Solidified sludge additive optimization method for planting soil

Abstract

The invention provides an optimization method of a solidified sludge additive for planting soil, and belongs to the technical field of solidified sludge, and the optimization method comprises the following steps: firstly, obtaining different formulas of the solidified sludge additive and corresponding experimental effect data, including indexes of solidification strength, water retention capacity, nutrient release, heavy metal stabilization and the like; on the basis of the data, an effect prediction equation set is established, and various performance indexes and the synergistic effect of components of the performance indexes are covered. And training the equation set by adopting a machine learning algorithm to form a preliminary formula effect relation model. And based on the model, traversing to generate multiple groups of possible formula data, and performing clustering analysis to obtain a representative clustering center formula. And selecting the clustering center formulas to carry out small-scale planting experiments to obtain actual effect data. And by using a genetic algorithm and taking actual effect data as a fitness function, optimizing to obtain a final optimal additive formula. And finally, carrying out a large-scale planting verification experiment on the formula to determine an optimal solidified sludge additive formula.

Description

Technical Field

[0001] The invention belongs to the technical field of solidified sludge, and in particular relates to a method for optimizing a solidified sludge additive for planting soil. Background Art

[0002] The sludge from marine aquaculture contains a large amount of organic matter and nutrients, which can be used as an additive for agricultural soil improvement. However, directly applying untreated sludge to farmland will cause problems such as soil environmental pollution and accumulation of heavy metals in crops. Therefore, it is necessary to develop a reasonable formula and optimization method for solidified sludge additives to ensure its safety and effectiveness in planting soil.

[0003] At present, the research on solidified sludge additives mainly focuses on the following aspects: 1) additive formula design, that is, determining the appropriate combination ratio of inorganic minerals, organic matter and active ingredients; 2) optimization of additive preparation process, such as hot pressing, alkali activation, etc.; 3) evaluation of the application effect of additives in planting soil, including indicators such as solidification strength, water retention, nutrient release, and heavy metal stabilization. Most of the existing studies use the "experience-trial and error" method for formula design and process optimization, which is time-consuming and lacks systematicity and predictability. In addition, most of the existing studies are limited to small-scale laboratory experiments and fail to fully consider the complex factors of engineering applications.

[0004] In view of the above problems, it is urgent to establish a scientific and reasonable optimization method for solidified sludge additives. Summary of the invention

[0005] In view of this, the present invention provides a method for optimizing solidified sludge additives for planting soil, which can solve the technical problem that most existing studies use an "experience-trial and error" approach to carry out formula design and process optimization, resulting in a long time consumption.

[0006] The present invention is achieved in that:

[0007] The present invention provides a method for optimizing a solidified sludge additive for planting soil, which comprises the following steps:

[0008] S10, obtaining formula data of different formulas of the solidified sludge additive and effect data obtained from specific experiments corresponding to each formula, wherein the effect data includes solidification strength, water retention performance, nutrient release rate, heavy metal stabilization degree and environmental compatibility index;

[0009] S20, establishing an effect prediction equation group of the solidified sludge additive, wherein the effect prediction equation group includes a solidification reaction equation, a water retention performance equation, a nutrient release equation, a heavy metal stabilization equation, an environmental compatibility equation, and a component synergy equation;

[0010] S30, using a machine learning algorithm to train the effect prediction equation group using the formulation data of the different formulations to obtain a preliminary formulation-effect relationship model;

[0011] S40, generating multiple sets of formula data by traversal according to the preset formula component range and ratio step, and calculating the multiple sets of formula data by using the preliminary formula effect relationship model to obtain theoretical effect data;

[0012] S50, performing cluster analysis on the multiple groups of recipe data based on the theoretical effect data to obtain multiple recipe cluster centers;

[0013] S60, selecting the formula corresponding to each of the formula cluster centers to form an experimental formula set, conducting a small-scale planting experiment, and collecting actual effect data;

[0014] S70, selecting multiple groups of recipes with the best actual effects from the experimental recipe set as the initial population, normalizing the actual effect data and setting it as a fitness function, establishing a genetic algorithm, and obtaining the optimal recipe through multi-generation iterative optimization;

[0015] S80, conducting a large-scale planting verification experiment on the optimal formula to determine the final solidified sludge additive formula.

[0016] On the basis of the above technical solution, the optimization method of solidified sludge additive for planting soil of the present invention can also be improved as follows:

[0017] The specific experiment corresponding to each of the formulas is specifically a small-scale planting experiment, including the following steps:

[0018] Step 1, preparing test soil and solidified sludge additives;

[0019] Step 2, mixing the solidified sludge additive with soil according to different formulas;

[0020] Step 3: Plant indicator plants in a controlled environment;

[0021] Step 4: Regularly monitor and record plant growth;

[0022] Step 5: measuring plant biomass and soil physical and chemical properties after harvest;

[0023] Step 6: Analyze the curing strength, water retention performance, nutrient release rate, heavy metal stabilization degree and environmental compatibility index.

[0024] The solidification reaction equation is used to predict the solidification strength of the solidified sludge additive after mixing with the soil, the input is the additive component content and the reaction conditions, and the output is the solidification strength value.

[0025] The water retention performance equation is used to predict the water retention capacity of the soil-additive mixture, the input is the additive component content and the basic properties of the soil, and the output is the water retention coefficient.

[0026] The nutrient release equation is used to predict the release rate of nutrients in additives, the input is the content of additive components and environmental factors, and the output is the nutrient release rate.

[0027] The heavy metal stabilization equation is used to predict the stabilization effect of additives on heavy metals in soil, the input is the additive component content and the initial concentration of heavy metals, and the output is the heavy metal stabilization rate.

[0028] The environmental compatibility equation is used to evaluate the impact of additives on soil ecosystems, the input is the additive component content and soil biological indicators, and the output is the environmental compatibility index.

[0029] The component synergy equation is used to describe the interaction effect between the components of the additive, the input is the content of each component, and the output is the synergy coefficient.

[0030] The machine learning algorithm adopts a random forest regression algorithm, and the step of using the machine learning algorithm to train the effect prediction equation group using the formulation data of the different formulations specifically includes:

[0031] Data preprocessing: standardize and normalize the recipe data and effect data;

[0032] Feature selection: Use principal component analysis to select the main features;

[0033] Model building: Build a random forest regression model and set appropriate hyperparameters;

[0034] Model training: Use training set data to train the model;

[0035] Model validation: Use cross-validation methods to evaluate model performance;

[0036] Model optimization: Optimize the model by adjusting hyperparameters and feature engineering;

[0037] Model testing: Use the test set data to evaluate the prediction performance of the final model.

[0038] Furthermore, the indicator plant is wheat or corn.

[0039] Specifically, the step S10 specifically includes: first, through a small-scale laboratory planting test, the performance data of solidified sludge additives with different formulations in the planting soil are obtained. These performance data include indicators such as solidification strength, water retention performance, nutrient release rate, heavy metal stabilization degree and environmental compatibility. Each formula needs to be repeated at least 3 times to ensure the reliability of the data. Both the formula data and the experimental test data need to be recorded in detail to establish a standardized database. Among them, the acquisition of the solidification strength index needs to refer to the relevant standards, and the compressive strength of the mixture is measured by a pressure testing machine; the water retention performance is reflected by measuring the water retention coefficient of the mixture; the nutrient release rate is obtained by the regular leaching-determination method; the heavy metal stabilization rate is determined by the continuous extraction method; the environmental compatibility index needs to comprehensively consider indicators such as soil microorganisms, enzyme activity, and root activity. After sufficient experimental accumulation, the specific values ​​of various performance indicators under different formulations were obtained.

[0040] Specifically, step S20 includes: first, establishing corresponding prediction equations for the five key performance indicators of curing strength, water retention, nutrient release, heavy metal stabilization and environmental compatibility. These equations need to consider multiple influencing factors such as the content of the components of the additive, soil properties, environmental conditions, etc. At the same time, it is also necessary to establish an equation to describe the interaction between the various components of the additive. By establishing such a complete set of effect prediction equations, it is possible to predict various performance indicators of the additive in the planting soil based on the formula and environmental conditions of the additive. This provides a strong mathematical model support for the subsequent optimization design.

[0041] Specifically, the step S30 specifically includes: first, standardizing and normalizing the recipe data and effect data to eliminate the influence of dimensional differences. Then, the principal component analysis (PCA) method is used to perform feature selection on the influencing factors of each prediction indicator to determine the main independent variables. Next, a prediction model based on the random forest regression algorithm is constructed. The random forest regression model is trained using the experimental data set obtained in the aforementioned step S10. By continuously optimizing hyperparameters and feature engineering, a set of optimal model parameters is finally determined. This provides a reliable mathematical model support for subsequent recipe optimization.

