Method for realizing rapid composting based on prediction and reverse adjustment

Abstract

The invention provides a method for realizing rapid composting based on prediction and reverse adjustment, which comprises the following steps: 1) carrying out Meta analysis, comprehensively judging eight maturity indexes inside and outside a pile body, and providing an optimized data set for establishing an optimal prediction model for machine learning; 2) testing the performance of the model in a real scene, and collecting data of eight indexes; 3) utilizing Python to integrate prediction and reverse adjustment in the charcoal composting process into the optimal prediction model, and setting upper and lower limit values of process indexes in a thermophilic period, a high-temperature period and a cooling period; 4) inputting the collected eight index data into a model fusing prediction and reverse adjustment in the biochar composting process, and optimizing upper and lower limit values of process indexes in three periods; and 5) finally verifying the composting days based on prediction and reverse adjustment in the charcoal composting process in a real scene experiment so as to realize rapid composting. According to the method, the multi-target prediction model is established, different indexes of compost can be rapidly predicted, and the cost and time of manual determination are reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of biocomposting, and particularly relates to a method for rapid composting based on prediction and reverse regulation. Background Art

[0002] Composting is recognized as a sustainable and eco-friendly biochemical technology due to its economic efficiency, simplicity of operation, and adaptability to the increasing treatment of organic solid waste. It has been widely used in the resource utilization of organic waste. This process relies on microorganisms to degrade and transform biodegradable organic matter into stable substances similar to humus, thereby achieving harmless treatment. Its final product can be used as a soil conditioner to optimize the physical and chemical properties of the soil. However, traditional aerobic composting methods still face many challenges, such as temperature fluctuations during the composting process, long degradation cycles, unstable product quality, and problems such as nitrogen loss caused by ammonia volatilization and greenhouse gas emissions. Therefore, it is crucial to develop efficient composting technologies to overcome these limitations. To optimize the composting environment and improve the degradation efficiency, researchers have explored various methods, and among them, incorporating exogenous additives is one of the most common strategies. Exogenous additives mainly include three categories: chemical, physical, and microbial. Among them, compared with chemical or microbial additives, physical additives such as biochar, zeolite, and MnO2 show more significant advantages in reducing nitrogen loss and ammonia emissions, making their application potential in enhancing the composting process highly concerned. In the past three decades, the application of biochar in traditional solid waste composting has become a research hotspot. Biochar can effectively promote humification, extend the heat-resistant period, accelerate maturation, and reduce greenhouse gas emissions, thereby improving the quality of compost. Nevertheless, the optimal usage conditions of biochar still lack systematic research. Most existing experiments are limited to the analysis of single variables, mainly focusing on reducing gas emissions or improving compost characteristics (such as pH value, moisture content, carbon-nitrogen ratio, etc.) while ignoring their comprehensive effects.

[0003] In addition, due to the high heterogeneity of the physicochemical properties of biochar (such as raw material source, pyrolysis temperature, particle size, total porosity) and initial composting conditions (including C / N ratio, water content, pH value), the experimental results are highly variable and even contradictory, which to some extent limits the in-depth understanding of the key factors affecting the composting maturity process. Therefore, it is of great significance to quantitatively evaluate the effect of biochar on composting effect. Meta-analysis, as a powerful statistical method, can systematically evaluate the effect of biochar addition on composting indicators using existing research data. Previous meta-analyses have examined the effects of biochar characteristics and initial composting conditions on internal maturity indicators (such as pH, conductivity, carbon and nitrogen loss, ammonia nitrogen, nitrate nitrogen) and external maturity indicators (such as N2O, CH4, NH3). However, many existing meta-analyses did not conduct heterogeneity tests, publication bias tests, or significance analyses based on one-way analysis of variance (ANOVA), which affected their reliability. Therefore, integrating internal and external maturity indicators and comprehensively evaluating the impact of the high heterogeneity of biochar properties and composting conditions on compost maturity are important issues that need to be urgently addressed in current research.

[0004] Rapid composting is the use of control methods such as reactors, process optimization, and exogenous additives to artificially strengthen the key controllable factors of composting, promote the humification process, and shorten the composting cycle [see reference 1 for details]. However, composting is a complex biochemical process that is easily affected by various conditions, which poses a challenge to accurately adjust parameters in the process of achieving rapid composting. Therefore, developing a convenient, fast, and user-friendly method to predict compost maturity indicators and achieve rapid composting is the key to achieving efficient compost production. As a tool with powerful computing power for processing and analyzing big data, machine learning can quickly and accurately predict compost maturity indicators through modeling, and reversely adjust and optimize composting process indicators to achieve rapid composting. Although existing machine learning has been studied in predicting maturity indicators (C / N, GI) in the composting process [see reference 2 for details], there is still a lack of research on predicting multiple indicators and reversely adjusting to achieve rapid composting.

[0005] Technical solution of prior art 1 It is crucial for Shunxi Zhou et al. [see Reference 3 for details] to quantitatively evaluate the effects of biochar properties and composting conditions on the quality (physical, chemical, and nutritional properties) and ecological risks of the final compost product through (Meta) analysis. To comprehensively understand the mechanism of action of biochar in the composting process and its optimized application, the research team constructed a systematic analysis framework. The team conducted a global Meta-analysis based on 84 studies to quantitatively evaluate the overall effect size of biochar on composting indicators, including physical, chemical, and nutritional properties, as well as ecological risk indicators. According to the changes in biochar properties and composting conditions, the average response of the target composting indicators to the biochar addition rate was further evaluated. Finally, the mechanism of action of the key properties of biochar on the quality of the compost product and ecological risks was explored, and the optimal use conditions of biochar were proposed, that is, it is recommended to add straw biochar at a rate of 10-15% to achieve the best production mode of composting.

