Dynamic simulation and self-adaptive regulation and control method and system for diesel power system of unmanned ship
By constructing a control method for diesel power systems based on multimodal data fusion, the problems of response speed and adaptability of unmanned surface vessel diesel power systems in complex marine environments were solved, achieving high-precision simulation and intelligent adaptive control, thereby improving the reliability and mission efficiency of the system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ARMY ENG UNIV OF PLA
- Filing Date
- 2026-01-22
- Publication Date
- 2026-04-21
AI Technical Summary
Traditional unmanned surface vessel (USV) diesel power systems suffer from slow control response, poor adaptability, and insufficient fault diagnosis capabilities, making it difficult to meet mission requirements in complex marine environments.
By collecting multimodal data in real time, constructing physical and data-driven models, using deep learning algorithms for data fusion analysis, generating optimal control strategies, and adjusting the operating parameters of the diesel power system in real time to achieve adaptive regulation.
It improves the simulation accuracy and adaptability of the diesel power system of unmanned surface vessels, enhances the fault diagnosis capability, improves the reliability and safety of the system, and enhances the effectiveness of maritime missions.
Smart Images
Figure CN121900179A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of unmanned surface vessel (USV) power control and intelligent operation and maintenance technology, and in particular to a dynamic simulation and adaptive control method and system for an USV diesel power system. Background Technology
[0002] Unmanned surface vessels (USVs), as a new type of maritime mission platform, are playing an increasingly important role in modern maritime operations. The diesel propulsion system is a core component of USVs, and its performance directly affects their maneuverability, endurance, and mission effectiveness. However, traditional diesel propulsion control systems suffer from slow response times, poor adaptability, and insufficient fault diagnosis capabilities, making it difficult to meet the mission requirements in complex marine environments.
[0003] In traditional solutions, the control of diesel power systems is mainly based on fixed rules or simple PID control algorithms, lacking the ability to adapt to complex operating conditions. Meanwhile, traditional simulation models have limited accuracy, making it difficult to accurately predict the operating state of diesel power systems in dynamic environments. Furthermore, existing control systems typically only consider data from a single mode, lacking the fusion and utilization of multi-source information, resulting in insufficient accuracy and robustness of control decisions. Summary of the Invention
[0004] The purpose of this invention is to solve at least one technical problem in the background art and to provide a method and system for dynamic simulation and adaptive control of unmanned surface vessel diesel power systems.
[0005] To achieve the above objectives, this invention provides a method for dynamic simulation and adaptive control of an unmanned surface vessel's diesel power system, comprising: Real-time acquisition of operating parameters, environmental data, and hull status information of the unmanned surface vessel's diesel power system to generate multimodal data; A physical model of the working process of the diesel power system is constructed, and a data-driven model is built based on the historical operating data of the unmanned surface vessel (USV) diesel power system. Dynamic simulation of the diesel power system is carried out based on the physical model and the data-driven model, and the system state prediction data and corresponding confidence scores of the USV diesel power system are output. Based on system state prediction data and corresponding confidence levels, a deep learning algorithm is used to fuse and analyze multimodal data to generate the optimal control strategy for the diesel power system. The operating parameters of the diesel power system are adjusted in real time according to the obtained optimal control strategy to achieve adaptive regulation.
[0006] According to one aspect of the present invention, the real-time acquisition of operating parameters of the unmanned surface vessel's diesel propulsion system, environmental data, and hull status information to form multimodal data includes: Collect diesel engine speed, temperature, pressure, fuel consumption rate, vibration, and exhaust parameters; Collect marine environmental parameters such as temperature, humidity, wind speed, wave height, sea temperature, and visibility; Collect parameters such as speed, heading, attitude, position, acceleration, and angular velocity of the unmanned surface vessel.
[0007] According to one aspect of the present invention, the physical model for constructing the working process of the diesel power system includes: Based on the first and second laws of diesel engine thermodynamics and the Navier-Stokes equations of fluid dynamics, a mathematical model of the working process of a diesel engine is established, including: an intake system model, a compression process model, a combustion process model, an expansion process model, and an exhaust system model. The data-driven model built based on historical operating data of the unmanned surface vessel's diesel power system includes: A data-driven model based on a Long Short-Term Memory (LSTM) network and a Convolutional Neural Network (CNN) is trained using historical operating data of a diesel power system. The LSTM network is used to capture the temporal dynamic characteristics of the historical operating data of the diesel power system, and the CNN network is used to extract the spatial features of the historical operating data. The historical operating data includes diesel power system operating data under normal operating conditions, extreme operating conditions, and fault operating conditions.
[0008] According to one aspect of the present invention, the dynamic simulation of the diesel power system based on a physical model and a data-driven model, outputting system state prediction data and corresponding confidence levels of the unmanned surface vessel's diesel power system, includes: The weights of the physical model and the data-driven model are dynamically adjusted based on real-time operating conditions. At the same time, a residual connection mechanism is introduced to use the output of the physical model as one of the input features of the data-driven model. A variable step-size numerical integration algorithm is used for dynamic simulation of diesel power systems. During the simulation, the physical constraints and operational boundary conditions of the physical model of the diesel power system are monitored in real time to ensure that the simulation results are within the physically feasible range. The simulation results output includes the predicted values of the diesel power system state and the confidence assessment of the predicted values.
