Energy integration management optimization method and system for LNG hybrid power ship

Abstract

The invention relates to an energy integration management optimization method and system for an LNG hybrid power ship. The method comprises the following steps that a mathematical model of an energy system of a target LNG hybrid power ship is established; constructing a multi-target optimization model according to the mathematical model of the energy system of the target LNG hybrid power ship; a hierarchical collaborative optimization framework is adopted to solve the multi-target optimization model, a global optimal solution meeting all constraint conditions is obtained, the global optimal solution serves as an optimized energy management strategy of the target LNG hybrid power ship, and the hierarchical collaborative optimization framework comprises two layers of optimization; and according to the optimized energy management strategy of the target LNG hybrid power ship, realizing optimal management of the energy of the target LNG hybrid power ship. Compared with the prior art, the economical efficiency, the environmental protection property and the robustness of energy integration management under all working conditions and the adaptability and the effectiveness of an optimization strategy are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of marine power system integration technology, and in particular to an energy integration management optimization method and system for LNG hybrid ships. Background Technology

[0002] Due to the enormous amount of greenhouse gases produced by traditional ships, the shipbuilding industry needs to improve onboard energy efficiency to reduce pollutant emissions. While traditional diesel engine ships are technologically mature, their high emissions make it difficult to meet increasingly stringent environmental regulations. Therefore, the shipbuilding industry is actively seeking solutions. Liquefied natural gas (LNG), as a clean energy source, has the advantage of low carbon emissions, while lithium batteries and supercapacitors can provide efficient energy storage and rapid power response. Hybrid power systems combining LNG, lithium batteries, and supercapacitors are becoming an important development direction for future ship propulsion systems.

[0003] However, energy management is one of the key technologies for achieving efficient operation of hybrid-powered ships. A reasonable energy management strategy can optimize the power allocation of LNG, lithium batteries, and supercapacitors, improve ship energy efficiency, reduce operating costs, and ensure ship reliability and safety. Currently, existing energy management strategies have significant limitations: First, most methods rely on pre-set simple rules or single optimization algorithms, making it difficult to simultaneously coordinate multiple conflicting optimization objectives such as minimizing fuel consumption, maximizing energy storage lifespan, optimizing dynamic response performance, reducing carbon emissions, and controlling battery degradation costs. Second, ships face uncertainties such as rapid fluctuations in power demand and variable environmental conditions during actual navigation. Existing strategies are often based on deterministic models and lack explicit handling capabilities for uncertainties, resulting in insufficient optimization performance and robustness under complex operating conditions. Furthermore, traditional methods typically fail to fully utilize the spatiotemporal dynamic characteristics inherent in ship operation data and lack effective means to handle insufficient or poor-quality data samples, limiting the adaptive capability and accuracy of optimization strategies. For example, patent application CN120509525A discloses an energy optimization management method and system for range-extended hybrid electric vehicles. This method predicts load demand through a fusion model of temporal convolutional networks and long short-term memory, and uses an improved multi-objective genetic algorithm to generate a Pareto front solution set to optimize the energy allocation between the battery and the range extender. This method is only applicable to a dual-source architecture of battery and fuel range extender, and cannot be matched with a three-energy coupling system of liquefied natural gas, lithium battery and supercapacitor; it can only optimize three deterministic objectives: fuel consumption, battery degradation and carbon emissions, and cannot explicitly handle uncertainties such as power fluctuations and environmental disturbances and incorporate them into the objectives of dynamic response performance and economic cost of battery degradation; it can only rely on online knowledge distillation for passive model fine-tuning, and cannot actively extract spatiotemporal dynamic features to achieve robust adaptation in scenarios with insufficient or poor data quality.

[0004] Therefore, developing an efficient, intelligent, and robust energy management system is of great significance for the promotion and application of LNG hybrid power ships. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and provide an energy integration management optimization method and system for LNG hybrid ships, which significantly improves the economy, environmental protection and robustness of energy integration management under all operating conditions, as well as the adaptability and effectiveness of optimization strategies.

[0006] The objective of this invention can be achieved through the following technical solutions: An energy integration management optimization method for LNG hybrid ships includes the following steps: Establish a mathematical model of the energy system of the target LNG hybrid power ship; A multi-objective optimization model is constructed based on the mathematical model of the energy system of the target LNG hybrid power ship; A hierarchical collaborative optimization architecture is adopted to solve the multi-objective optimization model. This architecture comprises two layers of optimization. The first layer includes a data preprocessing and enhancement module, a spatiotemporal feature extraction module, and a feature fusion and initial scheme generation module. The data preprocessing and enhancement module enhances ship operation data to generate synthetic sample data. The spatiotemporal feature extraction module extracts spatial and temporal dynamic features from the synthetic sample data. The feature fusion and initial scheme generation module fuses the spatial and temporal dynamic features and generates an initial optimization scheme through a fully connected layer mapping. The second layer includes an uncertainty handling module and a distributed solution and global coordination module. The uncertainty handling module simulates various uncertain scenarios in ship operation and generates candidate optimization schemes based on the initial optimization scheme. The distributed solution and global coordination module decomposes the candidate optimization schemes into sub-problems and generates a globally optimal solution that satisfies all constraints through alternating updates and coordination mechanisms. This globally optimal solution is then used as the optimized energy management strategy for the target LNG hybrid power ship. The optimized energy management strategy for the target LNG hybrid vessel is used to achieve optimized energy management of the target LNG hybrid vessel.