[0042] Specifically, the step S40 specifically includes: first, according to the previous experimental accumulation and literature data, determine the reasonable use range of each additive component. Then, set an appropriate change step within these ranges to generate a large amount of data on different formula combinations. Next, use the random forest regression prediction model established in the aforementioned step S30 to calculate these formula data one by one. Through this series of calculations, a large amount of theoretical effect data is obtained, covering a wide range of formula combinations. These rich theoretical data provide valuable information reference for subsequent optimization design.

[0043] Specifically, the step S50 specifically includes: first, standardizing a large amount of theoretical effect data generated in step S40. Then, selecting a suitable clustering algorithm for classification, and the K-Means algorithm can be used here. Through K-Means clustering, a large amount of theoretical formula data is divided into several clusters. Each cluster has a cluster center, which represents the typical characteristics of this type of formula. The formulas corresponding to these cluster centers become the objects that need to be focused on and experimentally verified.

[0044] Specifically, the step S60 specifically includes: first, according to the cluster analysis results of step S50, the formula composition corresponding to each cluster center is selected as the experimental formula set. Then, the soil and solidified sludge additives for the experiment are prepared. Next, the additives are mixed with the soil according to different formulas. Under well-controlled environmental conditions, the mixed planting matrix is ​​placed in a small pot and indicator plants are planted. The growth of the plants is monitored and recorded regularly. Finally, harvesting and analysis tests are carried out, including plant biomass, soil physical and chemical properties, nutrient content, heavy metal content and soil biological indicators. Through this series of small-scale planting tests, various performance data of each experimental formula under actual planting conditions will be obtained.

[0045] Specifically, the step S70 specifically includes: first, from the small-scale planting experiment of step S60, select several groups of formulas with the best performance as the initial population. Encode the formula data of these formulas into genotypes as the initial population of the genetic algorithm. Next, the actual effect data obtained in step S60 is normalized as the fitness function of the genetic algorithm. Then, the operation process of the genetic algorithm is established, including the steps of selection, crossover and mutation. Through multiple generations of iterative optimization, the genetic algorithm will continuously improve the fitness level of the population. When the preset number of iterations or convergence conditions are reached, the algorithm outputs the individual with the highest fitness in the current population, which is the optimal additive formula.

[0046] Specifically, the step S80 specifically includes: first, according to the optimal formula obtained in step S70, solidified sludge additives are prepared in large quantities. Then, a suitable test site is selected for large-scale planting. Additives are applied and crops are planted in the test area. During the entire growth period, it is necessary to regularly observe and record the growth of crops and changes in the physical and chemical properties of the soil. After the crops mature, they are harvested and fully analyzed and tested. Only when the large-scale planting verification results show that the optimal formula can continuously and stably meet various performance requirements, can the additive formula be finally determined as the best solution.

[0047] The following is a detailed description of the formula expression of each equation in the effect prediction equation group:

[0048] The curing reaction equation is specifically expressed as follows:

[0049]

[0050] Where, S is the curing strength (MPa); C 1 ,C 2 ,C 3 are the mass fractions of the main additive components (%); T is the reaction temperature (℃); H is the relative humidity (%); t is the curing time (d); k 1 ,k 2 ,k 3 ,k 4 ,k 5 ,k 6 is the unknown coefficient; α ij is the interaction coefficient between components; n is the number of additive components; ε S is the error term.

[0051] Parameter acquisition method:

[0052] C 1 ,C 2 ,C 3 Determined according to formula design; T, H, t are measured by environmental monitoring instruments; the curing strength S is determined by a pressure testing machine. The specific steps are: 1. Prepare standard size test blocks; 2. Curing to the specified age; 3. Use a pressure testing machine to determine the compressive strength.

[0053] The water retention performance equation is specifically expressed as follows:

[0054]

[0055] Where, W is the water retention coefficient (g / g); C 1 ,C 2 ,C 3 Same as above; ρ is soil bulk density (g / cm 3 ), φ is soil porosity (%), OM is organic matter content (%), a 1 ,a 2 ,a 3 ,b 1 ,b 2 ,b 3 is the unknown coefficient; β ij is the interaction coefficient between components; ε W is the error term.

[0056] Parameter acquisition method:

[0057] ρ is determined by the ring knife method: 1. Collect undisturbed soil samples; 2. Dry and weigh; 3. Calculate bulk density.

[0058] φ is calculated by the formula: where ρ s is the soil particle density.

[0059] OM is determined by the potassium dichromate volumetric method: 1. Soil sample pretreatment; 2. Potassium dichromate oxidation; 3. Titration to determine the organic carbon content; 4. Conversion to organic matter content.

[0060] The nutrient release equation is specifically expressed as follows:

[0061]

[0062] In the formula, R i is the release rate of the ith nutrient (mg / kg / d); C 1 ,C 2 ,C 3 Same as above; pH is soil acidity; T is temperature (℃); M is soil moisture content (%); f i (t) is the time function, expressed as f i (t) = p i ln(t+1)+q i ;m 1 ,m 2 ,m 3 ,n 1 ,n 2 ,n 3 ,p i ,q i is the unknown coefficient; jk is the interaction coefficient between components; is the error term.

[0063] Parameter acquisition method:

[0064] pH is determined by potentiometric method: 1. Prepare soil suspension; 2. Calibrate pH meter; 3. Measure pH value.

[0065] M is determined by the drying method: 1. Weigh the fresh soil sample; 2. Dry at 105℃ to constant weight; 3. Calculate the moisture content.

[0066] Nutrient release rate R i Obtained through the extraction-determination method: 1. Collect soil samples regularly; 2. Extract soluble nutrients; 3. Use analytical instruments (such as atomic absorption spectrophotometer) to measure nutrient concentration; 4. Calculate the release rate.

[0067] The heavy metal stabilization equation is specifically expressed as follows:

[0068]

[0069] In the formula, S jis the stabilization rate of the jth heavy metal (%); C 1 ,C 2 ,C 3 , pH, OM are the same as above; CEC is the cation exchange capacity (cmol / kg); t is the time (d); x 1 ,x 2 ,x 3 ,y 1 ,u 2 ,u 3 ,z 1 ,λ j is the unknown coefficient; δ kl is the interaction coefficient between components; is the error term.

[0070] Parameter acquisition method:

[0071] CEC is determined by the ammonium acetate exchange-distillation method: 1. Soil sample pretreatment; 2. Ammonium acetate extraction; 3. Distillation; 4. Titration to calculate the CEC value.

[0072] Heavy metal stabilization rate S j Determination by continuous extraction method: 1. Collect soil samples; 2. Perform Tessier five-step continuous extraction; 3. Determine the content of heavy metals in each form by atomic absorption spectrophotometer; 4. Calculate the stabilization rate.

[0073] The environmental compatibility equation is specifically expressed as follows:

[0074]

[0075] Where, E is the environmental compatibility index (dimensionless); C 1 ,C 2 ,C 3 Same as above; B is soil microbial biomass carbon (mg / kg); E is soil enzyme activity (μmol / g / h); R is plant root activity (μg / g / h); u 1 ,u 2 ,u 3 ,v 1 ,v 2 ,v 3 ,w 1 ,ω, is the unknown coefficient; θ mo is the interaction coefficient between components; ε E is the error term.

[0076] Parameter acquisition method:

[0077] R through chloroform fumigation -K 2 SO 4Extraction method: 1. Soil sample pretreatment; 2. Chloroform fumigation; 3. K 2 SO 4 Extraction; 4. TOC analyzer to determine the organic carbon content; 5. Calculate the microbial biomass carbon.

[0078] E is determined by the dehydrogenase activity method: 1. The soil sample is mixed with TTC (2,3,5-triphenyltetrazolium chloride) and cultured; 2. TTF (triphenylmethyl sulfone) is extracted with acetone; 3. The absorbance is measured by a spectrophotometer; 4. The enzyme activity is calculated.

[0079] R was determined by TTC method: 1. Take fresh roots; 2. Soak in TTC solution; 3. Extract TTF with ethanol; 4. Measure absorbance with spectrophotometer; 5. Calculate root activity.

[0080] The component synergy equation is specifically expressed as follows:

[0081]

[0082] Where, I is the synergy coefficient (dimensionless); C i ,C j ,C k is the mass fraction of each component of the additive (%); η ij is the second-order interaction coefficient; ζ ijk is the third-order interaction coefficient; ε I is the error term.

