A method for predicting the potential for hydrothermal humification of lignocellulosic waste
By establishing the PSO-LS-SVM model, the hydrothermal humification potential of lignocellulosic waste was predicted, solving the problem of low humic acid yield, achieving efficient conversion and accurate prediction, applicable to various working conditions, and reducing experimental costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TSINGHUA UNIVERSITY
- Filing Date
- 2024-08-06
- Publication Date
- 2026-05-19
AI Technical Summary
In existing technologies, the hydrothermal humification pathway of lignocellulosic waste is complex, the yield of humic acid is low, and there is a lack of effective prediction models, resulting in low treatment efficiency and potential environmental pollution.
A particle swarm optimization least squares support vector machine model (PSO-LS-SVM) was established. By identifying the quantitative relationship between the characteristics of key intermediate products and humic acid production, a process parameter database was constructed to predict the hydrothermal humification potential of lignocellulosic waste.
It improves the efficiency of converting lignocellulosic waste into humic acid, provides a quantitative evaluation tool, reduces experimental costs and time, and is applicable to a variety of complex working conditions.
Smart Images
Figure FT_1 
Figure FT_2 
Figure FT_3
Abstract
Description
Technical Field
[0001] This invention relates to the field of waste resource utilization and modeling algorithms, and in particular to a method for predicting the hydrothermal humification potential of lignocellulosic waste. Background Technology
[0002] Lignocellulosic waste, due to its high cellulose and lignin content, degrades slowly in nature, requiring a series of physical, chemical, and biological processes to eventually be stored in the soil as humus. It has been reported that humus accounts for up to 70% of the total organic matter in soil, promoting plant growth and metabolism. Furthermore, because humus is rich in oxygen-containing functional groups, it can act as an anion affinity agent, adsorbent, and modified catalyst, with a wide range of applications. With the accelerated development of intensive agriculture and the improvement of urban greening levels, a large amount of lignocellulosic waste (such as agricultural and forestry waste, landscaping waste, and fruit and vegetable waste) is generated. Current mainstream treatment technologies, such as landfill and incineration, not only disrupt the biogeochemical cycle of resources but also pollute the environment to a certain extent. Therefore, exploring a clean and efficient method to convert lignocellulosic waste into humus has significant social, environmental, and economic implications.
[0003] Hydrothermal technology can directionally convert lignocellulosic waste into humic products within hours. However, the one-step hydrothermal method for converting lignocellulosic waste into humic acid has a low yield, less than 5%. To address this, our team proposed a two-step acid-base hydrothermal humification technology for lignocellulosic waste, successfully overcoming the current bottleneck of low hydrothermal humification efficiency and further optimizing the process. Two invention patents have been granted (ZL 2021 1 0756585.X and ZL 2022 10885763.3).
[0004] However, the hydrothermal humification pathway of lignocellulosic waste is highly complex, and humic acid yield is influenced by numerous factors. Quantitative relationships in the hydrothermal humification of lignocellulosic waste have not yet been identified, and predictive models for its yield are lacking. In the current era of advanced artificial intelligence, this invention utilizes existing data on the hydrothermal humification of lignocellulosic waste, combined with process parameters from numerous literature studies on the hydrothermal carbonization of lignocellulosic waste, to establish a particle swarm optimization least squares support vector machine (PSO-LS-SVM) algorithm for predicting the hydrothermal humification potential of lignocellulosic waste. Summary of the Invention
[0005] The purpose of this invention is to propose a method for predicting the hydrothermal humification potential of lignocellulosic waste. This method first identifies the main pathways of hydrothermal humification of lignocellulosic waste, sets quantitative evaluation indicators to classify the operating conditions of the main hydrothermal humification pathways, and establishes the quantitative relationship between the characteristics of key intermediate products and the yield of humic acid under each operating condition. By constructing a database of process parameters for the hydrothermal preparation of humic acid from lignocellulosic waste via key reaction intermediates, a particle swarm optimization least squares support vector machine (PSO-LS-SVM) algorithm is established for predicting the hydrothermal humification potential of lignocellulosic waste.
