A method and system for controlling the conversion of main and auxiliary power of a main and auxiliary power hybrid ship

Abstract

The present invention relates to the field of ship propulsion technology, specifically a method and system for controlling the conversion of main and auxiliary power for a main and auxiliary hybrid ship. The method comprises the following steps: predicting the ship's future power demand based on multi-source information during the ship's voyage; establishing a dynamic power optimization model based on the predicted power demand and adjusting the optimal power conversion scheme in real time; the dynamic power optimization model can continuously optimize the control strategy based on historical operating data; and accurately predicting power demand using a deep learning algorithm or a machine learning algorithm (particularly a long short-term memory network (LSTM)) by collecting multi-source information during the ship's voyage, including navigation status, weather and sea conditions, and real-time performance parameters of the main and auxiliary engines. This helps improve the operating efficiency of the power system and reduce unnecessary energy consumption.

Description

Technical Field

[0001] The present invention relates to the technical field of ship propulsion, and in particular to a main-auxiliary power conversion control method and system for a main-auxiliary power hybrid ship. Background Art

[0002] As a primary means of water transportation, ships are increasingly valued for their energy efficiency and environmental performance. Especially in the current context of global climate change and heightened awareness of environmental protection, improving ship energy efficiency and reducing emissions have become key development directions for the shipping industry. Hybrid ships, as a new type of vessel, combine the main engine and auxiliary power sources to optimize power under varying sailing conditions, thereby improving energy efficiency and environmental performance.

[0003] Power conversion control for hybrid vessels with main and auxiliary propulsion is typically based on fixed control strategies or simple rules, making it difficult to adapt to complex and changing navigation conditions. For example, traditional control methods may adjust the power system output based solely on the ship's speed or load, while ignoring the impact of external factors such as weather and sea conditions on power demand. Furthermore, existing control systems often lack real-time capabilities and intelligence, making them unable to accurately predict and adapt to changes in the ship's power demand under varying conditions.

[0004] The problems existing in the prior art are summarized as follows:

[0005] Existing technologies are usually unable to accurately predict the power requirements of ships under different sailing conditions, resulting in inaccurate control of the power system and the inability to fully utilize energy efficiency and environmental performance.

[0006] Traditional control methods lack intelligence and adaptability, and are unable to dynamically adjust control strategies based on real-time data to adapt to complex and changing navigation conditions.

[0007] The fault detection and emergency response mechanisms in existing technologies are usually not timely enough to provide effective early warning and intervention before potential faults occur, increasing the risk of ship operation. Summary of the Invention

[0008] In view of the technical problems existing in the prior art, the present invention provides a main-auxiliary power conversion control method and system for a main-auxiliary power hybrid ship to solve the problem that the existing control strategies or simple rules are difficult to adapt to complex and changeable navigation conditions.

[0009] The present invention solves the above technical problems with the following technical solutions: A method for controlling the conversion of main and auxiliary power of a main and auxiliary power hybrid ship comprises the following steps:

[0010] Based on multi-source information during the ship's voyage, the ship's power demand in the future is predicted;

[0011] Based on the predicted power demand, a dynamic power optimization model is established to adjust the optimal power conversion plan in real time;

[0012] The power dynamic optimization model can continuously optimize the control strategy based on historical operating data;

[0013] Real-time monitoring of the main and auxiliary power operating status, and use of big data analysis technology to identify and analyze abnormal signals to detect potential faults;

[0014] When the detected abnormal signal indicates a potential fault, the preset emergency response plan is automatically triggered;

[0015] The power conversion status, power demand prediction results and optimized power conversion plan are transmitted to the terminal display device, allowing remote monitoring and control.

[0016] On the basis of the above technical solution, the present invention can also be improved as follows.

[0017] Furthermore, the multi-source information includes ship navigation status information, weather and sea condition information, and real-time performance parameters of main and auxiliary engines.

[0018] Furthermore, deep learning algorithms or machine learning algorithms are used to combine and process ship navigation status information, meteorological and sea condition information, and real-time performance parameters of main and auxiliary engines from multiple data sources to predict the power demand of the ship within a predetermined time period in the future.

[0019] Furthermore, the deep learning algorithm or machine algorithm is a long short-term memory network (LSTM), and the prediction step includes:

[0020] Collect multi-source information during the ship's voyage;

[0021] Clean the collected multi-source information data to remove outliers and missing values;

[0022] Normalize the data to improve the convergence speed and prediction accuracy of the algorithm;

[0023] Utilizing statistical analysis and data mining techniques, characteristic variables related to ship power requirements are extracted from pre-processed data;

[0024] Constructing a long short-term memory (LSTM) model, wherein the model consists of an input layer, a hidden layer, and an output layer;

[0025] The input layer receives the preprocessed feature variables as input.

[0026] The hidden layer contains multiple LSTM units to capture long-term dependencies in time series data.

[0027] The output layer outputs the predicted ship power demand value;

[0028] The LSTM model is trained using historical navigation data, and the model parameters are adjusted through the back-propagation algorithm to minimize the prediction error.

