Optimization method for environmental management in whole process of water transportation engineering construction

Abstract

The invention relates to the technical field of water transportation engineering environment management and data processing technologies, and discloses an optimization method for water transportation engineering construction whole-process environment management, which comprises the following steps: collecting real-time data and historical data of water transportation engineering construction whole-process environment data through a sensor, and carrying out data preprocessing on the real-time data and the historical data; storing the preprocessed real-time data and historical data in a distributed database, and respectively performing classification and tagging management to form first category feature data and second category feature data; constructing an environment change prediction model based on a long short-term memory network, and performing learning training of environment change prediction on the second type of feature data; and importing the first type of feature data into the trained environment change prediction model, optimizing the parameter precision, and predicting the environment change trend. According to the method, the real-time performance of environment monitoring, the intelligent level of data processing and the accuracy of prediction are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of water transport engineering environmental management and data processing technology, and in particular to an optimization method for environmental management of the entire process of water transport engineering construction. Background Art

[0002] In recent years, with the rapid development of China's economy and the acceleration of globalization, water transport engineering construction activities have increased significantly, such as port expansion, channel dredging, dock construction and river management. While promoting economic development, these projects have had a profound impact on the water environment ecosystem. The construction process of water transport projects usually involves large-scale earth excavation, dredging and reclamation activities, which lead to water pollution, ecological habitat destruction and changes in hydrodynamic characteristics.

[0003] Traditional environmental management methods mainly rely on manual on-site monitoring and regular sampling, which has many shortcomings. First, the frequency of manual sampling is limited, making it difficult to achieve continuous monitoring of environmental parameters, resulting in the lack of real-time data on environmental changes; environmental factors often have a high degree of spatiotemporal variability, and instantaneous changes may lead to serious environmental events that cannot be captured in time through regular sampling; secondly, the laboratory analysis cycle is long, and it may take days or even weeks from sampling to obtaining results, which cannot meet the needs of rapid response to environmental anomalies; in addition, the manual monitoring process consumes a lot of manpower and material resources, is costly, and is not conducive to large-scale, long-term environmental monitoring.

[0004] As the scale of water transport projects expands, the amount and complexity of environmental data are also increasing dramatically. Traditional environmental management methods are difficult to handle such a huge amount of data, especially in the collection, transmission, storage and analysis of data, where there are technical bottlenecks; existing data management systems usually lack systematization and automation, and the data format is not unified and the storage is scattered, making it difficult to conduct comprehensive analysis; this leads to the inability to guarantee the timeliness and accuracy of environmental data, which directly affects the scientific nature of environmental management decisions.

[0005] In terms of data analysis, traditional methods mainly rely on simple statistical analysis and empirical judgment, which makes it difficult to capture the complex nonlinear relationship between environmental parameters. The environmental system is highly dynamic and complex, and is affected by many factors, such as meteorological conditions, hydrodynamic conditions, and human interference. Simple statistical methods cannot accurately predict environmental change trends and cannot meet the needs of modern environmental management for early warning and prediction. This limits managers' early identification and response to environmental risks, leading to the occurrence of environmental accidents.

[0006] In addition, existing technologies are insufficient in the mining and utilization of environmental data. A large amount of environmental data has not been fully analyzed and interpreted, and potential information and patterns have not been mined. This not only wastes valuable data resources, but also limits the improvement of environmental management levels. Summary of the invention

[0007] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract of the specification and the title of the invention of this application to avoid blurring the purpose of this section, the abstract of the specification and the title of the invention, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0008] In view of the above existing problems, the present invention is proposed.

[0009] Therefore, the technical problem solved by the present invention is: the deficiencies of the prior art in real-time monitoring of environmental data, data processing and environmental prediction.

[0010] In order to solve the above technical problems, the present invention provides the following technical solutions: collecting real-time data and historical data of environmental data of the whole process of water transport engineering construction through sensors, and performing data preprocessing on them;

[0011] The pre-processed real-time data and the historical data are stored in a distributed database, and the real-time data and the historical data are classified and labeled to form first-category feature data and second-category feature data;

[0012] Building an environmental change prediction model based on a long short-term memory network, and performing learning and training on the environmental change prediction for the second category of feature data;

[0013] The first category feature data is then imported into the environmental change prediction model after training to optimize parameter accuracy and predict environmental change trends.

[0014] As a preferred solution of the optimization method for environmental management of the whole process of water transport engineering construction described in the present invention, the environmental data collected during the whole process of water transport engineering construction at least includes hydrodynamic environmental parameters, water quality parameters, biological indicators and soil bottom parameters.

[0015] As a preferred solution of the optimization method for environmental management of the entire process of water transport engineering construction described in the present invention, the data preprocessing converts data from different sources into data in a unified format by performing data cleaning and standardization on the real-time data and the historical data.

[0016] As a preferred solution of the optimization method for environmental management of the entire process of water transport engineering construction described in the present invention, the real-time data and the historical data are classified and labeled respectively, including:

[0017] Dividing the real-time data and the historical data into first physical data, second physical data, first chemical data, second chemical data, first biological data, second biological data, and first soil data and second soil data according to the attributes and time domain of the real-time data and the historical data;

[0018] Based on a set threshold, adding status labels to the first physical data, the second physical data, the first chemical data, the second chemical data, the first biological data, the second biological data, the first soil data, and the second soil data;

[0019] The status labels include normal, warning, and danger.

