Energy management method of hybrid power ship

Abstract

The invention discloses an energy management method of a hybrid power ship. The method comprises a management model construction process and a real-time use process. The construction process of the management model comprises the following steps of: A1, acquiring historical navigation data of a ship; a2, carrying out working condition mode division on the historical navigation data; step A3, constructing a layered framework model as a management model; step A4, generating an optimal working mode corresponding to each working condition through a mixed integer linear programming model; the real-time use process comprises the following steps: step B1, acquiring real-time navigation data of a ship, and selecting a working condition corresponding to the real-time navigation data of the ship; and B2, performing energy management prediction by model prediction control in the management model according to the optimal working mode corresponding to the working condition, and optimizing ship energy control according to a prediction result. The energy management efficiency of the hybrid power ship is improved, the fuel economy, the environmental protection property and the equipment reliability are improved, and the method is suitable for more scenes.

Description

Technical Field

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

[0002] Hybrid electric vehicles (HEVs) combine the advantages of diesel engines and electric motors, offering significant potential for energy conservation and emissions reduction. They can significantly improve energy efficiency in fixed-route vessels such as city ferries and inland waterway vessels, particularly in scenarios involving frequent starts and stops and intermittent charging.

[0003] Traditional ship energy management systems mostly use rule-based control or static optimization models, which are difficult to adapt to the dynamic load demands of cyclic operation modes. Existing systems often lead to fuel waste and battery life loss when dealing with scenarios such as frequent starts and stops and intermittent availability of charging stations. The real-time tracking of the energy storage system's SOC and the global optimization of the generator set's start-stop strategy lack coordination, making it difficult to balance short-term response and long-term efficiency. Traditional dynamic programming methods are computationally complex and difficult to apply in real time; a single optimization model is difficult to cope with sudden changes in operating scenarios. Existing methods mostly rely on fixed rules and are unable to adjust optimization targets based on historical operating conditions, resulting in insufficient energy efficiency potential. Summary of the Invention

[0004] In response to the deficiencies in the existing technology, the present invention provides an energy management method for hybrid ships to solve the technical problem that the existing technology relies on fixed rules and cannot adjust the optimization targets according to historical operating conditions, resulting in the energy efficiency potential not being fully released.

[0005] The present invention provides an energy management method for a hybrid ship, comprising: a management model building process and a real-time use process;

[0006] The management model building process includes:

[0007] Step A1: Obtain historical ship navigation data;

[0008] Step A2: Divide the historical navigation data into operating modes and identify several operating modes;

[0009] Step A3: Construct a hierarchical framework model as a management model, wherein the upper layer is a mixed integer linear programming model; the lower layer is a model predictive control;

[0010] Step A4: The historical navigation data and corresponding operating conditions in step A2 are used as inputs to the management model. The mixed integer linear programming model in the management model generates the optimal operating mode corresponding to each operating condition.

[0011] The real-time usage process includes:

[0012] Step B1: Acquire real-time ship navigation data and select the working condition corresponding to the real-time ship navigation data;

[0013] Step B2: The selected operating condition is used as the input of the management model. The model predictive control in the management model performs energy management prediction based on the optimal working mode corresponding to the operating condition, and optimizes the control of ship energy based on the prediction results.

[0014] Furthermore, in step A2, the historical navigation data is divided into operating mode patterns using the K-means clustering method.

[0015] Furthermore, the objective function of the K-means clustering method is:

[0016]

[0017] Where K is the number of cluster categories dynamically optimized by the silhouette coefficient method; is the mean speed; σ P is the load power variance; t charget is the charging station stop time; C i refers to the i-th cluster; μ i Refers to the i-th cluster center.

[0018] Furthermore, the objective function of the mixed integer linear programming model is:

[0019]

[0020] Where, F fuel (t) is the fuel consumption of the generator at time t; is the carbon emission cost; Cdegradation is the battery life degradation cost; y(t) is the generator start-stop state variable; α is the fuel consumption cost weight coefficient; β is the carbon emission cost weight coefficient; γ is the SOC tracking deviation weight coefficient; λ is the generator start-stop penalty weight coefficient; SOC ref is the reference trajectory of SOC.

