CN-122020507-A - Ship hybrid power system performance evaluation method and system with predictive dynamic weights

CN122020507ACN 122020507 ACN122020507 ACN 122020507ACN-122020507-A

Abstract

The invention belongs to the field of control of ship hybrid power systems, and particularly discloses a ship hybrid power system performance evaluation method and system with predictive dynamic weights. The method comprises the steps of constructing a multidimensional performance evaluation index system and ship perception data, constructing a digital twin model mapped with a physical ship hybrid power system to deduce a system response state in a future preset time window, generating a future state vector comprising a future battery electric quantity trend, a future equipment heat load state and a future energy consumption emission predicted value, constructing a predictive dynamic weight decision module based on LSTM and depth DDPG to make decisions, carrying out weighted summation calculation on the predictive dynamic weight vector and each index standardized value, and generating a targeted control optimization instruction according to the calculated value. The invention can provide comprehensive performance evaluation and optimization support which covers the full life cycle, has a prospective view and gives consideration to multi-objective cooperation for the ship hybrid power system.

Inventors

SUN HAIMING
AI WENXIN
LAN GUOYUAN
ZHANG ZUNHUA
LUO CHUJUN
ZHOU MENGNI
Yu Yangwanqing

Assignees

中国葛洲坝集团股份有限公司
武汉理工大学

Dates

Publication Date: 20260512
Application Date: 20251219

Claims (10)

