A ship energy efficiency optimization control method driven by artificial intelligence for navigation working condition perception

Abstract

The present application relates to the technical field of ship energy efficiency control, and discloses a ship energy efficiency optimization control method driven by artificial intelligence for navigation working condition sensing, which comprises the following steps: adopting a sliding time window mode to slide and intercept and encode process a standardized heading working condition state time sequence to obtain a navigation working condition state feature, and generating a working condition boundary wandering feature and a wave direction change feature by using an artificial intelligence time sequence network model; correcting a working condition determination result of frequent jump of working condition based on the working condition boundary wandering feature and the wave direction change feature, and matching to obtain a basic ship energy efficiency control strategy; generating a working condition stability index and a wave direction disturbance compensation factor, and performing weighted compensation and boundary constraint processing on the basic ship energy efficiency control strategy. The present application realizes intelligent feature extraction, boundary jump working condition smoothing correction and energy efficiency strategy dynamic compensation output, fully suppresses control buffeting caused by working condition switching, and achieves adaptive smooth energy-saving navigation control of the ship.

Description

Technical Field

[0001] This invention relates to data acquisition and control, and more particularly to the field of ship energy efficiency control, specifically to an artificial intelligence-driven ship energy efficiency optimization control method based on navigation condition perception. Background Technology

[0002] With increasingly stringent environmental regulations in the global shipping industry, energy conservation, emission reduction, and refined energy efficiency management have become core necessities for shipping operations. During navigation, waves couple with the ship's course, creating different relative wave direction conditions that directly alter hull resistance and propulsion system efficiency. Conventional industry control approaches rely on environmental sensor data to identify typical conditions such as bow-angled waves, cross waves, and stern-angled waves, matching corresponding optimal speeds, main engine loads, and adjustable pitch, among other energy-saving control parameters. However, in complex navigation scenarios such as near-shore island maneuvers and frequent steering changes, the relative wave direction can easily linger at the boundary between two operating conditions. Even minor wave direction disturbances can cause frequent jumps in the operating condition identification results, leading to continuous adjustments in control commands and resulting in chattering. This significantly reduces energy efficiency and exacerbates power equipment wear. Therefore, an intelligent energy efficiency optimization solution is urgently needed that can suppress jumps in operating condition boundary identification and output stable control commands.

[0003] There are already numerous research achievements related to ship energy efficiency optimization technologies. Among them, patent CN110110943B discloses a comprehensive intelligent optimization management system and optimization method for fleet energy efficiency based on big data. This solution relies on big data statistical algorithms to mine navigation environment and ship energy efficiency historical data, complete the prediction of navigation environment and ship operating conditions, build a joint dynamic optimization model of route and speed based on the prediction results, and combine swarm intelligence algorithms to solve the model to output a unified speed and route decision for the fleet. The optimization instructions are then remotely issued to ships in operation, realizing remote control of the overall energy efficiency of the fleet. This patent focuses on long-term rolling optimization of macro-level routes and speeds at the fleet level, and completes global planning based on massive historical data. It can take into account the spatiotemporal differences of the environment to improve the overall energy-saving level of the fleet. However, it is only applicable to long-term route scheduling of the fleet and cannot perform fine perception and dynamic control correction for the real-time instantaneous wave direction and operating conditions of individual ships.

[0004] Besides, other traditional technologies for identifying single-ship operating conditions generally use fixed angle thresholds to discretely divide navigation conditions, collect wave direction, wave height, and ship speed sensor data, and then simply classify them, directly retrieving the corresponding fixed control parameters. This type of solution is simple to build and the control logic is easy to implement. However, it relies solely on sensor values ​​at a single moment to determine the operating condition, fails to explore the continuous evolution characteristics of wave direction within a short time window, and lacks a quantitative identification mechanism for the hovering state at the boundary of the operating condition.

[0005] To address this issue, this invention proposes an AI-driven method for optimizing ship energy efficiency control based on navigation conditions. This method suppresses frequent flipping of the operating condition label when the relative wave direction hovers around the operating condition boundary threshold, thereby improving the stability and feasibility of energy-saving control for ships under complex sea conditions. Summary of the Invention

[0006] This invention proposes an AI-driven method for optimizing ship energy efficiency control based on navigation condition perception. Step S1 collects multi-source navigation condition sensor data and performs standardization processing to generate a standardized time-series sequence of heading and operating conditions. This unifies the dimensions and time-series benchmarks of various sensor data and can completely preserve the ship's wave exposure and power equipment's full-dimensional time-series observation information. Step S2 uses a sliding time window to extract encoding and combines it with an AI time-series network model to analyze the navigation condition characteristics, simultaneously outputting the characteristics of boundary oscillations and wave direction changes. This allows for refined mining of short-term wave direction fluctuations and long-term evolution patterns, accurately quantifying the degree to which the ship approaches the operating condition boundary and the wave direction oscillation trend, solving the problems of not being able to identify minute boundary oscillations and the difficulty in quantifying and extracting the dynamic evolution patterns of wave direction. Technical issues: Step S3 determines the degree of operating condition jump by using the boundary lingering characteristics of the operating condition, and corrects the jump judgment result by combining the wave direction change characteristics to generate a vibration-suppressing operating condition trajectory. Then, it matches the corresponding basic ship energy efficiency control strategy, which can avoid frequent switching of operating conditions caused by instantaneous noise and solve the technical problem of discrete abrupt changes in the operating condition trajectory. Step S4 generates operating condition stability index and wave direction disturbance compensation factor by operating condition boundary lingering characteristics and wave direction change characteristics. It implements weighted compensation and multi-source sensor parameter boundary constraints on the basic ship energy efficiency control strategy. The compensation intensity can be dynamically adjusted according to the degree of operating condition stability. At the same time, it relies on the measured sea state and dynamic parameters to limit the control output range, which solves the technical problem of large abrupt changes in control parameters in the boundary section and control commands exceeding the safe operating range of the equipment.

[0007] To achieve the above objectives, the present invention provides an artificial intelligence-driven ship energy efficiency optimization control method based on navigation condition perception, comprising the following steps: S1: Collect multi-source operating condition sensor data during the ship's navigation process, standardize the multi-source operating condition sensor data, and generate a standardized heading operating condition state time sequence. S2: The standardized heading condition state time sequence is trunculated and encoded using a sliding time window method. The navigation condition state features of the ship in the time period corresponding to the sliding time window are extracted. The navigation condition state features are received using an artificial intelligence time series network model, and the condition boundary lingering features and wave direction change features are generated. S3: Extract the ship's operating condition determination result during the time period corresponding to the sliding time window from the operating condition boundary wandering characteristics, and identify whether the operating condition determination result changes frequently. Correct the operating condition determination result with frequent changes based on the operating condition boundary wandering characteristics and wave direction change characteristics to obtain a stable and continuous vibration suppression operating condition trajectory. Obtain the basic ship energy efficiency control strategy based on the vibration suppression operating condition trajectory matching. S4: Based on the boundary lingering characteristics and wave direction change characteristics of the operating conditions, generate the operating condition stability index and wave direction disturbance compensation factor, and perform weighted compensation and boundary constraint processing on the basic ship energy efficiency control strategy to obtain the ship energy efficiency control strategy during the ship's navigation process.