[0029] Input multi-source information collected in real time into the trained LSTM model to perform real-time prediction of power demand;

[0030] According to the prediction results, the operating status of the main and auxiliary power systems is adjusted to meet the power requirements of the ship. Furthermore, the LSTM unit state update formula includes:

[0031] f t =σ(W f ·[h t-1 , x t ]+b f )

[0032] i t =σ(W i ·[h t-1 , x t ]+b i )

[0033]

[0034] o t =σ(W o ·[h t-1 , x t ]+b o )

[0035] h t =o t *tanh(C t );

[0036] Among them, f t 、i t 、o t They are the activation function outputs of the forget gate, input gate, and output gate respectively;

[0037] C t 、 are the current and candidate unit states respectively;

[0038] h t is the hidden layer output at the current moment;

[0039] W f 、W i 、W c 、W o are the weight matrices of the forget gate, input gate, candidate unit state, and output gate respectively;

[0040] b f 、b i 、b c 、b o are the corresponding bias terms respectively;

[0041] σ is the Sigmoid activation function; tanh is the hyperbolic tangent activation function;

[0042] x t is the input feature vector at the current moment;

[0043] h t-1 is the hidden layer output of the previous moment;

[0044] C t-1 is the unit state at the previous moment;

[0045] W and b are the weight and bias parameters learned during model training.

[0046] Furthermore, the power dynamic optimization model adopts a hybrid model that combines a multi-objective optimization algorithm with a machine learning prediction model. The steps of real-time adjustment and determination of the optimal power conversion solution include:

[0047] Collect real-time data during the ship's voyage;

[0048] Preprocess the collected data;

[0049] Define the optimization objective function, including minimizing energy consumption cost, minimizing total emissions, and maximizing system stability;

[0050] A multi-objective optimization model is constructed using a non-dominated sorting genetic algorithm (NSGA-II) multi-objective optimization algorithm. The multi-objective optimization model uses power combination as a decision variable and energy consumption cost, total emissions, and system stability as optimization objectives.

[0051] Leveraging historical navigation data, we build a machine learning prediction model to predict energy consumption, emissions, and system stability under different powertrain combinations.

[0052] Input real-time data into the multi-objective optimization model to generate a series of feasible power combination solutions;

[0053] Using machine learning prediction models, the generated solutions are predicted in terms of energy consumption, emissions, and system stability.

[0054] Based on the prediction results, the Pareto frontier analysis method is used to select the power conversion solution with the lowest energy consumption, the lowest emissions and the best system stability;

[0055] Collect new navigation data in real time and use it to update machine learning prediction models and multi-objective optimization models;

[0056] Furthermore, the multi-objective optimization objective function is:

[0057] minF(x)={f1(x), f2(x), f3(x)};

[0058] Among them, x is the decision variable of the power combination scheme; f1(x) is the energy cost function, which represents the energy cost under the power combination scheme; f2(x) is the total emission function, which represents the total emission under the power combination scheme; f3(x) is the system stability function, which represents the system stability under the power combination scheme;

[0059] The machine learning prediction model is a neural network, and the neural network formula is:

[0060] y=f(W,b,x);

[0061] Among them, y is the prediction result;

[0062] f is the neural network model;

[0063] W and b are the weight and bias of the neural network respectively;

[0064] x is the input feature.

[0065] Furthermore, the triggering of the emergency response plan includes at least one of switching to a backup power source and adjusting the navigation plan.

[0066] Furthermore, it also includes a user-friendly human-computer interaction interface for displaying the power conversion status, the predicted results of power demand and the optimized power conversion plan, and allowing users to remotely monitor and control through the interface.

[0067] Furthermore, the human-computer interaction interface also includes a fault warning prompt function, which reminds the user visually and / or auditorily when a potential fault is detected.

[0068] The present invention also provides a main-auxiliary power conversion control system for a main-auxiliary power hybrid ship, comprising:

[0069] The power demand prediction module is used to predict the power demand of the ship in the future based on multi-source information during the ship's navigation process;

[0070] A power dynamics optimization module is used to establish a power dynamics optimization model based on predicted power demand, and to adjust and determine the optimal power conversion solution in real time. The power dynamics optimization model can continuously optimize the control strategy based on historical operating data;

[0071] The status monitoring and fault detection module is used to monitor the operating status of the main and auxiliary power systems in real time, and use big data analysis technology to identify and analyze abnormal signals to detect potential faults;

[0072] Emergency response module, which automatically triggers the preset emergency response plan when the detected abnormal signal indicates a potential fault;

[0073] Human-machine interface module, used to display power conversion status, power demand forecast results and optimized power conversion plan, and allows remote monitoring and control

[0074] Furthermore, the multi-source information includes ship navigation status information, weather and sea condition information, and real-time performance parameters of main and auxiliary engines;

[0075] The power demand prediction module uses a deep learning algorithm or a machine learning algorithm to combine and process the multi-source information from multiple data sources to predict the power demand of the ship within a predetermined future time period;

[0076] The power demand forecasting module uses the long short-term memory network (LSTM) algorithm to predict power demand, including a data collection and preprocessing unit, a feature extraction unit, an LSTM model construction and training unit, and a real-time prediction and adjustment unit;

[0077] The power dynamic optimization module adopts a hybrid model that combines a multi-objective optimization algorithm with a machine learning prediction model to adjust and determine the optimal power conversion plan in real time;.

