Dam upstream navigation management prediction method based on multi-feature factor research

Abstract

The invention discloses a dam upstream navigation management prediction method based on multi-feature factor research, and relates to the technical field of dam upstream navigation management. According to the method, multi-dimensional data of ship foundation, traffic flow, scheduling management and environment are collected, and key feature factors are extracted after preprocessing; constructing a navigation efficiency evaluation model fusing the cargo load, the ex-warehouse flow and the waiting time, and determining a weight coefficient by adopting a maximum variance method; a navigation state prediction model is constructed based on a random forest algorithm, and prediction of traffic flow density, collision risk and navigation efficiency is realized; and finally, generating an air route planning, scheduling adjustment and emergency management optimization scheme. According to the method, the limitation of single factor consideration of a traditional method is overcome, the comprehensiveness of navigation efficiency evaluation and the accuracy of navigation state prediction are improved, scientific decision support can be provided for navigation management of the upstream water area of the dam, the collision risk is effectively reduced, and the navigation efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of navigation management, and particularly relates to a dam upstream navigation management prediction method based on multi-feature factor research. BACKGROUND

[0002] The dam upstream water area, especially the junction of the main stream and the branch stream, is a key node of water transportation and undertakes a large number of navigation tasks. With the growth of shipping demand, such water areas often face problems such as complex traffic flow, ship congestion, high collision risk and low navigation efficiency. The traditional navigation management method relies on artificial experience and static planning, lacks comprehensive consideration of multi-feature factors such as ship properties, traffic flow operation, dispatching management and environment, and is difficult to accurately predict the navigation state, resulting in insufficient pertinence of the management scheme and inability to effectively respond to the dynamically changing navigation environment.

[0003] In the prior art, although some studies focus on navigation efficiency evaluation, such as measuring navigation efficiency by a single factor (such as ship navigation time and waiting time), the key influencing factors such as cargo capacity and dam discharge flow are not fully integrated, and the evaluation results are one-sided. At the same time, the navigation state prediction mostly uses simple statistical methods, which has low prediction accuracy and is difficult to meet the management needs of the complex water area upstream of the dam. Therefore, there is an urgent need for a navigation management method that integrates multi-feature factors and has precise evaluation and prediction functions to improve the navigation safety and efficiency of the dam upstream water area. SUMMARY

[0004] The technical problem to be solved by the present application is to provide a dam upstream navigation management prediction method based on multi-feature factor research, which realizes accurate prediction of the navigation state of the dam upstream water area by comprehensively collecting multi-dimensional data, extracting key feature factors, constructing a navigation efficiency evaluation model and a state prediction model, and generating a targeted management optimization scheme, thereby effectively improving the navigation management level.

[0005] To solve the above technical problems, the technical solution adopted by the present application is: A dam upstream navigation management prediction method based on multi-feature factor research, comprising the following steps: Step 1: Collect ship basic data, traffic flow data, dispatching management data and environmental data of the target water area upstream of the dam; Step 2: Preprocess the collected data, eliminate outliers and standardize the data, and construct a standardized database; Step 3: Based on the standardized database, extract ship navigation influence feature factors, including ship property characteristics, traffic flow operation characteristics, dispatching management characteristics and environmental characteristics; Step four, establish a navigation efficiency evaluation model, take the extracted characteristic factors as input, and get the navigation efficiency evaluation index through weight distribution and function calculation; Step five, build a navigation state prediction model, based on historical navigation data and navigation efficiency evaluation index, train the model by machine learning algorithm, realize the prediction of the navigation state of the target water area in the future preset time period; Step six, according to the navigation state prediction result, generate the corresponding navigation management optimization scheme.

[0006] The ship basic data in the above step one includes ship name, ship type, ship age, main engine power, cargo carrying capacity; traffic flow data includes ship departure time, arrival time at target intersection water area, ship navigation speed, traffic flow density, traffic flow direction; scheduling management data includes anchorage scheduling scheme, lock scheduling interval, ship security time, ship moving time; environmental data includes dam discharge flow, visibility, wind grade.

