Construction method of cucumber seedling growth simulation model based on effective accumulated temperature

By combining the particle swarm algorithm and support vector regression model, the problem of insufficient prediction accuracy and stability of cucumber seedling growth in the existing technology is solved, and higher prediction accuracy and fitting ability are achieved, especially When dealing with complex environments and nonlinear problems.

CN119006201APending Publication Date: 2024-11-22YUNNAN AGRICULTURAL UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411048334.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-01
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The existing cucumber seedling growth prediction model is based on traditional statistical methods, with poor accuracy and stability, which is difficult to meet actual production needs, especially when dealing with complex environments and nonlinear problems.

Method used

Using a cucumber seedling growth simulation model based on effective accumulation temperature, combined with particle swarm algorithm (PSO) and support vector regression model (SVR), the PSO-SVR model is constructed, using effective accumulation temperature as the input variable, through data acquisition, normalization processing and iteration Optimize, establish a growth simulation model for cucumber seedlings, and use the model evaluation indicators RMSE and R2 for evaluation.

Benefits of technology

The PSO-SVR model demonstrates high accuracy and effectiveness in the prediction of cucumber seedling growth, which improves prediction accuracy and stability, especially in the prediction of indicators such as fresh quality, leaf length, leaf width, plant height and stem thickness. Performance improved significantly.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119006201A_ABST
    Figure CN119006201A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of cucumber seedling growth prediction, and discloses a construction method of a cucumber seedling growth simulation model based on effective accumulated temperature. Comprising the following steps: data acquisition and preprocessing: collecting effective accumulated temperature, leaf length, leaf width, leaf area, plant height, fresh quality and stem diameter of cucumber seedlings planted in a test greenhouse in the growth process, and carrying out normalization processing on the data; and S3, taking the effective accumulated temperature as a model input variable, respectively constructing six cucumber seedling growth simulation models of leaf length, leaf width, leaf area, plant height, fresh quality and stem diameter by adopting the first batch of cucumber seedling planting data, and evaluating the cucumber seedling growth simulation model established in the step S3 by adopting a model evaluation index root mean square error RMSE and a determination coefficient R2. The PSO-SVR model shows relatively high precision and effectiveness in cucumber seedling growth prediction.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of cucumber seedling growth prediction, and in particular to a method for constructing a cucumber seedling growth simulation model based on effective accumulated temperature. Background Art

[0002] Cucumber is an important economic crop, and its growth directly affects its yield and quality.

[0003] Most of the existing cucumber seedling growth prediction models are based on traditional statistical methods, which have poor accuracy and stability and are difficult to meet actual production needs. At the same time, the existing technology shows great limitations when dealing with complex environments and nonlinear problems. Summary of the invention

[0004] 1. Technical issues to be solved

[0005] In view of the deficiencies in the prior art, the present invention provides a method for constructing a cucumber seedling growth simulation model based on effective accumulated temperature, which has the advantages of accurate prediction and solves the above-mentioned technical problems.

[0006] (II) Technical solution

[0007] To achieve the above object, the present invention provides the following technical solution: a method for constructing a cucumber seedling growth simulation model based on effective accumulated temperature, comprising the following steps:

[0008] S1. Data collection and preprocessing: effective accumulated temperature, leaf length, leaf width, leaf area, plant height, fresh weight and stem diameter of cucumber seedlings planted in the experimental greenhouse were collected during their growth, and the data were normalized;

[0009] S2, dividing the cucumber seedlings in step S1 into two batches, the first batch as a construction group for building a model, and the second batch as a verification group for model verification;

[0010] S3. Establish a cucumber seedling growth simulation model during the cucumber growth process, and use the effective accumulated temperature as the model input variable. Use the first batch of cucumber seedling planting data to construct a total of six cucumber seedling growth simulation models for leaf length, leaf width, leaf area, plant height, fresh weight and stem thickness;

[0011] S4. Use the model evaluation indicators root mean square error RMSE and determination coefficient R 2 The cucumber seedling growth simulation model established in step S3 is evaluated.