[0020] As a preferred solution of the optimization method for environmental management of the entire process of water transport engineering construction described in the present invention, the first physical data, the second physical data, the first chemical data, the second chemical data, the first biological data, the second biological data and the first soil data, the second soil data are converted into numerical features through label coding to obtain first category feature data and second category feature data, respectively.

[0021] As a preferred solution of the optimization method for environmental management of the entire process of water transport engineering construction described in the present invention, the first category feature data is data generated by classifying and labeling the real-time data, and the first category feature data includes numerical features converted from the first physical data, the first chemical data, the first biological data, and the first soil data;

[0022] The second category feature data is data generated by classifying and labeling the historical data, and the second category feature data includes numerical features converted from the second physical data, the second chemical data, the second biological data and the second soil data.

[0023] As a preferred solution of the optimization method for environmental management of the whole process of water transport engineering construction described in the present invention, an environmental change prediction model is constructed based on a long short-term memory network to capture the time series characteristics of environmental data. The environmental change prediction model uses a cross entropy loss function as an objective function and uses an adaptive estimation optimization algorithm to adjust the model parameters to improve the convergence speed and prediction accuracy of the model. Its mathematical expression is as follows:

[0024]

[0025] Among them, θ t+1 is the model parameter of the t+1th iteration, θ t is the model parameter of the tth iteration, η is the learning rate, which controls the step size of parameter update, is the loss function with respect to the parameter θ t The gradient of

[0026] In the learning training, the learning rate η=0.001, the number of hidden layer units is 64, and the number of training rounds is 100.

[0027] As a preferred solution of the optimization method for environmental management of the whole process of water transport engineering construction described in the present invention, using the optimized environmental change prediction model to predict the environmental change trend includes:

[0028] Define the input as real-time data x t , using the optimized model parameters θ * , calculate the hidden state h through forward propagation t and output P t , its mathematical expression formula is as follows:

[0029] h t =LSTM(x t ,h t-1 θ * )

[0030]

[0031] Among them, P t is the predicted value of the environmental state at the tth time step, indicating the probability that the current environment is abnormal. is the weight matrix of the output layer, is the bias vector of the output layer, σ is the Sigmoid activation function, x t is the input data of the tth time step, h t-1 is the hidden state at the previous time step.

[0032] As a preferred solution of the optimization method for environmental management of the whole process of water transport engineering construction described in the present invention, by outputting the predicted value of the environmental state P t Make a judgment and get the environmental status:

[0033] When P≥0.8, it means that there is a significant change or abnormality in the environmental state, which is a dangerous state and requires immediate response measures;

[0034] When 0.5≤P<0.8, it indicates that the environmental status has a trend of change, which is a warning state and requires enhanced monitoring;

[0035] When P < 0.5, it means that the environmental state is stable, with no obvious trend of change, and is in a normal state.

[0036] Beneficial effects of the invention: The method of the invention solves the problems of poor real-time performance, insufficient data processing capability and low prediction accuracy of traditional environmental management methods through sensor technology, data processing method and deep learning algorithm. It not only optimizes the environmental management process of water transport engineering construction, but also significantly improves the real-time performance of environmental monitoring, the intelligent level of data processing and the accuracy of prediction, achieving the beneficial effects of improving the real-time performance of environmental monitoring, data management efficiency and the accuracy of environmental prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0038] Figure 1 It is a flow chart of the optimization method for environmental management of the whole process of water transport engineering construction shown in the present invention;

[0039] Figure 2 The calculation domain and grid distribution diagram of the sea area engineering shown in the present invention;

[0040] Figure 3 It is a schematic diagram of the flow velocity change of the local sea area before and after the water transport project shown in the present invention during the rapid rise and high tide period;

[0041] Figure 4 It is a schematic diagram of the flow velocity changes in the local sea area before and after the water transport project shown in the present invention during the ebb and flow period and the high tide period. DETAILED DESCRIPTION

[0042] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0043] Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without making any creative work should fall within the scope of protection of the present invention.

[0044] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0045] In view of the deficiencies in the above-mentioned prior art, an embodiment of the present invention proposes an optimization method for environmental management of the entire process of water transport engineering construction. The method deploys a multi-type sensor network at the water transport engineering construction site to collect real-time environmental data including hydrodynamic environmental parameters (water level, flow rate), water quality parameters (pH value, dissolved oxygen, pollutant concentration), biological indicators (plankton species and quantity) and soil bottom parameters (particle size distribution, organic matter content), and preprocesses these real-time data with historical data to eliminate the influence of outliers and missing values, ensure data quality, and provide accurate and effective environmental data quality and reliability for subsequent data analysis and model construction.

