Method for predicting microalgae magnetic flocculation harvesting efficiency
By extracting the key parameters of microalgae magnetic flocculation and building multiple HE prediction models, the problem of the prediction model of microalgae magnetic flocculant in the existing technology ignores physical and chemical parameters, and the accurate prediction of the harvesting efficiency of different microalgae and magnetic flocculant is achieved, expanding the scope of application of the model and improving the prediction accuracy.
Patent Information
- Application Number
- CN202411781589.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-05
- Publication Date
- 2025-03-28
AI Technical Summary
The prediction model of microalgae magnetic flocculants in the prior art ignores the important physical and chemical parameters of microalgae and magnetic flocculants, resulting in the inability to accurately predict the harvesting efficiency of other types of microalgae and magnetic flocculants.
Key parameters of microalgae magnetic flocculation were extracted, including the physical and chemical properties parameters of microalgae, the physical and chemical properties parameters of magnetic flocculant, and the magnetic flocculation process parameters, and multiple microalgae magnetic flocculation HE prediction models were constructed, and data analysis and model evaluation were used to determine the best prediction model.
It realizes accurate prediction of the harvesting efficiency of different microalgae and magnetic flocculants, expands the scope of application of the model, improves prediction accuracy and accuracy, and has stronger generalization capabilities.
Smart Images
Figure CN119724409A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of harvesting efficiency prediction, and in particular to a method for predicting the harvesting efficiency of microalgae magnetic flocculation. Background Art
[0002] Microalgae is an economical and efficient biomass raw material. Compared with other biomass raw materials, microalgae has the advantages of not occupying arable land, high growth efficiency, short growth cycle and high oil content. It is one of the important raw materials for high value-added products such as food, health products, biofuels, medicines and chemicals. However, since microalgae live in water bodies and have the characteristics of small size, strong stability and low cell density, it requires certain technical means to separate microalgae from water bodies and efficiently harvest microalgae. It is estimated that the cost of harvesting microalgae can reach 20-30% of the total cost of the microalgae biomass utilization process. Therefore, an efficient and low-cost microalgae harvesting technology is needed.
[0003] In the existing technology, the separation and collection process of microalgae from water bodies includes different technologies such as centrifugation, filtration, sedimentation, flotation and flocculation. Among them, the microalgae magnetic flocculation technology in the flocculation technology utilizes the high affinity between nano-scale (diameter not exceeding 1000nm) magnetic flocculants and microalgae, and can achieve efficient collection of microalgae in a few minutes under the action of an external magnetic field. It has a very high collection efficiency, and the magnetic flocculants (MNPs) used in magnetic flocculation can also be recycled, and the collection cost is low. Therefore, it has received widespread attention in recent years. There are a variety of magnetic flocculants available on the market, which are mainly divided into two categories for microalgae magnetic flocculation: non-coated magnetic flocculants (NMNPs) and functionalized magnetic flocculants (FMNPs). The former are composed only of magnetic iron oxides (such as Fe3O4), while the latter are obtained by coating functional materials on NMNPs. The functional materials include nitrogen-containing cationic functional materials (polyethyleneimine, polyacrylamide, polyamidoamine and quaternized plant polyphenols, etc.) and anionic plant tannins. The use of these functional materials can improve the harvesting efficiency (HE) of magnetic flocculants.
