Intelligent guiding system and method for predicting ship navigation disturbance release based on machine learning

Abstract

The invention discloses an intelligent guidance system and method for predicting ship navigation disturbance release based on machine learning. The intelligent guidance system comprises a data acquisition module, a machine learning analysis module and a ship navigation prediction feedback module. A channel water pollution prediction model is constructed through a machine learning algorithm, the concentration of pollutants in overlying water of a channel after a ship passes is predicted in real time, and an optimal route is recommended after water environment protection and economic benefits are comprehensively considered, so that the ship is guided to sail. The problem that endogenous pollution release caused by ship navigation disturbance is difficult to predict and control is solved, ecological protection is considered while ship transportation is developed, the method has the advantages of automation, high efficiency, accuracy, flexibility and the like, and technical support is provided for controlling water environment risks in a navigation channel.

Description

Technical Field

[0001] The present invention relates to an intelligent guidance system and method for predicting the release of ship navigation disturbance based on machine learning, belonging to the field of water ecological environment protection. Background Art

[0002] Plain river network areas are characterized by prosperous industry, dense population, dense water network, and highly developed shipping industry. Due to the relatively shallow water depth of the river channel, the tail propeller of the ship sucks in the surrounding water flow and accelerates it to discharge it quickly to the rear when sailing, which causes scouring of the riverbed and is accompanied by the rotation and diffusion of the water flow. This process has the characteristics of high intensity, high frequency, and strong uncertainty. The induced complex hydrodynamics can easily destroy the original water-sediment balance system, causing the re-migration and transportation of pollutants in the riverbed, and have an adverse impact on the aquatic ecosystem of the regional waterway. However, there is currently little understanding of the release of endogenous pollutants caused by ship navigation, and there is a lack of effective methods to predict the release of pollutants caused by ship navigation disturbances in actual engineering. In addition, there are many factors that affect the intensity of ship disturbances, including ship tonnage, type, propeller diameter, rotation speed, speed, load, water flow velocity, etc. The relationship between them and the intensity of ship disturbances is complex and all have an impact on the release of pollutants, but it is still difficult to summarize and analyze them with a quantitative formula. Machine learning is a branch of artificial intelligence that can find patterns in large amounts of data and apply these patterns to predict new data. It has the advantages of automated decision-making and prediction, continuous learning and optimization, discovery of hidden patterns, and improved efficiency and accuracy. How to apply it to predict ship disturbance release will effectively control water environment risks in inland waterways and is of great significance for coordinating the development of ship shipping and ecological environmental protection. Summary of the invention

[0003] Purpose of the invention: In order to overcome the deficiencies in the prior art, the present invention provides an intelligent guidance system and method for predicting the release of ship navigation disturbance based on machine learning. The present invention constructs a waterway water pollution prediction model through a machine learning algorithm, predicts the concentration of pollutants in the overlying water of the channel after the ship passes in real time, and recommends the best route after comprehensively considering water environment protection and economic benefits, thereby guiding the navigation of the ship, solving the problem that the release of endogenous pollution caused by ship navigation disturbance is difficult to predict and control, reducing the water environment risk in the waterway, taking into account ecological protection while developing ship transportation, and has the advantages of automation, efficiency, accuracy, and flexibility.

[0004] Technical solution: To solve the above technical problems, the present invention provides an intelligent guidance system based on machine learning to predict the release of ship navigation disturbances, including a data acquisition module, a machine learning analysis module and a ship travel prediction feedback module, wherein the data acquisition module includes a ship information scanner, a water level sensor, a flow rate sensor, a displacement sensor, a water quality multi-parameter sensor and a sediment multi-parameter sensor, the ship information scanner is arranged on the shore, the ship information scanner is used to scan the ship number, the water level sensor, the flow rate sensor, the displacement sensor, the water quality multi-parameter sensor and the sediment multi-parameter sensor are all arranged in the waterway, the ship information scanner, the water level sensor, the flow rate sensor, the displacement sensor, the water quality multi-parameter sensor and the sediment multi-parameter sensor are all connected to the controller signal; the data signal collected by the controller is transmitted to the machine learning analysis module, the result of the analysis by the machine learning analysis module is transmitted to the ship travel prediction feedback module, the ship travel prediction feedback module transmits the result to the controller, and the controller is connected to the display module signal.

