Energy optimization management method and system for extended-range hybrid power ship

Abstract

The invention relates to the technical field of ship power systems and energy management. The invention provides an energy optimization management method and system for an extended-range hybrid power ship. The method comprises the following steps: collecting ship multi-source heterogeneous data in real time, and constructing a multi-source heterogeneous data set; based on the multi-source heterogeneous data set, constructing a load prediction model based on a time sequence convolutional network and long and short term memory fusion, and dynamically outputting propulsive power demand probability distribution in a future navigation period; a multi-objective optimization model is established, and an improved multi-objective genetic algorithm is adopted to generate a Pareto frontier solution set; and on the basis of a rolling time domain optimization framework, in combination with real-time navigation situation awareness data, carrying out online correction on the Pareto solution set, and generating and executing an optimal energy distribution instruction set. The problems of an existing extended-range hybrid power ship energy management technology in the aspects of working condition adaptability, multi-energy coupling efficiency, real-time state sensing and predicting capacity and emission and economical efficiency balance are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of ship power systems and energy management, and in particular to an energy optimization management method and system for a range-extended hybrid ship. Background Art

[0002] With the acceleration of the green transformation of the global shipping industry, energy conservation, emission reduction and efficient operation of inland vessels have become important development directions. Inland waterways have complex operating conditions: the water is shallow during the dry season, there are many bends and shoals, and frequent operations such as beaching and sharp turns are required, which places stringent requirements on the flexibility, propulsion efficiency and energy economy of the ship's power system. As a transitional solution between traditional fuel ships and pure electric ships, extended-range hybrid ships combine the efficient electric drive of power batteries with the endurance supplementation capability of range extenders (such as small fuel engines or fuel cells). They can not only meet the energy needs of ships for long-distance navigation, but also significantly reduce carbon emissions and operating costs. They have become a research hotspot in the current field of ship power.

[0003] The core of extended-range hybrid ships lies in the optimized management of the energy system. Its core goal is to achieve the rational distribution of energy flow and maximize conversion efficiency while ensuring the ship's power performance. However, existing technologies still have the following key issues in energy management:

[0004] Insufficient adaptability to working conditions: Ships face complex and changeable load demands during navigation (such as full-speed sailing, low-speed cruising, cargo loading and unloading, port berthing, etc.). Traditional energy management strategies mostly use fixed threshold logic control (such as "start the range extender when the battery is exhausted"), which is difficult to dynamically match the real-time working conditions. As a result, the range extender frequently starts and stops or works in the low-efficiency range for a long time, exacerbating fuel consumption and mechanical losses.

[0005] Low multi-energy coupling efficiency: The energy coupling between the power battery and the range extender lacks global optimization, failing to fully consider the synergy between the battery's charge and discharge characteristics and the range extender's operating economy. For example, under high-load conditions, relying solely on the range extender for direct power supply may lead to overload operation; under low-load conditions, the range extender's excess energy is not effectively stored, resulting in energy waste.

[0006] Lack of real-time status perception and prediction capabilities: The existing system has insufficient monitoring accuracy for key parameters such as remaining battery capacity, range extender health status, and ship propulsion efficiency, and lacks forward-looking analysis of the navigation environment (such as wind and wave resistance and route planning). This leads to lagging energy allocation strategies and the inability to achieve preventive optimization.

[0007] Difficulty in balancing emissions and economy: Carbon emissions, fuel costs and power response speed must be taken into account in ship operations, but existing management methods are mostly guided by a single goal (such as maximizing cruising range or minimizing fuel consumption) and have not built a multi-objective optimization model, making it difficult to achieve optimal comprehensive benefits in different navigation stages (such as ocean shipping and inland waterway shipping). Summary of the Invention

[0008] The purpose of the present invention is to provide an energy optimization management method and system for extended-range hybrid ships, aiming to solve the problems existing in the energy management technology of extended-range hybrid ships in terms of adaptability to working conditions, multi-energy coupling efficiency, real-time state perception and prediction capabilities, and balance between emissions and economy.

[0009] The present invention is achieved through the following technical solutions:

[0010] An energy optimization management method for a range-extended hybrid ship comprises the following steps:

[0011] Real-time collection of multi-source heterogeneous ship data, including operating load parameters, power battery remaining capacity matrix, range extender efficiency characteristic curve, propulsion system efficiency map, and navigation environment parameters, to construct a multi-source heterogeneous data set;

[0012] Based on multi-source heterogeneous data sets, a load prediction model based on the fusion of time series convolutional networks and long-short-term memory is constructed to dynamically output the probability distribution of propulsion power demand in the future navigation cycle;

[0013] A multi-objective optimization model was established, with the range extender fuel consumption rate, battery health attenuation coefficient, and equivalent carbon emission index as optimization variables, and an improved multi-objective genetic algorithm was used to generate the Pareto frontier solution set.

[0014] Based on the rolling horizon optimization framework and combined with real-time navigation situation awareness data, the Pareto solution set is corrected online to generate and execute the optimal energy allocation instruction set.

[0015] Optionally, the specific process of constructing a load prediction model based on a time series convolutional network and a fusion of long and short-term memory based on a multi-source heterogeneous data set and dynamically outputting the probability distribution of propulsion power demand in the future navigation cycle is as follows:

[0016] Normalize the propeller torque-speed time series in the operating load parameters, the charge-discharge current-temperature correlation tensor in the power battery remaining capacity state matrix, and the river depth-flow velocity three-dimensional point cloud data in the navigation environment parameters. Use a sliding window mechanism to uniformly map the different-frequency sampling data to the same time base.

[0017] A load prediction model consisting of parallel branches was built. The first branch used a stack of dilated causal convolutions to extract high-frequency fluctuations in the propulsion system efficiency map. The second branch used a bidirectional gated recurrent unit to capture long-range dependencies in the range extender efficiency curve. The third branch processed the spatial topology of the navigation environment parameters using a three-dimensional sparse convolutional network.

[0018] The feature vectors output by each branch are subjected to multi-scale feature fusion to generate cross-modal fusion features. The contribution ratio of the cross-modal fusion features to the features of each branch is dynamically adjusted through a learnable gating vector, and high-dimensional features are adaptively weighted and spliced.