[0012] As a preferred technical solution of the present invention, the cucumber seedling growth simulation model constructed in step S3 is a PSO-SVR model constructed by combining the PSO algorithm and the SVR model. The specific process of establishing the PSO-SVR model is as follows:

[0013] S3.1, initialize the particle swarm parameters, randomly initialize the position and velocity of each particle, set the particle swarm size, maximum number of iterations and particle swarm parameters;

[0014] S3.2, iterative optimization. During the iterative optimization process, each particle uses its position parameters to build an SVR model, and uses the first batch of construction group data for training. The second batch of validation group data is used to evaluate the prediction performance of the SVR model, and the prediction performance is used as the fitness value of the particle. In each iteration, if the fitness value of the current particle is better than its individual optimal fitness value, its individual optimal position is updated, and the speed and position of each particle are updated through the PSO algorithm until the maximum number of iterations is reached;

[0015] S3.3, when the maximum number of iterations is reached, output the optimal parameters;

[0016] S3.4, and construct the PSO-SVR model with the optimal parameters in step S3.3.

[0017] As a preferred technical solution of the present invention, the parameters output in step S3.3 are the optimal regularization parameter C and the width parameter gamma of the optimal radial basis function.

[0018] As a preferred technical solution of the present invention, the value range of the initial regularization parameter in the SVR model in step S3.2 is set to 1-100, and the value range of the width parameter of the initial radial basis function in the SVR model in step S3.2 is set to 0.01-20.

[0019] Compared with the prior art, the present invention provides a method for constructing a cucumber seedling growth simulation model based on effective accumulated temperature, which has the following beneficial effects:

[0020] The present invention uses the model evaluation index R 2 The established PSO-SVR model was evaluated by RMS / RMSE. The PSO-SVR model showed high accuracy and effectiveness in predicting the growth of cucumber seedlings. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a schematic diagram of a 1:1 straight line graph of the measured value and the predicted value of the cucumber seedlings supported by the vector regression of the present invention;

[0022] Figure 2 This is a schematic diagram of the PSO-SVR process of the present invention;

[0023] Figure 3 Schematic diagram of the fitness curve of different growth indicators of PSO-SVR of the present invention;

[0024] Figure 4This is a schematic diagram of the particle position change trend of each growth index of PSO-SVR of the present invention;

[0025] Figure 5 It is a schematic diagram of a 1:1 straight line graph of the measured value and the predicted value of the PSO-SVR cucumber seedling of the present invention;

[0026] Figure 6 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0027] 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.

[0028] See also Figure 1 - Figure 6 A method for constructing a cucumber seedling growth simulation model based on effective accumulated temperature comprises the following steps:

[0029] S1. Data collection and preprocessing: effective accumulated temperature, leaf length, leaf width, leaf area, plant height, fresh weight and stem diameter of cucumber seedlings planted in the experimental greenhouse were collected during their growth, and the data were normalized;

[0030] S2, the cucumber seedlings in step S1 are divided into two batches, the first batch is used as a construction group for building a model, and the second batch is used as a verification group for model verification, WPS is used to process the test data, and Python programming language is used for model construction and drawing;

[0031] S3. Establish a cucumber seedling growth simulation model during the cucumber growth process, and use the effective accumulated temperature as the model input variable. Use the first batch of cucumber seedling planting data to construct a total of six cucumber seedling growth simulation models for leaf length, leaf width, leaf area, plant height, fresh weight and stem thickness;

[0032] The cucumber seedling growth simulation model constructed in step S3 is a PSO-SVR model constructed by combining the PSO algorithm and the SVR model. The specific process of establishing the PSO-SVR model is as follows:

[0033] S3.1, initialize the particle swarm parameters, randomly initialize the position and velocity of each particle, set the particle swarm size, maximum number of iterations and particle swarm parameters;

[0034] S3.2, iterative optimization. During the iterative optimization process, each particle uses its position parameters to build an SVR model, and uses the first batch of construction group data for training. The second batch of validation group data is used to evaluate the prediction performance of the SVR model, and the prediction performance is used as the fitness value of the particle. In each iteration, if the fitness value of the current particle is better than its individual optimal fitness value, its individual optimal position is updated, and the speed and position of each particle are updated through the PSO algorithm until the maximum number of iterations is reached;

[0035] The value range of the initial regularization parameter in the SVR model in step S3.2 is set to 1-100, and the value range of the width parameter of the initial radial basis function in the SVR model in step S3.2 is set to 0.01-20;

[0036] S3.3, when the maximum number of iterations is reached, output the optimal parameters;

[0037] S3.4, and construct the PSO-SVR model with the optimal parameters in step S3.3;

