A method and system for multi-condition energy management and range optimization of functional ships

Abstract

This invention discloses a method and system for multi-condition energy management and endurance optimization of functional vessels, comprising: A1: collecting vessel power operation sequence data, inland waterway operation task attribute data, and navigation space control geographic data, and preprocessing them respectively; A2: sequentially extracting condition transition enhancement features and condition expression features; A3: generating environmental space constraint features through vessel positioning coding and spatial distance constraint feature extraction; A4: calculating the condition features after fusing environmental constraint information and the environmental constraint features after fusing condition information; A5: sequentially generating condition-aware task priority coding features, task energy demand features, and emergency reserve features; A6: sequentially generating multi-source energy allocation weight vectors and power output command vectors for each energy source. This invention can solve the problems of insufficient environmental constraints, insufficient emergency power reserve, and poor adaptability to condition switching in traditional inland waterway functional vessel energy management methods under multi-condition conditions.

Description

Technical Field

[0001] This invention relates to the field of energy management technology for functional ships, and in particular to a method and system for multi-condition energy management and range optimization for functional ships. Background Technology

[0002] Energy management and range optimization for inland waterway hybrid power vessels are primarily aimed at inland waterway vessels equipped with diesel-lithium battery hybrid power systems. By controlling the power distribution and operating mode switching of different power sources, and combining mission requirements, vessel status, and navigation environment, energy can be rationally allocated and efficiently utilized.

[0003] Energy management for functional vessels typically begins with sensors collecting operational information such as speed, battery level, load, and location. A controller then generates a power allocation plan based on predetermined logic or algorithms, and finally, the power system executes the corresponding output and mode switching. Unlike the energy management logic of ocean-going vessels, ordinary inland waterway freight ships, and land-based hybrid vehicles, inland waterway functional vessels, as core carriers for multi-functional maritime operations, do not solely aim for minimum energy consumption or optimal economy. Instead, they must simultaneously meet multiple rigid requirements, including task execution and environmental compliance. Specifically, inland waterway functional vessels need to switch between extreme and diverse operational conditions, such as low-speed cruising, idling at fixed points, and instantaneous full-load emergency acceleration. This places extremely high demands on the sensitivity of the energy management system's operational condition perception and the real-time performance of its power response. Furthermore, their navigation areas cover various types of controlled waterways, constrained by both channel boundaries and tiered emission controls in different waterways, resulting in complex and dynamically changing spatial constraints that exacerbate decision-making difficulties.

[0004] As my country's inland waterway ecological environment protection regulations become increasingly comprehensive, key areas such as drinking water source protection zones and ecologically sensitive waters are subject to strict legal constraints on ship emissions, including zero emissions and power limits. At the same time, the urgency of inland waterway operations and patrols is becoming increasingly apparent, placing higher demands on ship response speed, power continuity, and operational continuity. The rationality and adaptability of energy management strategies directly affect ship endurance, emission compliance, and the quality of operational tasks.

[0005] In actual operation, inland waterway vessels need to switch frequently and irregularly between various working conditions, and must cope with uncertainties such as task adjustments and dynamic changes in controlled areas, which leads to three types of problems: First, poor adaptability to spatial constraints, making it impossible to dynamically adjust the constraint intensity according to the vessel's real-time position, which can easily lead to violations of emissions in sensitive areas or insufficient power in ordinary waterways; Second, difficulty in balancing conflicting demands from multiple dimensions, as there are often conflicts between operational power requirements and environmental constraints, as well as between equipment limits and emergency impacts, which can easily lead to wavering decisions, control fluctuations, or even deadlocks; Third, unreasonable allocation of multiple power sources, as the physical constraints of devices such as diesel generator sets and energy storage battery packs differ significantly, which can easily lead to pain points such as insufficient range, lack of pure electric power in sensitive areas, and lack of emergency power.

[0006] Traditional energy management relies on fixed thresholds or simple rules for power switching, which can only cope with fixed operating conditions and is difficult to adapt to dynamic operating scenarios. Its effectiveness in environmental compliance, emergency power reserve, and range improvement is limited. At present, energy management methods based on deep learning mostly focus on reducing energy consumption and extending range as optimization goals. They do not fully consider scenario-related factors such as the urgency of the task and environmental restrictions, nor do they establish a correspondence between task priority and energy allocation or a power reserve mechanism for emergency tasks. They are still biased towards the energy-saving ideas of civilian ships and have low matching degree with the actual use characteristics of inland waterway vessels, making it difficult to meet their comprehensive needs for emission compliance, task execution, and range optimization. Summary of the Invention

[0007] In view of this, the present invention aims to provide a method and system for multi-condition energy management and endurance optimization of functional vessels, so as to solve the problems of insufficient environmental constraints, insufficient emergency power reserve, and poor adaptability to switching conditions in traditional energy management methods for inland waterway functional vessels under multi-condition conditions.

[0008] A method for multi-condition energy management and range optimization of a functional ship includes: A1: Collect ship power operation sequence data, inland waterway operation task attribute data, and navigation space control geographic data, and preprocess them respectively to obtain preprocessed ship power operation sequence data, preprocessed operation task attribute data, and preprocessed navigation space control geographic data. A2: Based on the preprocessed ship power operation sequence data, extract the working condition transition enhancement features and working condition expression features in sequence; A3: Based on the preprocessed navigation space control geographic data, environmental protection space constraint features are generated through ship positioning codes and spatial distance constraint feature extraction; A4: Based on the characteristics of working conditions and environmental space constraints, a bidirectional complementary fusion is performed through an adaptive cross-modal gating fusion layer to obtain the working condition characteristics after fusing environmental constraint information and the environmental constraint characteristics after fusing working condition information. A5: Based on the preprocessed task attribute data, the working condition characteristics after integrating environmental constraint information, and the environmental constraint characteristics after integrating working condition information, the task priority coding characteristics, task energy demand characteristics, and emergency reserve characteristics of working condition perception are generated in sequence. A6: Based on the operating condition characteristics after integrating environmental protection constraint information, the environmental protection constraint characteristics after integrating operating condition information, the task energy demand characteristics, and the emergency reserve characteristics, a multi-source energy allocation weight vector and a power output command vector for each energy source are generated sequentially.

[0009] Furthermore, step A1 also includes: A11: Collect ship power operation sequence data. The data type is time-series numerical data, including remaining battery power, real-time diesel engine output power, real-time ship speed, and total ship electrical load. The ship power operation sequence data is preprocessed using outlier removal based on interquartile range, missing value completion using cubic spline interpolation, and Z-score standardization to obtain preprocessed ship power operation sequence data. A12: Collect inland waterway operation task attribute data. The data type is a mixture of structured numerical and text data, including the planned route, task type, task urgency level, and operation point information. Preprocess the inland waterway operation task attribute data according to field type: For the planned route field, convert the route coordinate point sequence to the WGS84 coordinate system and perform equidistant resampling along the route path to ensure consistent coordinate point sequence lengths across different routes; for the task type field, use one-hot encoding to convert the category text into a binary numerical vector; for the task urgency level field, assign incremental integer values ​​from low to high to establish an ordered numerical mapping; for the operation point information field, convert the latitude and longitude coordinates of each operation point to the WGS84 coordinate system and convert the attribute text labels of the operation points into integer numbers using label encoding, thus obtaining the preprocessed operation task attribute data. A13: Collect navigation space control geographic data, which is vector geographic data, including real-time ship positioning data, environmental sensitive area fence polygons, and waterway boundary coordinate sequences; preprocess the navigation space control geographic data using WGS84 coordinate system and Gauss-Kruger plane projection transformation, time-series positioning data resampling at equal time intervals, and timestamp alignment methods to obtain preprocessed navigation space control geographic data.

