Green low-carbon arctic route intelligent planning method and system

Abstract

This invention provides a green and low-carbon intelligent planning method and system for Arctic shipping routes. The method includes: constructing a gridded environmental model of the Arctic sea area; establishing a mapping relationship between ship navigation resistance and propulsion power based on a semi-empirical ice resistance formula to construct an ice resistance-power-carbon emission coupled model, and determining the carbon emissions per unit distance; constructing a multi-objective reward function, including a carbon emission penalty term based on the estimated carbon emissions and a time penalty term based on navigation time; and performing path training and strategy optimization based on the gridded environmental model and the multi-objective reward function to output the optimal route that satisfies the multi-objective constraints. This invention, by constructing an ice resistance-power-carbon emission coupled model and introducing an equivalent turning time mechanism, achieves synergistic optimization of navigation efficiency and carbon emission control in complex ice-covered environments, improving the engineering feasibility and green and low-carbon level of planned routes, and is suitable for intelligent decision-making in Arctic mixed ice and water navigation scenarios.

Description

Technical Field

[0001] This invention relates to the field of route planning technology, specifically to a green and low-carbon intelligent planning method and system for Arctic routes. Background Technology

[0002] As global climate change intensifies, Arctic sea ice cover is showing a seasonal decreasing trend, gradually extending the navigation window for Arctic shipping routes. Compared to the traditional Suez Canal route, the Arctic route can significantly shorten the voyage, reducing transportation time and fuel consumption. However, the Arctic shipping area has a complex environment, with problems such as uneven spatial distribution of sea ice concentration, significant variations in ice thickness, and frequent ice-water mixing.

[0003] Existing route planning methods, such as Dijkstra's algorithm, A* algorithm, and fast expanding random trees, suffer from high computational complexity, difficulty in designing heuristic functions, and poor smoothness of generated paths when dealing with the high-dimensional, dynamically changing Arctic environment. More importantly, these traditional methods mostly focus on single-objective optimization (such as shortest distance or minimum risk), making it difficult to effectively integrate carbon emission factors and address their dynamic coupling with navigation efficiency. Although some studies have attempted to introduce multi-objective optimization, most rely on static weights or simplified energy consumption models, lacking the ability to adaptively learn from dynamic environmental changes and failing to flexibly and quantitatively balance economic and environmental benefits during navigation.

[0004] Furthermore, existing methods do not adequately consider factors such as maneuvering lag time and equivalent turning time loss during course changes, leading to operational deviations in the planned routes during engineering applications. Therefore, there is an urgent need to propose an intelligent route planning method that can comprehensively consider the impact of sea ice environment, propulsion energy consumption, and carbon emissions, and possess adaptive decision-making capabilities. Summary of the Invention

[0005] The present invention aims to provide an intelligent planning method and system for Arctic routes that can dynamically balance navigation efficiency and carbon emission targets, so as to overcome the shortcomings of existing technologies in achieving multi-objective collaborative optimization under complex ice conditions.

[0006] To address the aforementioned technical problems, the present invention adopts the following technical solution: A green and low-carbon intelligent planning method for Arctic shipping routes includes: A gridded environmental model of the Arctic Ocean is constructed, which includes sea ice parameters characterizing ice conditions. An ice resistance-power-carbon emission coupled model is constructed. The coupled model establishes the mapping relationship between ship navigation resistance and propulsion power based on the semi-empirical ice resistance formula, and determines the carbon emission per unit distance based on the proportional relationship between propulsion power and fuel consumption rate. Construct a multi-objective reward function, which includes a carbon emission penalty term constructed based on the carbon emission estimate and a time penalty term constructed based on the sailing time, wherein the sailing time includes the ship's normal sailing time and the equivalent turning time caused by the ship changing its course; Based on the gridded environment model and the multi-objective reward function, path training and strategy optimization are performed to output the optimal route that satisfies the multi-objective constraints.

[0007] Furthermore, the ice resistance-power-carbon emission coupling model is a semi-empirical physical model based on ship ice class, ice thickness, and sea ice concentration, used to determine the carbon emission coefficient per unit voyage under different ice conditions.

[0008] Furthermore, the ice conditions include at least open water, thin ice zones, and thick ice zones, and different sailing speeds are preset for different ice zones based on the ship's ice class and / or icebreaking capability.

[0009] Furthermore, the sea ice parameters include sea ice concentration and / or ice thickness, and the gridded environment model is constructed through an ice zone risk assessment model or a direct determination rule based on ice thickness, and divides the navigable area into navigable and non-navigable areas.