[0006] Disadvantages of the prior art I When conducting the Meta-analysis, the dataset only collected data from 84 pieces of literature, lacking comprehensiveness; when conducting the Meta-analysis, it focused on the indicators inside the compost, and did not provide a more comprehensive description of the gas indicators outside the compost; when proposing the optimal use conditions of biochar, it only considered the effects of Bio-Cu and Bio-Zn of heavy metals inside the compost.

[0007] Technical solution of the prior art II Xin Wan et al. [see Reference 2 for details] adopted four machine learning (ML) techniques to achieve automated composting by predicting compost maturity and providing process regulation. Machine learning algorithms such as random forest (RF), extreme gradient boosting (XGBoost), Light gradient boosting machine (LightGBM), and Multilayer Perceptron (MLP) were used to predict the seed germination index (GI) and C / N ratio. Based on the best fusion models for multi-task prediction of GI and C / N ratio (R 2 were 0.977 and 0.986 respectively), the key factors and their interactions with maturity were determined through SHAP analysis. In addition, the machine learning model was verified on the compost reactor, and the genetic algorithm (GA) was used to find the optimal solution (i.e., the optimal operating conditions) of the composting operating conditions (i.e., temperature, pH, EC, and aeration) based on the model. During this process, compost batches that reached maturity within 15 days were screened, the compost maturity curve was plotted, and it was determined whether the compost situation needed to be adjusted. Based on the above prediction applications of machine learning, regulation can be provided to ensure the decomposition of food and garbage within the specified time.

[0008] Disadvantages of the prior art II It only predicts the maturity index for a single system, namely food and garbage; it only predicts two physical and chemical properties in the compost pile, namely C / N and germination index (GI) of seeds, to judge the compost, and does not involve the gas emission index outside the compost pile; it does not involve analyzing the types and properties of additives on the final compost; in the process index reverse adjustment stage, only the compost batches that reach maturity within 15 days are screened, and the screening principle is relatively incomplete, which is likely to lead to problems such as unrepresentative data. In the process index reverse adjustment stage, it does not adjust according to different composting processes, but only uses the GA genetic algorithm to adjust the composting process as a whole, and cannot achieve precise adjustment. Summary of the Invention

[0009] The purpose of the present invention is to solve the above-mentioned defects existing in the prior art and provide a method for rapid composting based on prediction and reverse adjustment.

[0010] Due to its unique porous structure, large specific surface area and rich surface functional groups, biochar is widely used as an additive in the composting process to promote humification, improve maturity and optimize composting performance. However, due to the diversity of the characteristics of biochar itself and the initial composting conditions, its impact on compost maturity still has many uncertainties. Therefore, accurately predicting the compost maturity index and optimizing the composting process based on key parameters are the keys to achieving efficient composting.

[0011] To solve this problem, it is necessary to comprehensively analyze the key indicators inside and outside the compost pile, use the Meta-analysis method to screen the key biochar characteristics and initial compost properties that affect compost maturity, and build a machine learning model based on these core variables to accurately optimize the composting process parameters and achieve rapid composting. At the same time, to improve the operability of the application, the present invention constructs a visual prediction interface through a web page, enabling users to intuitively judge the compost maturity and dynamically adjust the composting parameters. This method not only reduces the manual detection cost and simplifies the operation process, but also provides scientific decision-making support for compost production practice, provides a theoretical basis for the efficient resource utilization of biomass waste, and promotes the development of composting technology towards environmental protection, high quality and large scale.

[0012] The present invention adopts the following technical solutions: A method for rapid composting based on prediction and reverse adjustment, comprising the following steps; Step 1. Conduct a Meta-analysis to comprehensively evaluate 8 maturity indicators inside and outside the compost pile, screen the key biochar properties and initial compost properties that affect the final compost, and provide an optimized data set for establishing an optimal prediction model for machine learning, where the indicators inside the compost pile are C / N, GI, NO3 - -N, NH4 +-N, the off-heap indicators are NH3, N2O, CO2, and CH4.

[0013] Based on the results of the forest plot, the influence relationships between biochar properties, compost properties, and maturity indicators were analyzed through structural equation modeling. The standardized total effects of the structural equation model were used to judge the influence degree of property indicators on maturity indicators and rank them, thereby verifying the reliability of the forest plot results and revealing the importance of key biochar properties and key compost properties, providing optimization indicators for the machine learning model to establish a dataset. When establishing the machine learning model, compost process indicators were added to establish a dataset including 8 maturity indicators. The basic machine learning algorithms extra trees, random forest, gradient boosting, and eXtremeGradient Boosting were selected. An aggregated model of five models was constructed through Python editing. The data records of the same experiment were grouped and then stratified into a training set and a test set in a ratio of 8:2; Bayesian optimization was used to perform hyperparameter tuning to R 2 , root mean square error, and mean absolute error were used as the criteria for evaluating the algorithm to screen out the algorithm with the best performance that can predict the in-heap maturity indicators and off-heap gas indicators; 10 different random states were set to evaluate the performance of the five models, using R 2 (squared score), RMSE (root mean square error), and MAE (mean absolute error) were used to compare the prediction accuracies of the five models, and the optimal prediction model was used as the best prediction model.