[0010] Furthermore, step A2 also includes: A21: Based on the preprocessed ship dynamics timing data, the operating condition transition enhancement features are extracted through causal dilation convolution and a condition transition sensitivity enhancement mechanism. The calculation method is as follows: ; ; ; in, For the first The basic characteristics of the operating conditions at any given time, where t is the time index. This is a causal dilation convolution operation. For the first Preprocessed ship dynamics timing data at specific times. For the first The first-order time difference component of the operating characteristics at any given time. For the first The basic characteristics of the operating conditions at any given time. For the first The characteristics of the transition between operating conditions at any given moment are enhanced. For the Sigmoid function, It is a multilayer perceptron. For Hadamah accumulation; A22: Based on the enhanced characteristics of the transition between operating conditions, the characteristics of the operating condition expression are calculated. The calculation method is as follows: ; in, For the first The characteristics of the working condition at any given moment.

[0011] It should be further explained that, unlike ordinary cargo ships, inland waterway vessels need to switch between operating conditions such as low-speed cruising, idling fixed-point duty, and instantaneous full-load acceleration. If the system cannot capture the sudden change in operating conditions in time, it is easy to cause a delay in power switching. The diesel engine will only slowly intervene after emergency acceleration has already started, resulting in insufficient power output and affecting the efficiency of the operation. This invention addresses this problem through the following methods: First, causal dilated convolution is used to process the preprocessed ship dynamics timing data. Causal dilated convolution utilizes only historical data from the current moment and earlier, expanding the receptive field without increasing the number of parameters through a dilation mechanism. This allows the network to capture the evolution of operating conditions within a longer time window, following temporal causal relationships, thus obtaining the basic operating characteristics containing historical dynamic information. Second, the first-order time difference component of the basic operating characteristics at adjacent moments is calculated. This difference component directly reflects the magnitude and direction of changes in the operating condition state—when a functional vessel suddenly transitions from low-speed cruising to acceleration, the difference component exhibits a significant response, shifting the system's focus from the absolute state of the operating condition to its dynamic trend. Third, a mechanism to enhance sensitivity to operating condition transitions is applied to the difference component. The components undergo adaptive gating: the differential component is mapped to a gating value between 0 and 1 through a multilayer perceptron and a sigmoid function. The gating value reflects the intensity of the change in operating conditions. When the operating conditions change drastically, the gating value is close to one, fully preserving the differential signal. When the operating conditions are stable, the gating value is close to zero, suppressing noise interference. The gating value is multiplied element-wise with the differential component to achieve adaptive amplification of significant operating condition transitions and filtering of small fluctuations during stable periods. Combined with the basic operating condition characteristics, this ensures that the enhanced operating condition transition features highlight both the information of sudden changes in operating conditions and preserve the basic steady-state representation. Finally, the enhanced operating condition transition features are processed by a multilayer perceptron and added to themselves. The residual connection ensures the stable transmission of features, enabling the final operating condition expression features to accurately represent the dynamic operating condition evolution of functional ships in complex mission scenarios. In existing technologies, traditional ship energy management methods determine operating conditions based on the absolute value of current power or speed, focusing only on the static state of the operating condition while ignoring dynamic evolution trends. When facing the switching of operating conditions of functional ships, due to the lack of perception of the direction and rate of change of operating conditions, the system can only respond passively after the switching of operating conditions, resulting in lag. This invention, through the calculation of causal dilatation convolution and first-order time difference components, combined with adaptive gating of the operating condition transition sensitivity enhancement mechanism, enables the system to capture change signals and predict the trend of operating conditions at the very beginning of the transition. This provides a forward-looking operating condition expression feature for the early reconfiguration and energy pre-allocation of the power system, avoiding the problem of untimely power response caused by perception lag in traditional methods.

[0012] Furthermore, step A3 also includes: ; ; ; ; ; in, Let be the ship's positioning coding feature at time t. This refers to the real-time ship positioning data in the preprocessed navigation space control geographic data at time t. Let be the distance vector from the ship to each boundary vertex of the channel at time t. Calculated using Euclidean distance. Let be the directed distance from a point to the i-th polygon, where is zero inside the polygon and positive outside, and is the polygon index. Where m is the total number of polygons; These are the first to nth coordinates in the channel boundary coordinate sequence from the preprocessed navigation space control geographic data, where n is the number of channel boundary coordinates. Contribute a weight vector to the channel boundary. For the Softplus function, For temperature parameters, Let i be the scalar quantity of the constraint strength of the i-th environmentally sensitive area. It is an exponential function. The polygon representing the fence of the i-th environmentally sensitive area in the preprocessed geographic data for airspace control. The standard deviation of distance attenuation. Let be the environmental space constraint characteristics at time t. Let m be the constraint strength scalar of the m-th environmentally sensitive area. This is a weight vector projection operation.

[0013] It should be further explained that the navigation of functional vessels is subject to multiple spatial constraints. On the one hand, vessels must navigate within designated waterways, and the location of the waterway boundary directly determines the vessel's operational space. On the other hand, there are environmentally sensitive areas, such as ecologically sensitive waters, where the energy use strategies of vessels need to be strictly regulated and energy allocation cannot be carried out according to the standards of ordinary waterways. If the system cannot dynamically adjust the intensity of environmental constraints based on the specific location of the vessel in the navigation space, it may lead to high-power energy output in environmentally sensitive areas, causing pollution, or excessive constraints in ordinary waterways, resulting in insufficient power. This invention accurately characterizes environmental space constraints through geospatial coding and dynamic calculation of constraint strength. First, real-time ship positioning data is processed using a multilayer perceptron to obtain ship positioning coding features. These features convert geographic coordinates into a learnable vector representation, laying the foundation for subsequent spatial constraint calculations. Second, the distances from the ship to each boundary vertex of the channel are calculated. These distance values ​​reflect the ship's spatial relationship relative to the channel boundary—the closer the ship is to the boundary, the closer it is to the boundary. These distance values ​​are then processed using the Softplus function and temperature parameters to obtain a channel boundary contribution weight vector. This weight vector measures the degree of constraint influence of each boundary vertex on the current position. Adjusting the temperature parameter makes the weight calculation process controllable and flexible. Third, the real-time ship positioning data is processed to each boundary vertex of the channel boundary to obtain the spatial constraints. The directed distance within the sensitive zone is processed, utilizing the characteristic of directed distance—zero inside the zone and positive outside—to determine whether a vessel has entered the environmentally sensitive zone. The square of the distance is processed using an exponential function and the distance decay standard deviation to obtain a scalar of the environmentally sensitive zone constraint strength. This scalar decays rapidly as the distance between the vessel and the environmentally sensitive zone increases, achieving a smooth transition from strict to lenient constraints and avoiding drastic fluctuations in energy distribution caused by abrupt constraint switching. Finally, the vessel's positioning coding features are multiplied element-wise with the environmentally sensitive zone constraint strength scalar to strengthen the representation of the position coding within the environmentally sensitive zone. Simultaneously, the product of the position coding and the channel boundary contribution weight is added, fusing the geometric constraint information of the channel boundary. The resulting environmental space constraint features can simultaneously reflect the superposition effect of channel boundary constraints and environmentally sensitive zone constraints. In existing technologies, a fixed geofencing approach is generally used for constraint—a strict constraint strategy is immediately activated upon entering a certain area and immediately restored upon leaving. This method ignores the spatial continuity of constraints; the constraint intensity changes abruptly when a ship approaches the boundary of the constraint area, which can easily lead to inconsistencies in energy management strategies and instability in the propulsion system. This invention calculates the real-time distance from the ship to the channel boundary and the directed distance to the environmentally sensitive area, and uses an exponential decay function to make the constraint intensity change smoothly with distance. This allows the environmental spatial constraint characteristics to reflect the constraint attributes of the ship's location—gradually increasing the constraint as the ship approaches the constraint area and gradually decreasing the constraint as the ship moves away from the constraint area, avoiding abrupt changes in constraint switching. This provides a precise spatial constraint information basis for subsequent environmental constraint modulation and rational energy allocation.