[0007] The data preprocessing in the above step two specifically includes: using the Relyda criterion to eliminate the abnormal values in the ship basic data, traffic flow data, scheduling management data and environmental data; using Min-Max standardization method to process continuous data such as cargo carrying capacity, main engine power and dam discharge flow, the formula is: ; Wherein is the standardized data, is the original data, is the minimum value of this kind of data, is the maximum value of this kind of data.

[0008] The ship's own attribute characteristics in the above step three include standardized cargo carrying capacity, main engine power and ship age; traffic flow operation characteristics include ship navigation time, traffic flow density and traffic flow intersection times; scheduling management characteristics include average waiting time of ships in anchorage, lock scheduling interval length, ship security pass rate; environmental characteristics include standardized dam discharge flow, visibility grade and wind grade.

[0009] The calculation formula of the navigation efficiency evaluation model in the above step four is: ; Wherein is the navigation efficiency evaluation index, is the standardized ship cargo carrying capacity, is the dam discharge flow, is the reference flow value, is the waiting time of the ship in the anchorage, , is the waiting time influence coefficient, and is a weight coefficient, and , and is determined by the maximum variance method, that is, a unit eigenvector corresponding to the maximum eigenvalue of the characteristic factor covariance matrix is calculated, and the vector elements are and .

[0010] The machine learning algorithm in the above step five is a random forest algorithm, and the training process of the navigation state prediction model includes: taking historical standardized data and corresponding navigation efficiency evaluation indexes as a training set, and taking traffic flow density, ship collision risk level, and navigation efficiency evaluation index range in a future preset time period as a prediction target; the number of decision trees and the maximum tree depth parameters of the random forest algorithm are optimized by 5-fold cross-validation, and the model training is completed.

[0011] The navigation management optimization scheme in the above step six includes a route planning scheme, a scheduling adjustment scheme, and an emergency management scheme; the route planning scheme is based on the predicted traffic flow direction and density to demarcate a "ring island type" route or set a virtual navigation mark; the scheduling adjustment scheme is based on the predicted navigation efficiency low valley period to adjust the ship sailing batch or lock interval; the emergency management scheme is based on the predicted low visibility, strong wind, and other adverse environmental periods to develop a ship warning and release plan after the ban is lifted.

[0012] The above step five further includes precision verification of the navigation state prediction model, and the mean absolute error and the determination coefficient are used as evaluation indexes, wherein is the number of verification samples, is the actual value, is the predicted value, is the average value of the actual value, when the MAE is less than a preset threshold and is greater than 0.8, the model meets the use requirements.

[0013] The determination method of the reference flow value in the above step four is: the dam discharge flow data of the target water area in the past three years is counted, the average value is calculated after removing the maximum and minimum values, and the average value is taken as ; the determination method of the waiting time influence coefficient is: based on the correlation analysis of the waiting time and the navigation efficiency in the historical data, the linear regression fitting is used to obtain the value, so that can accurately reflect the negative impact of the waiting time on the navigation efficiency.

[0014] The data collection in step one is achieved by cooperation of multiple sources, including collecting ship position, speed, departure and arrival time data by automatic identification system (AIS), collecting discharge flow data by dam hydrological monitoring station, collecting waiting time and moving data by anchorage monitoring system, collecting visibility and wind grade data by weather station, and transmitting data to the data collection terminal in real time through data interface.

[0015] The dam upstream navigation management prediction method based on multi-feature factor research has the following beneficial effects: 1. Multi-feature factor fusion: The present application integrates four characteristic factors of ship itself, traffic flow, dispatching management and environment, overcomes the limitations of traditional single factor consideration, and makes the navigation efficiency evaluation more comprehensive and objective.

[0016] 2. Precise evaluation and prediction: By constructing a scientific navigation efficiency evaluation model and a random forest prediction model, the quantitative evaluation of navigation efficiency and the precise prediction of navigation state are realized, providing data support for management decision.