[0009] According to one aspect of the present invention, the step of using a deep learning algorithm to fuse and analyze multimodal data to generate an optimal control strategy for a diesel power system includes: Deep learning networks based on attention mechanisms intelligently fuse multimodal data to form fused data. The attention mechanism includes self-attention and cross-modal attention mechanisms. The self-attention mechanism is used to process time series data within the same modality and capture long-term dependencies. The cross-modal attention mechanism is used to handle the correlation between data from different modalities and achieve effective information complementarity. The fusion process adopts a strategy that combines feature-level fusion and decision-level fusion. Based on a deep Q-network architecture, a control policy network is constructed. This network is trained using fused data and outputs an initial policy. The control policy network employs a dual-network structure, including a policy network and a value network. Experience replay and target network techniques are used to improve learning stability. The reward function design comprehensively considers fuel economy, emission control, power performance, and system reliability. A designed reward shaping technique guides the agent to learn the optimal control policy. For the balance between exploration and utilization, an adaptive ε-greedy strategy is adopted to achieve intelligent exploration, focusing on exploration in the early stages of training and gradually shifting to utilizing learned knowledge in the later stages. A multi-objective particle swarm optimization algorithm is used to further optimize the initial strategy output by the control strategy network. This includes: constructing a multi-objective optimization problem to find the Pareto optimal solution among multiple objectives such as fuel economy, emission control, power performance and system reliability; considering the physical constraints and operational boundary conditions of the physical model of the diesel power system during the optimization process to ensure that the generated control strategy is within the physically feasible range.
[0010] According to one aspect of the present invention, the step of adjusting the operating parameters of the diesel power system in real time according to the obtained optimal control strategy to achieve adaptive control includes: Based on the optimal control strategy, a multivariable coordinated control algorithm is used to precisely control the fuel injection system, intake system and exhaust system of the diesel power system. A hybrid architecture combining convolutional neural networks and long short-term memory networks is adopted to monitor the operating status of diesel power systems in real time, including fault status monitoring and fault severity assessment. Fault status includes sensor failure, actuator failure, and component performance degradation. Fault severity assessment uses a fuzzy logic algorithm to classify the fault according to its impact on system safety and performance. When a fault is detected, a backup control strategy is automatically activated, employing system reconfiguration technology to maintain the basic functions of the diesel power system by reconfiguring the remaining healthy components: for sensor faults, soft sensor technology is used to estimate the faulty sensor signal based on the relevant parameters of other normal sensors; for actuator faults, a control reallocation algorithm is used to transfer the function of the faulty actuator to other actuators; for severe faults, a degraded operation mode or a safety shutdown procedure is initiated to ensure system safety.
[0011] To achieve the above objectives, the present invention also provides a dynamic simulation and adaptive control system for an unmanned surface vessel's diesel power system, comprising: The multimodal data acquisition module collects real-time operating parameters, environmental data, and hull status information of the unmanned surface vessel's diesel power system, forming multimodal data. The dynamic simulation module constructs a physical model of the working process of the diesel power system and builds a data-driven model based on the historical operating data of the unmanned surface vessel's diesel power system. Based on the physical model and the data-driven model, it performs dynamic simulation of the diesel power system and outputs the system state prediction data and corresponding confidence levels of the unmanned surface vessel's diesel power system. The AI decision-making module, based on system state prediction data and corresponding confidence levels, uses deep learning algorithms to fuse and analyze multimodal data to generate the optimal control strategy for the diesel power system. The adaptive control module adjusts the operating parameters of the diesel power system in real time according to the obtained optimal control strategy to achieve adaptive control.
[0012] To achieve the above objectives, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the dynamic simulation and adaptive control method for the unmanned surface vessel diesel power system as described above.
[0013] To achieve the above objectives, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the dynamic simulation and adaptive control method for the unmanned surface vessel diesel power system as described above.
[0014] According to the present invention, the present invention can realize high-precision dynamic simulation and intelligent adaptive control of the diesel power system of unmanned surface vessels, thereby improving the maritime mission efficiency and reliability of unmanned surface vessels. This invention improves the accuracy and robustness of control decisions by fully utilizing the operating parameters of the diesel power system, environmental data, and hull status information through multimodal data fusion. This invention employs a dynamic simulation method that combines physical models and data-driven approaches, thereby improving the accuracy and generalization capability of diesel power system simulation. The AI decision-making module based on deep reinforcement learning in this invention can adaptively learn the optimal control strategy, thereby improving the operating efficiency and adaptability of the diesel power system. The adaptive control module of this invention has fault diagnosis and fault-tolerant control capabilities, which improves the reliability and security of the system. This invention enables virtual-real interaction and collaborative optimization through a digital twin platform, providing a novel approach for the design, testing, and maintenance of unmanned surface vessel diesel power systems. This invention employs a federated learning framework to enable collaborative learning among multiple unmanned surface vessels while protecting data privacy, thereby accelerating the model optimization process. Attached Figure Description
[0015] Figure 1The flowchart illustrates a dynamic simulation and adaptive control method for an unmanned surface vessel diesel power system according to one embodiment of the present invention. Detailed Implementation
[0016] The invention will now be discussed with reference to exemplary embodiments. It should be understood that the described embodiments are merely intended to enable those skilled in the art to better understand and thus implement the invention, and are not intended to imply any limitation on the scope of the invention.
[0017] As used herein, the term "comprising" and its variations are to be interpreted as open-ended terms meaning "including but not limited to". The term "based on" is to be interpreted as "at least partially based on". The terms "one embodiment" and "an embodiment" are to be interpreted as "at least one embodiment".