1. The ship hybrid power system performance evaluation method with predictive dynamic weight is characterized by comprising the following steps of: Firstly, constructing a multidimensional performance evaluation index system and ship perception data, wherein the evaluation index system comprises economy, environmental protection, technical performance and reliability, the ship perception data comprises current state data and external environment data of a ship hybrid power system, and the data is subjected to standardized processing; Step two, a digital twin model mapped with a physical ship hybrid power system is constructed, the model comprises a ship motion dynamics model and an energy flow model, a future environment prediction data sequence obtained according to weather forecast and a future sailing task sequence of a ship sailing plan are used as initial boundary conditions, the current state data, the future environment prediction data sequence and the future sailing task sequence are input into the digital twin model together, so that a system response state in a future preset time window is deduced, and a future state vector comprising a future battery electric quantity trend, a future equipment thermal load state and a future energy consumption emission predicted value is generated; step three, establishing a predictive dynamic weight decision module based on LSTM and DDPG, fusing the current state data, the external environment data and the future state vector generated by the deduction in the step two, constructing an enhanced state vector, extracting the time sequence characteristics of the enhanced state vector by utilizing an LSTM network, and deciding by utilizing an Actor-Critic network architecture; And step four, carrying out weighted summation calculation on the output predictive dynamic weight vector and each index standardized value, and generating a targeted control optimization instruction according to the calculated value so as to complete closed-loop management from evaluation to optimization of the ship.
2. The method for estimating the performance of the ship hybrid power system with the predictive dynamic weight according to claim 1, wherein in the first step, the current state data comprises the power battery health state, the fuel battery performance attenuation degree, the engine real-time efficiency point and the current task stage code which are obtained by a cabin monitoring system, and the external environment data comprises the real-time wind speed, the sea wave grade and the water flow speed which are obtained by a ship weather station and a sensor, and the real-time fuel oil price, the liquefied natural gas price and the emission control area state identification of the current navigation area which are received by a satellite communication system.
3. The method for evaluating the performance of the ship hybrid power system with the predictive dynamic weight according to claim 2, wherein the real-time original data of each performance index is obtained by adopting a dynamic range normalization method to carry out dimensionless treatment, the normalization value of the benefit type index is calculated, the inverse normalization value of the cost type index is calculated, and in this way, all indexes are uniformly mapped to the numerical intervals of [0,1 ].
4. The method for estimating the performance of a ship hybrid system with predictive dynamic weights according to claim 1, wherein the second step comprises the steps of: (21) In a shore end or high-performance edge computing unit, a digital twin model corresponding to the entity ships one by one is deployed, and the model comprises: A hull resistance model which calculates real-time resistance according to the parameters of the speed, draft and stormy waves based on Holtrop-Mennen method or CFD pre-calculation database; the propulsion system model comprises a propeller open water characteristic curve, a main machine universal characteristic curve, a motor efficiency Map and a battery equivalent circuit model; the energy management strategy model is that energy management logic completely consistent with a real ship is operated in a twin body; (22) And reading the navigation plan and the wind wave flow data on the navigation path in the electronic chart system, and iteratively solving the digital twin model by using a Dragon-Kutta method with the actual measurement state at the current moment as a starting point to generate a future state vector comprising a future battery electric quantity trend, a future equipment heat load state and a future energy consumption emission predicted value.
5. The method for estimating a performance of a marine hybrid system with predictive dynamic weights as recited in claim 4, wherein said process of deriving a digital twin model comprises: let the current time be The prediction step length is Constructing future state vectors Is a computational model of (a): Wherein, the A digital twin dynamics mapping function is represented, For the acquired actual measurement of the system state at the current time, For weather environment prediction data within a future preset time window, Planning data for future voyages.
6. The method for estimating the performance of a ship hybrid system with predictive dynamic weights according to claim 1, wherein the third step comprises the steps of: (31) Fusing the current state data and the external environment data obtained in the first step with the future state vector generated by deduction in the second step to construct an enhanced state vector, so that a decision module perceives the current situation and the future evolution trend of the system at the same time; (32) The LSTM network is utilized to extract time sequence characteristics of the enhanced state vector, decision is made through an Actor-Critic network architecture, wherein the Actor network outputs four-dimensional weight vectors meeting normalization conditions according to the input time sequence characteristics, the four-dimensional weight vectors respectively correspond to real-time weights of economy, environmental protection, technicality and reliability, the Critic network evaluates expected accumulated rewards of the weight actions in current and future deduction states, and the weight generation strategy is continuously optimized by maximizing accumulated rewards.
7. The method of predictive dynamic weighted marine hybrid system performance assessment of claim 6, wherein said enhanced state vector The concrete construction form of (a) is as follows: In the formula, For real-time status vectors containing current wind speed, wave height, oil price, mission phase and equipment health, For the deduced future system state, Is a future environmental change rate characteristic.
8. The method for estimating performance of a ship hybrid system with predictive dynamic weights as recited in claim 6, wherein said predictive dynamic weight decision module rewards functions during training thereof A trend penalty term based on twin deduction is introduced: Wherein, the In order to standardize the cost of operation, In order to standardize the total amount of emissions, The comprehensive efficiency of the system is achieved; Is a basic weight coefficient; For a prejudgment penalty function based on a digital twin deduction result, Are bonus coefficients.
9. A method of predictive dynamic weighted marine hybrid system performance evaluation in accordance with any one of claims 1-8, further comprising: when the evaluation result shows that the current comprehensive score is lower than a preset threshold value, automatically identifying key indexes causing low scores: If the main reason is that the environmental protection score is too low and the digital twin deduction display is about to enter an emission control area, the automatic generation instruction is preferably switched to a fuel cell or lithium battery power mode, and if the main reason is that the economic score is too low and the system is in an open sea area, the system is recommended to be switched to a diesel engine direct drive mode and optimized to the economic navigational speed.
10. A predictive dynamic weighted marine hybrid system performance evaluation system, comprising: The data acquisition and multidimensional sensing module is used for acquiring power system state data, task data, market data and environment data of the ship in real time through the sensor network, the weather station and the communication interface; The digital twin deduction module is connected with the data acquisition and multidimensional sensing module and is used for storing a digital twin model of the ship hybrid power system, carrying out super real-time simulation deduction on the future running state of the system by combining the received weather forecast and the navigation plan, and outputting a future state vector; The predictive dynamic weight decision module is respectively connected with the data acquisition and multidimensional sensing module and the digital twin deduction module, is internally provided with an agent based on an LSTM-DDPG algorithm and is used for receiving current state data and future state vectors, and the predictive dynamic weight of each evaluation dimension is calculated and output in real time through an enhanced state space analysis and reinforcement learning strategy; the data standardization processing module is used for receiving the performance index original data and carrying out extremely poor standardization processing on the data by utilizing a dynamic boundary based on the health state of the equipment; And the weighted comprehensive evaluation module is used for receiving the standardized index data and the predictive dynamic weight, calculating the comprehensive performance score, and displaying the score through a user interface or sending an optimization instruction to the ship control system.