[0021] Furthermore, the constraints of the mixed integer linear programming model include: battery dynamic constraints, battery safety constraints, generator set operation constraints, and load power balance constraints;

[0022] in,

[0023] Battery dynamic constraints:

[0024]

[0025] Battery safety constraints:

[0026] 0.2≤SOC(t)≤0.9,0≤Pbat (t)≤160kW

[0027] Generator set operating constraints:

[0028] 100kW≤P gen (t)≤300kW,ΔP gen (t)≤15kW / s

[0029] Load power balance constraints:

[0030] P load (t) = P gen (t)+P dis (t)-P ch (t)

[0031] Where n ch Battery charging efficiency; n dis is the battery discharge efficiency; E bat is the total capacity of the battery; P ch (t) is the charging power of the battery at time t; P dis (t) is the discharge power of the battery at time t; P bat The maximum power allowed by the battery; P gen (t) is the output power of the generator set; ΔP gen (t) is the power change rate limit of the generator set; Δt is the battery time step; P load (t) is the load demand power.

[0032] Furthermore, the objective function of the model predictive control is:

[0033]

[0034] Where k is the discrete time step index in the prediction time domain; H is the number of prediction time domain steps; SOC(k) is the battery state of charge predicted in the kth step; SOC ref (k) is the SOC reference trajectory of the kth step; P gen (k) is the output power of the generator set predicted in step k; P gen,opt (t) is the optimal power of the generator at step k.

[0035] Furthermore, the constraints of the model predictive control include:

[0036] Battery dynamic constraint recursion:

[0037]

[0038] Generator set power constraints:

[0039] P gen (k)≤300kW,ΔPgen (k)≤15kW / s

[0040] Where, P ch (k) is the battery charging power at step k; P dis (k) is the battery discharge power at step k; ΔP gen (k) is the power change rate limit of the generator set.

[0041] Furthermore, in step B1, the specific method for selecting the working condition corresponding to the real-time navigation data of the ship is:

[0042] The distance between the real-time ship navigation data and the cluster center corresponding to each working condition is calculated by the weighted cosine similarity algorithm, and the working condition with the closest distance is selected as the working condition corresponding to the current real-time ship navigation data.

[0043] Furthermore, the distance is calculated as follows:

[0044]

[0045] Where x j is the real-time feature vector; μ i,j is the jth feature of the i-th cluster center; w = [0.3, 0.5, 0.2] is the feature weight vector, corresponding to the weights of the mean speed, load power variance, and charging time, respectively.

[0046] Furthermore, after step B2, the following steps are further included:

[0047] Step B3: Determine whether steps A2-A4 need to be executed based on the optimization trigger conditions, and update the operating condition classification and the corresponding optimal operating mode. The optimization trigger conditions include: when the difference between the planned charging time and the actual charging time is greater than the preset difference, steps A2-A4 need to be executed.

[0048] Furthermore, in B3, the optimization triggering condition includes: when the difference between the planned charging time and the actual charging time is greater than a preset difference, or the relative deviation of the load power is greater than a preset deviation, steps A2-A4 need to be executed.

[0049] Furthermore, during the updating process of step B3, the SOC reference trajectory is:

[0050] SOC ref,new (t) = SOC ref,old (t)+0.01·δ charge

[0051] Beneficial effects of the present invention:

[0052] By combining a hierarchical optimization framework with a dynamic adaptive mechanism, the present invention significantly improves the energy management efficiency of hybrid ships, achieving a coordinated improvement in fuel economy, environmental protection, and equipment reliability. It is suitable for fixed-route scenarios such as urban ferries and inland ships, and has significant engineering application value.

[0053] Based on cluster analysis of historical navigation data, the present invention systematically and accurately divides typical operating modes, providing reliable input for global optimization; the mixed integer linear programming model in the present invention generates the optimal strategy for each operating condition, optimizes the start-stop plan of the generator set and the coordinated control of battery charging and discharging, reduces redundant operation and balances fuel consumption and carbon emissions; the model predictive control in the present invention relies on the dynamic model to adjust the battery power in real time, ensures that the state of charge tracks the reference trajectory, and quickly responds to load fluctuations; the present invention adopts the weighted cosine similarity algorithm to efficiently match the real-time operating conditions, and improves the accuracy and real-time performance of the strategy call; the present invention adopts an adaptive update mechanism to dynamically correct the reference trajectory and the operating condition library to enhance the system's adaptability to charging time deviations and load mutations; the present invention constrains the charging and discharging power and the state of charge fluctuation range through the battery life degradation cost function, and effectively delays battery aging. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the present invention in any way. In the accompanying drawings:

[0055] Figure 1 It is a flow chart of a specific embodiment of the present invention. DETAILED DESCRIPTION

[0056] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0057] The present invention will be further described below with reference to specific examples. Those skilled in the art will appreciate that these examples are intended only to illustrate the present invention and are not intended to limit the scope of the present invention, and that modifications to various equivalent forms of the present invention fall within the scope defined by the appended claims.

[0058] like Figure 1 As shown, the present invention provides an energy management method for a hybrid ship, including: a management model building process and a real-time use process;

[0059] The management model building process includes:

[0060] Step A1: Obtain historical navigation data of the ship, including: speed v(t), load power P load (t), charging station stop time t charge , battery charge and discharge power P bat (t) and energy storage system charging state data SOC(t);

[0061] Step A2: Using the K-means clustering method, the historical navigation data is divided into operating mode categories to create an operating condition library that includes at least three classic operating conditions, such as high load, low load, and charging, where each operating condition corresponds to a working mode.

[0062] The objective function of the K-means clustering method is:

[0063]

[0064] Where K is the number of cluster categories dynamically optimized by the silhouette coefficient method; is the mean speed; σ P is the load power variance; tcharget is the charging station stop time; Ci refers to the i-th cluster; μi refers to the i-th cluster center.

[0065] Among them, the silhouette coefficient method evaluates the clustering quality through the silhouette coefficient and dynamically determines the optimal number of clusters K to balance the intra-class compactness and inter-class separation.

[0066] Step A3: Construct a hierarchical framework model as a management model, where the upper layer is a mixed integer linear programming model; the lower layer is model predictive control. The mixed integer linear programming model (MILP) is used to calculate generator fuel consumption, carbon emission costs, and generator set start and stop state variables. A battery life degradation cost function is introduced to optimize the generator set start and stop plan and SOC reference trajectory throughout the entire route cycle. Model predictive control (MPC) is used to adjust the battery charge and discharge power in real time.

[0067] The standard mathematical expression of the MILP problem is as follows:

[0068] minc T x

[0069] stAx≤b

[0070]

[0071] Where x is a decision variable vector, containing integer variables x i and the continuous variable x j ;

[0072] c is the coefficient vector of the objective function;

[0073] A is the constraint coefficient matrix, b is the constraint vector;

[0074] I is the index set of integer variables, and J is the index set of continuous variables.

[0075] The objective function of the mixed integer linear programming model is:

[0076]

[0077] Where, is the fuel consumption of the generator at time t;

[0078] c co2 (t)=3.17·F fuel (t), is the carbon emission cost;

[0079] The cost of battery life degradation;

[0080] y(t)∈{0,1}, is the generator start-stop state variable; α is the fuel consumption cost weight coefficient; β is the carbon emission cost weight coefficient; γ is the SOC tracking deviation weight coefficient; λ is the generator start-stop penalty weight coefficient; SOC ref is the reference trajectory of SOC.

[0081] The constraints of the MILP model include battery dynamic constraints, battery safety constraints, generator set operation constraints, and load power balance.

[0082] Among them, the battery dynamic constraint is:

[0083]

[0084] n ch =0.93,n dis =0.95,E bat =500kWh

[0085] The battery safety constraints are:

[0086] 0.2≤SOC(t)≤0.9,0≤P bat (t)≤160kW

[0087] P bat,max =200kW×0.8=160kW

[0088] The operating constraints of the generator set are:

[0089] 100kW≤P gen (t)≤300kW,ΔP gen(t)≤15kW / s

[0090] The load power balance is:

[0091] P load (t) = P gen (t)+P dis (t)-P ch (t)

[0092] Where n ch Battery charging efficiency; n dis is the battery discharge efficiency; E bat is the total capacity of the battery; P ch (t) is the charging power of the battery at time t; P dis (t) is the discharge power of the battery at time t; P bat The maximum power allowed by the battery; P gen (t) is the output power of the generator set; ΔP gen (t) is the power change rate limit of the generator set; Δt is the battery time step; P load (t) is the load demand power.