[0005] Preferably, the machine learning analysis module includes a data entry module, a data screening module, an intelligent fitting module and a deep analysis module. The data entry module enters all data information and performs data identification; the data screening module uses preset rules to automatically eliminate noise and outliers; the intelligent fitting module uses machine learning algorithms to fit and verify data patterns; the deep analysis module mines the hidden laws and trends behind the data, finds the response relationship between various influencing factors under ship navigation disturbance and the pollutant concentration in the overlying water, and constructs a waterway water pollution prediction model.

[0006] Preferably, the ship prediction feedback module predicts the concentration of pollutants in the water above the channel that will be caused by the navigable ship passing through the current route, generates corresponding route suggestions, and feeds back to the ship operator through the display module.

[0007] A guidance method of an intelligent guidance system based on machine learning to predict the release of ship navigation disturbances, comprising the following steps:

[0008] Step 1: According to the grid division of regional river and lake shipping, the system is deployed at the central node of the grid, and data acquisition modules are installed at important sections of all regional waterways. Water level sensors, flow rate sensors, displacement sensors, water quality multi-parameter sensors, and sediment multi-parameter sensors are simultaneously set up with a collection frequency of 2 times / ship;

[0009] Step 2: When a ship passes through the channel, the ship information scanner is triggered to start, scan the ship number and obtain the ship's tonnage, ship type, and propeller information. At the same time, the water level sensor measures the height of the ship from the bottom, the flow rate sensor measures the water flow rate, the displacement sensor measures the ship's speed, and the water quality multi-parameter sensor measures the concentration of pollutants in the overlying water. The water quality multi-parameter sensor is used to measure the COD, ammonia nitrogen, total phosphorus, total nitrogen and turbidity values ​​of the water body. The sediment multi-parameter sensor measures the concentration of pollutants in the sediment;

[0010] Step 3. When the ship completely passes through the section, the data acquisition module transmits the real-time recorded ship tonnage, ship type, propeller parameters (number, propeller diameter, rotation speed, thrust coefficient), ship height from the bottom, water flow velocity, ship speed, background concentration of pollutants in sediments, and pollutant concentration in overlying water to the machine learning analysis module through the communication link for preprocessing, encodes and identifies the data information to ensure that each piece of data can be uniquely identified and its source and collection time can be traced, and the preset rules are applied to automatically remove noise and outliers, including checking whether the data range is reasonable (such as ship tonnage, upper and lower limits of speed, water flow velocity), identifying and deleting those data points that obviously deviate from the normal distribution, and then standardizing or normalizing the numerical variables so that variables of different dimensions can be compared and calculated on the same scale;

[0011] Step 4: Extract characteristic variables related to the concentration of pollutants in the overlying water from the preprocessed data using correlation analysis and other methods, and further transform the characteristic variables as needed, such as calculating the ratio of ship speed to water flow velocity, the product of propeller speed and thrust coefficient, etc., to better describe the potential influencing factors of the concentration of pollutants in the overlying water, and further screen out the characteristic variables most useful for model prediction through feature importance evaluation (such as model-based feature importance scoring) or feature selection algorithms (such as recursive feature elimination, principal component analysis);

[0012] Step 5. Considering the advantages of neural networks in processing complex nonlinear relationships, the neural network algorithm is selected as the basis of the waterway water pollution prediction model. According to the characteristics of the data, the appropriate neural network architecture (such as multi-layer perceptron, convolutional neural network or recurrent neural network) is selected. The processed data is divided into training set, validation set and test set (ratio is 70%, 15%, 15%). The neural network model is trained using the training set data. The model parameters are optimized by back propagation algorithm and gradient descent method to minimize the prediction error. The validation set is used regularly to evaluate the model performance when training the model. The hyperparameters (such as learning rate, number of hidden layers, number of nodes, activation function, optimizer, etc.) are adjusted as needed to prevent overfitting, and the generalization ability of the model is improved by using methods such as cross-validation.