[0010] Furthermore, the equivalent turning time is related to the angle of change of the ship's course and / or the ice conditions in the ice zone.

[0011] Furthermore, the expression for the multi-objective reward function is: R = αR1 + βR2 + γR3 R1, R2, and R3 correspond to the path length, carbon emissions, and travel time, respectively, while α, β, and γ are the corresponding weighting coefficients.

[0012] Furthermore, by adjusting the relative magnitudes of the weighting coefficients α, β, and γ, route strategies with different optimization objectives are generated. These route strategies with different optimization objectives include at least an economy-first strategy, an emissions-first strategy, and a comprehensive balance strategy.

[0013] Furthermore, a reinforcement learning algorithm is used for path training and policy optimization, including deep Q-networks, policy gradients, or the Actor-Critic algorithm.

[0014] Furthermore, the navigability of the gridded environment model is assessed based on the polar operation limitation assessment risk index model, and the navigable area is classified into ice zone types according to the ice thickness threshold.

[0015] On the other hand, the present invention provides a green and low-carbon intelligent Arctic route planning system, comprising: An environmental model building module is used to build a gridded environmental model of the Arctic Ocean, which includes sea ice parameters characterizing ice conditions. The coupling calculation module is used to construct an ice resistance-power-carbon emission coupled model. The coupled model establishes the mapping relationship between ship navigation resistance and propulsion power based on the semi-empirical ice resistance formula, and determines the carbon emissions per unit distance based on the proportional relationship between propulsion power and fuel consumption rate. The reward function construction module is used to construct a multi-objective reward function, which includes a carbon emission penalty term constructed based on the carbon emission estimate and a time penalty term constructed based on the sailing time, wherein the sailing time includes the ship's normal sailing time and the equivalent turning time caused by the ship changing its course; The reinforcement learning decision module is used to perform path training and policy optimization based on the gridded environment model and the multi-objective reward function, so as to output the optimal route that satisfies the multi-objective constraints.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention achieves a refined and dynamic estimation of carbon emissions in different ice zones by constructing an ice resistance-power-carbon emission coupling model, breaking through the limitations of traditional methods that rely on static coefficients, and making carbon emission estimation more consistent with actual navigation conditions.

[0017] 2. By introducing an equivalent turning time mechanism, this invention penalizes the maneuvering lag time caused by the ship changing course, which significantly improves the engineering operability and actual navigation efficiency of the planned path and avoids the problems of frequent turning of the planned path and inability to actually track it.

[0018] 3. This invention sets adjustable weighting coefficients, realizing a dynamic and quantitative trade-off between navigation efficiency and carbon emission reduction targets, solving the problem of multi-objective conflict, and providing decision-makers with a set of diversified route options from "economic priority" to "emission reduction priority" and then to "comprehensive balance" to meet the needs of different operational strategies.

[0019] 4. The model trained by this invention can optimize the route under changing ice conditions and has the potential to provide real-time or near-real-time decision support. Attached Figure Description

[0020] Figure 1 This is a flowchart of a method according to an embodiment of the present invention; Figure 2 This is a sea ice data diagram from an embodiment of the present invention; Figure 3 This is a schematic diagram of a rasterized environmental model of the Arctic Ocean, as described in an embodiment of the present invention. Figure 4 This is a block diagram illustrating the principle of the method in an embodiment of the present invention; Figure 5 This is a schematic diagram comparing multi-objective optimization paths in an embodiment of the present invention; Figure 6 This is the output diagram of the path planning for three random days under different ice conditions according to an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0022] Example 1 This embodiment provides a green and low-carbon intelligent planning method for Arctic routes. Its core lies in combining ice resistance models with deep reinforcement learning to achieve intelligent optimization of multi-objective routes. For example... Figure 1 As shown, it includes the following steps: Step 1: Construct a gridded environmental model of the Arctic Ocean, which includes sea ice parameters characterizing ice conditions; Step 2: Construct an ice resistance-power-carbon emission coupled model. The coupled model establishes the mapping relationship between ship navigation resistance and propulsion power based on the semi-empirical ice resistance formula, and determines the carbon emissions per unit distance based on the proportional relationship between propulsion power and fuel consumption rate. Step 3: Construct a multi-objective reward function, which includes a carbon emission penalty term constructed based on the carbon emission estimate and a time penalty term constructed based on the sailing time, wherein the sailing time includes the ship's normal sailing time and the equivalent turning time caused by the ship changing its course; Step 4: Based on the gridded environment model and the multi-objective reward function, perform path training and strategy optimization to output the optimal route that satisfies the multi-objective constraints.