[0046] It should be noted that, in terms of data processing, the embodiment of the present invention adopts a classification and labeling management method, which divides the data into physical data, chemical data, biological data and soil data according to the attributes and time domain of the data, and further divides the data into real-time data and historical data. Based on the set threshold, a status label is added to each category of data, such as normal, warning and danger. Through label encoding, the data is converted into numerical features to form first category feature data and second category feature data, thereby realizing efficient storage and management of a large amount of environmental data; among them, the distributed database improves the scalability and access speed of data storage, and the classification and labeling management makes the data structured, facilitates subsequent data calls and processing, and improves data management efficiency and availability.

[0047] Furthermore, in terms of data analysis and prediction, the embodiment of the present invention constructs an environmental change prediction model through a long short-term memory network (LSTM). LSTM can capture the time series characteristics of environmental data and accurately learn and predict environmental change trends. The LSTM model can effectively learn complex nonlinear relationships and improve the accuracy of environmental change predictions. The model uses the cross entropy loss function as the objective function and uses the adaptive moment estimation (Adam) optimization algorithm to adjust the model parameters, thereby improving the convergence speed and prediction accuracy of the model. By inputting the first category feature data into the trained model, real-time prediction and early warning of the environmental status are achieved.

[0048] According to an embodiment of the present invention, Figure 1 The flowchart shown is an optimization method for environmental management of the entire process of water transport engineering construction, which specifically includes the following steps:

[0049] S1. Collect real-time and historical data of the environmental data of the entire process of water transport engineering construction through sensors, and perform data preprocessing. Among them, what needs to be explained in this step is:

[0050] Arrange a multi-type sensor network at the water transport project construction site to collect real-time data and historical data of environmental data throughout the water transport project construction process, and the multi-type sensor network includes at least an ultrasonic water level meter, a Doppler flow meter, a water quality monitor, an automatic sampler, a microscope and a soil analyzer;

[0051] The environmental data collected during the entire process of water transport engineering construction shall at least include hydrodynamic environmental parameters (water level, flow velocity, flow direction), water quality parameters (pH value, suspended sediment content, harmful chemical content), biological indicators (plankton species and quantity), and soil bottom parameters (soil elevation, soil particle size, organic matter content).

[0052] In an optional implementation, the environmental data is classified according to its attributes and divided into physical data, chemical data, biological data and soil data.

[0053] As an example, the physical data at least include hydrodynamic environment parameters (water level, flow velocity, flow direction).

[0054] As an example, the chemical data include at least water quality parameters (pH value, suspended sediment content, content of harmful chemicals).

[0055] As an example, the biological data includes at least biological indicators (plankton species and quantities).

[0056] As an example, soil data at least includes soil substrate parameters (soil elevation, soil particle size, organic matter content).

[0057] It should be noted that data preprocessing converts data from different sources into data in a unified format by cleaning and standardizing real-time data and historical data.

[0058] As an example, the box plot method is used to identify outliers. For each parameter, its quartiles Q1 and Q3 are calculated, and the upper limit is set to Q3+1.5×IQR, and the lower limit is set to Q1-1.5×IQR, where IQR=Q3-Q1 is the interquartile range. Data exceeding the upper and lower limits are regarded as outliers; outliers are replaced by the average of adjacent legitimate data, or interpolated and supplemented according to historical trends.

[0059] As an example, for missing values ​​in a time series, linear interpolation is used to find the missing values ​​based on the adjacent known data points (x i-1 ,y i-1 ) and (x i+1 ,y i+1 ), calculate the missing point x i Values:

[0060]

[0061] Among them, (x i-1 ,y i-1 ) is the independent variable and dependent variable of the known data point before the missing value, (x i+1 ,y i+1 ) is the independent variable and dependent variable of the known data point after the missing value, x i is the independent variable corresponding to the missing value, y i is the dependent variable with missing values ​​calculated by interpolation.

[0062] As an example, converting data of different dimensions to the interval [0,1], the mathematical expression formula is as follows:

[0063]

[0064] Among them, X is the original data, X min and X max are the minimum and maximum values ​​of the original data, respectively, and X′ is the standardized data.

[0065] It should be noted that the parameters collected in this embodiment cover multiple fields such as physics, chemistry, biology and soil, which can fully reflect the environmental conditions, ensure the comprehensiveness and systematicness of environmental monitoring, and provide rich data support for environmental impact assessment and management. Abnormal data and noise are eliminated through data cleaning, and the problems of dimensional differences and scale inconsistency are eliminated through standardization processing. This allows data from different sensors and sources to be fused together, providing high-quality data input for model training, and ultimately achieving the beneficial effect of improving data quality and model training effect.

[0066] S2. Store the pre-processed real-time data and historical data in a distributed database, classify and label the real-time data and historical data, respectively, to form first-category feature data and second-category feature data. It should be noted that this step classifies and labels the real-time data and historical data, respectively, including:

[0067] Store pre-processed real-time data and historical data in a distributed database to ensure high availability and scalability of data;

[0068] According to the attributes and time domain of the real-time data and the historical data, the real-time data and the historical data are divided into first physical data, second physical data, first chemical data, second chemical data, first biological data, second biological data, and first soil data, second soil data;

[0069] Based on the set threshold, adding status labels to the first physical data, the second physical data, the first chemical data, the second chemical data, the first biological data, the second biological data, the first soil data, and the second soil data;

[0070] Among them, the status labels include normal, warning, and danger;

[0071] The first physical data, the second physical data, the first chemical data, the second chemical data, the first biological data, the second biological data, the first soil data, and the second soil data are converted into numerical features by label encoding to obtain first category feature data and second category feature data respectively;

[0072] The first category feature data is data generated by classifying and labeling real-time data, and the first category feature data includes numerical features converted from first physical data, first chemical data, first biological data, and first soil data;

[0073] The second category feature data is data generated by classifying and labeling historical data. The second category feature data includes numerical features converted from second physical data, second chemical data, second biological data, and second soil data.