[0014] Furthermore, step A4 also includes: A4: Based on the characteristics of operating conditions and environmental constraints, a bidirectional complementary fusion is performed through an adaptive cross-modal gating fusion layer to obtain the operating condition characteristics after fusing environmental constraint information and the environmental constraint characteristics after fusing operating condition information. The calculation method is as follows: ; ; ; in, For the first Adaptive cross-modal gating weights at time intervals, For linear layers, This is a dimension splicing operation. To integrate operating condition characteristics with environmental constraints information, It is the hyperbolic tangent function. The environmental constraints are characteristics after integrating operating condition information.

[0015] It is worth noting that inland waterway vessels face conflicts between operational requirements and environmental constraints during daily navigation operations. For example, a vessel may be in a constant-speed cruising mode to ensure a predetermined schedule, requiring the diesel engine to operate at a high load to maintain the preset speed and ensure timely passage through navigation gates and arrival at ports to complete loading and unloading operations. However, it may also be about to enter legally controlled emission areas such as inland drinking water source protection zones, requiring a switch to zero-emission pure electric operation mode to meet environmental compliance requirements. In such scenarios, if a simple rule engine is used for hard decision-making, the system is prone to frequent switching and indecisiveness between operational requirements and environmental constraints, leading to oscillations in the power system control and even logical confusion. This invention addresses this conflict through an adaptive cross-modal gating fusion mechanism. First, it concatenates the operating condition characteristics and environmental space constraint characteristics, calculating adaptive cross-modal gating weights using a linear layer and a sigmoid function. This weight calculation integrates information from both types of features—when operating condition characteristics strongly indicate high-load operating speed requirements, while environmental features show the vessel is approaching legally controlled emission areas, the gating weights are dynamically adjusted based on the combined information, rather than simply taking the absolute value of a single feature. Second, it performs bidirectional fusion of operating condition characteristics: on one hand, it retains the original operating condition characteristics multiplied by the gating weights, representing the basic requirements of the operating condition. On the one hand, the environmental constraint characteristics are multiplied by the complement of the gating weight after nonlinear transformation, so that the environmental constraint information is gradually integrated into the operating condition decision. When the operating condition and the constraint conflict, the gating weight will automatically obtain a balance value between zero and one, so that the fused operating condition characteristics retain the necessary speed and power response requirements, and introduce the compliance consideration of environmental constraints in proportion. At the same time, a symmetrical fusion strategy is adopted for the environmental constraint characteristics, so that they are fused with the operating condition information in reverse, forming a two-way complementary fusion effect. The fused characteristics obtained in this way are neither a simple weighted average, nor an either-or choice, but a flexible balance based on the real-time operating condition-constraint conflict intensity. In existing technologies, fixed-priority decision-making or threshold-triggered hard switching schemes are generally adopted. Either environmental compliance is prioritized, directly limiting main engine output and suppressing operational demand, leading to ship delays, impacting navigation efficiency and transport fulfillment capabilities; or transportation timeliness is prioritized, allowing the main engine to operate at high loads, resulting in excessive emissions in controlled waters and facing environmental compliance penalties. This rigid decision-making approach lacks flexibility in complex conflict scenarios, easily leading to a two-way imbalance or decision deadlock between navigation timeliness and environmental compliance. This invention, through a gating fusion mechanism, enables the system to adaptively switch between operational demand and environmental constraints. It performs proportional information fusion based on the real-time conflict level of the two types of characteristics, ensuring that the final ship energy management decision meets both the timeliness and fulfillment requirements of trunk line transportation and strictly considers emission constraints in environmental protection spaces. This significantly improves the robustness, adaptability, and compliance of decisions in complex navigation scenarios.

[0016] Furthermore, step A5 also includes: A51: Generate condition-aware task priority coding features based on the preprocessed job attribute data; A52: Based on the task priority coding characteristics perceived by the working condition, generate basic energy demand characteristics, working condition modulation coefficient, environmental protection constraint modulation coefficient, and task energy demand characteristics; A53: Based on the basic energy demand characteristics, mission energy demand characteristics, operating condition characteristics after integrating environmental constraints, and environmental constraints after integrating operating condition information, emergency reserve characteristics are generated.

[0017] Furthermore, step A5 also includes: First, the preprocessed task attribute data is processed by a multilayer perceptron to obtain task attribute encoding features. Then, the condition features after fusing environmental constraint information and the environmental constraint features after fusing condition information are concatenated and processed by a linear layer to obtain condition-constraint fusion features. Next, the task attribute encoding features and the condition-constraint fusion features are concatenated and processed by a linear layer and a Sigmoid function to obtain task-condition interaction weights. Finally, the task attribute encoding features and the task-condition interaction weights are multiplied by a Hadamard product, and then added to the condition-constraint fusion features and the Hadamard product of 1 minus the task-condition interaction weights to obtain the condition-aware task priority encoding features. The task priority encoding features of the working condition perception are processed sequentially by a linear layer and a hyperbolic tangent function to obtain the basic energy demand features; the working condition features after incorporating environmental constraint information are processed sequentially by average pooling, a linear layer, and a sigmoid function to obtain the working condition modulation coefficients; the environmental constraint features after incorporating working condition information are processed sequentially by average pooling, a linear layer, and a sigmoid function to obtain the environmental constraint modulation coefficients; the basic energy demand features, the working condition modulation coefficients, and the environmental constraint modulation coefficients are multiplied by a Hadamard product, and then added to the result of the task priority encoding features of the working condition perception processed by a multilayer perceptron to obtain the task energy demand features; The task urgency index is obtained by concatenating the task energy demand characteristics, the operating condition characteristics after integrating environmental constraints, and the environmental constraints characteristics after integrating operating condition information, and then processing them sequentially through average pooling, a linear layer, and a Sigmoid function. The task urgency index is then processed sequentially through a linear layer and a Sigmoid function to obtain the emergency reserve ratio vector. The emergency reserve ratio vector is then multiplied by the basic energy demand characteristics, and added to the result of subtracting the Hadamard product of the emergency reserve ratio vector and the task energy demand characteristics processed by the multilayer perceptron from 1.

