of trees and obtain their growth rate and health status.

[0013] Optionally, the data processing unit is a Raspberry Pi, which is used to pre-process the environment and growth data collected in real time by the sensor module.

[0014] Optionally, the prediction model unit is an XGBoost model, the input of the XGBoost model is tree growth rate, temperature and humidity, and light intensity data, the output is a predicted value of carbon fixation, and the XGBoost model is stored on the GPU of the Raspberry Pi for predicting carbon fixation for preprocessed environment and growth data.

[0015] Optionally, the display and interaction unit includes a display screen and a touch screen / external keyboard;

[0016] The display screen is connected to the data processing unit to display the current carbon fixation prediction results and environmental and growth data in real time;

[0017] The touch screen / external keyboard is used to query the current carbon fixation prediction results, environment and growth data and the historical carbon fixation prediction results, environment and growth data.

[0018] A method for accurately predicting carbon sequestration based on a Bayesian optimization algorithm, comprising:

[0019] Step 1: Start the real-time and accurate prediction system for carbon fixation, and collect environmental and growth data in real time through the temperature and humidity sensor, photoresistor and linear displacement sensor in the sensor module;

[0020] Step 2: Transmit the real-time collected environmental and growth data to the data processing unit for preprocessing;

[0021] Step 3: Train the XGBoost model and optimize its parameters, input the preprocessed environmental and growth data into the trained and optimized XGBoost model to obtain the carbon fixation prediction results;

[0022] Step 4: The carbon fixation prediction results and the environment and growth data are displayed and stored on the display screen in the display and interaction unit. The user exports the carbon fixation prediction results and the pre-processed environment and growth data through the touch screen / external keyboard, and sets the data collection frequency of the sensor module and the carbon fixation warning threshold.

[0023] Optionally, the preprocessing steps in step 2 include:

[0024] Raspberry Pi is used to clean the environment and growth data, remove outliers and noise in the environment and growth data, and normalize the cleaned environment and growth data to complete the preprocessing of the environment and growth data.

[0025] Optionally, the step of obtaining the carbon sequestration prediction result in step 3 includes:

[0026] Step 3.1: Obtain historical carbon fixation data, environmental data, and growth rate data through query;

[0027] Step 3.2: Perform correlation analysis on historical carbon fixation data, environmental data and growth rate data by Pearson correlation analysis method, and retain environmental factors and growth rate factors with Pearson coefficients greater than preset values;

[0028] Step 3.3: Fill missing values ​​and process outliers for environmental factors and growth rate factors, and use the environmental factors and growth rate factors after missing value filling and outlier processing to build the XGBoost model;

[0029] Step 3.4: Optimize the parameters of the XGBoost model through the Bayesian optimization algorithm, use the current preprocessed environmental and growth data as input data, and output the carbon fixation prediction results.

[0030] The steps for optimizing the parameters of the XGBoost model using the Bayesian optimization algorithm in step 3.4 include:

[0031] Step 3.4.1: Use Latin hypercube sampling to obtain uniformly distributed initial points in the parameter space;

[0032] Step 3.4.2: Set n random initial points, define the expected variance and optimal value of the objective function for each random initial point, and select the acquisition function f(x) based on the PI strategy;

[0033] Step 3.4.3: Use BO optimizer to iteratively optimize the acquisition function f(x) to obtain the hyperparameter combination x * ;

[0034] Step 3.4.4: Compare the prediction accuracy of each set of hyperparameter combinations, select the set of hyperparameter combinations with the highest prediction accuracy, and calculate the posterior probability through Gaussian regression;

[0035] Step 3.4.5: Determine the next sampling point through the acquisition function and determine whether the preset accuracy is achieved. If so, output the optimal hyperparameter combination. If not, repeat steps 3.4.1 to 3.4.4 until the preset accuracy is achieved.

[0036] The calculation formula of the acquisition function f(x) is:

[0037]

[0038] In formula (1), Φ is the probability function, μ(x) is the expectation, σ(x) is the variance, and f(x + ) is the optimal value of the objective function, ξ is a set of hyperparameters;

[0039] Hyperparameter combination x * The calculation formula is:

[0040]

[0041] In formula (2), x * is a combination of hyperparameters.