[0008] As a further improvement of the present invention: Further, in step S1, the multi-source operating condition sensor data is standardized to generate a standardized heading operating condition state time sequence, including: S11: The multi-source operating condition sensing data includes the original sequence of gyrocompass heading angle, the original sequence of wave direction dominated by wave radar, the original sequence of significant wave height, the original sequence of ship speed from speedometer, the original sequence of main engine load setting, and the original sequence of adjustable pitch setting. S12: Based on the acquisition time range of the multi-source working condition sensor data, generate a globally unified sampling time sequence, extract the data values ​​of the multi-source working condition sensor data at the sampling time in the globally unified sampling time sequence, and obtain globally time-aligned multi-source working condition sensor data. S13: Perform two-layer outlier removal processing on the global time-aligned multi-source operating condition sensing data, and use linear interpolation to perform interpolation completion processing on the removed outliers to obtain standardized multi-source operating condition sensing data. S14: Extract the original sequence of the gyrocompass heading angle and the original sequence of the dominant wave direction of the wave radar from the standardized multi-source working condition sensing data, calculate the difference sequence of the extracted sequences, and map the difference in the difference sequence to 0° to 180° as the relative encountered wave direction sequence, and calculate the first-order difference sequence of the relative encountered wave direction sequence as the relative encountered wave direction change rate sequence. S15: The original sequence of effective wave height, the original sequence of speedometer speed, the original sequence of main engine load setting, the original sequence of adjustable pitch setting, and the relative encounter wave direction sequence and the relative encounter wave direction change rate sequence calculated in step S14 are spliced ​​together to form a standardized heading condition state time sequence from the standardized multi-source operating condition sensor data.

[0009] Further, in step S2, a sliding time window method is used to perform sliding truncation and encoding processing on the standardized heading condition state time sequence, extracting the navigation condition state features of the ship during the time period corresponding to the sliding time window, including: S21: Construct a sliding time window with a window length of 2k+1, and take any sequence value in the standardized heading condition state time sequence as the center of the sliding time window to obtain the standardized heading condition state sequence segment associated with the sequence value. S22: Extract the relative encounter wave direction sequence and the relative encounter wave direction change rate sequence from the standardized heading condition state sequence segment, and perform dual-channel encoding on the relative encounter wave direction sequence and the relative encounter wave direction change rate sequence to obtain the navigation condition state features of the sliding time window associated with the standardized heading condition state sequence segment.

[0010] Furthermore, step S2 utilizes an artificial intelligence temporal network model to receive navigation condition status characteristics and generate condition boundary lingering characteristics and wave direction change characteristics, and also includes: S23: The artificial intelligence temporal network model includes an input layer, a front-end temporal convolutional layer, a bidirectional gated temporal correlation layer, a dual-branch mapping layer, and an output layer. The input layer receives navigation condition status features. S24: The front-end temporal convolutional layer uses sliding convolution to extract features of the navigation condition status and obtain local wave texture feature vectors; Specifically, the front-end temporal convolutional layer includes a first-level temporal convolutional layer Conv1 and a second-level temporal convolutional layer Conv2, and the navigation condition state features... The feature extraction formula is: ; ; ; in, Indicates the characteristics of navigation operating conditions The convolution calculation results of the first-level temporal convolutional layer Conv1, This represents the local wave-shaped texture feature vector. This indicates the activation function (LeakyReLU is selected by default). These represent the trainable convolutional weight matrix and bias in the first-level temporal convolutional layer Conv1, respectively. These represent the trainable convolutional weight matrix and bias in the second-level temporal convolutional layer Conv2, respectively. express The residual correction term, Represents the residual fusion coefficient. Indicates the characteristics of navigation operating conditions The convolution calculation results of the first-level temporal convolutional layer Conv1, Represents sequence value The associated navigation condition characteristics of the sliding time window This represents the nth sequence value in the standardized heading condition time sequence. N represents the sequence length of multi-source operating condition sensor data; S25: The bidirectional gated temporal association layer performs forward and reverse bidirectional GRU gated temporal encoding on the local wave-shaped texture feature vector to obtain the comprehensive representation vector of the center of the sliding time window; S26: The dual-branch mapping layer includes a working condition attribution branch and a boundary response branch. The working condition attribution branch receives the comprehensive representation vector and outputs the working condition category weight at the center of the sliding time window. The boundary response branch receives the comprehensive representation vector and the working condition category weight respectively and outputs the working condition boundary proximity. The formula for calculating the proximity of the working condition boundary is: ; ; in, Indicates the characteristics of navigation operating conditions Center of associated sliding time window The boundary response value, This represents the nth sequence value in the standardized heading condition time sequence. Indicates The navigation condition characteristics of the time period corresponding to the center of the sliding time window. This represents the weight vector of the boundary response branch, where T denotes the transpose. Indicates the characteristics of navigation operating conditions Center of associated sliding time window The comprehensive representation vector, These represent the centers of the sliding time windows, respectively. Weights for bow-dipping waves, transverse waves, and stern-dipping waves. This represents the weighted aggregation penalty coefficient. Indicates the characteristics of navigation operating conditions Center of associated sliding time window The proximity of the working condition boundary It represents an exponential function with the natural constant as the base. The working condition boundary proximity is used to quantify how close the working condition judgment result of the sampling time associated with the center of the sliding time window is to the working condition boundary. The closer the working condition boundary proximity is to 1, the closer the working condition judgment result is to the boundary between the two working condition categories, and the more likely the working condition category will jump. S27: The output layer generates the working condition boundary lingering feature and wave direction change feature of the center of the sliding time window based on the working condition category weight and the working condition boundary proximity of the center of the sliding time window. The working condition boundary lingering feature includes the working condition category weight and working condition lingering parameter of the center of the sliding time window in the three working condition categories.

[0011] Further, in step S3, the ship's operating condition determination result during the time period corresponding to the sliding time window is extracted from the operating condition boundary lingering features, and the frequent changes in the operating condition determination result are identified, including: Extract the weights of the three working condition categories of the center of the sliding time window in the working condition boundary wandering features, and use them as the working condition determination results of the ship in the time period corresponding to the sliding time window; Extract the working condition lingering parameter at the center of the sliding time window from the working condition boundary lingering feature. If the working condition lingering parameter is higher than the preset lingering threshold, mark the sampling time associated with the center of the sliding time window as the jump time. The percentage of abrupt changes in the globally unified sampling time series is statistically analyzed. If the percentage is higher than the allowable percentage threshold, it indicates that the ship's operating condition judgment results in the globally unified sampling time series are frequently abruptly changed.