[0019] A Monte Carlo Dropout mechanism is introduced after the fully connected layer to perform multi-step rolling forecasts on the propulsion power demand time series. Multiple forward propagation sampling is used to generate a probability density function of the power demand value, and a probability distribution cloud diagram containing the mean, variance, and confidence interval is output.

[0020] Based on the propulsion motor current-voltage feedback signal obtained in real time during navigation, the sliding window statistics of the prediction error are calculated. When the mean square error continuously exceeds the dynamic threshold, the model parameters are fine-tuned. The online knowledge distillation algorithm is used to migrate the newly added data features to the pre-trained load prediction model.

[0021] Optionally, the specific process of performing multi-scale feature fusion on the feature vectors output by each branch to generate cross-modal fusion features is as follows:

[0022] The high-frequency fluctuation feature vector of the first branch, the long-range dependency feature vector of the second branch, and the spatial topology feature vector of the third branch are linearly transformed to generate corresponding query vectors, construction vectors, and value vectors;

[0023] A multi-head attention mechanism is used to split the query vector, the construction vector, and the value vector into several sub-heads, and each sub-head independently calculates the attention weight matrix;

[0024] After concatenating the outputs of each sub-head, cross-modal fusion features are generated through residual connection and layer normalization.

[0025] Optionally, the specific process of establishing a multi-objective optimization model, taking the range extender fuel consumption rate, battery health attenuation coefficient, and equivalent carbon emission index as optimization variables, and using an improved multi-objective genetic algorithm to generate a Pareto frontier solution set is as follows:

[0026] A multi-objective optimization model with time coupling constraints is constructed, and the objective function set is shown in the following formula (1):

[0027]

[0028] Among them, minf1 represents the minimum fuel consumption rate; k f Indicates the slope of the range extender fuel consumption curve; P gen (t) represents the output power of the range extender at time t; b f represents the fixed fuel consumption during start-stop or no-load; δ represents the start-stop indicator function of the range extender; represents the total time step of the optimization time domain; minf2 represents the minimum battery health attenuation coefficient; I bat (t) represents the charge and discharge current of the battery at time t; SOC(t) represents the remaining capacity of the power battery at time t; α represents the reference coefficient of the influence of current on aging; β represents the nonlinear influence index of current amplitude; γ represents the adjustment coefficient of the remaining capacity of the power battery to aging attenuation; minf3 represents the minimum equivalent carbon emission index; c co2 Indicates the carbon emissions per unit power output of the range extender; Indicates the rate of change of battery state of charge; E grid represents the carbon emission factor of grid electricity; λ represents the conversion coefficient of grid charging carbon emissions, which is used to balance the carbon emission contributions of different energy sources;

[0029] The constraint condition set is shown in formula (2):

[0030]

[0031] Among them, P prop (t) represents the propulsion system power demand at time t; P bat (t) represents the power provided by the power battery at time t; P loss (t) represents the system power loss at time t; SOC min Indicates the battery's lowest state of charge; SOC max Indicates the highest state of charge of the battery; Indicates the maximum output power of the range extender; Indicates the rated maximum charge and discharge current of the battery;

[0032] An improved multi-objective genetic algorithm is used to solve the problem; the dynamic crossover probability is calculated as shown in the following formula (3):

[0033]

[0034] Among them, p c represents the crossover probability after dynamic adjustment; p c0 represents the initial crossover probability; g represents the current number of iterations; G max represents the maximum number of iterations; Δp c Indicates the crossover probability adjustment amplitude; d i represents the average Euclidean distance from the i-th individual to other individuals in the population; d maxrepresents the maximum individual distance in the population; N represents the total number of individuals in each generation;

[0035] The adaptive mutation intensity is calculated as shown in the following formula (4):

[0036]

[0037] Among them, σ m represents the adaptive mutation intensity; σ m0 represents the initial mutation intensity; f dom (x) represents the dominance strength of individual x; max(f dom ) represents the maximum dominance strength value of all individuals in the current population;

[0038] Fast non-dominated sorting based on the Kriging surrogate model, where the individual dominance strength is calculated as follows:

[0039]

[0040] Among them, x represents the current individual; y represents the population P op Other individuals in P op represents the current population; l represents a constant, which is used to control the influence of distance;

[0041] Through the elite retention strategy and crowding distance screening, a Pareto optimal solution set that satisfies the following formula (6) is generated:

[0042]

[0043] Among them, x * represents the individual in the Pareto optimal solution set; X pareto represents the Pareto optimal solution set; and there is at least one i such that f i (y)≤f i (x * ).

[0044] Optionally, the specific process of performing online correction on the Pareto solution set based on the rolling horizon optimization framework and combining the real-time navigation situation awareness data to generate and execute the optimal energy allocation instruction set is as follows:

[0045] A rolling time window is established to divide the optimization time domain into a fixed-length prediction interval and an execution interval. Within each control cycle, the embedded edge computing unit obtains the latest ship attitude angle, the instantaneous gradient of the remaining battery capacity, and the heat map of the channel obstacle distribution to build a real-time situation awareness vector.

[0046] Based on the real-time situational awareness vector, an adaptive Kalman filter algorithm is used to compensate for the error in the output of the load prediction model and generate a corrected probability distribution of propulsion power demand.

[0047] The corrected probability distribution is input into the multi-objective optimization model. Based on the deviation between the measured value and the predicted value of the current battery health attenuation coefficient, the time-varying weight coefficients of β and γ in the objective function are dynamically adjusted to generate an updated Pareto solution set.

[0048] The system evaluates the satisfaction of each solution with respect to fuel consumption rate, battery health, and carbon emissions based on a fuzzy membership function. Poor solutions with membership below a dynamic threshold are eliminated. A cosine similarity matching algorithm is then used to select the candidate solution that is closest to the historical optimal energy allocation pattern from the remaining solutions.

[0049] The energy allocation instructions corresponding to the selected candidate solution are decomposed into the range extender power setting value, battery charge and discharge rate, and propulsion motor torque distribution ratio, and are sent to each actuator via the CAN bus protocol;

[0050] At the end of the execution interval, the deviation matrix between the actual energy consumption data and the predicted value is collected, and the back-propagation reinforcement learning algorithm is used to update the weight parameters of the rolling horizon optimization framework.

[0051] Optionally, perform hierarchical energy routing control:

[0052] During the propulsion power mutation stage, the power battery pack is called upon first for power compensation, and during the steady-state cruising stage, it switches to the range extender's high-efficiency range constant power output mode, and realizes adaptive feedback storage of regenerative braking energy and surplus energy of the range extender through a bidirectional DC / DC converter.

