CN105654203A - Cucumber whole-course photosynthetic rate predicting model based on support vector machine, and establishing method - Google Patents

Abstract

The invention relates to a cucumber whole-course photosynthetic rate predicting model based on a support vector machine. Photosynthetic rate test data of cucumber seedlings is obtained by utilizing a multi-factor nested experiment; model training is carried out by adopting a LM training method; a cucumber whole-course photosynthetic rate model fused with multiple growing periods is established; the cucumber whole-course photosynthetic rate model is subjected to comparison verification with a single-growing-period photosynthetic rate model and a whole-growing-period cucumber photosynthetic rate model respectively by adopting an abnormal check manner; the result shows that the whole-course photosynthetic rate model, which is established by adding the growing period as the one-dimensional input quantity, can effectively cross a local flat region, have obvious superiority, and satisfy training requirements that the error is less than 0.0001; the determination coefficient of a predicted value and a practically measured value of the model is 0.993; the error is less than 6.253%; both the training effect and the model-fitting degree of the model are superior to the mixed-growing period model; the precision of the model is similar to that of the single-growing-period photosynthetic rate model; and thus, the model disclosed by the invention can provide theoretical basis and technical support for light environment adjustment and control of facility crops.

Description

A prediction model and establishment method of whole-range photosynthetic rate of cucumber based on support vector machine

technical field

The invention belongs to the technical field of intelligent agriculture, and in particular relates to a support vector machine-based prediction model and establishment method of the whole-process photosynthetic rate of cucumber.

Background technique

Cucumber is one of the main vegetables cultivated in my country. The quality and yield of cucumber are inseparable from its photosynthetic ability. Photosynthetic rate has a significant relationship with multiple factors such as chlorophyll content, temperature, _CO2 concentration, light intensity, and relative humidity. Among them, chloroplast is the basic organelle for photosynthesis of green plants, and chlorophyll is the basic component of chloroplast, which is very important in plant photosynthesis, and its content is an important indicator of plant photosynthesis ability, nutritional status and growth status. Temperature It affects the activity of Rubisco activating enzyme and stomatal conductance in crops. The concentration of CO ₂ directly affects the dark reaction rate of crops and the accumulation of dry matter. The intensity of light is the direct driving force and driving force of photosynthesis. Relative humidity affects the stomatal conductance of leaves, etc. And there are interactions among the factors. Therefore, comprehensive consideration of the influence of multiple factors and the establishment of a multi-factor coupling full-range photosynthetic rate prediction model play an important role in optimizing the cucumber light environment.

Many foreign scholars and research institutions have conducted in-depth research on plant photosynthesis, and based on this, they have established a large number of models for controlling the internal environment of greenhouses and for plant growth. In the 1970s, Charles-Edwards proposed that the physiological model of plant photosynthesis is one of the initial models for establishing leaf photosynthetic models, and the physiological model includes photorespiration, dark respiration and oxygen effects. On the basis of this related research, relevant scholars have established a variety of photosynthetic models, including Cartesian hyperbolic models, non-Cartesian hyperbolic models, and exponential relationship models, but the model parameters are not easy to obtain, which brings certain difficulties to the application of the model. Based on the above physiological model, ZipiaoYe et al. proposed a photosynthetic rate model based on electron transport, etc. proposed a steady-state model of photosynthetic rate, J.Z.XU et al. conducted a study on the photosynthetic model under different nitrogen, Y.LANG et al. used different leaves, Carried out related research and exploration of photosynthetic rate model. Therefore, it is more necessary to select a plant photosynthesis model with good performance and determine more accurate related parameters to regulate the environment of plant growth and crop cultivation. However, the current model of net photosynthesis of crops in solar greenhouses in China still needs continuous improvement.

