A chlorella biomass soft measurement method combining kinetics and xgboost

Abstract

The present application relates to the technical field of chlorella biomass soft measurement, and discloses a chlorella biomass soft measurement method combining kinetics and XGBoost, which solves the problem of insufficient measurement accuracy of traditional biomass. Firstly, a characteristic vector containing key environmental parameters such as nitrogen concentration, light intensity and culture temperature is constructed; through an XGBoost regression model, the maximum biomass of chlorella in the stable phase is accurately predicted in combination with experimental data, and at the same time, with the help of a multi-factor correction model, the intrinsic growth rate of chlorella is calculated; relying on a Logistic growth kinetics model, the parameters are organically combined to realize accurate prediction of the biomass of chlorella in any culture stage. The present application combines the interpretability of mechanism model and the high-precision advantage of machine learning, reduces experimental interference and cost, improves prediction stability through sample expansion and parameter optimization, can quickly respond to the biomass prediction demand of different culture conditions, and has strong practicability and good popularization.

Description

Technical Field

[0001] This invention relates to the field of soft measurement technology for Chlorella biomass, specifically to a soft measurement method for Chlorella biomass that combines kinetics and XGBoost. Background Technology

[0002] Chlorella, as a highly efficient photosynthetic microorganism, has significant application value in carbon emission reduction, functional foods, bio-feed, and biomass energy. Its growth process is influenced by multiple environmental factors such as nitrogen and phosphorus concentrations, light intensity, and temperature, exhibiting highly nonlinear, dynamic, time-varying, and multi-factor interactive characteristics. Accurate biomass acquisition and real-time estimation are crucial for efficient regulation of the cultivation process. Currently, biomass detection in Chlorella cultivation is mainly based on offline sampling and measurement, including dry weight methods, optical density methods, and cell counting methods. These methods generally suffer from long detection cycles, significant lag, susceptibility to contamination, and lack of real-time feedback, failing to meet the needs of online status estimation and real-time optimization control in both laboratory and industrial cultivation.

[0003] Traditional mechanistic models (such as Monod and Logistic models) have clear physical meanings and strong interpretability. Existing publicly available microalgal growth kinetic modeling methods often rely on static fitting or empirical settings for key parameters like the half-saturation constant, making them unable to adapt to complex and variable culture conditions and limiting their prediction accuracy in multi-factor collaborative scenarios. Data-driven models, such as neural networks and gradient boosting trees, while possessing strong fitting capabilities and fast prediction speeds, are mostly black-box models lacking growth kinetic constraints, exhibiting poor generalization ability and low extrapolation reliability in sparse data regions. Current technologies focus on the semi-continuous culture process of *Chlorella vulgaris*, neglecting biomass soft measurement and prediction modeling; they emphasize the identification of *Chlorella vulgaris* productive traits but lack real-time estimation capabilities for the growth process; and they focus on the carbon fixation effect of culture conditions but have not established high-precision soft measurement methods. Existing microalgal biomass detection and soft measurement technologies generally suffer from limited applicability, poor model generalization and robustness, and weak online real-time estimation capabilities. Summary of the Invention

[0004] This invention relates to a soft measurement method for Chlorella biomass that combines kinetics and XGBoost, in order to solve the problem of poor generalization and robustness of existing microalgal biomass detection and soft measurement technology models.

[0005] This invention employs the following technical solution: a soft measurement method for Chlorella biomass combining kinetics and XGBoost, comprising the following steps: S1. Select key environmental and operational parameters in the Chlorella cultivation process and construct feature vectors for model input; S2. Construct a kinetic model to describe the growth pattern of Chlorella, a multi-factor correction model to correct the growth rate, and an XGBoost regression model to predict the maximum biomass, respectively. S3. The Chlorella culture experiment dataset was preprocessed, features were constructed and normalized, and the XGBoost regression model was trained after dividing it into training and test sets. At the same time, the optimal parameter set of the dynamic model was identified through parameter optimization methods. S4. Input the target culture environment conditions, obtain the corresponding parameters through the trained XGBoost regression model and multi-factor correction model respectively, and substitute them into the kinetic model to realize the prediction of Chlorella biomass.

[0006] Furthermore, the feature vector for: ; In the formula, Nitrogen concentration, Phosphorus concentration, Light intensity, To cultivate temperature, This represents the initial inoculation density, i.e., the absorbance value at 750 nm.