[0012] Furthermore, step S3, which corrects the frequently changing operating condition determination results based on the boundary oscillation characteristics and wave direction change characteristics to obtain a stable and continuous vibration-suppressing operating condition trajectory, also includes: Extract the wave direction change characteristics at the center of the sliding time window associated with the jump moment, and generate the working condition category of the jump moment based on the wave direction change characteristics; For sampling times that are not abrupt changes, the working condition category with the highest weight is selected as the working condition category at the corresponding sampling time center of the sliding time window; The selected operating condition categories are sorted according to the order of sampling times to obtain a stable and continuous vibration suppression operating condition trajectory.

[0013] Furthermore, the basic ship energy efficiency control strategy obtained in step S3 based on the vibration suppression trajectory matching also includes: Extract the operating condition categories contained in the vibration suppression operating condition trajectory, and obtain the basic energy efficiency control strategy corresponding to the extracted operating condition category as the basic ship energy efficiency control strategy.

[0014] Furthermore, step S4, which generates operating condition stability indices and wave direction disturbance compensation factors based on the operating condition boundary lingering characteristics and wave direction change characteristics, includes: S41: Obtain the working condition lingering parameters at the center of the sliding time window associated with different working condition categories, and calculate the average value of the working condition lingering parameters for different working condition categories; S42: Obtain the wave direction change characteristics at the center of the sliding time window, and perform absolute value processing on the wave direction change characteristics to calculate the mean value of the wave direction change characteristics after absolute value processing of the center of the sliding time window associated with different working conditions. S43: Based on the calculation results of S41 and S42, the working condition stability index and wave direction disturbance compensation factor for different working conditions are calculated.

[0015] Specifically, the calculation formulas for the operating condition stability index and wave direction disturbance compensation factor for the different operating condition categories are as follows: ; ; in, This represents the stability index for the j-th operating condition category. The first to third operating conditions are bow-dipping waves, transverse waves, and stern-dipping waves, respectively. This represents the mean of the condition oscillation parameter for the j-th condition category. This represents the wave disturbance compensation factor for the j-th operating condition category. Indicates the disturbance sensitivity coefficient. It represents the mean value of the wave direction change characteristics after taking the absolute value of the center of the sliding time window associated with the j-th working condition category.

[0016] Furthermore, step S4 involves weighted compensation and boundary constraint processing of the basic ship energy efficiency control strategy to obtain the ship energy efficiency control strategy during navigation, and also includes: S44: Based on the operating condition stability index and wave disturbance compensation factor of the different operating condition categories, the weighting coefficients of the different operating condition categories are calculated. S45: Based on the weighting coefficients of different operating condition categories, the basic energy efficiency control strategies for the corresponding operating condition categories in the basic ship energy efficiency control strategies are weighted to obtain the weighted basic ship energy efficiency control strategies. S46: Extract the original sequence of effective wave height, the original sequence of speedometer speed, the original sequence of main engine load setting, and the original sequence of adjustable pitch setting from the standardized heading condition state time sequence, and perform boundary constraint processing on the weighted basic ship energy efficiency control strategy to obtain the ship energy efficiency control strategy during the ship's navigation process.

[0017] Compared with existing technologies, this invention proposes an artificial intelligence-driven ship energy efficiency optimization control method based on navigation condition perception. This technology has the following beneficial effects: First, this invention marks the jump moments by using operating condition lingering parameters, and determines the overall degree of operating condition jump by combining the global proportion threshold of the jump moments, thus distinguishing between stable navigation and operating condition boundary jitter sections. During non-jump periods, the operating condition category with the highest weight is directly selected to ensure that the control strategy under normal operating conditions conforms to the ship's actual wave-affected state. For jump periods, wave direction change characteristics are introduced to constrain the operating condition category, locking the operating condition according to the long-term unidirectional evolution trend of the waves, preventing instantaneous noise from causing back-and-forth switching of operating conditions, and suppressing frequent label jumps from a temporal trend perspective.

[0018] Meanwhile, this invention statistically analyzes the mean values ​​of the wave direction variation characteristics after absolute value processing for three different operating conditions. It independently solves for the operating condition stability index and wave direction disturbance compensation factor for each operating condition, distinguishing the differences in boundary fluctuations under different wave conditions and avoiding over- or under-compensation in a single operating condition due to globally uniform parameters. The operating condition stability index uses a difference mapping between 1 and the mean value of the wave direction variation parameters, resulting in a regular value range that intuitively reflects the stability of each operating condition category. The wave direction disturbance compensation factor is coupled with the mean value of the wave direction variation characteristics after absolute value processing and the mean value of the wave direction variation parameters, simultaneously considering both the intensity of wave direction oscillations and the degree of boundary wave direction variation. The disturbance sensitivity coefficient is uniformly calibrated and adjustable in engineering. Under stable operating conditions, the wave direction disturbance compensation factor approaches 1, while under boundary fluctuation conditions, the compensation intensity is automatically increased, achieving differentiated smooth correction for different operating conditions, reducing the amplitude of control parameter jumps, and balancing navigation energy efficiency and the stability of power equipment regulation. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating a ship energy efficiency optimization control method driven by artificial intelligence for navigation condition perception, provided in an embodiment of the present invention. Figure 2 This is a structural diagram of an artificial intelligence temporal network model provided in an embodiment of the present invention; Figure 3 This is an experimental comparison diagram provided for one embodiment of the present invention. Detailed Implementation

[0020] The realization of the objectives, functional characteristics, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0021] This invention provides an AI-driven ship energy efficiency optimization control method based on navigation condition perception. The executing entity of this AI-driven ship energy efficiency optimization control method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this invention: a server, a terminal, etc. In other words, the AI-driven ship energy efficiency optimization control method can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0022] Reference Figure 1 as well as Figure 2 Embodiment 1 of the present invention is as follows: An artificial intelligence-driven ship energy efficiency optimization control method based on navigation condition perception, the method comprising: S1: Collect multi-source operating condition sensor data during the ship's navigation process, standardize the multi-source operating condition sensor data, and generate a standardized heading operating condition time sequence.