[0053] Optionally, a cost function for starting and stopping the range extender is established. When the remaining capacity of the power battery enters the buffer threshold interval, the timing of the range extender intervention is dynamically adjusted according to the historical start-stop frequency and current emission constraints, so that mechanical loss and fuel economy achieve a nonlinear balance.

[0054] Based on the same inventive concept, the present invention further provides an energy optimization management system for a range-extended hybrid vessel, which is used to implement the energy optimization management method for the range-extended hybrid vessel, comprising:

[0055] The multi-source heterogeneous data acquisition module is used to collect multi-source heterogeneous data of ships in real time, including operating load parameters, power battery remaining capacity state matrix, range extender efficiency characteristic curve, propulsion system efficiency map and navigation environment parameters, and construct a multi-source heterogeneous data set;

[0056] A cross-modal load prediction module is used to build a load prediction model based on a fusion of a temporal convolutional network and long-short-term memory based on multi-source heterogeneous data sets, dynamically outputting the probability distribution of propulsion power demand within the future navigation cycle;

[0057] A multi-objective dynamic optimization module is used to establish a multi-objective dynamic programming function, using the range extender fuel consumption rate, battery health attenuation coefficient, and equivalent carbon emission index as optimization variables, and using an improved multi-objective genetic algorithm to generate a Pareto frontier solution set;

[0058] The rolling horizon correction module is used to perform online correction on the Pareto solution set based on the rolling horizon optimization framework and combined with real-time navigation situation awareness data to generate the optimal energy allocation instruction set;

[0059] A hierarchical energy routing controller is used to implement hierarchical energy routing control. During the propulsion power mutation phase, the power battery pack is prioritized for power compensation. During the steady-state cruise phase, the power output mode is switched to the range extender's high-efficiency range constant power output mode. A bidirectional DC / DC converter is used to adaptively store regenerative braking energy and excess energy from the range extender.

[0060] The range extender start-stop decision module is used to establish a cost function for starting and stopping the range extender. When the remaining capacity of the power battery enters the buffer threshold range, the range extender intervention timing is dynamically adjusted based on the historical start-stop frequency and current emission constraints to achieve a nonlinear balance between mechanical loss and fuel economy.

[0061] Among them, the output end of the multi-source heterogeneous data acquisition module is connected to the input end of the cross-modal load prediction module, the output end of the cross-modal load prediction module is respectively connected to the input end of the multi-objective dynamic optimization module and the input end of the rolling time domain correction module, the output end of the multi-objective dynamic optimization module is connected to the input end of the rolling time domain correction module, the output end of the rolling time domain correction module is respectively connected to the input end of the hierarchical energy routing controller and the input end of the cross-modal load prediction module, and the hierarchical energy routing controller is connected to the range extender start-stop decision module.

[0062] Based on the same inventive concept, the present invention also provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the above-mentioned energy optimization management method for extended-range hybrid ships.

[0063] Based on the same inventive concept, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which implements the above-mentioned energy optimization management method for extended-range hybrid ships when the computer program is executed by a processor.

[0064] The technical solution of the present invention has at least the following advantages and beneficial effects:

[0065] Significantly improve the dynamic adaptability of working conditions: By collecting multi-source heterogeneous data in real time and combining the load prediction model with the fusion of time series convolutional network and long-short-term memory, it can accurately capture the dynamic load demand of ships under complex working conditions such as beaching, sharp turns, loading and unloading, and predict the probability distribution of propulsion power in the future navigation cycle in advance; compared with the traditional fixed threshold control strategy, it can effectively avoid the frequent start-stop or inefficient operation of the range extender due to sudden changes in working conditions, reduce fuel consumption and mechanical losses, and improve the adaptability and stability of the power system in extreme working conditions such as shallow water, many bends and sharp shoals during the dry season.

[0066] Significantly improve the efficiency of multi-energy coupling: Construct a multi-objective optimization model with the range extender's fuel consumption rate, battery health attenuation coefficient, and equivalent carbon emissions as variables, generate a Pareto frontier solution set through an improved multi-objective genetic algorithm, and achieve global coordinated optimization of the energy flow of the power battery and the range extender; under high-load conditions, the system can intelligently coordinate the power output of the range extender and the battery to avoid overloading the range extender; under low load conditions, surplus energy is recovered through energy storage strategies to reduce energy waste; fully consider the coupling relationship between the battery charging and discharging characteristics and the operating economy of the range extender to maximize the energy conversion efficiency, fundamentally solving the problem of lack of global optimization of energy coupling in traditional systems.

[0067] Enhanced real-time status perception and forward-looking optimization capabilities: Based on a real-time monitoring system with multi-source data fusion, it accurately obtains key parameters such as remaining battery capacity, range extender health status, and propulsion system efficiency. Combined with forward-looking analysis of the navigation environment (such as wind and wave resistance, route planning), it uses a rolling time domain optimization framework to perform online corrections to the Pareto solution set. This breaks through the limitations of the traditional system's state perception lag, realizes dynamic adjustment and preventive optimization of energy allocation strategies, and can respond to changes in operating conditions in advance, ensuring that the range extender always operates in the high-efficiency range, and improving the reliability of system operation and energy utilization efficiency.

[0068] Achieve a multi-objective balance among emissions, economy, and power performance: Build a multi-objective optimization model, incorporate indicators such as fuel costs, carbon emissions, and battery health degradation into a unified optimization system, and cover the comprehensive benefit requirements of different navigation stages (such as full-speed sailing, low-speed cruising, and port berthing) through the Pareto frontier solution set; while ensuring the ship's power response speed, it takes into account both low-carbon emissions and economic goals. It is especially suitable for scenarios that are sensitive to environmental protection and operating costs, such as inland shipping, and provides an efficient and sustainable energy management solution for the green transformation of ships. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 This is a flow chart of an energy optimization management method for a range-extended hybrid ship according to an embodiment of the present invention;

[0070] Figure 2Schematic diagram of the structure of an energy optimization management system for a range-extended hybrid ship according to an embodiment of the present invention. DETAILED DESCRIPTION

[0071] The following is a specific implementation method with reference to the accompanying drawings.