[0017] 3. Dynamic management optimization: Based on the prediction results, targeted route planning, dispatching adjustment and emergency management scheme are generated, which can dynamically adapt to the changes of navigation environment in the dam upstream water area, effectively reduce the collision risk and improve the navigation efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0018] The present application will be further described below in conjunction with the drawings and examples: Figure 1 The flowchart of the dam upstream navigation management prediction method based on multi-feature factor research in the present application; Figure 2 The schematic diagram of random forest prediction model training and verification in the present application; Figure 3 The schematic diagram of "ring island type" route planning in the embodiment of the present application. DETAILED DESCRIPTION

[0019] The technical solutions of the present application will be described in detail below in conjunction with the drawings and examples.

[0020] As shown in Figure 1 A dam upstream navigation management prediction method based on multi-feature factor research, comprising the following steps: Step one, collecting ship basic data, traffic flow data, dispatching management data and environmental data in the target water area upstream of the dam; Step two, pre-processing the collected data, eliminating outliers and standardizing, and constructing a standardized database; Step three, based on the standardized database, extract the ship navigation influence characteristic factors, including ship attribute characteristics, traffic flow operation characteristics, scheduling management characteristics and environmental characteristics; Step four, establish a navigation efficiency evaluation model, take the extracted characteristic factors as input, and get the navigation efficiency evaluation index through weight distribution and function calculation; Step five, build a navigation state prediction model, based on historical navigation data and navigation efficiency evaluation index, use machine learning algorithm to train the model, realize the prediction of the navigation state of the target water area in the future preset time period; Step six, according to the navigation state prediction result, generate the corresponding navigation management optimization scheme.

[0021] The ship basic data in step one includes ship name, ship type, ship age, main engine power, cargo carrying capacity; traffic flow data includes ship departure time, arrival time at target intersection water area, ship navigation speed, traffic flow density, traffic flow direction; scheduling management data includes anchorage scheduling scheme, lock scheduling interval, ship security time, ship moving time; environmental data includes dam discharge flow, visibility, wind grade.

[0022] The data preprocessing in step two specifically includes: using the Relyda criterion to eliminate outliers in ship basic data, traffic flow data, scheduling management data and environmental data; using Min-Max standardization method to process continuous data such as cargo carrying capacity, main engine power and dam discharge flow, the formula is: ; Among them is the standardized data, is the original data, is the minimum value of this kind of data, is the maximum value of this kind of data.

[0023] The ship attribute characteristics in step three include standardized cargo carrying capacity, main engine power and ship age; traffic flow operation characteristics include ship navigation time, traffic flow density and traffic flow intersection times; scheduling management characteristics include average waiting time of ships in anchorage, lock scheduling interval and ship security pass rate; environmental characteristics include standardized dam discharge flow, visibility grade and wind grade.

[0024] The calculation formula of navigation efficiency evaluation model in step four is: ; Among them is the navigation efficiency evaluation index, is the standardized ship cargo carrying capacity, is the dam discharge flow, is a reference flow value, is a waiting time of the ship in the anchorage, , is a waiting time influence coefficient, and is a weight coefficient, and , and is determined by the maximum variance method, that is, a unit eigenvector corresponding to the maximum eigenvalue of the characteristic factor covariance matrix is calculated, and the vector elements are and .

[0025] The machine learning algorithm in the above step five is a random forest algorithm, and the training process of the navigation state prediction model includes: taking the historical standardized data and the corresponding navigation efficiency evaluation index as the training set, and taking the traffic flow density, ship collision risk level and navigation efficiency evaluation index range in the future preset time period as the prediction target; the number of decision trees and the maximum tree depth parameters of the random forest algorithm are optimized by 5-fold cross-validation, and the model training is completed.

[0026] The navigation management optimization scheme in the above step six includes a route planning scheme, a scheduling adjustment scheme and an emergency management scheme; the route planning scheme is based on the predicted traffic flow direction and density, and a "ring island type" route is delimited or a virtual beacon is set; the scheduling adjustment scheme is based on the predicted navigation efficiency trough period, and the ship departure batch or lock passage interval is adjusted; the emergency management scheme is based on the predicted low visibility, strong wind and other adverse environmental periods, and a ship warning and release plan after the release is formulated.

[0027] The above step five further includes precision verification of the navigation state prediction model, and the mean absolute error and the coefficient of determination are used as evaluation indexes, wherein is the number of verification samples, is the actual value, is the predicted value, is the average value of the actual value, when the MAE is less than a preset threshold and is greater than 0.8, the model meets the use requirements.