[0083] Parameter acquisition method:

[0084] The synergistic effect coefficient I was obtained through multi-factor orthogonal experiment: 1. Designing orthogonal experiment scheme; 2. Carrying out formulation experiment; 3. Determining various effect indicators; 4. Calculating the significance of interaction through variance analysis; 5. Obtaining the interaction coefficient through regression analysis.

[0085] Compared with the prior art, the method for optimizing the solidified sludge additive for planting soil provided by the present invention has the following beneficial effects:

[0086] 1. A reliable experimental data foundation has been established. Through small-scale planting tests, the performance data of different formula additives in key indicators such as curing strength, water retention, nutrient release, heavy metal stabilization, and environmental compatibility were comprehensively measured. These rich experimental data provide a solid foundation for subsequent model establishment and optimization.

[0087] 2. A prediction model based on physical and chemical principles was constructed. For each performance index, the present invention established corresponding mathematical prediction equations. These equations take into account multiple influencing factors such as additive composition, soil properties, environmental conditions, etc., and can be used to accurately predict the performance of additives in planting soil.

[0088] 3. The prediction model is optimized by using machine learning technology. The present invention uses the random forest regression algorithm to train and optimize the above prediction equation, which greatly improves the prediction accuracy of the model. This provides a reliable mathematical support for subsequent formula optimization.

[0089] 4. The best formula is obtained through intelligent optimization algorithm. The present invention uses genetic algorithm to perform intelligent optimization search on a large number of theoretical formulas, and achieves the best in various performance indicators. This systematic optimization method overcomes the limitations of the traditional "experience-trial and error" method.

[0090] 5. Strict engineering verification has been carried out. The present invention has fully verified the optimal formula in large-scale planting experiments to ensure its reliability and stability in practical applications. This provides a strong guarantee for the application of sludge additives in engineering.

[0091] In general, the present invention constructs a systematic research method that integrates experimental testing, mathematical modeling, machine learning, and intelligent optimization, which solves the technical problem that most existing studies use the "experience-trial and error" method to carry out formula design and process optimization, resulting in a long time consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] Figure 1 A flow chart of the method provided by the present invention;

[0093] Figure 2 To optimize the release kinetics of nutrients (nitrogen, phosphorus, potassium) in the formula;

[0094] Figure 3 This is a graph showing the changes in the contents of three heavy metals, copper, zinc and cadmium, in the soil before and after treatment;

[0095] Figure 4 This is a comparison chart of the differences in many aspects between the treatment area and the control area. DETAILED DESCRIPTION

[0096] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0097] like Figure 1 As shown, it is a flow chart of a method for optimizing a solidified sludge additive for planting soil provided by the present invention, and the method comprises the following steps:

[0098] S10, obtaining formula data of different formulas of solidified sludge additives and effect data obtained from specific experiments corresponding to each formula, the effect data including solidification strength, water retention performance, nutrient release rate, heavy metal stabilization degree and environmental compatibility index;

[0099] S20, establishing an effect prediction equation group of solidified sludge additives, the effect prediction equation group including a solidification reaction equation, a water retention performance equation, a nutrient release equation, a heavy metal stabilization equation, an environmental compatibility equation, and a component synergistic effect equation;

[0100] S30, using a machine learning algorithm, using formulation data of different formulations to train the effect prediction equation group, and obtaining a preliminary formulation effect relationship model;

[0101] S40, generating multiple sets of formula data by traversal according to the preset formula component range and ratio step, and calculating the multiple sets of formula data by using the preliminary formula effect relationship model to obtain theoretical effect data;

[0102] S50, based on the theoretical effect data, performing cluster analysis on multiple groups of recipe data to obtain multiple recipe cluster centers;

[0103] S60, selecting the formula corresponding to each formula cluster center to form an experimental formula set, conducting a small-scale planting experiment, and collecting actual effect data;

[0104] S70, selecting multiple groups of recipes with the best actual effects from the experimental recipe set as the initial population, normalizing the actual effect data and setting it as the fitness function, establishing a genetic algorithm, and obtaining the optimal recipe through multi-generation iterative optimization;

[0105] S80. Conduct large-scale planting verification experiments on the optimal formula to determine the final formula of solidified sludge additives.

[0106] The specific implementation methods of the above steps are described in detail below:

[0107] Step S10, obtaining the formula data of different formulas of the solidified sludge additive and the effect data obtained from the specific experiments corresponding to each formula. The purpose of this step is to establish a reliable experimental data foundation to provide a basis for subsequent model establishment and optimization. The specific implementation method is as follows:

[0108] First, through a small-scale laboratory planting test, the performance data of solidified sludge additives with different formulations in planting soil were obtained. These performance data include indicators such as solidification strength, water retention performance, nutrient release rate, heavy metal stabilization degree and environmental compatibility. Each formulation needs to be repeated at least 3 times to ensure the reliability of the data. The formulation data and experimental test data need to be recorded in detail to establish a standardized database. Among them, the acquisition of the solidification strength index needs to refer to the relevant standards, and the compressive strength of the mixture is measured by a pressure testing machine; the water retention performance is reflected by measuring the water retention coefficient of the mixture; the nutrient release rate is obtained by the regular leaching-determination method; the heavy metal stabilization rate is determined by the continuous extraction method; the environmental compatibility index needs to comprehensively consider indicators such as soil microorganisms, enzyme activity, and root activity. These test methods need to be carried out strictly in accordance with the established standard operating procedures. After sufficient experimental accumulation, the specific values ​​of various performance indicators under different formulations were obtained.

[0109] Through this step of experiment and data collection, a rich and reliable experimental database was established, which provided a solid foundation for subsequent model building and optimization. These data included the formulation information of additives and the corresponding experimental test results, providing necessary support for subsequent optimization design.

[0110] Step S20, establish a set of equations for predicting the effect of solidified sludge additives. The purpose of this step is to establish a set of mathematical models based on physical and chemical principles to predict the various properties of additives in planting soil. The specific implementation is as follows:

[0111] First, corresponding prediction equations were established for the five key performance indicators of curing strength, water retention, nutrient release, heavy metal stabilization, and environmental compatibility. These equations need to consider multiple influencing factors such as the content of additive components, soil properties, and environmental conditions.

[0112] For the curing strength prediction equation, the following linear regression model can be used:

[0113]

[0114] Where, S is the curing strength (MPa); C 1 ,C 2 ,C 3 are the mass fractions of the main additive components (%); T is the reaction temperature (℃); H is the relative humidity (%); t is the curing time (d); k 1 ,k 2 ,k 3 ,k 4 ,k 5 ,k 6 is the unknown coefficient; α ijis the interaction coefficient between components; n is the number of additive components; ε S is the error term. This equation can be used to predict the curing strength after the additive is mixed with the soil.

[0115] For the water retention performance prediction equation, the following multivariate linear regression model can be used:

[0116]

[0117] Where W is the water retention coefficient (g / g); ρ is the soil bulk density (g / cm 3 ), φ is soil porosity (%), OM is organic matter content (%), a 1 ,a 2 ,a 3 ,b 1 ,b 2 ,b 3 is the unknown coefficient; β ij is the interaction coefficient between components; ε W is the error term. This equation can be used to predict the effect of additives on soil water retention performance.

[0118] For the nutrient release prediction equation, the following nonlinear regression model can be used:

[0119]

[0120] In the formula, R i is the release rate of the ith nutrient (mg / kg / d); pH is the soil acidity; T is the temperature (℃); M is the soil moisture content (%); f i (t) is the time function, expressed as f i (t) = p i ln(t+1)+q i ;m 1 ,m 2 ,m 3 ,n 1 ,n 2 ,n 3 ,p i ,q i is the unknown coefficient; jk is the interaction coefficient between components; is the error term. This equation can be used to predict the effect of additives on soil nutrient release.

[0121] For the heavy metal stabilization prediction equation, the following exponential decay model can be used:

[0122]

[0123] In the formula, Sj is the stabilization rate of the jth heavy metal (%); CEC is the cation exchange capacity (cmol / kg); t is the time (d); x 1 ,x 2 ,x 3 ,y 1 ,y 2 ,y 3 ,z 1 ,λ j is the unknown coefficient; δ kl is the interaction coefficient between components; is the error term. This equation can be used to predict the effect of additives on soil heavy metal stabilization.

[0124] For the environmental compatibility prediction equation, the following semi-empirical model can be used:

[0125]

[0126] Where, E is the environmental compatibility index (dimensionless); B is the soil microbial biomass carbon (mg / kg); E is the soil enzyme activity (μmol / g / h); R is the plant root activity (μg / g / h); u 1 ,u 2 ,u 3 ,v 1 ,v 2 ,v 3 ,w 1 ,ω, is the unknown coefficient; θ mo is the interaction coefficient between components; ε E is the error term. This equation can be used to predict the impact of additives on soil ecosystems.