[0078] The power dynamic optimization module includes a data collection and preprocessing unit, a multi-objective optimization model construction unit, a machine learning prediction model construction unit, an optimization scheme generation and selection unit, and a model update and optimization unit.

[0079] The beneficial effects of the present invention are:

[0080] 1. By collecting multi-source information during the ship's voyage, including navigation status, weather and sea conditions, and real-time performance parameters of main and auxiliary engines, deep learning or machine learning algorithms (particularly long short-term memory networks (LSTMs)) are used to accurately predict power demand. This helps improve the operating efficiency of the power system and reduce unnecessary energy consumption.

[0081] 2. Based on the predicted power demand, a dynamic power optimization model is established. A hybrid model combining a multi-objective optimization algorithm (such as NSGA-II) with machine learning prediction models (such as SVM, RF, and NN) is used to adjust and determine the optimal power conversion plan in real time. This hybrid model comprehensively considers multiple objectives, including energy consumption costs, total emissions, and system stability, to achieve global optimization of the power system.

[0082] 3. Real-time monitoring of the operating status of the main and auxiliary power sources, utilizing big data analytics to identify and analyze abnormal signals and detect potential faults. Upon detecting an anomaly, the system automatically triggers pre-defined emergency response plans, such as switching to a backup power source or adjusting the navigation plan, to ensure the safe operation of the vessel. Furthermore, the fault warning function within the human-machine interface promptly alerts users to potential faults, enhancing system safety and reliability.

[0083] 4. By accurately predicting power demand and optimizing power conversion plans in real time, the system can rationally allocate power resources based on the ship's actual needs, avoiding unnecessary energy consumption and emissions. This helps reduce ship operating costs while minimizing environmental impact and achieving green sailing.

[0084] 5. Provides intuitive information display and remote control functions. Users can view the power conversion status, power demand forecast results, and optimized power conversion plan through the human-computer interaction interface, and perform necessary remote monitoring and control. This design improves the system's intelligence level and user experience, and enhances the interactivity and convenience between users and the system.

[0085] 6. Provides intuitive information display and remote control functions. Users can view the power conversion status, power demand forecast results, and optimized power conversion plans through the human-computer interaction interface, and perform necessary remote monitoring and control. This design improves the system's intelligence level and user experience, and enhances the interactivity and convenience between users and the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 It is a schematic diagram of the arrangement process of the present invention;

[0087] Figure 2 A flow chart showing power requirements within a predetermined time period of the present invention;

[0088] Figure 3 This is a schematic diagram of the principle of the power dynamic optimization model of the present invention. DETAILED DESCRIPTION

[0089] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0090] like Figure 1-3 The present invention relates to a method for controlling the conversion of main and auxiliary power of a main and auxiliary power hybrid ship, comprising the following steps:

[0091] Based on multi-source information during the ship's voyage, the ship's power demand in the future is predicted;

[0092] Based on the predicted power demand, a dynamic power optimization model is established to adjust the optimal power conversion plan in real time;

[0093] The power dynamic optimization model can continuously optimize the control strategy based on historical operating data;

[0094] Real-time monitoring of the main and auxiliary power operating status, and use of big data analysis technology to identify and analyze abnormal signals to detect potential faults;

[0095] When the detected abnormal signal indicates a potential fault, the preset emergency response plan is automatically triggered;

[0096] The power conversion status, power demand prediction results and optimized power conversion plan are transmitted to the terminal display device, allowing remote monitoring and control.

[0097] This method predicts future power requirements based on multi-source information during a vessel's voyage. It establishes a dynamic power optimization model to adjust the optimal power conversion plan in real time, and continuously optimizes the control strategy based on historical data. Simultaneously, it monitors the operating status of the main and auxiliary power sources in real time, uses big data analysis to detect potential faults, and automatically triggers emergency response plans. All power conversion status, predictions, and optimization plans are displayed on a terminal, enabling remote monitoring and control.

[0098] As an embodiment, the multi-source information includes ship navigation status information (specifically, including but not limited to the ship's speed, heading, and load conditions), meteorological and sea condition information (specifically, including but not limited to wind speed, wave height, and water flow speed), and real-time performance parameters of the main and auxiliary engines.

[0099] As a supplement, deep learning algorithms or machine learning algorithms are used to combine and process ship navigation status information, meteorological and sea condition information, and real-time performance parameters of main and auxiliary engines from multiple data sources to predict the power demand of the ship within a predetermined time period in the future.

[0100] Specifically, the deep learning algorithm or machine algorithm is a long short-term memory network (LSTM), and the prediction step includes:

[0101] Collect multi-source information during the ship's voyage (including but not limited to the ship's speed, heading, load, wind speed, wave height, water flow speed, and real-time performance parameters of main and auxiliary engines, etc.);

[0102] Clean the collected multi-source information data to remove outliers and missing values; ensure the accuracy and completeness of the data;

[0103] Normalize the data to improve the convergence speed and prediction accuracy of the algorithm;

[0104] Utilize statistical analysis and data mining techniques to extract characteristic variables related to ship power requirements from pre-processed data (specifically, these characteristic variables include ship acceleration, deceleration, turning rate, wind and wave resistance coefficient, etc.)