[0018] Figure 1 The flowchart illustrates a method for dynamic simulation and adaptive control of an unmanned surface vessel's diesel propulsion system according to an embodiment of the present invention. Figure 1 As shown, in this embodiment, the dynamic simulation and adaptive control method for the diesel power system of an unmanned surface vessel includes: Real-time acquisition of operating parameters, environmental data, and hull status information of the unmanned surface vessel's diesel power system to generate multimodal data; A physical model of the working process of the diesel power system is constructed, and a data-driven model is built based on the historical operating data of the unmanned surface vessel (USV) diesel power system. Dynamic simulation of the diesel power system is carried out based on the physical model and the data-driven model, and the system state prediction data and corresponding confidence scores of the USV diesel power system are output. Based on system state prediction data and corresponding confidence levels, a deep learning algorithm is used to fuse and analyze multimodal data to generate the optimal control strategy for the diesel power system. The operating parameters of the diesel power system are adjusted in real time according to the obtained optimal control strategy to achieve adaptive regulation.
[0019] Furthermore, according to one embodiment of the present invention, the operating parameters of the unmanned surface vessel's diesel power system, environmental data, and hull status information are collected in real time to form multimodal data, including: Collect diesel engine speed, temperature, pressure, fuel consumption rate, vibration, and exhaust parameters; Collect marine environmental parameters such as temperature, humidity, wind speed, wave height, sea temperature, and visibility; Collect parameters such as speed, heading, attitude, position, acceleration, and angular velocity of the unmanned surface vessel.
[0020] In this embodiment, various sensors are used to acquire the aforementioned data parameters. These include: a magnetoelectric speed sensor with a measurement accuracy of ±1 rpm; a PT100 platinum resistance temperature sensor with a measurement range of -50℃ to 300℃; and a piezoresistive pressure sensor with a range of 0-5 MPa and an accuracy of 0.5%FS. Environmental data acquisition integrates multiple marine environmental monitoring sensors: an ultrasonic anemometer with a measurement range of 0-60 m / s and an accuracy of ±0.3 m / s; a wave altimeter using radar ranging with a measurement range of 0-20 m and an accuracy of ±0.1 m; and a thermistor for seawater temperature with a measurement range of -5℃ to +35℃ and an accuracy of ±0.1℃. All environmental data is transmitted via an RS485 bus and undergoes temperature and humidity compensation correction. The hull status acquisition utilizes a high-precision navigation and positioning system. The GPS positioning system employs a BeiDou / GPS dual-mode receiver with a positioning accuracy of 2 meters. The IMU (Inertial Measurement Unit) includes a three-axis gyroscope and a three-axis accelerometer, with a gyroscope range of ±300° / s and zero-bias stability of 0.5° / h. The attitude sensor uses MEMS technology, achieving roll and pitch measurement accuracy of ±0.1°. Navigation data is transmitted via Ethernet with an update frequency of 100Hz. The multimodal data acquisition module uses a unified timestamp system to ensure the synchronization of data from different sources.
[0021] After the above data parameters are collected, data preprocessing is performed, which includes steps such as outlier removal, data smoothing, and unit standardization.
[0022] Furthermore, according to one embodiment of the present invention, constructing a physical model of the working process of a diesel power system includes: Based on the first and second laws of diesel engine thermodynamics and the Navier-Stokes equations of fluid dynamics, a mathematical model of the working process of a diesel engine is established, including: an intake system model, a compression process model, a combustion process model, an expansion process model, and an exhaust system model. A data-driven model is constructed based on historical operational data of the unmanned surface vessel's diesel propulsion system, including: A data-driven model based on a Long Short-Term Memory (LSTM) network and a Convolutional Neural Network (CNN) is trained using historical operating data of a diesel power system. The LSTM network is used to capture the temporal dynamic characteristics of the historical operating data of the diesel power system, and the CNN network is used to extract the spatial features of the historical operating data. The historical operating data includes diesel power system operating data under normal operating conditions, extreme operating conditions, and fault operating conditions.
[0023] In this embodiment, the core function and processing of the above physical model include: (1) As the foundation of digital twins, it supports high-fidelity dynamic simulation. The physical model provides a mechanism-level system representation and can simulate the dynamic response of diesel engines under different operating conditions, such as transient load changes and combustion fluctuations under environmental disturbances, which is difficult to achieve with a purely data-driven model. In the digital twin platform, the physical model is coupled with the environment and the hull model to realize virtual-real interactive simulation of the entire system, which is used for early control strategy verification and optimization, and reduces the risk of actual ship debugging.
[0024] (2) Deep integration with data-driven models to improve prediction accuracy and generalization ability. The output of the physical model (such as in-cylinder temperature and pressure curves) can be used as one of the input features of the data-driven model to enhance its understanding of mechanistic behavior. Through the residual connection mechanism, the data-driven model learns the "black box" parts not covered by the physical model, such as wear, carbon deposits and other degradation effects, forming a "white box + gray box" hybrid modeling, which improves the prediction reliability under extreme or unseen operating conditions.
[0025] (3) Provide an interpretable mechanistic basis for control strategies. In the AI decision-making process, the state variables output by the physical model (such as combustion efficiency and emission indicators) can be used as mechanistic reward terms in the reward function, guiding the agent to learn control behaviors that conform to physical laws and avoiding strategies that are "feasible but unreasonable". In multi-objective optimization, the physical model provides mechanistic constraints, such as maximum combustion temperature and maximum cylinder pressure, to ensure that the optimization results are within the engineering feasible range.
[0026] (4) Mechanism analysis supporting fault diagnosis and health management. The physical model can be used to generate fault simulation datasets to simulate the impact of various faults (such as injector blockage and turbocharger efficiency reduction) on the system state, and assist data-driven models in achieving early fault identification. In fault-tolerant control, system reconfiguration simulation based on the physical model can quickly verify the effectiveness of backup control strategies.