[0014] Step 2. Conduct experiments on the performance of the best prediction model in a real composting scenario and obtain data of the corresponding 8 indicators. Samples were collected on days 0, 4, 9, 16, 23, 37, and 51, including biochar property parameters, initial compost property parameters, compost process indicators, and 8 maturity indicator parameters measured in the experiment at different stages of composting.

[0015] Step 3. Incorporate the genetic algorithm into the optimal prediction model in Step 1 using Python, and preliminarily determine the upper and lower limits of the process indicators for 3 periods. The 3 periods of the composting process are the thermophilic period, the high-temperature period, and the cooling period.

[0016] Step 4. Use the data of the 8 maturity indicators obtained in Step 2 to feed into the optimal prediction model incorporated with the genetic algorithm to optimize the upper and lower limits of the process indicators for the 3 periods of composting. In-heap indicators of the compost: C / N and NH4 + -N shows a decreasing trend with fluctuations during the composting process, and the larger the GI and NO3 - -N, the better. At the same time, for the off-heap indicators of the compost: CO2, CH4, N2O, and NH3, the smaller the emissions, the better.

[0017] The Meta-analysis in Step 1 includes: S101. By setting keywords, collecting and screening literature on the database, and through heterogeneity test, publication bias test, one-way ANOVA combined with post hoc least significant analysis for subgroup grouping significance verification; S102. Using Meta software to explore the influence of biochar properties and initial compost properties on eight compost maturity indicators through forest plots, evaluating and proposing the key biochar properties and initial compost properties affecting maturity, calculating the average effect size of each classification subgroup through the experimental group and control group of each indicator and the corresponding standard deviation and sample size, and comparing the subgroup average effect size with the overall effect size calculated only considering the addition or not of biochar to judge and recommend the indicators with better classification subgroup effects.

[0018] Furthermore, Step 1 also includes using cosine similarity to establish the application domain of the model. When determining the application domain, the training set is used as the reference data set, the test set is used as the exploration set, and cosine similarity is used to calculate the similarity between the data in the test set and the data used for modeling in the training set, forming different parts of the test set. Data below a certain threshold is data outside the application domain, and data above a certain threshold is data within the application domain. For each pre-set cosine similarity, set a gradient to discard data outside the application domain from the test data set and recalculate R 2 , RMSE and the amount of discarded data, discarding the threshold data with the lowest R 2 and the highest RMSE value. After ensuring that the data outside the applicable range is small, data greater than the threshold will be identified as the optimal threshold, and the Shapley additive explanation method is used to quantitatively evaluate the contribution of input feature values to the compost maturity indicators and verify the prediction accuracy of the optimal prediction model through experiments.

[0019] Furthermore, Step 2 includes compost treatments with different biochar addition rates and compares the prediction results with the actual data collected during the compost maturity process to prove the performance of the best prediction model.

[0020] Furthermore, in Step 3, the genetic algorithm is combined with actual composting. Compared with the control group without biochar addition, biochar is added at a ratio of 15%, and the application domain is used to significantly reduce the number of mature days.

[0021] Furthermore, the genetic algorithm optimization of the composting process in Step 4 includes: (1). Initial population generation: Input the experimental data obtained in step 2 into the best prediction model. The eigenvalue includes biochar, initial compost properties, and composting process indicators. The genetic algorithm randomly generates 50 to 100 different combinations of composting process parameters, including composting cycle, pH value, temperature, and moisture content. These initial parameters constitute the initial population, providing a basis for optimization.

[0022] (2). Fitness evaluation for each parameter combination: Replace the original process indicators with the generated process parameter combinations and perform model prediction again. If the new predicted maturity index is better than the original predicted value, fitness evaluation is carried out. The higher the fitness value, the better the effect of the new predicted value in optimizing the original predicted value.

[0023] (3). Individual selection: When all the process parameter combinations are re-predicted, it is regarded as one round of iteration. A total of 15 iterations are carried out. In each iteration process, select the parameter combination with the highest fitness as the parent generation each time, and retain these optimal combinations for the next generation inversion to continuously optimize the process parameters.

[0024] (4). Gene exchange: Randomly select a parameter position for crossover among the selected parameter combinations to introduce new parameter combinations, enhance the diversity of the population, and increase the possibility of finding a better solution.

[0025] (5). Gene mutation: Randomly modify the value of a certain process parameter to increase the diversity of solutions, prevent the algorithm from falling into a local optimal solution, and improve the global search ability.

[0026] (6). Optimal individual selection: Through continuous optimization of selection, crossover, and mutation, finally output the optimal combination of composting process parameters to maximize the optimization of the current values of 8 maturity indicators. Adjust and optimize the process parameters at each stage of compost sampling to make the maturity indicators at this stage reach the optimal value, and maximize the shortening of the compost maturity time and improve the composting effect during the entire composting period.

[0027] Further, the establishment of the dataset includes: Using cosine similarity to establish the application domain of the model. When determining the application domain, use the training set as the reference dataset and the test set as the exploration set. Use cosine similarity to calculate the similarity between the data in the test set and the data used for modeling in the training set. Based on the similarity measure, different parts of the test set are formed. Data below a certain threshold is data outside the application domain, and data above a certain threshold is data within the application domain. For each preset cosine similarity, set a gradient to discard the data outside the application domain from the test dataset and recalculate R 2 , the root mean square error and the amount of discarded data, discard the threshold data with the lowest R 2 and the highest root mean square error value. After ensuring that the data outside the applicable range is small, the data greater than this threshold will be identified as the optimal threshold.