[0105] Constructing a long short-term memory (LSTM) model, wherein the model consists of an input layer, a hidden layer, and an output layer;

[0106] The input layer receives the preprocessed feature variables as input.

[0107] The hidden layer contains multiple LSTM units to capture long-term dependencies in time series data.

[0108] The output layer outputs the predicted ship power demand value;

[0109] The LSTM model is trained using historical navigation data, and the model parameters are adjusted through the back-propagation algorithm to minimize the prediction error.

[0110] Specifically, during the training process, methods such as cross-validation can be used to evaluate the generalization ability of the model and prevent overfitting;

[0111] Input the multi-source information collected in real time into the trained LSTM model to make real-time predictions of power demand;

[0112] According to the prediction results, the operating status of the main and auxiliary power systems is adjusted to meet the power requirements of the ship. As a supplement, the LSTM unit state update formula includes:

[0113] f t =σ(W f ·[h t-1 , x t ]+b f )

[0114] i t =σ(W i ·[h t-1 , x t ]+b i )

[0115]

[0116] o t =σ(W o ·[h t-1 , x t ]+b o )

[0117] h t =o t *tanh(C t );

[0118] Among them, f t 、i t 、o t They are the activation function outputs of the forget gate, input gate, and output gate respectively;

[0119] C t 、 are the current and candidate unit states respectively;

[0120] h t is the hidden layer output at the current moment;

[0121] W f 、W i 、W c 、W o are the weight matrices of the forget gate, input gate, candidate unit state, and output gate respectively;

[0122] b f 、b i 、b c 、b o are the corresponding bias terms respectively;

[0123] σ is the Sigmoid activation function; tanh is the hyperbolic tangent activation function;

[0124] x t is the input feature vector at the current moment;

[0125] h t-1 is the hidden layer output of the previous moment;

[0126] C t-1 is the unit state at the previous moment;

[0127] W and b are the weight and bias parameters learned during model training.

[0128] The present invention also provides a specific embodiment to illustrate how to use the long short-term memory network (LSTM) algorithm to predict the power demand of a ship under different navigation conditions.

[0129] Example 1:

[0130] This embodiment provides a method for predicting the power demand of a hybrid ship with main and auxiliary propulsion, based on a long short-term memory (LSTM) algorithm. This method collects multi-source information during a ship's voyage and uses the LSTM algorithm for training and prediction to accurately predict the ship's power demand under different navigation conditions.

[0131] Step 1: Data collection and preprocessing

[0132] First, we collected multi-source data on a specific type of hybrid vessel operating over a period of time, including but not limited to the vessel's speed, heading, load, wind speed, wave height, current velocity, and real-time performance parameters of the main and auxiliary engines. This data was derived from the vessel's automatic monitoring system, ensuring its real-time and accuracy.

[0133] Next, we cleaned the collected data, removing outliers and missing values ​​to ensure data integrity and reliability. We then normalized the data, converting data of different dimensions to the same scale to improve the algorithm's convergence speed and prediction accuracy.

[0134] Step 2: Feature extraction

[0135] Based on data preprocessing, we use statistical analysis and data mining techniques to extract characteristic variables related to ship power requirements from the preprocessed data. These characteristic variables include the ship's acceleration, deceleration, turning rate, wind and wave resistance coefficient, etc., which can reflect the ship's dynamic characteristics and power requirements under different navigation conditions.

[0136] Step 3: Model building and training

[0137] We constructed a long short-term memory (LSTM) model consisting of an input layer, a hidden layer, and an output layer. The input layer receives preprocessed feature variables as input, the hidden layer contains multiple LSTM units to capture long-term dependencies in time series data, and the output layer outputs the predicted ship power demand value.

[0138] We then trained the LSTM model using historical navigation data. During training, we employed methods such as cross-validation to evaluate the model's generalization capabilities and prevent overfitting. We used backpropagation to adjust model parameters to minimize prediction error. After multiple iterations of training, we developed a well-performing LSTM prediction model.

[0139] Step 4: Real-time prediction and adjustment

[0140] During a vessel's voyage, we collect multi-source information in real time and input it into a trained LSTM model to predict power demand in real time. The prediction results reflect the changes in power demand under different sailing conditions, providing an important basis for controlling the vessel's power system.

[0141] Based on the prediction results, we promptly adjust the operating status of the main and auxiliary power systems to meet the ship's power needs. When we predict an increase in power demand, we increase the output of the main engine or activate the auxiliary power source. When we predict a decrease in power demand, we reduce the output of the main engine or shut down the auxiliary power source. This allows us to optimize control of the ship's power system, improve navigation efficiency, and reduce energy consumption and emissions.

[0142] Implementation Effect

[0143] Using the method in this example, we successfully implemented power demand forecasting for hybrid ships with main and auxiliary propulsion systems based on a long short-term memory (LSTM) algorithm. The prediction results are accurate and reliable, providing strong support for ship power system control. Furthermore, this method offers advantages such as strong adaptability and high prediction accuracy, making it widely applicable to power demand forecasting for various types of ships.