[0007] Furthermore, the energy system of the LNG hybrid power ship includes an LNG fuel subsystem, a lithium battery energy storage subsystem, and a supercapacitor subsystem. The constraints of the energy system of the LNG hybrid power ship include power balance constraints, power source power constraints, and energy constraints of lithium batteries and supercapacitors.

[0008] Furthermore, the optimization objectives of the multi-objective optimization model include minimizing fuel consumption, maximizing the lifespan of lithium batteries and supercapacitors, optimizing ship dynamic response performance, reducing carbon emissions, and controlling battery degradation costs.

[0009] Furthermore, the data preprocessing and enhancement module uses GAN to preprocess the data, enhance the ship operation data, and generate synthetic sample data. The spatiotemporal feature extraction module includes CNN and LSTM. The CNN is used to extract spatial features from the synthetic sample data, and the LSTM is used to capture the temporal dynamic features of the synthetic sample data and predict the ship's power demand. The synthesized sample data includes LNG consumption rate, remaining energy of lithium batteries, remaining energy of supercapacitors, and ship power requirements.

[0010] Furthermore, the GAN includes a generator and a discriminator, and the specific steps of the GAN training process include: The generator receives a random noise vector and generates synthetic sample data. The synthesized sample data and real ship operation data are input into the discriminator to determine the authenticity of the data; The results of the data authenticity assessment are compared with the actual source of the data to calculate the adversarial loss. The weights of the discriminator and generator are updated respectively using the backpropagation algorithm based on the adversarial loss. Training ends when the adversarial process reaches a dynamic equilibrium.

[0011] Furthermore, the uncertainty processing module uses the initial optimization scheme as the initial iteration point, optimizes the initial optimization scheme using SGMHC, and introduces random noise in each scheme iteration to simulate various uncertain scenarios in ship operation and generate multiple candidate optimization schemes covering different operating conditions.

[0012] Furthermore, the specific formula for the iterative process of the proposed solution is as follows: In the formula, For the first Sample solutions for the next iteration For the first Sample solutions for the next iteration Step size, For the first The probability distribution function of the sample schemes in the next iteration. It is a random noise vector.

[0013] Furthermore, the distributed solution and global coordination module uses ADMM to decompose the candidate optimization scheme into subproblems, and generates a globally optimal solution that satisfies all constraints through alternating updates and coordination mechanisms. Specific steps include: The multi-objective optimization problem of the candidate optimization schemes is formulated in the standard ADMM form; The global optimization problem is decomposed into local subproblems for the LNG fuel subsystem, lithium battery energy storage subsystem, and supercapacitor subsystem, and the optimal power allocation for each subproblem is solved independently. By alternately updating local variables and global multipliers, the solutions to the subproblems are coordinated, and a globally optimal solution that satisfies all constraints is obtained.

[0014] Furthermore, the update formula for the local variables is: The update formula for the global multiplier is: In the formula, For power allocation local variables related to LNG engines, For the first Local variables related to power allocation of the LNG engine in the next iteration For the local objective function related to the LNG engine, For power allocation local variables related to energy storage, For the first Local variables related to power allocation in the next iteration of energy storage For the first Local variables related to power allocation in the next iteration of energy storage For the local objective function related to electrical energy storage, For penalty parameters, , The coefficient matrix, For the power requirements of ships, As a global multiplier, For the first The global multiplier value at the next iteration. For the first The global multiplier value at the next iteration.