[0027] Furthermore, according to one embodiment of the present invention, dynamic simulation of the diesel power system is performed based on a physical model and a data-driven model, outputting system state prediction data and corresponding confidence levels of the unmanned surface vessel's diesel power system, including: The weights of the physical model and the data-driven model are dynamically adjusted based on real-time operating conditions. At the same time, a residual connection mechanism is introduced to use the output of the physical model as one of the input features of the data-driven model. A variable step-size numerical integration algorithm is used for dynamic simulation of diesel power systems. During the simulation, the physical constraints and operational boundary conditions of the physical model of the diesel power system are monitored in real time to ensure that the simulation results are within the physically feasible range. The simulation results output includes the predicted values of the diesel power system state and the confidence assessment of the predicted values.
[0028] In this embodiment, the constructed physical model of the diesel engine power system working process is not only used for physical constraint verification during the simulation process, but also deeply collaborates with the data-driven model and control system in the following ways: (1) Mechanism-data hybrid prediction. In dynamic simulation, the key state variables output by the physical model (such as in-cylinder temperature and pressure) are fused with the prediction results of the data-driven model by confidence weighting to achieve state estimation with both high accuracy and strong generalization. (2) Mechanism guidance of control strategy. In the AI decision-making process, the mechanism indicators provided by the physical model (such as indicated thermal efficiency and NOx generation rate) are incorporated into the reward function design to ensure that the learned control strategy optimizes performance indicators and conforms to the combustion and emission mechanism. (3) Fault injection and fault tolerance verification. Typical fault modes are injected into the digital twin environment using the physical model to generate a labeled fault-response dataset for training the fault diagnosis model; at the same time, the feasibility of the reconstruction strategy is quickly verified by simulation during the adaptive control process. (4) Online model calibration. Based on real-time operational data, key parameters in the physical model (such as combustion heat release coefficient and turbine efficiency) are calibrated online to maintain consistency between the model and the actual system and improve the accuracy of long-term simulation and control.
[0029] Furthermore, according to one embodiment of the present invention, a deep learning algorithm is used to fuse and analyze multimodal data to generate an optimal control strategy for the diesel power system, including: Deep learning networks based on attention mechanisms intelligently fuse multimodal data to form fused data. The attention mechanism includes self-attention and cross-modal attention mechanisms. The self-attention mechanism is used to process time series data within the same modality and capture long-term dependencies. The cross-modal attention mechanism is used to handle the correlation between data from different modalities and achieve effective information complementarity. The fusion process adopts a strategy that combines feature-level fusion and decision-level fusion. Based on a deep Q-network architecture, a control policy network is constructed. This network is trained using fused data and outputs an initial policy. The control policy network employs a dual-network structure, including a policy network and a value network. Experience replay and target network techniques are used to improve learning stability. The reward function design comprehensively considers fuel economy, emission control, power performance, and system reliability. A designed reward shaping technique guides the agent to learn the optimal control policy. For the balance between exploration and utilization, an adaptive ε-greedy strategy is adopted to achieve intelligent exploration, focusing on exploration in the early stages of training and gradually shifting to utilizing learned knowledge in the later stages. A multi-objective particle swarm optimization algorithm is used to further optimize the initial strategy output by the control strategy network. This includes: constructing a multi-objective optimization problem to find the Pareto optimal solution among multiple objectives such as fuel economy, emission control, power performance and system reliability; considering the physical constraints and operational boundary conditions of the physical model of the diesel power system during the optimization process to ensure that the generated control strategy is within the physically feasible range.
[0030] Specifically, in this embodiment, fusion analysis is performed based on multimodal data, and the control strategy is optimized, including: 1. Multimodal data fusion analysis (1) Intelligent weight allocation based on cross-modal attention mechanism For dynamic importance identification of different modal data: Through a cross-modal attention mechanism, the contribution of different modal data under different operating conditions is automatically identified and weighted. For example, under severe sea conditions, the influence weight of wave height and hull attitude data on fuel injection strategy will be significantly increased; in fault diagnosis scenarios, the weight of vibration and exhaust parameters will be enhanced.
[0031] Dual focus on temporal and spatial features: self-attention mechanism handles temporal dependencies within the same modality, such as speed sequence, while cross-modal attention mechanism integrates data from different sources, such as the influence of ambient temperature on diesel engine temperature, to achieve spatiotemporal feature synergy.
[0032] (2) Hybrid architecture that integrates feature level and decision level Feature-level fusion: Using convolutional neural networks to extract spatial features of different modal data, such as vibration spectrum and exhaust component distribution, to generate a unified feature representation.
[0033] Decision-level fusion: In the control policy network, multimodal features and predicted state confidence are combined to generate decision outputs with redundancy and robustness, avoiding decision failure caused by anomalies in a single data source.
[0034] (3) Multimodal anomaly detection and fault tolerance mechanism In response to sensor failures or data loss, the system uses soft sensor technology and multimodal correlation modeling to reconstruct missing signals in real time, such as estimating temperature based on rotational speed and vibration, to ensure the integrity of the fused input.
[0035] 2. Control strategy generation and optimization based on multimodal fusion (1) Multimodal sensing design of reward function The reward function is not only based on traditional indicators such as fuel economy and emissions, but also introduces modal consistency rewards to encourage strategies to obtain higher rewards when multimodal data such as environment, hull, and power are coordinated and consistent, thereby improving the overall coordination of the system.
[0036] (2) Environmental perception adjustment of adaptive exploration strategy During reinforcement learning training, the exploration rate is dynamically adjusted according to the environmental modality, such as sea state level. In calm sea conditions, the focus is on utilizing existing strategies, while in severe sea conditions, exploration is increased to learn adaptive behaviors.