[0023] In step 1 of this embodiment, Arctic sea ice thickness and concentration data for August 2025 are obtained (e.g., Figure 2As shown, based on the ice class of PC7 class vessels, the POLARIS model is used to evaluate the navigability (RIO value) of each grid cell, generating a grid environment that includes navigable areas (RIO>0) and non-navigable areas (RIO<0). The navigable areas are then divided into ice zones based on ice thickness (thin ice zone, thick ice zone, open water). Taking the PC7 vessel as an example, different grid areas are divided into open water (ice thickness ≤10cm), thin ice zone (ice thickness 20-40cm), and thick ice zone (ice thickness 60-80cm).

[0024] In step 2 of this embodiment, a semi-empirical physical coupling model of ice resistance, power, and carbon emissions is constructed. Based on the theory of ship ice resistance and combined with the empirical values ​​of PC7-class ships navigating polar regions, a positive correlation between ice resistance and propulsion power is established. Then, by correlating propulsion power with fuel consumption rate, a quantitative estimate of carbon emissions is achieved. Specifically, the propulsion power requirements under different ice conditions are determined based on the ship's ice class, ice thickness, and sea ice concentration. This includes: firstly, modeling carbon emissions and time costs. Since direct measurement of carbon emissions is difficult, this embodiment uses a semi-empirical ice resistance model for indirect quantification. Based on the ice resistance theory formula... ,in, Sea ice thickness, in meters. The result reflects the nonlinear amplification effect of ice thickness on resistance, which is derived from the fracture mechanics theory in ice mechanics and the fitting results of ship model test data. The ship's speed is expressed in meters per second. The characteristic velocity is taken as 4.0 m / s (approximately 8 knots). The term characterizes the linear amplification effect of speed on resistance; the faster the ship's speed, the greater the amount of ice that is impacted and broken per unit time. This represents sea ice concentration, with values ​​ranging from 0 to 1. The ice concentration index, taken as 1.2, is an empirical value derived from regression analysis of ship model test data by Mellor et al., reflecting the power-law relationship between resistance and ice concentration. Simultaneously, considering the typical speeds of PC7-class vessels in different ice zones (12 knots in open water, 8 knots in thin ice, and 4 knots in thick ice), it is estimated that the propulsion power requirements increase by approximately 10% and 75% respectively in thin and thick ice zones compared to open water. Based on this, the carbon emissions per unit distance in thin ice zones are determined to be 1.10 times that in open water, and 1.75 times that in thick ice zones. Therefore, by combining the IMO standard fuel carbon factor and benchmark fuel consumption, the carbon emission cost per unit step length (per nautical mile) in different ice zones can be calculated. Furthermore, the cost per unit step time is composed of both sailing time and equivalent turning time.

[0025] To accurately quantify ships Discharge is defined by setting a single grid cell to represent a nautical mile of travel. The baseline for each grid cell a ship moves is... Emissions Calculate using the following formula:

[0026] Among them, the fuel carbon factor based on the IMO standard ( = 3.114 kg CO2 / kg fuel) and fuel consumption per unit distance for PC7 class vessels in open water ( = 0.12 tons / nautical mile), the benchmark was calculated. The emissions were 374 kg / step. Based on the relative ratio of propulsion power, the carbon emissions per unit distance in the thin ice zone and the thick ice zone were further determined to be 1.10 times and 1.75 times the baseline values, respectively, i.e., 441.4 kg / step and 654.5 kg / step.

[0027] single-step travel time Based on the distance of each grid cell (1 nautical mile) and ship speed The calculation formula is as follows:

[0028] Based on the PC7 ship's speed in different ice zones (12 knots in open water, 8 knots in thin ice, and 4 knots in thick ice), the calculated single-step navigation times are 0.083 h, 0.125 h, and 0.25 h, respectively.

[0029] Meanwhile, an equivalent turning time is introduced to characterize the maneuvering delay caused by course changes. This time is set based on the experience value of PC7-class ships in polar navigation, and is positively correlated with the ship's course change angle and negatively correlated with the speed in ice areas. It is taken as 0.05 to 0.2 times the corresponding speed according to the ice type. The course change angle only considers the 90° turning angle caused by the agent's up, down, left, and right movements. The relevant parameters of PC7-class ships in different ice areas are shown in Table 1 below.