[0074] As an example, a status label is added to each category of data. The status labels include "normal" (0), "warning" (1), and "dangerous" (2). The specific judgment conditions are as follows:

[0075] If data X≤threshold T 正常 , then the label is 0;

[0076] If the threshold T 正常 <Data X≤Threshold T 警告 , then the label is 1;

[0077] If data X> threshold T 警告 , then the label is 2;

[0078] Add status labels to each category of data. The labels and judgment conditions are as follows:

[0079] Normal (0): The parameter value is within the safe range and no action is required;

[0080] Warning (1): The parameter value is close to the critical value and needs to be observed;

[0081] Danger (2): The parameter value is outside the safe range and immediate action is required.

[0082] As an example, the embodiment of the present invention uses One-Hot Encoding to convert the state label into a numerical feature, where:

[0083] Normal (0): 1, 0, 01, 0, 01, 0, 0;

[0084] Warning (1): 0,1,00,1,00,1,0;

[0085] Danger (2): 0, 0, 10, 0, 10, 0, 1;

[0086] Through encoding, first-category feature data (feature vectors of real-time data) and second-category feature data (feature vectors of historical data) are formed.

[0087] It should be noted that the embodiments of the present invention are conducive to capturing the characteristics and changing laws of different environmental factors through attribute and time domain classification, providing conditions for model learning, training and prediction, and improving the model prediction accuracy and sensitivity to environmental changes.

[0088] Preferably, the addition of status labels makes the environmental data indicative, facilitates the identification of potential environmental risks, and enhances the environmental risk identification and early warning capabilities. The use of label encoding to convert classification information into numerical inputs acceptable to the model enables the model to make full use of the category information in the data, improves the effectiveness and predictive ability of model training, and enhances the model's learning ability and accuracy.

[0089] S3, build an environmental change prediction model based on the long short-term memory network, perform learning and training on the second category feature data for environmental change prediction, and then import the first category feature data into the trained environmental change prediction model to optimize parameter accuracy. Among them, what needs to be explained in this step is:

[0090] An environmental change prediction model is constructed based on the long short-term memory network to capture the time series characteristics of environmental data. The model architecture of the environmental change prediction model is:

[0091] Input layer: receives input data with feature dimension n, where n is the number of features, including numerical features of physical, chemical, biological and soil data;

[0092] Hidden layer: set a number of LSTM units, the number of hidden layer neurons is h, such as h = 64;

[0093] Output layer: one neuron, outputting the predicted value of the environment state P t , the value range is [0,1].

[0094] Specifically, the learning and training includes:

[0095] The second category feature data (historical data) is used as the training data set, and the training set and the validation set are divided in a ratio of 8:2;

[0096] Set the input sequence length, for example, use the data from the past T time steps to predict the environment state at the next time step;

[0097] Forward propagation: input historical data sequence x t-T+1 ,x t-T+2 ,…,xt ;

[0098] Calculate the hidden state h through the LSTM unit t ;

[0099] Calculate output prediction value

[0100] Loss calculation: Using the true label y t (based on historical environmental status markers) and predicted values Calculate the loss L;

[0101] As an example, the embodiment of the present invention selects to use the cross entropy loss function, which is suitable for classification problems. Its mathematical expression formula is:

[0102]

[0103] Among them, L is the loss function value, NNN is the number of samples, y i is the true label of the i-th sample, which takes a value of 0 or 1. is the model prediction value of the i-th sample, ranging from [0,1];

[0104] Calculate the gradient of the loss function with respect to the model parameters

[0105] θ t are all the parameters of the model at the tth iteration, including W, U, and b;

[0106] The Adam optimization algorithm is used to update parameters and fine-tune optimization parameters to improve the convergence speed and prediction accuracy of the model. Its mathematical expression is as follows:

[0107]

[0108] Among them, θ t+1 is the model parameter of the t+1th iteration, θ t is the model parameter of the tth iteration, η is the learning rate, which controls the step size of parameter update, is the loss function with respect to the parameter θ t The gradient of

[0109] In the learning training, the learning rate η = 0.001, the number of hidden layer units is 64, and the number of training rounds is 100;

[0110] After training, the weight matrix and bias vector W, U, b of the LSTM model are updated to the optimized parameters, denoted as W * , U * 、b * , the mathematical expression of the model is:

[0111] Input Gate:

[0112]

[0113] Forget Gate:

[0114]

[0115] Output Gate:

[0116]

[0117] Candidate cell states:

[0118]

[0119] Cell status update:

[0120]