[0006] This invention provides a method for predicting the hydrothermal humification potential of lignocellulosic waste. The specific steps of the method are as follows: (1) The lignocellulosic waste is hydrothermally carbonized to prepare a reaction intermediate - hydrothermal carbon, and then the hydrothermal carbon is hydrothermally treated to prepare humic acid; (2) Using the average molecular formula unsaturation of the hydrothermal carbon as a quantitative evaluation index, the hydrothermal conversion of the lignocellulosic waste into humic acid is divided into two working conditions: insufficient hydrothermal carbonization and sufficient hydrothermal carbonization; (3) Establishing the quantitative relationship between the average molecular formula unsaturation of the hydrothermal carbon and the yield of humic acid under each working condition; (4) Constructing a process parameter database for the hydrothermal preparation of humic acid from lignocellulosic waste using the hydrothermal carbon; (5) Using the hydrothermal conditions as input and the humic acid yield as output, establishing a particle swarm optimization least squares support vector machine model (PSO-LS-SVM) algorithm for predicting the hydrothermal humification potential of the lignocellulosic waste.
[0007] Furthermore, the lignocellulosic waste refers to biomass waste rich in hemicellulose, cellulose, and lignin (referred to as the three elements), such as agricultural and forestry waste, landscaping waste, and fruit and vegetable waste. The conditions for preparing hydrothermal carbon from the lignocellulosic waste raw materials by hydrothermal carbonization are as follows: hydrothermal temperature range of 160℃~250℃, reaction time range of 1h~8h, liquid-solid ratio of 2:1~20:1, and hydrothermal pH of 0~7. Subsequently, the conditions for preparing humic acid from the hydrothermal carbon alkali are as follows: hydrothermal temperature range of 160℃~250℃, reaction time range of 1h~8h, liquid-solid ratio of 2:1~20:1, and hydrothermal pH of 12~14.
[0008] Furthermore, the average molecular formula unsaturation of the hydrothermal carbon is calculated from the carbon, hydrogen, and nitrogen content of the hydrothermal carbon: number of carbon atoms + 1 - (number of hydrogen atoms - number of nitrogen atoms) / 2.
[0009] Further, sufficient hydrothermal carbonization means that the average molecular formula unsaturation of the hydrothermal carbon is ≤4, and the difference between the average molecular formula unsaturation of the hydrothermal carbon and that of the lignocellulosic waste raw material is ≥1; insufficient hydrothermal carbonization means that the difference between the average molecular formula unsaturation of the hydrothermal carbon and that of the lignocellulosic waste raw material is ≤1.
[0010] Furthermore, under conditions of sufficient hydrothermal carbonization, the average molecular formula unsaturation value (x) of the hydrothermal carbon is positively correlated with the humic acid yield (y) of the product, which can be calculated by the following formula: y = 21.5x - 16.8; under conditions of insufficient hydrothermal carbonization, the humic acid yield of the product is related to the amount of hemicellulose, cellulose, and lignin contained in the hydrothermal carbon: y = amount of hemicellulose contained in the hydrothermal carbon × 0.5 + cellulose content × 0.6 + lignin × 61.6.
[0011] Furthermore, the process database for preparing humic acid from lignocellulosic waste via hydrothermal carbonization includes the characteristics of the lignocellulosic waste raw materials, the conditions for hydrothermal carbonization, the characteristics of the hydrothermal carbon, and the humic acid yield calculated based on two working conditions: sufficient and insufficient hydrothermal carbonization. The characteristics of the lignocellulosic waste raw materials include dosage, content of the three elements (carbon, hydrogen, nitrogen, and oxygen), and average molecular formula unsaturation. The conditions for hydrothermal carbonization include reaction temperature, reaction time, liquid-to-solid ratio, hydrothermal pH, and container volume. The characteristics of the hydrothermal carbon include carbon, hydrogen, nitrogen, and oxygen content, average molecular formula unsaturation, total yield, ash content, and organic matter yield.
[0012] Furthermore, the PSO-LS-SVM model uses hydrothermal conditions as input, specifically the amount of lignocellulosic waste raw materials used, the content of the three elements, and the hydrothermal carbonization conditions (reaction temperature, reaction time, liquid-to-solid ratio, hydrothermal pH, and container volume).