[0013] Step 6: Use the test set data to perform a final evaluation of the trained model, calculate and compare various performance indicators (such as mean square error, root mean square error, R 2 ) to evaluate the accuracy of the model's predictions, and further optimize the model based on the evaluation results, such as adjusting the grid structure, adding regularization terms, using more complex feature conversion methods, etc. The trained neural network model implements an automated monitoring mechanism, regularly (for example, once a week) collects new data and uses it for model retraining and verification, and continuously iterates and optimizes the model based on actual needs and data changes to improve prediction accuracy and reliability.

[0014] Step 7: Use the constructed waterway water pollution prediction model for real-time monitoring. When a ship enters the regional waterway grid, the system initially proposes all possible routes according to the ship's transportation destination, and automatically inputs the ship's current tonnage, ship type, number of propellers, propeller diameter, speed, thrust coefficient, ship height from the bottom, water flow velocity, and ship speed information. At the same time, it obtains the sediment pollutant concentration in all possible routes of the current ship. Based on the above information, the model automatically predicts the pollutant concentration C in the overlying water after the ship passes through each of the above possible routes. i ;

[0015] Step 8: For each navigable route, i and length L i Sort them from small to large and give them scores in equal proportions. i and L i The smaller the score, the higher the score. i and L i The weights A and B are used to calculate the total score G of each route. i , G i =A×C i +B×L i Based on the scoring results, the route with the highest score is taken as the best route, and feedback is given to the ship operator through an intuitive interface to guide the ship to sail along the best route.

[0016] Beneficial effects: Compared with the prior art, the advantages of the present invention are:

[0017] (1) The present invention automatically constructs a waterway water pollution prediction model through a machine learning algorithm, thereby realizing real-time prediction of pollutant concentrations in the waterway after a ship passes through. The automated and intelligent processing method greatly reduces the need for human intervention, improves the efficiency and accuracy of prediction and control, and reduces the risk of human error.

[0018] (2) The present invention establishes a direct and efficient connection mechanism, which closely links the physical parameters such as ship tonnage, draft, propeller speed, and speed, which are very easy to obtain in daily navigation and have a significant impact on navigation disturbances, with the changes in waterway water quality caused by the ship's disturbance. It simplifies the tedious processes such as complex physical model calculations and intermediate force solutions that are traditionally required, avoids the error accumulation caused by too many intermediate links, makes the prediction and evaluation process more direct, rapid and accurate, and provides a more intuitive and convenient basis for decision-making.

[0019] (3) The present invention can comprehensively consider water environment protection and economic benefits, and quickly calculate and recommend the best route by dynamically adjusting the route, thereby effectively guiding the navigation of ships. This not only ensures that the navigation process is environmentally friendly, but also enables the system to adapt to different navigation conditions and requirements, reflecting the concept of sustainable development. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a schematic diagram of the present invention;

[0021] Figure 2 This is the workflow of the present invention. DETAILED DESCRIPTION

[0022] The present invention will be further described below in conjunction with the accompanying drawings.

[0023] like Figure 1 The intelligent guidance system for predicting the release of ship navigation disturbance based on machine learning shown includes a data acquisition module 1, a machine learning analysis module 2, and a ship travel prediction feedback module 3. The data acquisition module 1 includes a ship information scanner 1-1, a water level sensor 1-2, a flow rate sensor 1-3, a displacement sensor 1-4, a water quality multi-parameter sensor 1-5, and a sediment multi-parameter sensor 1-6. The ship information scanner 1-1 is connected to the ship information platform, and can automatically obtain basic information such as ship tonnage, ship type (type, length, width), propeller (number, diameter, thrust coefficient, speed) by scanning the ship number; the water level sensor 1-2 is used to measure the height of the ship from the bottom; the flow rate sensor 1-3 is used to measure the flow rate of the water flow; the displacement sensor 1-4 is used to measure the speed of the ship; the water quality multi-parameter sensor 1-5 is used to measure the concentration of pollutants in the overlying water; the sediment multi-parameter sensor 1-6 is used to measure the concentration of pollutants in the sediment;