[0004] HE is a key indicator for evaluating the magnetic flocculation performance of MNPs on microalgae, which determines whether the MNPs used can be used for microalgae harvesting. Currently, more than 40,000 species of microalgae have been discovered, many of which are potential species for obtaining high-value biological products. There are many types of microalgae, and there are also many types of magnetic flocculants. In addition, due to the large differences in the physical and chemical properties of different varieties of microalgae and magnetic flocculants, the harvesting efficiency (HE) of the same microalgae using different magnetic flocculants (MNPs) varies greatly, and the same magnetic flocculant also shows different harvesting efficiencies for different microalgae. Therefore, when developing a new magnetic flocculant, how does the magnetic flocculant perform on the harvesting efficiency (HE) of different microalgae; when facing a new variety of microalgae, which magnetic flocculant has the best harvesting efficiency, the existing technical means are difficult to predict. It is a time-consuming and laborious project to determine the HE of magnetic flocculants for different microalgae through batch experiments, or to screen the HE of different magnetic flocculants for new varieties of microalgae through batch experiments. The workload is huge and the efficiency is extremely low. A potential solution is to use computer models to predict the HE of microalgae magnetic flocculation process. However, only a few researchers have published the use of simple polynomial models to predict HE, such as Ajala et al., SN Appl. Sci. 2020, 387, Boli et al., Sep. Sci. Technol. 2021, 730-737, Yin et al., Sep. Purif. Technol. 2021, 119676 all disclose prediction models for microalgae HE. Although these models introduce parameters such as microalgae concentration, magnetic flocculation pH, MNPs dosage and flocculant dosage mass ratio (MNPs dosage / microalgae dry weight), they only simulate and predict the HE of a specific MNPs or a specific microalgae strain, focusing too much on the specificity of the microalgae strain and ignoring the physical and chemical properties of the microalgae. As a result, there is model overfitting. These models cannot accurately predict the HE of other types of MNPs and algae species, which in turn limits the wide application of these prediction models. Summary of the invention
[0005] The purpose of the present invention is to provide a method for predicting the recovery efficiency of microalgae magnetic flocculation, so as to solve the technical problem that the important physical and chemical parameters of microalgae and magnetic flocculation itself are ignored in the prediction model of microalgae-magnetic flocculant in the prior art, resulting in the prediction model being unable to or unable to accurately predict the HE of other types of MNPs and algae species.
[0006] In order to solve the above problems, the technical solution adopted by the present invention is as follows: A method for predicting the efficiency of microalgae magnetic flocculation recovery comprises the following steps: Step 1, extracting key parameters: extracting key parameters of microalgae magnetic flocculation that affect the microalgae magnetic flocculation process and HE, wherein the key parameters of microalgae magnetic flocculation include three categories: microalgae physicochemical property parameters, MNPs physicochemical property parameters and magnetic flocculation process parameters; The physical and chemical property parameters of the microalgae include the dry weight of the microalgae, the zeta potential of the microalgae cells, and the diameter of the microalgae cells; the physical and chemical property parameters of the MNPs include the concentration of the MNPs, the coating ratio, the nitrogen content of the coating material, the quaternary ammonium nitrogen content, the oxygen content, the zeta potential of the MNPs, and the diameter of the MNPs; the magnetic flocculation process parameters include the pH of the microalgae magnetic flocculation, the magnetic separation time t of the flocs, and the magnetic flocculation mass ratio; Step 2, data collection and processing: Relevant research paper data were searched by keywords in Web of Science. When the proportion of missing key parameters of microalgae magnetic flocculation data in the paper was greater than 1 / 3, they were removed and the rest were retained; Step 3, removing redundant parameters: using the Pearson correlation coefficient r to judge the key parameters of the microalgae magnetic flocculation that have completed data collection and processing in step 2, obtaining the redundant parameters deleted by the user, and removing the strongly correlated redundant parameters; Step 4, constructing a data set: constructing a microalgae magnetic flocculation data set based on the key parameters of microalgae magnetic flocculation obtained after processing in step 3, the data set includes the key parameters of microalgae magnetic flocculation and HE; Step 5, model establishment: using the microalgae magnetic flocculation dataset, multiple microalgae magnetic flocculation HE prediction models are established based on one or more machine learning regression algorithms; Step 6: All prediction models in step 5 are evaluated using the determination coefficient R2, root mean square error RMSE and mean absolute error MAE, and the best prediction model for HE is determined accordingly.