[0093] The expression formula of MPC is as follows:

[0094] MPC relies on a dynamic model of the system, typically a discrete-time state-space model, to predict future states:

[0095] x k+1 =f(x k ,u k )

[0096] The common form of a linear time-invariant (LTI) system is:

[0097] x k+1 =Ax k +Bu k

[0098] y k =Cx k

[0099] Where, state vector;

[0100] Control input;

[0101] Output vector.

[0102] The rolling optimization objective function of MPC is:

[0103]

[0104] Among them, the prediction time domain H = 10, corresponding to 30 seconds; the control period Δt MPC =50ms;

[0105] Where k is the discrete time step index in the prediction time domain; H is the number of prediction time domain steps; 0.7 is the SOC tracking weight coefficient; SOC(k) is the battery state of charge predicted at the kth step; SOC ref (k) is the SOC reference trajectory of the kth step; 0.3 is the power generation tracking weight coefficient; P gen (k) is the output power of the generator set predicted in step k; P gen,opt (t) is the optimal power of the generator at step k.

[0106] Dynamic model constraints include:

[0107] Battery dynamic constraint recursion:

[0108]

[0109] Generator set power constraints:

[0110] P gen (k)≤300kW,ΔP gen (k)≤15kW / s

[0111] Where, P ch (k) is the battery charging power at step k; P dis (k) is the battery discharge power at step k; ΔP gen (k) is the power change rate limit of the generator set; 0.93 is the battery charging efficiency; 0.95 is the battery discharging efficiency; 0.05 (h) is the battery time step, that is, 3 minutes; 500 (kWh) is the total capacity of the battery pack; 300 (kW) is the maximum output power of the generator set.

[0112] Step A4: The historical navigation data and corresponding operating conditions in step A2 are used as inputs to the management model. The mixed integer linear programming model in the management model generates the optimal operating mode corresponding to each operating condition.

[0113] The real-time usage process includes:

[0114] Step B1: Obtain the real-time navigation data of the ship, calculate the distance between the real-time navigation data of the ship and the cluster center corresponding to each working condition by using the weighted cosine similarity algorithm, and select the working condition with the closest distance as the working condition corresponding to the current real-time navigation data of the ship;

[0115] The distance calculation formula is:

[0116]

[0117] Select the working condition with the highest similarity and call the corresponding MILP optimization strategy.

[0118] Step B2: The selected operating condition is used as the input of the management model. The model predictive control in the management model performs energy management prediction based on the optimal working mode corresponding to the operating condition and optimizes the control of ship energy based on the prediction results.

[0119] Step B3: Determine whether steps A2-A4 need to be executed based on the optimization trigger condition, and update the operating condition classification and the corresponding optimal operating mode. Steps A2-A4 need to be executed when the difference between the planned charging time and the actual charging time is greater than a preset difference, or when the relative deviation of the load power is greater than a preset deviation.

[0120] The difference between the planned charging time and the actual charging time is:

[0121] δ charge =|t charge,plan -t charge,real |

[0122] The preset difference can be adjusted according to actual needs, preferably 2 minutes.

[0123] The relative deviation of load power is:

[0124]

[0125] The preset deviation can be adjusted according to actual needs, preferably 5%.

[0126] During the update process, the formula for adaptively updating the SOC reference trajectory is:

[0127] SOC ref,new (t) = SOC ref,old (t)+0.01·δ charge

[0128] Among them, δ charge The unit is minutes.

[0129] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for energy management of a hybrid ship, characterized in that: include: Manage the model building process and its real-time use; The management model building process includes: Step A1: Obtain historical ship navigation data; Step A2: Divide the historical navigation data into operating modes and identify several operating modes; Step A3: Construct a hierarchical framework model as a management model, wherein the upper layer is a mixed integer linear programming model; the lower layer is a model predictive control; Step A4: The historical navigation data and corresponding operating conditions in step A2 are used as inputs to the management model. The mixed integer linear programming model in the management model generates the optimal operating mode corresponding to each operating condition. The real-time usage process includes: Step B1: Acquire real-time ship navigation data and select the working condition corresponding to the real-time ship navigation data; Step B2: The selected operating condition is used as the input of the management model. The model predictive control in the management model performs energy management prediction based on the optimal working mode corresponding to the operating condition, and optimizes the control of ship energy based on the prediction results.

