Real-time energy management control system and method for ship hybrid power system based on edge computing

Abstract

The application discloses a kind of based on edge computing's ship hybrid power system real-time energy management control system, it includes edge computing architecture module and constructs distributed real-time control environment by edge computing node and ship local area network;Real-time data processing module collects the original operation data of each edge computing node corresponding power unit and carries out pre-processing, utilizes pre-processing data to each power unit by fault prediction model and carries out anomaly detection, generates exception prediction information;Distributed decision module generates the local power instruction corresponding to each power unit based on exception prediction information, carries out coordination and fusion to all local power instructions by information interaction between each edge computing node, generates energy management control signal.The application can realize the real-time, optimization and reliable management of ship hybrid power system energy.

Description

Technical Field

[0001] This invention relates to the field of marine hybrid power system control technology, specifically to a real-time energy management control system and method for marine hybrid power systems based on edge computing. Background Technology

[0002] Marine hybrid power systems are a core technology for meeting the energy conservation and emission reduction needs of the shipping industry. They achieve efficient energy utilization by integrating traditional internal combustion engines with new energy power units. The energy management system is the core control unit of the hybrid power system. Existing marine energy management systems generally adopt a centralized computing architecture, requiring the aggregation of sensor data from all shipboard sensors to a central server for calculation and decision-making before issuing control commands. As the complexity of ship systems increases, this architecture exposes problems such as high computational latency, heavy communication load, significant single-point failure risk, and substantial data security vulnerabilities. While existing technologies have been explored for optimization, blockchain solutions suffer from high computational overhead and network latency, and are unsuitable for real-time ship control scenarios. Optimization solutions for centralized architectures only improve algorithm accuracy and cannot fundamentally solve the single-point failure and communication bottleneck defects. Summary of the Invention

[0003] The purpose of this invention is to provide a real-time energy management control system and method for marine hybrid power systems based on edge computing. By constructing a distributed edge computing architecture, this invention improves the real-time performance, reliability, resource utilization, and environmental adaptability of energy management in marine hybrid power systems.

[0004] To achieve this objective, the present invention provides a real-time energy management and control system for a marine hybrid power system based on edge computing, comprising: The edge computing architecture module is used to build a distributed real-time control environment through edge computing nodes and ship local area networks; The real-time data processing module is used to collect the raw operating data of the power units corresponding to each edge computing node in the distributed real-time control environment, preprocess the raw operating data to obtain preprocessed data, and use the preprocessed data to perform anomaly detection on each power unit through the fault prediction model to generate anomaly prediction information. The distributed decision module is used to generate local power commands for each power unit based on the anomaly prediction information. Through information interaction between edge computing nodes, all local power commands are coordinated and fused to generate energy management control signals.

[0005] Preferably, the resource optimization and allocation module is used to monitor the computing resource usage status of each edge computing node, and dynamically schedule the allocation and migration of computing tasks among each edge computing node based on the computing resource usage status, the anomaly prediction information, and the real-time operating conditions of the ship.

[0006] Preferably, the fault tolerance and security mechanism module is used to migrate the computing tasks of the faulty edge computing node to the normal node when a fault is detected, based on the anomaly prediction information and the status monitoring between each edge computing node, and to provide security protection for the communication between each edge computing node.

[0007] Preferably, the specific process of constructing a distributed real-time control environment through edge computing nodes and ship local area networks includes: Embedded edge computing nodes with built-in neural network processing units are deployed in the ship's power unit. All edge computing nodes are connected into a peer-to-peer network through the ship's local area network based on the real-time industrial communication network. Each edge computing node can dynamically allocate local computing resources according to the resource management platform with a multi-level priority task scheduling mechanism, thereby building a physically distributed and logically collaborative real-time control environment.

[0008] Preferably, the specific method for preprocessing the raw operating data of the power unit corresponding to each edge computing node to obtain preprocessed data includes: Each edge computing node obtains the raw operating data of the corresponding power unit through the CAN bus or analog interface. The raw operating data is denoised and smoothed using the Kalman filter algorithm to obtain filtered time-series data. The filtered time-series data is then transformed from the time domain to the frequency domain using the fast Fourier transform to extract key frequency domain features. The resulting low-dimensional feature vector is the preprocessed data.