[0015] According to another aspect of the present invention, an energy integrated management system for LNG hybrid power ships is provided, comprising: The mathematical model building module is used to build a mathematical model of the energy system of the target LNG hybrid power ship; A multi-objective optimization model construction module is used to construct a multi-objective optimization model based on the mathematical model of the energy system of the target LNG hybrid power ship; A multi-objective optimization model solving module is used to solve the multi-objective optimization model using a hierarchical collaborative optimization architecture. This architecture comprises two optimization layers. The first layer includes a data preprocessing and enhancement module, a spatiotemporal feature extraction module, and a feature fusion and initial scheme generation module. The data preprocessing and enhancement module enhances ship operation data. The spatiotemporal feature extraction module extracts spatial and temporal dynamic features from the enhanced ship operation data. The feature fusion and initial scheme generation module fuses the spatial and temporal dynamic features and generates an initial optimization scheme through a fully connected layer mapping. The second layer includes an uncertainty handling module and a distributed solving and global coordination module. The uncertainty handling module simulates various uncertain scenarios in ship operation and generates candidate optimization schemes based on the initial optimization scheme. The distributed solving and global coordination module decomposes the candidate optimization schemes into sub-problems and generates a globally optimal solution that satisfies all constraints through an alternating update and coordination mechanism. This globally optimal solution is then used as the optimized energy management strategy for the target LNG hybrid power ship. The energy optimization management module is used to optimize the energy management of the target LNG hybrid vessel according to the optimized energy management strategy of the target LNG hybrid vessel.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention employs a hierarchical collaborative optimization architecture. In the first layer, generative adversarial networks and spatiotemporal neural networks are used for data augmentation and feature fusion to generate high-quality initial solutions. In the second layer, the stochastic gradient hybrid Monte Carlo method is used to simulate uncertainties, and the alternating direction multiplier method is used for distributed global solution. This achieves systematic and refined optimization of complex ship energy management problems, solving the problem that existing methods, which rely on a single algorithm or rule, cannot simultaneously handle multiple objectives, strong coupling constraints, and operational uncertainties. This significantly improves the economy, environmental friendliness, and robustness of energy integrated management under all operating conditions.

[0017] 2. This invention constructs a data augmentation module based on generative adversarial networks and combines convolutional neural networks and long short-term memory networks for joint extraction and fusion of spatiotemporal features, thereby achieving deep perception and representation of ship operation dynamics and multi-source energy states. It solves the problems of weak model generalization ability and poor dynamic response performance caused by insufficient data quality or limited feature extraction capabilities in traditional optimization methods. It provides a more accurate and comprehensive information foundation for energy management decisions and improves the adaptability and effectiveness of optimization strategies.

[0018] 3. This invention introduces a stochastic gradient hybrid Monte Carlo method in the second-level optimization to actively simulate various uncertain scenarios during navigation and generate a series of robust candidate solutions. This enables explicit modeling and proactive response to random fluctuations in the operating environment, solves the risk of performance drop and strategy failure of traditional deterministic optimization methods under actual uncertain conditions, and significantly enhances the anti-interference capability of energy integrated management and its stable performance in actual complex sea conditions.

[0019] 4. This invention decomposes the global multi-objective problem into multiple subsystem subproblems for parallel and collaborative solution by using the alternating direction multiplier method in the second-level optimization. This achieves rapid global coordinated optimization of multiple objectives such as fuel consumption, carbon emissions, and equipment lifespan under the premise of strictly satisfying various equipment constraints and power balance. It solves the problem that centralized algorithms are computationally complex and prone to getting trapped in local optima when solving high-dimensional and strongly coupled problems such as ship hybrid power systems. This improves optimization efficiency while ensuring solution accuracy. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating an energy integration management optimization method for LNG hybrid ships proposed in this invention. Figure 2 A schematic diagram of the hierarchical collaborative optimization architecture; Figure 3 A schematic diagram of the training process for an adversarial generative network; Figure 4 A flowchart illustrating the process of processing time-series data using a Long Short-Term Memory (LSTM) network. Detailed Implementation

[0021] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0022] The following English abbreviations are involved: Liquefied Natural Gas (LNG) Generative Adversarial Networks (GANs) Convolutional Neural Networks (CNN) Long Short-Term Memory (LSTM) networks Stochastic Gradient Mixture Monte Carlo (SGHMC) Alternating Direction Multiplier Method (ADMM) Example 1 This embodiment provides an energy integration management optimization method for LNG hybrid power ships, such as... Figure 1 As shown, it includes the following steps: S1. Establish a mathematical model of the energy system of the target LNG hybrid power ship.

[0023] The energy system of an LNG hybrid ship includes an LNG fuel subsystem, a lithium battery energy storage subsystem, and a supercapacitor subsystem.

[0024] The LNG fuel subsystem is one of the main power sources for ships. Its working principle involves vaporizing liquefied natural gas and then burning it, converting chemical energy into mechanical energy through an engine. The combustion process of LNG can be described by the law of conservation of energy. The LNG consumption rate is closely related to the engine output power, thermal efficiency, and the lower calorific value of LNG. The mathematical model of the LNG fuel subsystem is as follows: In the formula, This represents the mass flow rate consumed by LNG. This refers to the output power of the LNG engine. For the efficiency of LNG engines, It has the low calorific value of LNG.

[0025] Lithium-ion batteries, as energy storage devices, play a role in balancing power demand and storing excess energy in hybrid-powered ships. The charging and discharging process of a lithium-ion battery energy storage subsystem can be described using electrochemical equations and circuit models. The mathematical model of the charging and discharging process of a lithium-ion battery energy storage subsystem is as follows: In the formula, For lithium batteries State of charge at time t, For lithium batteries State of charge at time t, This represents the battery current; it is positive during discharge and negative during charging. For time step, This refers to the battery capacity of the lithium battery.

