A Wetland Ecological Restoration Method Combining Water Ecological Restoration Monitoring

Abstract

This invention discloses a wetland ecological restoration method combining water ecological restoration monitoring, belonging to the field of ecological restoration technology. The invention collects water depth, nitrogen and phosphorus concentrations, dissolved oxygen, and biodiversity indices in the deep water, shallow water, and riparian zones of the wetland using a sensor network to generate an initial dataset. It then uses a K-means clustering algorithm to divide the wetland into regions and constructs a classification regional feature model. Based on the nitrogen and phosphorus concentrations and water depths of each region, plant species are selected to simulate water flow corridors and optimize pollutant removal efficiency. The riparian zone data is integrated, and a genetic algorithm is used to optimize the interception efficiency of exogenous pollution to obtain the optimal matching parameters. A collaborative ecological restoration framework is generated by iteratively adjusting plant parameters, and a comprehensive numerical evaluation model is established. Based on the evaluation results, a management dataset is generated to achieve quantitative optimization of the wetland ecological restoration effect. This invention can effectively improve pollutant removal efficiency and ensure the collaborative restoration of ecological functions.

Description

Technical Field

[0001] This invention relates to the field of ecological environment restoration and monitoring technology, specifically a wetland ecological restoration method that combines water ecological restoration and monitoring. Background Technology

[0002] Water purification and ecological restoration of wetland ecosystems are crucial areas of environmental governance, involving the coordinated management of multi-gradient zones such as deep water areas, shallow water areas, and riparian zones. The core challenge lies in achieving a balance between efficient pollutant removal and biodiversity restoration. Currently, the main technical challenge in wetland restoration is how to comprehensively consider the dynamic coupling effects of water depth, nitrogen and phosphorus concentrations, water flow velocity, and biodiversity indices within multi-gradient zones to achieve efficient pollutant removal and long-term stability of ecological functions. Existing methods often focus on planting vegetation or deploying artificial facilities in single areas, lacking comprehensive control over inter-regional water flow dynamics and pollutant exchange rates, making it difficult to simultaneously optimize purification efficiency and ecological restoration effects.

[0003] The invention patent with publication number CN119498161A discloses a wetland ecological restoration technology. The wetland ecological restoration technology includes the following steps: Step 1), mixing *Syntrophus esculenta*, *Flavobacterium*, *Pseudomonas saccharophilus*, nutrient substrate, and water, and culturing for 7-9 days to obtain a culture solution; Step 2), cutting reed rhizomes into sections and soaking them in the culture solution for 1-2 days to obtain pretreated rhizomes; Step 3), inserting the pretreated rhizomes into the sewage-covered soil for planting, allowing the reeds to grow and purify the sewage.

[0004] Patent CN110163423A discloses a river ecological restoration system, including a river ecological data analysis system and an ecological implementation system. The analysis method of the river ecological data analysis system specifically includes the following steps: S1, water balance calculation; S2, seepage calculation; S3, separation and reasoning of the impacts of climate change and human activities on water resources. This invention relates to the field of watershed resource protection and utilization technology. This river ecological restoration system compares and analyzes the degree of watershed ecological water protection before and after comprehensive watershed management, as well as the changes in natural vegetation cover and soil drought characteristics. It quantitatively assesses the effectiveness of comprehensive watershed management models in ensuring ecological water supply under the background of climate change. By constructing a quantitative relationship model between river runoff, seepage, and standard tree-ring chronology, it proposes a new approach to reconstruct river seepage using tree-ring chronology, providing a basis for optimizing water resource management in similar watersheds in arid regions.

[0005] The invention patent with publication number CN104628140A discloses a comprehensive water purification method based on shallow and deep-water plants and riparian vegetation. This method addresses eutrophication by using artificial floating beds in deep water, planting aquatic purification plants in shallow water, and planting vegetation along the riparian shore to absorb nutrients. However, this method fails to adequately consider the impact of differences in water flow velocity between regions on pollutant removal efficiency. It also lacks a basis for dynamically adjusting plant species and density, making it difficult to adapt to complex scenarios with high nitrogen and phosphorus concentrations in deep water or varying water depths in shallow water. Furthermore, it does not address the quantitative assessment of biodiversity indices or the synergistic optimization of ecological functions between regions, thus limiting the stability and durability of the restoration effect. When nitrogen and phosphorus concentrations are high in deep water, plant configuration is difficult to adapt to water depth and flow characteristics, resulting in low pollutant removal efficiency. In shallow water and riparian zones, differences in water flow velocity mean that the selection of plant species and density lacks a dynamic adjustment basis, affecting pollutant interception. Simultaneously, the insufficient matching between pollutant exchange rates and water flow velocities between regions makes it difficult to assess the long-term stability of the biodiversity index using a single indicator. Summary of the Invention

[0006] The purpose of this invention is to provide a wetland ecological restoration method that combines water ecological restoration monitoring, which can effectively improve pollutant removal efficiency and ensure the coordinated restoration of ecological functions.

[0007] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0008] A wetland ecological restoration method combining water ecological restoration monitoring includes the following steps:

[0009] Step 1: Collect water depth, nitrogen and phosphorus concentrations, dissolved oxygen, and biodiversity index from deep water areas, shallow water areas, and riparian zones of the wetland using a sensor network to generate an initial ecological and pollution dataset;

[0010] Step 2: Based on the initial ecological and pollution dataset, the K-means clustering algorithm is used to divide the wetland into deep water zone, shallow water zone and riparian zone according to the water depth, nitrogen and phosphorus concentration and biodiversity index, and a classification regional feature model is generated.

[0011] Step 3: Extract nitrogen and phosphorus concentration and water depth data for each region from the classification region feature model, and configure the plant species data for each region based on the nitrogen and phosphorus concentration and water depth to generate a regional plant configuration dataset;

[0012] Step 4: Based on the regional plant configuration dataset, simulate the water flow corridor between the deep water area and the shallow water area, calculate the water flow velocity Vw and the nitrogen and phosphorus removal efficiency Enp. If the Enp is lower than the preset threshold Et, adjust the plant density data to generate an optimized water flow and pollutant removal model.

[0013] Step 5: Based on the optimized water flow and pollutant removal model, integrate the riparian vegetation coverage and soil adsorption rate data, use a genetic algorithm to optimize the external pollution interception efficiency, calculate the best matching parameters between the water flow velocity Vw and the interception efficiency Ei, and generate a set of pollution interception and water flow regulation parameters.

[0014] Step 6: Extract the inter-regional water flow velocity Vw and pollutant exchange rate Re from the pollution interception and water flow regulation parameter set. If the Re is lower than the preset threshold Rt or the Ei is lower than the preset threshold, iteratively adjust the plant species and density data to generate a collaborative ecological restoration framework.