In recent years, many scholars have carried out relevant research on the establishment of photosynthetic rate models. The above studies have considered the relationship between different environmental factors, but there are deficiencies such as low fitting degree, complex fitting formula, and large error. The neural network has the advantages of nonlinear mapping and adaptive learning capabilities, and is suitable for fitting and predicting nonlinear complex system models. Therefore, the modeling of photosynthetic rate based on neural networks has become a research hotspot. Recently, the photosynthetic rate model based on Hopfield network, the stomatal conductance model of greenhouse tomato leaves based on BP neural network, and the net photosynthetic rate prediction model of single leaf in tomato flowering period based on WSN have appeared. The above studies have applied neural networks to photosynthetic rate from different angles. Modeling, but the impact of different growth periods on crops has not been considered, and a full-range cucumber photosynthetic rate prediction model has not been established, and there are shortcomings in the slow training process and large differences in training errors.

Support vector machine is a new general machine learning method based on the framework of statistical learning theory. It can solve the highly nonlinear classification and regression problems in the sample space, and is an effective method to deal with nonlinear classification and nonlinear regression. Photosynthetic rate prediction contains a large number of non-linear factors. Traditional statistical forecasting methods require significant linear correlations between factors and forecasted objects when establishing forecasting models, and the minimum linear correlation between factors is required. The complexity and nonlinearity of many factors that cause photosynthetic rate changes determine the nonlinear correlation between predictors and predictors. Therefore, traditional model prediction methods are difficult to solve the prediction problem that is essentially a nonlinear relationship. Rate prediction provides a feasible and efficient way.

Contents of the invention

In order to overcome the above-mentioned shortcoming of the prior art, the object of the present invention is to provide a kind of cucumber full-range photosynthetic rate prediction model based on support vector machine, design multi-factor nested experiment, adopt support vector machine modeling after data normalization process, A cucumber photosynthetic rate prediction model that only distinguishes the seedling stage, flowering and fruiting stage, full growth stage, and added growth stage as one-dimensional input factors was established respectively, and a whole-process cucumber photosynthetic rate prediction model was established through comparison and verification. Establish the basis for environmental regulation.

In order to achieve the above object, the technical scheme adopted in the present invention is:

A prediction model of whole-process photosynthetic rate of cucumber based on support vector machine, the model formula is: Among them, the output f(x) represents the predicted photosynthetic rate, the input signal X'=(X ₁ 'X ₂ '…X ₅ ') ^T , X ₁ ', X ₂ ', X ₃ ', X ₄ ', X ₅ 'respectively temperature, CO ₂ concentration, light intensity, relative humidity and chlorophyll content, w is the weight vector, b is the bias, Φ(x) is the nonlinear mapping function, l is the training set sample pair {( _xi , y _i ), the number of training samples in i=1,2,3,...,l}, x _i is the input column vector of the i-th training sample, y _i is the corresponding output value, y _i ∈ R, is the i×d-dimensional real number field, d is the column vector dimension, a _i and a _i ^* are the optimal solutions of the following formula:

is the kernel function, σ is the width parameter, ε is the abort training error, and c is the penalty factor.

The establishment method of the whole-range photosynthetic rate prediction model of cucumber based on support vector machine comprises the steps:

Step 1, obtain experimental data, the process is as follows:

Use a nutrient pot to grow seedlings. When the cucumber seedlings grow into two leaves and one heart, select cucumber seedlings with uniform growth, a stem diameter between 0.6 and 0.8 cm, and a plant height of less than 10 cm for the test. Select 150 strong cucumber seedlings as test samples. Treat that the cucumber is in the flowering and fruiting period, choose 150 plants whose flowering nodes are about 50 centimeters away from the faucet as the test sample of the flowering and fruiting period;

Measure the net photosynthetic rate. During the process, use the temperature control module to set ₅ temperature gradients of 16, 20, 24, 28, and 32°C; There are 5 gradients in L; a total of 11 photon flux density gradients of 0, 20, 50, 100, 200, 300, 500, 700, 1000, 1200, 1500μmol/(m ² s) were obtained by using the LED light source module, to embedding A total of 275 groups of experiments were carried out in this way, and each group of experiments was repeatedly tested on 3 randomly selected plants. During the experiment, the relative humidity of the leaf chamber was recorded, and the chlorophyll content of the measured leaves was recorded, so as to form a model based on chlorophyll content, temperature, and _CO2 concentration. , light intensity and relative humidity are input, and net photosynthetic rate is output 1650 sets of experimental data, namely 825 sets at the seedling stage and 825 sets at the flowering and fruiting stage;