[0007] The beneficial effects of this embodiment are: 1. The microalgae prediction model in the prior art only introduces parameters such as microalgae concentration, magnetic flocculation pH, MNPs dosage and flocculant dosage mass ratio (MNPs dosage / microalgae dry weight). It is just a point-to-point study of a simple model, and does not focus on the physical and chemical properties of microalgae and magnetic flocculants themselves. It comprehensively introduces other important parameters that can significantly affect the magnetic flocculation of microalgae to form a comprehensive system for big data analysis, such as: the size and zeta potential of MNPs, the zeta potential of microalgae, the nitrogen and oxygen content of the coating material, etc., while this application systematically considers the factors that may affect the magnetic flocculant and microalgae in the process of magnetic flocculation of microalgae harvesting, and screens out all the important key parameters about microalgae magnetic flocculation from the physical and chemical properties of microalgae, the physical and chemical properties of magnetic flocculants MNPs and the magnetic flocculation process conditions, which can better and more comprehensively reflect the differences in the effects of different microalgae strains, magnetic flocculants and magnetic flocculation process conditions on the HE of microalgae magnetic flocculation, so its prediction results are closer to the actual value and more accurate.
[0008] 2. Most of the models in the prior art only simulate and predict the HE of a specific MNPs or a specific microalgae strain, and pay too much attention to the specificity of the microalgae strain, so there is model overfitting. Moreover, since these models do not have relevant parameters of other microalgae and MNPs, they cannot accurately predict the HE of other types of MNPs or algae species, which limits the wide application of these prediction models. However, when extracting data, the prediction model proposed in this application ignores the variety of microalgae and the type of MNPs, and only extracts the relevant physicochemical parameter data of microalgae and MNPs. These physicochemical property parameters are used to fit the commonalities of the effects of microalgae and MNPs on HE in the process of microalgae magnetic flocculation, and these physicochemical parameter data are associated with the microalgae magnetic flocculation HE. Therefore, it is only necessary to know the physicochemical property parameters of microalgae (microalgae dry weight, microalgae cell zeta potential, microalgae cell diameter) and the relevant physicochemical property parameters of MNPs (MNPs concentration, coating ratio, coating material nitrogen content, quaternary ammonium nitrogen content, oxygen content, MNPs zeta potential, MNPs diameter d2). Regardless of the algae strain of microalgae or MNPs, the final microalgae magnetic flocculation HE can be predicted by the above model, and its scope of application is wider than that of existing models. The existing HE prediction model is only applicable to one microalgae species and magnetic flocculant, while the prediction model proposed in the present invention can theoretically predict the HE of different algae strains and different magnetic flocculants, which is helpful for predicting which magnetic flocculants are suitable for harvesting specific algae strains from a variety of magnetic flocculants, and can also quickly estimate the harvesting effect of specific magnetic flocculants on different algae strains.
[0009] 3. The prediction accuracy and precision of the model proposed in the present invention are better than those of the existing prediction models, and it has stronger generalization ability. Compared with the prediction models in the prior art (Boli et al., 2021, Sep. Sci. Technol., 730-737; Yin et al., 2021, Sep. Purif. Technol., 119676), the model constructed in the present invention shows better HE prediction effect on the MNPs test set (50 training set and test set combinations: R2 = 0.901 ± 0.019, RMSE = 8.56% ± 0.94%, MAE = 5.06% ± 0.48%), which is significantly better than Ajala et al. (50 training set and test set combinations: R2 = -1.112 ± 0.298, RMSE = 39.66% ± 3.53%, MAE = 25.85% ± 1.97%), Boli et al. (50 training and test set combinations: R2 = -0.587 ± 0.063, RMSE = 34.45% ± 1.87%, MAE = 23.35% ± 1.48%) and Yin et al. (50 training and test set combinations: R2 = -28820.381 ±15528.487, RMSE =4413.97% ± 1280.39%, MAE = 645.42% ± 265.33%). The above results show that the prediction method proposed in the present invention can accurately predict the HE value.