[0026] A supercapacitor is a high-power-density energy storage device capable of providing or absorbing large amounts of energy in a short time. The charging and discharging process of a supercapacitor subsystem can be described by a circuit model. The mathematical model of the supercapacitor subsystem is as follows: In the formula, This refers to the charging and discharging power of the supercapacitor. This refers to the remaining energy of the supercapacitor. For time.

[0027] Energy management requires consideration of various constraints to ensure the safe operation of the ship and the rational use of its energy systems. These constraints include power balance constraints and power and energy limits for each energy system. By defining these constraints, the optimization process can be ensured to proceed within a practically feasible framework.

[0028] Energy management requires consideration of various constraints to ensure the safe operation of the vessel and the rational use of its energy systems. Constraints on the energy systems of LNG hybrid-powered vessels include power balance constraints, power source power constraints, and energy constraints related to lithium batteries and supercapacitors. Defining these constraints ensures that the optimization process remains within practically feasible limits.

[0029] The ship's total power requirements are met by LNG engines, lithium batteries, and supercapacitors. This constraint ensures that the ship's propulsion system can meet navigation needs while avoiding power excess or deficiency. The power balance constraint is: In the formula, For the total power demand of the ship, This refers to the output power of the LNG engine. This refers to the net output power of the lithium battery subsystem. This represents the net output power of the supercapacitor subsystem.

[0030] The output power of LNG engines, lithium batteries, and supercapacitors is limited by their inherent physical characteristics (such as combustion efficiency, battery capacity, and capacitor capacity) and design parameters (such as rated power and maximum current). By defining these constraints, it is ensured that each energy system operates stably within safe thresholds, thereby avoiding equipment damage or performance degradation due to exceeding design limits. Overload or damage must be avoided. The power source constraints are: In the formula, This represents the maximum permissible output power of the LNG engine. This represents the maximum permissible output power of the lithium battery. This represents the maximum permissible output power of the supercapacitor.

[0031] The remaining energy of lithium-ion batteries and supercapacitors is a key factor affecting a ship's range and safety. By defining minimum and maximum remaining energy thresholds, it is possible to ensure that batteries and capacitors maintain reasonable energy levels during operation, avoiding over-discharge or over-charging. The energy constraints for lithium-ion batteries and supercapacitors are as follows: In the formula, This refers to the instantaneous remaining energy of a lithium battery. This represents the maximum usable energy of a lithium battery. This refers to the instantaneous residual energy of the supercapacitor. This represents the maximum energy that a supercapacitor can store.

[0032] S2. Construct a multi-objective optimization model based on the mathematical model of the energy system of the target LNG hybrid ship.

[0033] The optimization objectives of the multi-objective optimization model include minimizing fuel consumption, maximizing the lifespan of lithium batteries and supercapacitors, optimizing ship dynamic response performance, reducing carbon emissions, and controlling battery degradation costs.

[0034] Fuel consumption is a significant component of ship operating costs. Optimizing LNG consumption can reduce ship operating costs while simultaneously decreasing carbon emissions. The objective function for minimizing fuel consumption is: In the formula, Total fuel consumption The start time of the optimized time interval is, To optimize the end time of the time interval, The lifespan of lithium batteries and supercapacitors has a significant impact on the economy and reliability of ships. Optimizing charge and discharge control strategies can reduce the peak current and lifespan loss of lithium batteries. Furthermore, energy management strategies based on SOC limit protection can achieve reasonable load power allocation, improving the stability and economy of the ship's hybrid power system.

[0035] In hybrid-electric ships, the rapid charge-discharge characteristics of supercapacitors can reduce the charge-discharge frequency and depth of lithium batteries, thereby extending their lifespan. The objective function for maximizing the lifespan of both lithium batteries and supercapacitors is: In the formula, This is the comprehensive service life evaluation value for the energy storage system. This refers to the cumulative energy throughput of the lithium battery during its current operating cycle. This refers to the maximum permissible total energy throughput of a lithium battery over its lifespan. This represents the cumulative energy throughput of the supercapacitor during its current operating cycle. This represents the maximum permissible total energy throughput of a supercapacitor over its lifespan.

[0036] During ship operation, power demand exhibits significant and rapid dynamic changes under different operating conditions. Specifically, during complex conditions such as acceleration, turning, wave impact, and emergency avoidance, power demand fluctuates dramatically within a short period. For example, during ship acceleration, power demand may increase several times from the baseline level within seconds; while under wave impact, power demand may peak within milliseconds. To effectively cope with these rapid changes in power demand, the ship's propulsion system must possess high dynamic response performance. This requires the propulsion system to quickly sense changes in power demand and adjust its output power within a very short time to achieve rapid and stable power supply. The objective function for optimizing the ship's dynamic response performance is: In the formula, For dynamic response performance indicators, This refers to the actual total output power of the ship's propulsion system. This represents the ship's instantaneous total power demand.