[0028] Furthermore, the screening rules in S101 are as follows: (1) The inclusion criteria require original research, and the quantitative research results show that the application of biochar leads to changes in compost maturity, that is, the data obtained by the author through experiments. Moreover, the theme must be (or include) literature on the impact of biochar as an additive on compost maturity, rather than review articles, machine learning articles, or Meta-analysis articles that need to be collected. (2) Each study needs to include a control group and an experimental group, consisting of multiple groups for comparison. (3) The selected studies are expected to include biochar properties and initial compost properties. (4) Obtain the mean, standard deviation, and the number of control and experimental groups from the table, or digitize them from the chart using the GetDate image digitizer. For studies that only provide the standard error, use the formula SD = SE × , calculate SD, where n represents the sample size. For studies that do not provide SD and SE, SD is calculated as 1 / 10 of the mean.

[0029] Furthermore, the process indicators are: process moisture: MP_P, process temperature: TEMP_P, process pH: pH_P, process time: Day_P.

[0030] Furthermore, the upper and lower limits of the process indicators for the three preliminary periods in step 3 are as follows: The process time in the thermophilic period is 1 - 7 days, the process temperature is 25 - 50 °C, the process pH is 6.5 - 8.5, and the process moisture is 40% - 65%; the process time in the high-temperature period is 7 - 15 days, the process temperature is 50 - 70 °C, the process pH is 6.5 - 8.5, and the process moisture is 45% - 55%; the process time in the cooling period is greater than 15 days, the process temperature is 40 - 60 °C, the process pH is 6.5 - 8.5, and the process moisture is 40% - 55%.

[0031] Advantages of the present invention: 1. The present invention combines the maturity index inside the compost pile and the gas index outside the compost pile to comprehensively judge the impact of the addition of biochar on compost maturity, and recommends the best properties and the best range that can improve compost maturity, providing theoretical guidance for high-quality biochar composting.

[0032] 2. The present invention establishes a multi-objective prediction model, which can quickly predict different indicators of compost, so as to reduce the cost and time invested in manual measurement.

[0033] 3. By combining the best prediction model with the prediction and reverse regulation in the biochar composting process, the present invention can accurately control the process indicators at different composting stages and finally achieve the rapid maturity of compost. Description of the Drawings

[0034] Figure 1 Recommend the characteristics of biochar and initial compost for the forest plot analysis of the impact based on compost maturity indicators. Recommendation level: strongly recommended in green circles with white text; generally recommended in light green circles with black text; red circles with yellow text indicate the opposite advice.

[0035] Figure 2 Rank the importance of the key characteristics affecting compost maturity based on the structural equation model.

[0036] Figure 3 For the comparison and establishment of the optimal prediction model for maturity prediction, (a)-(d) are the comparisons of RF, XGB, ET, GB, and FUSION models. (a) is C / N, (b) is GI, (c) is NO3 - -N, (d) is NH4 + -N; (e)-(h) are the establishment of the optimal prediction model. (e) is C / N, (f) is GI, (g) is NO3 - -N, (h) is NH4 + -N.

[0037] Figure 4 For the comparison and establishment of the optimal prediction model for maturity prediction, (a)-(d) are the comparisons of RF, XGB, ET, GB, and FUSION models; (a) is CO2, (b) is CH4, (c) is N2O, (d) is NH3; (e)-(h) are the establishment of the optimal prediction model, (e) is CO2, (f) is CH4, (g) is N2O, (h) is NH3.

[0038] Figure 5 For the comparison of the prediction results and actual data of different biochar addition amounts on the internal maturity indicators of compost, (a-1)-(a-3) are C / N, (b-1)-(b-3) are GI, (c-1)-(c-3) are NO3 - -N, (d-1)-(d-3) are NH4 + -N.

[0039] Figure 6 For the comparison of the prediction results and actual data of different biochar addition amounts on the internal maturity indicators of compost, (a-1)-(a-3) are CO2, (b-1)-(b-3) are CH4, (c-1)-(c-3) are N2O, (d-1)-(d-3) are NH3.

[0040] Figure 7 Application of genetic algorithm to optimize the composting process indicators.

[0041] Figure 8 Flowchart of the steps of the present invention. Detailed implementation manners

[0042] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions in the present invention will be described clearly and completely below. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without any creative work belong to the scope of protection of the present invention.

[0043] As Figure 1 、 Figure 8 shown, a method for achieving rapid composting based on prediction and reverse regulation of the present invention includes: First, conduct a Meta-analysis to comprehensively evaluate eight maturity indicators in the compost body (C / N, GI, NO3 - N, NH4 + -N) and outside the compost body (NH3, N2O, CO2, CH4), and screen the key biochar properties and initial compost properties that affect the final composting, providing an optimized dataset for establishing a model by machine learning.

[0044] By setting keywords (biochar or black carbon or char) and (maturity or mature) and (composting or compost), excluding review and meta-analysis. Set the search time from January 1, 1990, to January 31, 2024, and collect literature on databases such as Web of Science, Google Scholar, Elsevier Science Direct, and PubMed and screen them according to certain screening rules. The screening rules are as follows: First, the inclusion criteria require original research, and the quantitative research results show that the application of biochar leads to changes in compost maturity. Second, each study needs to include a control group (without biochar) and an experimental group (with biochar), consisting of multiple groups for comparison. Third, the selected studies are expected to include biochar characteristics and initial compost. In addition, obtain the mean, standard deviation (SD), and the number of control groups and experimental groups from the table, or digitize them from the chart using the GetDate image digitizer. For studies that only provide the standard error (SE), use the formula SD = SE × , calculate SD, where n represents the sample size. For studies that do not provide SD and SE, SD is calculated as 1 / 10 of the mean. Finally, 125 relevant papers are screened out, including a total of 269 groups of valid data for subsequent Meta-analysis.