[0009] Preferably, the specific process of using a fault prediction model to perform anomaly detection on each power unit using preprocessed data and generating anomaly prediction information is as follows: A fault prediction model based on a long short-term memory network, deployed on edge computing nodes, is used to perform real-time fitting calculations on the input preprocessed data. The fault prediction model based on long short-term memory networks uses preprocessed data to obtain predicted values ​​for the power unit. The residual is then calculated by comparing the predicted values ​​with the actual values ​​measured by sensors. The calculation formula is as follows: in, For residuals, This is the actual value of the power unit. These are the predicted values ​​for the power unit; residual Compared with a preset dynamic threshold, when the residual When the dynamic threshold is exceeded, it indicates that the operating parameters of the power unit have an abnormal trend, and an early warning analysis is initiated. The fault prediction model based on the long short-term memory network uses a multi-step recursive extrapolation method to predict the health status evolution of the power unit whose operating parameters have an abnormal trend within a preset time period. The specific process is as follows: The fault prediction model based on a long short-term memory network at time t outputs the predicted value of the power unit at time t+1. The features are input into a fault prediction model based on a long short-term memory network to obtain the predicted value of the power unit at time t+2. By iterating n times, a continuous sequence of future states can be derived: in, The predicted value of the dynamic unit at time t+n; By analyzing the trend of future state sequences, it is possible to determine whether a power unit with abnormal operating parameters has a potential risk of failure, and generate anomaly prediction information.

[0010] Preferably, the specific process of generating local power commands for each power unit based on the anomaly prediction information, and coordinating and fusing all local power commands through information interaction between edge computing nodes to generate energy management control signals is as follows: Each edge computing node generates a local power command corresponding to each power unit based on the anomaly prediction information, combined with the preprocessed data and the received local load demand; Each edge computing node exchanges Lagrange multipliers containing local power commands and global consistency constraints via the ship's local area network, and iteratively solves the problem using the alternating direction multiplier method. The iterative formula is as follows: in, This represents the updated local power instruction for the power unit corresponding to the i-th edge computing node in the (k+1)-th iteration. This represents the local power command of the power unit corresponding to the i-th edge computing node; For penalty parameters; Let represent the local objective function of the i-th edge computing node; The square of the Euclidean norm of a vector is given by . This represents the globally consistent variable reached by all edge computing nodes in the k-th iteration; This represents the Lagrange multiplier of the i-th edge computing node in the k-th iteration; In each iteration, each edge computing node determines the global consistency variables it receives. The local solution is updated by optimizing the Lagrange multipliers of the remaining edge computing nodes. and Lagrange multipliers And the updated Lagrange multipliers The power commands are sent to the remaining edge computing nodes via the ship's local area network. Through multiple distributed iterations, the local power commands of all edge computing nodes converge to the global optimum, ultimately generating the power management control signal.

[0011] A real-time energy management and control method for a ship hybrid power system based on edge computing includes the following steps: A distributed real-time control environment is built using edge computing nodes and ship local area networks; The raw operating data of the power units corresponding to each edge computing node in the distributed real-time control environment is collected, the raw operating data is preprocessed to obtain preprocessed data, and the preprocessed data is used by the fault prediction model to detect anomalies in each power unit and generate anomaly prediction information. Based on the anomaly prediction information, local power commands corresponding to each power unit are generated. Through information interaction between each edge computing node, all local power commands are coordinated and fused to generate energy management control signals.

[0012] A computer program product includes a computer program that, when executed by a processor, implements the steps of the above-described method.

[0013] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method.