[0072] Reference Figure 1 , an energy optimization management method for a range-extended hybrid ship, comprising the following steps:

[0073] Step 1: Real-time acquisition of multi-source heterogeneous ship data, including operating load parameters, the remaining battery capacity matrix, the range extender efficiency characteristic curve, the propulsion system efficiency map, and navigation environment parameters, to construct a multi-source heterogeneous dataset. A torque sensor (such as a strain gauge torque sensor) and a speed encoder are installed on the propeller shaft to collect real-time torque and speed time series at a sampling frequency of 100 Hz. The propulsion motor's real-time power is acquired through the motor controller, and motor input parameters are monitored using voltage and current sensors. The voltage, charge and discharge current, and temperature of each battery cell are collected to construct a three-dimensional state matrix at a sampling frequency of 50 Hz. The diesel generator's output power, fuel consumption rate, and speed are collected, and an efficiency curve is pre-fitted through bench testing and uploaded to the ship's central control system in real time. The motor speed, torque, and efficiency are monitored to construct a two-dimensional efficiency map, which is updated in real time using dynamic test data. A multibeam sonar and electromagnetic current meter are installed to acquire three-dimensional point cloud data with a resolution of 0.5 m and a sampling frequency of 10 Hz. Transmit the data from each sensor to the edge computing unit via the ship's local area network (CAN bus or Ethernet). Use the GPS timing module to add a unified timestamp to all data to ensure time alignment across modal data. For heterogeneous sampling data (such as low-frequency sampling of environmental parameters and high-frequency sampling of battery data), resample using a sliding window with a fixed time interval (such as 1s) and uniformly map it to the same time base through linear interpolation or mean filtering. Standardize or normalize numerical data (such as current, temperature, and power) to eliminate dimensionality effects. Use the isolation forest or sliding median method to identify and repair abnormal sensor data to prevent noise from interfering with subsequent model training.

[0074] Step 2: Based on multi-source heterogeneous data sets, a load prediction model based on the fusion of time series convolutional network and long short-term memory is constructed to dynamically output the probability distribution of propulsion power demand in the future navigation cycle.

[0075] In some embodiments, based on multi-source heterogeneous data sets, a load prediction model based on a time series convolutional network and long short-term memory fusion is constructed to dynamically output the probability distribution of propulsion power demand in the future navigation cycle. The specific process is as follows:

[0076] Normalize the propeller torque-speed time series in the operating load parameters, the charge-discharge current-temperature correlation tensor in the power battery remaining capacity state matrix, and the river depth-flow velocity three-dimensional point cloud data in the navigation environment parameters. Use a sliding window mechanism to uniformly map the different-frequency sampling data to the same time base.

[0077] A load prediction model consisting of parallel branches was built. The first branch used a stack of dilated causal convolutions to extract high-frequency fluctuations in the propulsion system efficiency map. The second branch used a bidirectional gated recurrent unit to capture long-range dependencies in the range extender efficiency curve. The third branch processed the spatial topology of the navigation environment parameters using a three-dimensional sparse convolutional network.

[0078] The feature vectors output by each branch are subjected to multi-scale feature fusion to generate cross-modal fusion features. The contribution ratio of the cross-modal fusion features to the features of each branch is dynamically adjusted through a learnable gating vector, and high-dimensional features are adaptively weighted and spliced.

[0079] A Monte Carlo Dropout mechanism is introduced after the fully connected layer to perform multi-step rolling forecasts on the propulsion power demand time series. Multiple forward propagation sampling is used to generate a probability density function of the power demand value, and a probability distribution cloud diagram containing the mean, variance, and confidence interval is output.

[0080] Based on the propulsion motor current-voltage feedback signal obtained in real time during navigation, the sliding window statistics of the prediction error are calculated. When the mean square error continuously exceeds the dynamic threshold, the model parameters are fine-tuned. The online knowledge distillation algorithm is used to migrate the newly added data features to the pre-trained load prediction model.

[0081] In some embodiments, the specific process of performing multi-scale feature fusion on the feature vectors output by each branch to generate cross-modal fusion features is as follows:

[0082] The high-frequency fluctuation feature vector of the first branch, the long-range dependency feature vector of the second branch, and the spatial topology feature vector of the third branch are linearly transformed to generate corresponding query vectors, construction vectors, and value vectors;

[0083] A multi-head attention mechanism is used to split the query vector, the construction vector, and the value vector into several sub-heads, and each sub-head independently calculates the attention weight matrix;

[0084] After concatenating the outputs of each sub-head, cross-modal fusion features are generated through residual connection and layer normalization.

[0085] Step 3: Establish a multi-objective optimization model, use the range extender fuel consumption rate, battery health attenuation coefficient, and equivalent carbon emission index as optimization variables, and use an improved multi-objective genetic algorithm to generate the Pareto frontier solution set.

[0086] In some embodiments, a multi-objective optimization model is established, with the range extender fuel consumption rate, battery health attenuation coefficient, and equivalent carbon emission index as optimization variables, and the specific process of generating a Pareto frontier solution set using an improved multi-objective genetic algorithm is as follows:

[0087] A multi-objective optimization model with time coupling constraints is constructed, and the objective function set is shown in the following formula (1):

[0088]

[0089] Among them, minf1 represents the minimum fuel consumption rate; k f Indicates the slope of the range extender fuel consumption curve; P gen (t) represents the output power of the range extender at time t; b f represents the fixed fuel consumption during start-stop or no-load; δ represents the start-stop indicator function of the range extender; represents the total time step of the optimization time domain; minf2 represents the minimum battery health attenuation coefficient; I bat (t) represents the charge and discharge current of the battery at time t; SOC(t) represents the remaining capacity of the power battery at time t; α represents the reference coefficient of the influence of current on aging; β represents the nonlinear influence index of current amplitude; γ represents the adjustment coefficient of the remaining capacity of the power battery to aging attenuation; minf3 represents the minimum equivalent carbon emission index; c co2 Indicates the carbon emissions per unit power output of the range extender; Indicates the rate of change of battery state of charge; E grid represents the carbon emission factor of grid electricity; λ represents the conversion coefficient of grid charging carbon emissions, which is used to balance the carbon emission contributions of different energy sources;

[0090] The constraint condition set is shown in formula (2):

[0091]

[0092] Among them, P prop (t) represents the propulsion system power demand at time t; P bat (t) represents the power provided by the power battery at time t; P loss (t) represents the system power loss at time t; SOC min Indicates the battery's lowest state of charge; SOC max Indicates the highest state of charge of the battery; Indicates the maximum output power of the range extender;