Step 2, build a model

Randomly select training samples and test samples from the experimental data obtained in step 1, then select the support vector machine kernel function type as radial basis function, and use the grid method to determine the support vector machine penalty factor c and radial basis kernel function parameter g In the optimization interval, the genetic algorithm is used to further optimize the parameters c and g based on the optimization interval until the maximum number of iterations is reached, the optimization results of the parameters c and g are output, and the whole-process photosynthetic rate prediction model of cucumber based on the support vector machine is constructed.

In the step 2,

The experimental data was normalized to the [0.2,0.9] interval to obtain the whole set of sample data, and the sample number was randomly selected according to the ratio of 4:1 to construct the training sample set and test sample set.

In the step 2,

The optimization interval of penalty factor c and radial basis kernel function parameter g is expressed as lgc∈[2,3], lgg∈[0,1].

In the step 2,

The genetic algorithm is used to further optimize the optimization interval of the penalty factor c and the radial basis kernel function parameter g. The process is as follows: set the genetic algorithm parameters, the number of individuals NIND = 100, the maximum genetic algebra MAXGEN = 50, the generation gap GGAP = 0.95, The crossover probability and mutation probability are 0.8, 0.25, c and g are used as variables, respectively represented by 10-bit binary; define the objective function as the model mean square error value MSE; genetic operator operations: selection, crossover, mutation; calculate the target value of offspring function and reinsert the child to the parent to obtain a new population; reach the maximum number of iterations and output the optimization results of c and g.

In the step 2,

The built model is: Among them, the input signal X'=(X ₁ 'X ₂ '…X ₅ ') ^T , X ₁ ', X ₂ ', X ₃ ', X ₄ ', X ₅ ' are temperature, CO ₂ concentration, light intensity , relative humidity and chlorophyll content, w is a weight vector, b is a bias, and Φ(x) is a nonlinear mapping function.

After obtaining the full-range photosynthetic rate prediction model of cucumber based on support vector machine, input sample data for photosynthetic efficiency prediction, if the accuracy requirement is met, then output the prediction result, otherwise return to the genetic algorithm step, based on the optimization interval for c and g further optimization.

Compared with prior art, the beneficial effect of the present invention is:

1) Build a model based on the support vector machine algorithm. The topology of the support vector machine is determined by the support vectors, and the kernel function is used to map the nonlinear transformation to the linear transformation in the high-dimensional space, which not only ensures that the model has good generalization ability, but also solves the problem of "curse of dimensionality". It avoids the problem that the traditional neural network needs to try and determine the network structure. In addition, the model verification results show that the prediction effect of the SVM model is significantly better than that of the BP neural network in the regression fitting problem of small samples, and the local minimum problem of the BP neural network can be effectively avoided.

2) Optimization of model parameters by combining grid method and genetic algorithm. The traditional support vector machine parameter determination mostly adopts the empirical method or K-fold cross-validation method. Due to the different types of sample data, there are large differences in the selection of parameter values, which often leads to large model errors or long optimization calculation time. The grid method pre-training determines the optimization range of the model parameters, and then uses the genetic algorithm to optimize the above range. After verification, under the same sample data randomly selected, the mean square error between the measured value and the simulated value of the model constructed by this method is 0.000409, and the coefficient of determination is 0.9915, while the mean square error between the measured value and the simulated value of the model constructed by the K-fold cross-validation method is 0.000860, the coefficient of determination is 0.9769, and the prediction accuracy is effectively improved.