[0030]

[0031] Combining semi-empirical ice resistance models and Figure 2 The ice concentration data is used to assign a relative carbon emission coefficient to each grid cell, such as... Figure 3 As shown, the table in the lower right corner extracts the relative carbon emission coefficient values ​​corresponding to different sea ice concentrations.

[0032] In step 3 of this embodiment, to achieve multi-objective planning of flight routes, this embodiment designs a comprehensive reward function that integrates carbon emission penalties and time penalties, the expression of which is:

[0033] in, The path length term is a unit penalty (-0.5) incurred for each grid cell the ship moves, used to guide the agent to move in the target direction and find the shortest feasible path; For carbon emissions, the value is the amount generated per grid cell moved by the ship. The negative incentive corresponding to emissions, in kg / step; The travel time term is the negative reward corresponding to the sum of the single-step travel time and the equivalent turning time, in hours per step. , , These are weighting coefficients for path length, carbon emissions, and travel time, used to adjust the relative importance of different optimization objectives in the decision-making process. Set to 1 and adjust. , This enables training to achieve multi-objective optimization.

[0034] By adjusting the weighting coefficients , Based on the relative size, this invention can generate route strategies with different optimization objectives: when Approaching 0 The value of makes and When the magnitudes are similar, agents tend to learn economic-first strategies; when Approaching 0 When the value approaches 1, the agent tends to learn emission reduction priority strategies; when... , All values ​​make and When they are at the same level, the intelligent agent learns a comprehensive balancing strategy to achieve multi-objective collaborative optimization of navigation efficiency and carbon emission control.

[0035] In step 4 of this embodiment, the deep reinforcement learning model construction and training step is performed. This method uses the DQN framework (e.g., Figure 4 As shown in the figure, its specific implementation is as follows: the state space consists of the current position coordinates of the ship's grid. It is composed of ice condition types (open water, thin ice, thick ice, and non-navigable areas); the action space consists of four discrete directions: up, down, left, and right; the network structure adopts a two-layer fully connected neural network, with the input layer dimension being the size of the state space, the hidden layers containing 128 and 64 neurons respectively, and the output layer dimension being 4, corresponding to the Q-values ​​of the four actions. A target network and an experience replay mechanism are also introduced to improve training stability. The agent starts from the starting point, based on... The strategy selects an action, executes it, enters the next state, and obtains the reward from the aforementioned comprehensive reward function. The calculated instant reward, through continuous iterative updates of network parameters, ultimately outputs the optimal route.

[0036] pass The strategy control exploration and utilization involves using the target network for stable training to bring the network parameters and Q-values ​​together, ultimately obtaining a deep neural network model that can predict the optimal action value (i.e., the optimal route) based on the environmental state.

[0037] Finally, there's the route planning step. Once the model is trained, it can be used for actual route planning. For example... Figure 5 As shown, in a 78×34 grid environment constructed based on real data (starting point (10,8), ending point (70,24)), by setting different weighting coefficients, the method of this invention successfully planned three routes with different characteristics: setting , , =5 Economy Priority routes (fewest steps, shortest time, but highest carbon emissions), setting , , =0 emission reduction priority routes (lowest carbon emissions, but longest routes and longest travel times) and setting , , =5 is a well-balanced route (achieving a good compromise in terms of steps, time, and carbon emissions). Furthermore, as... Figure 6 As shown, this method can quickly (within seconds) plan a reasonable flight path under ice condition data on different dates, demonstrating its good adaptability and potential practical application value.

[0038] Example 2 This embodiment provides a green, low-carbon intelligent planning system for Arctic routes, including: An environmental model building module is used to build a gridded environmental model of the Arctic Ocean, which includes sea ice parameters characterizing ice conditions. The coupling calculation module is used to construct an ice resistance-power-carbon emission coupled model. The coupled model establishes the mapping relationship between ship navigation resistance and propulsion power based on the semi-empirical ice resistance formula, and determines the carbon emissions per unit distance based on the proportional relationship between propulsion power and fuel consumption rate. The reward function construction module is used to construct a multi-objective reward function, which includes a carbon emission penalty term constructed based on the carbon emission estimate and a time penalty term constructed based on the sailing time, wherein the sailing time includes the ship's normal sailing time and the equivalent turning time caused by the ship changing its course; The reinforcement learning decision module is used to perform path training and policy optimization based on the gridded environment model and the multi-objective reward function, so as to output the optimal route that satisfies the multi-objective constraints.