[0024] The machine learning analysis module 2 integrates data fitting and analysis functions, specifically covering four modules: data entry 2-1, data screening 2-2, intelligent fitting 2-3, and in-depth analysis 2-4. The data entry module 2-1 can input all data information and perform data identification; the data screening module 2-2 uses preset rules to automatically remove noise and outliers to ensure data quality; the intelligent fitting module 2-3 uses machine learning algorithms, such as decision trees, random forests, neural networks, etc., to fit and verify data patterns; the in-depth analysis module 2-4 mines the hidden laws and trends behind the data, finds the response relationship between various influencing factors and pollutant concentrations in the overlying water under ship navigation disturbances, and constructs a waterway water pollution prediction model;

[0025] The ship prediction feedback module 3 includes a real-time prediction module 3-1 and a route feedback module 3-2. The real-time prediction module 3-1 predicts the concentration of pollutants in the water above the channel caused by the navigable ship 4 passing through the current route; the route feedback module 3-2 generates corresponding route suggestions and feeds back to the ship operator through an intuitive interface;

[0026] The data acquisition module 1, the machine learning analysis module 2 and the ship travel prediction feedback module 3 work together through signal connection; Figure 1 The arrow direction in is the sailing direction of the ship 4.

[0027] like Figure 2 As shown, an intelligent guidance method for predicting the release of ship navigation disturbance based on machine learning includes the following steps:

[0028] Step 1: According to the grid division of regional river and lake navigation, the system is deployed at the central node of the grid, and the data acquisition module 1 is installed at the important sections of all regional waterways (R1, R2, R3, R4, R5, R6), and the water level sensor 1-2, flow velocity sensor 1-3, displacement sensor 1-4, water quality multi-parameter sensor 1-5, and sediment multi-parameter sensor 1-6 are simultaneously set with a collection frequency of 2 times / ship;

[0029] Step 2: When a ship passes through the channel, the ship information scanner 1-1 is triggered to start, scan the ship number and obtain the basic information of the ship's tonnage, ship type (type, length, width), and propeller (number, diameter, thrust coefficient, speed). At the same time, the water level sensor 1-2 measures the height of the ship from the bottom, the flow rate sensor 1-3 measures the water flow rate, the displacement sensor 1-4 measures the ship's speed, the water quality multi-parameter sensor 1-5 measures the pollutant concentration in the overlying water, and the sediment multi-parameter sensor 1-6 measures the pollutant concentration in the channel sediment;

[0030] Step 3. When the ship completely passes through the section, the data acquisition module transmits the real-time recorded ship tonnage, ship type, propeller parameters (number, propeller diameter, rotation speed, thrust coefficient), ship height from the bottom, water flow velocity, ship speed, background concentration of pollutants in sediments, and pollutant concentration in overlying water to the machine learning analysis module through the communication link for preprocessing, encodes and identifies the data information to ensure that each piece of data can be uniquely identified and its source and collection time can be traced, and the preset rules are applied to automatically remove noise and outliers, including checking whether the data range is reasonable (such as ship tonnage, upper and lower limits of speed, water flow velocity), identifying and deleting those data points that obviously deviate from the normal distribution, and then standardizing or normalizing the numerical variables so that variables of different dimensions can be compared and calculated on the same scale;

[0031] Step 4: Extract characteristic variables related to the concentration of pollutants in the overlying water from the preprocessed data using correlation analysis and other methods, and further transform the characteristic variables as needed, such as calculating the ratio of ship speed to water flow velocity, the product of propeller speed and thrust coefficient, etc., to better describe the potential influencing factors of the concentration of pollutants in the overlying water, and further screen out the characteristic variables most useful for model prediction through feature importance evaluation (such as model-based feature importance scoring) or feature selection algorithms (such as recursive feature elimination, principal component analysis);

[0032] Step 5: Select a neural network algorithm as the basis of the waterway water pollution prediction model. Select a suitable neural network architecture (such as a multi-layer perceptron, convolutional neural network or recurrent neural network) according to the characteristics of the data. The processed data is divided into a training set, a validation set and a test set (with a ratio of 70%, 15%, and 15%). Use the training set data to train the neural network model. Optimize the model parameters through the back propagation algorithm and the gradient descent method to minimize the prediction error. When training the model, regularly use the validation set to evaluate the model performance, and adjust the hyperparameters (such as learning rate, number of hidden layers, number of nodes, activation function, optimizer, etc.) as needed to prevent overfitting, and use methods such as cross-validation to improve the generalization ability of the model.