[0037] (3) Modal constraint integration in multi-objective optimization In multi-objective particle swarm optimization, multi-modal constraints are introduced, such as the influence of ambient temperature on emissions and the constraint of hull attitude on fuel efficiency, to ensure that the control strategy is feasible at both the physical and environmental levels.
[0038] Furthermore, according to one embodiment of the present invention, the operating parameters of the diesel power system are adjusted in real time according to the obtained optimal control strategy to achieve adaptive regulation, including: Based on the optimal control strategy, a multivariable coordinated control algorithm is used to simultaneously and precisely control the fuel injection system (corresponding to the combustion process model), intake system and exhaust system of the diesel power system. A hybrid architecture combining convolutional neural networks and long short-term memory networks is adopted to monitor the operating status of diesel power systems in real time, including fault status monitoring and fault severity assessment. Fault status includes sensor failure, actuator failure, and component performance degradation. Fault severity assessment uses a fuzzy logic algorithm to classify the fault according to its impact on system safety and performance. When a fault is detected, a backup control strategy is automatically activated, employing system reconfiguration technology to maintain the basic functions of the diesel power system by reconfiguring the remaining healthy components: for sensor faults, soft sensor technology is used to estimate the faulty sensor signal based on the relevant parameters of other normal sensors; for actuator faults, a control reallocation algorithm is used to transfer the function of the faulty actuator to other actuators; for severe faults, a degraded operation mode or a safety shutdown procedure is initiated to ensure system safety.
[0039] According to the above-described scheme of the present invention, the present invention can realize high-precision dynamic simulation and intelligent adaptive control of the diesel power system of unmanned surface vessels, thereby improving the maritime mission efficiency and reliability of unmanned surface vessels. This invention improves the accuracy and robustness of control decisions by fully utilizing the operating parameters of the diesel power system, environmental data, and hull status information through multimodal data fusion. This invention employs a dynamic simulation method that combines physical models and data-driven approaches, thereby improving the accuracy and generalization capability of diesel power system simulation. The AI decision-making module based on deep reinforcement learning in this invention can adaptively learn the optimal control strategy, thereby improving the operating efficiency and adaptability of the diesel power system. The adaptive regulation of this invention has fault diagnosis and fault-tolerant control capabilities, which improves the reliability and safety of the system.
[0040] Furthermore, to achieve the above objectives, the present invention also provides a dynamic simulation and adaptive control system for an unmanned surface vessel (USV) diesel power system that implements the above-mentioned dynamic simulation and adaptive control method for USV diesel power systems, comprising: The multimodal data acquisition module collects real-time operating parameters, environmental data, and hull status information of the unmanned surface vessel's diesel power system, forming multimodal data. The dynamic simulation module constructs a physical model of the working process of the diesel power system and builds a data-driven model based on the historical operating data of the unmanned surface vessel's diesel power system. Based on the physical model and the data-driven model, it performs dynamic simulation of the diesel power system and outputs the system state prediction data and corresponding confidence levels of the unmanned surface vessel's diesel power system. The AI decision-making module, based on system state prediction data and corresponding confidence levels, uses deep learning algorithms to fuse and analyze multimodal data to generate the optimal control strategy for the diesel power system. The adaptive control module adjusts the operating parameters of the diesel power system in real time according to the obtained optimal control strategy to achieve adaptive control.
[0041] Furthermore, according to one embodiment of the present invention, the multimodal data acquisition module includes three core units: a diesel engine operating parameter acquisition unit, an environmental data acquisition unit, and a hull status acquisition unit. The diesel engine operating parameter acquisition unit monitors the engine's operating status in real time through a sensor network installed on key parts of the diesel engine. The speed sensor is a magnetoelectric sensor with a measurement accuracy of ±1 rpm; the temperature sensor is a PT100 platinum resistance thermometer with a measurement range of -50℃ to 300℃; and the pressure sensor is a piezoresistive sensor with a range of 0-5 MPa and an accuracy of 0.5%FS. These sensors transmit data to the central processing unit in real time via a CAN bus.
[0042] The environmental data acquisition unit integrates multiple marine environmental monitoring sensors. The meteorological sensor uses an ultrasonic anemometer with a measurement range of 0-60 m / s and an accuracy of ±0.3 m / s; the wave altimeter uses radar wave ranging with a measurement range of 0-20 m and an accuracy of ±0.1 m; and the seawater temperature sensor uses a thermistor with a measurement range of -5℃ to +35℃ and an accuracy of ±0.1℃. All environmental data is transmitted via an RS485 bus and undergoes temperature and humidity compensation correction.
[0043] The hull status acquisition unit employs a high-precision navigation and positioning system. The GPS positioning system uses a BeiDou / GPS dual-mode receiver with a positioning accuracy of 2 meters. The IMU (Inertial Measurement Unit) includes a three-axis gyroscope and a three-axis accelerometer; the gyroscope has a range of ±300° / s and a zero-bias stability of 0.5° / h. The attitude sensor uses MEMS technology, with roll and pitch measurement accuracy of ±0.1°. Navigation data is transmitted via Ethernet at a 100Hz update frequency.
[0044] The multimodal data acquisition module employs a unified timestamp system to ensure the synchronization of data from different sources. Data processing is performed on the acquired data parameters, including outlier removal, data smoothing, and unit standardization. The preprocessed multimodal data is then transmitted to the dynamic simulation module via a high-speed data bus, providing accurate input data for system simulation and control decisions. This module features a highly reliable design, employing redundant sensor configurations and fault self-diagnosis capabilities. When a sensor fails, the system automatically switches to a backup sensor and issues a maintenance alarm, ensuring the continuity and reliability of data acquisition.