[0127] Finally, an equation describing the interaction between the additive components needs to be established:

[0128]

[0129] Where, I is the synergy coefficient (dimensionless); η ij is the second-order interaction coefficient; ζ ijk is the third-order interaction coefficient; ε I is the error term. This equation can be used to describe the interaction between the various components of the additive.

[0130] By establishing such a complete set of effect prediction equations, it is possible to predict various performance indicators of additives in planting soil based on the formulation and environmental conditions of the additives, which provides a strong mathematical model support for subsequent optimization design.

[0131] Step S30, using a machine learning algorithm, using the formula data of the different formulas to train the effect prediction equation group. The purpose of this step is to use machine learning technology to optimize the parameters of the above prediction equations and improve their prediction accuracy. The specific implementation is as follows:

[0132] First, the recipe data and effect data are standardized and normalized to eliminate the impact of dimensional differences. Then, the principal component analysis (PCA) method is used to select the influencing factors of each prediction indicator and determine the main independent variables. Next, a prediction model based on the random forest regression algorithm is constructed. Random forest regression is a non-parametric ensemble learning method that can effectively improve the stability and accuracy of predictions by generating multiple decision tree models and averaging their results. During the model construction process, some hyperparameters need to be adjusted, such as the maximum depth of the decision tree, the minimum number of samples, etc., to achieve the best prediction performance.

[0133] The random forest regression model is trained using the experimental data set obtained in the above step S10. Specifically, the data set is divided into a training set and a validation set. The model is first fitted using the training set data, and then the prediction effect of the model is evaluated using the validation set data. By continuously optimizing hyperparameters and feature engineering, a set of optimal model parameters is finally determined. Finally, the model is finally tested and verified using the entire data set to ensure that its prediction accuracy meets the requirements.

[0134] Through this step of machine learning modeling, we obtained optimized prediction equation group parameters, which greatly improved the prediction accuracy of various performance indicators. This provides a reliable mathematical model support for subsequent formula optimization.

[0135] Step S40, based on the preset formula component range and ratio step length, multiple sets of formula data are generated by traversal, and the preliminary formula effect relationship model is used to calculate the multiple sets of formula data to obtain theoretical effect data. The purpose of this step is to obtain a large amount of theoretical performance data through exhaustive formula traversal combined with the aforementioned prediction model, so as to provide sufficient samples for subsequent optimization. The specific implementation is as follows:

[0136] First, based on previous experimental accumulation and literature data, determine the reasonable use range of each additive component. For example, the main component C 1 The usage range is set to 10% to 30%, component C 2 5% to 15%, component C 3 Then, set an appropriate change step within these ranges, such as changing every 2%. In this way, you can generate a large number of data for different recipe combinations, such as (10%, 5%, 1%), (12%, 7%, 2%), (14%, 9%, 3%), etc.

[0137] Next, the random forest regression prediction model established in the above step S30 is used to calculate these formula data one by one. The model will automatically predict the theoretical performance indicators of each formula in terms of curing strength, water retention, nutrient release, heavy metal stabilization, environmental compatibility, etc. Through this series of calculations, a large amount of theoretical effect data is obtained, covering a wide range of formula combinations.

[0138] These rich theoretical data provide valuable information reference for subsequent optimization design. By analyzing and mining these data, we can understand the influence of different formulas on various performance indicators and lay the foundation for determining the optimal formula.

[0139] Step S50, based on the theoretical effect data, cluster analysis is performed on the multiple groups of formula data to obtain multiple formula cluster centers. The purpose of this step is to classify a large amount of theoretical formula data through an unsupervised learning method to find key representative formulas. The specific implementation method is as follows:

[0140] First, a large amount of theoretical effect data generated in step S40 is standardized to eliminate the influence of dimensional differences. Then, a suitable clustering algorithm is selected for classification. Here, the K-Means algorithm can be used, which is a distance-based partitioning clustering method. By iteratively optimizing the position of the cluster center, each recipe data is finally divided into different clusters.

[0141] When implementing the K-Means algorithm, it is necessary to determine the initial position of the cluster center and the value of the cluster number K. The initial center can be initialized randomly or using several pre-selected representative recipes as seeds. The selection of the cluster number K needs to consider the evaluation indicators of the clustering effect, such as the average silhouette coefficient, Calinski-Harabasz index, etc. Generally speaking, the value of K should make the clusters compact and the clusters far apart. After many attempts, a suitable K value was finally determined.

[0142] Through K-Means clustering, a large amount of theoretical formula data is divided into several clusters. Each cluster has a cluster center, which represents the typical characteristics of this type of formula. The formulas corresponding to these cluster centers become the objects that need to be focused on and experimentally verified. This lays the foundation for subsequent small-scale planting experiments.

[0143] Step S60, select the formula corresponding to each of the formula cluster centers to form an experimental formula set, conduct a small-scale planting experiment, and collect actual effect data. The purpose of this step is to verify the reliability of the aforementioned theoretical prediction through actual planting experiments and obtain more realistic performance data. The specific implementation method is as follows:

[0144] First, according to the cluster analysis results of step S50, the recipe compositions corresponding to the cluster centers are selected as the experimental recipe set. These recipes should cover various typical situations in theoretical prediction.

[0145] Then, prepare the soil and solidified sludge additives for the experiment. The soil can be common planting soil or agricultural soil. The additives are prepared according to the selected experimental formula.

[0146] Next, the additives are mixed with the soil according to different formulas. A small mixer can be used to mix thoroughly to ensure that the additives are fully in contact with the soil.

[0147] Under well-controlled environmental conditions (such as temperature 25 ± 2 ° C, relative humidity 60 ± 5%), put the mixed planting medium into small pots and plant indicator plants (such as wheat or corn). Regularly monitor and record the growth of plants, such as seedling height, leaf color, biomass, etc.

[0148] When the plants grow to the appropriate stage, they are harvested and analyzed and tested. The tests include:

[0149] 1. Plant biomass: Plants were harvested and fresh and dry weights were determined.

[0150] 2. Soil physical and chemical properties: determination of solidification strength, water retention, pH value, organic matter content, etc.

[0151] 3. Nutrient content: Determine the concentration of various nutrient elements in the plant.

[0152] 4. Heavy metal content: Determine the concentration and distribution of various heavy metals in the soil.

[0153] 5. Soil biological indicators: determination of microbial biomass, enzyme activity, root activity, etc.

[0154] Through this series of small-scale planting tests, we will obtain various performance data of each experimental formula under actual planting conditions. These measured data will be compared and verified with the aforementioned theoretical prediction results, providing a reliable basis for the next step of optimization design.

[0155] Step S70, select multiple groups of formulas with the best actual effects from the experimental formula set as the initial population, normalize the actual effect data and set it as the fitness function, establish a genetic algorithm, and obtain the optimal formula through multiple generations of iterative optimization. The purpose of this step is to use the genetic algorithm for intelligent optimization to find the best additive formula. The specific implementation is as follows:

[0156] First, from the small-scale planting experiment in step S60, several groups of recipes with the best performance are selected as the initial population. These recipes should have good performance in various indicators. The recipe data of these recipes are encoded into genotypes as the initial population of the genetic algorithm.

[0157] Next, the actual effect data obtained in step S60 is normalized and used as the fitness function of the genetic algorithm. Here, various performance indicators, such as curing strength, water retention, nutrient release, heavy metal stabilization, environmental compatibility, etc., need to be comprehensively considered, given appropriate weights, and an overall fitness evaluation index is constructed.

[0158] Then, the operation process of the genetic algorithm is established. First, the selection operation is performed to select individuals with higher fitness from the current population as parents; second, the crossover operation is performed to randomly combine the genotypes of the parents to generate new individuals; finally, the mutation operation is performed to randomly change the genotype of the individual to increase the diversity of the population.

[0159] Through multiple generations of iterative optimization, the genetic algorithm will continuously improve the fitness level of the population. When the preset number of iterations or convergence conditions are reached, the algorithm outputs the individual with the highest fitness in the current population, which is the optimal additive formula.

[0160] This optimization process requires a lot of computing resources and time. In the specific implementation, parallel computing can be used to speed up the optimization. In addition, some improvement measures can be introduced, such as adaptive mutation probability, elite retention strategy, etc., to further improve the convergence and stability of the algorithm.