[0093] Indicates the rated maximum charge and discharge current of the battery;

[0094] An improved multi-objective genetic algorithm is used to solve the problem; the dynamic crossover probability is calculated as shown in the following formula (3):

[0095]

[0096] Among them, p c represents the crossover probability after dynamic adjustment; p c0 represents the initial crossover probability; g represents the current number of iterations; G max represents the maximum number of iterations; Δp c Indicates the crossover probability adjustment amplitude; d i represents the average Euclidean distance from the i-th individual to other individuals in the population; d max represents the maximum individual distance in the population; N represents the total number of individuals in each generation;

[0097] The adaptive mutation intensity is calculated as shown in the following formula (4):

[0098]

[0099] Among them, σ m represents the adaptive mutation intensity; σ m0 represents the initial mutation intensity; f dom (x) represents the dominance strength of individual x; max(f dom ) represents the maximum dominance strength value of all individuals in the current population;

[0100] Fast non-dominated sorting based on the Kriging surrogate model, where the individual dominance strength is calculated as follows:

[0101]

[0102] Among them, x represents the current individual; y represents the population P op Other individuals in P op represents the current population; l represents a constant, which is used to control the influence of distance;

[0103] Through the elite retention strategy and crowding distance screening, a Pareto optimal solution set that satisfies the following formula (6) is generated:

[0104]

[0105] Among them, x * represents the individual in the Pareto optimal solution set; X pareto represents the Pareto optimal solution set; and there is at least one i such that f i (y)≤f i (x * ).

[0106] Step 4: Based on the rolling horizon optimization framework and combined with real-time navigation situation awareness data, the Pareto solution set is corrected online to generate and execute the optimal energy allocation instruction set.

[0107] In some embodiments, based on the rolling horizon optimization framework and combined with real-time navigation situational awareness data, the Pareto solution set is corrected online to generate and execute the optimal energy allocation instruction set. The specific process is as follows:

[0108] A rolling time window is established to divide the optimization time domain into a fixed-length prediction interval and an execution interval. Within each control cycle, the embedded edge computing unit obtains the latest ship attitude angle, the instantaneous gradient of the remaining battery capacity, and the heat map of the channel obstacle distribution to build a real-time situation awareness vector.

[0109] Based on the real-time situational awareness vector, an adaptive Kalman filter algorithm is used to compensate for the error in the output of the load prediction model and generate a corrected probability distribution of propulsion power demand.

[0110] The corrected probability distribution is input into the multi-objective optimization model. Based on the deviation between the measured value and the predicted value of the current battery health attenuation coefficient, the time-varying weight coefficients of β and γ in the objective function are dynamically adjusted to generate an updated Pareto solution set.

[0111] The system evaluates the satisfaction of each solution with respect to fuel consumption rate, battery health, and carbon emissions based on a fuzzy membership function. Poor solutions with membership below a dynamic threshold are eliminated. A cosine similarity matching algorithm is then used to select the candidate solution that is closest to the historical optimal energy allocation pattern from the remaining solutions.

[0112] The energy allocation instructions corresponding to the selected candidate solution are decomposed into the range extender power setting value, battery charge and discharge rate, and propulsion motor torque distribution ratio, and are sent to each actuator via the CAN bus protocol;

[0113] At the end of the execution interval, the deviation matrix between the actual energy consumption data and the predicted value is collected, and the back-propagation reinforcement learning algorithm is used to update the weight parameters of the rolling horizon optimization framework.

[0114] In some embodiments, hierarchical energy routing control is implemented:

[0115] During the propulsion power mutation stage, the power battery pack is called upon first for power compensation, and during the steady-state cruising stage, it switches to the range extender's high-efficiency range constant power output mode, and realizes adaptive feedback storage of regenerative braking energy and surplus energy of the range extender through a bidirectional DC / DC converter.

[0116] When it is detected that the rate of change of the propulsion power demand exceeds a preset threshold (such as an instantaneous power fluctuation ≥ 20% of the rated power), it is determined to have entered the power mutation stage (such as ship acceleration, sharp turns, encountering sudden water impact, etc.). The power battery pack is activated first as the main energy supply source, and the bidirectional DC / DC converter quickly responds to power demand and directly provides instantaneous power compensation to the propulsion motor; the range extender maintains the current operating state or enters idle standby mode to avoid mechanical loss and emission surges due to frequent starts and stops; the battery discharge current is monitored in real time to ensure that it does not exceed the rated maximum charge and discharge current, and the discharge rate is dynamically adjusted through the battery management system to prevent over-discharge from damaging the battery health.

[0117] When the propulsion power demand fluctuates by ≤5% of the rated power for 5 consecutive minutes, it is determined that the vehicle has entered the steady-state cruise phase (such as a straight and steady-speed driving condition). The range extender is switched to the main energy supply source. According to the efficiency characteristic curve of the range extender, the fixed output power is set to the midpoint value of the high-efficiency range, and the power stability is maintained through closed-loop PID control. The power battery pack enters the floating charge state or the low-current charge and discharge mode (charge and discharge current ≤10% of the rated current), keeping the remaining capacity of the power battery in the middle optimization range (such as 40%-60%) to avoid capacity attenuation caused by deep charge and discharge. The operating parameters of the range extender (such as speed, torque, and fuel consumption rate) are monitored in real time, and the efficiency map is dynamically corrected through an adaptive filtering algorithm to ensure that the operating point is always in the optimal fuel economy area.

[0118] When the ship slows down or brakes, the propulsion motor switches to generator mode, and the regenerative braking energy (power range is 10%-30% of the rated power) is converted into a voltage / current signal suitable for battery storage through a bidirectional DC / DC converter. It is first stored in the power battery pack and fully recovered when the remaining capacity is less than 80%. When the remaining capacity is greater than or equal to 80%, it is partially recovered or used to power auxiliary equipment. When the output power of the range extender is greater than the current propulsion demand (such as the range extender's rated power output is greater than the actual load during steady-state cruising), the surplus energy is charged to the battery pack through the bidirectional DC / DC converter. The charging current is dynamically adjusted according to the remaining capacity. For example, when the remaining capacity is less than 50%, 0.5C fast charging is used, and when the remaining capacity is greater than or equal to 50% and less than 80%, 0.2C slow charging is used. Through a hierarchical control strategy, the system can automatically identify different operating conditions such as power mutation, steady-state cruising, and low-speed mooring. It can complete energy routing switching without human intervention and adapt to complex and changing navigation environments (such as sharp turns, shallows, sudden water flow changes, etc.)