[0014] The beneficial effects of this invention are: This invention offloads energy management computing tasks to the edge computing nodes of the power unit, enabling on-site data acquisition, processing, and decision-making. This meets the millisecond-level real-time control requirements of ship propulsion systems, reduces the bandwidth burden on the ship's local area network, weakens dependence on strong communication environments, and improves stability in weak signal scenarios. The invention employs a physically distributed peer-to-peer network architecture, coupled with node fault tolerance and hot backup mechanisms, eliminating the single-point-of-failure risk of centralized architectures. The failure of a single edge computing node does not affect overall operation, enhancing security capabilities. Furthermore, by monitoring node load in real time and dynamically scheduling computing resources based on ship operating conditions, allocating computing power according to task priority, this invention can reduce the overall energy consumption of edge computing nodes while ensuring control performance. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the structure of the present invention; Figure 2 This is a flowchart of the present invention. Detailed Implementation

[0016] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments: Example 1 A real-time energy management and control system for a marine hybrid power system based on edge computing, such as Figure 1 As shown, it includes: The edge computing architecture module is used to build a distributed real-time control environment through edge computing nodes (which are dedicated embedded hardware devices deployed near the power units of the ship) physically deployed near the power units of the ship and the ship's local area network connecting the edge computing nodes. This design establishes a peer-to-peer network without a central server by connecting the edge computing nodes and the ship's local area network, pushing the computing tasks down to the near end of the power units. This eliminates the single point of failure risk of centralized architecture from the hardware level, enables data to be processed locally, and improves real-time performance and reliability. For the aforementioned edge computing architecture module, some optimized technical solutions include: the edge computing architecture module includes multiple embedded edge computing nodes physically deployed near the various power units of the ship (such as next to the main diesel engine, in the battery room, or in the power distribution cabinet). Each edge computing node uses an industrial-grade microcontroller with a built-in neural network processing unit (NPU) as its core and integrates a CAN bus and analog signal acquisition interface for directly reading sensor data from the corresponding power unit (such as generators, energy storage batteries, and frequency converters). All edge computing nodes are connected to a marine industrial Ethernet network based on the Modbus TCP / IP protocol via shielded twisted-pair cables, forming a peer-to-peer network. Each edge computing node runs a resource management platform with a multi-level priority task scheduling mechanism, which can dynamically allocate local computing resources according to the urgency of the task (such as propulsion commands having higher priority than monitoring tasks), thereby physically constructing a distributed real-time control environment without a central server and with low latency.

[0017] The real-time data processing module is used to collect raw operating data (including voltage, current, speed and temperature) of the power units corresponding to each edge computing node in the distributed real-time control environment. The raw operating data is preprocessed to obtain preprocessed data. The preprocessed data is then used by the fault prediction model to detect anomalies in each power unit (such as the main diesel engine, energy storage battery and frequency converter) and generate anomaly prediction information. This design preprocesses the raw operating data, which can reduce the computational overhead of subsequent algorithms. Based on the fault prediction model, it can realize early identification of equipment anomaly trends and minute-level fault warning, which can provide reliable data support for energy management decisions. The distributed decision module is used to generate local power commands for each power unit based on the anomaly prediction information. Through information interaction between edge computing nodes, all local power commands are coordinated and fused to generate a globally consistent energy management control signal. This design generates local power commands and completes decision fusion through information interaction between edge computing nodes, perfectly adapting to the dynamic operating conditions of the ship. While ensuring real-time control, it achieves optimal global energy allocation, thereby realizing distributed and coordinated scheduling of the ship's energy.

[0018] Regarding the energy management control signal, some optimized technical solutions include: the energy management control signal is directly sent to each actuator of the ship's hybrid power system (such as generators, energy storage converters, propulsion inverters, etc.) to control the operating status of the actuators (e.g., setting the generator speed, adjusting the inverter PWM pulse width, and setting the battery charging and discharging power), thereby realizing real-time and optimized allocation and scheduling of the ship's energy flow.

[0019] In the above technical solution, the resource optimization and allocation module is used to monitor the computing resource usage status of each edge computing node. Based on the computing resource usage status, the anomaly prediction information, and the real-time operating conditions of the ship, it dynamically schedules the allocation and migration of computing tasks among the edge computing nodes. The above design, by monitoring the computing resource usage status of edge computing nodes in real time, combined with anomaly prediction information and the real-time operating conditions of the ship, dynamically completes the allocation and migration of computing tasks among edge computing nodes. In the resource-constrained shipboard environment, it maximizes the utilization rate of computing resources and can prioritize the execution of high real-time and safe tasks, thereby achieving dynamic and efficient scheduling of computing resources of edge computing nodes.