[0121] Hide status updates:

[0122] h t =o t *tanh(C t )

[0123] Finally, the output of the model is t Through a fully connected layer and activation function, we get:

[0124]

[0125] Among them, W * is the weight matrix of the output layer, b * is the bias vector of the output layer, x t is the input data of the tth time step, h t-1 is the hidden state of the previous time step, i t 、f t , o t are the activation values ​​of the input gate, forget gate, and output gate, respectively, with a value range of [0,1], C t is the cell state at the current time step, is the candidate cell state, W, U, b are the weight matrix and bias vector to be learned, are the weight matrices input to each gate and candidate state, respectively. are the weight matrices from hidden state to each gate and candidate state, are the bias vectors of each gate and candidate state respectively, and σ is the Sigmoid activation function.

[0126] It should be noted that the embodiment of the present invention constructs an environmental change prediction model based on a long short-term memory network to capture the time series characteristics of environmental data. The model uses the cross entropy loss function as the objective function and adopts an adaptive estimation optimization algorithm to adjust the model parameters, thereby achieving deep learning and effective training of environmental data.

[0127] Preferably, the LSTM network can handle the long-term dependency of time series data, the cross entropy loss function is suitable for classification problems, and the Adam optimization algorithm improves the convergence speed and stability of the model. Through learning and training, the training parameters of the model are optimized, ensuring that the model achieves good training results within a reasonable time and improving the prediction accuracy and generalization ability of the model.

[0128] S4. Predicting the environmental change trend using the environmental change prediction model after parameter tuning. It should be noted that predicting the environmental change trend using the optimized environmental change prediction model includes:

[0129] Define the input as real-time data x t , using the optimized model parameters θ * (i.e., the parameter C used to update the cell state in step S3 above) t ), calculate the hidden state h through forward propagation t and output P t , its mathematical expression formula is as follows:

[0130] h t =LSTM(x t ,h t-1 θ * )

[0131]

[0132] Among them, P t is the predicted value of the environmental state at the tth time step, indicating the probability that the current environment is abnormal. is the weight matrix of the output layer, is the bias vector of the output layer, σ is the Sigmoid activation function, x t is the input data of the tth time step, h t-1 is the hidden state at the previous time step.

[0133] By outputting the predicted value of the environmental state P t Make a judgment and get the environmental status:

[0134] When P≥0.8, it means that there is a significant change or abnormality in the environmental state, which is a dangerous state and requires immediate response measures;

[0135] When 0.5≤P<0.8, it indicates that the environmental status has a trend of change, which is a warning state and requires enhanced monitoring;

[0136] When P < 0.5, it means that the environmental state is stable, with no obvious trend of change, and is in a normal state.

[0137] It should be noted that the embodiment of the present invention utilizes the latest real-time data, combines the trained model parameters, and uses the optimized environmental change prediction model to predict the environmental change trend, thereby achieving real-time prediction of the current environmental status.

[0138] It should also be noted that the embodiments of the present invention utilize sensor technology, data processing methods and deep learning algorithms to solve the problems of poor real-time performance, insufficient data processing capabilities and low prediction accuracy of traditional environmental management methods, thereby improving the real-time performance of environmental monitoring, data management efficiency and environmental prediction accuracy.

[0139] In order to verify the effectiveness of the optimization method for environmental management of the entire process of water transport engineering construction proposed in the embodiment of the present invention, this embodiment selected a large coastal water transport engineering construction project as the experimental object. The project plans to build a new deep-water channel to connect the port with the open sea, involving large-scale dredging and reclamation projects, which may have a significant impact on the surrounding water environment.

[0140] By applying this method in actual water transport engineering projects, its performance in environmental data collection, processing, storage, prediction and management is tested. By real-time monitoring, data preprocessing, classification and labeling management of environmental data in the construction area, an environmental change prediction model is constructed to predict the impact of the construction process on the water environment, thereby verifying the effectiveness and superiority of the method of the present invention.

[0141] [Experimental subjects and conditions]

[0142] This large-scale coastal water transport project plans to build a new deep-water channel in the sea area surrounded by points A, B, C and the coastline. The construction content mainly includes large-scale dredging. Since the construction may cause changes in water depth and hydrodynamic conditions, which will affect the surrounding water environment, it is necessary to conduct detailed analysis and management. In order to evaluate and optimize the impact on the water environment during the construction process, the optimization method for environmental management of the entire process of water transport project construction proposed in this invention is adopted, and experimental tests are carried out in combination with the MIKE21FM plane two-dimensional numerical model. Among them:

[0143] Geographical scope: Reference Figure 2, is the calculation domain range of the sea area mathematical model established in the above example (the plane two-dimensional numerical model MIKE21FM is used for prediction and analysis, and the model uses unstructured triangular mesh to divide the calculation domain), that is, the sea area surrounded by the three points A, B, and C in the figure and the coastline. This area has complex coastline and hydrodynamic characteristics. The coordinates of each boundary point are shown in Table 1.