[0018] In maritime operations, the energy allocation logic of hybrid power systems faces resource conflicts between cruising missions and sudden emergency missions. Traditional energy management strategies often prioritize overall ship fuel economy as the sole optimization objective, leading to a tendency for ships to continuously consume battery power to replace the inefficient operation of diesel engines during long periods of low-speed cruising or standby. Once the system receives a high-priority command, and the vessel happens to be in an area requiring low-noise operation or subject to environmental restrictions (requiring pure electric or range-extended mode), the battery state of charge (SOC) is often already low, making it difficult to support instantaneous high power output demands. This lack of mission-predictive energy supply mode exposes the power system to the risk of energy shortages at critical moments. To address the aforementioned problems, this invention achieves differentiated scheduling of power resources through deep mapping of task priorities and an energy reserve mechanism. The specific logic is as follows: This invention addresses this problem through a task priority awareness and emergency reserve mechanism. First, the task attribute data is processed using a multilayer perceptron to obtain task attribute encoding features. These features convert task priority, urgency, and other information into a learnable vector representation. The task attribute encoding features are then dimensionally concatenated with the fused operating condition-constraint features. A linear layer and a sigmoid function are used to calculate the task-operating condition interaction weight. This weight reflects the degree of matching between the current task attributes and the operating condition—the interaction weight increases significantly when a sudden high-priority task occurs and the operating condition requires pure electric intervention. Based on this weight, the task attribute encoding features and the operating condition-constraint fusion features are weighted and fused to obtain the operating condition-aware task priority encoding features. These features retain the priority information of the task itself while incorporating the context of the current operating condition. Secondly, energy demand characteristics are generated based on the task priority coding features of operating condition perception. Basic energy demand characteristics are obtained through nonlinear transformation, and then operating condition modulation coefficients and environmental constraint modulation coefficients are calculated. These two modulation coefficients extract the main information from the operating condition characteristics and environmental constraint characteristics, respectively. The operating condition modulation coefficient reflects the basic energy demand level of the current operating condition, and the environmental constraint modulation coefficient reflects the limiting strength of environmental constraints on energy allocation. The basic energy demand characteristics are multiplied element-wise with the two modulation coefficients to obtain the modulated energy demand, which is then added to the result of the task priority coding processed by the multilayer perceptron to finally obtain the task energy demand characteristics, which are used to reflect the actual energy demand of the current task under the current operating conditions and constraints. Next, the emergency reserve mechanism is calculated. The task energy demand characteristics are dimensionally concatenated with the fused operating condition characteristics and environmental constraints. The task urgency index is calculated using average pooling, linear layers, and the Sigmoid function. This index integrates three types of information: the task's energy demand, the dynamic characteristics of the current operating condition, and the strength of environmental constraints. When the system detects an increase in the task urgency index, it indicates a potential high-priority emergency task requiring rapid pure electric intervention. Based on the task urgency index, the emergency reserve ratio is further calculated—this ratio determines how much battery power the system must retain before meeting daily energy output. When the urgency index is high, the emergency reserve ratio increases accordingly, and the system locks in more power. When the urgency index is low, the emergency reserve ratio decreases, allowing more energy for daily operation. The emergency reserve characteristic is obtained by weighted fusion of the basic energy demand characteristics and the task energy demand characteristics using the emergency reserve ratio—ensuring both the satisfaction of daily energy needs and the mandatory withholding of emergency power during periods of high urgency. In existing technologies, systems typically employ fixed reserve strategies or completely disregard reserve issues, resulting in a lack of foresight in energy allocation. This invention, through dynamic calculation of task urgency and adaptive adjustment of emergency reserve ratio, enables the system to lock in energy in advance when it senses potentially high-priority tasks in the future. In this way, even if a sudden emergency task occurs, the battery can maintain sufficient available power to support rapid startup in pure electric mode, avoiding the problem of no power available in emergencies caused by traditional indiscriminate power supply strategies, and ensuring the executability of the task.

[0019] Furthermore, step A6 also includes: The fused decision state vector is obtained by concatenating the operating condition characteristics after integrating environmental constraint information, the environmental constraint characteristics after integrating operating condition information, the task energy demand characteristics, and the emergency reserve characteristics. The fused decision state vector is then processed sequentially through a linear layer and a ReLU function to obtain intermediate decision features. Finally, the intermediate decision features are processed sequentially through a linear layer and a Softmax function to obtain a multi-source energy allocation weight vector. The modulated basic energy demand characteristic is obtained by performing a Hadamard product on the basic energy demand characteristic, the operating condition modulation coefficient, and the environmental constraint modulation coefficient. The modulated basic energy demand characteristic is then processed sequentially by a linear layer and a ReLU function to obtain the total power demand. The operating condition characteristic after incorporating environmental constraint information is concatenated with the environmental constraint characteristic after incorporating operating condition information and then processed sequentially by average pooling, a linear layer, and a Sigmoid function to obtain the constraint satisfaction coefficient. The total power demand is multiplied element-wise by the constraint satisfaction coefficient to obtain the constrained modulated power demand. The multi-source energy allocation weight vector and the constrained modulated power demand are then processed by a Hadamard product to obtain the power output command vector of each energy source. In the power output command vector of each energy source, a truncation function is used to ensure that the value of each component is not greater than the power output upper limit of the corresponding energy source.

[0020] It should be further explained that inland waterway vessels are typically equipped with multiple energy devices such as diesel generator sets, energy storage battery packs, and shaft generators. The available power output range of each energy device varies significantly under different operating conditions and is subject to multiple physical constraints such as mechanical wear limits, thermal management boundaries, and electrochemical safety thresholds. At the same time, the scheduling demands identified in the preceding stages, such as sudden changes in navigation conditions, compliance requirements in environmentally sensitive waters, instantaneous energy impacts of emergency operations, and bottom-line guarantees of safety reserves, often present contradictory or even fierce conflicts. For example, a task may require the instantaneous release of maximum power, but if the vessel is in an emission control zone, environmental constraints may require limiting diesel engine output; if the energy storage battery's state of charge is close to its lower limit, safety reserves may require maintaining a margin. The intersection of these multiple constraints is extremely narrow. In traditional ship energy management practices, multi-source power allocation usually relies on rule-based lookup table methods or simple proportional allocation strategies. That is, a fixed power allocation ratio is pre-set for each typical operating condition, and when the actual operation deviates from the preset scenario, the ship's engineer manually intervenes and adjusts it based on experience. In the above steps, this invention first concatenates the operating condition characteristics after integrating environmental constraint information, the environmental constraint characteristics after integrating operating condition information, the task energy demand characteristics, and the safety reserve characteristics to construct a fusion decision state vector that includes considerations of all business dimensions. Based on this, a multi-source energy allocation weight vector is generated through linear transformation and activation operations. This weight vector, through a normalization function, ensures that the sum of the allocation ratios of each energy source is always one, thus guaranteeing the completeness and mutual exclusivity of energy allocation mathematically. This ensures that each power demand is clearly and completely allocated to the corresponding energy device. In the specific generation process of the power output command, this invention does not simply distribute the calculated total power demand according to the weights, but introduces a constraint satisfaction coefficient as a safety modulation mechanism. This coefficient is obtained by performing mean pooling and nonlinear transformation on the fused operating condition characteristics and environmental constraint characteristics. The output value range is strictly limited to between zero and one, which essentially represents the maximum safe proportion of the actual release power of the system to the theoretical power demand under the joint constraints of the current operating conditions and environmental compliance requirements. By multiplying the constraint satisfaction coefficient element by the total power demand, flexible compression modulation of the power command is achieved. This ensures that the final power demand after constraint modulation fully responds to the real needs of sudden changes in operating conditions, environmental compliance, and emergency tasks identified in the preceding stages, while also being constrained within the safe intersection space of the multidimensional physical boundaries. In addition, this invention applies a truncation function to the final energy power output command vector to hard upper limit the value of each energy component, ensuring that the power command issued to each specific power device does not exceed the maximum continuous output power of the device, thus forming the last rigid line of defense against equipment overload. Compared to existing technologies that rely on fixed rule tables or manual experience for multi-source power allocation, this invention drives the adaptive generation of allocation weights by fusing all-dimensional business features into a unified decision state vector. This allows the allocation strategy to be dynamically adjusted according to the actual operating situation at each moment, rather than being limited to a limited set of preset scenario templates. At the same time, the dual safety guarantee mechanism composed of constraint satisfaction coefficients and truncation functions achieves joint modulation of multi-dimensional operating boundaries at the soft constraint level and limits the physical limits of single devices at the hard constraint level. The synergistic effect of the two effectively avoids the risk of equipment overload shutdown caused by incomplete constraint consideration or untimely response in traditional methods, ensuring the continuous and reliable operation of the power system of inland waterway vessels in complex and ever-changing mission scenarios.