[0037] To further reduce carbon emissions, this invention introduces a carbon emission penalty model, with the objective function for reducing carbon emissions being: In the formula, The objective function value of the emission penalty term. The penalty coefficient is... for Instantaneous nitrogen oxide emission rate at any given time for The maximum allowable emission rate threshold at any given time. For rectified linear unit function, if A result of 0 indicates that the limit has not been exceeded and no penalty will be imposed. The result was an excess, indicating that a penalty was required.

[0038] By using a penalty model, violations of emission standards can be punished, thereby incentivizing the optimization process to reduce carbon emissions.

[0039] To account for the degradation cost of lithium batteries, this invention introduces a battery degradation cost model, and the objective function for controlling the battery degradation cost is: In the formula, For the cost of battery degradation, This is the battery degradation cost coefficient. for The incremental degradation of the battery over time. The degradation rate coefficient, for Battery current at all times for The state of charge of the battery at any given time.

[0040] The battery degradation cost model is used to quantify the aging process of lithium batteries during use. By incorporating degradation costs, the optimization process can more comprehensively consider the battery's lifespan and avoid rapid aging caused by overcharging and discharging.

[0041] S3. A hierarchical collaborative optimization architecture is adopted to solve the multi-objective optimization model, obtain the global optimal solution that satisfies all constraints, and use the global optimal solution as the energy management strategy for the optimized target LNG hybrid power ship.

[0042] like Figure 2 As shown, the hierarchical collaborative optimization architecture includes two layers of optimization. The first layer of optimization includes a data preprocessing and enhancement module, a spatiotemporal feature extraction module, and a feature fusion and initial scheme generation module. The data preprocessing and enhancement module is used to enhance ship operation data and generate synthetic sample data. The spatiotemporal feature extraction module is used to extract spatial features and temporal dynamic features from the synthetic sample data. The feature fusion and initial scheme generation module is used to fuse spatial features and temporal dynamic features and generate an initial optimization scheme through a fully connected layer mapping.

[0043] The data preprocessing and enhancement module uses GAN to preprocess data, enhance ship operation data, and generate synthetic sample data. The spatiotemporal feature extraction module includes CNN and LSTM. CNN is used to extract spatial features from synthetic sample data, and LSTM is used to capture the temporal dynamic features of synthetic sample data and predict ship power demand. The synthetic sample data includes LNG consumption rate, lithium battery remaining energy, supercapacitor remaining energy, and ship power demand.

[0044] GANs consist of a generator and a discriminator. The generator produces synthetic data from random noise, while the discriminator attempts to distinguish between real and synthetic data. Through adversarial training, the generator continuously learns to generate more realistic data, while the discriminator learns to more accurately identify real and fake data. This mechanism not only generates high-quality synthetic data but also significantly expands the training dataset, thereby improving the model's generalization ability and robustness. The specific steps of the GAN training process are as follows: Figure 3 As shown, it includes: The generator receives a random noise vector and generates synthetic sample data. The synthesized sample data and real ship operation data are input into the discriminator to determine the authenticity of the data; The results of the data authenticity assessment are compared with the actual source of the data to calculate the adversarial loss. The weights of the discriminator and generator are updated separately using the backpropagation algorithm based on the adversarial loss. Training ends when the adversarial process reaches a dynamic equilibrium.

[0045] After data preprocessing, key features are extracted from the data. Convolutional Neural Networks (CNNs) are deep learning models that excel at processing data with spatial structure, such as images and time-series data. CNNs first extract local features through convolutional layers. These convolutional layers use multiple convolutional kernels (filters) to slide across the input data, generating feature maps. The mathematical expression for this is: .in, It is a convolution kernel. It is the input data. It is a bias term. It is an activation function. This represents a convolution operation. Then, pooling layers reduce the dimensionality of the feature map, decreasing computation while preserving important features. Common pooling operations include max pooling and average pooling. Max pooling extracts the maximum value from a local region, retaining the salient part of the feature. In this embodiment, the CNN is used to extract spatial features such as the changing trend of LNG consumption rate and the remaining energy levels of lithium batteries and supercapacitors. Specifically, convolutional layers extract local features through convolution operations, while pooling layers reduce the dimensionality of the feature map, decreasing computation while preserving important features. Through these operations, the CNN can effectively extract information useful for optimization from high-dimensional data, supporting the subsequent optimization process.

[0046] Besides spatial characteristics, the dynamic properties of time series data are equally important for the optimization process. For example... Figure 4 As shown, Long Short-Term Memory (LSTM) networks are a type of recurrent neural network specifically designed to handle long-term dependencies in time-series data. LSTMs, through their unique gating mechanisms, effectively capture dynamic changes in time-series data. An LSTM unit includes an input gate, a forget gate, and an output gate. These gating mechanisms allow the network to selectively retain or discard information, thus better handling long-term dependencies in time-series data.