[0015] Step 7: Based on the aforementioned collaborative ecological restoration framework, establish a comprehensive numerical assessment model covering water quality indicators, plant growth parameters, and biodiversity indices. By periodically monitoring key parameters of each gradient zone, a quantitative evaluation of the restoration effect is formed, and the output data of the assessment model is obtained.

[0016] Step 8: Based on the output data of the evaluation model, generate plant planting and management datasets for each region. Verify the balance between water flow velocity Vw, nitrogen and phosphorus removal efficiency Enp, and biodiversity index Bd through long-term simulation. If Bd is lower than the preset threshold Bt, adjust the management dataset to generate the final wetland restoration dataset. Specifically, Bt is the critical threshold for the biodiversity index, with a value of 0.6, set based on the average Shannon-Wiener index of biodiversity in historical healthy wetlands in the study area. If it is lower than this value, the management strategy needs to be adjusted to improve biodiversity.

[0017] Furthermore, the specific steps of step 1 are as follows:

[0018] S101: Acquire the initial dataset through a sensor network; if the initial dataset is missing or abnormal, fill in the missing values ​​using linear interpolation.

[0019] S102: Based on the complete dataset, the K-means clustering algorithm is used to perform partitioned clustering of the data in the deep water area, shallow water area and coastal zone to determine the regional feature distribution;

[0020] S103: Principal component analysis was used to extract the main contributing factors of water depth, nitrogen and phosphorus concentrations, dissolved oxygen and biodiversity index, and key environmental factors were obtained.

[0021] If the factor value exceeds a preset threshold, the trend of factor change is predicted by time series analysis; the wetland ecosystem and pollution status are classified by decision tree algorithm to determine the ecological health level and pollution degree; and a dynamic environmental monitoring report is generated.

[0022] Linear interpolation was used to fill in missing or outlier data, ensuring the integrity and accuracy of the initial dataset and providing a reliable data foundation for subsequent analysis. K-means clustering was employed for partitioned clustering, enabling more accurate determination of regional characteristic distributions. Principal component analysis extracted key environmental factors, simplifying data dimensions and highlighting the main factors affecting wetland ecology. When factor values ​​exceeded preset thresholds, time series analysis was used to predict trends, facilitating proactive countermeasures. Decision tree algorithms were used to classify ecosystems and pollution states, quickly determining ecological health levels and pollution levels. The generated dynamic environmental monitoring reports provided timely and effective evidence for wetland restoration decisions, addressing issues of low data quality, unclear key factors, and delayed ecological status assessments.

[0023] Furthermore, step 2 includes:

[0024] S201: Merge the initial ecological and pollution datasets into a unified dataset to obtain a standardized dataset;

[0025] S202: Use the K-means clustering algorithm to determine the initial wetland zoning; calculate the mean and standard deviation of water depth data in each zoning; if they exceed the preset threshold, readjust the clustering parameters to optimize the zoning.

[0026] S203: Principal component analysis algorithm is used to reduce the dimensionality of nitrogen and phosphorus concentrations and biodiversity index to obtain the partition feature vector;

[0027] S204: Calculate the distance between the feature vectors of the partitions. If it is less than a preset threshold, merge the adjacent partitions.

[0028] S205: Construct a regional classification feature model and generate classification model parameters; and use the support vector machine algorithm to optimize the classification boundary.

[0029] Step 2 refines the wetland zoning process, making it more scientific and rational. Merging and standardizing datasets eliminates dimensional differences between data points, improving the accuracy of cluster analysis. Calculating the mean and standard deviation of water depth data and adjusting clustering parameters optimizes the initial zoning results. Principal component analysis (PCA) is used to reduce the dimensionality of nitrogen and phosphorus concentrations and biodiversity indices, resulting in zoning feature vectors that better reflect the essential characteristics of the regions. Calculating the distance between zoning feature vectors and merging adjacent zonings avoids overly detailed or unreasonable zoning. Constructing a regional classification feature model and using a support vector machine (SVM) algorithm to optimize classification boundaries improves the accuracy of regional classification, solving the problems of inaccurate wetland zoning and lack of prominent regional characteristics. This provides a precise basis for subsequent steps such as plant configuration.

[0030] Furthermore, in step 3, the plant species data for each region is configured, including:

[0031] If the nitrogen and phosphorus concentration in the deep water area is higher than the preset threshold Nt=5mg / L, then configure the data on submerged plant species.

[0032] If the nitrogen and phosphorus concentration in the deep water area is lower than the Nt, then configure the floating plant species data;

[0033] The shallow water area and the riparian zone are configured with emergent plant species data according to the water depth;

[0034] The plant configuration subsets are integrated using a data fusion algorithm to generate the regional plant configuration dataset; and the priority ranking of plant configurations in each region is determined using a cluster analysis algorithm.

[0035] The rules for plant species configuration in each area were clearly defined. Submerged or floating plants were configured according to the different nitrogen and phosphorus concentrations in deep water areas, making plant species selection more targeted and improving the efficiency of nitrogen and phosphorus removal. Emergent plants were configured in shallow water areas and riparian zones according to water depth, adapting to the environmental characteristics of different areas. A data fusion algorithm was used to integrate plant configuration subsets, ensuring data consistency and integrity; a cluster analysis algorithm determined the priority ranking of plant configurations in each area, enabling reasonable resource allocation and prioritizing the plant configuration problems of key areas. This solved the problems of blind plant species selection and unreasonable configuration, improving the effectiveness of phytoremediation.

[0036] Furthermore, in step 4, the plant density data is adjusted to generate an optimized water flow and pollutant removal model, including:

[0037] S401: Calculate the water flow velocity Vw using a fluid dynamics model;

[0038] S402: The nitrogen and phosphorus removal efficiency Enp is calculated using a pollutant removal model; if the nitrogen and phosphorus removal efficiency Enp is lower than the preset threshold Et=70%, the plant density in the deep water area and the shallow water area is adjusted using a linear regression algorithm;

[0039] S403: Update water flow gallery parameters using hydrological simulation algorithms;

[0040] S404: Recalculate the nitrogen and phosphorus removal efficiency Enp using the pollutant removal model; and generate the optimized water flow and pollutant removal model using data fusion technology.

[0041] By calculating water flow velocity and nitrogen and phosphorus removal efficiency using fluid dynamics and pollutant removal models respectively, a scientific basis for adjusting plant density was provided. When nitrogen and phosphorus removal efficiency fell below a preset threshold, a linear regression algorithm was used to adjust plant density, accurately finding suitable density parameters to improve nitrogen and phosphorus removal efficiency. Hydrological simulation algorithms updated water flow corridor parameters, ensuring the rationality of the water flow corridor design. Recalculating nitrogen and phosphorus removal efficiency and using data fusion technology to generate an optimized model ensured the model's accuracy and reliability, resolving the mismatch between water flow velocity and nitrogen and phosphorus removal efficiency, as well as the blind adjustment of plant density, thus improving the wetland's water purification capacity.