[0020] For the aforementioned computing task, some optimized technical solutions include: the computing task refers to the various algorithm programs or processing procedures running on the edge computing node for realizing real-time energy management of the ship's hybrid power system.

[0021] In the above technical solution, the fault tolerance and security mechanism module is used to monitor the status between each edge computing node based on the anomaly prediction information. When an edge computing node failure is detected, the computing tasks of the failed edge computing node are migrated to the normal node, and the communication between each edge computing node is protected. The above design ensures continuous and stable operation by monitoring the survival status of edge computing nodes in real time and smoothly migrating the computing tasks of the failed edge computing node to the normal node when a failure occurs. Through digital certificate security authentication and hardware-level data encryption, the operational safety of the ship hybrid power system is comprehensively improved.

[0022] For the resource optimization allocation module and the fault tolerance and security mechanism module, some optimized technical solutions include: the resource optimization allocation module monitors the usage status of computing resources (CPU, memory) of each edge computing node in real time through a monitoring script running at the kernel layer, and uses a Markov chain model to predict future computing load; with the objective function of minimizing the weighted sum of task decision latency and execution energy consumption, an improved particle swarm optimization algorithm is used to iteratively solve the optimal allocation and migration scheme of computing tasks among edge computing nodes under the hard constraints of computing resources, memory bandwidth, and task real-time performance; the fault tolerance and security mechanism module monitors the liveness status of edge computing nodes by periodically sending and receiving heartbeat detection messages; when an edge computing node failure is detected, the computing tasks of the failed edge computing node (such as the power calculation task of a specific battery pack) are immediately hot-migrated to the normal edge computing node with the lowest load through the resource optimization allocation module; at the same time, security protection is provided for communication between edge computing nodes through security authentication based on digital certificates and hardware-level data encryption.

[0023] The specific process of constructing a distributed real-time control environment through edge computing nodes and ship local area networks in the above technical solution includes: Embedded edge computing nodes with built-in Neural Processing Units (NPUs) are deployed in the ship's power unit. All edge computing nodes are connected in a peer-to-peer network through a ship's local area network (LAN) based on a real-time industrial communication network (Modbus TCP / IP). Each edge computing node can dynamically allocate local computing resources according to a resource management platform with a multi-level priority task scheduling mechanism, thereby constructing a physically distributed and logically collaborative real-time control environment. The above design enables the embedded edge computing nodes with built-in neural processing units to have independent AI computing power, which can support the local closed-loop operation of complex algorithms. The ship's LAN based on industrial Ethernet enables low-latency communication between nodes, and the multi-level priority task scheduling mechanism dynamically allocates computing power according to the urgency of tasks, ensuring that edge nodes can independently complete local real-time control and global collaborative decision-making, thereby improving distributed collaborative capabilities and real-time control performance.

[0024] In the above technical solution, the specific method for preprocessing the raw operating data of the power unit corresponding to each edge computing node to obtain preprocessed data includes: Each edge computing node obtains the raw operating data of the corresponding power unit through a CAN bus or analog interface. The raw operating data is then denoised and smoothed using a Kalman filter algorithm to obtain filtered time-series data. A Fast Fourier Transform (FFT) is then used to convert the filtered time-series data from the time domain to the frequency domain, extracting key frequency domain features and compressing the high-dimensional raw data stream into a low-dimensional feature vector. This low-dimensional feature vector is the preprocessed data. This design achieves denoising and smoothing of the raw operating data signal using the Kalman filter algorithm and converts the time-domain data into a frequency-domain feature vector using the FFT. While retaining key state information, it transforms high-dimensional raw data into low-dimensional features, reducing the computational load of subsequent decision-making algorithms. This design is well-suited to the resource-constrained hardware characteristics of edge computing nodes, effectively improving data quality and reducing edge computing overhead.