[0161] Through this step of genetic algorithm optimization, we obtained the additive formula that achieved the best performance in all indicators, which laid the foundation for the subsequent large-scale planting verification experiment.

[0162] Step S80, conduct a large-scale planting verification experiment on the optimal formula to determine the final solidified sludge additive formula. The purpose of this step is to fully verify the optimal formula obtained by the above optimization under larger-scale planting conditions to ensure its feasibility and stability in practical applications. The specific implementation method is as follows:

[0163] First, according to the optimal formula obtained in step S70, solidified sludge additives are prepared in large quantities. The composition and production process of the additives need to be strictly controlled to ensure stable quality.

[0164] Then, select a suitable test site for large-scale planting. The test site should have representative soil conditions and be large enough to ensure the reliability of the experiment. Different test plots should be divided according to the needs of the experiment.

[0165] Next, the additives are applied and crops are planted in the test area. The optimally formulated additives are fully mixed with the soil and the target crops are planted. The planting and management of the crops must be carried out strictly in accordance with standardized agricultural technical regulations.

[0166] During the entire growth period, it is necessary to regularly observe and record the growth of crops, such as seedling condition, plant height, leaf color, biomass, etc. At the same time, it is also necessary to monitor changes in the physical and chemical properties of the soil, such as nutrient content, heavy metal status, enzyme activity, etc.

[0167] When the crops are mature, they will be harvested and fully analyzed and tested. Not only will the yield and quality of the crops be measured, but also the various indicators of the soil will be evaluated. These test data will be compared and analyzed with the results of the previous small-scale tests.

[0168] Only when the results of large-scale planting verification show that the optimal formula can continuously and stably meet various performance requirements can the additive formula be finally determined as the best solution.

[0169] Through this series of rigorous experimental verification processes, the reliability and effectiveness of the developed solidified sludge additive formula in practical applications are ensured, which provides strong technical support for subsequent engineering applications.

[0170] Specifically, the principle of the present invention is:

[0171] 1. Experimental data accumulation and effect prediction modeling

[0172] First, through small-scale planting tests, the present invention comprehensively measured the specific performance data of different formula additives in terms of curing strength, water retention, nutrient release, heavy metal stabilization, environmental compatibility, etc. These rich experimental data laid the foundation for subsequent mathematical modeling.

[0173] For each performance index, the present invention has established corresponding prediction equations. These equations take into account multiple influencing factors such as additive formula, soil properties, environmental conditions, etc., and are modeled based on physical and chemical principles. For example, the solidification strength prediction equation uses a linear regression model, the water retention performance uses a multivariate linear regression model, and the nutrient release uses a nonlinear regression model. Through this complete set of prediction equations, the various performances of additives in planting soil can be accurately predicted.

[0174] 2. Model optimization based on machine learning

[0175] In order to improve the accuracy of the prediction model, the present invention uses the random forest regression algorithm to train and optimize these prediction equations. Random forest is an ensemble learning method that can effectively improve the stability and accuracy of prediction by generating multiple decision tree models and averaging their results. During the training process, the hyperparameters and feature engineering of the model are continuously adjusted and optimized, and finally an optimized prediction model is obtained.

[0176] This model optimization method based on machine learning overcomes the limitations of traditional empirical models. Due to the numerous and complex influencing factors such as additive components, soil conditions, and environmental factors, it is difficult to accurately model simply by relying on empirical formulas. By using machine learning technology to intelligently analyze and optimize the influencing factors, the performance of the prediction model can be greatly improved.

[0177] 3. Recipe optimization based on intelligent algorithms

[0178] The present invention uses a genetic algorithm for intelligent optimization of a large number of possible additive formula combinations. First, a large amount of theoretical formula data is generated based on the preset formula component range and change step size. Then, the performance of these formula data is predicted one by one using the aforementioned optimized prediction model.

[0179] Next, the best performing ones are selected from these theoretical formulas as the initial population, and a genetic algorithm is established for optimization. The genetic algorithm continuously iterates and optimizes the fitness level of the population through operations such as selection, crossover and mutation, and finally obtains the additive formula with the best performance indicators.

[0180] This optimization method based on intelligent algorithms overcomes the limitations of traditional "experience-trial and error" formula design. Genetic algorithms can systematically search and optimize in a wide range of possible solution spaces to obtain the true global optimal solution. This provides reliable technical support for determining the best additive formula that meets engineering application requirements.

[0181] 4. Strict engineering verification

[0182] To ensure the reliability of the optimal formula in practical applications, the present invention has fully verified it in large-scale planting tests. Under representative planting soil conditions, additives were prepared in large quantities and engineering-scale application and planting tests were carried out. Through long-term monitoring and testing of indicators such as crop growth conditions and soil physical and chemical properties, the stability and effectiveness of the optimal formula in engineering applications were ensured.

[0183] Only through such rigorous verification can we finally determine that the additive formula obtained is the truly optimal solution. This systematic research method can not only quickly produce the optimal formula design solution, but also ensure its reliable application in actual engineering. This provides a strong technical guarantee for the resource utilization of solidified sludge.

[0184] A specific embodiment 1 of the present invention is provided below. The specific implementation of each step in this embodiment 1 is described in detail as follows: In step S10, it is necessary to obtain performance data of solidified sludge additives of different formulations in planting soil. First, through a small-scale planting test, the performance of each formulation in terms of the following indicators can be determined:

[0185] 1. Curing strength S (unit: MPa):

[0186]

[0187] Among them, C 1 ,C 2 ,C 3 are the main component contents of the additives (unit: %); T is the reaction temperature (unit: °C); H is the relative humidity (unit: %); t is the curing time (unit: d); k 1 ,k 2 ,k 3 ,k 4 ,k 5 ,k 6 is the unknown coefficient; α ij is the interaction coefficient between components; n is the number of additive components; ε S is the error term. This formula can be used to predict the curing strength after the additive is mixed with the soil.

[0188] 2. Water retention performance W (unit: g / g):

[0189]

[0190] Where ρ is the soil bulk density (unit: g / cm 3 ), φ is soil porosity (unit: %), OM is organic matter content (unit: %), a 1 ,a 2 ,a 3 ,b 1 ,b 2 ,b 3 is the unknown coefficient; β ij is the interaction coefficient between components; ε W is the error term. This formula can be used to predict the effect of additives on soil water retention performance.

[0191] 3. Nutrient release rate R i(Unit: mg / kg / d):

[0192]

[0193] Wherein, pH is soil acidity and alkalinity; M is soil water content (unit: %); f i (t) is the time function, expressed as f i (t) = p i ln(t+1)+q i ;m 1 ,m 2 ,m 3 ,n 1 ,n 2 ,n 3 ,p i ,q i is the unknown coefficient; jk is the interaction coefficient between components; is the error term. This formula can be used to predict the effect of additives on soil nutrient release.

[0194] 4. Heavy metal stabilization rate S j (unit:%):

[0195]

[0196] Wherein, CEC is the cation exchange capacity (unit: cmol / kg); t is the time (unit: d); x 1 ,x 2 ,x 3 ,y 1 ,y 2 ,y 3 ,z 1 ,λ j is the unknown coefficient; δ kl is the interaction coefficient between components; is the error term. This formula can be used to predict the effect of additives on soil heavy metal stabilization.

[0197] 5. Environmental compatibility index E (dimensionless):

[0198]

[0199] Wherein, B is soil microbial biomass carbon (unit: mg / kg); E is soil enzyme activity (unit: μmol / g / h); R is plant root activity (unit: μg / g / h); u 1 ,u 2 ,u 3 ,v 1 ,v 2 ,v 3 ,w 1,ω, is the unknown coefficient; θ mo is the interaction coefficient between components; ε E is the error term. This formula can be used to evaluate the impact of additives on soil ecosystems.

[0200] 6. Interaction between additive components I (dimensionless):

[0201]

[0202] Among them, η ij is the second-order interaction coefficient; ζ ijk is the third-order interaction coefficient; ε I is the error term. This formula describes the interaction between the various components of the additive.

[0203] By measuring the above series of indicators, specific values ​​of various formulations in different performance aspects were obtained, providing basic data for subsequent model establishment and optimization.

[0204] In step S20, a complete set of effect prediction equations needs to be established based on the above experimental data. The details are as follows:

[0205] 1. Curing strength prediction equation S:

[0206] This equation takes the form of a linear regression model and can be used to predict the curing strength of the additive after mixing with the soil. 1 ,C 2 ,C 3 represents the content of the main components of the additive, T, H, t are the reaction temperature, relative humidity and curing time, k 1 ,k 2 ,k 3 ,k 4 ,k 5 ,k 6 is the unknown coefficient, α ij is the interaction coefficient between components, ε S is the error term.