[0007] Furthermore, the kinetic model is a Logistic growth kinetic model; The Logistic growth kinetic model is as follows: ; In the formula, Microalgal biomass is expressed as absorbance at 750 nm. The incubation period refers to the time from inoculation to measurement. The intrinsic growth rate is calculated using the modified model. Environmental carrying capacity is physically represented by the biomass of Chlorella, which is predicted by a machine learning model. The multi-factor correction model is a multi-factor intrinsic growth rate correction model. The multi-factor intrinsic growth rate correction model is as follows: ; In the formula, For theoretical maximum intrinsic growth, Nitrogen concentration is an influencing factor. Factors affecting phosphorus concentration, The light intensity influencing factor, To cultivate the influence of temperature, Factors influencing initial vaccination density; Nitrogen concentration influencing factors for: ; In the formula, This represents the actual nitrogen concentration. It is the nitrogen half-saturation constant; The phosphorus concentration influencing factors for: ; In the formula, This represents the actual phosphorus concentration. It is the phosphorus half-saturation constant; The light intensity influencing factor for: ; In the formula, Light intensity, It is the light half-saturation constant. It is the light suppression constant; The influencing factors of culture temperature for: ; In the formula, and For frequency factors, For the activity, For deactivation, The gas constant is... Absolute temperature; The initial vaccination density influencing factor for: ; In the formula, Initial inoculation density, This is the self-inhibition constant.

[0008] Furthermore, the input environmental variables of the XGBoost regression model The output is the biomass of Chlorella. ; The hyperparameters of the XGBoost regression model are set as follows: number of trees 300, maximum depth 5, learning rate 0.03, subsampling 0.8, column sampling 0.8, L1 regularization coefficient 0.01, and L2 regularization coefficient 0.01.

[0009] Furthermore, the preprocessing, feature construction, and normalization of the Chlorella culture experimental dataset, followed by dividing it into training and test sets, and then training the XGBoost regression model, includes grouping valid samples according to the same environmental factor conditions and extracting the maximum biomass during the stable period of each group. Label data used for training the XGBoost regression model; if the initial sample size is less than 50, the sample size is expanded to 15 times using interpolation, and then used in adjacent samples. and 14 points were evenly inserted between the points; nitrogen concentration was selected. Phosphorus concentration Light intensity Incubation temperature Initial vaccination density As a fundamental feature, the nitrogen-phosphorus ratio is constructed. Light and temperature interaction items As interactive features, the QuantileTransformer normalizer was used to normalize the quantiles of all features. The feature-engineered dataset was divided into training and test sets in an 8:2 ratio. test_size=0.2 and random_state=42 were set. The model performance evaluation metrics included the test set determination coefficient R² and root mean square error RMSE.

[0010] Furthermore, the identification of the optimal parameter set of the kinetic model through parameter optimization includes: the kinetic parameter identification adopts nonlinear least squares iterative optimization, with the goal of minimizing the residual between the measured biomass value and the model prediction value, to obtain the optimal kinetic parameter set; the maximum number of iterations is set to 8000, the function convergence threshold ftol=0.00000001, the gradient convergence threshold gtol=0.00000001, and the upper and lower bound constraints and initial values ​​of the parameters are set to ensure the physiological rationality and numerical stability of the parameters; The parameters include , , , , , , , , and .

[0011] Furthermore, S4 specifically includes: inputting environmental conditions and predicting stable biomass under the corresponding environment using an XGBoost model. Calculated using a multi-factor intrinsic growth rate correction model ,Will and By incorporating the Logistic growth model, it can quickly predict biomass at any time point to achieve offline / online soft measurement; it also supports prediction under new conditions and can output biomass time series, maximum biomass during the stationary period, and growth curve.

[0012] Compared with the prior art, the present invention has the following beneficial effects: This invention combines a Logistic kinetic model with an XGBoost regression model, preserving the strong interpretability of the kinetic model while leveraging machine learning to improve biomass prediction accuracy, thus addressing the limitations of traditional single-model approaches. The XGBoost model accurately predicts the maximum biomass during the stationary phase, enabling non-invasive, sampling-free measurement of Chlorella biomass, reducing experimental costs and interference. Furthermore, through sample interpolation expansion and standardized training, it maintains high prediction accuracy even in small-sample scenarios, adapting to various culture conditions and providing reliable data support for optimizing Chlorella culture processes. It is highly practical and easy to promote. Attached Figure Description

[0013] Figure 1 This is a flowchart of a soft biomass measurement method for the growth process of Chlorella.

[0014] Figure 2 The training results of the Chlorella XGBoost model are shown.

[0015] Figure 3 The fitting effect diagram of the Chlorella growth model.