Description

Ship hybrid power system performance evaluation method and system with predictive dynamic weights Technical Field The invention belongs to the field of control of ship hybrid power systems, and particularly relates to a ship hybrid power system performance evaluation method and system with predictive dynamic weights. Background With the increasing strictness of International Maritime Organization (IMO) emissions regulations and the increasing demand for "green shipping", marine hybrid systems (integrated diesel, LNG, lithium batteries, fuel cells, etc.) have become the industry's core development. However, performance assessment of such systems involves multiple mutually constrained dimensions of economy, environmental protection, technology, etc., and is extremely complex. In the prior art, for example, chinese patent CN120509525A, although an energy optimization management method is proposed, the method focuses on the generation of real-time control instructions, and a simple feedback mechanism with fixed weight or based on the current state is mostly adopted. These prior art techniques have the major disadvantage of assessing hysteresis, namely, merely adjusting weights based on "current" or "historical" data, and the inability to predict future conditions (e.g., about to enter an emissions control zone or about to encounter severe sea conditions), resulting in a slow response of the system in the event of a sudden change in conditions, and the inability to achieve "preventative" optimization. Baseline rigidification-in the data normalization process, a fixed maximum/minimum value is generally used as a baseline, and objective facts of aging of ship equipment (such as battery capacity fading) along with time are ignored, so that the evaluation result of the old ship is distorted, and the relative optimal performance of the old ship under the current health state cannot be reflected. The lack of global vision is that the training sample of the single ship intelligent body is limited, and the capability of overall planning of future voyages is lacking, so that the global optimization of the whole voyages is difficult to realize. Therefore, there is a need for a performance assessment method that can integrate future state predictions, have prospective decision making capabilities, and accommodate full life cycle changes in the device. Disclosure of Invention Aiming at the defects or improvement demands of the prior art, the invention provides a ship hybrid power system performance evaluation method and system with predictive dynamic weight, which are used for carrying out super-real-time deduction on future sailing working conditions by constructing a digital twin model, integrating a deduction result into an LSTM-DDPG-based reinforcement learning decision mechanism, realizing the basic crossing of an evaluation system from 'current-situation-based reactive evaluation' to 'future-based predictive evaluation', and simultaneously combining a dynamic standardization method along with the drift of the health state of equipment, effectively solving the problems of decision errors and evaluation distortion caused by evaluation lag and reference rigidification in the prior art, and providing comprehensive performance evaluation and optimization support which cover a full life cycle, have forward-looking vision and take into consideration of multi-objective cooperation for the ship hybrid power system. To achieve the above object, according to one aspect of the present invention, there is provided a ship hybrid system performance evaluation method of predictive dynamic weights, comprising the steps of: Firstly, constructing a multidimensional performance evaluation index system and ship perception data, wherein the evaluation index system comprises economy, environmental protection, technical performance and reliability, the ship perception data comprises current state data and external environment data of a ship hybrid power system, and the data is subjected to standardized processing; Step two, a digital twin model mapped with a physical ship hybrid power system is constructed, the model comprises a ship motion dynamics model and an energy flow model, a future environment prediction data sequence obtained according to weather forecast and a future sailing task sequence of a ship sailing plan are used as initial boundary conditions, the current state data, the future environment prediction data sequence and the future sailing task sequence are input into the digital twin model together, so that a system response state in a future preset time window is deduced, and a future state vector comprising a future battery electric quantity trend, a future equipment thermal load state and a future energy consumption emission predicted value is generated; Step three, a predictive dynamic weight decision module based on LSTM and DDPG is built, the current state data and the external environment data obtained in the step S1 a