3) The model integrates time series, comprehensively considers the changes in plant photosynthetic rate in different growth stages, and effectively distinguishes the difference in photosynthetic rate values between the seedling stage and the flowering and fruiting stage of cucumber under different conditions by adding one-dimensional growth stage variables, and constructs the whole photosynthetic rate. predictive model. The prediction results show that the coefficient of determination between the predicted value and the measured value is 0.993, the maximum error is 3.571, and the relative error is less than 6.253%. The prediction accuracy is higher than that of the prediction model in the mixed growth period.

The full-range photosynthetic rate prediction model proposed by the invention can provide a theoretical basis for the regulation and control of the cucumber light environment, and can be expanded and applied to the establishment of photosynthetic optimization regulation models for different crops to improve the photosynthetic capacity of greenhouse crops.

Description of drawings

Fig. 1 is a flow chart of the present invention based on a support vector machine algorithm.

Fig. 2 is the prediction effect figure of different kernel function type models of the present invention, wherein Fig. 2 (a) represents linear kernel function prediction value, Fig. 2 (b) represents polynomial kernel function prediction value, Fig. 2 (c) represents radial basis kernel function Predicted value, Figure 2(d) shows the predicted value of sigmoid kernel function.

Fig. 3 is that the present invention determines model parameter optimization range based on the grid method, wherein Fig. 3 (a) represents the change of the mean square error of the training set sample with c, g, and wherein Fig. 3 (b) represents the variation of the mean square error of the test set sample with c, g changes.

Fig. 4 is a flowchart of optimizing model parameters based on genetic algorithm in the present invention.

Fig. 5 is a diagram of the evolution process of the parameters of the optimization model based on the genetic algorithm in the present invention.

Fig. 6 is a schematic diagram of the correlation between the measured value and the simulated value of the photosynthetic rate in the model verification of the present invention. Among them, Figure 6 (a) is a schematic diagram of the correlation between the measured photosynthetic rate and the simulated value in the photosynthetic rate model verification of plant seedling stage; Figure 6 (b) is the photosynthetic rate measured value and Schematic diagram of the correlation between the simulated values; Figure 6(c) is a schematic diagram of the correlation between the measured photosynthetic rate and the simulated value in the model verification of the photosynthetic rate of plants in the mixed growth period; Figure 6(d) is the photosynthetic rate in the mixed growth period Schematic diagram of the correlation between the measured and simulated values of photosynthetic rate in rate model validation.

detailed description

The implementation of the present invention will be described in detail below in conjunction with the drawings and examples.

The establishment process of a kind of whole-range photosynthetic rate prediction model based on neural network of cucumber of the present invention is as follows:

1. Test materials and methods

This experiment was carried out in the scientific research greenhouse of Northwest A&F University from April to July 2014. The cucumber variety to be tested is "Changchun Mici". The swollen seeds were germinated in a petri dish, and treated at low temperature when they were about to germinate. Seedlings were raised in a 50-hole (540mm280mm50mm) plug tray with a nutrient pot. The seedling raising substrate is a special substrate for agricultural seedling raising. During seedling cultivation, keep sufficient water and fertilizer, wait for cucumber seedlings to grow into two leaves and one heart, choose cucumber seedlings with uniform growth, stem diameter between 0.6 ~ 0.8 cm, and plant height within 10 cm for the test. 150 strong cucumber seedlings were selected as test samples. During the test period, normal field cultivation management was carried out without spraying any pesticides and hormones. When the cucumbers were in the flowering and fruiting stage, 150 plants whose flowering nodes were about 50 cm away from the faucet were selected as test samples for the flowering and fruiting stage.