[0045] Heterogeneity tests, publication bias tests, and one-way analysis of variance (ANOVA) combined with post hoc least significant difference (LSD) analysis were performed to verify the significance of subgrouping and ensure the reliability of the Meta-analysis results. Finally, the biochar properties and initial compost properties are shown in Table 1, the classification of biochar properties is shown in Table 2, and the classification of initial compost properties is shown in Table 3.

[0046] Table 1 Nomenclature and abbreviations of complete biochar characteristics, initial compost characteristics, and compost maturity indices

[0047] Table 2 Subgroup classification of biochar properties and their thresholds

[0048] Table 3 Subgroup classification of initial compost properties and their thresholds

[0049] The detailed impact of classified biochar properties and initial compost properties on eight final compost maturity indices was explored through a forest plot using MetaWin 2.1 software. The key biochar properties and initial compost properties affecting the final maturity were comprehensively evaluated. The average effect size of each classification subgroup was calculated through the experimental and control groups of each index, as well as the corresponding standard deviation and sample size. By comparing the subgroup average effect size with the calculated overall effect size, the indices with better classification subgroup effects were judged and recommended in turn. As Figure 1 shown, the results indicate that specific biochar characteristics such as FT (straw), PT (>400), C / N (100 - 200), PV (<0.02), SA (<100), and AR (>12) have a significant positive impact on compost maturity and can be used as recommended indices to improve compost maturity. In terms of initial compost characteristics, ICN (24 - 28), WT (sludge), and IMC (55 - 60) show good signs of improving compost maturity.

[0050] Recommendation level: A white font in a green circle indicates a strong recommendation; a black font in a light green circle indicates a general recommendation; a red circle with a yellow font indicates the opposite suggestion.

[0051] Based on the results of the forest plot, the influence relationships among biochar properties, compost properties, and maturity indices were analyzed through a structural equation model. The influence degree of property indices on maturity indices was judged and ranked through the standardized total effect of the structural equation model. This was used to verify the reliability of the forest plot results and reveal the importance of key biochar properties and key compost properties, providing optimization indices for the machine learning model to establish a dataset. As Figure 2As shown in the figure, the key traits affecting compost maturity were ranked. The results showed that the property PV of biochar had the most significant effect on compost maturity, followed by FT, AR, and IMC, WT, IC / N, C / N, IpH, PT, and SA.

[0052] Based on the key properties from the Meta-analysis, to precisely regulate and achieve rapid composting, composting process indicators (process moisture: MP_P, process temperature: TEMP_P, process pH: pH_P, process time: Day_P) were added when establishing the machine learning model, and a dataset including 8 maturity indicators was established. Among them, there were 949 groups of data for predicting C / N, 909 groups of data for predicting GI, - 1210 groups of data for predicting NO3 + -N, 1204 groups of data for predicting NH4 2 -N, 734 groups of data for predicting CO2, 493 groups of data for predicting CH4, and 451 groups of data for predicting N2O, and 974 groups of data for predicting NH3. The basic machine learning algorithms extra trees (ET), random forest (RF), gradient boosting (GB), and eXtreme Gradient Boosting (XGB) were selected, and a combined model (Fusion) of five models was constructed through Python editing. To reduce the risk of data leakage, the data records of the same experiment were grouped and then stratified into a training set and a test set at a ratio of 8:2 to reduce the risk of model overfitting; Bayesian optimization was used to perform hyperparameter tuning with - R, RMSE, and MAE as the criteria for evaluating the algorithm, and the algorithm with the best performance for predicting in-compost maturity indicators (C / N, GI, NO3 + -N, NH4 2 -N) and out-of-compost gas indicators (NH3, N2O, CO2, CH4) was selected; the performance of the ML model was evaluated with 10 different random states, and the R-squared score (R Figure 3 ), root mean square error (RMSE), and mean absolute error (MAE) were used to compare the prediction accuracy of the prediction models; an optimal prediction model with a prediction performance close to the average was output as the engineering application model, as shown in Figure 4 As shown in the figure. The results showed that during the model comparison process, as shown in Figure 3 in (a)-(d), to enhance the robustness of the model, the average performance of the models (across 10 random states that affect the training and test splits) was compared. An optimal model with a prediction performance close to the average was selected, as shown in Figure 3 in (e)-(f). The results showed that the GB model performed well in predicting the carbon-nitrogen ratio, while the ET model performed well in predicting GI and NO3 --N is 0.77, NH4 + -N is 0.76. In addition, when predicting 4 indicators outside the compost, as Figure 4 shown, the ET model made excellent predictions for the emissions of CH4, NH3 and N2O; while the RF model was superior to other models in predicting CO2 emissions, and R 2 were 0.52, 0.58, 0.33 and 0.48 respectively.

[0053] To ensure the reasonable application of the compost maturity prediction model, the application domain (AD) of the model was established using cosine similarity. When determining the AD, the training set was used as the reference dataset and the test set was used as the exploration set. Cosine similarity was used to calculate the similarity between the data in and the data modeled in the training set. Based on this similarity measure, different parts of the test set were formed, and the data below a certain threshold were the data outside the AD, and the data above a certain threshold were the data inside the AD. For each pre-set cosine similarity, the gradient was set to discard the data outside the AD from the test dataset and recalculate R 2 , RMSE and the amount of discarded data. After discarding the threshold data with the lowest R 2 and the highest RMSE value, ensure that the data outside the applicable range is as small as possible. Those greater than this threshold will be identified as the optimal threshold; the Shapley additive explanation method (SHAP) was used to quantitatively evaluate the contribution of input feature values to the compost maturity index and the prediction accuracy of the optimal prediction model was verified through experiments.