[0042] Furthermore, in step 5, a set of pollution interception and water flow regulation parameters is generated, including:

[0043] S501: Obtain data on riparian vegetation cover and soil adsorption rate;

[0044] S502: Construct an initial model for pollution interception efficiency; if the pollution interception efficiency of the initial model is lower than a preset threshold, then use a genetic algorithm to optimize the vegetation coverage and soil adsorption rate parameters.

[0045] S503: Calculate the matching relationship between the water flow velocity and the pollution interception efficiency, and generate water flow velocity adjustment parameters;

[0046] S504: Simulate a water flow regulation scheme using water flow velocity adjustment parameters; if the water flow velocity adjustment value does not match the pollutant concentration in the environmental data, rerun the genetic algorithm.

[0047] By acquiring data on riparian vegetation cover and soil adsorption rate, foundational data was provided for constructing a pollution interception efficiency model. When the initial model efficiency fell below a preset threshold, a genetic algorithm was used to optimize the parameters, effectively improving the interception efficiency of exogenous pollution and avoiding the problem of the genetic algorithm getting trapped in local optima. The matching relationship between water flow velocity and pollution interception efficiency was calculated, and adjustment parameters were generated, achieving synergistic optimization of water flow velocity and pollution interception efficiency. Simulating water flow adjustment schemes and re-running the genetic algorithm when mismatches occurred ensured the rationality of the parameter set, solving the problems of unstable exogenous pollution interception efficiency and mismatch between water flow velocity and interception efficiency in the riparian zone, and enhancing the wetland's ability to control exogenous pollution.

[0048] Furthermore, in step 6, the plant species and density data are iteratively adjusted to generate a collaborative ecological restoration framework, including:

[0049] S601: If the pollutant exchange rate Re is lower than the preset threshold Rt=0.1m / s or the interception efficiency Ei is lower than 60%, then a suitable plant species will be queried from the plant species database to determine a list of candidate plant species.

[0050] S602: The random forest algorithm is used to predict the impact of different combinations of plant species and densities on the pollutant exchange rate Re, and the optimized plant species and density configuration is obtained.

[0051] S603: The water flow velocity adjustment value between regions is calculated using a numerical simulation method; if the water flow velocity adjustment parameter set does not match the pollutant exchange rate Re, the plant density data is adjusted iteratively.

[0052] S604: The gradient descent algorithm is used to optimize the matching relationship between the water flow velocity Vw and the pollutant exchange rate Re, thereby generating the collaborative ecological restoration framework.

[0053] When pollutant exchange rates or interception efficiencies are insufficient, suitable plant species are queried from a plant species database, providing a rich selection for plant species adjustment. The random forest algorithm predicts the impact of different plant species and density combinations on pollutant exchange rates, quickly identifying optimal configuration schemes. Numerical simulation methods calculate water flow velocity adjustment values ​​and iteratively adjust plant density when mismatches occur, ensuring coordination between water flow velocity and pollutant exchange rates. The gradient descent algorithm optimizes the matching relationship between the two, generating a collaborative ecological restoration framework. This addresses issues such as poor inter-regional pollutant exchange, low interception efficiency, and mismatch between plant configuration and pollutant migration, achieving collaborative restoration across different wetland areas.

[0054] Furthermore, in step 7, a comprehensive numerical evaluation model is established to form a quantitative evaluation of the recovery effect, including:

[0055] Step 701: Obtain monitoring data for each gradient partition;

[0056] Step 702: Normalize dissolved oxygen, nitrogen and phosphorus content, plant coverage, and root development status using standardized treatment methods;

[0057] Step 703: Use principal component analysis algorithm to extract the main features of water quality indicators, plant growth parameters and biodiversity index; if the variance contribution rate of the feature vector set is greater than a preset threshold, calculate the comprehensive evaluation score by weighted linear combination;

[0058] Step 704: Use the K-means clustering algorithm to classify each gradient partition to obtain the partition recovery level distribution; if there are low-level partitions in the partition recovery level distribution, use the regression analysis algorithm to predict the changing trends of dissolved oxygen and nitrogen and phosphorus content in the low-level partitions; generate optimization parameters for the recovery effect of each gradient partition.

[0059] By acquiring and normalizing monitoring data from each gradient zone, the dimensional differences between various parameters are eliminated, facilitating comprehensive evaluation. Principal component analysis extracts key features, highlighting crucial factors influencing restoration effectiveness. When the variance contribution rate of the feature vector set exceeds a preset threshold, a weighted linear combination is used to calculate the comprehensive evaluation score, fully reflecting the wetland's restoration status. K-means clustering is employed to classify the restoration levels of each zone, providing a clear understanding of their restoration progress. For low-level zones, the changing trends of dissolved oxygen and nitrogen / phosphorus content are predicted, generating optimized parameters that provide a basis for targeted restoration. This addresses the limitations of the comprehensive evaluation model's prediction accuracy and the lack of quantification in evaluating restoration effectiveness, improving the accuracy and scientific rigor of wetland restoration assessment.

[0060] Furthermore, in step 8, the management dataset is adjusted to generate the final wetland restoration dataset, including:

[0061] Step 801: Obtain plant planting and management datasets for each region;

[0062] Step 802: Normalize the plant species, planting density, and management frequency using standardized processing methods;

[0063] Step 803: Use principal component analysis algorithm to extract the main features of plant planting and management; if the variance contribution rate of the feature vector set is greater than a preset threshold, calculate the comprehensive score of plant management in each region by weighted linear combination;

[0064] Step 804: Use the K-means clustering algorithm to classify each region and obtain the regional plant management level distribution; if there are low-level regions in the regional plant management level distribution, use the linear regression analysis algorithm to predict the changing trends of water flow velocity Vw and nitrogen and phosphorus removal efficiency Enp in the low-level regions, and adjust the plant planting density and management frequency parameters of the low-level regions.

[0065] Standardizing the plant planting and management datasets facilitates data analysis and comparison. Principal component analysis (PCA) extracts key features, identifying crucial factors influencing plant management effectiveness. Weighted linear combination calculations yield comprehensive scores, enabling a complete evaluation of plant management levels across regions. K-means clustering is employed to classify regional plant management levels, identifying areas with weak management practices. For low-level areas, trends in water flow velocity and nitrogen and phosphorus removal efficiency are predicted, and planting density and management frequency parameters are adjusted accordingly. This improves the targetedness and effectiveness of plant management, addresses issues of low biodiversity indices and inadequate plant management, and ensures the long-term stable development of wetland ecosystems.