[0023] Specifically, step S1 involves standardizing the multi-source operating condition sensor data to generate a standardized heading operating condition time series, including: As an embodiment of the present invention, during ship navigation, multi-source operating condition sensor data is periodically acquired through an onboard data acquisition network. Specifically, a gyrocompass is used to output the ship's heading angle relative to geographic north in real time, forming a raw sequence of gyrocompass heading angles in chronological order; a wave radar is used to scan and analyze the sea surface wave field around the ship, obtaining the dominant wave propagation direction and forming a raw sequence of dominant wave directions from the wave radar, while simultaneously acquiring effective wave height data within a statistical time window and forming a raw sequence of effective wave heights; a speedometer is used to measure the ship's speed relative to the water or land, forming a raw sequence of speedometer speeds; the ship's power control system reads the current load setting value output by the main engine controller, forming a raw sequence of main engine load settings; and an adjustable pitch propulsion control system reads the current pitch angle setting value of the propeller blades, forming a raw sequence of adjustable pitch settings. Specifically, the gyrocompass (including fiber optic compasses) is a core navigation device that all seagoing vessels are required to be equipped with. It is a standard marine instrument that can output the ship's heading angle in real time, 24 / 7. Ships are equipped with speedometers, which are divided into underwater acoustic type, electromagnetic type and other types. They can measure the ship's speed relative to the water in real time and output the continuous sailing speed. Specifically, the gyrocompass heading angle sequence is used to characterize the ship's heading change state and heading stability; the wave radar dominant wave direction sequence is used to characterize the main wave propagation direction and sea state direction change characteristics; the significant wave height sequence is used to characterize the current wave disturbance intensity and sea state level; the speedometer speed sequence is used to characterize the ship's actual operating speed; the main engine load setting sequence is used to characterize the power output demand level of the ship's propulsion system; and the adjustable pitch setting sequence is used to characterize the current thrust adjustment state and propulsion efficiency adjustment state of the propeller. All these original sequences collectively reflect the ship's heading state, sea state, and power control state, providing basic data support for subsequent navigation condition identification, extraction of condition boundary lingering features, and generation of ship energy efficiency control strategies.

[0024] S11: The multi-source operating condition sensing data includes the original sequence of gyrocompass heading angle, the original sequence of wave direction dominated by wave radar, the original sequence of significant wave height, the original sequence of ship speed from speedometer, the original sequence of main engine load setting, and the original sequence of adjustable pitch setting. S12: Based on the acquisition time range of the multi-source working condition sensor data, generate a globally unified sampling time sequence, extract the data values ​​of the multi-source working condition sensor data at the sampling time in the globally unified sampling time sequence, and obtain globally time-aligned multi-source working condition sensor data. Specifically, during ship navigation, multi-source operating condition sensor data is periodically collected, and the collection time range is extracted and configured according to a preset time interval. The acquisition time range is divided into a globally unified sampling time sequence, which consists of consecutive sampling times, with a time interval between adjacent sampling times. The default acquisition period for the multi-source operating condition sensor data is set to 5 minutes. S13: Perform two-layer outlier removal processing on the global time-aligned multi-source operating condition sensing data, and use linear interpolation to perform interpolation completion processing on the removed outliers to obtain standardized multi-source operating condition sensing data. Specifically, the two-layer outlier removal process includes: marking data values ​​that exceed the parameter range as outliers, identifying outliers in the sequence using the 3sigma principle, and removing the identified outliers. S14: Extract the original sequence of the gyrocompass heading angle and the original sequence of the dominant wave direction of the wave radar from the standardized multi-source working condition sensing data, calculate the difference sequence of the extracted sequences, and map the difference in the difference sequence to 0° to 180° as the relative encountered wave direction sequence, and calculate the first-order difference sequence of the relative encountered wave direction sequence as the relative encountered wave direction change rate sequence. As an embodiment of the present invention, the formula for generating the relative encountered wave direction sequence is: ; ; in, Indicates the relative wave direction sequence. Indicates the relative encounter wave direction sequence The N data values ​​in the data. Indicates the relative encounter wave direction sequence The nth data value represents the relative encounter wave angle at the nth sampling time. The relative encounter wave angle is the smallest spatial angle between the wave's direction of approach and the ship's current heading, used to describe the direction from which the wave approaches the ship. N represents the sequence length of multi-source operating condition sensor data. This represents the nth data value in the original sequence of the gyrocompass heading angle after standardization. This represents the nth data value in the original sequence of the dominant wave direction from the wave radar after standardization. Indicates selection The minimum value in; The formula for generating the relative encounter wave direction change rate sequence is: ; ; in, This represents the sequence of relative rates of change of the direction of the encountered wave. Represents the relative wave direction change rate sequence The N data values ​​in the data. Represents the relative wave direction change rate sequence The nth data value in This represents the time interval between the sampling times associated with adjacent data values; It should be noted that this invention relies on the standardized gyrocompass heading angle original sequence and the wave radar dominant wave direction original sequence to calculate the minimum included angle to obtain the relative encountered wave direction in the range of 0° to 180°, avoiding the numerical change problem caused by the 360° angle crossing zero jump, accurately quantifying the real-time wave-receiving position of the ship, and simultaneously obtaining the first-order difference of the wave direction to obtain the rate of change sequence, intuitively representing the instantaneous fluctuation speed of the wave direction.

[0025] S15: The original sequence of effective wave height, the original sequence of speedometer speed, the original sequence of main engine load setting, the original sequence of adjustable pitch setting, and the relative encounter wave direction sequence and the relative encounter wave direction change rate sequence calculated in step S14 are spliced ​​together to form a standardized heading condition state time sequence from the standardized multi-source operating condition sensor data.

[0026] Specifically, the standardized heading condition state time sequence is represented as follows: ; ; in, This represents a time sequence of standardized heading and operating conditions. Represents the time sequence of standardized heading condition states. The N sequence values ​​in the data. Represents the time sequence of standardized heading condition states. The nth sequence value in the sequence, These represent the nth data value in the following sequences, respectively: relative wave direction sequence, relative wave direction change rate sequence, standardized original effective wave height sequence from multi-source operating condition sensor data, original speedometer speed sequence, original main engine load setting sequence, and original adjustable pitch setting sequence.

[0027] S2: The standardized heading condition state time sequence is truncated and encoded using a sliding time window method. The navigation condition state features of the ship in the time period corresponding to the sliding time window are extracted. The navigation condition state features are received using an artificial intelligence time series network model, and the condition boundary wandering features and wave direction change features are generated.

[0028] Specifically, in step S2, a sliding time window method is used to perform sliding truncation and encoding processing on the standardized heading condition state time sequence, extracting the navigation condition state features of the ship during the time period corresponding to the sliding time window, including: S21: Construct a sliding time window with a window length of 2k+1, and take any sequence value in the standardized heading condition state time sequence as the center of the sliding time window to obtain the standardized heading condition state sequence segment associated with the sequence value. Specifically, the sequence value The standardized heading condition state sequence segment of the associated sliding time window is as follows: The default setting for k is 4; S22: Extract the relative encounter wave direction sequence and the relative encounter wave direction change rate sequence from the standardized heading condition state sequence segment, and perform dual-channel encoding on the relative encounter wave direction sequence and the relative encounter wave direction change rate sequence to obtain the navigation condition state features of the sliding time window associated with the standardized heading condition state sequence segment.

[0029] Specifically, the standardized heading condition state sequence segment The encoding formula for the navigation condition status features of the associated sliding time window is as follows: ; ; ; ; in, Represents a standardized heading condition state sequence segment The associated navigation condition characteristics of the sliding time window Represents a standardized heading condition state sequence segment Middle sequence value The relative encounter wave direction encoding results, Represents a standardized heading condition state sequence segment Middle sequence value The relative encounter wave direction change rate coding results, This indicates the preset maximum rate of change (the default setting is 12° / second). This indicates the preset minimum rate of change (the default setting is -12° / second). Indicates selection The minimum value in, Indicates selection The maximum value in.