[0042] The beneficial effects of the present invention are:

[0043] 1. This invention integrates advanced sensor technology and XGBoost machine learning models to achieve real-time monitoring and processing of environmental and growth data. Compared with existing technologies, this system can collect and analyze data in real time and provide highly accurate carbon sequestration predictions through the XGBoost model. This is achieved by utilizing XGBoost's powerful nonlinear modeling capabilities to timely reflect the impact of environmental changes on carbon sequestration, thereby providing more accurate data to support forest management and carbon credit assessment.

[0044] 2. All sensors and prediction models are integrated on a single Raspberry Pi platform, using a user-friendly interface for data display and interaction. The present invention greatly simplifies the operation process through automated data collection and processing, reduces the dependence on professional technicians, and reduces the maintenance cost of the system.

[0045] 3. The system design of the present invention takes into account the scalability of data and the compatibility of interfaces, and is easy to add new sensors and update models. The design of the system of the present invention allows for easy integration of more types of sensors and upgraded prediction models to adapt to different application requirements and environmental changes. This flexibility allows the system to continuously adapt to new scientific research developments and market demands, and effectively support forest resource management and environmental monitoring in the long term.

[0046] 4. The present invention uses a graphical user interface (GUI) to directly display real-time data and prediction results on the display screen of the Raspberry Pi. Unlike the traditional method that requires professional equipment or software to view data, the user interface provided by the present invention is intuitive and easy to understand. Even non-professional users can easily operate and interpret data, which significantly improves the user experience and the interactivity of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A structural block diagram of a carbon fixation accurate prediction system based on a Bayesian optimization algorithm provided by the present invention;

[0048] Figure 2 A schematic diagram of data collection of the sensor module provided by the present invention;

[0049] Figure 3 A schematic diagram of the selection and packaging of the carbon fixation prediction model provided by the present invention;

[0050] Figure 4 An integrated diagram of the carbon fixation measurement system provided by the present invention;

[0051] Figure 5 A flow chart of a method for accurately predicting carbon fixation based on a Bayesian optimization algorithm provided by the present invention;

[0052] Figure 6 A comparison chart of the carbon sequestration prediction result and the actual value of carbon sequestration provided by the present invention;

[0053] Figure 7 A heat map of the Pearson correlation analysis between tree growth rate and carbon sequestration provided by the present invention;

[0054] Figure 8 A flowchart of the Bayesian optimization algorithm provided by the present invention;

[0055] Fig. 9 A graph showing the carbon fixation prediction results provided by the present invention;

[0056] In the figure, True Values ​​is the actual value of carbon sequestration, Predicted Values ​​is the predicted value of carbon sequestration, growth is the tree growth rate, and y is the carbon sequestration. Example

[0057] Example 1

[0058] Combination Figure 1-2 This embodiment is described as follows. Figure 1As shown, the structure of a precise carbon fixation prediction system based on the Bayesian optimization algorithm provided in this embodiment includes: a sensor module, a data processing unit, a display and interaction unit and a prediction model unit. The sensor module is electrically connected to the data processing unit, the data processing unit is electrically connected to the prediction model unit and the display and interaction unit, and the prediction model unit is electrically connected to the display and interaction unit.

[0059] like Figure 2 As shown, the sensor module includes a temperature and humidity sensor, a linear displacement sensor and a photoresistor. The temperature sensor is the common DS18B20, with a temperature measurement range of -40°C to 60°C, which is suitable for most areas. The humidity measurement range is 0% RH to 95% RH (non-frosting). The linear displacement sensor is a resistive sensor, which is fixed to the tree trunk by a clamping device and fixed by an inelastic steel rope. It will surround the tree trunk. The inelastic steel rope will drive the linear displacement sensor to change as the tree diameter grows. The resistance change will be converted into the displacement change. The tree diameter change can be obtained through formula conversion, and then the growth rate can be obtained. After collecting factors such as temperature, humidity, light intensity, and growth rate, the data is transmitted to the CPU of the Raspberry Pi.