[0207] 2. Water retention performance prediction equation W:

[0208] This equation takes the form of a multiple linear regression model and can be used to predict the effect of additives on soil water retention. Where ρ, φ, and OM represent soil bulk density, porosity, and organic matter content, respectively. 1 ,a 2 ,a 3 ,b 1 ,b 2 ,b 3 is the unknown coefficient, β ijis the interaction coefficient between components, ε W is the error term.

[0209] 3. Nutrient release prediction equation R i :

[0210] This equation takes the form of a nonlinear regression model and can be used to predict the effect of additives on soil nutrient release. Where pH, T, and M represent soil acidity, temperature, and water content, respectively. i (t) is the time function, m 1 ,m 2 ,m 3 ,n 1 ,n 2 ,n 3 ,p i ,q i is the unknown coefficient, γ jk is the interaction coefficient between components, is the error term.

[0211] 4. Heavy metal stabilization prediction equation S j :

[0212] This equation adopts the form of an exponential decay model and can be used to predict the effect of additives on soil heavy metal stabilization. Where CEC represents cation exchange capacity, t is time, x 1 ,x 2 ,x 3 ,y 1 ,y 2 ,y 3 ,z 1 ,λ j is the unknown coefficient, δ kl is the interaction coefficient between components, is the error term.

[0213] 5. Environmental compatibility prediction equation E:

[0214] This equation adopts the form of a semi-empirical model and can be used to evaluate the impact of additives on soil ecosystems. Among them, B, E, and R represent soil microbial biomass carbon, enzyme activity, and root activity, respectively, and u 1 ,u 2 ,u 3 ,v 1 ,v 2 ,v 3 ,w 1 ,ω, is the unknown coefficient, θ mo is the interaction coefficient between components, ε E is the error term.

[0215] 6. Component synergy equation I:

[0216] This equation describes the interaction effect between the additive components. ij is the second-order interaction coefficient, ζ ijk is the third-order interaction coefficient, ε I is the error term.

[0217] By establishing this complete set of effect prediction equations, we can predict various performance indicators of additives in planting soil based on the formulation and environmental conditions of the additives, which provides a strong mathematical model support for subsequent optimization design.

[0218] In step S30, a machine learning algorithm is required to train the above effect prediction equations. The details are as follows:

[0219] First, the formula data and effect data were standardized and normalized to eliminate the influence of dimensional differences. Then, the principal component analysis (PCA) method was used to select the influencing factors of each prediction index and determine the main independent variables.

[0220] Next, we build a prediction model based on the random forest regression algorithm. Random forest regression is an ensemble learning method that can effectively improve the stability and accuracy of predictions by generating multiple decision tree models and averaging their results. During the model building process, some hyperparameters need to be adjusted, such as the maximum depth of the decision tree and the minimum number of samples, to achieve the best prediction performance.

[0221] Specifically, the training process of the random forest regression model is as follows:

[0222] 1. Data preprocessing:

[0223] standardization:

[0224] Normalization:

[0225] Among them, μ and σ are the mean and standard deviation of feature X, respectively. min and X max are the minimum and maximum values ​​of feature X.

[0226] 2. Feature selection: Using the principal component analysis (PCA) method, the principal components whose cumulative variance contribution rate reaches more than 95% are selected as the main features.

[0227] 3. Model construction: Build a random forest regression model and set appropriate hyperparameters, such as the maximum depth of the tree, the minimum number of samples, etc.

[0228] 4. Model training: Use the training set data obtained in step S10 to fit the model.

[0229] 5. Model validation: The cross-validation method is used to evaluate the predictive performance of the model.

[0230] 6. Model optimization: Continuously optimize the prediction accuracy of the model by adjusting hyperparameters and feature engineering.

[0231] 7. Model testing: Use independent test set data to evaluate the predictive effect of the final model.

[0232] Through the above steps, an optimized random forest regression prediction model was obtained. This model can accurately predict the performance of additives in various performance indicators and provide reliable mathematical support for subsequent formula optimization.

[0233] In step S40, a large amount of theoretical formula data needs to be generated in a traversal manner according to the preset formula component range and ratio step, and its performance index is calculated using the aforementioned optimized prediction model.

[0234] The details are as follows:

[0235] 1. Determine the reasonable use range of additive components: For example, C 1 The range is 10% to 30%, C 2 The range is 5% to 15%, C 3 The range is 1% to 5%.

[0236] 2. Set an appropriate change step: Assuming a change every 2%, the following formula combinations can be generated: (10%, 5%, 1%), (12%, 7%, 2%), (14%, 9%, 3%), and so on.

[0237] 3. Use the random forest regression model established in step S30 to calculate these formula data one by one: Input C 1 ,C 2 ,C 3 ,T,H,t,ρ,φ,OM,pH,M and other parameters to predict the curing strength S, water retention W, nutrient release R i , Heavy Metal Stabilization S j And environmental compatibility E and other indicators.

[0238] Through this series of calculations, a large amount of theoretical effect data was obtained, covering a wide range of formula combinations. These rich theoretical data provide valuable information reference for subsequent optimization design.

[0239] In step S50, cluster analysis is required on these theoretical effect data to find representative formula samples. The details are as follows:

[0240] 1. Data standardization: Min-Max normalization: Eliminate the impact of dimensional differences on clustering results.

[0241] 2. Select clustering algorithm: K-Means algorithm is used here, which is a distance-based partitioning clustering method.

[0242] 3. Determine the initial position of the cluster center: You can use random initialization or select several representative recipes as seed points.

[0243] 4. Set the number of clusters K: It is necessary to use evaluation indicators such as silhouette coefficient and Calinski-Harabasz index to determine the appropriate number of clusters K. Usually, a K value is selected that makes the clusters compact and the clusters far apart.

[0244] 5. Iteratively optimize cluster centers: Use the K-Means algorithm to iteratively optimize the location of cluster centers until convergence.

[0245] Through K-Means clustering, a large amount of theoretical formula data is divided into several clusters. Each cluster has a cluster center, which represents the typical characteristics of this type of formula. The formulas corresponding to these cluster centers become the objects that need to be focused on and experimentally verified.

[0246] In step S60, it is necessary to conduct small-scale planting experiments on these representative formulas to obtain actual effect data. The details are as follows:

[0247] 1. Prepare test soil and additives: Select representative soil samples and prepare additives according to the cluster center formula obtained in step S50.

[0248] 2. Mix additives with soil: Use a mixer to mix the additives with the soil thoroughly.

[0249] 3. Planting indicator plants: In a controlled environment (temperature 25±2°C, humidity 60±5%), fill the mixed soil into small pots and plant indicator plants (such as wheat or corn).

[0250] 4. Monitor plant growth: regularly observe and record plant indicators such as seedling condition, plant height, leaf color, and biomass.

[0251] 5. Harvest and analysis test: After the plants mature, they are harvested and analyzed. The test content includes:

[0252] Plant biomass: fresh weight and dry weight were measured;

[0253] Soil physical and chemical properties: solidification strength S, water retention performance W, pH value, organic matter content, etc.;

[0254] Nutrient content: Determine the concentration of various nutrient elements in the plant;

[0255] Heavy metal content: Determine the concentration and form distribution of various heavy metals in the soil;

[0256] Soil biological indicators: microbial biomass B, enzyme activity E, root activity R, etc.;

[0257] Through this series of small-scale planting tests, we will obtain various performance data of each experimental formula under actual planting conditions. These measured data will provide a reliable basis for subsequent optimization design.

[0258] In step S70, it is necessary to select the best-performing samples from the aforementioned experimental formula set as the initial population, and establish a genetic algorithm for intelligent optimization. The details are as follows:

[0259] 1. Selecting the initial population: From the small-scale planting experiment in step S60, select several formulations with the best performance indicators as the initial population.

[0260] 2. Encode into genotype: The recipe data of these recipes (C 1 ,C 2 ,C 3 etc.) are encoded as genotypes and used as the input of the genetic algorithm.

[0261] 3. Design fitness function: The actual effect data (S, W, R) obtained in step S60 i ,S j ,E, etc.) are normalized and used as the fitness function of the genetic algorithm.

[0262] 4. Establish genetic algorithm process:

[0263] Selection operation: select individuals with higher fitness from the current population as parents;

[0264] Crossover operation: randomly combining the genotypes of the parents to generate new individuals;

[0265] Mutation operation: randomly change the genotype of individuals to increase the diversity of the population;

[0266] 5. Iterative optimization: Through multiple generations of iterations, the genetic algorithm will continuously improve the fitness level of the population.