[0045] Furthermore, according to one embodiment of the present invention, the workflow of the dynamic simulation module includes five main steps: physical model construction, data-driven modeling, model fusion and optimization, high-precision dynamic simulation, and simulation result output. Specifically, the physical model construction is based on the first and second laws of diesel engine thermodynamics and the Navier-Stokes equations of fluid dynamics, establishing a mathematical model of the diesel engine's working process, including an intake system model, a compression process model, a combustion process model, an expansion process model, and an exhaust system model. The physical model parameters are calibrated using diesel engine design parameters and experimental data to ensure the basic accuracy of the model.
[0046] Data-driven modeling employs deep learning methods, using LSTM (Long Short-Term Memory) networks to capture the temporal dynamic characteristics of the diesel power system and CNN (Convolutional Neural Network) to extract spatial features from multi-sensor data. Training data is derived from historical operational data, including system behavior data under normal, extreme, and fault conditions, ensuring the model's generalization ability.
[0047] Model fusion and optimization employs a confidence-based adaptive weighted fusion algorithm, dynamically adjusting the weights of the physical model and the data-driven model according to the current operating conditions. Under steady-state conditions, the physical model has a higher weight; under transient conditions, the data-driven model has a higher weight. A residual connection mechanism is also introduced, using the output of the physical model as one of the input features of the data-driven model.
[0048] High-precision dynamic simulation employs a variable-step numerical integration algorithm. When the system state changes drastically, the step size is automatically reduced to ensure simulation accuracy; when the state is stable, the step size is increased to improve computational efficiency. During the simulation, the physical constraints and boundary conditions of the system's physical model are monitored in real time to ensure that the simulation results remain within physically feasible limits.
[0049] The simulation results output not only include predicted system states but also provide confidence assessments of the predictions, offering a basis for the AI decision-making module. Simultaneously, the simulation module possesses self-diagnostic capabilities, detecting model drift and triggering timely model retraining. This dynamic simulation module, through deep integration of physical and data-driven models, maintains the interpretability of the physical model while possessing the adaptability of the data-driven model, significantly improving the accuracy and real-time performance of diesel power system simulations.
[0050] Furthermore, according to one embodiment of the present invention, the algorithm architecture of the AI decision-making module includes three core units: a multimodal data fusion unit, a deep reinforcement learning unit, and a decision optimization unit. The multimodal data fusion unit employs a deep learning network based on an attention mechanism to intelligently fuse multi-source data from different sensors. This unit includes self-attention and cross-modal attention mechanisms, enabling it to automatically learn the importance weights of data from different modalities. The self-attention mechanism is used to process time-series data within the same modality, capturing long-term dependencies; the cross-modal attention mechanism is used to handle the correlations between data from different modalities, achieving effective information complementarity. The fusion process employs a strategy combining feature-level fusion and decision-level fusion to ensure the accuracy and robustness of the fusion results.
[0051] The deep reinforcement learning unit is based on a deep Q-network (DQN) architecture to construct a control policy network. This network employs a dual-network structure, including a policy network and a value network, and enhances learning stability through experience replay and target network techniques. The reward function design comprehensively considers multiple objectives such as fuel economy, emission control, power performance, and system reliability, guiding the agent to learn the optimal control policy through a carefully designed reward shaping technique. Regarding the balance between exploration and exploitation, the system adopts an adaptive ε-greedy strategy for intelligent exploration, focusing on exploration in the early stages of training and gradually shifting towards utilizing learned knowledge in later stages.
[0052] The decision optimization unit employs a multi-objective particle swarm optimization algorithm to further optimize the initial strategy output by the deep reinforcement learning unit. This unit constructs a multi-objective optimization problem, seeking a Pareto optimal solution among multiple objectives such as fuel economy, emission control, power performance, and system reliability. During the optimization process, the physical constraints and operational boundary conditions of the system's physical model are considered to ensure that the generated control strategy is physically feasible and reliable.
[0053] In this embodiment, the AI decision-making module employs a learning strategy combining online learning and offline training. During the offline phase, historical data is used for pre-training to establish a basic control strategy; during the online phase, fine-tuning is performed based on real-time operational data to adapt to changes in specific operating conditions. The module has a self-evaluation function, enabling it to monitor the decision-making effect in real time and perform self-optimization, ensuring consistently excellent control performance in complex marine environments. This AI decision-making module, through the organic combination of multimodal data fusion, deep reinforcement learning, and multi-objective optimization, achieves intelligent adaptive control of the diesel power system, significantly improving the unmanned surface vessel's mission capabilities and execution efficiency in complex marine environments.
[0054] Furthermore, according to one embodiment of the present invention, the control logic of the adaptive control module includes three core parts: a parameter adjustment unit, a fault diagnosis unit, and a fault-tolerant control unit. It achieves precise control of the diesel power system through intelligent decision-making and closed-loop feedback. The parameter adjustment unit is responsible for executing the optimal control strategy generated by the AI decision-making module. This unit employs a multivariable coordinated control algorithm to precisely control multiple actuators, including the fuel injection system, intake system, and exhaust system. Fuel injection control uses high-pressure common rail technology, with an injection pressure up to 2000 bar and an injection accuracy of ±0.1 mg. Injection timing control is based on crankshaft position signals, achieving a control accuracy of ±0.1° crankshaft angle. Boost pressure control uses electric wastegate and variable geometry turbine technology, with a response time of less than 50 ms. All control parameters are optimized in real time to ensure optimal performance under different operating conditions.