[0144] As an embodiment, the power dynamic optimization model adopts a hybrid model that combines a multi-objective optimization algorithm with a machine learning prediction model. The steps of real-time adjustment and determination of the optimal power conversion solution include:

[0145] Collect real-time data during the ship's voyage (including but not limited to the ship's speed, heading, load, wind and wave conditions, engine speed, fuel consumption rate, emission concentration, etc.).

[0146] Preprocess the collected data (specifically, including data cleaning, outlier processing, data normalization, etc. to ensure data accuracy and consistency.)

[0147] Define the optimization objective function (specifically, minimizing energy consumption costs, minimizing total emissions, and maximizing system stability);

[0148] A multi-objective optimization model was constructed using the non-dominated sorting genetic algorithm (NSGA-II) multi-objective optimization algorithm. The multi-objective optimization model used the power combination (specifically, the main engine output power, the auxiliary power source status, etc.) as the decision variable, and energy consumption cost, total emissions, and system stability as the optimization objectives.

[0149] Leveraging historical navigation data, we build a machine learning prediction model to predict energy consumption, emissions, and system stability under different powertrain combinations.

[0150] Specifically, machine learning models include support vector machines (SVM), random forests (RF), neural networks (NN), etc. By training the model, it can accurately predict energy consumption, emissions and system stability in the future.

[0151] Input real-time data into the multi-objective optimization model to generate a series of feasible power combination solutions;

[0152] Using machine learning prediction models, the generated solutions are predicted in terms of energy consumption, emissions, and system stability.

[0153] Based on the prediction results, the Pareto frontier analysis method is used to select the power conversion solution with the lowest energy consumption, the lowest emissions and the best system stability;

[0154] Collect new navigation data in real time and use it to update machine learning prediction models and multi-objective optimization models;

[0155] Furthermore, through continuous learning and optimization, the prediction accuracy and optimization performance of the model can be improved.

[0156] As a supplement, the multi-objective optimization objective function is:

[0157] minF(x)={f1(x), f2(x), f3(x)};

[0158] Among them, x is the decision variable of the power combination scheme; f1(x) is the energy cost function, which represents the energy cost under the power combination scheme; f2(x) is the total emission function, which represents the total emission under the power combination scheme; f3(x) is the system stability function, which represents the system stability under the power combination scheme;

[0159] The machine learning prediction model is a neural network, and the neural network formula is:

[0160] y=f(W,b,x);

[0161] Where y is the prediction result (energy consumption, emissions, system stability, etc.);

[0162] f is the neural network model;

[0163] W and b are the weight and bias of the neural network respectively;

[0164] x is the input feature (real-time data).

[0165] The present invention is used to illustrate how to use a hybrid model that combines a multi-objective optimization algorithm with a machine learning prediction model in actual operation to achieve optimal power conversion control for a main-auxiliary hybrid ship. It also provides the following optimized implementation examples of power conversion control for a main-auxiliary hybrid ship.

[0166] Example 2:

[0167] This embodiment focuses on a specific main-auxiliary hybrid ship, and aims to determine and adjust its optimal power conversion scheme in real time by adopting a hybrid model that combines a multi-objective optimization algorithm with a machine learning prediction model.

[0168] 1. Data collection and preprocessing stage

[0169] Data Collection: We collected comprehensive real-time data on the vessel during its voyage, covering key indicators such as speed, heading, load status, wind and wave conditions, engine speed, fuel consumption rate, and emission concentration. This data was derived from the vessel's automatic monitoring system, ensuring its real-time and accuracy.

[0170] Data preprocessing: After collecting the raw data, we cleaned it to remove outliers and missing data to ensure the completeness and accuracy of the dataset. We then normalized the data to eliminate discrepancies between data of different dimensions, thereby improving the processing efficiency and prediction accuracy of subsequent algorithms.

[0171] 2. Multi-objective optimization model construction stage

[0172] Defining optimization objectives: We identified three core optimization objectives: minimizing energy costs, minimizing total emissions, and maximizing system stability. These objective functions will serve as the foundation for subsequent multi-objective optimization algorithms.

[0173] Building a Multi-Objective Optimization Model: We constructed a multi-objective optimization model based on the non-dominated sorting genetic algorithm (NSGA-II). This model uses the powertrain combination (including main engine output and auxiliary power source status) as the decision variable and aims to find the optimal balance between energy consumption costs, total emissions, and system stability. Through genetic operations (such as selection, crossover, and mutation) and a non-dominated sorting mechanism, we generated and screened a series of excellent powertrain combination solutions.

[0174] 3. Machine Learning Prediction Model Construction Phase

[0175] Data preparation: We use historical navigation data as a training set. After preprocessing, this data is used to build a machine learning prediction model.

[0176] Model Selection and Training: We selected a neural network (NN) as the prediction model due to its excellent nonlinear fitting capabilities and generalization performance. Through training, we enabled the model to accurately predict energy consumption, emissions, and system stability under different powertrain combinations. During training, we employed methods such as cross-validation to evaluate model performance and optimized parameters to improve prediction accuracy.

[0177] 4. Optimization solution generation and selection stage

[0178] Generating power mix solutions: We input real-time data into a multi-objective optimization model to generate a series of feasible power mix solutions that achieve different trade-offs between energy costs, total emissions, and system stability.

