Energy management method of multi-stack fuel cell vehicle based on hierarchical distributed model predictive control

Abstract

The invention relates to an energy management method of a multi-stack fuel cell automobile based on hierarchical distributed model predictive control, and belongs to the field of new energy automobiles. The method comprises the following steps: firstly, configuring and modeling a fuel cell automobile power system, and establishing a complete mathematical model comprising a fuel cell module, a cell unit and a converter; and then constructing a centralized model predictive control optimization problem based on the model, and converting the centralized model predictive control optimization problem into a distributed model predictive control architecture of gain scheduling by introducing a virtual power vector and consistency constraint. The method effectively solves the problems that an existing centralized energy management strategy is insufficient in flexibility, poor in expansibility, high in single-point fault risk and the like, distributed collaborative optimization management of all energy units is achieved, and the reliability, the modularization degree, the dynamic response performance and the overall energy efficiency of the system are remarkably improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of new energy vehicles, and relates to an energy management method for a multi-stack fuel cell vehicle based on a hierarchical distributed model predictive control. BACKGROUND

[0002] With the increasing trend of global warming and the growing challenge of climate change caused by the use of fossil fuels, the transportation industry, as one of the important factors leading to this problem, urgently needs to transition from traditional internal combustion engine vehicles to new energy vehicles. Fuel cell vehicles, with their high energy efficiency and zero emission characteristics, have become one of the most promising alternatives. Such vehicles use green hydrogen produced by the electrolysis of water from renewable energy sources as fuel, with the significant advantages of environmental friendliness and sustainability.

[0003] Fuel cells, as the core power source of fuel cell vehicles, are power generation devices that use oxygen as an oxidant and hydrogen as a fuel. They have the advantages of high power density, low noise, light weight, fast startup speed, and stable operation, and are widely used in the field of fuel cell vehicles and have become the main power source. The performance and efficiency of fuel cell vehicles depend largely on the effectiveness of their energy management strategies.

[0004] Currently, the focus of research in this field is mainly on centralized energy management systems. Such systems usually use a single control unit to manage and coordinate all energy units of the vehicle. Although centralized control can achieve global optimization under certain conditions, its inherent lack of flexibility limits the actual application effect. Especially in modularly designed fuel cell vehicles, when it is necessary to add or reduce fuel cell stack modules, the centralized system lacks the ability to plug and play, resulting in poor system scalability and adaptability. In addition, the centralized control architecture has a single point of failure risk, and the system reliability is challenged.

[0005] Model predictive control, as an advanced control strategy, shows good application prospects in the field of energy management. However, existing energy management methods based on model predictive control mostly adopt a centralized framework, failing to fully exploit the potential advantages of distributed control in flexibility, reliability, and real-time performance. Therefore, there is an urgent need in the field for a hierarchical distributed energy management method that combines the advantages of model predictive control to solve the problem of energy coordination and distribution in multi-stack fuel cell vehicles under complex operating conditions, while improving the modularity and robustness of the system. SUMMARY

[0006] In view of the above, the purpose of the present application is to provide a kind of energy management method of multi-stack fuel cell vehicle based on hierarchical decentralized model predictive control.The hierarchical decentralized energy management system (Dec-EMS) based on model predictive control (Model Predictive Control, MPC) proposes a kind of modular FCV power system for being composed of two parallel proton exchange membrane fuel cells and an energy storage system.Gain scheduling makes the proposed Dec-EMS controller more effective in its performance.

[0007] To achieve the above purpose, the present application provides the following technical solutions: A kind of energy management method of multi-stack fuel cell vehicle based on hierarchical decentralized model predictive control, comprising the following steps: S1: fuel cell vehicle power system configuration and modeling; S2: based on centralized model predictive control (Centralized Model Predictive Control, Cen-MPC) it is converted into gain scheduling decentralized model predictive control (Decentralized Model Predictive Control, Dec-MPC).

[0008] Further, the fuel cell vehicle power system configuration and modeling of step S1, specifically includes: S11: the power balance equation of FC module and battery cell is established; S12: in the balance equation established in step S11, the upper limit, lower limit and conversion rate of the instantaneous power of each module fuel cell are limited; S13: the first-order RC model of battery pack and the power and conversion rate applied to battery cell Limit; S14: the calculation formula of battery state of charge (State of Charge, SOC), and the constraint condition of battery SOC level; S15: two converter modeling.

[0009] Further, the power balance equation of step S11 is:

[0010] Wherein, The power of each module Is represented by P m, P b Represent the power provided by battery cell, P req It is the requested power from propulsion system.

[0011] Limits are imposed on the upper and lower bounds and slew rate of the instantaneous power of each module fuel cell, which are:

[0012]

[0013]

[0014] where and are the minimum and maximum limits, and are the slew rate bounds, denotes the time step.

[0015] The first-order RC model of the battery pack and the cell Power and slew rate limits are imposed: The first-order RC model of the battery pack is represented by the following equation:

[0016] where is the battery pack current, is the open circuit voltage, is the series ohmic resistance, is the terminal voltage, denotes the polarization resistance, is the polarization capacitor. The cell PB, k power and slew rate limits are as follows:

[0017]

[0018]

[0019] where and are the minimum and maximum limits, and are the slew rate bounds of .