[0038] Six different models were constructed using PSO-SVR for the six growth indicators of leaf length, leaf width, leaf area, plant height, fresh weight, and stem diameter. To ensure the comparability of the experiments, the parameters used were kept consistent. The modeling process for each growth indicator described in the article remained the same, and a total of six models were established;

[0039] Take the leaf length model as an example:

[0040] 1. Initialize the particle swarm parameters, randomly initialize the position and velocity of each particle, set the particle swarm size, maximum number of iterations, and particle swarm parameters. Specifically, the particle position is expressed as the parameters of the SVR model (C and gamma), and the initial value ranges are set to 1-100 and 0.01-20 respectively;

[0041] 2. Iterative optimization. During the iterative optimization process, each particle uses its position parameters to build an SVR model, and uses the first batch of construction group data for training. The second batch of validation group data is used to evaluate the prediction performance of the SVR model, and the prediction performance is used as the fitness value of the particle. In each iteration, if the fitness value of the current particle is better than its individual optimal fitness value, its individual optimal position is updated, and the speed and position of each particle are updated through the PSO algorithm until the maximum number of iterations is reached. The following is the specific optimization process of the leaf length model:

[0042] Initialization: Randomly generate several sets of parameter combinations (C and gamma) and use these parameter combinations as the positions of the initial particles.

[0043] Calculate fitness: For each particle position, use the SVR model for training and calculate the RMSE of the model on the validation group data as the fitness value of the particle.

[0044] Update particle position: Update the position and velocity of each particle according to the PSO algorithm to gradually approach the global optimal solution.

[0045] Loop iteration: Repeat the above steps until the preset maximum number of iterations is reached;

[0046] 3. When the maximum number of iterations is reached, the optimal parameters (C and gamma) are output. Through multiple iterations, the PSO algorithm can effectively search for the parameter combination that makes the model perform best;

[0047] 4. The final PSO-SVR leaf length model is constructed with the best parameter combination. The final model parameters will be used to predict the growth of cucumber seedling leaf length;

[0048] Then, RMSE and R 2 The above leaf length-based model was evaluated;

[0049] S4. Use the model evaluation indicators root mean square error RMSE and determination coefficient R 2 The cucumber seedling growth simulation model established in step S3 is evaluated.

[0050] The present invention is further described in detail in conjunction with specific embodiments. By taking the effective accumulated temperature as the main input variable, the growth simulation model of cucumber seedling leaf length, leaf width, leaf area, plant height, fresh weight and stem diameter is constructed by support vector regression SVR and PSO-SVR methods respectively. Data collection is carried out during the growth stage of cucumber seedlings, and the model evaluation indicators RMSE and R 2 The models were compared and selected, and a 1:1 straight line graph of the measured value and the predicted value of cucumber seedlings was drawn to prove the high accuracy and effectiveness of the model of the present invention.

[0051] like Figure 1 Table 1. Support vector regression (SVR) was used to construct the cucumber seedling growth simulation model. The effective accumulated temperature was used as the model input variable. The first batch of cucumber seedling planting data was used to construct the growth simulation models of leaf length, leaf width, leaf area, plant height, fresh weight and stem diameter. The model evaluation indicators RMSE and R 2 The models were compared and selected, and a 1:1 straight line graph was drawn between the measured value and the predicted value of cucumber seedlings;

[0052] The support vector regression model showed a high determination coefficient (R 2 ). Fresh quality R 2The highest value is 0.937, indicating that the SVR model has a very high prediction accuracy in this aspect. 2 The values ​​were 0.915 and 0.908, respectively, which also showed strong fitting ability. 2 The value is 0.904, and the R 2 The value is 0.872, which indicates that the model has a high accuracy in predicting these indicators. 2 The value is 0.866, and the prediction accuracy is not high. Since the support vector regression model has certain limitations in dealing with complex nonlinear problems, the prediction errors of some growth indicators are large, so further optimization is needed;

[0053] From the perspective of root mean square error (RMSE), the SVR model showed some variability in the prediction of different growth indicators. The RMSE of fresh weight was 0.983 g, indicating that the model had a small error in predicting fresh weight. The RMSE of leaf length and leaf width were 0.852 cm and 1.147 cm, respectively, indicating that the model also had a low error in predicting these indicators. The RMSE of plant height was 1.824 cm, and the RMSE of stem diameter was 0.420 mm, indicating that the SVR model had relatively small errors in predicting these growth indicators and had good prediction effects. However, the RMSE of leaf area was 30.364 cm 2 , the error is relatively large, and there is still room for improvement;