[0179] Prediction and Evaluation: Using the trained machine learning prediction model, we predict the energy consumption, emissions, and system stability of the generated solutions. The prediction results provide us with an intuitive understanding of the performance of each solution.

[0180] Selecting the optimal solution: Based on the forecast results, we employed methods such as Pareto frontier analysis to select the power conversion solution with the lowest energy consumption, lowest emissions, and best system stability. This solution was determined to be the optimal choice under current sailing conditions.

[0181] 5. Model update and optimization stage

[0182] As the vessel continues to sail, we continuously collect new navigation data and use it to update our machine learning prediction model and multi-objective optimization model. Through continuous learning and optimization, we improve the model's prediction accuracy and optimization performance, ensuring the continued effectiveness and adaptability of the power conversion control solution.

[0183] As an embodiment, the triggering of the emergency response plan includes at least one of switching to a backup power source and adjusting the navigation plan.

[0184] When a potential fault is detected, the system automatically takes measures such as switching to a backup power source or adjusting the navigation plan to ensure the safe operation of the ship and avoid adverse effects of the fault on navigation.

[0185] As an embodiment, it also includes a user-friendly human-computer interaction interface for displaying the power conversion status, the predicted results of power demand and the optimized power conversion plan, and allowing users to remotely monitor and control through the interface.

[0186] The human-computer interaction interface also includes a fault early warning prompt function, which reminds the user visually and / or auditorily when a potential fault is detected.

[0187] The interface can clearly display the power conversion status, power demand forecast results and optimized power conversion plan, providing users with intuitive information display.

[0188] More importantly, the interface also features a fault warning function. Once the system detects a potential fault, it immediately alerts the user through visual (such as a flashing indicator light, a red warning message) and / or auditory (such as an alarm) means, ensuring that the user can react quickly and take the necessary measures to ensure the safe operation of the ship. In this way, the method not only improves the intelligence level of the ship's power system, but also enhances the interactivity and safety between users and the system.

[0189] Based on the above main-auxiliary power conversion control method of the main-auxiliary power hybrid ship, the present invention also proposes a main-auxiliary power conversion control system of the main-auxiliary power hybrid ship.

[0190] The present invention relates to a main-auxiliary power conversion control system for a main-auxiliary power hybrid ship, comprising:

[0191] The power demand prediction module is used to predict the power demand of the ship in the future based on multi-source information during the ship's navigation process;

[0192] A power dynamics optimization module is used to establish a power dynamics optimization model based on predicted power demand, and to adjust and determine the optimal power conversion solution in real time. The power dynamics optimization model can continuously optimize the control strategy based on historical operating data;

[0193] The status monitoring and fault detection module is used to monitor the operating status of the main and auxiliary power systems in real time, and use big data analysis technology to identify and analyze abnormal signals to detect potential faults;

[0194] Emergency response module, which automatically triggers the preset emergency response plan when the detected abnormal signal indicates a potential fault;

[0195] The human-machine interaction interface module is used to display the power conversion status, the predicted results of power demand and the optimized power conversion plan, and allows remote monitoring and control.

[0196] The system integrates multiple key modules to achieve intelligent management and optimization of the ship's power system. First, the power demand forecasting module leverages multi-source information during the ship's voyage to accurately predict future power demand. Next, the power dynamic optimization module builds a dynamic optimization model based on these predictions and makes real-time adjustments to determine the optimal power conversion solution. This model also continuously optimizes control strategies based on historical operating data, ensuring the power system remains in optimal operation.

[0197] The status monitoring and fault detection module also monitors the operating status of the main and auxiliary power systems in real time, using big data analytics to identify and analyze abnormal signals to promptly identify potential faults. If an anomaly is detected, the emergency response module automatically triggers pre-defined emergency response plans, such as switching to a backup power source or adjusting the navigation plan, to ensure the safe operation of the vessel.

[0198] Finally, the human-machine interface module provides users with intuitive information display and remote control capabilities. Through this interface, users can view power conversion status, power demand forecasts, and optimized power conversion plans, and perform necessary remote monitoring and control. This design not only enhances the system's intelligence but also enhances user interactivity and convenience.

[0199] As an embodiment, the multi-source information includes ship navigation status information (specifically, including but not limited to the ship's speed, heading, and load condition), meteorological and sea condition information (specifically, including but not limited to wind speed, wave height, and water flow speed), and real-time performance parameters of the main and auxiliary engines; as well as real-time performance parameters of the main and auxiliary engines, so as to fully understand the current status of the ship and the external environment.

[0200] The power demand prediction module uses a deep learning algorithm or a machine learning algorithm to combine and process the multi-source information from multiple data sources to predict the power demand of the ship within a predetermined future time period;

[0201] The power demand forecasting module uses the long short-term memory network (LSTM) algorithm to predict power demand, including a data collection and preprocessing unit, a feature extraction unit, an LSTM model construction and training unit, and a real-time prediction and adjustment unit;

[0202] To predict the ship's power requirements within a predetermined timeframe, the system employs deep learning or machine learning algorithms, specifically the Long Short-Term Memory (LSTM) algorithm. The LSTM algorithm collects and preprocesses data through a data collection and preprocessing unit, extracts key features through a feature extraction unit, builds and trains an LSTM model, and ultimately implements real-time prediction and adjustment. This process accurately predicts a ship's future power requirements, providing a foundation for power system optimization.