[0267] 6. Output the optimal formula: When the preset number of iterations or convergence conditions are reached, the algorithm outputs the individual with the highest fitness in the current population, which is the optimal additive formula.

[0268] This optimization process requires a lot of computing resources and time. In the specific implementation, parallel computing can be used to speed up the optimization. In addition, improvements such as adaptive mutation probability and elite retention strategy can be introduced to further improve the convergence and stability of the algorithm.

[0269] Through this step of genetic algorithm optimization, we obtained the additive formula that achieved the best performance in all indicators, which laid the foundation for the subsequent large-scale planting verification experiment.

[0270] In step S80, it is necessary to conduct a large-scale planting test on the optimal formula to determine the final additive formula. The details are as follows:

[0271] 1. Mass production of additives: According to the optimal formula obtained in step S70, the solidified sludge additive is mass produced.

[0272] 2. Select the test site: Choose representative planting soil and divide it into different test plots.

[0273] 3. Carry out large-scale planting: Mix the additives with the soil thoroughly and plant the target crops.

[0274] 4. Monitor growth conditions: Regularly observe and record crop growth conditions, such as seedling condition, plant height, leaf color, biomass, etc.; at the same time, monitor changes in the physical and chemical properties of the soil, such as nutrient content, heavy metal status, enzyme activity, etc.

[0275] 5. Harvest and analytical testing: After the crops mature, comprehensive harvest and analytical testing will be carried out; the test content includes crop yield and quality indicators, as well as various soil indicators.

[0276] 6. Evaluate the final solution: Only when the large-scale planting verification results show that the optimal formula can continuously and stably meet various performance requirements, can the additive formula be finally determined as the best solution.

[0277] Through this series of rigorous experimental verification process, the reliability and effectiveness of the developed solidified sludge additive formula in practical applications are ensured.

[0278] In order to better understand and implement the present invention, Example 2 of a specific application scenario of the present invention is provided below: This Example 2 describes the specific application of a solidified sludge additive in agricultural soil. A large amount of sludge is generated in a shrimp pond on a beach, and a reasonable resource recovery plan is urgently needed. In response to this demand, the optimization method proposed in the present invention is used to systematically design and verify the additive formula.

[0279] 1. Experimental Data Collection

[0280] First, a small-scale planting test was conducted according to step S10 to obtain performance data of different formula additives in planting soil. The specific process is as follows:

[0281] (1) Preparation of test soil and additives

[0282] Typical local agricultural soil was selected as the test matrix. The basic physical and chemical properties of the soil are shown in Table 1.

[0283] Table 1 Basic physical and chemical properties of the test soil

[0284]

[0285]

[0286] The sludge was collected from a local beach shrimp pond with a moisture content of 75%. Under laboratory conditions, the sludge was pretreated by hot air drying and crushing to obtain dry sludge. Then, according to the formula design, three solidified sludge additives with different ratios were prepared, and their formulas are shown in Table 2.

[0287] Table 2 Additive formula design

[0288] serial number Sludge content (%) Fly ash content (%) Activated carbon content (%) A 50 30 20 B 60 25 15 C 55 35 10

[0289] (2) Small-scale planting trials

[0290] Each additive was mixed with soil at a ratio of 5%. The mixed soil was then placed in small pots (20 cm in diameter and 15 cm in depth) and wheat was sown as an indicator plant. The test was conducted in a controlled environment with a temperature of 25 ± 2 °C and a relative humidity of 60 ± 5%.

[0291] During the plant growth process, we regularly observed and recorded indicators such as plant height, leaf color, and biomass. When the wheat matured, we harvested it and conducted comprehensive analytical tests, including:

[0292] Plant biomass: fresh weight and dry weight were measured;

[0293] Soil properties: solidification strength, water holding capacity, pH value, organic matter content, etc.;

[0294] Nutrient content: Determine the nutrients such as N, P, K in plants and soil;

[0295] Heavy metal content: Determine the distribution of heavy metals such as Cu, Zn, Cd in the soil;

[0296] Through this series of experimental tests, specific data on various performance indicators of different additive formulations were obtained. These data provide a basis for subsequent model establishment and optimization.

[0297] 2. Establishment of performance prediction model

[0298] According to step S20, a corresponding mathematical prediction model is established for each performance indicator.

[0299] (1) Curing strength prediction model

[0300] The linear regression model was used to fit the relationship between the curing strength S and the additive component content and environmental conditions:

[0301] S=0.36C1+0.28C 2 +0.12C 3 +0.015T+0.011H+0.023t+0.008C 1 C 2 +ε S ;

[0302] Among them, C 1 , C 2 , C 3 are the mass fractions of sludge, fly ash and activated carbon (%) respectively; T is the reaction temperature (℃); H is the relative humidity (%); t is the curing time (d); ε S is the error term. The model can be used to predict the curing strength after the additive is mixed with the soil.

[0303] (2) Water retention performance prediction model

[0304] The multivariate linear regression model was used to fit the relationship between the water retention coefficient W and the additive components and soil properties:

[0305] W=0.42C 1 +0.35C 2 +0.23C 3 +0.18ρ-0.26φ+0.15OM+0.012C 1 C 2 +ε W ;

[0306] Where ρ is the soil bulk density (g / cm 3 ), φ is porosity (%), OM is organic matter content (%), ε W is the error term. This model can be used to predict the effect of additives on soil water retention.

[0307] (3) Nutrient release prediction model

[0308] The nonlinear regression model was used to fit the nutrient release rate R i Relationship with additive components, soil properties and environmental factors:

[0309]

[0310] Wherein, pH is soil acidity and alkalinity; T is temperature (℃); M is water content (%); t is time (d); is the error term. The model can be used to predict the effect of additives on soil nutrient release.

[0311] (4) Heavy metal stabilization prediction model

[0312] The exponential decay model was used to fit the heavy metal stabilization rate S j Relationship with additive components, soil properties and time:

[0313]

[0314] Wherein, CEC is the cation exchange capacity (cmol / kg); t is the time (d); is the error term. This model can be used to predict the effect of additives on soil heavy metal stabilization.

[0315] (5) Environmental compatibility prediction model

[0316] The semi-empirical model was used to fit the relationship between the environmental compatibility index E and the additive components and soil biological indicators:

[0317] E=0.36C 1 +0.28C 2 +0.18C 3 +0.12B+0.15E+0.10R+0.42sin(0.08t+1.2)+ε E ;

[0318] Wherein, B is soil microbial biomass carbon (mg / kg); E is enzyme activity (μmol / g / h); R is root activity (μg / g / h); t is time (d); ε E is the error term. This model can be used to evaluate the impact of additives on soil ecosystems.

[0319] By establishing the above series of prediction models, the performance of additives in various performance indicators can be accurately predicted, which provides reliable mathematical support for subsequent formulation optimization.

[0320] 3. Recipe optimization and traversal

[0321] According to step S40, a large amount of theoretical formula data is generated in a traversal manner within the preset component content range. The details are as follows:

[0322] (1) Setting the component content range

[0323] According to the results of previous experiments, the reasonable use range of each component of the additive was determined: sludge content C 1 :45%-65%; Fly ash content C 2 :20%-40%; Activated carbon content C 3 :5%-15%;

[0324] (2) Generate theoretical formula data

[0325] Within the above range, a large number of formula combinations were generated with a step size of 2%. For example: (45%, 20%, 5%), (47%, 22%, 7%), (49%, 24%, 9%), (51%, 26%, 11%), etc.

[0326] (3) Performance prediction calculation

[0327] Using the prediction model established above, the performance of these theoretical formulas was comprehensively calculated, and the prediction data of each formula in terms of curing strength, water retention, nutrient release, heavy metal stabilization, environmental compatibility, etc. were obtained.

[0328] Through this series of formulation traversal and performance prediction, a large amount of theoretical effect data was obtained, which provided rich references for subsequent optimization design.

[0329] 4. Cluster analysis and optimization

[0330] According to steps S50 and S70, cluster analysis and genetic algorithm optimization are performed on the theoretical effect data, and the best additive formula is finally determined.

[0331] (1) Cluster analysis

[0332] First, the theoretical effect data was normalized by Min-Max to eliminate the influence of dimensional differences. Then, the K-Means algorithm was used to cluster the data. After many attempts, it was finally determined that the cluster number K = 5 was appropriate.

[0333] Through K-Means clustering, a large number of theoretical recipes are divided into 5 clusters. Each cluster has a cluster center, which represents the typical characteristics of this type of recipe. The recipe composition and prediction performance corresponding to these 5 cluster centers are shown in Table 3.