[0066] Compared with the prior art, the present invention has the following beneficial effects:

[0067] This invention achieves systematic and synergistic wetland ecological restoration. It comprehensively collects data through a sensor network, providing a foundation for subsequent analysis. Then, it divides the area into regions to ensure the targeted nature of restoration measures. Plant species are configured according to regional characteristics to achieve an initial layout for ecological restoration. By simulating and optimizing water flow and pollutant removal models, nitrogen and phosphorus removal efficiency is improved. The efficiency of external pollution interception is optimized, enhancing the wetland's pollution control capabilities. Iterative adjustments to plant species and density ensure synergistic effects between regions. More importantly, a comprehensive evaluation model is established to achieve quantitative evaluation of restoration effects. Finally, a management dataset is generated and dynamically adjusted to ensure the long-term stability of the wetland ecosystem. This invention solves the problem of coordinating multiple regions and parameters in wetlands, achieving synergistic optimization of water purification, pollutant interception, and biodiversity enhancement, thus improving the efficiency and stability of wetland ecological restoration. Attached Figure Description

[0068] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0069] Figure 1 This is a schematic diagram of the overall process of the present invention. Detailed Implementation

[0070] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the embodiments of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive. The following description is in conjunction with the accompanying drawings. Figure 1 The embodiments of the present invention will be described in detail below.

[0071] Example 1:

[0072] This embodiment discloses a wetland ecological restoration method that combines water ecological restoration monitoring. The specific implementation process is as follows:

[0073] First, a sensor network is deployed in the wetland environment to collect key environmental data in deep water areas, shallow water areas, and the riparian zone. The sensor network includes multiple distributed sensor nodes, each of which is connected to the central data processing unit through a wireless communication module to ensure the stability of real-time data transmission.

[0074] The sensor network collects data on water depth, nitrogen and phosphorus concentrations, dissolved oxygen levels, and biodiversity index parameters, and stores this data as an initial ecological and pollution dataset.

[0075] In practical applications, if data is missing or exhibits abnormal fluctuations, linear interpolation is used to fill in the missing data. Simultaneously, principal component analysis is employed to extract key environmental factors and generate time-series prediction models, thereby eliminating the impact of incomplete data on subsequent analysis. Sensor networks, as the core component of data acquisition, achieve comprehensive coverage of environmental data across multiple regions through their high-density deployment.

[0076] After data collection was completed, the initial ecological and pollution datasets were partitioned based on the K-means clustering algorithm, dividing the wetland into three main areas: deep water zone, shallow water zone, and riparian zone.

[0077] The specific steps for partitioning are to first calculate the mean and standard deviation of water depth, nitrogen and phosphorus concentration, and biodiversity index in each region. If the standard deviation of a certain region exceeds the preset threshold, the clustering parameters are readjusted until the conditions are met.

[0078] Subsequently, a support vector machine (SVM) algorithm was used to optimize the classification boundaries, ensuring the accuracy of the partitioning results. Through the combined action of K-means clustering and SVM algorithms, a classification region feature model was generated by comprehensively analyzing data from multiple dimensions. This model not only includes key feature parameters for each region but also labels the transition boundaries between different regions, providing a scientific basis for subsequent plant configuration and water flow control.

[0079] Based on the classification region feature model, nitrogen and phosphorus concentration and water depth data of each region are further extracted, and regional plant configuration datasets are generated accordingly.

[0080] For deep water areas, if the nitrogen and phosphorus concentration is higher than the preset threshold Nt=5mg / L, submerged plant species will be prioritized; if it is lower than the threshold, floating plant species will be selected.

[0081] In shallow water areas and riparian zones, emergent plant species are selected based on water depth data. The selection of plant species integrates subset information through data fusion algorithms and uses cluster analysis algorithms to determine the priority ranking of plant configuration in each area.

[0082] In this process: a list of suitable plant species from the plant database is retrieved, and the random forest algorithm is used to predict the impact of different combinations of plant species and densities on the pollutant exchange rate, thereby generating an optimized plant species and density configuration scheme. The random forest algorithm and the gradient descent algorithm work together to optimize the matching relationship between water flow velocity regulation parameters and pollutant exchange rates, ensuring effective control of pollutant diffusion between regions.

[0083] To improve the efficiency of external pollution interception, an initial pollution interception efficiency model was constructed. This model integrated data on riparian vegetation cover and soil adsorption rate, and optimized these parameters using a genetic algorithm. When calculating the relationship between water flow velocity and pollution interception efficiency, if the initial model's efficiency was lower than a preset threshold, the genetic algorithm was re-run to adjust the relevant parameters until the target value was reached. The generation of water flow velocity adjustment parameters relied on the simulation of the relationship between water flow velocity and pollution interception efficiency. If the simulation results did not match the pollutant concentrations in the actual environmental data, the genetic algorithm was iteratively run again to optimize the parameters.

[0084] After optimizing plant configuration and pollution interception, a comprehensive numerical assessment model was established to quantify wetland restoration effects. This model encompasses multi-dimensional data including water quality indicators, plant growth parameters, and biodiversity indices. Dissolved oxygen, nitrogen and phosphorus content, plant cover, and root development were normalized using standardized processing methods. Principal component analysis (PCA) was used to extract key features. If the variance contribution rate of the feature vector set exceeded a preset threshold, a weighted linear combination was used to calculate the comprehensive assessment score. K-means clustering was employed to classify each gradient partition, generating a partition restoration level distribution. If low-level partitions existed within the restoration level distribution, regression analysis was used to predict the trends in dissolved oxygen and nitrogen and phosphorus content in these low-level partitions, generating optimization parameters for restoration effects. PCA and K-means clustering worked together in the multi-indicator quantitative assessment process to ensure the scientific rigor and accuracy of the restoration effect evaluation.

[0085] Based on the output data of the comprehensive numerical assessment model, plant planting and management datasets for each region were generated. Long-term simulations were used to verify the balance of water flow velocity (Vw), nitrogen and phosphorus removal efficiency (Enp), and biodiversity index (Bd) among regions. If the biodiversity index (Bd) was lower than a preset threshold (Bt), the management dataset was adjusted to generate the final wetland restoration dataset. Plant species, planting density, and management frequency were normalized using standardization methods. Principal component analysis was used to extract the main characteristics of plant planting and management, and K-means clustering was used to classify each region to generate a regional plant management level distribution. If low-level regions existed in the regional plant management level distribution, linear regression analysis was used to predict the changing trends of water flow velocity (Vw) and nitrogen and phosphorus removal efficiency (Enp) in these low-level regions, and the corresponding planting density and management frequency parameters were adjusted accordingly.

[0086] To enable those skilled in the art to fully understand and implement this invention, the specific implementation principles of this invention are further supplemented below in conjunction with specific application scenarios.

[0087] This embodiment focuses on an urban wetland park with an area of ​​approximately 20 hm², where ecological restoration is being carried out. The wetland suffers from problems such as excessive nitrogen and phosphorus pollution and declining biodiversity. The ecological restoration is achieved through the method of this invention.