[0060] Example 2

[0061] Combination Figure 3 and Figure 6 This embodiment is described as follows. Figure 3 As shown, in order to combine machine learning to accurately predict carbon fixation, this embodiment first uses the existing climate factors and growth rate carbon fixation data of the research group to establish an XGBoost prediction model, and optimizes the XGBoost algorithm based on the Bayesian algorithm so that it can achieve an accuracy of 90%, and obtains a carbon fixation prediction model based on BO-XGBoost. The specific steps are as follows:

[0062] Step 1: Organize all existing data.

[0063] Step 2: Use SPSS software to conduct correlation analysis and use the Pearson correlation analysis method to find the climate factor most correlated with carbon sequestration;

[0064] Pearson correlation analysis can calculate multiple sets of data for two values ​​x and y, recorded as (, y) (i = 1, 2, ..., n), and define the Pearson correlation coefficient as the quotient of the covariance and standard deviation of the estimated sample 3. The formula for Pearson correlation analysis is:

[0065]

[0066] In formula (1), x and y are the means of n test values, r~(-1,1), if r>0, then z and y are positively correlated; if r<0, then x and y are negatively correlated, the larger the r number, the more correlated; if r=0, it means that there is no relationship between z and y. Generally, |r|~(0,0.2) is irrelevant or very weakly correlated, |r|~(0.2,0.4) is weakly correlated, |r|~(0.4,0.6) is moderately correlated, |r|~(0.6,0.8) is strongly correlated, and |r|~(0.8,1) is very strongly correlated.

[0067] like Figure 7 As shown, this embodiment uses the Pearson correlation coefficient method to verify the correlation between a model input value and a model output value, taking the most important model input: tree growth rate and model output: carbon fixation as two quantities, Figure 7 The correlation coefficient between tree growth rate and carbon fixation is |r|=0.63~(0.6,0.8), and the tree growth rate and carbon fixation are strongly correlated. At the same time, the correlation coefficient between tree growth rate and carbon fixation is r<0, indicating that there is a negative correlation between the two. Therefore, there is a strong negative correlation between tree growth rate and carbon fixation, which is consistent with the actual relationship between growth rate and carbon fixation, that is, trees with a small ring age have a fast growth rate, but a small wood volume and thus a small carbon fixation; trees with a large ring age have a slow growth rate, but a large wood volume and thus a large carbon fixation. Therefore, the model input and output selection is reasonable, and the model can be further established.

[0068] Step 3: The climate factors with strong correlation obtained in step 2 are temperature, humidity, and light intensity.

[0069] Step 4: The climate factors and growth rate data obtained in step 3 are processed for missing values ​​and outliers to ensure the accuracy of the data. This step can improve the model performance to a certain extent.

[0070] Step 5: After step 4, the processed data is used to establish an XGBoost carbon fixation prediction model. Temperature, humidity, light intensity, and growth rate are used as feature inputs, and carbon fixation is used as feature output. A carbon fixation prediction model with high enough accuracy is established.

[0071] Assume that a given data set has n samples and m-dimensional features and D = {(x i ,y i )}(D|=n,x i ∈R m ,y i ∈R), the calculation formula of the output result of the XGBoost model is:

[0072]

[0073] In formula (2), F is the set of CART trees established; f k is the kth tree; f k (x i ) is the kth tree for sample x i The prediction results; It is the displacement value in the elevation direction predicted by the model.

[0074] The objective function to be optimized consists of two parts: the regularization term and the loss function. The solution formula of the objective function is:

[0075]

[0076] In formula (3), l is a second-order differentiable loss function, which is used to evaluate the gap between the predicted value and the monitored value; y i is the monitoring value of the displacement of the elevation equation, γ and λ are the coefficients of the regular terms L1 and L2 respectively, T is the number of leaf nodes on the decision tree, ω is the leaf node value vector, and the learning strategy of the boosting method is to learn the current optimal model. This algorithm is to learn the current optimal CART tree. That is, to obtain the minimum value of the loss function (that is, the previous decision tree has found the optimal tree, and continue to find the next optimal decision tree). After the second-order Taylor expansion and derivation of the objective function, the optimal solution of the leaf node and the minimum value of the loss function can be obtained.