[0030] It should be noted that this invention uses a sliding time window of odd length to extract time-series segments, focusing on the sequence value at the sampling moment while also considering the evolution information of the wave direction before and after, thus completely capturing the short-term wave direction fluctuation process. Furthermore, this invention only extracts data from the relative encountered wave direction and the wave direction change rate, eliminating redundant equipment parameters and reducing the amount of feature computation. The two channels respectively employ fixed-interval normalization and amplitude-limiting scaling encoding, uniformly mapping the relative encountered wave direction and change rate to the 0-1 interval, eliminating dimensional differences and avoiding interference from extreme abrupt data in model inference. Specifically, the lightweight encoding logic is adapted to shipborne embedded hardware, ensuring real-time extraction efficiency. Normalized features amplify small boundary oscillation signals, improving the sensitivity of subsequent time-series networks in recognizing boundary oscillation behavior, reducing the probability of noise misjudgment at the feature input level, and providing regular, highly discriminative time-series feature input for accurately quantifying the degree of boundary jumps.

[0031] Step S2 utilizes an artificial intelligence temporal network model to receive navigation condition status characteristics and generate boundary lingering characteristics and wave direction change characteristics. It also includes: S23: See below Figure 2 The diagram shows the structure of an artificial intelligence temporal network model. The artificial intelligence temporal network model includes an input layer, a front-end temporal convolutional layer, a bidirectional gated temporal correlation layer, a bi-branch mapping layer, and an output layer. The input layer receives navigation condition status features. S24: The front-end temporal convolutional layer uses sliding convolution to extract features of the navigation condition status and obtain local wave texture feature vectors; Specifically, the front-end temporal convolutional layer includes a first-level temporal convolutional layer Conv1 and a second-level temporal convolutional layer Conv2, and the navigation condition state features... The feature extraction formula is: ; ; ; in, Indicates the characteristics of navigation operating conditions The convolution calculation results of the first-level temporal convolutional layer Conv1, This represents the local wave-shaped texture feature vector. This indicates the activation function (LeakyReLU is selected by default). These represent the trainable convolutional weight matrix and bias in the first-level temporal convolutional layer Conv1, respectively. These represent the trainable convolutional weight matrix and bias in the second-level temporal convolutional layer Conv2, respectively. express The residual correction term, This represents the residual fusion coefficient (default setting is 0.3). Indicates the characteristics of navigation operating conditions The convolution calculation results of the first-level temporal convolutional layer Conv1, Represents sequence value The navigation condition characteristics of the associated sliding time window; S25: The bidirectional gated temporal association layer performs forward and reverse bidirectional GRU gated temporal encoding on the local wave-shaped texture feature vector to obtain the comprehensive representation vector of the center of the sliding time window; Specifically, the local wave-shaped texture feature vector is a vector with a length of 2k+1; S26: The dual-branch mapping layer includes a working condition attribution branch and a boundary response branch. The working condition attribution branch receives the comprehensive representation vector and outputs the working condition category weight at the center of the sliding time window. The boundary response branch receives the comprehensive representation vector and the working condition category weight respectively and outputs the working condition boundary proximity. The formula for calculating the proximity of the working condition boundary is: ; ; in, Indicates the characteristics of navigation operating conditions Center of associated sliding time window The boundary response value, This represents the nth sequence value in the standardized heading condition time sequence. Indicates The navigation condition characteristics of the time period corresponding to the center of the sliding time window. This represents the weight vector of the boundary response branch, where T denotes the transpose. Indicates the characteristics of navigation operating conditions Center of associated sliding time window The comprehensive representation vector, These represent the centers of the sliding time windows, respectively. Weights for bow-dipping waves, transverse waves, and stern-dipping waves. This represents the weight aggregation penalty coefficient (default setting is 0.8). Indicates the characteristics of navigation operating conditions Center of associated sliding time window The proximity of the working condition boundary It represents an exponential function with the natural constant as the base. The working condition boundary proximity is used to quantify how close the working condition judgment result of the sampling time associated with the center of the sliding time window is to the working condition boundary. The closer the working condition boundary proximity is to 1, the closer the working condition judgment result is to the boundary between the two working condition categories, and the more likely the working condition category will jump. Specifically, the working condition attribution branch is a fully connected layer structure. The working condition attribution branch receives the comprehensive representation vector and outputs the fully connected score of the sliding time window center in the three working condition categories. The fully connected score is converted into normalized weight coefficients using the softmax function, which are used as the working condition category weights of the sliding time window center. The three working condition categories are bow-sloping waves, transverse waves, and stern-sloping waves, respectively. Specifically, when a ship is navigating, its navigation state is divided into different operating conditions based on the relative angle of the encountered wave direction. Each operating condition corresponds to a specific energy efficiency control strategy, and the area where the threshold angles of the two operating conditions intersect is the operating condition boundary. In sections of navigation with frequent turning, such as near the coast and around islands, the external wave direction is stable, but the ship's course changes slightly. This will cause the relative encountered wave direction to linger within the threshold range for a long time, resulting in a navigation state that is between two typical operating conditions.

[0032] As an embodiment of the present invention, by acquiring multiple sets of navigation condition status features and the corresponding real condition categories of the navigation condition status features, with the goal of maximizing the weight of the navigation condition status features output by the artificial intelligence time series network model in the real condition category, a training loss function is constructed, and the stochastic gradient descent algorithm is used to optimize and train the trainable parameters in the artificial intelligence time series network model. It should be noted that the two-level temporal convolutional layer of this invention introduces a convolutional residual coupling correction term for adjacent windows, which enhances the small-amplitude oscillation characteristics of the wave direction by utilizing the residual fusion coefficient, amplifies the weak fluctuation signal at the boundary of the operating condition, and effectively distinguishes between measurement noise and the reciprocating oscillation of the real boundary. Compared with ordinary temporal convolution, the sensitivity to capturing subtle changes in the boundary is significantly improved. The bidirectional GRU simultaneously mines the complete evolution law of the wave direction along the forward and reverse temporal directions, and generates a comprehensive representation vector of the center by fusing the temporal correlation information before and after the window, avoiding the loss of long-term wave direction change trend by unidirectional encoding and improving the integrity of temporal features. The dual-branch mapping layer shares the lightweight structure of the backbone network, reducing the computational power consumption of the shipborne controller inference. The boundary response branch introduces a cross-product penalty term of the operating condition weights to quantify the degree of fuzzy coexistence of multiple operating condition weights. It is combined with the feature vector to jointly calculate the boundary response value, and the boundary proximity in the 0-1 interval is obtained by mapping with the Sigmoid function. S27: The output layer generates the working condition boundary lingering feature and wave direction change feature of the center of the sliding time window based on the working condition category weight and the working condition boundary proximity of the center of the sliding time window. The working condition boundary lingering feature includes the working condition category weight and working condition lingering parameter of the center of the sliding time window in the three working condition categories.