[0016] Figure 4 This is a growth curve of Chlorella under given growth conditions. Detailed Implementation

[0017] The present invention will be further illustrated below with reference to embodiments. These embodiments are for illustrative purposes only and are not intended to limit the invention in any way. It should be understood that the described embodiments are merely some, not all, of the embodiments described in this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0018] like Figure 1 As shown, a soft measurement method for Chlorella biomass combining kinetics and XGBoost includes the following steps: Chlorella sp. was cultured in BG11 medium, and biomass was determined by measuring absorbance at 750 nm. The UV-Vis spectrophotometer was zeroed using sterile deionized water as a blank control. Take 2–5 mL of aseptic sample from the culture system according to the set culture time (experimental cycle of 16 days, daily measurement), shake well and inject into a quartz cuvette; if the algal solution concentration is too high, dilute appropriately with sterile deionized water. Absorbance was measured at a wavelength of 750 nm. Three parallel samples were set up for each group, and the average value was taken as the Chlorella biomass (OD) at that time. 750 ); Data from Chlorella culture experiments under different environmental conditions were collected, with the following variable range: Nitrogen concentration is 0–494 mg / L; Phosphorus concentration ranged from 0 to 14.26 mg / L; Light intensity is 0–1000 μmol・m⁻²・s⁻¹; Temperature range: 0–38℃; Initial inoculation density, i.e., OD 750 The value is 0.01–0.1. The final biomass of the data collected in the specific experiments is shown in Table 1.

[0019] Table 1 Final biomass of the experiment .

[0020] Based on the final biomass data in Table 1, an XGBoost regression model was constructed, with environmental factors as input. Maximum biomass during the stationary period This is the output.

[0021] To improve model stability, the sample size was increased by 15 times using interpolation; subsequently, it was divided into training and test sets in an 8:2 ratio. The model hyperparameters were set as follows: Number of trees: 300; maximum depth: 5; learning rate: 0.03; subsampling rate: 0.8; column sampling rate: 0.8; L1 regularization coefficient: 0.01; L2 regularization coefficient: 0.01. Training results show that the XGBoost model has a training set determination coefficient R² = 0.9997 and a root mean square error RMSE = 0.0190; and a validation set determination coefficient R² = 0.9854 and a root mean square error RMSE = 0.1639, indicating excellent mapping fit between environmental factors and parameters. Figure 2 The training results of the Chlorella XGBoost model are shown in the image. Figure 2 As shown.

[0022] Based on all experimental data, the key parameters of the Logistic kinetic model and the multi-factor intrinsic growth rate correction model were globally identified using the nonlinear least squares method. The optimal parameter set was solved with the goal of minimizing the residual between the measured biomass value and the model prediction value.

[0023] To ensure the physiological rationality and numerical stability of the parameters, the upper and lower bound constraints and initial values ​​of the parameters are set as follows: ; ; ; Parameter identification results showed that the overall growth model had a determination coefficient R² = 0.9332 and a root mean square error RMSE = 0.4225. The kinetic model accurately matched experimental biomass data under different environmental conditions, and the model fit was as follows: Figure 3 As shown.

[0024] A set of culture conditions not directly appearing in Table 1 was selected for validation. The input parameters are as follows: Nitrogen concentration = 247 mg / L, phosphorus concentration = 7.13 mg / L, light intensity = 100 μmol・m⁻²・s⁻¹, temperature = 26℃, initial inoculation density = 0.01; The method of this invention first uses a trained XGBoost model to predict the maximum biomass during the stationary phase under this condition. The intrinsic growth rate was then calculated using a multi-factor correction model, and finally substituted into the Logistic kinetic model to predict the biomass change of Chlorella under this condition during a 16-day culture period.

[0025] The model predicts that the biomass after 16 days will be 4.249 (OD). 750 The growth curve of Chlorella under the given growth conditions in the model is shown in the figure below. Figure 4 As shown.

[0026] The model-predicted 16-day biomass was compared with the measured value of 4.594 (OD) under the same conditions. 750 By comparison, the relative error was calculated to be 7.510%. The results show that the prediction accuracy of the kinetic and XGBoost fusion model proposed in this invention is significantly better than that of the traditional static mechanism model, and it can realize accurate and rapid soft measurement of Chlorella biomass under different culture conditions.

[0027] The invention utilizes a fusion framework of "growth kinetics + XGBoost machine learning" to achieve high-precision and high-reliability soft measurement of Chlorella biomass. It can quickly estimate the time-varying patterns of biomass under different culture conditions and can be applied to laboratory research, process analysis, and optimization of the culture process of Chlorella.

[0028] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A soft measurement method for Chlorella biomass combining kinetics and XGBoost, characterized in that, Includes the following steps: S1. Select key environmental and operational parameters in the Chlorella cultivation process and construct feature vectors for model input; S2. Construct a kinetic model to describe the growth pattern of Chlorella, a multi-factor correction model to correct the growth rate, and an XGBoost regression model to predict the maximum biomass, respectively. S3. The Chlorella culture experiment dataset was preprocessed, features were constructed and normalized, and the XGBoost regression model was trained after dividing it into training and test sets. At the same time, the optimal parameter set of the dynamic model was identified through parameter optimization methods. S4. Input the target culture environment conditions, obtain the corresponding parameters through the trained XGBoost regression model and multi-factor correction model respectively, and substitute them into the kinetic model to realize the prediction of Chlorella biomass.