[0010] Furthermore, the machine learning algorithm is any one of linear regression, decision tree, random forest, gradient boosting, lightweight gradient boosting machine and extreme gradient boosting, or a combination of two or more machine learning algorithms.
[0011] Furthermore, in step 5, before constructing the HE prediction model, the microalgae magnetic flocculation dataset is first divided into a training set and a test set at a ratio of 80%:20%.
[0012] Furthermore, the two key parameters of microalgae magnetic flocculation in step 3 whose absolute values of the Pearson correlation coefficient r are greater than or equal to 0.80 are redundant parameters. The redundant parameters deleted by the user are obtained, and only one of them is retained.
[0013] Furthermore, in step 2, when the missing key parameters of microalgae magnetic flocculation are less than or equal to 1 / 3, interpolation is performed using typical values or microalgae mean values, wherein the typical value is the average value of the missing parameters corresponding to the same type of microalgae in the extracted data, and the microalgae mean value is the average value of the missing parameters corresponding to all microalgae. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] Figure 1The figure shows the modeling prediction process of the present invention. DETAILED DESCRIPTION
[0015] The following is further described in detail through specific implementation methods: Implementation example Figure 1 As shown: Step 1, extract key parameters: extract key parameters of microalgae magnetic flocculation that affect HE in the microalgae magnetic flocculation process. These parameters include three categories: microalgae physicochemical property parameters, MNPs physicochemical property parameters and magnetic flocculation process parameters; There are 13 key parameters of microalgae magnetic flocculation that affect microalgae magnetic flocculation in step 1, including three parameters of microalgae physicochemical properties: microalgae dry weight (m1), microalgae cell zeta potential ζ1 (mV), and microalgae cell diameter d1 (μm); There are 7 physical and chemical property parameters of MNPs: MNPs concentration (m2), coating ratio α (coating material mass / NMNPs mass), coating material nitrogen content N%, quaternary ammonium nitrogen content (QN%), oxygen content (O%), MNPs zeta potential ζ2 (mV), and MNPs diameter d2 (μm).
[0016] There are three magnetic flocculation process parameters: microalgae magnetic flocculation pH, floc magnetic separation time t (s), and magnetic flocculation mass ratio β (MNPs concentration / microalgae concentration).
[0017] The essence of flocculation is to destroy the balance between the dispersed matter and the dispersant in the dispersed system. The important parameters for maintaining these balances are the diameter and zeta potential of the dispersed matter. In this scheme, microalgae and MNPs are included as important physical and chemical property parameters in the equilibrium process of the dispersed matter into the main extraction data.
[0018] Step 2, data collection and processing: Relevant research paper data were searched by keywords in Web of Science. When a set of data on microalgae magnetic flocculation data in the paper was missing more than 1 / 3 of the key parameters of microalgae magnetic flocculation, it was eliminated and the rest was retained.
[0019] Based on the 13 key parameters of microalgae magnetic flocculation extracted in step 1, papers with key parameters of microalgae magnetic flocculation were mined in the paper database Web of Science, and the key parameters of microalgae magnetic flocculation in the papers were extracted. The papers in the paper database are from publicly published peer-reviewed research papers.
[0020] When the missing parameters of a group of key parameters of microalgae magnetic flocculation data in the mined papers are greater than 1 / 3, they are directly removed as noise data. If they are less than or equal to 1 / 3, the missing values of the parameters are first interpolated using the typical values in the extracted data. If there is no typical value, the mean value of the microalgae is used for interpolation. The typical value is the average value of the missing parameters corresponding to the same type of microalgae in the extracted data, and the mean value of the microalgae is the average value of the missing parameters corresponding to all microalgae in all extracted data.