[0088] Step 1: Initial Ecological and Pollution Data Collection:

[0089] S101: Deploy a monitoring network consisting of 50 sensors in the wetland, including 15 in deep water (depth > 1.5m), 20 in shallow water (0.5-1.5m), and 15 in the riparian zone (< 0.5m), and continuously collect water depth (H), nitrogen concentration (TN), phosphorus concentration (TP), dissolved oxygen (DO), and biodiversity index (Bd) for 30 days.

[0090] Bd is calculated using the Shannon-Wiener exponent, and the formula is as follows:

[0091] ;

[0092] Where S represents the total number of species. This represents the proportion of individuals of the i-th species to the total number of individuals, with a value ranging from 0 to 3. A higher value indicates richer biodiversity.

[0093] Data from some sensors was missing on day 10. Linear interpolation was used to fill in the missing data, using the following formula:

[0094] ;

[0095] in For missing values, , For adjacent known values, , and For example, the TN data of sensor A in deep water:

[0096] The concentration was 6.2 mg / L on day 9 and 5.8 mg / L on day 11. The missing data for day 10 was calculated as (6.2 + 5.8) / 2 = 6.0 mg / L.

[0097] S102: After imputation, the complete dataset is obtained. K-means clustering (K=3) is used to partition and cluster the data, with the objective function being to minimize the sum of squares within each cluster.

[0098] ;

[0099] in, Let K be the objective function, and K be the number of clusters (K=3 here). For the k-th cluster, As the cluster center, The distance is Euclidean. The characteristic distribution of TN (total nitrogen) was obtained for the deep water area (7.2 mg / L), the shallow water area (4.5 mg / L), and the coastal zone (3.1 mg / L).

[0100] S103: Principal component analysis was used to extract key environmental factors and calculate the contribution rate of each indicator: water depth (28%), TN (32%), DO (21%), and Bd (19%). The cumulative contribution rate reached 100%, and they were identified as key factors.

[0101] In the deep water area, the total nitrogen (TN) exceeded the preset threshold of 5 mg / L. The TN change trend over the next three months was predicted using ARIMA(p,d,q) time series analysis. The core formula is:

[0102] ;

[0103] in For time series data (such as TN concentration). For a d-order difference operator, These are the autoregressive coefficients. The moving average coefficient is... It is white noise. This is a model constant term used to correct the baseline level of the time series. , ,…, is the autoregressive coefficient, where p is the autoregressive order, reflecting the weight of historical data on the current value; , ,…, is the moving average coefficient, where q is the moving average order, used to correct for the impact of model prediction errors.

[0104] The decision tree algorithm is used to classify the ecological health level, determining that the deep water area is "poor", the shallow water area is "average" and the coastal zone is "good", and a dynamic monitoring report is generated.

[0105] Step 2: Generation of classification region feature model. The classification region feature model is a dataset describing the characteristics of wetland zones, including the mean water depth, nitrogen and phosphorus concentration range, biodiversity index interval and classification boundary parameters of each zone, which is used to guide the regional differentiated design of subsequent plant configuration and water flow regulation.

[0106] S201: Standardize the initial dataset (mean 0, standard deviation 1) using the Z-score standardization formula:

[0107] ;

[0108] in This is the original data. The mean of the dataset. denoted as , and z as the standardized value. For example, the standardized TN value for deep water is (7.2 - 5.2) / 1.8 = 1.11, where 5.2 is the total mean and 1.8 is the total standard deviation.

[0109] S202: The K-means clustering algorithm (iterations = 50, error threshold = 0.01) is used to determine the initial partitions. The water mean of each partition is calculated as follows: deep water 1.8m (standard deviation 0.2m), shallow water 1.0m (standard deviation 0.3m), and coastal zone 0.3m (standard deviation 0.1m). None of them exceed the preset threshold (standard deviation < 0.5m), so no adjustment of the clustering parameters is required.

[0110] S203: Dimensionality reduction of nitrogen and phosphorus concentrations and biodiversity index yielded the following zonal feature vectors: deep water zone (0.82, 0.35, -0.21), shallow water zone (0.41, 0.22, 0.15), and coastal zone (0.12, -0.18, 0.63) (corresponding to the principal components of TN, TP, and Bd, respectively).

[0111] S204: Calculate the feature vector distance. The distance between the deep water area and the shallow water area is 0.53, and the distance between the shallow water area and the shoreline is 0.72. Both are greater than the preset threshold of 0.3, so there is no need to merge the partitions.

[0112] S205: Construct a regional classification feature model. The model parameters include cluster center coordinates and boundary thresholds. Use the support vector machine algorithm to optimize the classification boundary and obtain the final partitioning model.

[0113] Step 3: Generation of Regional Plant Configuration Dataset

[0114] Plant species were selected based on nitrogen and phosphorus concentrations and water depth data for each region:

[0115] The average TN value in the deep water area is 7.2 mg / L > Nt = 5 mg / L. The submerged plant species are: Hydrilla verticillata (coverage 30%) and Potamogeton crispus (coverage 20%). Nt is the critical threshold for nitrogen concentration in the deep water area (unit: mg / L), which is set based on the nitrogen concentration limit for Class V water bodies in the "Surface Water Environmental Quality Standard" (GB3838-2002) and is used to distinguish the configuration conditions of submerged plants and floating plants.

[0116] The shallow water area has a depth of 1.0m, and the emergent plant species are configured as follows: reeds (40% coverage) and cattails (30% coverage).

[0117] The riparian zone has a water depth of 0.3m, and the following emergent plant species are configured: sweet flag (35% coverage) and iris (25% coverage).

[0118] The above configurations are integrated using a data fusion algorithm, employing a weighted average fusion formula:

[0119] ;

[0120] in, Assign a comprehensive value to the plants in region j. Let i be the weight of plant i in region j. Assign a suitability score (0-100) to plant i. Generate a regional plant configuration dataset and determine priorities through cluster analysis: deep water area *Hydrilla verticillata* > *Potamogeton crispus*, shallow water area *Phragmites australis* > *Typha orientalis*, riparian zone *Acorus calamus* > *Iris*.

[0121] Step 4: Optimize water flow and pollutant removal model generation

[0122] S401: Calculation of water flow velocity using Manning's formula:

[0123] ;

[0124] in, The velocity of the water flow is (m / s). The roughness coefficient, The hydraulic radius is (m). This represents the hydraulic gradient. With an initial plant density of n=0.03, R=0.8m, and S=0.001, we obtain Vw=1 / 0.03×0.8 (2 / 3) ×0.001 (1 / 2) =0.21m / s.

[0125] S402: Nitrogen and phosphorus removal efficiency was calculated using a plant absorption kinetics model.