[0334] Table 3 Recipes corresponding to cluster centers and prediction performance

[0335]

[0336] These five typical formulas will become the objects that need to be verified by subsequent experiments.

[0337] (2) Genetic algorithm optimization

[0338] Three of the five cluster center formulas are selected as initial populations and encoded as genotypes. The actual effect data obtained in step S60 is normalized and used as the fitness function of the genetic algorithm.

[0339] Then, the operation process of the genetic algorithm was established, including the steps of selection, crossover and mutation. After 400 generations of iterative optimization, the genetic algorithm finally converged and output the individual with the highest fitness, which is the optimal additive formula.

[0340] The optimal formula consists of 55% sludge, 32% fly ash and 13% activated carbon. It is predicted that the formula has reached the optimal level in various performance indicators: curing strength 2.75MPa, water retention coefficient 0.44g / g, nutrient release rate 9.3 / 4.0 / 1.9mg / kg / d, heavy metal stabilization rate 95 / 93 / 90%, and environmental compatibility index 0.86.

[0341] 5. Large-scale verification

[0342] According to step S80, the above optimal formula is verified by large-scale planting. The specific situation is as follows:

[0343] (1) Preparation of additives

[0344] According to the optimal formula composition ratio, solidified sludge additives were produced in large quantities. The quality stability of the additive is good, and all indicators meet the requirements.

[0345] (2) Large-scale planting trials

[0346] Select a representative agricultural land (2000m2) in the local area. 2 ), divided into different test plots. The produced additives were thoroughly mixed with the soil and wheat was sown as an indicator crop.

[0347] During the entire growth period, the growth of crops was regularly monitored and recorded. At the same time, the soil physical and chemical properties, nutrient status, heavy metal content, etc. were also tracked and tested.

[0348] (3) Harvest and Analysis

[0349] After the wheat matured, each test area was fully harvested and analyzed. The test results are shown in Table 4.

[0350] Table 4 Results of large-scale planting test

[0351] index Additive treatment area Control area Wheat yield (kg / ha) 6750 5800 Soil solidification strength (MPa) 2.62 1.84 Soil water holding coefficient (g / g) 0.41 0.32 Soil pH 6.8 6.2 Organic matter content (%) 3.1 2.5 N / P / K content (mg / kg) 210 / 28 / 180 160 / 22 / 145 Soluble heavy metal content (mg / kg) Cu 0.38 / Zn 1.2 / Cd 0.05 Cu 0.54 / Zn 1.7 / Cd 0.08 Soil microbial biomass (mg / kg) 520 430 Soil enzyme activity (μmol / g / h) 34 26 Root activity (μg / g / h) 72 58

[0352] It can be seen from the results that the use of the optimal additive formula optimized by the present invention for large-scale planting not only significantly improves the physical and chemical properties of the soil, but also increases the yield and quality of crops. At the same time, the additive also effectively reduces the bioavailability of heavy metals in the soil and maintains a good soil ecological environment. In order to more intuitively display the relevant data of this Example 2, a number of charts are provided below:

[0353] Figure 2 The release kinetics curve of nutrients (nitrogen, phosphorus, and potassium) in the optimized formula is shown. The changing trend of nutrient release within 31 days is shown in the form of multiple curves. It can be seen that nitrogen releases the fastest, followed by phosphorus, and potassium releases the slowest. This gradient release characteristic is conducive to the continuous absorption and utilization of crops.

[0354] Figure 3 The grouped bar chart shows the changes in the content of three heavy metals, copper, zinc and cadmium, in the soil before and after treatment. The figure clearly shows that after the additive treatment, the content of the three heavy metals has been significantly reduced, especially cadmium has the largest reduction, indicating that the additive has a significant passivation effect on heavy metals.

[0355] Figure 4 The radar chart comprehensively shows the differences between the treatment area and the control area in soil ecological indicators such as microbial biomass, enzyme activity, root activity, organic matter content and pH value. It can be seen intuitively from the figure that the treatment area is better than the control area in all indicators, especially in terms of microbial biomass and root activity.

[0356] In summary, this method can not only quickly determine the optimal additive formula that meets the requirements of engineering applications, but also ensure its stability and environmental safety in actual planting. This provides strong technical support for the resource utilization of solidified sludge in farmland.

[0357] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for optimizing solidified sludge additives for planting soil, characterized in that: The following steps are involved: S10, obtaining formula data of different formulas of the solidified sludge additive and effect data obtained from specific experiments corresponding to each formula, wherein the effect data includes solidification strength, water retention performance, nutrient release rate, heavy metal stabilization degree and environmental compatibility index; S20, establishing an effect prediction equation group of the solidified sludge additive, wherein the effect prediction equation group includes a solidification reaction equation, a water retention performance equation, a nutrient release equation, a heavy metal stabilization equation, an environmental compatibility equation, and a component synergy equation; S30, using a machine learning algorithm to train the effect prediction equation group using the formulation data of the different formulations to obtain a preliminary formulation-effect relationship model; S40, generating multiple sets of formula data by traversal according to the preset formula component range and ratio step, and calculating the multiple sets of formula data by using the preliminary formula effect relationship model to obtain theoretical effect data; S50, performing cluster analysis on the multiple groups of recipe data based on the theoretical effect data to obtain multiple recipe cluster centers; S60, selecting the formula corresponding to each of the formula cluster centers to form an experimental formula set, conducting a small-scale planting experiment, and collecting actual effect data; S70, selecting multiple groups of recipes with the best actual effects from the experimental recipe set as the initial population, normalizing the actual effect data and setting it as a fitness function, establishing a genetic algorithm, and obtaining the optimal recipe through multi-generation iterative optimization; S80, conducting a large-scale planting verification experiment on the optimal formula to determine the final solidified sludge additive formula.

2. The method for optimizing solidified sludge additives for planting soil according to claim 1, characterized in that: The specific experiment corresponding to each of the above-mentioned formulas is specifically a small-scale planting experiment, which includes the following steps: Step 1, preparing test soil and solidified sludge additives; Step 2, mixing the solidified sludge additive with soil according to different formulas; Step 3: Plant indicator plants in a controlled environment; Step 4: Regularly monitor and record plant growth; Step 5: measuring plant biomass and soil physical and chemical properties after harvest; Step 6: Analyze the curing strength, water retention performance, nutrient release rate, heavy metal stabilization degree and environmental compatibility index.

3. The method for optimizing solidified sludge additives for planting soil according to claim 1, characterized in that: The solidification reaction equation is used to predict the solidification strength after the solidification sludge additive is mixed with soil, the input is the additive component content and the reaction conditions, and the output is the solidification strength value.

4. The method for optimizing solidified sludge additives for planting soil according to claim 1, characterized in that: The water retention performance equation is used to predict the water retention capacity of soil-additive mixtures. The input is the additive component content and the basic properties of the soil, and the output is the water retention coefficient.

5. The method for optimizing solidified sludge additives for planting soil according to claim 1, characterized in that: The nutrient release equation is used to predict the release rate of nutrients in additives, the input is the content of additive components and environmental factors, and the output is the nutrient release rate.

6. The method for optimizing solidified sludge additives for planting soil according to claim 1, characterized in that: The heavy metal stabilization equation is used to predict the stabilization effect of additives on heavy metals in soil, the input is the additive component content and the initial concentration of heavy metals, and the output is the heavy metal stabilization rate.

7. The method for optimizing solidified sludge additives for planting soil according to claim 1, characterized in that: The environmental compatibility equation is used to evaluate the impact of additives on soil ecosystems, the input is the additive component content and soil biological indicators, and the output is the environmental compatibility index.

8. The method for optimizing solidified sludge additives for planting soil according to claim 1, characterized in that: The component synergy equation is used to describe the interaction effect between the components of the additive, the input is the content of each component, and the output is the synergy coefficient.

9. The method for optimizing solidified sludge additives for planting soil according to claim 1, characterized in that: The machine learning algorithm adopts a random forest regression algorithm. The step of adopting the machine learning algorithm and using the formula data of the different formulas to train the effect prediction equation group specifically includes: Data preprocessing: standardize and normalize the recipe data and effect data; Feature selection: Use principal component analysis to select the main features; Model building: Build a random forest regression model and set appropriate hyperparameters; Model training: Use training set data to train the model; Model validation: Use cross-validation methods to evaluate model performance; Model optimization: Optimize the model by adjusting hyperparameters and feature engineering; Model testing: Use the test set data to evaluate the prediction performance of the final model.

10. The method for optimizing solidified sludge additives for planting soil according to claim 2, characterized in that: The indicator plant is wheat or corn.

Images