[0144] Table 1. Boundary point coordinate statistics

[0145] dot number north latitude East longitude A 21°30′3.38″ 108°13′6.45″ B 20°33′57.83″ 108°19′23.48″ C 20°38′0.59″ 109°44′33.54″

[0146] The simulation uses an unstructured triangular mesh. The engineering simulation area consists of 68,531 nodes and 36,582 triangular elements. The minimum spatial step is about 30 m, the minimum time step is 1.0 s, and the sea area calculation grid is as follows: Figure 2 As shown;

[0147] Sensor network: Multiple types of sensors, including ultrasonic water level meters, Doppler flow meters, water quality monitors, automatic samplers, microscopes, and soil analyzers, are deployed at the construction site, covering the entire construction area and surrounding environment;

[0148] Data collection cycle: Real-time data is collected once every hour, and historical data is collected from the past year;

[0149] Environmental parameters: hydrodynamic environmental parameters (water level, flow velocity, flow direction), water quality parameters (pH value, suspended sediment content, harmful chemical content), biological indicators (plankton species and quantity), soil bottom parameters (soil elevation, particle size, organic matter content);

[0150] Selected model: Considering the impact of engineering construction on the hydrodynamic environment, this embodiment uses the plane two-dimensional numerical model MIKE21FM for water environment analysis. The model uses unstructured triangular mesh to divide the calculation domain, which can accurately fit the complex land boundary. The mesh design is flexible and the mesh density can be controlled at will. The model has the following characteristics:

[0151] Mesh generation: The unstructured triangular mesh is used to accurately fit the complex land boundaries. The mesh design is flexible and the density of the mesh can be controlled at will. The simulation area consists of 68,531 nodes and 36,582 triangular units, and the minimum spatial step length is about 30 meters.

[0152] Calculation method: The standard Galerkin finite element method is used for horizontal space discretization, and the explicit upwind difference format is used for time discretization of momentum equation and transport equation;

[0153] Model parameter settings:

[0154] Meshing: The simulation area consists of 68,531 nodes and 36,582 triangular elements, with a minimum spatial step of about 30 meters;

[0155] Time step: dynamically adjusted according to CFL conditions, with a minimum time step of 1.0 second;

[0156] Water depth and shore boundary data: Select the latest 1:1000 surveyed water depth and topographic data of the sea chart and the sea area near the project area, and unify them to the national 85 benchmark elevation as the water depth and shore boundary conditions of the model;

[0157] Boundary conditions: The offshore tide prediction program (China Tide) provides open boundaries (the open boundary of the offshore is a given tide process line) and closed boundaries (the sea area and the coastline around the sea area are used as closed boundaries);

[0158] Substrate roughness: controlled by Manning coefficient, Manning coefficient n is 30~60m 1 / 3 / s.

[0159]

Experimental process

[0160] (1) A multi-type sensor network was deployed at the project site to collect environmental data throughout the construction process of the water transport project, including real-time data and historical data. The specific sensors and collection parameters are as follows:

[0161] Physical Data Sensors:

[0162] Ultrasonic water level meter: collect water level data;

[0163] Doppler flowmeter: collects flow velocity and direction data;

[0164] Chemical Data Sensors:

[0165] Water quality monitor: collects water quality parameters such as pH value, suspended sediment content, and harmful chemical content;

[0166] Automatic sampler: collects water samples regularly for laboratory analysis;

[0167] Biological data collection:

[0168] Microscope: Analyze the types and quantities of plankton in water samples;

[0169] Soil Data Sensors:

[0170] Soil Analyzer: measures soil bottom parameters such as soil elevation, particle size, and organic matter content;

[0171] (2) Preprocess the collected real-time data and historical data, including data cleaning and standardization;

[0172] According to the attributes of environmental data, it is divided into:

[0173] Physical data: water level, flow velocity, flow direction;

[0174] Chemical data: pH value, suspended sediment content, harmful chemical content;

[0175] Biological data: species and quantity of plankton;

[0176] Soil data: soil elevation, particle size, organic matter content;

[0177] (3) Store preprocessed real-time data and historical data in a distributed database to ensure high availability and scalability of data;

[0178] (4) Time domain classification:

[0179] The first category of characteristic data: real-time data, including first physical data, first chemical data, first biological data and first soil data;

[0180] The second category of characteristic data: historical data, including second physical data, second chemical data, second biological data and second soil data;

[0181] (5) Status label added:

[0182] Based on the set threshold, a status label is added to each category of data, and the labels include "normal 0", "warning 1" and "danger 2". The judgment conditions are shown in the above step S2;

[0183] For example, the threshold value of suspended sediment content is set as:

[0184] T 正常 =50mg / L

[0185] T 警告 ==100mg / L

[0186] Then use One-Hot Encoding to convert the state label into a numerical feature to form a feature vector for easy model processing;

[0187] (6) The model is trained and its parameters are updated using the method in step S3 above. The simulation results of the MIKE21FM model (such as the water level, flow velocity, and flow direction change data before and after the project) are used as additional features and integrated into the historical data to improve the prediction accuracy of the model.