[0013] Furthermore, the PSO-LS-SVM model is built using Python code, and the specific process is as follows:
[0014] Input conditions: characteristics of the lignocellulosic waste material and the conditions for the hydrothermal carbonization;
[0015] The results column outputs: the mass of hydrothermal humic acid;
[0016] Data preprocessing: Normalize the conditional data;
[0017] Dataset partitioning: The data is divided into a training set and a test set, wherein the training set accounts for 70% to 80% of the total data and the test set accounts for 20% to 30% of the total data;
[0018] Modeling method: Use the PSO-LS-SVM model to optimize the objective function R. 2 To maximize this, the particle swarm size (30–50), inertia weight (0.3–1), individual cognitive coefficient (1–2), and social experience coefficient (1–2) are adjusted.
[0019] Parameter optimization: Set the number of PSO iterations to 100–500 to search for the optimal combination of kernel function parameters γ and σ, such that R0 2 Optimization;
[0020] Performance evaluation: using R 2 The performance of the vector machine model is evaluated using metrics such as mean squared error (MAE), mean squared deviation (MSE), and root mean squared error (RMSE).
[0021] The stability test of the vector machine model is performed by running it 50 to 100 times, randomly setting up the training set and the test set from the database, and obtaining 50 to 100 sets of model prediction results.
[0022] The present invention has the following beneficial effects:
[0023] 1. The classification of working conditions for the hydrothermal conversion of lignocellulosic waste into humic acid involved in this method can provide data support for subsequent in-depth research and engineering applications;
[0024] 2. The PSO-LS-SVM prediction model constructed by this method has a wide range of applications and can assess many types of lignocellulosic waste as well as complex hydrothermal conditions.
[0025] 3. The PSO-LS-SVM prediction model constructed by this method can save users' experimental costs and effort. Attached Figure Description
[0026] Figure 1 A flowchart for predicting the hydrothermal humification potential of lignocellulosic waste using the PSO-LS-SVM model;
[0027] Figure 2 Figure 1 shows the evaluation results of the hydrothermal humification potential of lignocellulosic waste predicted by the PSO-LS-SVM model under conditions of sufficient hydrothermal carbonization. Figure (a) shows the statistical results of MAE value; Figure (b) shows the statistical results of MSE value; Figure (c) shows the statistical results of RMSE value; and Figure (d) shows the R 2 Value statistics results;
[0028] Figure 3The figures show the evaluation results of the PSO-LS-SVM model in Example 2, predicting the hydrothermal humification potential of lignocellulosic waste under conditions of insufficient hydrothermal carbonization. Figure (a) shows the MAE value statistics; Figure (b) shows the MSE value statistics; Figure (c) shows the RMSE value statistics; and Figure (d) shows the R... 2 Value statistics results. Detailed Implementation
[0029] All features disclosed in this specification, or steps in all methods or processes disclosed herein, may be combined in any way, except for mutually exclusive features and / or steps.
[0030] Any feature disclosed in this specification (including any appended claims and abstract) may be replaced by other equivalent or similar features, unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is merely one example of a series of equivalent or similar features.
[0031] Example 1
[0032] First, the program was configured to take the characteristics of the lignocellulosic waste raw materials used in hydrothermal treatment (including raw material mass, hemicellulose, cellulose and lignin content, and carbon, hydrogen, and oxygen content) and hydrothermal carbonization conditions (including hydrothermal reaction temperature, reaction time, hydrothermal pH, liquid-to-solid ratio, and container volume) as input conditions, and output the mass of hydrothermal humic acid as the output (see Tables 1-3). Second, based on the unsaturation of the raw materials and hydrothermal carbon, the program was configured to read data from the dataset that met the criteria for sufficient hydrothermal carbonization of lignocellulosic waste (Type 1). To unify the condition variable domain, all condition data (i.e., the characteristics of the lignocellulosic waste raw materials and hydrothermal carbonization conditions) were normalized between -1 and +1. Subsequently, 20% of the data conforming to Type 1 was extracted as the test set, and the remaining 80% was used as the training set. A PSO-LS-SVM model was constructed using the SVM module of the scikit-learn library and the PSO module of the Pyswarm library (the flowchart of the model is shown in...). Figure 1 Using the above conditions as input, predict the quality of hydrothermal humic acid derived from hydrothermal carbon and compare it with the values already stored in the results column. (Using R...) 2 To optimize the objective function, based on its numerical results (R... 2(The closer the value is to 1, the higher the model fit.) The particle swarm size, inertia weight, individual cognitive coefficient, and social experience coefficient are continuously adjusted. The final optimized parameters are as follows: particle swarm size is set to 40, and inertia weight, individual cognitive coefficient, and social experience coefficient are set to 0.3, 1.5, and 1.5, respectively. The number of PSO iterations is set to 200, meaning the algorithm updates the position and velocity of each particle, searching 200 times in the particle swarm space to find the optimal combination of kernel function parameters γ and σ (γ ranges from 0.001 to 1, and σ ranges from 0.01 to 10), making R0... 2 To maximize this, this study also uses mean squared error (MAE), mean squared deviation (MSE), and root mean squared error (RMSE) to test the model accuracy. The entire program is set to run 50 times, that is, to randomly select training and test sets from the constructed dataset to obtain 50 sets of model prediction results, which are used to test the stability of the PSO-LS-SVM prediction model.