2. The energy management method for a hybrid ship according to claim 1, characterized in that: In step A2, the historical navigation data is divided into operating mode patterns using the K-means clustering method.

3. The energy management method for a hybrid ship according to claim 2, characterized in that: The objective function of the K-means clustering method is: Where K is the number of cluster categories dynamically optimized by the silhouette coefficient method; is the mean speed; σ P is the load power variance; t charget is the charging station stop time; C i refers to the i-th cluster; μ i Refers to the i-th cluster center.

4. The energy management method for a hybrid ship according to claim 1, wherein: The objective function of the mixed integer linear programming model is: Where, F fuel (t) is the fuel consumption of the generator at time t; is the carbon emission cost; Cdegradation is the battery life degradation cost; y(t) is the generator start-stop state variable; α is the fuel consumption cost weight coefficient; β is the carbon emission cost weight coefficient; γ is the SOC tracking deviation weight coefficient; λ is the generator start-stop penalty weight coefficient; SOC ref is the reference trajectory of SOC.

5. The energy management method for a hybrid ship according to claim 1 or 4, characterized in that: The constraints of the mixed integer linear programming model include: battery dynamic constraints, battery safety constraints, generator set operation constraints, and load power balance constraints; in, Battery dynamic constraints: Battery safety constraints: 0.2≤SOC(t)≤0.9,0≤P bat (t)≤160kW Generator set operating constraints: 100kW≤P gen (t)≤300kW,ΔP gen (t)≤15kW / s Load power balance constraints: P load (t)=P gen (t)+P dis (t)-P ch (t) Where n ch Battery charging efficiency; n dis is the battery discharge efficiency; E bat is the total capacity of the battery; P ch (t) is the charging power of the battery at time t; P dis (t) is the discharge power of the battery at time t; P bat The maximum power allowed by the battery; P gen (t) is the output power of the generator set; ΔP gen (t) is the power change rate limit of the generator set; Δt is the battery time step; P load (t) is the load demand power.

6. The energy management method for a hybrid ship according to claim 1, wherein: The objective function of the model predictive control is: Where k is the discrete time step index in the prediction time domain; H is the number of prediction time domain steps; SOC(k) is the battery state of charge predicted in the kth step; SOC ref (k) is the SOC reference trajectory of the kth step; P gen (k) is the output power of the generator set predicted in step k; P gen,opt (t) is the optimal power of the generator at step k.

7. The energy management method for a hybrid ship according to claim 1 or 6, characterized in that: The constraints of the model predictive control include: Battery dynamic constraint recursion: Generator power constraints: P gen (k)≤300kW,ΔP gen (k)≤15kW / s Where, P ch (k) is the battery charging power at step k; P dis (k) is the battery discharge power at step k; ΔP gen (k) is the power change rate limit of the generator set.

8. The energy management method for a hybrid ship according to claim 1, wherein: In step B1, the specific method for selecting the working condition corresponding to the real-time navigation data of the ship is: The distance between the real-time ship navigation data and the cluster center corresponding to each working condition is calculated by the weighted cosine similarity algorithm, and the working condition with the closest distance is selected as the working condition corresponding to the current real-time ship navigation data. The distance calculation formula is: Where x j is the real-time feature vector; μ i,j is the jth feature of the i-th cluster center; w = [0.3, 0.5, 0.2] is the feature weight vector, corresponding to the weights of the mean speed, load power variance, and charging time, respectively.

9. The energy management method for a hybrid ship according to claim 1, wherein: After step B2, the following steps are further included: Step B3: Determine whether steps A2-A4 need to be executed based on the optimization triggering conditions, and update the working condition classification and the corresponding optimal working mode. The optimization triggering conditions include: When the difference between the planned charging time and the actual charging time is greater than a preset difference, or the relative deviation of the load power is greater than a preset deviation, steps A2 to A4 need to be performed.

10. The energy management method for a hybrid ship according to claim 9, characterized in that: During the updating process of step B3, the SOC reference trajectory is: SOCIETY ref,new (t)=SOC ref,old (t)+0.01·δ charhe 。

Images