The Li-6400XT portable photosynthetic instrument produced by American LI-COR Company was used to measure the net photosynthetic rate. During the test, multiple sub-modules selected by the photosynthetic instrument were used to control the temperature, CO2 concentration, light intensity and other parameters around the leaves as needed. Among them, use the temperature control module to set a total of 5 temperature gradients of 16, 20, 24, 28, and 32°C; use the _CO injection module to set a total of 5 gradients for the volume ratio of carbon dioxide to 300, 600, 900, 1200, and 1500 μL/L ;A total of 11 photon flux density (Photofluxdensity, PFD) gradients of 0, 20, 50, 100, 200, 300, 500, 700, 1000, 1200, 1500 μmol/(m ² s) were obtained by using the LED light source module, and embedded A total of 275 sets of tests were carried out in total, and each set of tests was repeated on 3 randomly selected plants. During the test, the relative humidity of the leaf chamber was recorded, and the chlorophyll of the measured leaves was recorded by the SPAD-502Plus chlorophyll meter of Japan Konica Company. Content, thus forming 1650 sets of experimental data with chlorophyll content, temperature, _CO2 concentration, light intensity, and relative humidity as input, and net photosynthetic rate as output, that is, 825 sets at the seedling stage and 825 sets at the flowering and fruiting stage.

2. Model building method

2.1 Basic theory of support vector machine regression

SVM was originally used to solve the classification problem of two types of samples in the case of linear separability (SVC). Its core idea is to find an optimal classification hyperplane to maximize the classification interval of two types of samples. When SVM is applied to regression fitting analysis, its basic idea is no longer to find an optimal classification plane to separate the two types of samples, but to find an optimal classification plane to minimize the error of all training samples from the optimal classification plane.

Without loss of generality, let the training set sample pair containing l training samples be {( _xi ,y _i ),i=1,2,3,…,l}, where, is the input column vector of the i-th training sample, d is the dimension of the column vector, is the i×d-dimensional real number field, y _i ∈ R is the corresponding output value.

Let the linear regression function established in the high-dimensional feature space be

f(x)=wΦ(x)+b(2-1)

Where x is the input vector, w is the weight vector, b is the bias, and Φ(x) is the nonlinear mapping function.

Define the ε linear insensitive loss function

Among them, f(x) is the predicted value returned by the regression function; y is the corresponding real value, which means that if the difference between the predicted value and the real value is less than or equal to ε, the loss is equal to 0.

For linear regression problems, the problem becomes to find an optimal hyperplane so that y can be fitted without error under the condition of a given accuracy (ε≥0), that is, the distance from all sample points to the optimal hyperplane is not greater than ε ; Considering the allowable error, the slack variable ξ _i can be introduced, and the optimization problem of ξ _i ^* ≥ 0 can be transformed into the corresponding quadratic programming problem as:

Among them, c is the penalty factor. The larger c means the greater the penalty for the sample whose training error is greater than ε. ε specifies the error requirement of the regression function. The smaller ε means the smaller the error of the regression function.

The solution problem of formula 2-3 can be transformed into a dual problem:

Among them, a _i and a _i ^* are the optimal solutions of formula (2-4).

Solving the above problems to solve the optimal regression function can be obtained as:

Where K( _xi ,x _j )=Φ( _xi )Φ(x _j ) is the kernel function.

2.2 Support vector machine kernel function selection

The kernel function of SVM is to convert nonlinearly separable samples into linearly separable feature space. Different selection of kernel function will make the classification hyperplane produced by SVM model different, resulting in large differences, which have a direct impact on the performance of SVM model. Impact. Therefore, the selection of kernel function is the key to affect the SVM prediction model. Commonly used kernel functions are: linear function, polynomial function, radial basis function, sigmoid function, etc. This fa.m performs model pre-training for the above four kernel functions, and uses the mean square error and the coefficient of determination as evaluation indicators. The prediction results of the model are shown in Table 1 and Figure 2.

Table 1 Effect of kernel function on model performance

Table 1 The impact on the performance of the model kernel

It can be seen from Table 1 that the radial basis function and the polynomial kernel function have smaller empirical errors than the linear function and the sigmoid function, but the generalization ability of the polynomial function is poor, and compared with the radial basis function, the polynomial function requires There are many parameters to be determined, and it can be seen from Figure 2 that the radial basis kernel function has the best fitting effect between the measured value and the simulated value. Therefore, considering the training effect and complexity, the radial basis function is selected as the kernel function of the SVM photosynthetic rate prediction model in this paper.