[0047] The forget gate determines which historical information to discard from long-term memory based on the current input and the previous state using the sigmoid function. The expression is: The input gate filters valuable new information from the current input using the Sigmoid function and combines this with candidate memories generated by the Tanh function to jointly determine which information to update to the cell state. The expression is: The cell state is updated by combining the discard decision from the forget gate and the new information from the input gate. Finally, the output gate determines the final output of the LSTM unit based on the updated cell state and the current input. The expression for the output gate is: In the formula, This is the output vector of the forget gate. The current hidden state. Let the input vector be the input vector at the current time. The output vector of the input gate. Let be the hidden state vector from the previous time step. For candidate memories, This represents the updated cell state at the current moment. This is the unit state vector from the previous time step. It is the Sigmoid activation function. The output vector of the output gate. The hyperbolic tangent activation function is used. As weight, This is the bias parameter.

[0048] To improve the model's generalization ability and prevent overfitting to the training data, this embodiment introduces a regularization module after the LSTM layer. This module randomly discards a portion of neurons during training to prevent overfitting. The mathematical expression is as follows: In the formula, This is the final output after Dropout processing. This is the raw output of the LSTM unit. A randomly generated mask vector. This indicates element-wise multiplication.

[0049] In this embodiment, the input to the LSTM network is historical time-series data, specifically including historical power demand and the status of each energy system (such as battery SOC, generator efficiency, etc.). After the above-mentioned gating mechanism and regularization processing, the network finally outputs dynamic prediction information, namely the changing trend of ship power demand and the response characteristics of each energy system, providing dynamic information for the optimization process.

[0050] After data preprocessing and feature extraction, a combination of CNN and LSTM is used to generate an initial optimization scheme. Specifically, the features extracted by CNN and the dynamic information predicted by LSTM are integrated into a fully connected layer to generate the final prediction result. The fully connected layer maps the extracted features to the target output space, generating the initial optimization scheme. This initial scheme provides an important reference for the second-layer optimization, ensuring that the optimization process is conducted on a basis that more closely approximates the actual operating state. The combination of CNN and LSTM enables comprehensive modeling of ship operation data. CNN is responsible for extracting spatial features from the data, while LSTM is responsible for capturing the dynamic characteristics in the time series data. This combination not only fully utilizes the advantages of CNN in spatial feature extraction but also leverages the powerful capabilities of LSTM in time series modeling, providing a more comprehensive and accurate feature representation for the optimization process.

[0051] The second-level optimization includes an uncertainty handling module and a distributed solution and global coordination module. The uncertainty handling module is used to simulate various uncertain scenarios in ship operation and generate candidate optimization schemes based on the initial optimization scheme. The distributed solution and global coordination module is used to decompose the candidate optimization schemes into sub-problems and generate a global optimal solution that satisfies all constraints through alternating updates and coordination mechanisms. The global optimal solution is then used as the energy management strategy for the optimized target LNG hybrid power ship.

[0052] In actual operation, the ship's power requirements, energy system response, and environmental conditions are all subject to uncertainty. The uncertainty handling module uses the initial optimization scheme as the starting point for iteration, employs SGMHC to optimize the initial scheme, and introduces random noise in each iteration to simulate various uncertain scenarios in ship operation, generating multiple candidate optimization schemes covering different operating conditions. These optimization schemes not only cover normal operating conditions but also consider possible abnormal situations, thereby improving the robustness of the optimization results.

[0053] The specific formula for the iterative solution is: In the formula, For the first Sample solutions for the next iteration For the first Sample solutions for the next iteration Step size, For the first The probability distribution function of the sample schemes in the next iteration. It is a random noise vector.

[0054] The distributed solution and global coordination module uses ADMM to decompose candidate optimization schemes into subproblems, and generates a globally optimal solution that satisfies all constraints through alternating updates and coordination mechanisms. Specific steps include: The multi-objective optimization problem of candidate optimization solutions is formulated in the standard ADMM form; The global optimization problem is decomposed into local subproblems for the LNG fuel subsystem, lithium battery energy storage subsystem, and supercapacitor subsystem, and the optimal power allocation for each subproblem is solved independently. By alternately updating local variables and global multipliers, the solutions to the subproblems are coordinated, and a globally optimal solution that satisfies all constraints is obtained.

[0055] The update formula for local variables is: The update formula for the global multipliers is: In the formula, For power allocation local variables related to LNG engines, For the first Local variables related to power allocation of the LNG engine in the next iteration For the local objective function related to the LNG engine, For power allocation local variables related to energy storage, For the first Local variables related to power allocation in the next iteration of energy storage For the first Local variables related to power allocation in the next iteration of energy storage For the local objective function related to electrical energy storage, For penalty parameters, , The coefficient matrix, For the power requirements of ships, As a global multiplier, For the first The global multiplier value at the next iteration. For the first The global multiplier value at the next iteration.

[0056] The globally optimal solution that satisfies all constraints is used as the optimized energy management strategy for the target LNG hybrid power ship.