[0188] The mass conservation equation:

[0189]

[0190] Momentum equation:

[0191]

[0192] Among them, the Coriolis force parameter f = 2ωsinφ, ω is the angular velocity of the earth's rotation, φ is the geographical latitude, ζ is the water level change, h is the still water depth, H is the total water depth, H = h + ζ, u and v are the vertical average flow velocities in the x and y directions respectively, t is time, x and y are spatial coordinates, g is the gravitational acceleration, C D : Shear stress coefficient (bottom friction coefficient), used to represent the resistance of bottom friction to fluid movement, unit is dimensionless;

[0193] The rate of change of flow in space; The partial derivative of total water depth with respect to time represents the rate of change of water depth with time; The rate of change of momentum per unit width in time; The convective change of momentum in its own direction represents the spatial change of momentum caused by the change of velocity;

[0194] Convective changes of momentum in cross directions; Water level gradient, which indicates the slope of the water surface; The modulus of the flow velocity indicates the magnitude of the fluid velocity;

[0195] Horizontal eddy viscosity coefficient: calculated using the Smagorinsky formula:

[0196]

[0197] Among them, v t : horizontal eddy viscosity coefficient, which represents the effect of turbulence on fluid momentum transfer; C s : Smagorinsky constant, an empirical coefficient, ranging from 0.1 to 0.2, used to adjust the strength of the turbulence model; l: characteristic mixing length, representing the scale of the turbulent eddy; u: velocity component of the fluid in the x direction (vertical average velocity); v: velocity component of the fluid in the y direction (vertical average velocity), The partial derivative of u with respect to x represents the spatial rate of change of the fluid velocity in the x direction; The partial derivative of v with respect to y represents the spatial rate of change of the fluid velocity in the y direction; The partial derivative of u with respect to y represents the rate of change of the fluid velocity in the x direction with respect to the y direction. The partial derivative of v with respect to x represents the rate of change of the fluid velocity in the y direction with respect to the x direction;

[0198] (7) Import the first category feature data (real-time data) into the environmental change prediction model after training, use the optimized model parameters, calculate the hidden state and output through forward propagation, and calculate the predicted value P according to the predicted value P. t Determine the state of the environment;

[0199] The simulation model was simulated using MATLB software to obtain the flow velocity and flow direction changes at representative points before and after the project. The results are shown in Tables 2 and 3:

[0200] Table 2. Statistics of flow velocity changes at representative points before and after the project

[0201]

[0202] Table 3. Statistical table of flow direction changes at representative points before and after the project

[0203]

[0204]

[0205] The simulation results show that the flow velocity and direction in some areas have changed after the project, especially near the starting point and axis of the waterway, where the flow velocity has shown an increasing trend;

[0206] It should be noted that the construction period of the water transport project is mainly dredging work. The reason for the change of the tidal field in the surrounding sea area is the change of flow velocity caused by the change of water depth. The change of tidal field flow velocity in the sea area around the project during the rapid rise and fall of the high tide before and after the project construction is as follows: Figure 3 , Figure 4 As shown, where:

[0207] The range of velocity changes before and after the construction of water transport projects is mainly concentrated in the offshore position of the channel. As the channel extends toward the offshore direction, the water depth before the project gradually approaches the construction depth. The water depth before and after the project changes in a gradually decreasing trend, and the velocity changes gradually become smaller;

[0208] After the project, the flow velocity at the starting point and near the axis of the channel showed an increasing trend, and the flow velocity outside the channel (mainly B1~M, M~A channel sections) showed a decreasing trend. The change range was generally lower than 0.02m / s. The degree of change in flow velocity before and after the project was small.

[0209] Furthermore, the environmental change prediction model that integrates the output results of the MIKE21FM model is used to predict the real-time monitoring data. The results are as follows:

[0210] Table 4: Environmental parameter prediction results

[0211]

[0212]

[0213] Referring to Table 4, it can be seen intuitively that the suspended sediment content: the real-time monitoring value is 85 mg / L, which exceeds the normal threshold, the status label is "Warning 1", and the predicted value P t =0.75, the model successfully predicted the environmental risk; harmful chemical content: the real-time monitoring value is 0.05 mg / L, close to the warning threshold, the status label is "Warning 1", and the predicted value P t =0.70, indicating the need for enhanced monitoring; other parameters: water level, flow velocity, flow velocity change, flow direction change and other parameters predicted value P t Both are less than 0.5, and the environmental status is judged to be “normal 0”, which is consistent with the actual situation.

[0214] It should be noted that the embodiment of the present invention analyzes the water environment after the construction of the water transport engineering project by introducing a planar two-dimensional numerical model MIKE21FM. The model uses an unstructured triangular grid to divide the calculation domain. The triangular grid can fit the land boundary well, the grid design is flexible, and the grid density can be controlled at will. The Smagorinsky formula (turbulence model) considering the subscale grid effect is used to calculate the horizontal eddy viscosity coefficient to more accurately simulate, predict, and describe the movement state of the fluid in a specific environment, that is, to simulate the movement of the fluid before and after the construction of the water transport engineering project, and predict its impact on the water environment, thereby improving the environmental management accuracy of the entire process of water transport engineering construction.