[0119] In some embodiments, a cost function for starting and stopping the range extender is established. When the remaining capacity of the power battery enters the buffer threshold interval, the timing of the range extender intervention is dynamically adjusted based on the historical start-stop frequency and current emission constraints, so that mechanical loss and fuel economy achieve a nonlinear balance.

[0120] The preset buffer threshold range of the remaining capacity (SOC) of the power battery is [SOC low ,SOC high ], where SOC low Indicates the minimum warning value for starting the range extender (such as 20%); SOC high Indicates the maximum warning value for shutting down the range extender (e.g. 80%). When the real-time collected SOC(t) enters this interval, the range extender start-stop decision mechanism is triggered.

[0121] Real-time acquisition of current SOC(t), battery charge and discharge current I bat (t), the historical start and stop frequency of the range extender and the current emission constraints, such as the regional carbon emission limit C limit .

[0122] Construct the range extender start-stop cost function as shown in the following formula (7):

[0123] J start-stop =ω1·J mech +ω2·J fuel +ω3·J emission +ω4·J freq (7)

[0124] Among them, J mech Represents the mechanical loss cost; J fuel Indicates fuel economy cost; J emission represents the emission constraint cost; J freq represents the historical frequency penalty; ω1, ω2, ω3 and ω4 represent the corresponding weights respectively;

[0125] Each start and stop introduces fixed mechanical losses (such as bearing wear and cylinder thermal stress), as shown in the following formula (8):

[0126] J mech =c mech ·δ(Δstate) (8)

[0127] Wherein, δ(Δstate) represents the start-stop state change indicator function (takes 1 when the start-stop state switches, otherwise takes 0); c mech Indicates the mechanical loss coefficient of a single start and stop;

[0128] Fuel consumption and output power P when the range extender is running gem (t) is positively correlated, combined with the start-stop no-load loss, as shown in the following formula (9):

[0129] J fuel =∫ t t+ΔT (k f ·P gen (t N )+bf ·δ(on))dt N (9)

[0130] Among them, δ(on) represents the range extender operation indicator function; t N Represents the time points that change continuously within the interval [t, t+ΔT], which is used to accumulate the fuel consumption of the range extender during this period;

[0131] If the equivalent carbon emissions f3 of the range extender exceed the regional limit C limit , introduce the penalty term J emission , the penalty term is shown in the following formula (10):

[0132] J emission =max(0,f3-C limit )·λ penalty (10)

[0133] Among them, λ penalty Indicates the penalty coefficient for exceeding emission standards;

[0134] Suppress frequent start and stop, based on the historical start and stop frequency N start-stop Dynamic adjustment is shown in the following formula (11):

[0135]

[0136] Among them, α freq Represents the frequency penalty weight; N max Indicates the maximum allowed start and stop frequency (e.g. 5 times per hour).

[0137] Start conditions:

[0138] When SOC(t)≤SOC low The range extender is started when one of the following conditions is met:

[0139] Condition 1: SOC(t) is predicted to continue to decrease to SOC within a preset time in the future min the following;

[0140] Condition 2: Current J start-stop Middle J fuel +J mech Less than the health degradation cost caused by deep discharge of the battery (i.e., item f2).

[0141] Closing conditions:

[0142] When SOC(t)≥SOC high And when the range extender is running J emission ≥C limit , or J freq When the threshold is exceeded, a shutdown is triggered.

[0143] According to the real-time emission constraints and historical start-stop data, the cost function weights ω1, ω2, ω3 and ω4 are dynamically adjusted (for example, ω3 is increased in emission-sensitive areas and ω4 is increased in high-frequency start-stop scenarios).

[0144] Start-stop decision-making is linked with hierarchical energy routing control: when the range extender is started, the range extender is given priority to provide steady-state power, and the battery only compensates for transient power fluctuations; when the range extender is turned off, regenerative braking energy is recovered to the battery through a bidirectional DC / DC converter.

[0145] Output control command: Generate range extender start / stop signal δ(on) and power setting value P gen (t), is sent to the actuator via the CAN bus.

[0146] Based on the same inventive concept, corresponding to any of the above embodiments, refer to Figure 2 The present invention provides an energy optimization management system for a range-extended hybrid ship, which is used to implement the aforementioned energy optimization management method for a range-extended hybrid ship, comprising:

[0147] The multi-source heterogeneous data acquisition module is used to collect multi-source heterogeneous data of ships in real time, including operating load parameters, power battery remaining capacity state matrix, range extender efficiency characteristic curve, propulsion system efficiency map and navigation environment parameters, and construct a multi-source heterogeneous data set;

[0148] A cross-modal load prediction module is used to build a load prediction model based on a fusion of a temporal convolutional network and long-short-term memory based on multi-source heterogeneous data sets, dynamically outputting the probability distribution of propulsion power demand within the future navigation cycle;

[0149] A multi-objective dynamic optimization module is used to establish a multi-objective dynamic programming function, using the range extender fuel consumption rate, battery health attenuation coefficient, and equivalent carbon emission index as optimization variables, and using an improved multi-objective genetic algorithm to generate a Pareto frontier solution set;

[0150] The rolling horizon correction module is used to perform online correction on the Pareto solution set based on the rolling horizon optimization framework and combined with real-time navigation situation awareness data to generate the optimal energy allocation instruction set;

[0151] A hierarchical energy routing controller is used to implement hierarchical energy routing control. During the propulsion power mutation phase, the power battery pack is prioritized for power compensation. During the steady-state cruise phase, the power output mode is switched to the range extender's high-efficiency range constant power output mode. A bidirectional DC / DC converter is used to adaptively store regenerative braking energy and excess energy from the range extender.

[0152] The range extender start-stop decision module is used to establish a cost function for starting and stopping the range extender. When the remaining capacity of the power battery enters the buffer threshold range, the range extender intervention timing is dynamically adjusted based on the historical start-stop frequency and current emission constraints to achieve a nonlinear balance between mechanical loss and fuel economy.