[0057] SGHMC generates multiple optimization scenarios through Monte Carlo sampling, while ADMM coordinates conflicts between different objectives through decomposition and alternating updates, ultimately generating a globally optimal solution as the optimized energy management strategy for the LNG hybrid power ship. This hierarchical optimization approach effectively addresses the complex operating conditions and uncertainties of hybrid power ship operations, achieving highly efficient optimization of the energy management system. This method not only improves the ship's energy efficiency and economy but also enhances operational reliability and safety.

[0058] S4. Implement optimized energy management for the target LNG hybrid power vessel based on the optimized energy management strategy for the target LNG hybrid power vessel.

[0059] After obtaining the globally optimal solution as the energy management strategy, this strategy is applied in real time to coordinate the power allocation of the LNG fuel subsystem, lithium battery energy storage subsystem, and supercapacitor subsystem. The optimization management process includes dynamically adjusting the output power of the LNG engine, the charge / discharge state of the lithium battery, and the charge / discharge power of the supercapacitor based on the ship's instantaneous power demand, ensuring that power balance constraints and the power constraints of each subsystem are met. In this way, multi-objective optimization is achieved, including minimizing fuel consumption, maximizing the lifespan of the energy storage system, optimizing dynamic response performance, reducing carbon emissions, and controlling battery degradation costs. Ultimately, this strategy can effectively cope with the complex operating conditions and uncertainties of ship operation, improving the ship's energy efficiency, economy, reliability, and safety.

[0060] Example 2 This embodiment provides an energy integrated management system for LNG hybrid power ships, including: The mathematical model building module is used to build a mathematical model of the energy system of the target LNG hybrid power ship; The multi-objective optimization model construction module is used to construct a multi-objective optimization model based on the mathematical model of the energy system of the target LNG hybrid power ship. The multi-objective optimization model solution module is used to solve the multi-objective optimization model using a hierarchical collaborative optimization architecture. This architecture comprises two layers of optimization. The first layer includes a data preprocessing and enhancement module, a spatiotemporal feature extraction module, and a feature fusion and initial scheme generation module. The data preprocessing and enhancement module enhances ship operation data; the spatiotemporal feature extraction module extracts spatial and temporal dynamic features from the enhanced ship operation data; and the feature fusion and initial scheme generation module fuses spatial and temporal dynamic features and generates an initial optimization scheme through a fully connected layer mapping. The second layer includes an uncertainty handling module and a distributed solution and global coordination module. The uncertainty handling module simulates various uncertain scenarios in ship operation and generates candidate optimization schemes based on the initial optimization scheme. The distributed solution and global coordination module decomposes the candidate optimization schemes into sub-problems and generates a globally optimal solution that satisfies all constraints through alternating updates and coordination mechanisms. This globally optimal solution is then used as the optimized energy management strategy for the target LNG hybrid power ship. The energy optimization management module is used to optimize the energy management of the target LNG hybrid vessel based on the optimized energy management strategy of the target LNG hybrid vessel.

[0061] The rest is the same as in Example 1.

[0062] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A method for optimizing energy integration management of LNG hybrid power ships, characterized in that, Includes the following steps: Establish a mathematical model of the energy system of the target LNG hybrid power ship; A multi-objective optimization model is constructed based on the mathematical model of the energy system of the target LNG hybrid power ship; The multi-objective optimization model is solved using a hierarchical collaborative optimization architecture, which includes two optimization layers. The first optimization layer includes a data preprocessing and enhancement module, a spatiotemporal feature extraction module, and a feature fusion and initial scheme generation module. The data preprocessing and enhancement module is used to enhance ship operation data and generate synthetic sample data. The spatiotemporal feature extraction module is used to extract spatial features and temporal dynamic features from the synthetic sample data. The feature fusion and initial scheme generation module is used to fuse the spatial features and temporal dynamic features and generate an initial optimization scheme through a fully connected layer mapping. The second-level optimization includes an uncertainty handling module and a distributed solution and global coordination module. The uncertainty handling module is used to simulate various uncertain scenarios in ship operation and generate candidate optimization schemes based on the initial optimization scheme. The distributed solution and global coordination module is used to decompose the candidate optimization schemes into sub-problems and generate a global optimal solution that satisfies all constraints through an alternating update and coordination mechanism. The global optimal solution is then used as the energy management strategy for the optimized target LNG hybrid power ship. The optimized energy management strategy for the target LNG hybrid vessel is used to achieve optimized energy management of the target LNG hybrid vessel.

2. The energy integration management optimization method for LNG hybrid power ships according to claim 1, characterized in that, The energy system of the LNG hybrid power ship includes an LNG fuel subsystem, a lithium battery energy storage subsystem, and a supercapacitor subsystem. The constraints of the energy system of the LNG hybrid power ship include power balance constraints, power source power constraints, and energy constraints of lithium batteries and supercapacitors.

3. The energy integration management optimization method for LNG hybrid power ships according to claim 1, characterized in that, The optimization objectives of the multi-objective optimization model include minimizing fuel consumption, maximizing the lifespan of lithium batteries and supercapacitors, optimizing ship dynamic response performance, reducing carbon emissions, and controlling battery degradation costs.