[0033] This embodiment uses box plots to statistically analyze the evaluation indices of the PSO-LS-SVM model for predicting the hydrothermal humification potential of lignocellulosic waste, namely MAE value, MSE value, RMSE value, and R. 2 Values. Under conditions of sufficient hydrothermal carbonization of lignocellulosic waste, the MAE of 50 PSO-LS-SVM prediction models ranged from 0.56 g to 1.80 g, with a median of 1.15 g; the MSE ranged from 1.06 g. 2 ~11.46g 2 Between, the median was 2.62g. 2 The RMSE ranged from 1.03g to 3.39g, with a median of 1.90g. Figure 2 The smaller these values, the better the overall performance of the model. R0 is the objective function of the model. 2 The values ranged from 0.74 to 0.96, with a median of 0.89, indicating that these models had a high degree of fit.
[0034] Example 2
[0035] The procedure was the same as in Example 1, and the dataset read was data consistent with incomplete hydrothermal carbonization of lignocellulosic waste (Type 2). For incomplete hydrothermal carbonization of lignocellulosic waste, the MAE of 50 PSO-LS-SVM prediction models ranged from 0.19g to 0.71g, with a median of 0.34g; the MSE ranged from 0.08g. 2 ~3.27g 2 Between, the median was 0.55g. 2 The RMSE ranged from 0.28g to 1.81g, with a median of 0.74g. Figure 3 R2 The range is between 0.78 and 0.97, with a median of 0.94, indicating that the model can accurately predict the hydrothermal humification potential of lignocellulosic waste even when hydrothermal carbonization is insufficient. Furthermore, within the established dataset, the program used ensures the stable performance of the PSO-LS-SVM model.
[0036]
[0037]
[0038]
Claims
1. A method for predicting the hydrothermal humification potential of lignocellulosic waste, characterized in that, include: (1) First, the lignocellulosic waste is hydrothermally carbonized to prepare hydrothermal carbon as a key reaction intermediate, and then the hydrothermal carbon is hydrothermally treated with alkali to prepare humic acid as a hydrothermal humification pathway. (2) Using the average molecular formula unsaturation of the hydrothermal carbon as a quantitative evaluation index, the working conditions of the hydrothermal conversion of the lignocellulosic waste into the humic acid are divided into insufficient hydrothermal carbonization and sufficient hydrothermal carbonization. (3) Establish the quantitative relationship between the average molecular formula unsaturation of the hydrothermal carbon and the yield of the humic acid product under each working condition; (4) Construct a database of process parameters for the preparation of humic acid from the lignocellulosic waste via hydrothermal carbon-alkali hydrothermal process; (5) Using hydrothermal conditions as input and humic acid production as output, a least squares support vector machine model optimized by particle swarm optimization is established for predicting the hydrothermal humification production of the lignocellulosic waste.
2. The method for predicting the hydrothermal humification potential of lignocellulosic waste according to claim 1, characterized in that, In step (1), the lignocellulosic waste refers to agricultural and forestry waste, kitchen waste, greening waste, biogas residue, etc., which are rich in hemicellulose, cellulose and lignin. The conditions for preparing the hydrothermal carbon by acid hydrolysis of the lignocellulosic waste raw materials are as follows: the hydrothermal temperature range is 160℃~250℃, the reaction time range is 1h~8h, the liquid-solid ratio is 2:1~20:1, and the hydrothermal pH is 0~7. The conditions for preparing the humic acid by hydrothermal preparation of the alkali from the hydrothermal carbon are as follows: the hydrothermal temperature range is 160℃~250℃, the reaction time range is 1h~8h, the liquid-solid ratio is 2:1~20:1, and the hydrothermal pH is 12~14.