[0021] Step 3, eliminating redundant parameters: using the Pearson correlation coefficient r to judge the key parameters of microalgae magnetic flocculation that completed data collection and processing in step 2, removing strongly correlated parameters to reduce the data dimension in the data set; For the 13 key parameters of microalgae magnetic flocculation that have completed data preprocessing in step 2, the linear relationship between the parameters is evaluated using the Pearson correlation coefficient r; the calculation formula of the Pearson correlation coefficient r is:
[0022] Where x, y represent input parameters, and represent their average values, i represents the sample ordinal number, and n represents the number of samples in the data set.
[0023] The above formula is used to calculate the Pearson correlation coefficient r between 13 different key parameters of microalgae magnetic flocculation. If the absolute value of the Pearson correlation coefficient r is greater than or equal to 0.80, it is considered that there are redundant parameters, and the redundant parameters deleted by the user are obtained. Delete one of them, so in the modeling process, for the two parameters with an absolute value of the Pearson correlation coefficient r greater than or equal to 0.80, only one of the parameters is retained according to the user's choice, and usually the one with a lower detection cost is retained.
[0024] After parameter screening according to the above rules, 11 key parameters of microalgae magnetic flocculation were finally retained: ζ1, d1, α, N%, QN%, O%, ζ2, d2, pH, t, β.
[0025] Step 4, constructing a data set: constructing a microalgae magnetic flocculation data set based on the 11 microalgae magnetic flocculation key parameters obtained after processing in step 3. The data set includes the 11 microalgae magnetic flocculation key parameters and HE.
[0026] Step 5, model establishment: multiple microalgae magnetic flocculation HE prediction models are established based on one or more machine learning regression algorithms. Before building the HE prediction model, the microalgae magnetic flocculation data set was first split into a training set and a test set at a ratio of 80%:20% to avoid overfitting of the prediction model. Considering that the splitting method of the training set and the test set will affect the prediction performance of the prediction model, and in order to avoid the high or low prediction performance of the model due to some special splitting, 50 random splitting methods (corresponding to 50 training set and test set combinations) were used in the data splitting process. Based on machine learning regression algorithms such as linear regression, decision tree, random forest, gradient boosting, lightweight gradient boosting machine and extreme gradient boosting, the microalgae magnetic flocculation HE prediction model was established using any one or a combination of two or more machine learning regression algorithms.
[0027] Step 6: All prediction models in step 5 are evaluated using the determination coefficient R2, root mean square error RMSE and mean absolute error MAE, and the best prediction model for HE is determined accordingly.
[0028] The 10-fold cross validation method was used in the model training process, and Bayesian optimization was used to determine the optimal hyperparameters of the machine learning algorithm. In the modeling process, the three indicators of determination coefficient R2, root mean square error RMSE and mean absolute error MAE were used to evaluate the performance of the prediction model. The calculation formulas of these three evaluation indicators are as follows:
[0029]
[0030]
[0031] In the above formula, i represents the sample ordinal number, represents the observed value of HE, represents the predicted value of HE, represents the average value of the observed value of HE, and n represents the number of samples in the data set. Obviously, when the determination coefficient R2 of the prediction model is closer to 1, the prediction ability of the model is stronger; if R2 is smaller, the goodness of fit of the model is worse. When the RMSE and MAE values are smaller, the prediction performance of the model is better. Therefore, in the modeling process, the algorithm with the largest determination coefficient R2 and smaller RMSE and MAE will be selected as the optimal model to predict the harvesting efficiency HE of microalgae magnetic flocculation.
[0032] The results of machine learning modeling show that the HE prediction model based on the XGBoost algorithm has good prediction performance. Compared with the prediction models proposed by Boli and Yin et al. (Boli et al., 2021, Sep. Sci. Technol., 730-737; Yin et al., 2021, Sep. Purif. Technol., 119676), the model constructed in the present invention showed better HE prediction effect on the MNPs test set (50 training and test set combinations: R2 = 0.901 ± 0.019, RMSE = 8.56% ± 0.94%, MAE = 5.06% ± 0.48%), significantly better than Ajala et al. (50 training and test set combinations: R2 = -1.112 ± 0.298, RMSE = 39.66% ± 3.53%, MAE = 25.85% ± 1.97%), Boli et al. (50 training and test set combinations: R2 = -0.587 ± The above results show that the prediction method proposed in the present invention can accurately predict the HE value.