2.3 Support vector machine kernel parameter selection

SVM is a forecasting model based on statistical theory, the difficulty of using it for forecasting lies in the selection of model parameters. Forecasters often select appropriate parameters based on experience and trial and error. This does not guarantee that the model can converge to the global minimum, and the prediction results are naturally not guaranteed to be optimal. For actual operators who do not have relevant theoretical foundations, it is even more difficult to select the optimal parameters, which also limits the application of the SVM model. Support vector machine parameters include penalty factor c, radial basis kernel function parameter g, order p, abort training error ε, etc.

Among them, the penalty factor c is a coefficient specified by the user, indicating how much penalty to add to the wrong points. When c is large, the wrong points will be less, but the overfitting situation may be more serious. , when c is small, there may be many misclassified points, and the resulting model will not be correct.

Depending on the selection of the kernel parameter g, the shape of the function will change accordingly, which in turn will cause changes in the SVM model.

The present invention adopts the grid method to roughly select the parameter range. The radial basis function is selected as the model kernel function, the mean square error MSE is selected as the evaluation index, and the penalty factor c and the kernel function parameter g are respectively optimized, and the optimization range is lgc∈[-5,5], lgg∈[- 5,5], the optimization results determine the range of model parameters c and g as lgc∈[2,3], lgg∈[0,1], as shown in Figure 3.

Based on the above range, the genetic algorithm is used to optimize and determine the specific values of c and g parameters. The specific steps are shown in Figure 4, and the genetic evolution process is shown in Figure 5.

2.4 Construction of photosynthetic rate model based on support vector machine

The same modeling method was used to establish four models according to the different growth stages of cucumbers, namely, the prediction model only for the cucumber seedling stage, the prediction model only for the flowering and fruiting stage of cucumber, the photosynthetic rate prediction model for the whole cucumber and the growth stage. The difference is used as a one-dimensional input to establish a forecast model for the whole process of cucumber. The input signal is X'=(X ₁ 'X ₂ '…X ₅ ') ^T , X ₁ ', X ₂ ', X ₃ ', X ₄ ', X ₅ ' are temperature, CO ₂ concentration, light intensity, For relative humidity and chlorophyll content, the fourth model adds the growth period as a one-dimensional input, and the output signal uses T _o to represent the photosynthetic rate calculated by the network, and the corresponding measured photosynthetic rate of each group is the teacher signal T _d . The photosynthetic rate model T _d '(X') of cucumber seedlings was established by the support vector machine training method.

3 Model Training Performance Analysis

Based on the above test sample set, the support vector machine algorithm was used for training, and four models were obtained. The cucumber prediction model established at the seedling stage had a training error of 0.000446 and a coefficient of determination of 0.9883;

The cucumber prediction model established during the flowering growth period has a training error of 0.000387 and a coefficient of determination of 0.9900; the whole-process cucumber prediction model that mixes two growth periods has a training error of 0.0022 and a coefficient of determination of 0.9419; adding the growth period as a one-dimensional independent variable factor, A cucumber forecasting model that combines two growth periods was established, with a training error of 0.00021128 and a coefficient of determination of 0.9943. Comparing and analyzing the training results, it can be found that the staged training model has reached the expected error level, and the measured value of the training set has a good correlation with the simulated value; the whole cucumber prediction model mixed with two growth periods has a large training error, and the training The measured value and the simulated value have a good correlation, but the fusion prediction model that adds the growth period as a one-dimensional independent variable factor, the training error and the coefficient of determination are better than the staged model, and the model performance is optimized.

Based on the above results, the model established by adding the growth period as a one-dimensional factor has a significant effect, which can provide a theoretical basis and technical support for light environment regulation, and simplify the operation of light environment equipment.