[0025] In the above technical solution, the specific process of using a fault prediction model to perform anomaly detection on each power unit using preprocessed data and generating anomaly prediction information is as follows: A fault prediction model based on a long short-term memory network, deployed on edge computing nodes, performs real-time fitting calculations on the input preprocessed data (i.e., low-dimensional feature vectors reflecting the operating status after filtering and feature extraction). The fault prediction model based on long short-term memory networks uses preprocessed data to obtain predicted values ​​for the power unit. The residual is then calculated by comparing the predicted values ​​with the actual values ​​measured by sensors. The calculation formula is as follows: in, For residuals, This is the actual value of the power unit. These are the predicted values ​​for the power unit; residual The residual is compared with a preset dynamic threshold (which is dynamically adjusted based on historical deviations and operating conditions). When the dynamic threshold is exceeded, it indicates that the operating parameters of the power unit have an abnormal trend, and an early warning analysis is initiated. The fault prediction model based on the long short-term memory network uses a multi-step recursive extrapolation method to predict the health status evolution of the power unit whose operating parameters have an abnormal trend within a preset time period. The specific process is as follows: The fault prediction model based on a long short-term memory network at time t outputs the predicted value of the power unit at time t+1. The features are input into a fault prediction model based on a long short-term memory network to obtain the predicted value of the power unit at time t+2. By iterating n times, a continuous sequence of future states can be derived: in, The predicted value of the dynamic unit at time t+n; By analyzing the trend of future state sequences, it is possible to determine whether a power unit with abnormal operating parameters has potential failure risks, and generate abnormal prediction information including failure type, severity, and predicted evolution trend. The above-mentioned design, based on a fault prediction model of long short-term memory network, identifies abnormal trends through residual calculation, and infers the future health status evolution of the equipment. It can issue early warnings within a minute-level window before a failure occurs, providing a critical time window for pre-fault handling and avoiding sudden failures in the power system, and achieving accurate early warning of potential failures of the power unit.

[0026] In the above technical solution, the specific process of generating local power commands for each power unit based on the anomaly prediction information, and coordinating and fusing all local power commands through information interaction between edge computing nodes to generate energy management control signals is as follows: Each edge computing node generates a local power command corresponding to each power unit based on the anomaly prediction information, combined with the preprocessed data and the received local load demand. The anomaly prediction information is taken into consideration as a key constraint (e.g., limiting the power of units at risk of failure). Each edge computing node exchanges Lagrange multipliers containing local power commands and global consistency constraints via the ship's local area network, and iteratively solves the problem using the alternating direction multiplier method. The iterative formula is as follows: in, This represents the updated local power instruction for the power unit corresponding to the i-th edge computing node in the (k+1)-th iteration. This represents the local power command of the power unit corresponding to the i-th edge computing node; This is a penalty parameter used to adjust the weight between local decisions and global consistency constraints; Let represent the local objective function of the i-th edge computing node; The square of the Euclidean norm of a vector is given by . This represents the globally consistent variable reached by all edge computing nodes in the k-th iteration (representing the global power allocation scheme that all nodes need to eventually reach a consensus on). This represents the Lagrange multiplier of the i-th edge computing node in the k-th iteration; In each iteration, each edge computing node determines the global consistency variables it receives. The local solution is updated by optimizing the Lagrange multipliers of the remaining edge computing nodes. and Lagrange multipliers And the updated Lagrange multipliers The power commands are sent to the remaining edge computing nodes via the ship's local area network. Through multiple distributed iterations, the local power commands of all edge computing nodes converge to the global optimum, ultimately generating the power management control signal. The above design uses the alternating direction multiplier method for iterative solution among edge computing nodes. The local power commands are matched with the global consistency constraints through Lagrange multiplier interaction. After multiple distributed iterations, all edge computing nodes reach a unified decision, which not only retains the low latency advantage of distributed control, but also achieves efficient and balanced energy distribution across the entire ship, ensuring that the power distribution of the entire ship converges to the global optimum quickly.

[0027] Regarding the specific process of iterative solution using the alternating direction multiplier method, some optimized technical solutions include: during the iterative solution process, if there is a conflict between local demands among edge computing nodes (such as a conflict between propulsion power demand and domestic electricity demand during power shortages), arbitration is carried out based on preset fixed priority rules (such as ensuring propulsion priority) or combined with abnormal prediction information (such as a certain unit needing power reduction protection). The consensus reached after arbitration is converted into specific energy management control signals that can be directly issued to each actuator (such as diesel engine governor, energy storage converter, propulsion inverter), such as speed setpoint or PWM wave duty cycle command, thereby completing a distributed real-time energy scheduling closed loop.