[0054] To further verify the robust prediction performance and practicality of the model, experiments were carried out in real compost scenarios. Compost treatments with different biochar addition rates (CK: 0%, T1: 8%, T2: 15%) were used, and the prediction results were compared with the actual data collected during the compost maturity process, as Figure 5 and Figure 6 shown. The results of experimental verification confirmed the prediction ability of the model, and the prediction ability for the four internal indicators of compost was relatively excellent. Detailed explanation: Pig manure (PM) and wheat straw (WS) were from the livestock farm and farmland of Sichuan Agricultural University base. Under laboratory conditions, bamboo biochar (BB) was prepared by pyrolysis at 550 °C for 2 hours in an N2 atmosphere. WS was chopped and uniformly mixed with PM at a ratio of 1:2 (based on the dry weight of PM), and the C / N ratio was about 25. The treatments with 8% and 15% BB added were named T1 and T2 respectively, and the control group without any additives was named CK. The experiment was carried out in a 10-liter reactor for 51 days. The moisture content was adjusted to about 60.0%. The temperature of the compost pile was measured every day. Samples were collected on days 0, 4, 9, 16, 23, 37, 51.

[0055] Composting is a process in which unstable organic matter decomposes into stable organic matter through various processes. The decomposition process of organic matter in aerobic composting can be specifically divided into three stages: 1. Thermophilic stage. In this stage, the microorganisms in the compost are mainly mesophilic and aerobic species. The most common ones are non-spore-forming bacteria, spore-forming bacteria, and molds. Under aerobic conditions, they vigorously decompose easily decomposable organic substances (such as simple sugars, starches, proteins, etc.), generating a large amount of heat and continuously increasing the compost temperature from about 20°C to 40°C. 2. High-temperature stage. As the temperature rises, thermophilic microorganisms gradually replace mesophilic species and play a dominant role. The temperature continues to rise and generally reaches above 50°C within a few days, entering the high-temperature stage. In the high-temperature stage, thermophilic actinomycetes and thermophilic fungi become the main species. They strongly decompose complex organic substances in the compost (such as cellulose, hemicellulose, and pectin substances, etc.), accumulating heat and raising the compost temperature to 60 - 70°C, and even up to 80°C. 3. Cooling stage. When the high-temperature stage lasts for a certain period of time, most of the cellulose, hemicellulose, and pectin substances have been decomposed, leaving complex components that are difficult to decompose (such as lignin) and newly formed humus. The activity of microorganisms weakens, and the temperature gradually decreases. When the temperature drops below 40°C, mesophilic microorganisms become the dominant species again. The compost temperature drops to slightly higher than the air temperature. At this time, the compost should be compacted to create an anaerobic state, weakening the mineralization of organic matter to facilitate fertilizer preservation.

[0056] At the same time, a large amount of CO2 is produced by the aerobic degradation of organic matter during the high-temperature stage. At the same time, aerobic microorganisms consume a large amount of O2 when degrading OM. Organic substances such as proteins are degraded in the early stage of composting. The high temperature in this stage stimulates the emission of NH3. The initial stage and high-temperature stage of composting are the main periods of ammonia volatilization and also the key periods for controlling ammonia volatilization. CH4 emissions mainly occur during the high-temperature stage of composting, accounting for 89.4% - 98% of the total CH4 emissions during the entire composting process. Therefore, CH4 emissions are mainly concentrated in the heating and high-temperature stages of composting. Larger emissions of N2O occur during the cooling stage of composting. The reason for this pattern is the mineralization of organic nitrogen and the formation of NH4 + -N under high temperature and suitable pH. Conversely, higher NH4 + -N restricts nitrifying bacteria, resulting in insufficient NO3 - -N to produce N2O under anaerobic conditions. Therefore, precise regulation of process indicators for different processes of composting is crucial for achieving rapid composting, controlling greenhouse gas emissions, and promoting more efficient decomposition of compost.

[0057] Genetic algorithms are a series of search algorithms inspired by the theory of natural evolution. By mimicking the processes of natural selection and reproduction, genetic algorithms can provide high-quality solutions to various problems involving search, optimization, and learning. At the same time, they are similar to natural evolution and can thus overcome some of the obstacles encountered by traditional search and optimization algorithms, especially for problems with a large number of parameters and complex mathematical representations.

[0058] The genetic algorithm (GA) and the optimal prediction model are used to dynamically optimize the performance of the composting process to achieve rapid composting. To achieve precise optimization, the performance of the composting process is divided into three stages: the thermophilic stage, the high-temperature stage, and the cooling stage. Based on the corresponding experimental studies, the maximum and minimum values of these performances in each stage are determined to provide a more reasonable range for the GA to optimize the composting process performance.