[0077] like Figure 6 As shown, the present invention utilizes the powerful nonlinear modeling ability of the XGBoost model to achieve an accuracy rate of ≥ 90% between the calculated carbon sequestration prediction results and the actual values, thereby timely reflecting the impact of environmental changes on carbon sequestration, thereby providing more accurate data to support forest management and carbon credit assessment.

[0078] Example 3

[0079] Combination Figure 4 To illustrate this example, the carbon sequestration prediction model established in Implementation 2 will be encapsulated into the CPU of the Raspberry Pi, and the Raspberry Pi will also control various sensors to collect climate factor data, and the linear displacement sensor will also collect the tree diameter growth rate. The collected climate factor data and tree diameter growth rate are used as the feature inputs required by the prediction model in Example 2, and the collected data can be displayed in real time on the display screen.

[0080] like Figure 4 As shown, the operation steps of the carbon fixation prediction structure described in this embodiment include:

[0081] Step 1: The temperature, humidity, light intensity, and growth rate data collected by various sensors are transmitted to the Raspberry Pi and displayed.

[0082] Step 2: Set the time interval and send the real-time monitoring data to the Raspberry Pi's CPU.

[0083] Step 3: Use the data in step 2 as the feature input of the established carbon sequestration prediction model, and then run the algorithm in the CPU to obtain the accurate carbon sequestration in real time.

[0084] Step 4: Display the carbon fixation amount obtained in step 3 on the screen of the Raspberry Pi in real time.

[0085] Example 4

[0086] Combination Figure 5 This embodiment is described as follows. Figure 5 The steps of the method for accurately predicting carbon sequestration based on the Bayesian optimization algorithm described in this embodiment include:

[0087] S1: Based on the temperature and humidity sensor, photoresistor and linear displacement sensor in the sensor module, climate factor data and wood material property data are collected in real time;

[0088] S2: Preprocess the climate factor data and wood property data, and perform correlation analysis on the historical carbon sequestration data, climate factor data, and wood property data to determine the input and output of the XGBoost model;

[0089] S4: Train and optimize the parameters of the XGBoost model;

[0090] Obtain historical carbon fixation data, environmental data, and growth rate data through query, perform correlation analysis on historical carbon fixation data, environmental data, and growth rate data through Pearson correlation analysis method, retain environmental factors and growth rate factors with Pearson coefficients greater than preset values, fill missing values ​​and process outliers for environmental factors and growth rate factors, use the environmental factors and growth rate factors after missing value filling and outlier processing to build an XGBoost model, and optimize the parameters of the XGBoost model through the Bayesian optimization algorithm;

[0091] Bayesian Optimization (BO) is a black box optimization algorithm and also an active optimization algorithm. Existing studies have shown that Bayesian optimization has better optimization methods than network search and random search, and Bayesian is not easy to fall into local optimality. In this embodiment, the BO optimizer iterates and performs limited experiments in the search space X where the problem is generated to achieve the purpose of optimizing f, and finally obtains the optimal hyperparameters. The hyperparameter combination x * The calculation formula is:

[0092]

[0093] In formula (4), x * is a combination of hyperparameters.

[0094] The prior function and the acquisition function mostly constitute the BO algorithm. The algorithm first calculates the mean and variance of the function at each point through the prior function with the help of the Gaussian regression process to obtain the posterior probability, and then constructs the acquisition function using the improved probability to find the next hyperparameter below the current local minimum of the function.

[0095] Commonly used acquisition functions include PI, EI and UCB. The PI strategy can use the adjustment parameter ξ to select suitable retrieval methods such as global retrieval, thereby preventing the overall situation from falling into local optimality. In this embodiment, the acquisition function uses PI, ignoring the dependence of the acquisition function on hyperparameters. The following formula shows the definition of f(x) using PI:

[0096]

[0097] In formula (5), Φ is the probability function, μ(x) is the expectation, σ(x) is the variance, and f(x + ) is the optimal value of the objective function, and ξ is a set of hyperparameters.