[0054] Fresh quality Leaf length Leaf Width Plant height Stem thickness Leaf area <![CDATA[R 2 ]]> 0.937 0.915 0.908 0.904 0.872 0.866 RMSE 0.983(g) 0.852(cm) 1.147(cm) 1.824(cm) 0.420(mm) <![CDATA[30.364(cm 2 )]]>

[0055] Table 1

[0056] like Figure 2 - Figure 5 The PSO-SVR was used to construct a cucumber seedling growth simulation model. The effective accumulated temperature was used as the model input variable. The first batch of cucumber seedling planting data was used to construct growth simulation models of leaf length, leaf width, leaf area, plant height, fresh weight and stem diameter. The model evaluation indicators RMSE and R 2 The model was evaluated and a 1:1 straight line graph of the measured and predicted values ​​of cucumber seedlings was drawn;

[0057] The first batch of planting data was used to construct a PSO-SVR cucumber seedling growth simulation model. In order to intuitively show the performance of the PSO-SVR model on different growth indicators, the fitness curves of stem diameter, fresh weight, leaf width, leaf area, leaf length and plant height were depicted, which reflected the changing trend of the difference between the model prediction value and the measured value. The decrease in fitness value indicates that the prediction accuracy improves with the iteration process. Through the fitness curves of different growth indicators, not only can the performance of the model in a single iteration process be evaluated, but also the optimization effects between the indicators can be compared, and then the overall performance of the model can be judged;

[0058] In order to more vividly demonstrate the dynamic behavior of the particle swarm algorithm from global random search to centralized positioning of the optimal solution in the search space, the initial and final positions of the particles for each growth index of the cucumber seedlings were plotted.

[0059] Each particle is a set of specific parameters (C and gamma), and the initial particle position diagram reflects the widely dispersed initial state of the particles in the high-dimensional parameter space. This distribution ensures that the model can explore a wide range of potential solutions in the early stage, avoiding locking the potential local optimal solution in the early stage of model prediction, thereby missing the better global solution. As the iteration continues, the algorithm refines its search focus, which is reflected in the final particle position diagram. The particles gradually gather from the initial scattered state to a smaller area, showing the convergence trajectory of the optimization process. This dynamic focusing behavior marks the gradual transition of the PSO algorithm from global exploration to local search of fine-tuning parameters to identify those parameter sets that can significantly reduce the model error. The position of the global optimal solution is marked with special symbols, highlighting the optimal parameter combination finally determined by the algorithm. The global optimal solution not only symbolizes the highest point of the model prediction performance, but also reflects the particle swarm algorithm's ability to accurately locate in the parameter space after multiple iterations. By comparing the initial and final particle distributions, we can gain insight into the core characteristics of the algorithm's optimization behavior and verify the effectiveness of PSO in parameter selection;

[0060] PSO-SVR result analysis,After the model was built, the second batch of planting data was used to verify the model. In the prediction of cucumber seedling growth indicators, the PSO-SVR model showed a high determination coefficient, showing its strong fitting ability. 2 The R value of leaf length and leaf width was 0.962, indicating that the PSO-SVR model can accurately predict the fresh weight change of cucumber seedlings. 2 The R values ​​of plant height were 0.930 and 0.922, respectively, indicating the excellent performance of the model in predicting these indicators. 2 is 0.924, and the R 2 0.879, which indicates that the model has a high accuracy in predicting these growth indicators. In addition, the R 2 It is 0.894, which is slightly lower than other indicators, but it still shows the reliability and stability of the PSO-SVR model as a whole. Overall, the PSO-SVR model shows excellent prediction ability in the prediction of various growth indicators, indicating that it can fit the growth data of cucumber seedlings well;

[0061] From the perspective of root mean square error (RMSE), the PSO-SVR model showed low errors in the prediction of each growth index, demonstrating its high-precision prediction ability. The RMSE of fresh weight was 0.766 g, indicating that the model had very small errors in predicting fresh weight. The RMSE of leaf length and leaf width were 0.775 cm and 1.056 cm, respectively, further indicating that the PSO-SVR model had small errors and good prediction effects in predicting these indicators. The RMSE of plant height was 1.630 cm, and the RMSE of stem diameter was 0.409 mm, indicating that the model had small errors in predicting these growth indicators and had high accuracy. The RMSE of leaf area was 26.962 cm 2 , although the error is relatively large, it is a significant improvement over other models.