[0039] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0040] Obviously, those skilled in the art can make various modifications and variations to the embodiments of the present invention without departing from the spirit and scope of the embodiments of the present invention. Thus, if these modifications and variations to the embodiments of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention also intends to include these modifications and variations.

[0041] All other parts not described in detail are existing technologies.

Claims

1. A green and low-carbon intelligent planning method for Arctic routes, characterized in that, include: A gridded environmental model of the Arctic Ocean is constructed, which includes sea ice parameters characterizing ice conditions. An ice resistance-power-carbon emission coupled model is constructed. The coupled model establishes the mapping relationship between ship navigation resistance and propulsion power based on the semi-empirical ice resistance formula, and determines the carbon emission per unit distance based on the proportional relationship between propulsion power and fuel consumption rate. Construct a multi-objective reward function, which includes a carbon emission penalty term constructed based on the carbon emission estimate and a time penalty term constructed based on the sailing time, wherein the sailing time includes the ship's normal sailing time and the equivalent turning time caused by the ship changing its course; Based on the gridded environment model and the multi-objective reward function, path training and strategy optimization are performed to output the optimal route that satisfies the multi-objective constraints.

2. The intelligent planning method for green and low-carbon Arctic routes according to claim 1, characterized in that, The ice resistance-power-carbon emission coupling model is a semi-empirical physical model based on ship ice class, ice thickness, and sea ice concentration, used to determine the carbon emission coefficient per unit distance under different ice conditions.

3. The intelligent planning method for green and low-carbon Arctic shipping routes according to claim 1, characterized in that, The ice conditions include at least open water, thin ice zones, and thick ice zones, and different sailing speeds are preset for different ice zones based on the ship's ice class and / or icebreaking capability.

4. The intelligent planning method for green and low-carbon Arctic shipping routes according to claim 1, characterized in that, The sea ice parameters include sea ice concentration and / or ice thickness. The gridded environment model is constructed through an ice zone risk assessment model or a direct determination rule based on ice thickness, and divides the navigable area into navigable and non-navigable areas.

5. The intelligent planning method for green and low-carbon Arctic shipping routes according to claim 1, characterized in that, The equivalent turning time is related to the angle of change of the ship's course and / or the ice conditions in the ice zone.

6. The intelligent planning method for green and low-carbon Arctic shipping routes according to claim 1, characterized in that, The expression for the multi-objective reward function is: R = αR1 + βR2 + γR3 R1, R2, and R3 correspond to the path length, carbon emissions, and travel time, respectively, while α, β, and γ are the corresponding weighting coefficients.

7. The intelligent planning method for green and low-carbon Arctic shipping routes according to claim 6, characterized in that, By adjusting the relative magnitudes of the weighting coefficients α, β, and γ, route strategies with different optimization objectives are generated. These route strategies with different optimization objectives include at least an economy-first strategy, an emissions-first strategy, and a comprehensive balance strategy.

8. The intelligent planning method for green and low-carbon Arctic shipping routes according to claim 1, characterized in that, Path training and policy optimization are performed using reinforcement learning algorithms, including deep Q-networks, policy gradients, or the Actor-Critic algorithm.

9. The intelligent planning method for green and low-carbon Arctic shipping routes according to claim 1, characterized in that, The navigability of the gridded environment model is assessed based on the polar operation limitation assessment risk index model, and the navigable area is classified into ice zone types according to the ice thickness threshold.

10. A green and low-carbon intelligent planning system for Arctic routes, used to execute the steps of the green and low-carbon intelligent planning method for Arctic routes according to any one of claims 1-9, characterized in that, include: An environmental model building module is used to build a gridded environmental model of the Arctic Ocean, which includes sea ice parameters characterizing ice conditions. The coupling calculation module is used to construct an ice resistance-power-carbon emission coupled model. The coupled model establishes the mapping relationship between ship navigation resistance and propulsion power based on the semi-empirical ice resistance formula, and determines the carbon emissions per unit distance based on the proportional relationship between propulsion power and fuel consumption rate. The reward function construction module is used to construct a multi-objective reward function, which includes a carbon emission penalty term constructed based on the carbon emission estimate and a time penalty term constructed based on the sailing time, wherein the sailing time includes the ship's normal sailing time and the equivalent turning time caused by the ship changing its course; The reinforcement learning decision module is used to perform path training and policy optimization based on the gridded environment model and the multi-objective reward function, so as to output the optimal route that satisfies the multi-objective constraints.

Images