[0059] Composting is a process in which unstable organic matter decomposes into stable organic matter through different processes. Therefore, for the different processes of composting, namely the thermophilic stage, the high-temperature stage, and the cooling stage, different optimization upper and lower limits are set for the process indicators (process moisture: MC_P, process temperature: TEMP_P, process pH: pH_P, process time: Day_P), so that precise phased control of the whole process can be achieved during the optimization of composting, in order to achieve more efficient and precise rapid composting (as shown in Table 4). During this process, the established optimal prediction model and the genetic algorithm are combined, and the following principles are followed for the optimization of different maturity indicators. Compost pile body indicators: C / N and NH4 + -N shows a decreasing trend during composting (the smaller the better), and GI and NO3 - -N are the larger the better. At the same time, for the compost pile external indicators: CO2, CH4, N2O, NH3, the smaller the emission is also the better. Therefore, during the optimization by the GA genetic algorithm, the optimization of the process indicators will be achieved based on the above principles.

[0060] The specific optimization method is as follows: (1) Initial population generation: By inputting experimental data into the model (the characteristic values include biochar properties, initial compost properties, and composting process indicators), the genetic algorithm randomly generates 50 to 100 different combinations of composting process parameters, including the composting cycle (Day_P), pH value (pH_P), temperature (TEMP_P), and moisture content (MC_P). These initial parameters form the initial population, providing a basis for optimization. (2) Fitness evaluation for each parameter combination: Replace the original process indicators with the generated combinations and then perform model prediction again. The purpose is to make the newly predicted maturity indicators better than the original predicted values (the principle for judging superiority or inferiority: "C / N and NH4 + -N during composting is the smaller the better, GI and NO3 --N is better when it is larger. At the same time, for the external indicators of the compost heap: CO2, CH4, N2O, and NH3, the smaller the emissions, the better.") For fitness evaluation, the higher the fitness value, the better the effect of the new predicted value in optimizing the original predicted value. (3) Individual selection: Re-predicting all combinations of process parameters (initial population) is one iteration. A total of 15 iterations are carried out. In each iteration, select the parameter combination with the highest fitness as the parent generation, and retain these optimal combinations for the next generation of reproduction, so as to continuously optimize the process parameters. (4) Gene exchange: Randomly select a parameter position for crossover among the selected parameter combinations to introduce new parameter combinations, enhance the diversity of the population, and increase the possibility of finding a better solution. (5) Gene mutation: Randomly modify the value of a certain process parameter to increase the diversity of solutions, prevent the algorithm from falling into a local optimal solution, and improve the global search ability. (6) Optimal individual selection: Through continuous optimization of selection, crossover, and mutation, finally output the optimal combination of composting process parameters to maximize the optimization of the current values of eight maturity indicators. Adjust and optimize the process parameters at each stage of compost sampling (0, 4, 9, 16, 23, 37, 51 days) so that the maturity indicators at this stage are at the optimal value, and the composting maturity time can be shortened to the greatest extent and the composting efficiency can be improved during the entire composting period. This optimization scheme combines adaptive selection, mutation, and evolutionary strategies to ensure the efficient optimization of composting process parameters, thereby achieving rapid composting and improving resource utilization efficiency. (Such as Figure 7 )

[0061] By combining the genetic algorithm with actual composting, compared with the control group without adding biochar, adding biochar at a ratio of 15% and using GA resulted in a significant reduction of about 30% in the number of mature days (as shown in Table 5).

[0062] Table 4 Optimization thresholds of composting process indicators at different composting stages

[0063] Table 5 Prediction and reverse regulation results during the implementation of biochar composting without adding biochar (A) and adding biochar at a ratio of 15% (B)

[0064] References: Xu Ziqi; Yan Zhong; Ge Yanjun; Wei Quanyuan; Huang Bo; Optimization of Process Parameters for Mechanical Strengthened Rapid Aerobic Fermentation of Organic Waste. Environmental Engineering. 2022. 40(08), 159 - 163 - 142.

[0065] Wan, X; Li, J; Xie, L; Wei, Z; Wu, J; Tong, Y. W; Wang, X; He, Y; Zhang, J, Machine Learning framework for intelligent prediction of compost maturity towards automation of food waste composting systerm Bioresourcetechnology 2022, 365, 128107。

[0066] Zhou, S; Kong, F; Lu, L; Wang, P; Jiang, Z, Biochar - An effective additive for improving quality and reducing ecological risk of compost: A global meta - analysis。 Science of The Total Environment 2022, 806, 151439。

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for achieving rapid composting based on prediction and reverse regulation, characterized in that, It includes the following steps; Step 1. Conduct a Meta-analysis to comprehensively evaluate eight maturity indicators inside and outside the compost pile, screen the key biochar properties and initial compost properties that affect the final composting, and provide an optimized dataset for establishing an optimal prediction model for machine learning. The indicators inside the compost pile are C / N, GI, NO3 - -N, NH4 + -N, and the indicators outside the compost pile are NH3, N2O, CO2, CH4; Based on the results of the forest plot, analyze the influence relationship between biochar properties, compost properties, and maturity indicators through structural equation modeling. Judge the influence degree of property indicators on maturity indicators and rank them through the standardized total effect of the structural equation model, thereby verifying the reliability of the forest plot results and revealing the importance of key biochar properties and key compost properties, providing optimization indicators for the machine learning model to establish a dataset; When establishing the machine learning model, add compost process indicators, establish a dataset including 8 maturity indicators, select basic machine learning algorithms such as extra trees, random forest, gradient boosting, and eXtreme Gradient Boosting, and construct an aggregated model of five models through Python editing. Group the data records of the same experiment, and then stratify them into a training set and a test set according to a ratio of 8:2; Perform hyperparameter tuning using Bayesian optimization to R 2 , the root mean square error and the mean absolute error are used as the criteria for evaluating the algorithm, and the algorithm with the best performance for predicting the maturity index in the heap and the gas index outside the heap is selected; Set 10 different random states to evaluate the performance of five models, and use R 2 , RMSE and MAE to compare the prediction accuracies of the five models, and take the optimal prediction model as the best prediction model; Step 2. Evaluate the performance of the optimal prediction model in the real scenario and obtain the data of the corresponding 8 indicators. Collect samples on the 0th, 4th, 9th, 16th, 23rd, 37th, and 51st days, including biochar property parameters, initial compost property parameters, compost process indicators, and 8 maturity indicator parameters measured in the experiment at different composting stages; Step 3. Incorporate the genetic algorithm into the optimal prediction model in Step 1 using Python, and preliminarily determine the upper and lower limits of the process indicators for 3 periods. The 3 periods of the composting process are the thermophilic period, the high-temperature period, and the cooling period; Step 4. Feed the data of the 8 maturity indicators obtained in Step 2 into the optimal prediction model integrated with the genetic algorithm to optimize the upper and lower limits of the process indicators in the three periods of composting. The indicators in the compost pile body: C / N and NH4 + -N shows a trend of fluctuating decrease during the composting process, and GI and NO3 - -N is better if it is larger. At the same time, for the indicators outside the compost pile: CO2, CH4, N2O, and NH3, the smaller the emission, the better.