[0033] Step 6: Use the test set data to perform a final evaluation of the trained model, calculate and compare various performance indicators (such as mean square error, root mean square error, R 2 ) to evaluate the accuracy of model predictions, and further optimize the model based on the evaluation results, such as adjusting the grid structure, adding regularization terms, using more complex feature conversion methods, etc. The trained neural network model implements an automated monitoring mechanism, regularly collects new data and uses it for model retraining and verification, and continuously iterates and optimizes the model based on actual needs and data changes to improve prediction accuracy and reliability;

[0034] Step 7: Use the constructed water pollution prediction model for the waterway to conduct real-time monitoring. When the ship 4 enters the regional waterway grid, the system preliminarily drafts all the navigable routes according to the shipping destination of the ship, including Route 1: R2→R6-1→R6-2, Route 2: R1-1→R3-1→R3-2→R6-2, Route 3: R1-1→R3-1→R5→R4-2, Route 4: R1-1→R1-2→R4-1→R4-2, Route 5: R1-1→R1-2→R4-1→R5→R3-2→R6-2, and automatically inputs the current tonnage, ship type, number of propellers, propeller diameter, rotational speed, thrust coefficient, distance of the ship from the bottom, water flow velocity, ship speed, etc. of the ship 4. At the same time, obtain the sediment pollutant concentrations in all the navigable routes of the current ship 4. According to the above information, the model automatically predicts that the pollutant concentrations in the overlying water after the ship passes through the above 5 routes are C1, C2, C3, C4, C5 respectively, and the lengths of the 5 routes are L1, L2, L3, L4, L5 respectively;

[0035] Step 8: Sort the pollutant concentrations C i and the lengths L i of the 5 navigable routes from small to large. Assume C2 < C1 < C4 < C3 < C5, L3 < L2 < L1 < L4 < L5. Score the C i and the length L i proportionally, with a full score of 100 points. That is, C2 gets 100 points, C5 gets 20 points, L3 gets 100 points, L5 gets 20 points. Assign a weight of 50% to both C i and L i . Calculate the total scores G i of the 5 routes according to the formula G i = 50%×C i + 50%×L i to be 70, 90, 70, 50, 20 respectively. According to the scoring results, Route 2 with the highest score is taken as the best route and is fed back to the ship operator through an intuitive interface to guide the ship to sail along the route of R1-1→R3-1→R3-2→R6-2.

[0036] The above is only the preferred implementation manner of the present invention. It should be noted that: for those of ordinary skill in the art in this technical field, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. An intelligent guidance system based on machine learning to predict the release of ship navigation disturbances, characterized in that: It includes a data acquisition module, a machine learning analysis module and a ship travel prediction feedback module. The data acquisition module includes a ship information scanner, a water level sensor, a flow rate sensor, a displacement sensor, a water quality multi-parameter sensor and a sediment multi-parameter sensor. The ship information scanner is arranged on the shore and is used to scan the ship number. The water level sensor, the flow rate sensor, the displacement sensor, the water quality multi-parameter sensor and the sediment multi-parameter sensor are all arranged in the waterway. The ship information scanner, the water level sensor, the flow rate sensor, the displacement sensor, the water quality multi-parameter sensor and the sediment multi-parameter sensor are all connected to the controller signal; the data signal collected by the controller is transmitted to the machine learning analysis module, the result of the analysis by the machine learning analysis module is transmitted to the ship travel prediction feedback module, the ship travel prediction feedback module transmits the result to the controller, and the controller is connected to the display module signal.