[0021] This invention also discloses a multi-condition energy management and range optimization system for functional ships, comprising: Data acquisition unit: Collects ship power operation sequence data, inland waterway operation task attribute data, and navigation space control geographic data, and preprocesses them respectively to obtain preprocessed ship power operation sequence data, preprocessed operation task attribute data, and preprocessed navigation space control geographic data; Ship dynamic condition extraction unit: Based on the preprocessed ship dynamic operation sequence data, extract condition transition enhancement features and condition expression features in sequence; Environmental space constraint extraction unit: Based on the preprocessed navigation space control geographic data, environmental space constraint features are generated by extracting ship positioning codes and spatial distance constraint features; Operating condition and environmental protection fusion unit: Based on the operating condition expression characteristics and environmental protection space constraint characteristics, bidirectional complementary fusion is performed through an adaptive cross-modal gating fusion layer to obtain the operating condition characteristics after fusing environmental protection constraint information and the environmental protection constraint characteristics after fusing operating condition information. Task feature extraction unit: Based on the preprocessed task attribute data, the working condition features after integrating environmental constraint information, and the environmental constraint features after integrating working condition information, the task priority coding features, task energy demand features, and emergency reserve features are generated sequentially for working condition perception. Energy allocation unit: Based on the operating condition characteristics after integrating environmental protection constraint information, the environmental protection constraint characteristics after integrating operating condition information, the task energy demand characteristics, and the emergency reserve characteristics, it sequentially generates a multi-source energy allocation weight vector and a power output command vector for each energy source.

[0022] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) This invention systematically solves the problem of insufficient adaptability of traditional methods in multi-condition operation of inland waterway functional vessels by establishing a decision-making framework of ship operating conditions, environmental constraints, propagation tasks, and power management. At the level of operating condition perception, this invention adopts a time-series dynamic analysis mechanism, which not only captures the current operating status of the ship, but also identifies the direction and rate of operating condition switching in advance, so that the system changes from a passive and lagging response to an active and forward-looking prediction, ensuring the timely and effective intervention of power in emergency acceleration and other operational tasks. At the level of spatial constraints, this invention introduces a dynamic distance attenuation mechanism, which smoothly adjusts the constraint intensity according to the real-time spatial relationship between the ship and the waterway boundary and the controlled water area, avoiding the hard switching jump of the traditional geofencing method, so that the constraint changes from absolute restriction to continuous and progressive flexible control, which can gradually enhance emission control when approaching sensitive areas and gradually release power when moving away. This invention enhances the stability and continuity of navigation. At the demand balancing level, it employs an adaptive gating fusion strategy, proportionally integrating operational power requirements with environmental constraints rather than making rigid trade-offs. This allows the system to flexibly weigh power demands based on real-time conflict intensity, satisfying both urgent mission power needs and legal emission control requirements. At the power allocation level, it constructs a unified vector for decision-making, driving the adaptive generation of each energy allocation weight. Through a dual mechanism of constraint modulation and hard upper limit protection, it ensures that power commands remain within the safe intersection of multi-dimensional physical boundaries, effectively preventing equipment overload shutdown risks. This represents an upgrade to traditional energy management, improving the system's environmental compliance, emergency response capabilities, and endurance efficiency, providing solid technical support for the reliable, efficient, and compliant operation of inland waterway vessels in complex and ever-changing working scenarios.

[0023] (2) To address the problem that traditional ship energy management relies solely on the absolute value of current power or speed for operational condition judgment, lacks awareness of dynamic evolution trends in operational conditions, and leads to delayed response during operational condition switching, this invention establishes a dynamic operational condition perception mechanism. First, causal dilated convolution is used to process the ship's power operation sequence data, expanding the receptive field without increasing the number of parameters, enabling the network to capture the operational condition evolution process within a longer time window and obtain the basic operational characteristics of operational conditions containing historical dynamic information. Second, by calculating the first-order time difference component of the basic operational characteristics of operational conditions at adjacent moments, the changes in operational condition status are directly reflected. The system firstly considers the amplitude and direction of the differential signal; secondly, it introduces a sensitive enhancement mechanism for operating condition transitions to perform adaptive gating processing on the differential components. By using a multilayer perceptron and a sigmoid function, the differential components are mapped to gating values. When the operating conditions change drastically, the differential signal is fully preserved, while noise interference is suppressed when the operating conditions are stable. Finally, residual connections are used to ensure that the final operating condition expression features can accurately characterize the dynamic operating condition evolution of the functional ship. This allows the system to capture changing signals and predict the operating condition trend at the initial stage of operating condition transitions, providing forward-looking features for the early reconfiguration of the power system and avoiding the problem of untimely power response caused by perception lag in traditional methods.

[0024] (3) To address the problems of existing technologies that use fixed geofencing for constraints, neglect the spatial continuity of constraints, and cause sudden changes in constraint strength and fluctuations in energy distribution, this invention establishes a quantitative mechanism for environmental space constraints. First, the real-time positioning data of ships is processed by a multilayer perceptron to obtain ship positioning coding features, and the geographic coordinates are converted into a vector representation that the network can learn. Second, the distance from the ship to each boundary vertex of the channel is calculated, and the channel boundary contribution weight vector is obtained by processing with the Softplus function and temperature parameters to measure the degree of constraint influence of each boundary on the current position. Third, the directed distance from the ship to each environmentally sensitive area is processed, and the distance is processed by an exponential function to obtain a scalar of constraint strength for environmentally sensitive areas, so that the constraint strength decays rapidly with the increase of distance, realizing a smooth transition from strict constraints to relaxed constraints. Finally, the positioning coding features are fused with the constraint strength scalar to obtain environmental space constraint features. The exponential decay function makes the constraint strength change smoothly with distance, avoiding abrupt changes in constraints, and providing an accurate spatial constraint information basis for subsequent environmental constraint modulation and rational energy distribution.

[0025] (4) To address the problems of existing technologies that employ fixed-priority decision-making or hard-switching schemes, lack flexibility when there is a conflict between operating conditions and environmental constraints, and are prone to decision imbalance or deadlock, this invention establishes an adaptive cross-modal gating fusion mechanism. First, the operating condition expression features and environmental spatial constraint features are dimensionally concatenated, and the adaptive cross-modal gating weights are calculated through a linear layer and a Sigmoid function. These weights are dynamically adjusted by integrating the joint information of the two types of features. Second, the operating condition features are bidirectionally fused, retaining the portion of the original operating condition features multiplied by the gating weights that represents the operating condition requirements. The system retains the basic amount of environmental constraints and multiplies them by the complement of the gating weights after nonlinear transformation, so that the environmental constraint information is gradually integrated. When the operating conditions and constraints conflict, the gating weights automatically reach a balance value between zero and one. The integrated features retain the necessary power response requirements and introduce compliance considerations of environmental constraints in proportion. This enables the system to perform adaptive soft switching between operating condition requirements and environmental constraints. The system performs proportional information fusion based on the real-time conflict degree of the two types of features, which improves the robustness, adaptability and compliance of decision-making in complex general aviation scenarios.

[0026] (5) To address the problems in existing technologies where multi-source power allocation relies on fixed rule tables or simple proportional allocation, cannot cope with complex multi-condition combinations, and suffers from incomplete constraint consideration leading to equipment overload, this invention establishes an intelligent and safe mechanism for multi-source energy allocation. First, the fused operating condition characteristics, environmental constraint characteristics, task energy demand characteristics, and safety reserve characteristics are concatenated to construct a fused decision state vector. A multi-source energy allocation weight vector is generated through linear transformation and activation operations, and a normalization function ensures that the sum of the allocation ratios is always one. Second, a constraint satisfaction coefficient is introduced as a safety modulation mechanism to characterize the actual release power of the system relative to the theoretical demand. The maximum safety ratio is multiplied element-by-element by the total power demand to achieve flexible compression modulation of power commands. Finally, a truncation function is applied to each energy power output command vector to perform hard upper limit pruning, ensuring that the power command issued to each power unit does not exceed the calibrated maximum output power. This mechanism drives the adaptive generation of allocation weights through the fusion of full-dimensional business characteristics, enabling the allocation strategy to be dynamically adjusted according to the actual operating situation at each moment. The dual safety guarantee mechanism composed of constraint satisfaction coefficient and truncation function works synergistically at the soft and hard constraint levels, effectively avoiding the risk of equipment overload shutdown and ensuring the continuous and reliable operation of the functional ship's power system. Attached Figure Description

[0027] Figure 1 A flowchart illustrating a multi-condition energy management and range optimization method for a functional ship provided by the present invention; Figure 2 This is a schematic diagram of the architecture of the causal dilated convolution module provided by the present invention; Detailed Implementation

[0028] The present invention will be further described below with reference to the accompanying drawings, but this is not intended to limit the present invention in any way. Any modifications or substitutions made based on the teachings of the present invention shall fall within the protection scope of the present invention.