2. The method according to claim 1, characterized in that The Meta-analysis in Step 1 includes: S101. Collect and screen literature on the database by setting keywords, and conduct subgroup grouping significance verification through heterogeneity test, publication bias test, one-way ANOVA combined with post hoc least significant analysis; S102. Explore the influence of biochar properties and initial compost properties on eight compost maturity indicators through a forest plot using Meta software, evaluate and propose the key biochar properties and initial compost properties affecting maturity, calculate the average effect size of each classification subgroup through the experimental group and control group of each indicator and the corresponding standard deviation and sample size, and compare the average effect size of the subgroup with the overall effect size calculated by only considering whether biochar is added or not to judge and recommend the indicators with better classification subgroup effects.

3. The method according to claim 2, characterized in that, The application domain of the model established using cosine similarity. When determining the application domain, the training set is used as the reference data set, and the test set is used as the exploration set. Cosine similarity is used to calculate the similarity between the data in the test set and the data used for modeling in the training set, forming different parts of the test set. Data below a certain threshold is data outside the application domain, and data above a certain threshold is data within the application domain. For each pre-set cosine similarity, set the gradient to discard the data outside the application domain from the test data set and recalculate R 2 , RMSE, and the amount of discarded data, discard the lowest R 2 After the threshold data with the highest RMSE value, ensure that the data outside the applicable range is minimized. Those greater than this threshold will be identified as the optimal threshold. The Shapley additive explanation method is used to quantitatively evaluate the contribution of input feature values to the compost maturity index, and the prediction accuracy of the optimal prediction model is verified through experiments.

4. The method according to claim 2, wherein The screening rules are as follows: (1). The inclusion criteria require original research, and the quantitative research results show that the application of biochar leads to changes in compost maturity; (2). Each study needs to include a control group and an experimental group, consisting of multiple groups for comparison; (3). The selected studies will include biochar characteristics and initial compost properties; (4). Obtain the mean, standard deviation, and the number of control and experimental groups, or digitize from the chart using the GetDate image digitizer; (5). For studies that only provide the standard error, use the formula SD = SE × , calculate SD, where n represents the sample size. For studies that do not provide SD and SE, SD is calculated as 1 / 10 of the mean.

5. The method according to claim 1, wherein The process indicators in Step 3 are: process moisture: MP_P, process temperature: TEMP_P, process pH: pH_P, process time: Day_P.

6. The method according to claim 1, wherein The upper and lower limits of the process indicators for the 3 periods preliminarily determined in Step 3 are as follows: the process time in the thermophilic period is 1 - 7 days, the process temperature is 25 - 50 °C, the process pH is 6.5 - 8.5, and the process moisture is 40% - 65%; the process time in the high-temperature period is 7 - 15 days, the process temperature is 50 - 70 °C, the process pH is 6.5 - 8.5, and the process moisture is 45% - 55%; the process time in the cooling period is more than 15 days, the process temperature is 40 - 60 °C, the process pH is 6.5 - 8.5, and the process moisture is 40% - 55%.

7. The method according to claim 1, wherein , Step 4 The genetic algorithm for optimizing the composting process includes: (1). Initial population generation: Input the experimental data obtained in Step 2 into the best prediction model. The eigenvalue includes biochar, initial compost properties, and composting process indicators. The genetic algorithm randomly generates 50 to 100 different combinations of composting process parameters, including composting cycle, pH value, temperature, and moisture content. The initial parameters form the initial population; (2). Fitness evaluation for each parameter combination: Replace the original process indicators with the generated process parameter combinations and perform model prediction again. The new predicted maturity index is better than the original predicted value; (3). Individual selection: Completing the re-prediction of all process parameter combinations is one round of iteration. A total of 15 iterations are performed. In each iteration during all iterative processes, select the parameter combination with the highest fitness as the parent generation each time, and retain these optimal combinations for the next generation inversion to continuously optimize the process parameters; (4). Gene exchange: Randomly select a parameter position for crossover among the selected parameter combinations; (5). Gene mutation: Randomly modify the value of a certain process parameter to increase the diversity of solutions; (6). Optimal individual selection: Through continuous optimization of selection, crossover, and mutation, finally output the optimal combination of composting process parameters.

Images