[0055] The fault diagnosis unit utilizes a deep learning-based anomaly detection algorithm to monitor the real-time operating status of the diesel power system. This unit employs a hybrid architecture combining convolutional neural networks and long short-term memory networks, enabling it to simultaneously process time-series data and instantaneous features. It monitors the diesel power system's operating status in real-time, including fault condition monitoring and fault severity assessment. Fault conditions include various fault modes such as sensor failure, actuator failure, and component performance degradation, with a diagnostic accuracy exceeding 99%. Fault severity assessment uses a fuzzy logic algorithm to classify faults based on their impact on system safety and performance, providing a basis for fault-tolerant control decisions.
[0056] The fault-tolerant control unit automatically activates a backup control strategy upon detecting a fault. This unit employs system reconfiguration technology to maintain basic system functionality by reconfiguring remaining healthy components: for sensor faults, soft sensor technology is used to estimate the faulty sensor signal based on other relevant parameters; for actuator faults, a control reallocation algorithm is used to transfer the function of the faulty actuator to other actuators; for severe faults, a degraded operating mode or a safety shutdown procedure is initiated to ensure system safety. The fault-tolerant control uses a model predictive control algorithm to optimize control performance while satisfying system constraints.
[0057] In this embodiment, the closed-loop feedback mechanism of the adaptive control module monitors the control effect in real time and feeds the performance data back to the AI decision-making module for control strategy optimization. The feedback data includes multiple indicators such as fuel economy, emission levels, power performance, vibration, and noise, and the control strategy is continuously improved through a multi-objective optimization algorithm. This module also has a self-learning function, enabling it to optimize control parameters based on historical control experience, improving control accuracy and response speed. Through the organic combination of intelligent parameter adjustment, precise fault diagnosis, and reliable fault-tolerant control, this adaptive control module achieves efficient, reliable, and safe operation of the diesel power system, significantly enhancing the adaptability and combat effectiveness of the unmanned surface vessel in complex marine environments.
[0058] Furthermore, according to one embodiment of the present invention, the dynamic simulation and adaptive control system for the unmanned surface vessel's diesel power system also includes: a digital twin platform, which realizes a virtual mapping of the unmanned surface vessel's diesel power system. Through the digital twin platform, the diesel power system can be tested and optimized in a virtual environment, reducing the cost and risk of actual testing.
[0059] The dynamic simulation and adaptive control system for unmanned surface vessel (USV) diesel power systems employs a federated learning framework to enable collaborative learning among multiple USVs. While protecting the data privacy of each USV, the system accelerates the optimization process of the AI model and improves the overall performance of the system through model parameter aggregation.
[0060] According to the above-described scheme of the present invention, the present invention can realize high-precision dynamic simulation and intelligent adaptive control of the diesel power system of unmanned surface vessels, thereby improving the maritime mission efficiency and reliability of unmanned surface vessels. This invention improves the accuracy and robustness of control decisions by fully utilizing the operating parameters of the diesel power system, environmental data, and hull status information through multimodal data fusion. This invention employs a dynamic simulation method that combines physical models and data-driven approaches, thereby improving the accuracy and generalization capability of diesel power system simulation. The AI decision-making module based on deep reinforcement learning in this invention can adaptively learn the optimal control strategy, thereby improving the operating efficiency and adaptability of the diesel power system. The adaptive control module of this invention has fault diagnosis and fault-tolerant control capabilities, which improves the reliability and security of the system. This invention enables virtual-real interaction and collaborative optimization through a digital twin platform, providing a novel approach for the design, testing, and maintenance of unmanned surface vessel diesel power systems. This invention employs a federated learning framework to enable collaborative learning among multiple unmanned surface vessels while protecting data privacy, thereby accelerating the model optimization process.
[0061] Furthermore, to achieve the above objectives, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the dynamic simulation and adaptive control method for the unmanned surface vessel diesel power system as described above.
[0062] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the dynamic simulation and adaptive control method for the unmanned surface vessel diesel power system as described above.
[0063] Those skilled in the art will recognize that the modules and algorithm steps described in conjunction with the embodiments disclosed herein can be implemented using electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
[0064] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described apparatus and equipment can be referred to the corresponding process in the foregoing method implementation, and will not be repeated here.
[0065] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.
[0066] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the objectives of the embodiments of the present invention, depending on actual needs.
[0067] In addition, the functional modules in the embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.
[0068] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the sending / receiving methods of various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.
[0069] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in this application.
[0070] It should be understood that the sequence number of each step in the invention and its embodiments does not absolutely imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
Claims
1. A method for dynamic simulation and adaptive control of diesel power systems for unmanned surface vessels, characterized in that, include: Real-time acquisition of operating parameters, environmental data, and hull status information of the unmanned surface vessel's diesel power system to generate multimodal data; A physical model of the working process of the diesel power system is constructed, and a data-driven model is built based on the historical operating data of the unmanned surface vessel (USV) diesel power system. Dynamic simulation of the diesel power system is carried out based on the physical model and the data-driven model, and the system state prediction data and corresponding confidence scores of the USV diesel power system are output. Based on system state prediction data and corresponding confidence levels, a deep learning algorithm is used to fuse and analyze multimodal data to generate the optimal control strategy for the diesel power system. The operating parameters of the diesel power system are adjusted in real time according to the obtained optimal control strategy to achieve adaptive regulation.
2. The dynamic simulation and adaptive control method for the diesel power system of an unmanned surface vessel according to claim 1, characterized in that, The real-time acquisition of operating parameters, environmental data, and hull status information of the unmanned surface vessel's diesel propulsion system forms multimodal data, including: Collect diesel engine speed, temperature, pressure, fuel consumption rate, vibration, and exhaust parameters; Collect marine environmental parameters such as temperature, humidity, wind speed, wave height, sea temperature, and visibility; Collect parameters such as speed, heading, attitude, position, acceleration, and angular velocity of the unmanned surface vessel.