[0153] Among them, the output end of the multi-source heterogeneous data acquisition module is connected to the input end of the cross-modal load prediction module, the output end of the cross-modal load prediction module is respectively connected to the input end of the multi-objective dynamic optimization module and the input end of the rolling time domain correction module, the output end of the multi-objective dynamic optimization module is connected to the input end of the rolling time domain correction module, the output end of the rolling time domain correction module is respectively connected to the input end of the hierarchical energy routing controller and the input end of the cross-modal load prediction module, and the hierarchical energy routing controller is connected to the range extender start-stop decision module.

[0154] Based on the same inventive concept, corresponding to any of the above embodiments, the present invention provides an electronic device, including a memory and a processor, the memory being used to store a computer program, and the processor running the computer program to enable the electronic device to execute the energy optimization management method for an extended-range hybrid ship of the embodiment.

[0155] Optionally, the above-mentioned electronic device may be a server.

[0156] In addition, this embodiment further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the energy optimization management method for a range-extended hybrid ship of the embodiment is implemented.

[0157] It is understood that the processor in the embodiments of the present invention may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0158] The method steps in the embodiments of the present invention can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be an integral part of the processor. The processor and the storage medium can be located in an ASIC.

[0159] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a storage medium or transmitted via a storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

Claims

1. An energy optimization management method for a range-extended hybrid ship, characterized in that: The following steps are involved: Real-time collection of multi-source heterogeneous ship data, including operating load parameters, power battery remaining capacity matrix, range extender efficiency characteristic curve, propulsion system efficiency map, and navigation environment parameters, to construct a multi-source heterogeneous data set; Based on multi-source heterogeneous data sets, a load prediction model based on the fusion of time series convolutional networks and long-short-term memory is constructed to dynamically output the probability distribution of propulsion power demand in the future navigation cycle; A multi-objective optimization model was established, with the range extender fuel consumption rate, battery health attenuation coefficient, and equivalent carbon emission index as optimization variables, and an improved multi-objective genetic algorithm was used to generate the Pareto frontier solution set. Based on the rolling horizon optimization framework and combined with real-time navigation situation awareness data, the Pareto solution set is corrected online to generate and execute the optimal energy allocation instruction set.

2. The energy optimization management method for a range-extended hybrid ship according to claim 1, characterized in that: The specific process of constructing a load prediction model based on a time series convolutional network and long short-term memory fusion based on multi-source heterogeneous data sets and dynamically outputting the probability distribution of propulsion power demand in the future navigation cycle is as follows: Normalize the propeller torque-speed time series in the operating load parameters, the charge-discharge current-temperature correlation tensor in the power battery remaining capacity state matrix, and the river depth-flow velocity three-dimensional point cloud data in the navigation environment parameters. Use a sliding window mechanism to uniformly map the different-frequency sampling data to the same time base. A load prediction model consisting of parallel branches was built. The first branch used a stack of dilated causal convolutions to extract high-frequency fluctuations in the propulsion system efficiency map. The second branch used a bidirectional gated recurrent unit to capture long-range dependencies in the range extender efficiency curve. The third branch processed the spatial topology of the navigation environment parameters using a three-dimensional sparse convolutional network. The feature vectors output by each branch are subjected to multi-scale feature fusion to generate cross-modal fusion features. The contribution ratio of the cross-modal fusion features to the features of each branch is dynamically adjusted through a learnable gating vector, and high-dimensional features are adaptively weighted and spliced. A Monte Carlo Dropout mechanism is introduced after the fully connected layer to perform multi-step rolling forecasts on the propulsion power demand time series. Multiple forward propagation sampling is used to generate a probability density function of the power demand value, and a probability distribution cloud diagram containing the mean, variance, and confidence interval is output. Based on the propulsion motor current-voltage feedback signal obtained in real time during navigation, the sliding window statistics of the prediction error are calculated. When the mean square error continuously exceeds the dynamic threshold, the model parameters are fine-tuned. The online knowledge distillation algorithm is used to migrate the newly added data features to the pre-trained load prediction model.

3. The energy optimization management method for a range-extended hybrid ship according to claim 2, characterized in that: The specific process of performing multi-scale feature fusion on the feature vectors output by each branch to generate cross-modal fusion features is as follows: The high-frequency fluctuation feature vector of the first branch, the long-range dependency feature vector of the second branch, and the spatial topology feature vector of the third branch are linearly transformed to generate corresponding query vectors, construction vectors, and value vectors; A multi-head attention mechanism is used to split the query vector, the construction vector, and the value vector into several sub-heads, and each sub-head independently calculates the attention weight matrix; After concatenating the outputs of each sub-head, cross-modal fusion features are generated through residual connection and layer normalization.

4. The energy optimization management method for a range-extended hybrid ship according to claim 1, characterized in that: The specific process of establishing a multi-objective optimization model, taking the range extender fuel consumption rate, battery health attenuation coefficient, and equivalent carbon emission index as optimization variables, and using an improved multi-objective genetic algorithm to generate a Pareto frontier solution set is as follows: A multi-objective optimization model with time coupling constraints is constructed, and the objective function set is shown in the following formula (1): Among them, minf1 represents the minimum fuel consumption rate; k f Indicates the slope of the range extender fuel consumption curve; P gen (t) represents the output power of the range extender at time t; b f represents the fixed fuel consumption during start-stop or no-load; δ represents the start-stop indicator function of the range extender; represents the total time step of the optimization time domain; minf2 represents the minimum battery health attenuation coefficient; I bat (t) represents the charge and discharge current of the battery at time t; SOC(t) represents the remaining capacity of the power battery at time t; α represents the reference coefficient of the influence of current on aging; β represents the nonlinear influence index of current amplitude; γ represents the adjustment coefficient of the remaining capacity of the power battery to aging attenuation; minf3 represents the minimum equivalent carbon emission index; c co2 Indicates the carbon emissions per unit power output of the range extender; Indicates the rate of change of battery state of charge; E grid represents the carbon emission factor of grid electricity; λ represents the conversion coefficient of grid charging carbon emissions, which is used to balance the carbon emission contributions of different energy sources; The constraint condition set is shown in formula (2): Among them, P prop (t) represents the propulsion system power demand at time t; P bat (t) represents the power provided by the power battery at time t; P loss (t) represents the system power loss at time t; SOC min Indicates the battery's lowest state of charge; SOC max Indicates the highest state of charge of the battery; Indicates the maximum output power of the range extender; Indicates the rated maximum charge and discharge current of the battery; An improved multi-objective genetic algorithm is used to solve the problem; the dynamic crossover probability is calculated as shown in the following formula (3): Among them, p c represents the crossover probability after dynamic adjustment; p c0 represents the initial crossover probability; g represents the current number of iterations; G max represents the maximum number of iterations; Δp c Indicates the crossover probability adjustment amplitude; d i represents the average Euclidean distance from the i-th individual to other individuals in the population; d max represents the maximum individual distance in the population; N represents the total number of individuals in each generation; The adaptive mutation intensity is calculated as shown in the following formula (4): Among them, σ m represents the adaptive mutation intensity; σ m0 represents the initial mutation intensity; f dom (x) represents the dominance strength of individual x; max(f dom ) represents the maximum dominance strength value of all individuals in the current population; Fast non-dominated sorting based on the Kriging surrogate model, where the individual dominance strength is calculated as follows: Among them, x represents the current individual; y represents the population P op Other individuals in P op represents the current population; l represents a constant, which is used to control the influence of distance; Through the elite retention strategy and crowding distance screening, a Pareto optimal solution set that satisfies the following formula (6) is generated: Among them, x * represents the individual in the Pareto optimal solution set; X pareto represents the Pareto optimal solution set; and there is at least one i such that f i (y)≤f i (x * ).