[0029] Example 1: A method for multi-condition energy management and range optimization of a functional ship, such as... Figure 1 As shown, it includes the following steps: A1: Collect vessel power operation sequence data, inland waterway operation task attribute data, and navigation space control geographic data, and preprocess them respectively to obtain preprocessed vessel power operation sequence data, preprocessed operation task attribute data, and preprocessed navigation space control geographic data, including: A11: Collect ship power operation sequence data. The data type is time-series numerical data, including remaining battery power, real-time diesel engine output power, real-time ship speed, and total ship electrical load. The ship power operation sequence data is preprocessed using outlier removal based on interquartile range, missing value completion using cubic spline interpolation, and Z-score standardization to obtain preprocessed ship power operation sequence data. A12: Collect inland waterway operation task attribute data. The data type is a mixture of structured numerical and text data, including the planned route, task type, task urgency level, and operation point information. Preprocess the inland waterway operation task attribute data according to field type: For the planned route field, convert the route coordinate point sequence to the WGS84 coordinate system and perform equidistant resampling along the route path to ensure consistent coordinate point sequence lengths across different routes; for the task type field, use one-hot encoding to convert the category text into a binary numerical vector; for the task urgency level field, assign incremental integer values ​​from low to high to establish an ordered numerical mapping; for the operation point information field, convert the latitude and longitude coordinates of each operation point to the WGS84 coordinate system and convert the attribute text labels of the operation points into integer numbers using label encoding, thus obtaining the preprocessed operation task attribute data. A13: Collect navigation space control geographic data, which is vector geographic data, including real-time ship positioning data, environmental sensitive area fence polygons, and waterway boundary coordinate sequences; preprocess the navigation space control geographic data using WGS84 coordinate system and Gauss-Kruger plane projection transformation, time-series positioning data resampling at equal time intervals, and timestamp alignment methods to obtain preprocessed navigation space control geographic data.

[0030] A2: Based on the preprocessed ship power operation sequence data, extract the enhanced features of operating condition transitions and the features representing operating conditions in sequence, including: A21: Based on the preprocessed ship dynamics timing data, the operating condition transition enhancement features are extracted through causal dilation convolution and a condition transition sensitivity enhancement mechanism. The calculation method is as follows: ; ; ; in, For the first The basic characteristics of the operating conditions at any given time, where t is the time index. This is a causal dilation convolution operation. For the first Preprocessed ship dynamics timing data at specific times. For the first The first-order time difference component of the operating characteristics at any given time. For the first The basic characteristics of the operating conditions at any given time. For the first The characteristics of the transition between operating conditions at any given moment are enhanced. For the Sigmoid function, It is a multilayer perceptron. For Hadamah accumulation; A22: Based on the enhanced characteristics of the transition between operating conditions, the characteristics of the operating condition expression are calculated. The calculation method is as follows: ; in, For the first The characteristics of the working condition at any given moment.

[0031] Specifically, for functional ships facing frequent abrupt changes in operating conditions and requiring stronger dynamic response capabilities under complex mission scenarios, this invention also provides a method for calculating operating condition representation features based on adaptive residual enhancement to replace step A22. The calculation method is as follows: ; ; ; ; in, The attention-weighted working condition transition feature at time t. The residual connection condition characteristics at time t are... Let be the feature fusion weight at time t.

[0032] A3: Based on the preprocessed navigation space control geographic data, environmental protection space constraint features are generated through ship positioning codes and spatial distance constraint feature extraction. The calculation method is as follows: ; ; ; ; ; in, Let be the ship's positioning coding feature at time t. This refers to the real-time ship positioning data in the preprocessed navigation space control geographic data at time t. Let be the distance vector from the ship to each boundary vertex of the channel at time t. Calculated using Euclidean distance. Let be the directed distance from a point to the i-th polygon, where is zero inside the polygon and positive outside, and is the polygon index. Where m is the total number of polygons; These are the first to nth coordinates in the channel boundary coordinate sequence from the preprocessed navigation space control geographic data, where n is the number of channel boundary coordinates. Contribute a weight vector to the channel boundary. For the Softplus function, For temperature parameters, Let i be the scalar quantity of the constraint strength of the i-th environmentally sensitive area. It is an exponential function. The polygon representing the fence of the i-th environmentally sensitive area in the preprocessed geographic data for airspace control. The standard deviation of distance attenuation. Let be the environmental space constraint characteristics at time t. Let m be the constraint strength scalar of the m-th environmentally sensitive area. This is a weight vector projection operation.

[0033] A4: Based on the characteristics of operating conditions and environmental constraints, a bidirectional complementary fusion is performed through an adaptive cross-modal gating fusion layer to obtain the operating condition characteristics after fusing environmental constraint information and the environmental constraint characteristics after fusing operating condition information. The calculation method is as follows: ; ; ; in, For the first Adaptive cross-modal gating weights at time intervals, For linear layers, This is a dimension splicing operation. To integrate operating condition characteristics with environmental constraints information, It is the hyperbolic tangent function. The environmental constraints are characteristics after integrating operating condition information.

[0034] A5: Based on the preprocessed task attribute data, the operating condition characteristics after integrating environmental constraint information, and the environmental constraint characteristics after integrating operating condition information, the following are generated sequentially: operating condition-aware task priority coding characteristics, task energy demand characteristics, and emergency reserve characteristics, including: A51: Based on the preprocessed job attribute data, generate task priority coding features for work condition awareness. The calculation method is as follows: ; ; ; ; in, For the first Task attribute encoding features at any given time. For the first Preprocessed job attribute data at any given time. For the first The condition-constraint fusion characteristics at any given time. For the first Time-based task-condition interaction weights For the first Task priority encoding features based on real-time operational conditions; A52: Based on the task priority coding characteristics perceived by operating conditions, generate basic energy demand characteristics, operating condition modulation coefficients, environmental constraint modulation coefficients, and task energy demand characteristics. The calculation method is as follows: ; ; ; ; in, For the first The basic energy requirements at any given moment. It is the hyperbolic tangent function. For the first Modulation coefficient at time of operation For average pooling, For the first Environmental constraint modulation coefficient at any time For the first The characteristics of task energy requirements at any given moment; A53: Based on the basic energy demand characteristics, task energy demand characteristics, operating condition characteristics after integrating environmental constraints, and environmental constraints after integrating operating condition information, emergency reserve characteristics are generated. The calculation method is as follows: ; ; ; in, For the first The task urgency index at any given moment For the first Emergency reserve ratio vector at any given time. For the first The characteristics of emergency reserves at all times.

[0035] A6: Based on the operating condition characteristics after integrating environmental constraint information, the environmental constraint characteristics after integrating operating condition information, the task energy demand characteristics, and the emergency reserve characteristics, a multi-source energy allocation weight vector and a power output command vector for each energy source are generated sequentially, including: A61: Based on the operating condition characteristics after integrating environmental constraints, the environmental constraints characteristics after integrating operating condition information, and the task energy demand characteristics, calculate the multi-source energy allocation weight vector. The calculation method is as follows: ; ; ; in, For the first The fusion decision state vector at each time step. As an intermediate feature in decision-making, Assign weight vectors to multi-source energy. For the Softmax function; A62: Based on the multi-source energy allocation weight vector, generate the power output command vector for each energy source. The calculation method is as follows: ; ; ; ; ; in, For the first The characteristics of the modulated basic energy demand at any given time. For the first Total power demand at any given time For the first The constraint satisfaction coefficient at time t. For the first The time-constrained power demand after modulation For the first The power output command vector of each energy source at each time point; in the power output command vector of each energy source, a truncation function is used to ensure that the value of each component is not greater than the upper limit of the power output of the corresponding energy source.