3. The dynamic simulation and adaptive control method for the diesel power system of an unmanned surface vessel according to claim 1, characterized in that, The physical model for constructing the working process of the diesel power system includes: Based on the first and second laws of diesel engine thermodynamics and the Navier-Stokes equations of fluid dynamics, a mathematical model of the working process of a diesel engine is established, including: an intake system model, a compression process model, a combustion process model, an expansion process model, and an exhaust system model. The data-driven model built based on historical operating data of the unmanned surface vessel's diesel power system includes: A data-driven model based on a Long Short-Term Memory (LSTM) network and a Convolutional Neural Network (CNN) is trained using historical operating data of a diesel power system. The LSTM network is used to capture the temporal dynamic characteristics of the historical operating data of the diesel power system, and the CNN network is used to extract the spatial features of the historical operating data. The historical operating data includes diesel power system operating data under normal operating conditions, extreme operating conditions, and fault operating conditions.
4. The dynamic simulation and adaptive control method for the diesel power system of an unmanned surface vessel according to claim 1, characterized in that, The dynamic simulation of the diesel power system based on the physical model and data-driven model outputs system state prediction data and corresponding confidence levels for the unmanned surface vessel's diesel power system, including: The weights of the physical model and the data-driven model are dynamically adjusted based on real-time operating conditions. At the same time, a residual connection mechanism is introduced to use the output of the physical model as one of the input features of the data-driven model. A variable step-size numerical integration algorithm is used for dynamic simulation of diesel power systems. During the simulation, the physical constraints and operational boundary conditions of the physical model of the diesel power system are monitored in real time to ensure that the simulation results are within the physically feasible range. The simulation results output includes the predicted values of the diesel power system state and the confidence assessment of the predicted values.
5. The dynamic simulation and adaptive control method for the diesel power system of an unmanned surface vessel according to claim 1, characterized in that, The method of using deep learning algorithms to fuse and analyze multimodal data to generate the optimal control strategy for the diesel power system includes: Deep learning networks based on attention mechanisms intelligently fuse multimodal data to form fused data. The attention mechanism includes self-attention and cross-modal attention mechanisms. The self-attention mechanism is used to process time series data within the same modality and capture long-term dependencies. The cross-modal attention mechanism is used to handle the correlation between data from different modalities and achieve effective information complementarity. The fusion process adopts a strategy that combines feature-level fusion and decision-level fusion. Based on a deep Q-network architecture, a control policy network is constructed. This network is trained using fused data and outputs an initial policy. The control policy network employs a dual-network structure, including a policy network and a value network. Experience replay and target network techniques are used to improve learning stability. The reward function design comprehensively considers fuel economy, emission control, power performance, and system reliability. A designed reward shaping technique guides the agent to learn the optimal control policy. For the balance between exploration and utilization, an adaptive ε-greedy strategy is adopted to achieve intelligent exploration, focusing on exploration in the early stages of training and gradually shifting to utilizing learned knowledge in the later stages. A multi-objective particle swarm optimization algorithm is used to further optimize the initial strategy output by the control strategy network. This includes: constructing a multi-objective optimization problem to find the Pareto optimal solution among multiple objectives such as fuel economy, emission control, power performance and system reliability; considering the physical constraints and operational boundary conditions of the physical model of the diesel power system during the optimization process to ensure that the generated control strategy is within the physically feasible range.
6. The dynamic simulation and adaptive control method for the diesel power system of an unmanned surface vessel according to any one of claims 1-5, characterized in that, The step of adjusting the operating parameters of the diesel power system in real time according to the obtained optimal control strategy to achieve adaptive control includes: Based on the optimal control strategy, a multivariable coordinated control algorithm is used to precisely control the fuel injection system, intake system and exhaust system of the diesel power system. A hybrid architecture combining convolutional neural networks and long short-term memory networks is adopted to monitor the operating status of diesel power systems in real time, including fault status monitoring and fault severity assessment. Fault status includes sensor failure, actuator failure, and component performance degradation. Fault severity assessment uses a fuzzy logic algorithm to classify the fault according to its impact on system safety and performance. When a fault is detected, a backup control strategy is automatically activated, employing system reconfiguration technology to maintain the basic functions of the diesel power system by reconfiguring the remaining healthy components: for sensor faults, soft sensor technology is used to estimate the faulty sensor signal based on the relevant parameters of other normal sensors; for actuator faults, a control reallocation algorithm is used to transfer the function of the faulty actuator to other actuators; for severe faults, a degraded operation mode or a safety shutdown procedure is initiated to ensure system safety.
7. A dynamic simulation and adaptive control system for an unmanned surface vessel's diesel power system, characterized in that, include: The multimodal data acquisition module collects real-time operating parameters, environmental data, and hull status information of the unmanned surface vessel's diesel power system, forming multimodal data. The dynamic simulation module constructs a physical model of the working process of the diesel power system and builds a data-driven model based on the historical operating data of the unmanned surface vessel's diesel power system. Based on the physical model and the data-driven model, it performs dynamic simulation of the diesel power system and outputs the system state prediction data and corresponding confidence levels of the unmanned surface vessel's diesel power system. The AI decision-making module, based on system state prediction data and corresponding confidence levels, uses deep learning algorithms to fuse and analyze multimodal data to generate the optimal control strategy for the diesel power system. The adaptive control module adjusts the operating parameters of the diesel power system in real time according to the obtained optimal control strategy to achieve adaptive control.
8. An electronic device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the dynamic simulation and adaptive control method for the diesel power system of an unmanned surface vessel as described in any one of claims 1-6.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the dynamic simulation and adaptive control method for the diesel power system of an unmanned surface vessel as described in any one of claims 1-6.