[0028] Example 2 A real-time energy management and control method for a ship hybrid power system based on edge computing, such as Figure 2 As shown, a distributed real-time control environment is constructed through edge computing nodes and a ship's local area network; raw operating data of the power units corresponding to each edge computing node are collected and preprocessed; anomaly detection of each power unit is performed using the preprocessed data through a fault prediction model, generating anomaly prediction information; local power commands corresponding to each power unit are generated based on the anomaly prediction information; and all local power commands are coordinated and fused through information interaction between each edge computing node to generate energy management control signals.

[0029] The specific method for real-time energy management and control of a ship hybrid power system based on edge computing includes the following steps: A distributed real-time control environment is built using edge computing nodes and ship local area networks; The raw operating data of the power units corresponding to each edge computing node in the distributed real-time control environment is collected, the raw operating data is preprocessed to obtain preprocessed data, and the preprocessed data is used by the fault prediction model to detect anomalies in each power unit and generate anomaly prediction information. Based on the anomaly prediction information, local power commands corresponding to each power unit are generated. Through information interaction between each edge computing node, all local power commands are coordinated and fused to generate energy management control signals.

[0030] Example 3 A computer program product includes a computer program that, when executed by a processor, implements the steps of the method described in Embodiment 2.

[0031] Example 4 A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method as described in Embodiment 2.

[0032] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0033] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A system that specifies functions in one or more boxes.

[0034] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including an instruction set implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0035] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0036] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit its scope of protection. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that after reading the present invention, they can still make various changes, modifications or equivalent substitutions to the specific implementation of the invention, but these changes, modifications or equivalent substitutions are all within the scope of protection of the pending claims of the invention.

[0037] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

Claims

1. A real-time energy management and control system for a marine hybrid power system based on edge computing, characterized in that, It includes: The edge computing architecture module is used to build a distributed real-time control environment through edge computing nodes and ship local area networks; The real-time data processing module is used to collect the raw operating data of the power units corresponding to each edge computing node in the distributed real-time control environment, preprocess the raw operating data to obtain preprocessed data, and use the preprocessed data to perform anomaly detection on each power unit through the fault prediction model to generate anomaly prediction information. The distributed decision module is used to generate local power commands for each power unit based on the anomaly prediction information. Through information interaction between edge computing nodes, all local power commands are coordinated and fused to generate energy management control signals.

2. The real-time energy management and control system for a marine hybrid power system based on edge computing according to claim 1, characterized in that, It also includes: The resource optimization and allocation module is used to monitor the computing resource usage status of each edge computing node, and dynamically schedule the allocation and migration of computing tasks among the edge computing nodes based on the computing resource usage status, the anomaly prediction information and the real-time operating conditions of the ship.

3. The real-time energy management and control system for a marine hybrid power system based on edge computing according to claim 1, characterized in that, It also includes: The fault tolerance and security mechanism module is used to migrate the computing tasks of the faulty edge computing node to the normal edge computing node when the faulty edge computing node is detected, based on the anomaly prediction information and the status monitoring between each edge computing node, and to provide security protection for the communication between each edge computing node.

4. The real-time energy management and control system for a marine hybrid power system based on edge computing according to claim 1, characterized in that: The specific process of building a distributed real-time control environment through edge computing nodes and ship local area networks includes: Embedded edge computing nodes with built-in neural network processing units are deployed in the ship's power unit. All edge computing nodes are connected into a peer-to-peer network through the ship's local area network based on the real-time industrial communication network. Each edge computing node can dynamically allocate local computing resources according to the resource management platform with a multi-level priority task scheduling mechanism, thereby building a physically distributed and logically collaborative real-time control environment.