[0098] S5: Save the XGBoost model to the GPU of the Raspberry Pi, use the current preprocessed environment and growth data as input data, and output the carbon sequestration prediction result;

[0099] The prediction model as a whole adopts the XGBoost model optimized by Bayesian. The model input is tree growth rate, temperature and humidity, and light intensity, and the model output is carbon sequestration, such as Figure 8 As shown in the figure, first, an initial point is randomly generated in the parameter space as the starting point. To ensure the diversity of the initial points, the Latin Hypercube Sampling (LHS) method is used, which can evenly distribute the initial points in the parameter space and reduce the concentration of the initial points. During the execution process, multiple random starting points are set to prevent falling into the local optimum. Each random starting point will independently perform the BO optimization process, and finally select the best parameter combination among multiple optimization results. This multi-starting point strategy increases the probability of searching for the global optimal solution. During the specific execution, 10 random starting points are set for parallel optimization, and the best performance set of parameters is selected by comparing the final results of each starting point. The posterior probability is calculated by Gaussian regression, and the acquisition function PI is constructed. The optimal next sampling point is determined by the acquisition function, and it is judged whether the preset target is reached. If the preset target is met, the optimized optimal parameter set is output; if not, the initial point is generated again and each step before is repeated until the target is met. Finally, the XGBoost model uses the optimized parameters to predict the amount of carbon sequestration.

[0100] In this embodiment, the existing 420 sets of data are divided into training set and test set according to 8 / 2, the training set is used to train the BO-XGBoost prediction model, and the test set is input into the model to test the accuracy. Through continuous debugging, the carbon fixation prediction results are as follows: Fig. 9 As shown, finally, according to Figure 6 As shown, the test set is brought into the trained BO-XGBoost carbon fixation prediction model to predict the carbon fixation of this part of the test set, and the prediction result is compared with the carbon fixation corresponding to the actual test set. The accuracy of the prediction results obtained in this embodiment is above 90%.

[0101] The above is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technician familiar with this profession can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent replacement and improvement made to the above embodiments without departing from the content of the technical solution of the present invention, based on the technical essence of the present invention, within the spirit and principles of the present invention, still fall within the protection scope of the technical solution of the present invention.

Claims

1. A carbon fixation accurate prediction system based on Bayesian optimization algorithm, characterized in that: The structure of the carbon fixation accurate prediction system based on the Bayesian optimization algorithm includes: Sensor module, data processing unit, display and interaction unit and prediction model unit; The sensor module is electrically connected to the data processing unit, the data processing unit is electrically connected to the prediction model unit and the display and interaction unit, and the prediction model unit is electrically connected to the display and interaction unit.

2. The carbon fixation accurate prediction system based on Bayesian optimization algorithm according to claim 1 is characterized in that: The sensor module includes a temperature and humidity sensor, a photoresistor and a linear displacement sensor, and the temperature and humidity sensor, the photoresistor and the linear displacement sensor are all electrically connected to the data processing unit; The temperature and humidity sensor is used to measure and record ambient temperature and humidity; The photoresistor is used to monitor the ambient light intensity; The linear displacement sensor is used to measure the radial growth condition of trees to obtain the growth rate and health status of the trees.

3. The carbon fixation accurate prediction system based on Bayesian optimization algorithm according to claim 1 is characterized in that: The data processing unit is a Raspberry Pi, which is used to pre-process the environment and growth data collected in real time by the sensor module.

4. The carbon fixation accurate prediction system based on Bayesian optimization algorithm according to claim 1 is characterized in that: The prediction model unit is an XGBoost model, the input of the XGBoost model is tree growth rate, temperature and humidity, and light intensity data, the output is the carbon fixation prediction value, and the XGBoost model is stored on the GPU of the Raspberry Pi and is used to predict the carbon fixation of the preprocessed environment and growth data.