4 Analysis of model verification results

The test sample set obtained by multi-factor nested experiment consists of 1650 two groups, and the samples are divided into training set and test set, of which 700 are used for model training, and the remaining 175 groups are used to form the test set, accounting for about 20% of the total samples , using the difference calibration method to verify the model, and obtain the correlation analysis between the measured value and the predicted value of photosynthetic rate, as shown in the figure. It can be seen from Figure 6 that the coefficient of determination of the correlation analysis between the measured value and the predicted value of the SVM machine model in Figure 6a is 0.988, the slope of the line is 0.984, the intercept is 0.2066, and the maximum prediction error is 1.4672. The measured value of the SVM model in Figure 6b is The coefficient of determination of the correlation analysis with the predicted value is 0.986, the slope of the line is 0.9864, the intercept is 0.3202, and the maximum prediction error is 2.7186. The coefficient of determination of the correlation analysis between the measured value and the predicted value of the SVM model in Figure 6c is 0.884, and the slope of the line is 0.9661 , the intercept is 0.7343, the maximum prediction error is 12.55, the coefficient of determination of the correlation analysis between the measured value and the predicted value of the SVM model in Figure 6d is 0.993, the slope of the straight line is 0.9923, the intercept is 0.05523, and the maximum prediction error is 3.575, considering the growth period to establish the model The linearity is significantly higher and the fitting degree is better.

The error analysis of the test results shows that the maximum relative error between the measured value and the simulated value of the whole photosynthetic rate prediction model established in consideration of the growth period is less than ±6.253%, indicating that the model established in this paper can predict the photosynthetic rate model in the whole growth period. good precision.

Claims (7)