5. The real-time energy management and control system for a marine hybrid power system based on edge computing according to claim 1, characterized in that: The specific methods for preprocessing the raw operating data of the power units corresponding to each edge computing node to obtain preprocessed data include: Each edge computing node obtains the raw operating data of the corresponding power unit through the CAN bus or analog interface. The raw operating data is denoised and smoothed using the Kalman filter algorithm to obtain filtered time-series data. The filtered time-series data is then transformed from the time domain to the frequency domain using the fast Fourier transform to extract key frequency domain features. The resulting low-dimensional feature vector is the preprocessed data.

6. The real-time energy management and control system for a marine hybrid power system based on edge computing according to claim 1, characterized in that: The specific process of using a fault prediction model to perform anomaly detection on each power unit using preprocessed data and generating anomaly prediction information is as follows: A fault prediction model based on a long short-term memory network, deployed on edge computing nodes, is used to perform real-time fitting calculations on the input preprocessed data. The fault prediction model based on long short-term memory networks uses preprocessed data to obtain predicted values ​​for the power unit. The residual is then calculated by comparing the predicted values ​​with the actual values ​​measured by sensors. The calculation formula is as follows: in, For residuals, This is the actual value of the power unit. These are the predicted values ​​for the power unit; residual Compared with a preset dynamic threshold, when the residual When the dynamic threshold is exceeded, it indicates that the operating parameters of the power unit have an abnormal trend, and an early warning analysis is initiated. The fault prediction model based on the long short-term memory network uses a multi-step recursive extrapolation method to predict the health status evolution of the power unit whose operating parameters have an abnormal trend within a preset time period. The specific process is as follows: The fault prediction model based on a long short-term memory network at time t outputs the predicted value of the power unit at time t+1. As input features, these features are fed into a fault prediction model based on a long short-term memory network to obtain the predicted value of the power unit at time t+2. By iterating n times, a continuous sequence of future states can be derived: in, The predicted value of the dynamic unit at time t+n; By analyzing the trend of future state sequences, it is possible to determine whether a power unit with abnormal operating parameters has a potential risk of failure, and generate anomaly prediction information.

7. The real-time energy management and control system for a marine hybrid power system based on edge computing according to claim 1, characterized in that: Based on the anomaly prediction information, local power commands are generated for each power unit. The energy management control signal is generated by coordinating and fusing all local power commands through information interaction between edge computing nodes. The specific process is as follows: Each edge computing node generates a local power command corresponding to each power unit based on the anomaly prediction information, combined with the preprocessed data and the received local load demand. Each edge computing node exchanges Lagrange multipliers containing local power commands and global consistency constraints via the ship's local area network, and iteratively solves the problem using the alternating direction multiplier method. The iterative formula is as follows: in, This represents the updated local power instruction for the power unit corresponding to the i-th edge computing node in the (k+1)-th iteration. This represents the local power command of the power unit corresponding to the i-th edge computing node; For penalty parameters; Let represent the local objective function of the i-th edge computing node; The square of the Euclidean norm of a vector is given by . This represents the globally consistent variable reached by all edge computing nodes in the k-th iteration; This represents the Lagrange multiplier of the i-th edge computing node in the k-th iteration; In each iteration, each edge computing node determines the global consistency variables it receives. The local solution is updated by optimizing the Lagrange multipliers of the remaining edge computing nodes. and Lagrange multipliers And the updated Lagrange multipliers The power commands are sent to the remaining edge computing nodes via the ship's local area network. Through multiple distributed iterations, the local power commands of all edge computing nodes converge to the global optimum, ultimately generating the power management control signal.

8. A real-time energy management and control method for a ship hybrid power system based on edge computing, characterized in that, It includes the following steps: A distributed real-time control environment is built by using edge computing nodes and ship local area networks; The raw operating data of the power units corresponding to each edge computing node in the distributed real-time control environment is collected, the raw operating data is preprocessed to obtain preprocessed data, and the preprocessed data is used by the fault prediction model to detect anomalies in each power unit and generate anomaly prediction information. Based on the anomaly prediction information, local power commands corresponding to each power unit are generated. Through information interaction between each edge computing node, all local power commands are coordinated and fused to generate energy management control signals.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method of claim 8.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it implements the steps of the method as described in claim 8.

Images