[0028] The determination method of the reference flow value in the above step four is: the dam discharge flow data of the target water area in the past three years is counted, the average value is calculated after removing the maximum value and the minimum value, and the average value is taken as ; the determination method of the waiting time influence coefficient is: based on the correlation analysis of the waiting time and the navigation efficiency in the historical data, the value of is obtained by linear regression fitting, so that The negative influence of waiting time on navigation efficiency can be accurately reflected.

[0029] The data collection in step one is achieved by cooperation of multiple sources, including collecting ship position, speed, departure and arrival time data by the automatic identification system (AIS), collecting discharge flow data by the dam hydrological monitoring station, collecting waiting time and moving data by the anchorage monitoring system, and collecting visibility and wind grade data by the weather station. The data is transmitted in real time to the data collection terminal through the data interface of each device.

[0030] Embodiment 1 As shown in Figure 1 and 2 , a dam upstream navigation management prediction method based on multi-feature factor research includes the following steps: 1. Data collection: Collect ship basic data (ship name, type, age, main engine power, cargo capacity), traffic flow data (departure time, arrival time, sailing speed, traffic flow density, flow direction), scheduling management data (anchorage scheduling scheme, lock interval, security time, moving time), and environmental data (dam discharge flow, visibility, wind power) in the target water area upstream of the dam. The data is collected by multiple sources such as AIS system, hydrological monitoring station, anchorage monitoring system and weather station.

[0031] 2. Data preprocessing: Adopt the Relyda criterion to eliminate outliers to avoid interference of abnormal data on subsequent analysis; for continuous data (such as cargo capacity and discharge flow), adopt Min-Max standardization processing to map the data to the [0, 1] interval, eliminate dimension differences, and construct a standardized database.

[0032] 3. Feature factor extraction: Extract ship property characteristics (standardized cargo capacity, main engine power, age), traffic flow operation characteristics (sailing time, traffic flow density, intersection frequency), scheduling management characteristics (average waiting time in anchorage, lock interval, security pass rate), and environmental characteristics (standardized discharge flow, visibility level, wind power level) from the standardized database to form a multi-dimensional feature set.

[0033] 4. Construction of navigation efficiency evaluation model: Based on the extracted feature factors, a navigation efficiency evaluation model is constructed . Among them, is the standardized cargo capacity, is the discharge flow, is the reference flow (the average of the discharge flow in the past 3 years, excluding extreme values), is the anchorage waiting time, ( is the waiting time influence coefficient obtained by linear regression fitting); and is the weight coefficient, which is determined by the maximum variance method (covariance matrix maximum eigenvalue corresponding to the unit eigenvector) to ensure that the model can comprehensively reflect the influence of various factors on the navigation efficiency.

[0034] 5. Navigation state prediction model construction: as shown in Figure 2 , the random forest algorithm is used, the historical standardized data and the navigation efficiency evaluation index are used as the training set, and the traffic flow density, collision risk level and navigation efficiency range in the future preset time period are used as the prediction target; the 5-fold cross-validation optimization algorithm parameters (number of decision trees, maximum tree depth) are used to complete the model training, and the MAE and are used to verify the model accuracy (MAE < preset threshold and R 2 >0.8).

[0035] 6. Management optimization scheme generation: according to the prediction results, a multi-dimensional management scheme is generated: in terms of route planning, "ring island type" route or virtual navigation mark is set based on traffic flow direction and density; in terms of dispatch adjustment, the launch batch and lock interval are optimized during the navigation efficiency low period; in terms of emergency management, the forbidden and limited navigation warning and release plan are formulated during the adverse environment period.

[0036] Example 2: As shown in Figure 3 , the following describes the implementation and effect of the present application by taking the actual channel management and its data on the upstream of a dam as the test target.

[0037] 1. Data collection: taking the Y river water area on the upstream of X dam as the target water area, collecting the ship basic data from January to December 2023 through the AIS system, such as "Zhuoteng 1068" ordinary cargo ship, ship age 13 years, main engine power 1396 kW, cargo carrying capacity 6100 tons, and traffic flow data, launch time 16:19, arrival time at Y river intersection water area 17:04; collecting the discharge flow data from X hydrological monitoring station, such as 17600 cubic meters / second; collecting the ship waiting time (average waiting time 2.5 hours) and moving data through the anchorage monitoring system; collecting the visibility (such as 10 km) and wind grade (2) data from the local weather station.