[0020] The formula for calculating the battery SOC in step S14, and the constraint conditions for the battery SOC level:

[0021]

[0022] where and denote the minimum and maximum limits of the SOC, respectively, the initial SOC level​ is , represents the battery capacity.

[0023] The two converters modeling in step S15:

[0024]

[0025] where and are the current and voltage of , represents the smoothing inductor inductance, is the smoothing inductor resistance, is the average efficiency, is the modulation ratio of the converter.

[0026] Further, the Cen-MPC-based step S2 is converted into a gain-scheduled Dec-MPC. Specifically, the following steps are included: S21: build a centralized energy management optimization problem based on convex optimization; S22: convert the Cen-MPC control method into the Dec-MPC control method based on step S21; S23: develop a gain-scheduled decentralized controller.

[0027] Further, the optimization problem of the centralized convex optimization-based energy management strategy in step S21 is established as follows: The centralized multi-objective problem based on convex optimization (such as Figure 3 ) can be modeled in the following order:

[0028]

[0029] where represents the hydrogen cost of , is the battery unit SOC penalty cost, represents the power of module m, and c apply powertrain and coupling constraints to the modules, respectively. is calculated by , where is a quadratic approximation function for calculating the hydrogen consumption cost, is the hydrogen price, representing the hydrogen price, represents the time step; is a penalty term for measuring the change in SOC level, which is defined as follows:

[0030] wherein is the initial SOC, is a significant positive coefficient.

[0031] Further, the Cen-MPC control is converted to Dec-MPC control in step S22 In the proposed Dec-MPC, the master problem is decomposed into m∈M sub-problems, and each sub-problem is assigned to a FC module control unit. Since the modules are coupled on the DC bus, the general problem is not decomposable. To solve this problem, is replicated into its neighboring modules as virtual power and coupled with the global power vector, The following constraint is added to guarantee the replicated variables are equal to each other, and the modified EMS converges to the same optimal optimization result.

[0032]

[0033]

[0034]

[0035]

[0036] Further, the gain-scheduled decentralized controller developed in step S23.

[0037] The gain-scheduled MPC coordinates the control of a complex system operating over a wide range of operating conditions by switching among a set of pre-designed MPC controllers. To establish a hierarchical distributed controller, the controllers are designed at four operating points based on the driving cycle information. These individual regulators are used to cover the entire driving data range. When the required reference power lies between two operating points, the percentage error between the reference points is linearly weighted and mixed, and multiplied by the corresponding control value.

[0038] The present application has the following advantages: (1) The present application discards the traditional centralized control architecture, and creatively adopts a hierarchical decentralized control structure. The method reasonably decomposes the complex global optimization problem into a plurality of relatively simple sub-problems, and parallelly processes by the fuel cell module control units distributed everywhere. This architecture fundamentally avoids the single-point failure risk that may be caused by a single control unit, greatly improves the reliability and robustness of the entire power system operation. Even if an abnormality occurs in a certain sub-control unit, the remaining units can still continue to work cooperatively, ensuring the basic operating ability of the vehicle and strengthening the driving safety.

[0039] (2) The other core advantage brought by the application is excellent flexibility and scalability, which perfectly realizes the function requirement of plug and play. When the power system needs to be adjusted modularly according to actual requirements, for example, the number of fuel cell stacks is increased or decreased, the architecture of the whole control system does not need to be reconstructed or complicatedly reprogrammed. After the new module is connected, the control unit can automatically integrate into the existing distributed collaborative management network, greatly simplifying the system maintenance and upgrading process, and laying a solid technical foundation for the serial development and personalized configuration of products.

[0040] (3) The application is not a simple distributed control, but deeply integrates the advanced concepts of gain scheduling and model predictive control. By presetting multiple optimization working points for different typical driving conditions and designing an intelligent scheduling strategy, the control system can cover a wide range of working conditions. When the vehicle operating state changes, the system can smoothly switch or weightedly integrate between different working modes, so as to ensure that nearly optimal energy distribution instructions can be generated in any condition. This method significantly improves the overall energy efficiency of the power system, and effectively ensures that the working of key components such as fuel cells and batteries is in the efficient and safe interval, prolonging the service life.

[0041] (4) The application successfully solves the coupling problem between distributed subsystems by introducing virtual power vectors and consistency constraints, ensuring that the local decision of distributed calculation can finally converge to the global optimal solution or suboptimal solution. This makes each fuel cell module not only respond quickly according to its own state, but also realize overall coordination through information interaction with neighbor modules, finally achieving the multi-objective optimization goal of vehicle energy management, including reducing hydrogen consumption, maintaining battery state of charge balance, and meeting power demand, etc.