1. a whole-range photosynthetic rate forecasting model of cucumber based on support vector machine, is characterized in that, model formula is: $f\left(x\right)=w\mathrm{\&Phi;}\left(x\right)+b=\underset{i=1}{\overset{l}{\&Sigma;}}(a_i-{a_i}^{*})\&Center\; Dot;\exp (-\frac{\left|\right|x-x_i\left|\right|}{2{\mathrm{\&sigma;}}^2})+b,$ Among them, the output f(x) represents the predicted photosynthetic rate, the input signal X'=(X ₁ 'X ₂ '…X ₅ ') ^T , X ₁ ', X ₂ ', X ₃ ', X ₄ ', X ₅ 'respectively temperature, CO ₂ concentration, light intensity, relative humidity and chlorophyll content, w is the weight vector, b is the bias, Φ(x) is the nonlinear mapping function, l is the training set sample pair {( _xi , y _i ), the number of training samples in i=1,2,3,...,l}, x _i is the input column vector of the i-th training sample, y _i is the corresponding output value, y _i ∈ R, is the i×d-dimensional real number field, d is the column vector dimension, a _i and a _i ^* are the optimal solutions of the following formula: $\left\{\begin{array}{c}\underset{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}^{<span\; class="notranslate">\; *</span>}}{\mathrm{<span\; class="notranslate">\; max</span>}}<span\; class="notranslate">\; \&lsqb;</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; l</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; l</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; K</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; l</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}<span\; class="notranslate">\; +</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; l</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&rsqb;</span>\\ \mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; .</span>\left\{\begin{array}{c}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; l</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; \&le;</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; c</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; \&le;</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; c</span>}\end{array}\right.\end{array}\right.$ is the kernel function, σ is the width parameter, ε is the abort training error, and c is the penalty factor. 2. the establishment method of the cucumber full-range photosynthetic rate prediction model based on support vector machine described in claim 1, is characterized in that, comprises the steps: Step 1, obtain experimental data, the process is as follows: Use a nutrient pot to grow seedlings. When the cucumber seedlings grow into two leaves and one heart, select cucumber seedlings with uniform growth, a stem diameter between 0.6 and 0.8 cm, and a plant height of less than 10 cm for the test. Select 150 strong cucumber seedlings as test samples. Treat that the cucumber is in the flowering and fruiting period, choose 150 plants whose flowering nodes are about 50 centimeters away from the faucet as the test sample of the flowering and fruiting period; Measure the net photosynthetic rate. During the process, use the temperature control module to set ₅ temperature gradients of 16, 20, 24, 28, and 32°C; There are 5 gradients in L; a total of 11 photon flux density gradients of 0, 20, 50, 100, 200, 300, 500, 700, 1000, 1200, 1500μmol/(m ² s) were obtained by using the LED light source module, to embedding A total of 275 groups of experiments were carried out in this way, and each group of experiments was repeatedly tested on 3 randomly selected plants. During the experiment, the relative humidity of the leaf chamber was recorded, and the chlorophyll content of the measured leaves was recorded, so as to form a model based on chlorophyll content, temperature, and _CO2 concentration. , light intensity and relative humidity are input, and net photosynthetic rate is output 1650 sets of experimental data, namely 825 sets at the seedling stage and 825 sets at the flowering and fruiting stage; Step 2, build a model Randomly select training samples and test samples from the experimental data obtained in step 1, then select the support vector machine kernel function type as radial basis function, and use the grid method to determine the support vector machine penalty factor c and radial basis kernel function parameter g In the optimization interval, the genetic algorithm is used to further optimize the parameters c and g based on the optimization interval until the maximum number of iterations is reached, the optimization results of the parameters c and g are output, and the whole-process photosynthetic rate prediction model of cucumber based on the support vector machine is constructed. 3. according to claim 2 based on the cucumber full-range photosynthetic rate prediction model based on support vector machine, the method for establishing, is characterized in that, in described step 2, The experimental data was normalized to the [0.2,0.9] interval to obtain the whole set of sample data, and the sample number was randomly selected according to the ratio of 4:1 to construct the training sample set and test sample set. 4. according to the described establishment method of the whole-range photosynthetic rate prediction model of cucumber based on support vector machine of claim 2, it is characterized in that, in described step 2, The optimization interval of penalty factor c and radial basis kernel function parameter g is expressed as lgc∈[2,3], lgg∈[0,1]. 5. according to the described establishment method of the cucumber full-range photosynthetic rate prediction model based on support vector machine of claim 2, it is characterized in that, in described step 2, The genetic algorithm is used to further optimize the optimization interval of the penalty factor c and the radial basis kernel function parameter g. The process is as follows: set the genetic algorithm parameters, the number of individuals NIND = 100, the maximum genetic algebra MAXGEN = 50, the generation gap GGAP = 0.95, The crossover probability and mutation probability are 0.8, 0.25, c and g are used as variables, respectively represented by 10-bit binary; define the objective function as the model mean square error value MSE; genetic operator operations: selection, crossover, mutation; calculate the target value of offspring Function and reinsert the child to the parent to obtain a new population; reach the maximum number of iterations and output the optimization results of c and g. 6. according to claim 2 based on the cucumber full-range photosynthetic rate prediction model based on support vector machine, the method for establishing, is characterized in that, in described step 2, The built model is: $f\left(x\right)=w\mathrm{\&Phi;}\left(x\right)+b=\underset{i=1}{\overset{l}{\&Sigma;}}(a_i-{a_i}^{*})\&CenterDot;\exp (-\frac{\left|\right|x-x_i\left|\right|}{2{\mathrm{\&sigma;}}^2})+b,$ Among them, the input signal X'=(X ₁ 'X ₂ '…X ₅ ') ^T , X ₁ ', X ₂ ', X ₃ ', X ₄ ', X ₅ ' are temperature, CO ₂ concentration, light intensity , relative humidity and chlorophyll content, w is a weight vector, b is a bias, and Φ(x) is a nonlinear mapping function. 7. according to claim 2 based on the whole-range photosynthetic rate prediction model of cucumber based on support vector machine, the establishment method is characterized in that, after obtaining the whole-range photosynthetic rate prediction model based on support vector machine, input sample data carries out photosynthetic efficiency prediction , if the accuracy requirement is met, output the prediction result, otherwise return to the genetic algorithm step, and further optimize c and g based on the optimization interval.