[0126] ;

[0127] in, Nitrogen and phosphorus removal efficiency (%) The plant absorption coefficient is 0.05. Plant density (plants / m²). The duration of stay (in days). The saturation coefficient is 0.02. Initial calculations show Enp = 0.05 × 20 × 3 × (1 - e^(-0.02 × 20)) = 30% < Et = 70%. A linear regression algorithm is used to adjust the plant density; the formula is:

[0128] ;

[0129] To adjust the plant density, For the initial density, To adjust the amount, The target threshold for nitrogen and phosphorus removal efficiency (70%) is set based on experimental data of typical purification efficiency in wetland ecological restoration projects. If the value is lower than this, the plant density needs to be adjusted to improve the purification capacity. For current efficiency, The efficiency density coefficient is 2% / plant·m². The calculated ΔC = 20 plants / m², and the adjusted density is 40 plants / m².

[0130] S403: Update the water flow corridor parameters. After increasing the plant density, the roughness n=0.04, and the calculated Vw=0.16m / s.

[0131] S404: Recalculate Enp = 0.05 × 50 × 3 × (1 - e^(-0.02 × 50)) = 75% ≥ 70%, and use data fusion technology to generate an optimized model.

[0132] Step 5: Generation of Pollution Interception and Water Flow Regulation Parameter Set

[0133] S501: Obtain the initial vegetation cover of the riparian zone as 60% and the soil adsorption rate as 0.4 kg / (m²・a). S502: Construct the initial interception model and optimize it using a genetic algorithm. The objective function is:

[0134] ;

[0135] For better interception efficiency, For vegetation coverage, For soil adsorption rate, and The weights are (0.3, 0.5).

[0136] Constraints: 0.5≤cov≤1.0, 0.3≤ads≤0.8. After optimization, the coverage rate is 90% and the adsorption rate is 0.65kg / (m²・a). At this time, Ei=0.3×0.9+0.5×0.65=0.605≥0.6.

[0137] S503: Calculate the matching relationship between Vw and Ei to obtain the water flow velocity adjustment parameters: Ei is optimal when Vw = 0.15 m / s, and generate the adjustment parameter set.

[0138] S504: Simulates the water flow regulation scheme to verify that the pollutant concentration matches the regulation parameters when Vw=0.15m / s, without needing to rerun the algorithm.

[0139] Step 6: Generation of the collaborative ecological restoration framework. The collaborative ecological restoration framework is a comprehensive solution that integrates plant configuration, water flow regulation, and pollution interception. It includes inter-regional plant species-density matching rules, collaborative optimization objectives for water flow velocity and pollutant exchange rate, and dynamic adjustment mechanisms to achieve the coordinated restoration of ecological functions in different wetland zones.

[0140] S601: Calculate the pollutant exchange rate Re=0.08m / s<Rt=0.1m / s, and search for candidate species from the plant database: Myriophyllum spicatum (submerged) and Thalia dealbata (emergent).

[0141] S602: The random forest algorithm (100 decision trees) is used to predict the combined effects. The regression-based random forest prediction formula is as follows:

[0142] ;

[0143] in, For the predicted pollutant exchange rate Re, The number of decision trees is 100. Let m be the predicted value for the m-th tree. The vector represents the plant species and density characteristics. When *Hydrilla verticillata* + *Myriophyllum spicatum* (density 50 + 20 plants / m²), Re = 0.12 m / s, which meets the requirements.

[0144] S603: Calculate the water flow velocity adjustment value to 0.14m / s, matching Re=0.12m / s.

[0145] S604: The gradient descent algorithm is used to optimize the matching relationship. The parameter update formula is as follows:

[0146] ;

[0147] Where θ is the parameter to be optimized (water flow velocity, plant density), and η is the learning rate (0.01). Let the gradient of the objective function be the objective function gradient.

[0148] In this objective function, Indicates the actual water flow velocity between regions (unit: m / s); This indicates the target flow velocity (unit: m / s), which is the optimal flow velocity that matches the pollutant exchange rate. Re represents the optimized target pollutant exchange rate (unit: m / s), i.e., the preset threshold Rt=0.1m / s, which is set based on the needs of pollutant diffusion and ecological connectivity between regions to ensure smooth material exchange in each wetland zone; Re is the pollutant exchange rate.

[0149] After 10 iterations, the optimal matching parameters are obtained, and a collaborative ecological restoration framework is generated.

[0150] Step 7: Establishment of a comprehensive numerical evaluation model

[0151] Step 701: Obtain monitoring data for each zone: Deep water zone DO=5.2mg / L, TN=4.8mg / L, vegetation coverage 75%; Shallow water zone DO=6.5mg / L, TN=3.2mg / L, coverage 80%; Coastal zone DO=7.8mg / L, TN=2.5mg / L, coverage 85%.

[0152] Step 702: Standardize the data (range 0-1), for example, the TN standardized value in deep water = (5-4.8) / (7.2-2.5) = 0.04.

[0153] Step 703: Principal component analysis is used to extract features. The variance contribution rate is 85% > 80%. A weighted linear combination is used to calculate the comprehensive score.

[0154] ;

[0155] Where S represents the overall score. The weighting of the indicators is as follows: water quality 40%, plants 30%, biodiversity 30%. The scores are the standardized indicators. The calculated scores are 72 for deep water, 81 for shallow water, and 89 for the coastal zone.

[0156] Step 704: K-means clustering yields the recovery level: "Medium" for deep water, "Good" for shallow water, and "Excellent" for coastal zone. It is predicted that the DO in deep water will rise to 5.8 mg / L in the next 3 months, and optimized parameters are generated.

[0157] Step 8: Generation of the final wetland restoration dataset

[0158] Step 801: Obtain plant planting management data: In deep water, the planting density of black algae is 50 plants / m², and it is harvested once a month; in shallow water, the density of reeds is 40 plants / m², and it is harvested once every 2 months.

[0159] Step 802: Standardize the management data, density standardization value = (50-30) / (60-20) = 0.5.

[0160] Step 803: Principal component analysis feature contribution rate 88% > 80%, calculate management comprehensive score:

[0161] ;

[0162] in, For management comprehensive scoring, Rate the plant's growth status. The suitability of management measures was scored. The scores were 75 for deep water, 83 for shallow water, and 90 for the riparian zone. Step 804: Clustering was used to determine the management level. The deep water area was rated "medium," and its Vw was predicted to increase to 0.16 m / s, and Enp to 78%. The planting density in the deep water area was adjusted to 55 plants / m², and the harvesting frequency was changed to once every 20 days. The final repair dataset was then generated.