[0033] As an embodiment of the present invention, the center of the sliding time window The generation process for the boundary hesitant characteristics and wave direction change characteristics of the working condition is as follows: S271: Get the center of the sliding time window We assign weights to the heading wave, cross wave, and stern wave categories, extract the two categories with the highest weights, and calculate the absolute value of the difference between the extracted category weights. ; S272: Combining the center of the sliding time window working condition boundary proximity The center of the sliding time window is calculated. Working condition boundary lingering characteristics : ; in, This indicates the preset weight threshold (the default value is 0.12). Represents a sign function, if ,but =1, otherwise =0, These represent the centers of the sliding time windows, respectively. Weights for the operating condition categories of bow diagonal waves, transverse waves, and stern diagonal waves; S273: Characteristics of navigation operating conditions Extract wave-oriented evolution features The mean value of the wave direction evolution characteristics was calculated. And calculate the center of the sliding time window. Wave direction change characteristics : ; in, A value of 1 indicates that the relative wave direction is continuously increasing, and the ship as a whole is shifting towards a stern-hinged wave condition. A value of -1 indicates that the relative wave direction is continuously decreasing, and the ship as a whole is shifting towards a bow-dipping wave condition. A value of 0 indicates that the wave direction changes very little and there is no stable unidirectional shift trend.

[0034] It should be noted that this invention quickly identifies the fuzzy coexistence state of operating condition weights by screening the weights of the top two operating condition categories and calculating the difference, combined with a fixed weight threshold. It then constructs operating condition boundary lingering features by combining boundary proximity, fully encapsulating the weights of the three operating condition categories as operating condition boundary lingering features. This allows for intuitive quantification of the operating condition boundary jitter risk at the current sampling moment. Based on the wave direction evolution features within a sliding time window, the mean is calculated, and segmented thresholds are set to generate wave direction change features, accurately distinguishing between continuous unidirectional offset and minor noise disturbances. These two types of features decouple the degree of operating condition ambiguity and the wave direction evolution trend, providing a clear quantitative basis for subsequent identification of abrupt operating conditions and correction of monotonic trend constraints. This effectively distinguishes between effective boundary oscillations and random measurement noise, reduces incorrect operating condition corrections, and ensures a smooth and reliable vibration suppression operating condition trajectory.

[0035] S3: Extract the ship's operating condition determination result during the time period corresponding to the sliding time window from the operating condition boundary wandering characteristics, and identify whether the operating condition determination result changes frequently. Based on the operating condition boundary wandering characteristics and wave direction change characteristics, correct the operating condition determination result with frequent changes to obtain a stable and continuous vibration suppression operating condition trajectory. Based on the vibration suppression operating condition trajectory matching, obtain the basic ship energy efficiency control strategy.

[0036] Specifically, step S3 involves extracting the ship's operational condition determination result within the sliding time window from the operational condition boundary lingering features, and identifying whether the operational condition determination result frequently changes, including: Extract the weights of the three working condition categories of the center of the sliding time window in the working condition boundary wandering features, and use them as the working condition determination results of the ship in the time period corresponding to the sliding time window; Extract the working condition lingering parameter at the center of the sliding time window from the working condition boundary lingering feature. If the working condition lingering parameter is higher than the preset lingering threshold (default setting is 0.4), then mark the sampling time associated with the center of the sliding time window as the jump time. The percentage of jump moments in the globally unified sampling time series is counted. If the percentage is higher than the allowable percentage threshold (default setting is 10%), it indicates that the ship's operating condition judgment results in the globally unified sampling time series have frequent jumps.

[0037] Furthermore, if the proportion is not higher than the allowable proportion threshold (the default setting is 10%), it means that the ship's operating condition judgment result does not frequently change within the globally unified sampling time sequence. The operating condition category with the highest weight is selected as the operating condition category at the corresponding sampling time center of the sliding time window. The selected operating condition categories are sorted according to the order of occurrence of the sampling time to obtain a stable and continuous vibration suppression operating condition trajectory.

[0038] Step S3, based on the boundary oscillation characteristics and wave direction change characteristics, corrects the condition determination result for frequent condition jumps to obtain a stable and continuous vibration suppression trajectory, and also includes: Extract the wave direction change characteristics at the center of the sliding time window associated with the jump moment, and generate the working condition category of the jump moment based on the wave direction change characteristics; Specifically, if the wave direction change feature is 1, the generated working condition category is stern-sloping wave; if the wave direction change feature is -1, the generated working condition category is bow-sloping wave; if the wave direction change feature is 0, the generated working condition category is consistent with the working condition category with the highest weight at the previous sampling time. For sampling times that are not abrupt changes, the working condition category with the highest weight is selected as the working condition category at the corresponding sampling time center of the sliding time window; The selected operating condition categories are sorted according to the order of sampling times to obtain a stable and continuous vibration suppression operating condition trajectory.

[0039] Step S3, which obtains the basic ship energy efficiency control strategy based on the vibration suppression trajectory matching, also includes: Extract the operating condition categories contained in the vibration suppression operating condition trajectory, and obtain the basic energy efficiency control strategy corresponding to the extracted operating condition category as the basic ship energy efficiency control strategy.

[0040] Specifically, if the operating condition category in the vibration suppression operating condition trajectory includes stern-sloping waves and bow-sloping waves, then the basic energy efficiency control strategies corresponding to stern-sloping waves and bow-sloping waves are extracted respectively to form a basic ship energy efficiency control strategy. The basic energy efficiency control strategy includes economic target speed, optimal main engine load, and adjustable pitch setting value.

[0041] S4: Based on the boundary lingering characteristics and wave direction change characteristics of the operating conditions, generate the operating condition stability index and wave direction disturbance compensation factor, and perform weighted compensation and boundary constraint processing on the basic ship energy efficiency control strategy to obtain the ship energy efficiency control strategy during the ship's navigation process.

[0042] Specifically, step S4, which generates operating condition stability indices and wave direction disturbance compensation factors based on the operating condition boundary lingering characteristics and wave direction change characteristics, includes: S41: Obtain the working condition lingering parameters at the center of the sliding time window associated with different working condition categories, and calculate the average value of the working condition lingering parameters for different working condition categories; S42: Obtain the wave direction change characteristics at the center of the sliding time window, and perform absolute value processing on the wave direction change characteristics to calculate the mean value of the wave direction change characteristics after absolute value processing of the center of the sliding time window associated with different working conditions. S43: Based on the calculation results of S41 and S42, the working condition stability index and wave direction disturbance compensation factor for different working conditions are calculated.

[0043] Specifically, the calculation formulas for the operating condition stability index and wave direction disturbance compensation factor for the different operating condition categories are as follows: ; ; in, This represents the stability index for the j-th operating condition category. The first to third operating conditions are bow-dipping waves, transverse waves, and stern-dipping waves, respectively. This represents the mean of the condition oscillation parameter for the j-th condition category. This represents the wave disturbance compensation factor for the j-th operating condition category. This represents the disturbance sensitivity coefficient (default setting is 0.6). It represents the mean value of the wave direction change characteristics after taking the absolute value of the center of the sliding time window associated with the j-th working condition category.