[0203] The power dynamic optimization module adopts a hybrid model that combines a multi-objective optimization algorithm with a machine learning prediction model to adjust and determine the optimal power conversion plan in real time; the multi-objective optimization algorithm is a non-dominated sorting genetic algorithm (NSGA-II), and the machine learning prediction model includes support vector machine (SVM), random forest (RF), neural network (NN), etc.

[0204] The power dynamic optimization module includes a data collection and preprocessing unit, a multi-objective optimization model construction unit, a machine learning prediction model construction unit, an optimization scheme generation and selection unit, and a model update and optimization unit.

[0205] In terms of dynamic power optimization, the system adopts a hybrid model that combines a multi-objective optimization algorithm with a machine learning prediction model. Among them, the multi-objective optimization algorithm is the non-dominated sorting genetic algorithm (NSGA-II), which can find a balance between multiple objectives and generate a series of optimal solutions. Machine learning prediction models include support vector machines (SVM), random forests (RF), neural networks (NN), etc., which are used to predict the effects of different power conversion schemes. The dynamic power optimization module includes a data collection and preprocessing unit, a multi-objective optimization model construction unit, a machine learning prediction model construction unit, an optimization scheme generation and selection unit, and a model update and optimization unit. These units work together to adjust and determine the optimal power conversion scheme in real time to ensure the efficient, safe and reliable operation of the ship's power system.

[0206] As an embodiment, the emergency response plan includes at least one of switching to a backup power source and adjusting the navigation plan; the emergency response module automatically executes the emergency response plan when a potential fault is detected.

[0207] The human-computer interaction interface module also includes a fault warning prompt function, which reminds the user visually and / or auditorily when a potential fault is detected; the human-computer interaction interface module adopts a user-friendly design, which facilitates remote monitoring and control by users.

[0208] The system integrates multiple key modules to enable comprehensive monitoring, prediction, optimization, and emergency response of the ship's power system. When the system detects a potential fault through the status monitoring and fault detection module, the emergency response module automatically executes a pre-set emergency response plan, including switching to a backup power source, adjusting the navigation plan, and at least one other measure to ensure the safe operation of the ship. Simultaneously, the human-machine interface module immediately alerts the user to potential faults through visual (e.g., flashing indicator lights, red warning messages) and / or auditory (e.g., alarm sounds) means. The user-friendly design facilitates remote monitoring and operation.

[0209] As a supplement, a data communication module is also included to realize data transmission and communication between modules to ensure the overall coordination and efficient operation of the system.

[0210] The data communication module plays a crucial role in the entire system, enabling data transmission and communication between modules to ensure overall system coordination and efficient operation. Through the data communication module, the power demand prediction module, power dynamic optimization module, condition monitoring and fault detection module, and emergency response module can share information in real time and work together to achieve intelligent management and optimization of the ship's power system.

[0211] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for controlling the conversion of main and auxiliary power of a main and auxiliary power hybrid ship, characterized in that: The following steps are involved: Based on multi-source information during the ship's voyage, the ship's power demand in the future is predicted; Based on the predicted power demand, a dynamic power optimization model is established to adjust the optimal power conversion plan in real time; The power dynamic optimization model can continuously optimize the control strategy based on historical operating data; Real-time monitoring of the main and auxiliary power operating status, and use of big data analysis technology to identify and analyze abnormal signals to detect potential faults; When the detected abnormal signal indicates a potential fault, the preset emergency response plan is automatically triggered; Transmit power conversion status, power demand prediction results, and optimized power conversion plans to terminal display devices, allowing remote monitoring and control; The power dynamic optimization model adopts a hybrid model that combines a multi-objective optimization algorithm with a machine learning prediction model. The steps of real-time adjustment and determination of the optimal power conversion solution include: Collect real-time data during the ship's voyage; Preprocess the collected data; Define the optimization objective function, including minimizing energy consumption cost, minimizing total emissions, and maximizing system stability; A multi-objective optimization model is constructed using a non-dominated sorting genetic algorithm (NSGA-II) multi-objective optimization algorithm. The multi-objective optimization model uses power combination as a decision variable and energy consumption cost, total emissions, and system stability as optimization objectives. Leveraging historical navigation data, we build a machine learning prediction model to predict energy consumption, emissions, and system stability under different powertrain combinations. Input real-time data into the multi-objective optimization model to generate a series of feasible power combination solutions; Use machine learning prediction models to predict energy consumption, emissions, and system stability for the generated solutions; Based on the prediction results, the Pareto frontier analysis method is used to select the power conversion solution with the lowest energy consumption, the lowest emissions and the best system stability; New navigation data is collected in real time and used to update machine learning prediction models and multi-objective optimization models.

2. The main-auxiliary power conversion control method of a main-auxiliary power hybrid ship according to claim 1 is characterized in that: The multi-source information includes ship navigation status information, weather and sea condition information, and real-time performance parameters of main and auxiliary engines.

3. The main-auxiliary power conversion control method of a main-auxiliary power hybrid ship according to claim 2 is characterized in that: Using deep learning algorithms or machine learning algorithms, the system combines and processes ship navigation status information, weather and sea condition information, and real-time performance parameters of main and auxiliary engines from multiple data sources to predict the power demand of the ship within a predetermined time period in the future.