[0215] It should be further explained that, in this embodiment, by combining the simulation results of the MIKE21FM model with the environmental change prediction model of the present invention, the LSTM model that integrates the MIKE21FM model data can accurately predict the changing trends of environmental parameters, especially in key environmental indicators, thereby improving the prediction accuracy. The model accurately identifies potential environmental risks, provides early warning information, and provides a scientific basis for the formulation and implementation of environmental management measures.

[0216] This experiment verifies the feasibility and effectiveness of the optimization method proposed in the present invention in actual engineering projects. Through the comprehensive collection, preprocessing, storage and intelligent analysis of environmental data, it realizes the optimization of environmental management of the entire process of water transport engineering construction, improves the accuracy of environmental monitoring and prediction, and provides technical support for the coordinated development of engineering construction and environmental protection.

[0217] The method for learning and training the above-mentioned environmental change prediction model through learning rate, number of hidden layer units and number of training rounds can be carried out using methods and means in the prior art, which will not be repeated in this example.

[0218] 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 preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An optimization method for environmental management of the entire process of water transport engineering construction, characterized in that: include: Collect real-time and historical data of the environmental data of the entire process of water transport engineering construction through sensors, and perform data preprocessing; The pre-processed real-time data and the historical data are stored in a distributed database, and the real-time data and the historical data are classified and labeled to form first-category feature data and second-category feature data; Building an environmental change prediction model based on a long short-term memory network, and performing learning and training on the environmental change prediction for the second category of feature data; The first category feature data is then imported into the environmental change prediction model after training to optimize parameter accuracy and predict environmental change trends.

2. The optimization method for environmental management of the entire process of water transport engineering construction according to claim 1 is characterized in that: The environmental data collected during the entire process of water transport engineering construction shall at least include hydrodynamic environmental parameters, water quality parameters, biological indicators and soil bottom parameters.

3. The optimization method for environmental management of the entire process of water transport engineering construction according to claim 1 is characterized in that: The data preprocessing converts data from different sources into data in a unified format by performing data cleaning and standardization on the real-time data and the historical data.

4. The optimization method for environmental management of the entire process of water transport engineering construction according to claim 1 is characterized in that: The real-time data and the historical data are classified and labeled respectively, including: Dividing the real-time data and the historical data into first physical data, second physical data, first chemical data, second chemical data, first biological data, second biological data, and first soil data, second soil data according to the attributes and time domain of the real-time data and the historical data; Based on a set threshold, adding status labels to the first physical data, the second physical data, the first chemical data, the second chemical data, the first biological data, the second biological data, the first soil data, and the second soil data; The status labels include normal, warning, and danger.

5. The optimization method for environmental management of the entire process of water transport engineering construction according to claim 4 is characterized in that: The first physical data, the second physical data, the first chemical data, the second chemical data, the first biological data, the second biological data, the first soil data, and the second soil data are converted into numerical features through label encoding to obtain first category feature data and second category feature data, respectively.

6. The optimization method for environmental management of the entire process of water transport engineering construction according to claim 5 is characterized in that: The first category feature data is data generated by classifying and labeling the real-time data, and the first category feature data includes numerical features converted from the first physical data, the first chemical data, the first biological data, and the first soil data; The second category feature data is data generated by classifying and labeling the historical data, and the second category feature data includes numerical features converted from the second physical data, the second chemical data, the second biological data and the second soil data.

7. The optimization method for environmental management of the entire process of water transport engineering construction according to claim 1 is characterized in that: An environmental change prediction model is constructed based on the long short-term memory network to capture the time series characteristics of environmental data. The environmental change prediction model uses the cross entropy loss function as the objective function and uses an adaptive estimation optimization algorithm to adjust the model parameters to improve the convergence speed and prediction accuracy of the model. Its mathematical expression is as follows: Among them, θ t+1 is the model parameter of the t+1th iteration, θ t is the model parameter of the tth iteration, η is the learning rate, which controls the step size of parameter update, is the loss function with respect to the parameter θ t The gradient of In the learning training, the learning rate η=0.001, the number of hidden layer units is 64, and the number of training rounds is 100.

8. The optimization method for environmental management of the entire process of water transport engineering construction according to claim 7 is characterized in that: Using the optimized environmental change prediction model to predict environmental change trends includes: Define the input as real-time data x t , using the optimized model parameters θ * , calculate the hidden state h through forward propagation t and output P t , its mathematical expression formula is as follows: h t =LSTM(x t ,h t-1 ;θ * ) Among them, P t is the predicted value of the environmental state at the tth time step, indicating the probability that the current environment is abnormal. is the weight matrix of the output layer, is the bias vector of the output layer, σ is the Sigmoid activation function, x t is the input data of the tth time step, h t-1 is the hidden state at the previous time step.

9. The optimization method for environmental management of the entire process of water transport engineering construction according to claim 8 is characterized in that: By outputting the predicted value P of the environmental state t Make a judgment and obtain the environmental status; When P≥0.8, it means that there is a significant change or abnormality in the environmental state, which is a dangerous state and requires immediate response measures; When 0.5≤P<0.8, it indicates that the environmental status has a trend of change, which is a warning state and requires enhanced monitoring; When P < 0.5, it means that the environmental state is stable, with no obvious trend of change, and is in a normal state.

Images