2. The intelligent guidance system for predicting ship navigation disturbance release based on machine learning according to claim 1 is characterized by: The machine learning analysis module includes a data entry module, a data screening module, an intelligent fitting module and a deep analysis module. The data entry module enters all data information and performs data identification; the data screening module uses preset rules to automatically eliminate noise and outliers; the intelligent fitting module uses machine learning algorithms to fit and verify data patterns; the deep analysis module mines the hidden laws and trends behind the data, finds the response relationship between various influencing factors under ship navigation disturbances and the pollutant concentration in the overlying water, and constructs a waterway water pollution prediction model.

3. The intelligent guidance system for predicting ship navigation disturbance release based on machine learning according to claim 1 is characterized in that: The ship prediction feedback module predicts the concentration of pollutants in the water above the channel that will be caused by the navigable ship passing through the current route, generates corresponding route suggestions, and feeds back to the ship operator through the display module.

4. The guidance method of an intelligent guidance system based on machine learning to predict the release of ship navigation disturbances according to any one of claims 1 to 3, characterized in that: The steps include: Step 1: According to the grid division of regional river and lake shipping, the system is deployed at the central node of the grid, and data acquisition modules are installed at important sections of all regional waterways. Water level sensors, flow rate sensors, displacement sensors, water quality multi-parameter sensors, and sediment multi-parameter sensors are simultaneously set up with a collection frequency of 2 times / ship; Step 2: When a ship passes through the channel, the ship information scanner is triggered to start, scan the ship number and obtain the ship's tonnage, ship type, and propeller information. At the same time, the water level sensor measures the height of the ship from the bottom, the flow rate sensor measures the water flow rate, the displacement sensor measures the ship's speed, the water quality multi-parameter sensor measures the pollutant concentration in the overlying water, and the sediment multi-parameter sensor measures the pollutant concentration in the sediment; Step 3. When the ship completely passes through the section, the data acquisition module transmits the real-time recorded ship tonnage, ship type, propeller parameters, ship height from the bottom, water flow velocity, ship speed, background concentration of pollutants in sediments, and pollutant concentration in overlying water to the machine learning analysis module through the communication link for preprocessing, encodes and identifies the data information, automatically removes noise and outliers using preset rules, and standardizes or normalizes the numerical variables; Step 4: Extract characteristic variables related to the concentration of pollutants in the overlying water from the preprocessed data using correlation analysis methods, and further select the characteristic variables most useful for model prediction through feature importance evaluation or feature selection algorithms; Step 5: Select a neural network algorithm as the basis of the waterway water pollution prediction model. Select the corresponding neural network architecture according to the data characteristics. Divide the processed data into training set, validation set and test set. Use the training set data to train the neural network model. Optimize the model parameters through the back propagation algorithm and gradient descent method to minimize the prediction error. Use the validation set to evaluate the model performance when training the model, and adjust the hyperparameters as needed. Step 6: Use the test set data to conduct a final evaluation of the trained model, calculate and compare various performance indicators to evaluate the accuracy of the model prediction, further optimize the model based on the evaluation results, implement an automated monitoring mechanism for the trained neural network model, regularly collect new data and use it for model retraining and verification, and continuously iterate and optimize the model based on actual needs and data changes; Step 7: Use the constructed waterway water pollution prediction model for real-time monitoring. When a ship enters the regional waterway grid, the system initially proposes all possible routes according to the ship's transportation destination, and automatically inputs the ship's current tonnage, ship type, number of propellers, propeller diameter, speed, thrust coefficient, ship height from the bottom, water flow velocity, and ship speed information. At the same time, it obtains the sediment pollutant concentration in all possible routes of the current ship. Based on the above information, the model automatically predicts the pollutant concentration C in the overlying water after the ship passes through each of the above possible routes. i ; Step 8: For each navigable route, i and length L i Sort them from small to large and give them scores in equal proportions. i and L i The smaller the score, the higher the score. i and L i The weights A and B are used to calculate the total score G of each route. i , G i =A×C i +B×L i Based on the scoring results, the route with the highest score is taken as the best route, and feedback is given to the ship operator through an intuitive interface to guide the ship to sail along the best route.

Images