[0163] In this embodiment, steps 3 and 4 employ a weighted average method for data fusion, with weights dynamically allocated based on plant suitability scores and regional pollution levels. Step 3 utilizes a hierarchical clustering algorithm to determine the priority ranking of plant configurations in each region based on two dimensions: plant purification efficiency and environmental adaptability. Steps 4 and 8 employ a univariate linear regression model, with plant density as the independent variable and nitrogen and phosphorus removal efficiency as the dependent variable (step 4); and management frequency as the independent variable and water flow velocity / nitrogen and phosphorus removal efficiency as the dependent variable (step 8), fitting the regression equation using the least squares method.

[0164] This invention significantly improves the rationality of wetland zoning and vegetation configuration, especially enhancing its ability to meet the differentiated needs of deep-water areas, shallow-water areas, and riparian zones. The combination of genetic algorithms and random forest algorithms achieves global optimization of exogenous pollution interception efficiency, effectively suppressing pollutant diffusion and improving water purification effects. The comprehensive numerical evaluation model quantifies restoration effects through multiple indicators, balancing short-term purification efficiency with long-term ecological stability, providing a scientific basis and technical support for wetland ecological restoration.

[0165] This invention breaks through the limitations of traditional wetland zoning methods that rely on experience or single indicators. It innovatively achieves precise zoning through dynamic monitoring of the entire area using a sensor network and multi-algorithm fusion analysis. The sensor network simultaneously collects data on water depth, nitrogen and phosphorus concentrations, dissolved oxygen, and biodiversity indices in deep water, shallow water, and the riparian zone, generating a complete initial ecological and pollution dataset. This solves the problems of incomplete and fragmented data collection in traditional methods. By combining K-means clustering (based on multiple parameters including water depth, nitrogen and phosphorus concentration, and biodiversity index) and support vector machine algorithms to optimize classification boundaries, a classification regional feature model is generated. This ensures that the zoning results accurately reflect the ecological and pollution characteristics of each region, providing a scientific basis for subsequent differentiated restoration.

[0166] Meanwhile, addressing the issues of blind plant configuration and insufficient adaptability in existing technologies, an innovative mechanism for precise matching of environmental parameters and plant functions is proposed. In deep water areas, plant types are dynamically selected based on nitrogen and phosphorus concentration thresholds (Nt=5mg / L): submerged plants are configured for high nitrogen and phosphorus concentrations to enhance pollutant absorption, while floating plants are configured for low concentrations to optimize the ecological niche. In shallow water areas and riparian zones, emergent plants are configured based on water depth differences. The plant configuration subsets are integrated through data fusion algorithms, and priorities are determined through cluster analysis, achieving precise matching between plant species and regional environmental characteristics, thereby improving pollutant removal efficiency and ecological adaptability.

[0167] Furthermore, by integrating hydrological simulation and ecological restoration parameters, a collaborative optimization model for water flow, pollutants, and vegetation is constructed to address the mismatch between water flow and pollution removal efficiency across regions in existing technologies. Based on a fluid dynamics model simulating water flow corridors in deep and shallow water areas, the flow velocity (Vw) and nitrogen and phosphorus removal efficiency (Enp) are calculated. When Enp falls below a threshold (Et=70%), a linear regression algorithm dynamically adjusts plant density to optimize the water flow and pollutant removal model. For riparian pollution, a genetic algorithm is used to optimize vegetation cover and soil adsorption rate parameters, calculate the optimal match between flow velocity and interception efficiency Ei, and generate a set of pollution interception and water flow regulation parameters, achieving collaborative optimization of external pollution control and hydrological conditions.

[0168] More importantly, this invention breaks through the limitations of single-area restoration, innovatively achieving synergistic restoration of the entire wetland through dynamic regulation of inter-regional material exchange. It extracts the inter-regional water flow velocity (Vw) and pollutant exchange rate (Re). When Re is below a threshold (Rt=0.1m / s) or Ei is not up to standard, candidate plants are screened using a plant species database. The impact of plant species density combinations on Re is predicted using a random forest algorithm, and parameters are iteratively adjusted. A gradient descent algorithm is employed to optimize the matching relationship between water flow velocity and pollutant exchange rate, generating a synergistic ecological restoration framework. This ensures smooth material exchange between deep water areas, shallow water areas, and the riparian zone, addressing the problem of fragmented ecological functions between regions. Simultaneously, a comprehensive assessment system covering water quality, plants, and biodiversity is constructed, forming a closed-loop management system encompassing monitoring, assessment, and adjustment throughout the entire lifecycle. This addresses the issues of insufficient quantification and long-term stability in evaluating the restoration effects of existing technologies. A comprehensive numerical assessment model was established. Principal component analysis was used to extract water quality indicators such as dissolved oxygen, nitrogen and phosphorus, plant growth parameters such as coverage and root development, and core features of the biodiversity index. A weighted linear combination was used to calculate the comprehensive score and generate the zonal restoration level. Based on the assessment results, a plant planting and management dataset was generated. Long-term simulation was used to verify the balance between water flow velocity, nitrogen and phosphorus removal efficiency and biodiversity index (Bd). When Bd is lower than the threshold (Bt=0.6), management strategies such as planting density and management frequency were dynamically adjusted to ensure the long-term stability of wetland ecological functions.

[0169] This invention organically integrates sensor monitoring, K-means clustering, random forest, genetic algorithm, hydrological simulation, and ecological restoration technologies, forming a comprehensive technical system encompassing data acquisition, zonal restoration, coordinated regulation, and effect evaluation. Compared to existing technologies that focus on single aspects, such as planting plants or monitoring a single indicator, this invention achieves a systematic breakthrough in data-driven, precise restoration, dynamic optimization, and quantitative assurance, effectively improving pollutant removal efficiency and the ability to synergistically restore ecological functions.

[0170] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0171] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A wetland ecological restoration method combining water ecological restoration monitoring, characterized in that, Specifically, the steps include the following: Step 1: Collect water depth, nitrogen and phosphorus concentrations, dissolved oxygen, and biodiversity index from deep water areas, shallow water areas, and riparian zones of the wetland using a sensor network to generate an initial ecological and pollution dataset; Step 2: Based on the initial ecological and pollution dataset, the K-means clustering algorithm is used to divide the wetland into deep water zone, shallow water zone and riparian zone according to the water depth, nitrogen and phosphorus concentration and biodiversity index, and a classification regional feature model is generated. Step 3: Extract nitrogen and phosphorus concentration and water depth data for each region from the classification region feature model, and configure the plant species data for each region based on the nitrogen and phosphorus concentration and water depth to generate a regional plant configuration dataset; Step 4: Based on the regional plant configuration dataset, simulate the water flow corridor between the deep water area and the shallow water area, calculate the water flow velocity Vw and the nitrogen and phosphorus removal efficiency Enp. If the Enp is lower than the preset threshold Et, adjust the plant density data to generate an optimized water flow and pollutant removal model. Step 5: Based on the optimized water flow and pollutant removal model, integrate the riparian vegetation coverage and soil adsorption rate data, use a genetic algorithm to optimize the external pollution interception efficiency, calculate the best matching parameters between the water flow velocity Vw and the interception efficiency Ei, and generate a set of pollution interception and water flow regulation parameters. Step 6: Extract the inter-regional water flow velocity Vw and pollutant exchange rate Re from the pollution interception and water flow regulation parameter set. If the Re is lower than the preset threshold Rt or the Ei is lower than the preset threshold, iteratively adjust the plant species and density data to generate a collaborative ecological restoration framework. Step 7: Based on the aforementioned collaborative ecological restoration framework, establish a comprehensive numerical assessment model covering water quality indicators, plant growth parameters, and biodiversity indices. By periodically monitoring key parameters of each gradient zone, a quantitative evaluation of the restoration effect is formed, and the output data of the assessment model is obtained. Step 8: Based on the output data of the evaluation model, generate plant planting and management datasets for each region. Verify the balance between water flow velocity Vw, nitrogen and phosphorus removal efficiency Enp, and biodiversity index Bd between regions through long-term simulation. If the biodiversity index Bd is lower than the preset threshold Bt, adjust the management dataset to generate the final wetland restoration dataset.