[0033] The above is only an embodiment of the present invention, and the common knowledge such as the known specific structure and characteristics in the scheme is not described in detail here. It should be pointed out that for those skilled in the art, several deformations and improvements can be made without departing from the structure of the present invention, which should also be regarded as the protection scope of the present invention, and these will not affect the effect of the implementation of the present invention and the practicality of the patent. The scope of protection required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.
Claims
1. A method for predicting the efficiency of microalgae magnetic flocculation harvesting, comprising the following steps: Step 1, extracting key parameters: extracting key parameters of microalgae magnetic flocculation that affect the microalgae magnetic flocculation process and HE, wherein the key parameters of microalgae magnetic flocculation include three categories: microalgae physicochemical property parameters, MNPs physicochemical property parameters and magnetic flocculation process parameters; The physical and chemical property parameters of the microalgae include the dry weight of the microalgae, the zeta potential of the microalgae cells, and the diameter of the microalgae cells; the physical and chemical property parameters of the MNPs include the concentration of the MNPs, the coating ratio, the nitrogen content of the coating material, the quaternary ammonium nitrogen content, the oxygen content, the zeta potential of the MNPs, and the diameter of the MNPs; the magnetic flocculation process parameters include the pH of the microalgae magnetic flocculation, the magnetic separation time t of the flocs, and the magnetic flocculation mass ratio; Step 2, data collection and processing: Relevant research paper data were searched by keywords in Web of Science. When the proportion of missing key parameters of microalgae magnetic flocculation data in the paper was greater than 1 / 3, they were removed and the rest were retained; Step 3, removing redundant parameters: using the Pearson correlation coefficient r to judge the key parameters of the microalgae magnetic flocculation that have completed data collection and processing in step 2, obtaining the redundant parameters deleted by the user, and removing the strongly correlated redundant parameters; Step 4, constructing a data set: constructing a microalgae magnetic flocculation data set based on the key parameters of microalgae magnetic flocculation obtained after processing in step 3, the data set includes the key parameters of microalgae magnetic flocculation and HE; Step 5, model establishment: using the microalgae magnetic flocculation dataset, multiple microalgae magnetic flocculation HE prediction models are established based on one or more machine learning regression algorithms; Step 6, all prediction models in step 5 are evaluated using the determination coefficient R2, root mean square error RMSE and mean absolute error MAE, and the best prediction model for HE is determined accordingly; In step 2, when the missing key parameters of microalgae magnetic flocculation are less than or equal to 1 / 3, the missing parameters are interpolated using typical values or microalgae mean values, wherein the typical value is the average value of the missing parameters corresponding to the same type of microalgae in all extracted data, and the microalgae mean value is the average value of the missing parameters corresponding to all microalgae in all extracted data.
2. The method for predicting the efficiency of microalgae magnetic flocculation recovery according to claim 1, characterized in that: The machine learning algorithm is any one of linear regression, decision tree, random forest, gradient boosting, lightweight gradient boosting machine and extreme gradient boosting, or a combination of two or more machine learning algorithms.
3. The method for predicting the efficiency of microalgae magnetic flocculation recovery according to claim 1, characterized in that: In step 5, before constructing the HE prediction model, the microalgae magnetic flocculation dataset is first divided into a training set and a test set at a ratio of 80%:20%.
4. The method for predicting the efficiency of microalgae magnetic flocculation recovery according to claim 1, characterized in that: In step 3, the two key parameters of microalgae magnetic flocculation whose absolute values of the Pearson correlation coefficient r are greater than or equal to 0.80 are redundant parameters. The redundant parameters deleted by the user are obtained, and only one of them is retained.