[0038] 2. Data preprocessing: using the Pauta criterion to eliminate the abnormal value of the ship main engine power, such as the data exceeding 3 times the standard deviation of the normal range; Min-Max standardization is performed on the cargo carrying capacity (6100 tons), assuming that the cargo carrying capacity range is 1000-8000 tons, then the standardized ; the discharge flow (17600 cubic meters / second) is standardized, assuming that the flow range is 10000-25000 cubic meters / second, then the standardized flow is .

[0039] 3. Feature factor extraction: the ship's own attribute features are , the main engine power normalized value 0.68 (assuming the main engine power range 500-2000kW, the original value 1396kW), the ship age normalized value 0.65 (ship age range 0-20 years); traffic flow operation features are navigation duration 45 minutes (17:04-16:19), traffic flow density 20 vessels / hour, intersection times 3 times; scheduling management features are anchorage average waiting time 2.5 hours, lock passage interval 1.5 hours, security check pass rate 90%; environmental features are normalized discharge flow 0.507, visibility level 1 (10km for level 1), wind power level 1 (level 2 for level 1).

[0040] 4. Navigation efficiency evaluation: determine cubic meters / second (average discharge flow in the past 3 years, excluding extreme values), (linear regression fitting), then ; the covariance matrix is calculated by the maximum variance method, and , (satisfies ); substituting the formula gives .

[0041] 5. Model training and prediction: take the data from January to November 2023 as the training set and the data in December as the validation set to train the random forest prediction model; the optimized decision tree number is 100 and the maximum tree depth is 8; the validation result (smaller than the preset threshold 0.05), (greater than 0.8), the model meets the requirements; the traffic flow density is predicted to be 25 vessels / hour from 8:00 to 10:00 on January 10, 2024, the collision risk level is medium, and the navigation efficiency evaluation index range is 0.58-0.62.

[0042] 6. Management scheme generation: based on the prediction results, the "roundabout" route is drawn (ships along the left bank from Z Bay to Y Creek, and ships along the right bank from Y Creek); adjust the anchorage departure batch (every 20 minutes, 4 vessels per batch); for the possible low visibility on January 10 (predicted visibility 5km), develop a 7:30-8:00 no-navigation warning, and after the ban is lifted, release the ships in 2 batches, as shown in Figure 3 .

Claims

1. A method for predicting navigation management upstream of a dam based on the study of multiple characteristic factors, characterized in that, Includes the following steps: Step 1: Collect basic data on vessels, traffic flow data, dispatch management data, and environmental data in the target waters upstream of the dam; Step 2: Preprocess the collected data, remove outliers and standardize the data to build a standardized database; Step 3: Based on a standardized database, extract the characteristic factors affecting ship navigation. These factors include ship's own attribute characteristics, traffic flow operation characteristics, scheduling and management characteristics, and environmental characteristics. Step 4: Establish a general aviation efficiency evaluation model, using the extracted feature factors as input, and obtain general aviation efficiency evaluation indicators through weight allocation and function calculation; Step 5: Construct a navigation status prediction model. Based on historical navigation data and navigation efficiency evaluation indicators, use machine learning algorithms to train the model and predict the navigation status of the target waterway within a preset time period in the future. Step 6: Generate corresponding air traffic management optimization plans based on the air traffic status prediction results.

2. The method for predicting navigation management upstream of a dam based on multi-feature factor research as described in claim 1, characterized in that, The basic ship data in step one includes ship name, ship type, ship age, main engine power, and cargo capacity; traffic flow data includes ship departure time, arrival time at the target intersection area, ship speed, traffic flow density, and traffic flow direction; dispatch management data includes anchorage dispatch plan, lock interval, ship security inspection time, and ship shifting time; and environmental data includes dam outflow, visibility, and wind force level.