[0044] Step S4 involves weighted compensation and boundary constraint processing of the basic ship energy efficiency control strategy to obtain the ship energy efficiency control strategy during navigation, and also includes: S44: Based on the operating condition stability index and wave disturbance compensation factor of the different operating condition categories, the weighting coefficients of the different operating condition categories are calculated. Specifically, the formula for calculating the weighting coefficients for the different operating condition categories is as follows: ; in, Denotes the weighting coefficient for the j-th working condition category, where This represents the distance of the j-th working condition category from the working condition boundary. The higher the value, the farther the j-th operating condition category is from the operating condition boundary, and the closer the weighted basic ship energy efficiency control strategy is to the basic ship energy efficiency control strategy for the j-th operating condition category; otherwise, use... Increase the wave direction disturbance compensation factor to improve the smoothing constraint on the weighting coefficients; S45: Based on the weighting coefficients of different operating condition categories, the basic energy efficiency control strategies for the corresponding operating condition categories in the basic ship energy efficiency control strategies are weighted to obtain the weighted basic ship energy efficiency control strategies. Specifically, the weighted formula for the basic energy efficiency control strategy of the aforementioned operating condition category is as follows: ; in, This indicates the weighted basic ship energy efficiency control strategy. This represents the basic energy efficiency control strategy for the j-th operating condition category; S46: Extract the original sequence of effective wave height, the original sequence of speedometer speed, the original sequence of main engine load setting, and the original sequence of adjustable pitch setting from the standardized heading condition state time sequence, and perform boundary constraint processing on the weighted basic ship energy efficiency control strategy to obtain the ship energy efficiency control strategy during the ship's navigation process.

[0045] Specifically, the boundary constraint processing flow for the weighted basic ship energy efficiency control strategy is as follows: Calculate the sequence mean of the original effective wave height sequence after the standardization process. If the sequence mean is higher than the wave height safety threshold (default setting is 5 meters), then the weighted basic ship energy efficiency control strategy is reduced to 0.8 times. The data value ranges of the original speed sequence of the speedometer, the original main engine load setting sequence, and the original adjustable pitch setting sequence after standardization are extracted respectively, and are used as the value limit ranges of the economic target speed, the optimal main engine load, and the adjustable pitch setting value in the weighted basic ship energy efficiency control strategy. As an embodiment of the present invention, if the economic target speed exceeds the maximum data value in the original speed sequence of the speedometer after standardization, the economic target speed is set to the maximum data value in the original speed sequence of the speedometer after standardization.

[0046] Example 2 As an embodiment of the present invention, this invention collects real-ship navigation data of near-shore container ships, covering four typical navigation processes: island-circling maneuvers, regular straight navigation, alternating wind and waves, and frequent minor steering maneuvers. The performance of the traditional fixed-threshold operating condition judgment and control method and the ship energy efficiency optimization control method driven by artificial intelligence for navigation operating condition perception of the present invention are compared in terms of four indicators: main engine load fluctuation amplitude, power adjustment frequency, and fuel consumption per 100 kilometers. (Refer to...) Figure 3 The experimental comparison diagram shows that, through artificial intelligence time-series network model, trend constraint correction for jump conditions, and dynamic compensation control strategy for different operating conditions, this invention reduces the instantaneous fluctuation amplitude of the main engine load by 71.3%, the adjustment frequency of the power mechanism per minute by 68.9%, and the fuel consumption per 100 kilometers by 8.4% compared with traditional control schemes. It has significant technical advantages and engineering practical value in complex wave-changing navigation scenarios in nearshore waters.

[0047] It should be noted that the terms "comprising," "including," or any other variations thereof used herein 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. Unless otherwise specified, 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.

[0048] 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.

[0049] 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 ship energy efficiency optimization control method driven by artificial intelligence for navigation condition perception, characterized in that, The method includes: S1: Collect multi-source operating condition sensor data during the ship's navigation process, standardize the multi-source operating condition sensor data, and generate a standardized heading operating condition state time sequence. S2: The standardized heading condition state time sequence is trunculated and encoded using a sliding time window method. The navigation condition state features of the ship in the time period corresponding to the sliding time window are extracted. The navigation condition state features are received using an artificial intelligence time series network model, and the condition boundary lingering features and wave direction change features are generated. S3: Extract the ship's operating condition determination result during the time period corresponding to the sliding time window from the operating condition boundary wandering characteristics, and identify whether the operating condition determination result changes frequently. Correct the operating condition determination result with frequent changes based on the operating condition boundary wandering characteristics and wave direction change characteristics to obtain a stable and continuous vibration suppression operating condition trajectory. Obtain the basic ship energy efficiency control strategy based on the vibration suppression operating condition trajectory matching. S4: Based on the boundary lingering characteristics and wave direction change characteristics of the operating conditions, generate the operating condition stability index and wave direction disturbance compensation factor, and perform weighted compensation and boundary constraint processing on the basic ship energy efficiency control strategy to obtain the ship energy efficiency control strategy during the ship's navigation process.

2. The ship energy efficiency optimization control method driven by artificial intelligence for navigation condition perception as described in claim 1, characterized in that, In step S1, the multi-source operating condition sensor data is standardized to generate a standardized heading operating condition time series, including: S11: The multi-source operating condition sensing data includes the original sequence of gyrocompass heading angle, the original sequence of wave direction dominated by wave radar, the original sequence of significant wave height, the original sequence of ship speed from speedometer, the original sequence of main engine load setting, and the original sequence of adjustable pitch setting. S12: Based on the acquisition time range of the multi-source working condition sensor data, generate a globally unified sampling time sequence, extract the data values ​​of the multi-source working condition sensor data at the sampling time in the globally unified sampling time sequence, and obtain globally time-aligned multi-source working condition sensor data. S13: Perform two-layer outlier removal processing on the global time-aligned multi-source operating condition sensing data, and use linear interpolation to perform interpolation completion processing on the removed outliers to obtain standardized multi-source operating condition sensing data. S14: Extract the original sequence of the gyrocompass heading angle and the original sequence of the dominant wave direction of the wave radar from the standardized multi-source working condition sensing data, calculate the difference sequence of the extracted sequences, and map the difference in the difference sequence to 0° to 180° as the relative encountered wave direction sequence, and calculate the first-order difference sequence of the relative encountered wave direction sequence as the relative encountered wave direction change rate sequence. S15: The original sequence of effective wave height, the original sequence of speedometer speed, the original sequence of main engine load setting, the original sequence of adjustable pitch setting, and the relative encounter wave direction sequence and the relative encounter wave direction change rate sequence calculated in step S14 are spliced ​​together to form a standardized heading condition state time sequence from the standardized multi-source operating condition sensor data.