5. The carbon fixation accurate prediction system based on Bayesian optimization algorithm according to claim 1 is characterized in that: The display and interaction unit includes a display screen and a touch screen / external keyboard; The display screen is connected to the data processing unit and is used to display the current carbon fixation prediction results and environment and growth data in real time; The touch screen / external keyboard is used to query current carbon fixation prediction results, environment and growth data and historical carbon fixation prediction results, environment and growth data.

6. A method for accurately predicting carbon sequestration based on a Bayesian optimization algorithm, applied to a system for accurately predicting carbon sequestration based on a Bayesian optimization algorithm as described in any one of claims 1 to 5, characterized in that: include: Step 1: Start the real-time and accurate prediction system for carbon fixation, and collect environmental and growth data in real time through the temperature and humidity sensor, photoresistor and linear displacement sensor in the sensor module; Step 2: Transmit the real-time collected environmental and growth data to the data processing unit for preprocessing; Step 3: Train the XGBoost model and optimize its parameters, input the preprocessed environmental and growth data into the trained and optimized XGBoost model to obtain the carbon fixation prediction results; Step 4: The carbon fixation prediction results and the environment and growth data are displayed and stored on the display screen in the display and interaction unit. The user exports the carbon fixation prediction results and the pre-processed environment and growth data through the touch screen / external keyboard, and sets the data collection frequency of the sensor module and the carbon fixation warning threshold.

7. The method for accurately predicting carbon sequestration based on Bayesian optimization algorithm according to claim 6, characterized in that: The preprocessing steps in step 2 include: Raspberry Pi is used to clean the environment and growth data, remove outliers and noise in the environment and growth data, and normalize the cleaned environment and growth data to complete the preprocessing of the environment and growth data.

8. The method for accurately predicting carbon sequestration based on Bayesian optimization algorithm according to claim 6, characterized in that: The steps for obtaining the carbon sequestration prediction results in step 3 include: Step 3.1: Obtain historical carbon fixation data, environmental data, and growth rate data through query; Step 3.2: Perform correlation analysis on historical carbon fixation data, environmental data and growth rate data by Pearson correlation analysis method, and retain environmental factors and growth rate factors with Pearson coefficients greater than preset values; Step 3.3: Fill missing values ​​and process outliers for the environmental factors and growth rate factors, and use the environmental factors and growth rate factors after the missing values ​​are filled and the outliers are processed to build an XGBoost model; Step 3.4: Optimize the parameters of the XGBoost model through the Bayesian optimization algorithm, input the current preprocessed environmental and growth data into the parameter-optimized XGBoost model, and output the carbon fixation prediction results.

9. The method for accurately predicting carbon sequestration based on Bayesian optimization algorithm according to claim 8, characterized in that: The steps for optimizing the parameters of the XGBoost model using the Bayesian optimization algorithm in step 3.4 include: Step 3.4.1: Use Latin hypercube sampling to obtain uniformly distributed initial points in the parameter space; Step 3.4.2: Set n random initial points, define the expected variance and optimal value of the objective function for each random initial point, and select the acquisition function f(x) based on the PI strategy; Step 3.4.3: Use BO optimizer to iteratively optimize the acquisition function f(x) to obtain the hyperparameter combination x * ; Step 3.4.4: Compare the prediction accuracy of each set of hyperparameter combinations, select the set of hyperparameter combinations with the highest prediction accuracy, and calculate the posterior probability through Gaussian regression; Step 3.4.5: Determine the next sampling point through the acquisition function and determine whether the preset accuracy is achieved. If so, output the optimal hyperparameter combination. If not, repeat steps 3.4.1 to 3.4.4 until the preset accuracy is achieved. The calculation formula of the acquisition function f(x) is: In formula (1), Φ is the probability function, μ(x) is the expectation, σ(x) is the variance, and f(x + ) is the optimal value of the objective function, ξ is a set of hyperparameters; Hyperparameter combination x * The calculation formula is: In formula (2), x * is a combination of hyperparameters.