3. The method for predicting navigation management upstream of a dam based on multi-feature factor research as described in claim 1, characterized in that, The data preprocessing in step two specifically includes: using the Laida criterion to remove outliers from ship basic data, traffic flow data, dispatch management data, and environmental data; and processing continuous data such as cargo load, main engine power, and dam outflow using the Min-Max standardization method, with the formula as follows: ； in For standardized data, The original data, It is the minimum value in this type of data. This represents the maximum value in this type of data.

4. The method for predicting navigation management upstream of a dam based on multi-feature factor research as described in claim 1, characterized in that, The ship's own attributes in step three include standardized cargo capacity, main engine power, and ship age; traffic flow operation characteristics include ship sailing time, traffic flow density, and number of traffic flow intersections; dispatching and management characteristics include average waiting time of ships at anchorage, lock interval time, and ship security inspection pass rate; and environmental characteristics include standardized dam outflow, visibility level, and wind force level.

5. The method for predicting navigation management upstream of a dam based on multi-feature factor research as described in claim 1, characterized in that, The calculation formula for the navigation efficiency evaluation model in step four is as follows: ； in It serves as an indicator for evaluating navigation efficiency. This refers to the standardized cargo capacity of ships. The outflow from the dam reservoir. For reference flow rate values, The waiting time for a vessel at anchorage. , The waiting time impact coefficient. and These are the weighting coefficients, and , and The maximum variance method is used to determine the eigenvector corresponding to the largest eigenvalue of the covariance matrix of the feature factors. The vector elements are respectively... and .

6. The method for predicting navigation management upstream of a dam based on multi-feature factor research as described in claim 1, characterized in that, In step five, the machine learning algorithm is the random forest algorithm. The training process of the navigation status prediction model includes: using historical standardized data and corresponding navigation efficiency evaluation indicators as the training set, and using traffic flow density, ship collision risk level, and navigation efficiency evaluation indicator range within a preset future time period as the prediction targets; using 5-fold cross-validation to optimize the number of decision trees and maximum tree depth parameters of the random forest algorithm to complete the model training.

7. The method for predicting navigation management upstream of a dam based on multi-feature factor research as described in claim 1, characterized in that, The navigation management optimization plan in step six includes a route planning plan, a scheduling adjustment plan, and an emergency management plan. The route planning plan is based on the predicted traffic flow and density to delineate "island-like" routes or set up virtual navigation marks. The scheduling adjustment plan is based on the predicted low navigation efficiency periods to adjust the departure batches of ships at anchorages or the intervals between lock passages. The emergency management plan is based on the predicted periods of low visibility, strong winds, and other adverse environmental conditions to formulate early warnings for ship navigation restrictions and a phased release plan after the restrictions are lifted.

8. The method for predicting navigation management upstream of a dam based on multi-feature factor research as described in claim 1, characterized in that, Step five also includes verifying the accuracy of the navigation status prediction model using the mean absolute error. and coefficient of determination As an evaluation indicator, among them To verify the sample size, This is the actual value. For predicted values, The average of the actual values; when MAE is less than a preset threshold and When the value is greater than 0.8, the model meets the usage requirements.

9. The method for predicting navigation management upstream of a dam based on multi-feature factor research as described in claim 5, characterized in that, The reference flow rate value in step four The method for determining the value is as follows: Collect data on the outflow from the dam in the target water area over the past three years, remove the maximum and minimum values, calculate the average value, and use this average value as the benchmark. Waiting time impact coefficient The determination method is as follows: based on the correlation analysis of waiting time and navigation efficiency in historical data, linear regression fitting is used to obtain... Value, making It can accurately reflect the negative impact of waiting time on air traffic efficiency.

10. A method for predicting navigation management upstream of a dam based on multi-feature factor research, as described in claim 5, is characterized in that... In step one, data acquisition is achieved through the collaboration of multiple sources of equipment, including the Automatic Identification System (AIS) for collecting data on ship position, speed, departure and arrival times, the dam hydrological monitoring station for collecting outflow data, the anchorage monitoring system for collecting data on ship waiting time and shifting, and the meteorological station for collecting data on visibility and wind force. Each device transmits data to the data acquisition terminal in real time through a data interface.

Images