[0036] In this embodiment, the specific parameter settings, training methods, and optimization strategies for each neural network module are as follows: Causal dilated convolution module: Employs a multi-layer stacked structure with progressively increasing dilation rates to extract features from preprocessed ship dynamics timing data. The first layer's dilated convolution kernel size is set to 3, with a dilation rate of 1; the second layer's dilation rate is 2; and the third layer's dilation rate is 4. The number of convolution output channels is set to 32, 64, and 128 respectively. Each convolution layer is followed by a ReLU activation function to introduce a nonlinear transformation, such as... Figure 2 As shown; Multilayer Perceptron (MLP) module: The number of hidden layer neurons is flexibly configured according to the feature dimensions of different functional locations; the MLP for extracting work condition expression features adopts a two-layer hidden layer structure, with 64 neurons in the first hidden layer and 32 neurons in the second hidden layer, and the output layer maintains the same dimension as the input layer; the MLP for task attribute encoding also adopts a two-layer structure, with 48 and 32 neurons in the hidden layer respectively, and the output dimension is aligned with the fused feature dimension; all hidden layers use the ReLU activation function, and batch normalization operations are added between layers to accelerate network convergence; Gated fusion layer: The linear layer in the adaptive cross-modal gating weight calculation is set as a single-layer structure, which maps the spliced ​​working condition features and environmental constraint features to scalar values, followed by the Sigmoid function; Configuration of parameterized functions: Temperature parameters in Softplus functions The initial value was set to 1.0 and remained constant during training to ensure the smoothness of the channel boundary constraints; the distance decay standard deviation... Set to 100 meters; The entire network adopts an end-to-end supervised learning training paradigm. The historical operational dataset of functional ships is divided into training, validation, and test sets in a ratio of 7:2:1. The Adam optimizer is used for parameter updates, with an initial learning rate set to 0.0005 and a step-decay strategy, decreasing to 0.1 times the original rate at the 30th and 60th epochs, respectively. The batch size is 32, and the total number of training epochs is 100. The loss function adopts a multi-task weighted loss design: the weight coefficients of energy allocation loss (using smooth L1 loss), constraint satisfaction loss (using binary cross-entropy), and operating condition prediction loss (using mean squared error) are set to 0.5, 0.3, and 0.2, respectively. The weighted combination achieves balanced optimization of multi-dimensional objectives. An early stopping mechanism is introduced to prevent overfitting; training automatically stops when the total loss on the validation set no longer improves for 10 consecutive epochs.

[0037] Example 2: This invention also discloses a multi-condition energy management and range optimization system for functional ships, comprising: Data acquisition unit: Collects ship power operation sequence data, inland waterway operation task attribute data, and navigation space control geographic data, and preprocesses them respectively to obtain preprocessed ship power operation sequence data, preprocessed operation task attribute data, and preprocessed navigation space control geographic data; Ship dynamic condition extraction unit: Based on the preprocessed ship dynamic operation sequence data, extract condition transition enhancement features and condition expression features in sequence; Environmental space constraint extraction unit: Based on the preprocessed navigation space control geographic data, environmental space constraint features are generated by extracting ship positioning codes and spatial distance constraint features; Operating condition and environmental protection fusion unit: Based on the operating condition expression characteristics and environmental protection space constraint characteristics, bidirectional complementary fusion is performed through an adaptive cross-modal gating fusion layer to obtain the operating condition characteristics after fusing environmental protection constraint information and the environmental protection constraint characteristics after fusing operating condition information. Task feature extraction unit: Based on the preprocessed task attribute data, the working condition features after integrating environmental constraint information, and the environmental constraint features after integrating working condition information, the task priority coding features, task energy demand features, and emergency reserve features are generated sequentially for working condition perception. Energy allocation unit: Based on the operating condition characteristics after integrating environmental protection constraint information, the environmental protection constraint characteristics after integrating operating condition information, the task energy demand characteristics, and the emergency reserve characteristics, it sequentially generates a multi-source energy allocation weight vector and a power output command vector for each energy source.

[0038] It should be noted that the sequence numbers of the above embodiments of the present invention are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0039] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0040] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for multi-condition energy management and range optimization of a functional ship, characterized in that, Includes the following steps: A1: Collect ship power operation sequence data, inland waterway operation task attribute data, and navigation space control geographic data, and preprocess them respectively to obtain preprocessed ship power operation sequence data, preprocessed operation task attribute data, and preprocessed navigation space control geographic data. A2: Based on the preprocessed ship power operation sequence data, extract the working condition transition enhancement features and working condition expression features in sequence; Step A2 includes: A21: Based on the preprocessed ship dynamics timing data, the operating condition transition enhancement features are extracted through causal dilation convolution and a condition transition sensitivity enhancement mechanism. The calculation method is as follows: ； ； ； in, For the first The basic characteristics of the operating conditions at any given time, where t is the time index. This is a causal dilation convolution operation. For the first Preprocessed ship dynamics timing data at specific times. For the first The first-order time difference component of the operating characteristics at any given time. For the first The basic characteristics of the operating conditions at any given time. For the first The characteristics of the transition between operating conditions at any given moment are enhanced. For the Sigmoid function, It is a multilayer perceptron. For Hadamah accumulation; A22: Based on the enhanced characteristics of the transition between operating conditions, the characteristics of the operating condition expression are calculated. The calculation method is as follows: ； in, For the first The characteristics of the working condition at any given moment; A3: Based on the preprocessed navigation space control geographic data, environmental protection space constraint features are generated through ship positioning codes and spatial distance constraint feature extraction; A4: Based on the operating condition expression characteristics and environmental space constraint characteristics, a bidirectional complementary fusion is performed through an adaptive cross-modal gating fusion layer to obtain the operating condition characteristics after fusing environmental constraint information and the environmental constraint characteristics after fusing operating condition information; step A4 includes: A4: Based on the characteristics of operating conditions and environmental constraints, a bidirectional complementary fusion is performed through an adaptive cross-modal gating fusion layer to obtain the operating condition characteristics after fusing environmental constraint information and the environmental constraint characteristics after fusing operating condition information. The calculation method is as follows: ； ； ； in, For the first Adaptive cross-modal gating weights at time points. For linear layers, This is a dimension splicing operation. To integrate operating condition characteristics with environmental constraints information, It is the hyperbolic tangent function. Environmental constraints characteristics after integrating operating condition information; A5: Based on the preprocessed task attribute data, the working condition characteristics after integrating environmental constraint information, and the environmental constraint characteristics after integrating working condition information, the task priority coding characteristics, task energy demand characteristics, and emergency reserve characteristics of working condition perception are generated in sequence. A6: Based on the operating condition characteristics after integrating environmental protection constraint information, the environmental protection constraint characteristics after integrating operating condition information, the task energy demand characteristics, and the emergency reserve characteristics, a multi-source energy allocation weight vector and a power output command vector for each energy source are generated sequentially.