5. The energy optimization management method for a range-extended hybrid ship according to claim 4, characterized in that: The specific process of performing online correction on the Pareto solution set based on the rolling horizon optimization framework and combining it with real-time navigation situation awareness data to generate and execute the optimal energy allocation instruction set is as follows: A rolling time window is established to divide the optimization time domain into a fixed-length prediction interval and an execution interval. Within each control cycle, the embedded edge computing unit obtains the latest ship attitude angle, the instantaneous gradient of the remaining battery capacity, and the heat map of the channel obstacle distribution to build a real-time situation awareness vector. Based on the real-time situational awareness vector, an adaptive Kalman filter algorithm is used to compensate for the error in the output of the load prediction model and generate a corrected probability distribution of propulsion power demand. The corrected probability distribution is input into the multi-objective optimization model. Based on the deviation between the measured value and the predicted value of the current battery health attenuation coefficient, the time-varying weight coefficients of β and γ in the objective function are dynamically adjusted to generate an updated Pareto solution set. The system evaluates the satisfaction of each solution with respect to fuel consumption rate, battery health, and carbon emissions based on a fuzzy membership function. Poor solutions with membership below a dynamic threshold are eliminated. A cosine similarity matching algorithm is then used to select the candidate solution that is closest to the historical optimal energy allocation pattern from the remaining solutions. The energy allocation instructions corresponding to the selected candidate solution are decomposed into the range extender power setting value, battery charge and discharge rate, and propulsion motor torque distribution ratio, and are sent to each actuator via the CAN bus protocol; At the end of the execution interval, the deviation matrix between the actual energy consumption data and the predicted value is collected, and the back-propagation reinforcement learning algorithm is used to update the weight parameters of the rolling horizon optimization framework.

6. The energy optimization management method for a range-extended hybrid ship according to claim 1, characterized in that: Implement hierarchical energy routing control: During the propulsion power mutation stage, the power battery pack is called upon first for power compensation, and during the steady-state cruising stage, it switches to the range extender's high-efficiency range constant power output mode, and realizes adaptive feedback storage of regenerative braking energy and surplus energy of the range extender through a bidirectional DC / DC converter.

7. The energy optimization management method for a range-extended hybrid ship according to claim 1, characterized in that: A cost function for starting and stopping the range extender is established. When the remaining capacity of the power battery enters the buffer threshold interval, the timing of the range extender intervention is dynamically adjusted according to the historical start-stop frequency and current emission constraints, so as to achieve a nonlinear balance between mechanical loss and fuel economy.

8. An energy optimization management system for a range-extended hybrid ship, used to implement the energy optimization management method for a range-extended hybrid ship according to any one of claims 1 to 7, characterized in that: include: The multi-source heterogeneous data acquisition module is used to collect multi-source heterogeneous data of ships in real time, including operating load parameters, power battery remaining capacity state matrix, range extender efficiency characteristic curve, propulsion system efficiency map and navigation environment parameters, and construct a multi-source heterogeneous data set; A cross-modal load prediction module is used to build a load prediction model based on a fusion of a temporal convolutional network and long-short-term memory based on multi-source heterogeneous data sets, dynamically outputting the probability distribution of propulsion power demand within the future navigation cycle; A multi-objective dynamic optimization module is used to establish a multi-objective dynamic programming function, using the range extender fuel consumption rate, battery health attenuation coefficient, and equivalent carbon emission index as optimization variables, and using an improved multi-objective genetic algorithm to generate a Pareto frontier solution set; The rolling horizon correction module is used to perform online correction on the Pareto solution set based on the rolling horizon optimization framework and combined with real-time navigation situation awareness data to generate the optimal energy allocation instruction set; A hierarchical energy routing controller is used to implement hierarchical energy routing control. During the propulsion power mutation phase, the power battery pack is prioritized for power compensation. During the steady-state cruise phase, the power output mode is switched to the range extender's high-efficiency range constant power output mode. A bidirectional DC / DC converter is used to adaptively store regenerative braking energy and excess energy from the range extender. The range extender start-stop decision module is used to establish a cost function for starting and stopping the range extender. When the remaining capacity of the power battery enters the buffer threshold range, the range extender intervention timing is dynamically adjusted based on the historical start-stop frequency and current emission constraints to achieve a nonlinear balance between mechanical loss and fuel economy. Among them, the output end of the multi-source heterogeneous data acquisition module is connected to the input end of the cross-modal load prediction module, the output end of the cross-modal load prediction module is respectively connected to the input end of the multi-objective dynamic optimization module and the input end of the rolling time domain correction module, the output end of the multi-objective dynamic optimization module is connected to the input end of the rolling time domain correction module, the output end of the rolling time domain correction module is respectively connected to the input end of the hierarchical energy routing controller and the input end of the cross-modal load prediction module, and the hierarchical energy routing controller is connected to the range extender start-stop decision module.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the energy optimization management method for an extended-range hybrid ship according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the energy optimization management method for a range-extended hybrid ship according to any one of claims 1 to 7 is implemented.

Images