4. The energy integration management optimization method for LNG hybrid power ships according to claim 1, characterized in that, The data preprocessing and enhancement module uses GAN to preprocess the data, enhance the ship operation data, and generate synthetic sample data. The spatiotemporal feature extraction module includes CNN and LSTM. The CNN is used to extract spatial features from the synthetic sample data, and the LSTM is used to capture the temporal dynamic features of the synthetic sample data and predict the ship's power demand. The synthesized sample data includes LNG consumption rate, remaining energy of lithium batteries, remaining energy of supercapacitors, and ship power requirements.

5. The energy integration management optimization method for LNG hybrid power ships according to claim 4, characterized in that, The GAN includes a generator and a discriminator, and the specific steps of the GAN training process include: The generator receives a random noise vector and generates synthetic sample data. The synthesized sample data and real ship operation data are input into the discriminator to determine the authenticity of the data; The results of the data authenticity assessment are compared with the actual source of the data to calculate the adversarial loss. The weights of the discriminator and generator are updated respectively using the backpropagation algorithm based on the adversarial loss. Training ends when the adversarial process reaches a dynamic equilibrium.

6. The energy integration management optimization method for LNG hybrid power ships according to claim 1, characterized in that, The uncertainty processing module uses the initial optimization scheme as the initial point of iteration, optimizes the initial optimization scheme using SGMHC, and introduces random noise in each scheme iteration to simulate various uncertain scenarios in ship operation and generate multiple candidate optimization schemes covering different operating conditions.

7. The energy integration management optimization method for LNG hybrid power ships according to claim 6, characterized in that, The specific formula for the iterative process is as follows: In the formula, For the first Sample solutions for the next iteration For the first Sample solutions for the next iteration Step size, For the first The probability distribution function of the sample schemes in the next iteration. It is a random noise vector.

8. The energy integration management optimization method for LNG hybrid ships according to claim 1, characterized in that, The distributed solution and global coordination module uses ADMM to decompose the candidate optimization scheme into subproblems, and generates a globally optimal solution that satisfies all constraints through alternating update and coordination mechanisms. Specific steps include: The multi-objective optimization problem of the candidate optimization schemes is formulated in the standard ADMM form; The global optimization problem is decomposed into local subproblems for the LNG fuel subsystem, lithium battery energy storage subsystem, and supercapacitor subsystem, and the optimal power allocation for each subproblem is solved independently. By alternately updating local variables and global multipliers, the solutions to the subproblems are coordinated, and a globally optimal solution that satisfies all constraints is obtained.

9. The energy integration management optimization method for LNG hybrid power ships according to claim 8, characterized in that, The update formula for the local variables is: The update formula for the global multiplier is: In the formula, For power allocation local variables related to LNG engines, For the first Local variables related to power allocation of the LNG engine in the next iteration For the local objective function related to the LNG engine, For power allocation local variables related to energy storage, For the first Local variables related to power allocation in the next iteration of energy storage For the first Local variables related to power allocation in the next iteration of energy storage For the local objective function related to electrical energy storage, For penalty parameters, , The coefficient matrix, For the power requirements of ships, As a global multiplier, For the first The global multiplier value at the next iteration. For the first The global multiplier value at the next iteration.

10. An energy integrated management system for LNG hybrid power ships, characterized in that, include: The mathematical model building module is used to build a mathematical model of the energy system of the target LNG hybrid power ship; A multi-objective optimization model construction module is used to construct a multi-objective optimization model based on the mathematical model of the energy system of the target LNG hybrid power ship; A multi-objective optimization model solving module is used to solve the multi-objective optimization model using a hierarchical collaborative optimization architecture. The hierarchical collaborative optimization architecture includes two layers of optimization. The first layer of optimization includes a data preprocessing and enhancement module, a spatiotemporal feature extraction module, and a feature fusion and initial scheme generation module. The data preprocessing and enhancement module is used to enhance ship operation data. The spatiotemporal feature extraction module is used to extract spatial features and temporal dynamic features from the enhanced ship operation data. The feature fusion and initial scheme generation module is used to fuse the spatial features and temporal dynamic features and generate an initial optimization scheme through a fully connected layer mapping. The second-level optimization includes an uncertainty handling module and a distributed solution and global coordination module. The uncertainty handling module is used to simulate various uncertain scenarios in ship operation and generate candidate optimization schemes based on the initial optimization scheme. The distributed solution and global coordination module is used to decompose the candidate optimization schemes into sub-problems and generate a global optimal solution that satisfies all constraints through an alternating update and coordination mechanism. The global optimal solution is then used as the energy management strategy for the optimized target LNG hybrid power ship. The energy optimization management module is used to optimize the energy management of the target LNG hybrid vessel according to the optimized energy management strategy of the target LNG hybrid vessel.

Images