3. The ship energy efficiency optimization control method driven by artificial intelligence for navigation condition perception as described in claim 1, characterized in that, In step S2, a sliding time window method is used to perform sliding truncation and encoding processing on the standardized heading condition time sequence, extracting the navigation condition characteristics of the ship during the time period corresponding to the sliding time window, including: S21: Construct a sliding time window with a window length of 2k+1, and take any sequence value in the standardized heading condition state time sequence as the center of the sliding time window to obtain the standardized heading condition state sequence segment associated with the sequence value. S22: Extract the relative encounter wave direction sequence and the relative encounter wave direction change rate sequence from the standardized heading condition state sequence segment, and perform dual-channel encoding on the relative encounter wave direction sequence and the relative encounter wave direction change rate sequence to obtain the navigation condition state features of the sliding time window associated with the standardized heading condition state sequence segment.

4. The ship energy efficiency optimization control method driven by artificial intelligence for navigation condition perception as described in claim 3, characterized in that, Step S2 utilizes an artificial intelligence temporal network model to receive navigation condition status characteristics and generate boundary lingering characteristics and wave direction change characteristics. It also includes: S23: The artificial intelligence temporal network model includes an input layer, a front-end temporal convolutional layer, a bidirectional gated temporal correlation layer, a dual-branch mapping layer, and an output layer. The input layer receives navigation condition status features. S24: The front-end temporal convolutional layer uses sliding convolution to extract features of the navigation condition status and obtain local wave texture feature vectors; S25: The bidirectional gated temporal association layer performs forward and reverse bidirectional GRU gated temporal encoding on the local wave-shaped texture feature vector to obtain the comprehensive representation vector of the center of the sliding time window; S26: The dual-branch mapping layer includes a working condition attribution branch and a boundary response branch. The working condition attribution branch receives the comprehensive representation vector and outputs the working condition category weight at the center of the sliding time window. The boundary response branch receives the comprehensive representation vector and the working condition category weight respectively and outputs the working condition boundary proximity. The formula for calculating the proximity of the working condition boundary is: ； ； in, Indicates the characteristics of navigation operating conditions Center of associated sliding time window The boundary response value, This represents the nth sequence value in the standardized heading condition time sequence. Indicates The navigation condition characteristics of the time period corresponding to the center of the sliding time window. This represents the weight vector of the boundary response branch, where T denotes the transpose. Indicates the characteristics of navigation operating conditions Center of associated sliding time window The comprehensive representation vector, These represent the centers of the sliding time windows, respectively. Weights for bow-dipping waves, transverse waves, and stern-dipping waves. This represents the weighted aggregation penalty coefficient. Indicates the characteristics of navigation operating conditions Center of associated sliding time window The proximity of the working condition boundary This represents an exponential function with the natural constant as its base. N represents the sequence length of multi-source operating condition sensor data; S27: The output layer generates the working condition boundary lingering feature and wave direction change feature of the center of the sliding time window based on the working condition category weight and the working condition boundary proximity of the center of the sliding time window. The working condition boundary lingering feature includes the working condition category weight and working condition lingering parameter of the center of the sliding time window in the three working condition categories.

5. The ship energy efficiency optimization control method driven by artificial intelligence for navigation condition perception as described in claim 1, characterized in that, In step S3, the ship's operating condition determination result for the time period corresponding to the sliding time window is extracted from the operating condition boundary lingering features, and the frequent changes in the operating condition determination result are identified, including: Extract the weights of the three working condition categories of the center of the sliding time window in the working condition boundary wandering features, and use them as the working condition determination results of the ship in the time period corresponding to the sliding time window; Extract the working condition lingering parameter at the center of the sliding time window from the working condition boundary lingering feature. If the working condition lingering parameter is higher than the preset lingering threshold, mark the sampling time associated with the center of the sliding time window as the jump time. The percentage of abrupt changes in the globally unified sampling time series is statistically analyzed. If the percentage is higher than the allowable percentage threshold, it indicates that the ship's operating condition judgment results in the globally unified sampling time series are frequently abruptly changed.

6. The ship energy efficiency optimization control method driven by artificial intelligence for navigation condition perception as described in claim 5, characterized in that, Step S3, based on the boundary oscillation characteristics and wave direction change characteristics, corrects the condition determination result for frequent condition jumps to obtain a stable and continuous vibration suppression trajectory, and also includes: Extract the wave direction change characteristics at the center of the sliding time window associated with the jump moment, and generate the working condition category of the jump moment based on the wave direction change characteristics; For sampling times that are not abrupt changes, the working condition category with the highest weight is selected as the working condition category at the corresponding sampling time center of the sliding time window; The selected operating condition categories are sorted according to the order of sampling times to obtain a stable and continuous vibration suppression operating condition trajectory.

7. The ship energy efficiency optimization control method driven by artificial intelligence for navigation condition perception as described in claim 6, characterized in that, Step S3, which obtains the basic ship energy efficiency control strategy based on the vibration suppression trajectory matching, also includes: Extract the operating condition categories contained in the vibration suppression operating condition trajectory, and obtain the basic energy efficiency control strategy corresponding to the extracted operating condition category as the basic ship energy efficiency control strategy.

8. The ship energy efficiency optimization control method driven by artificial intelligence for navigation condition perception as described in claim 1, characterized in that, Step S4 generates operating condition stability indices and wave direction disturbance compensation factors based on the operating condition boundary lingering characteristics and wave direction change characteristics, including: S41: Obtain the working condition lingering parameters at the center of the sliding time window associated with different working condition categories, and calculate the average value of the working condition lingering parameters for different working condition categories; S42: Obtain the wave direction change characteristics at the center of the sliding time window, and perform absolute value processing on the wave direction change characteristics to calculate the mean value of the wave direction change characteristics after absolute value processing of the center of the sliding time window associated with different working conditions. S43: Based on the calculation results of S41 and S42, the working condition stability index and wave direction disturbance compensation factor for different working conditions are calculated.

9. The ship energy efficiency optimization control method driven by artificial intelligence for navigation condition perception as described in claim 8, characterized in that, Step S4 involves weighted compensation and boundary constraint processing of the basic ship energy efficiency control strategy to obtain the ship energy efficiency control strategy during navigation, and also includes: S44: Based on the operating condition stability index and wave disturbance compensation factor of the different operating condition categories, the weighting coefficients of the different operating condition categories are calculated. S45: Based on the weighting coefficients of different operating condition categories, the basic energy efficiency control strategies for the corresponding operating condition categories in the basic ship energy efficiency control strategies are weighted to obtain the weighted basic ship energy efficiency control strategies. S46: Extract the original sequence of effective wave height, the original sequence of speedometer speed, the original sequence of main engine load setting, and the original sequence of adjustable pitch setting from the standardized heading condition state time sequence, and perform boundary constraint processing on the weighted basic ship energy efficiency control strategy to obtain the ship energy efficiency control strategy during the ship's navigation process.

Images