[0062] Overall, the PSO-SVR model has a small error in the prediction of most growth indicators, can well capture the changing patterns in the data, and provide accurate prediction results; Table 2 shows the analysis of the PSO-SVR model results;

[0063] Fresh quality Leaf length Leaf Width Plant height Stem thickness Leaf area <![CDATA[R 2 ]]> 0.962 0.930 0.922 0.924 0.879 0.894 RMSE 0.766(g) 0.775(cm) 1.056(cm) 1.630(cm) 0.409(mm) <![CDATA[26.962(cm 2 )]]>

[0064] Table 2

[0065] The two constructed models were compared and analyzed by R 2 The model fitting ability was evaluated. Comparing the results of the PSO-SVR model and the traditional SVR model, it can be seen that the PSO-SVR model has a better fitting goodness of fit than the SVR model in all growth indicators. For the indicators of stem diameter, fresh weight, leaf width, leaf area, leaf length and plant height, the R of the PSO-SVR model is 2 The values ​​are 0.007, 0.025, 0.014, 0.028, 0.015 and 0.020 higher than those of the SVR model, respectively. This shows that by using the particle swarm algorithm to optimize the parameters of the support vector regression model, the PSO-SVR model can more accurately predict the growth of cucumber seedlings, especially in terms of leaf area, leaf length and plant height. The improvement effect of the PSO-SVR model is more significant. In general, the PSO-SVR model significantly improves the accuracy of the prediction of cucumber seedling growth indicators by combining the global search capability of the PSO algorithm and the regression analysis capability of the SVR model. Table 3 is a comparison table of the models;

[0066] Fresh quality Leaf length Leaf Width Plant height Stem thickness Leaf area SVR 0.937 0.915 0.908 0.904 0.872 0.866 PSO-SVR 0.962 0.930 0.922 0.924 0.879 0.894

[0067] Table 3

[0068] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a cucumber seedling growth simulation model based on effective accumulated temperature. Features: The following steps are involved: S1. Data collection and preprocessing: effective accumulated temperature, leaf length, leaf width, leaf area, plant height, fresh weight and stem diameter of cucumber seedlings planted in the experimental greenhouse were collected during their growth, and the data were normalized; S2, dividing the cucumber seedlings in step S1 into two batches, the first batch as a construction group for building a model, and the second batch as a verification group for model verification; S3. Establish a cucumber seedling growth simulation model during the cucumber growth process, and use the effective accumulated temperature as the model input variable. Use the first batch of cucumber seedling planting data to construct a total of six cucumber seedling growth simulation models for leaf length, leaf width, leaf area, plant height, fresh weight and stem thickness; S4. Use the model evaluation indicators root mean square error RMSE and determination coefficient R 2 The cucumber seedling growth simulation model established in step S3 is evaluated.

2. A method for constructing a cucumber seedling growth simulation model based on effective accumulated temperature according to claim 1, Features: The cucumber seedling growth simulation model constructed in step S3 is a PSO-SVR model constructed by combining the PSO algorithm and the SVR model. The specific process of establishing the PSO-SVR model is as follows: S3.1, initialize the particle swarm parameters, randomly initialize the position and velocity of each particle, set the particle swarm size, maximum number of iterations and particle swarm parameters; S3.2, iterative optimization. During the iterative optimization process, each particle uses its position parameters to build an SVR model, and uses the first batch of construction group data for training. The second batch of validation group data is used to evaluate the prediction performance of the SVR model, and the prediction performance is used as the fitness value of the particle. In each iteration, if the fitness value of the current particle is better than its individual optimal fitness value, its individual optimal position is updated, and the speed and position of each particle are updated through the PSO algorithm until the maximum number of iterations is reached; S3.3, when the maximum number of iterations is reached, output the optimal parameters; S3.4, and construct the PSO-SVR model with the optimal parameters in step S3.

3.

3. A method for constructing a cucumber seedling growth simulation model based on effective accumulated temperature according to claim 2, Features: The parameters output in step S3.3 are the optimal regularization parameter C and the width parameter gamma of the optimal radial basis function.

4. A method for constructing a cucumber seedling growth simulation model based on effective accumulated temperature according to claim 3, Features: The value range of the initial regularization parameter in the SVR model in step S3.2 is set to 1-100, and the value range of the width parameter of the initial radial basis function in the SVR model in step S3.2 is set to 0.01-20.