2. The method for soft measurement of Chlorella biomass combining kinetics and XGBoost according to claim 1, characterized in that, The feature vector for: ； In the formula, Nitrogen concentration, Phosphorus concentration, Light intensity, To cultivate temperature, This represents the initial inoculation density, i.e., the absorbance value at 750 nm.

3. The method for soft measurement of Chlorella biomass combining kinetics and XGBoost according to claim 1, characterized in that, The kinetic model is a Logistic growth kinetic model; The Logistic growth kinetic model is as follows: ； In the formula, Microalgal biomass is expressed as absorbance at 750 nm. The incubation period refers to the time from inoculation to measurement. The intrinsic growth rate is calculated using the modified model. Environmental carrying capacity and Chlorella biomass are predicted by machine learning models. The multi-factor correction model is a multi-factor intrinsic growth rate correction model. The multi-factor intrinsic growth rate correction model is as follows: ； In the formula, For theoretical maximum intrinsic growth, Nitrogen concentration is an influencing factor. Factors affecting phosphorus concentration, The light intensity influencing factor, To cultivate temperature-related factors, Factors influencing initial vaccination density; Nitrogen concentration influencing factors for: ； In the formula, This represents the actual nitrogen concentration. It is the nitrogen half-saturation constant; The phosphorus concentration influencing factors for: ； In the formula, This represents the actual phosphorus concentration. It is the phosphorus half-saturation constant; The light intensity influencing factor for: ； In the formula, Light intensity, It is the light half-saturation constant. It is the light suppression constant; The influencing factors of culture temperature for: ； In the formula, and For frequency factors, For the activity, For deactivation, The gas constant is Absolute temperature; The initial vaccination density influencing factor for: ； In the formula, Initial inoculation density, This is the self-inhibition constant.

4. The method for soft measurement of Chlorella biomass combining kinetics and XGBoost according to claim 1, characterized in that, The input environmental variables of the XGBoost regression model The output is the biomass of Chlorella. ; The hyperparameters of the XGBoost regression model are set as follows: number of trees 300, maximum depth 5, learning rate 0.03, subsampling 0.8, column sampling 0.8, L1 regularization coefficient 0.01, and L2 regularization coefficient 0.

01.

5. The method for soft measurement of Chlorella biomass combining kinetics and XGBoost according to claim 1, characterized in that, The process of preprocessing, feature construction, and normalization of the Chlorella culture experiment dataset, followed by dividing it into training and test sets, and then training the XGBoost regression model, includes grouping valid samples according to the same environmental factors and extracting the maximum biomass during the stable period of each group. Label data used for training the XGBoost regression model; if the initial sample size is less than 50, the sample size is expanded to 15 times using interpolation, and then used in adjacent samples. and 14 points were evenly inserted between the points; nitrogen concentration was selected. Phosphorus concentration Light intensity Incubation temperature Initial vaccination density As a fundamental feature, the nitrogen-phosphorus ratio is constructed. Light and temperature interaction items As interactive features, the QuantileTransformer normalizer was used to normalize the quantiles of all features. The feature-engineered dataset was divided into training and test sets in an 8:2 ratio. test_size=0.2 and random_state=42 were set. The model performance evaluation metrics included the test set determination coefficient R² and root mean square error RMSE.

6. The method for soft measurement of Chlorella biomass combining kinetics and XGBoost according to claim 1, characterized in that, The method of identifying the optimal parameter set of the kinetic model through parameter optimization includes: using nonlinear least squares iterative optimization to minimize the residual between the measured biomass and the model prediction, thereby obtaining the optimal kinetic parameter set; setting a maximum number of iterations of 8000, a function convergence threshold of ftol=0.00000001, and a gradient convergence threshold of gtol=0.00000001, while setting upper and lower bound constraints and initial values ​​for the parameters to ensure the physiological rationality and numerical stability of the parameters; The parameters include , , , , , , , , and .

7. The method for soft measurement of Chlorella biomass combining kinetics and XGBoost according to claim 1, characterized in that, S4 specifically includes: inputting environmental conditions and predicting stable biomass under the corresponding environment using the XGBoost model. Calculated using a multi-factor intrinsic growth rate correction model ,Will and By incorporating the Logistic growth model, it can quickly predict biomass at any time point to achieve offline / online soft measurement; it also supports prediction under new conditions and can output biomass time series, maximum biomass during the stationary period, and growth curve.

Images