2. The method for multi-condition energy management and range optimization of functional ships according to claim 1, characterized in that, Step A1 includes: A11: Collect ship power operation sequence data. The data type is time-series numerical data, including remaining battery power, real-time diesel engine output power, real-time ship speed, and total ship electrical load. The ship power operation sequence data is preprocessed using outlier removal based on interquartile range, missing value completion using cubic spline interpolation, and Z-score standardization to obtain preprocessed ship power operation sequence data. A12: Collect inland waterway operation task attribute data. The data type is a mixture of structured numerical and text data, including the planned route, task type, task urgency level, and operation point information. Preprocess the inland waterway operation task attribute data according to field type: For the planned route field, convert the route coordinate point sequence to the WGS84 coordinate system and perform equidistant resampling along the route path to ensure consistent coordinate point sequence lengths across different routes; for the task type field, use one-hot encoding to convert the category text into a binary numerical vector; for the task urgency level field, assign incremental integer values ​​from low to high to establish an ordered numerical mapping; for the operation point information field, convert the latitude and longitude coordinates of each operation point to the WGS84 coordinate system and convert the attribute text labels of the operation points into integer numbers using label encoding, thus obtaining the preprocessed operation task attribute data. A13: Collect navigation space control geographic data, which is vector geographic data, including real-time ship positioning data, environmental sensitive area fence polygons, and waterway boundary coordinate sequences; preprocess the navigation space control geographic data using WGS84 coordinate system and Gauss-Kruger plane projection transformation, time-series positioning data resampling at equal time intervals, and timestamp alignment methods to obtain preprocessed navigation space control geographic data.

3. The method for multi-condition energy management and range optimization of functional ships according to claim 1, characterized in that, Step A3 includes: ； ； ； ； ； in, Let be the ship's positioning coding feature at time t. This refers to the real-time ship positioning data in the preprocessed navigation space control geographic data at time t. Let be the distance vector from the ship to each boundary vertex of the channel at time t. Calculated using Euclidean distance. Let be the directed distance from a point to the i-th polygon, where is zero inside the polygon and positive outside, and is the polygon index. , where m is the total number of polygons; These are the first to nth coordinates in the channel boundary coordinate sequence from the preprocessed navigation space control geographic data, where n is the number of channel boundary coordinates. Contribute a weight vector to the channel boundary. For the Softplus function, For temperature parameters, Let i be the scalar quantity of the constraint strength of the i-th environmentally sensitive area. It is an exponential function. The polygon representing the fence of the i-th environmentally sensitive area in the preprocessed geographic data for airspace control. The standard deviation of distance attenuation. Let be the environmental space constraint characteristics at time t. Let m be the constraint strength scalar of the m-th environmentally sensitive area. This is a weight vector projection operation.

4. The method for multi-condition energy management and range optimization of functional ships according to claim 1, characterized in that, Step A5 includes: A51: Generate condition-aware task priority coding features based on the preprocessed job attribute data; A52: Based on the task priority coding characteristics perceived by the working condition, generate basic energy demand characteristics, working condition modulation coefficient, environmental protection constraint modulation coefficient, and task energy demand characteristics; A53: Based on the basic energy demand characteristics, mission energy demand characteristics, operating condition characteristics after integrating environmental constraints, and environmental constraints after integrating operating condition information, emergency reserve characteristics are generated.

5. The method for multi-condition energy management and range optimization of functional ships according to claim 4, characterized in that, Step A5 includes: First, the preprocessed task attribute data is processed by a multilayer perceptron to obtain task attribute encoding features. Then, the condition features after fusing environmental constraint information and the environmental constraint features after fusing condition information are concatenated and processed by a linear layer to obtain condition-constraint fusion features. Next, the task attribute encoding features and the condition-constraint fusion features are concatenated and processed by a linear layer and a Sigmoid function to obtain task-condition interaction weights. Finally, the task attribute encoding features and the task-condition interaction weights are multiplied by a Hadamard product, and the result is added to the result of the condition-constraint fusion features and the Hadamard product of 1 minus the task-condition interaction weights to obtain the condition-aware task priority encoding features. The task priority encoding features of the working condition perception are processed sequentially by a linear layer and a hyperbolic tangent function to obtain the basic energy demand features; the working condition features after incorporating environmental constraint information are processed sequentially by average pooling, a linear layer, and a sigmoid function to obtain the working condition modulation coefficients; the environmental constraint features after incorporating working condition information are processed sequentially by average pooling, a linear layer, and a sigmoid function to obtain the environmental constraint modulation coefficients; the basic energy demand features, the working condition modulation coefficients, and the environmental constraint modulation coefficients are multiplied by a Hadamard product, and then added to the result of the task priority encoding features of the working condition perception processed by a multilayer perceptron to obtain the task energy demand features; The task urgency index is obtained by concatenating the task energy demand characteristics, the operating condition characteristics after integrating environmental constraints, and the environmental constraints characteristics after integrating operating condition information, and then processing them sequentially through average pooling, a linear layer, and a Sigmoid function. The task urgency index is then processed sequentially through a linear layer and a Sigmoid function to obtain the emergency reserve ratio vector. The emergency reserve ratio vector is then multiplied by the basic energy demand characteristics, and added to the result of subtracting the Hadamard product of the emergency reserve ratio vector and the task energy demand characteristics processed by the multilayer perceptron from 1, thus obtaining the emergency reserve characteristics.

6. The method for multi-condition energy management and range optimization of functional ships according to claim 5, characterized in that, Step A6 includes: The fused decision state vector is obtained by concatenating the operating condition characteristics after integrating environmental constraint information, the environmental constraint characteristics after integrating operating condition information, the task energy demand characteristics, and the emergency reserve characteristics. The fused decision state vector is then processed sequentially through a linear layer and a ReLU function to obtain intermediate decision features. Finally, the intermediate decision features are processed sequentially through a linear layer and a Softmax function to obtain a multi-source energy allocation weight vector. The modulated basic energy demand characteristic is obtained by performing a Hadamard product on the basic energy demand characteristic, the operating condition modulation coefficient, and the environmental constraint modulation coefficient. The modulated basic energy demand characteristic is then processed sequentially by a linear layer and a ReLU function to obtain the total power demand. The operating condition characteristic after incorporating environmental constraint information is concatenated with the environmental constraint characteristic after incorporating operating condition information and then processed sequentially by average pooling, a linear layer, and a Sigmoid function to obtain the constraint satisfaction coefficient. The total power demand is multiplied element-wise by the constraint satisfaction coefficient to obtain the constrained modulated power demand. The multi-source energy allocation weight vector and the constrained modulated power demand are then processed by a Hadamard product to obtain the power output command vector of each energy source. In the power output command vector of each energy source, a truncation function is used to ensure that the value of each component is not greater than the power output upper limit of the corresponding energy source.

7. A multi-condition energy management and range optimization system for a functional ship, characterized in that, include: Data acquisition unit: Collects ship power operation sequence data, inland waterway operation task attribute data, and navigation space control geographic data, and preprocesses them respectively to obtain preprocessed ship power operation sequence data, preprocessed operation task attribute data, and preprocessed navigation space control geographic data; Ship dynamic condition extraction unit: Based on the preprocessed ship dynamic operation sequence data, extract condition transition enhancement features and condition expression features in sequence; Environmental space constraint extraction unit: Based on the preprocessed navigation space control geographic data, environmental space constraint features are generated by extracting ship positioning codes and spatial distance constraint features; Operating condition and environmental protection fusion unit: Based on the operating condition expression characteristics and environmental protection space constraint characteristics, bidirectional complementary fusion is performed through an adaptive cross-modal gating fusion layer to obtain the operating condition characteristics after fusing environmental protection constraint information and the environmental protection constraint characteristics after fusing operating condition information. Task feature extraction unit: Based on the preprocessed task attribute data, the working condition features after integrating environmental constraint information, and the environmental constraint features after integrating working condition information, the task priority coding features, task energy demand features, and emergency reserve features are generated sequentially for working condition perception. Energy allocation unit: Based on the operating condition characteristics after integrating environmental constraint information, the environmental constraint characteristics after integrating operating condition information, the task energy demand characteristics, and the emergency reserve characteristics, it sequentially generates a multi-source energy allocation weight vector and a power output command vector for each energy source; in order to realize the multi-operating condition energy management and endurance optimization method for functional ships as described in any one of claims 1-6.

Images