4. The main-auxiliary power conversion control method of a main-auxiliary power hybrid ship according to claim 3 is characterized in that: The deep learning algorithm or machine algorithm is a long short-term memory network (LSTM), and the prediction steps include: Collect multi-source information during the ship's voyage; Clean the collected multi-source information data to remove outliers and missing values; Normalize the data to improve the convergence speed and prediction accuracy of the algorithm; Utilizing statistical analysis and data mining techniques, characteristic variables related to ship power requirements are extracted from pre-processed data; Constructing a long short-term memory network (LSTM) model, wherein the model consists of an input layer, a hidden layer, and an output layer; The input layer receives the preprocessed feature variables as input; The hidden layer contains multiple LSTM units to capture long-term dependencies in time series data; The output layer outputs the predicted ship power demand value; The LSTM model is trained using historical navigation data, and the model parameters are adjusted through the back-propagation algorithm to minimize the prediction error. Input multi-source information collected in real time into the trained LSTM model to perform real-time prediction of power demand; According to the prediction results, the operating status of the main and auxiliary power systems is adjusted to meet the power requirements of the ship.

5. The main-auxiliary power conversion control method of a main-auxiliary power hybrid ship according to claim 4, characterized in that: The LSTM unit state update formula includes: ； ； ； ； ； ； in, 、 、 They are the activation function outputs of the forget gate, input gate, and output gate respectively; 、 are the current and candidate unit states respectively; is the hidden layer output at the current moment; 、 、 、 are the weight matrices of the forget gate, input gate, candidate unit state, and output gate respectively; 、 、 、 are the corresponding bias terms respectively; is the Sigmoid activation function; tanh is the hyperbolic tangent activation function; is the input feature vector at the current moment; is the hidden layer output of the previous moment; is the unit state at the previous moment; W and b are the weight and bias parameters learned during model training.

6. The main-auxiliary power conversion control method of a main-auxiliary power hybrid ship according to claim 1, characterized in that: Multi-objective optimization objective function: ; Among them, x is the decision variable of the power combination scheme; is the energy consumption cost function, which represents the energy consumption cost under the power combination scheme; is the total emission function, which represents the total emission under the power combination scheme; is the system stability function, which indicates the system stability under the power combination scheme; The machine learning prediction model is a neural network, and the neural network formula is: ; Among them, y is the prediction result; f is the neural network model; W and b are the weight and bias of the neural network respectively; x is the input feature.

7. The main-auxiliary power conversion control method of a main-auxiliary power hybrid ship according to claim 1, characterized in that: Triggering the emergency response plan includes switching to the backup power source and adjusting at least one of the navigation plans.

8. The main-auxiliary power conversion control method of a main-auxiliary power hybrid ship according to claim 1, characterized in that: It also includes a user-friendly human-computer interaction interface for displaying the power conversion status, power demand prediction results and optimized power conversion plan, and allows users to remotely monitor and control through the interface.

9. The main-auxiliary power conversion control method of a main-auxiliary power hybrid ship according to claim 8, characterized in that: The human-computer interaction interface also includes a fault early warning prompt function, which reminds the user visually and / or auditorily when a potential fault is detected.

10. A main-auxiliary power conversion control system for a main-auxiliary power hybrid ship, based on the main-auxiliary power conversion control method for a main-auxiliary power hybrid ship according to any one of claims 1 to 9, characterized in that: include: The power demand prediction module is used to predict the power demand of the ship in the future based on multi-source information during the ship's navigation process; A power dynamics optimization module is used to establish a power dynamics optimization model based on predicted power demand, and to adjust and determine the optimal power conversion solution in real time. The power dynamics optimization model can continuously optimize the control strategy based on historical operating data; The status monitoring and fault detection module is used to monitor the operating status of the main and auxiliary power systems in real time, and use big data analysis technology to identify and analyze abnormal signals to detect potential faults; Emergency response module, which automatically triggers the preset emergency response plan when the detected abnormal signal indicates a potential fault; The human-machine interaction interface module is used to display the power conversion status, the predicted results of power demand and the optimized power conversion plan, and allows remote monitoring and control.

11. The main-auxiliary power conversion control system for a main-auxiliary power hybrid ship according to claim 10, characterized in that: The multi-source information includes ship navigation status information, weather and sea condition information, and real-time performance parameters of main and auxiliary engines; The power demand prediction module uses a deep learning algorithm or a machine learning algorithm to combine and process the multi-source information from multiple data sources to predict the power demand of the ship within a predetermined future time period; The power demand forecasting module uses the long short-term memory network (LSTM) algorithm to predict power demand, including a data collection and preprocessing unit, a feature extraction unit, an LSTM model construction and training unit, and a real-time prediction and adjustment unit; The power dynamic optimization module uses a hybrid model that combines a multi-objective optimization algorithm with a machine learning prediction model to adjust and determine the optimal power conversion solution in real time; The power dynamic optimization module includes a data collection and preprocessing unit, a multi-objective optimization model construction unit, a machine learning prediction model construction unit, an optimization scheme generation and selection unit, and a model update and optimization unit.

Images