3. The method for predicting the hydrothermal humification potential of lignocellulosic waste according to claim 1, characterized in that, In step (2), the average molecular formula unsaturation of the hydrothermal carbon is calculated from the carbon, hydrogen and nitrogen content of the hydrothermal carbon: number of carbon atoms + 1 - (number of hydrogen atoms - number of nitrogen atoms) / 2.
4. The method for predicting the hydrothermal humification potential of lignocellulosic waste according to claim 1, characterized in that, In step (2), sufficient hydrothermal carbonization means that the average molecular formula unsaturation of the hydrothermal carbon is ≤4 and the difference between the average molecular formula unsaturation of the hydrothermal carbon and the average molecular formula unsaturation of the lignocellulosic waste raw material is ≥1; insufficient hydrothermal carbonization means that the difference between the average molecular formula unsaturation of the hydrothermal carbon and the average molecular formula unsaturation of the lignocellulosic waste raw material is ≤1.
5. The method for predicting the hydrothermal humification potential of lignocellulosic waste according to claim 1, characterized in that, In step (3), when the hydrothermal carbonization is sufficient, the average molecular formula unsaturation value x of the hydrothermal carbon is positively correlated with the humic acid yield y of the product, which can be calculated by the following formula: y = 21.5x - 16.8; when the hydrothermal carbonization is insufficient, the humic acid yield of the product is related to the content of hemicellulose, cellulose and lignin contained in the hydrothermal carbon: y = hemicellulose content of the hydrothermal carbon × 0.5 + cellulose content × 0.6 + lignin × 61.
6.
6. The method for predicting the hydrothermal humification potential of lignocellulosic waste according to claim 1, characterized in that, In step (4), the database related to the process route for preparing humic acid from lignocellulosic waste via hydrothermal carbonization intermediates includes the characteristics of the lignocellulosic waste raw materials, the conditions of hydrothermal carbonization, the characteristics of the hydrothermal carbon, and the humic acid yield calculated based on two working conditions: sufficient and insufficient hydrothermal carbonization. The characteristics of the lignocellulosic waste raw materials include the dosage, content of the three elements (carbon, hydrogen, nitrogen, and oxygen), and average molecular formula unsaturation. The conditions of hydrothermal carbonization include the reaction temperature, reaction time, liquid-to-solid ratio, hydrothermal pH, and container volume. The characteristics of the hydrothermal carbon include the content of carbon, hydrogen, nitrogen, and oxygen, average molecular formula unsaturation, total yield, ash content, and organic matter yield.
7. The method for predicting the hydrothermal humification potential of lignocellulosic waste according to claim 6, characterized in that, In step (5), the vector machine model takes hydrothermal conditions as input, specifically the amount of lignocellulosic waste raw materials used, the content of the three elements, and the hydrothermal carbonization conditions.
8. The method for predicting the hydrothermal humification potential of lignocellulosic waste according to claim 6, characterized in that, In step (5), the vector machine model is built using Python code, and the specific process is as follows: Input conditions: characteristics of the lignocellulosic waste and the conditions for the hydrothermal carbonization; The results column outputs: the mass of hydrothermal humic acid; Data preprocessing: Normalize the conditional data; Dataset partitioning: The data is divided into a training set and a test set, wherein the training set accounts for 70% to 80% of the total data and the test set accounts for 20% to 30% of the total data; Modeling method: Using the vector machine model, optimize the objective function R. 2 To maximize this, the particle swarm size (30–50), inertia weight (0.3–1), individual cognitive coefficient (1–2), and social experience coefficient (1–2) are adjusted. Parameter optimization: Set the number of iterations for the vector machine model to 100–500, and search for the optimal combination of kernel function parameters γ and σ to achieve R0. 2 Optimization; Performance evaluation: using R 2 The performance of the vector machine model is evaluated using the indices of mean square error, mean square deviation, and root mean square error. The stability test of the vector machine model is performed by running it 50 to 100 times, randomly setting up the training set and the test set from the database, and obtaining 50 to 100 sets of model prediction results.