2. The wetland ecological restoration method combining water ecological restoration monitoring according to claim 1, characterized in that, Step 1: The specific steps are as follows: S101: Acquire the initial dataset through a sensor network; if the initial dataset is missing or abnormal, fill in the missing values ​​using linear interpolation. S102: Based on the complete dataset, the K-means clustering algorithm is used to perform partitioned clustering of the data in the deep water area, shallow water area and coastal zone to determine the regional feature distribution; S103: Principal component analysis was used to extract the main contributing factors of water depth, nitrogen and phosphorus concentrations, dissolved oxygen and biodiversity index, and key environmental factors were obtained. If the factor value exceeds a preset threshold, the trend of factor change is predicted by time series analysis; the wetland ecosystem and pollution status are classified by decision tree algorithm to determine the ecological health level and pollution degree; and a dynamic environmental monitoring report is generated.

3. The wetland ecological restoration method combining water ecological restoration monitoring according to claim 1, characterized in that, Step 2 includes: S201: Merge the initial ecological and pollution datasets into a unified dataset to obtain a standardized dataset; S202: Use the K-means clustering algorithm to determine the initial wetland zoning; calculate the mean and standard deviation of water depth data in each zoning; if they exceed the preset threshold, readjust the clustering parameters to optimize the zoning. S203: Principal component analysis algorithm is used to reduce the dimensionality of nitrogen and phosphorus concentrations and biodiversity index to obtain the partition feature vector; S204: Calculate the distance between the feature vectors of the partitions. If it is less than a preset threshold, merge the adjacent partitions. S205: Construct a regional classification feature model and generate classification model parameters; and use the support vector machine algorithm to optimize the classification boundary.

4. The wetland ecological restoration method combining water ecological restoration monitoring according to claim 1, characterized in that, In step 3, the plant species data for each region is configured, including: If the nitrogen and phosphorus concentration in the deep water area is higher than the preset threshold Nt, then configure the submerged plant species data; If the nitrogen and phosphorus concentration in the deep water area is lower than the preset threshold Nt, then the floating plant species data are configured. The shallow water area and the riparian zone are configured with emergent plant species data according to the water depth; The plant configuration subsets are integrated using a data fusion algorithm to generate the regional plant configuration dataset; and the priority ranking of plant configurations in each region is determined using a cluster analysis algorithm.

5. The wetland ecological restoration method combining water ecological restoration monitoring according to claim 1, characterized in that, In step 4, the plant density data is adjusted to generate an optimized water flow and pollutant removal model, including: S401: Calculate the water flow velocity Vw using a fluid dynamics model; S402: The nitrogen and phosphorus removal efficiency Enp is calculated using a pollutant removal model; if the nitrogen and phosphorus removal efficiency Enp is lower than a preset threshold Et, the plant density in the deep water area and the shallow water area is adjusted using a linear regression algorithm; S403: Update the flow corridor parameters using a hydrological simulation algorithm; S404: Recalculate the nitrogen and phosphorus removal efficiency Enp using the pollutant removal model; and generate the optimized water flow and pollutant removal model using data fusion technology.

6. The wetland ecological restoration method combining water ecological restoration monitoring according to claim 1, characterized in that, In step 5, a set of parameters for pollution interception and water flow regulation is generated, including: S501: Obtain data on riparian vegetation cover and soil adsorption rate; S502: Construct an initial model for pollution interception efficiency; if the pollution interception efficiency of the initial model is lower than a preset threshold, then use a genetic algorithm to optimize the vegetation coverage and soil adsorption rate parameters. S503: Calculate the matching relationship between the water flow velocity and the pollution interception efficiency, and generate water flow velocity adjustment parameters; S504: Simulate a water flow regulation scheme using water flow velocity adjustment parameters; if the water flow velocity adjustment value does not match the pollutant concentration in the environmental data, rerun the genetic algorithm.

7. The wetland ecological restoration method combining water ecological restoration monitoring according to claim 1, characterized in that, In step 6, the plant species and density data are iteratively adjusted to generate a collaborative ecological restoration framework, including: S601: If the pollutant exchange rate Re is lower than the preset threshold Rt or the interception efficiency Ei is lower than 60%, then a suitable plant species will be queried from the plant species database to determine a list of candidate plant species. S602: The random forest algorithm is used to predict the impact of different combinations of plant species and densities on the pollutant exchange rate Re, and the optimized plant species and density configuration is obtained. S603: The water flow velocity adjustment value between regions is calculated using a numerical simulation method; if the water flow velocity adjustment parameter set does not match the pollutant exchange rate Re, the plant density data is adjusted iteratively. S604: The gradient descent algorithm is used to optimize the matching relationship between the water flow velocity Vw and the pollutant exchange rate Re, thereby generating the collaborative ecological restoration framework.

8. The wetland ecological restoration method combining water ecological restoration monitoring according to claim 1, characterized in that, In step 8, the management dataset is adjusted to generate the final wetland restoration dataset, including: Step 801: Obtain plant planting and management datasets for each region; Step 802: Normalize the plant species, planting density, and management frequency using standardized processing methods; Step 803: Use principal component analysis to extract the main features of plant planting and management; if the variance contribution rate of the feature vector set is greater than the preset threshold, calculate the comprehensive score of plant management in each region by weighted linear combination; Step 804: Use the K-means clustering algorithm to classify each region and obtain the regional plant management level distribution; if there are low-level regions in the regional plant management level distribution, use the linear regression analysis algorithm to predict the changing trends of water flow velocity Vw and nitrogen and phosphorus removal efficiency Enp in the low-level regions, and adjust the plant planting density and management frequency parameters of the low-level regions.

Images