[0042] Other advantages, objects, and features of the application will be set forth in part in the description which follows, and in part will become apparent to those skilled in the art upon examination of the following specification or can be learned by practice of the application. The objects and other advantages of the application can be realized and attained by the methods and instrumentalities particularly pointed out in the description. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to make the objects, technical solutions and advantages of the application clearer, the preferred detailed description of the application will be combined with the drawings to describe the application, and the drawings are as follows: Figure 1 is the general logic diagram of the application; Figure 2 is the vehicle power system used in the application; Figure 3 is the general structure of the real-time algorithm of the energy management strategy optimization problem; Figure 4A gain-scheduled decentralized controller is developed for the invention. DETAILED DESCRIPTION

[0044] Other advantages and benefits of the present application will become apparent to those skilled in the art upon consideration of the disclosure or can be learned by practice of the application. The application can be realized and achieved by means of the structures and combinations of structures described in this specification and claims. Various embodiments of the present application can be realized and carried out in other different ways than those specifically described herein without departing from the spirit of the application. Therefore, the disclosure of the embodiments described herein using specific terminology is to be considered only as illustrative of the generic principles that should be considered to be encompassed by the disclosure. The disclosure is not limited to the embodiments described herein but is capable of being practiced with variations of and modifications to the embodiments described herein.

[0045] The drawings are only used for exemplary illustration, and the representation is only a schematic diagram, not a physical diagram, and should not be understood as a limitation on the present application. In order to better illustrate the embodiments of the present application, some components in the drawings may be omitted, enlarged or reduced, and do not represent the actual product size. It is understandable to those skilled in the art that some well-known structures and their descriptions in the drawings may be omitted.

[0046] The same or similar reference numerals in the drawings of the embodiments of the present application correspond to the same or similar components; in the description of the present application, it should be understood that if the terms "upper", "lower", "left", "right", "front", "back" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, therefore the terms describing the positional relationship in the drawings are only used for exemplary illustration, and should not be understood as a limitation on the present application. For those skilled in the art, the specific meanings of the above terms can be understood according to the specific circumstances.

[0047] Referring to Figures 1-4 , the present application provides an energy management method for a multi-stack fuel cell vehicle based on a hierarchical decentralized model predictive control. The method specifically comprises the following steps: Step S1: establish the configuration and modeling of the fuel cell vehicle power system as a reference, as shown in Figure 2 .

[0048] S11: the power balance equation of the module and the battery unit on the DC bus is represented as follows:

[0049] wherein, Pm represents the power of each module, Pb represents the power provided by the battery unit, ​is the requested power from the propulsion system S12: The upper and lower limits of the instantaneous power and the slew rate impose constraints:

[0050]

[0051]

[0052] where and are the minimum and maximum limits of and are the slew rate boundaries, denotes the time step.

[0053] S13: First-order RC model of the battery pack and the battery cell Power and slew rate limits are imposed: The first-order RC model of the battery pack is represented by the following equation:

[0054] where is the battery pack current, is the open-circuit voltage, is the series ohmic resistance, is the terminal voltage, denotes the polarization resistance, is the polarization capacitor. The battery cell Power and slew rate limits are as follows:

[0055]

[0056]

[0057] where and are the minimum and maximum limits of and are the slew rate boundaries of

[0058] S14: Formula for the calculation of the battery state of charge (SOC) and constraints on the battery SOC level:

[0059]

[0060] where ​​​​and respectively represent the minimum and maximum limits of S C, initial S C level is , represents the battery capacity.

[0061] S15: Two converter modeling:

[0062]

[0063] where and are the current and voltage of S respectively, represents the smoothing inductor inductance, is the smoothing inductor resistance, is the average efficiency, is the modulation ratio of the converter.

[0064] Step S2: Convert Cen-MPC to Dec-MPC, specifically comprising the following steps: S21: Build a centralized energy management optimization problem based on convex optimization: As shown in Figure 3 , the centralized energy management optimization problem based on convex optimization can be modeled in the following order:

[0065]

[0066] where represents the hydrogen cost of S C, is the battery unit S C penalty cost, represents the power of module m, and c apply the powertrain and coupling constraints to the module respectively. is calculated by , where is a quadratic approximation function for calculating the hydrogen consumption cost, is the hydrogen price, is a penalty term for measuring the change in S C level, which is defined as follows:

[0067] where is the initial SOC, is a significantly positive coefficient.

[0068] S22: Transform Cen-MPC to Dec-MPC: In the proposed Dec-MPC, the main problem is decomposed into m e M sub-problems, and each sub-problem is assigned to a FC module control unit. Since the modules are coupled on the DC bus, the general problem is not decomposable. To solve this problem, is replicated into its neighboring modules as virtual power and coupled with the global power vector, . The following constraints are added to guarantee the replicated variables are equal to each other, and the modified EMS converges to the same optimal optimization result.

[0069]

[0070]

[0071]

[0072]

[0073] S23: Gain scheduling structure in the constructed distributed predictive strategy The gain-scheduled MPC coordinates the control of a complex system operating over a wide range of working conditions by smoothly switching and blending between a set of pre-defined, optimally-tuned MPC controllers for different typical driving conditions.

[0074] To establish a hierarchical distributed controller, four representative reference power working points are selected to cover the whole driving data range from low load to high load based on the statistical analysis of historical driving cycle data. The four working points (WP) and their corresponding typical working conditions are defined as follows: WP1 (low load working point): the reference power is preset to P ref1 kW, corresponding to low-speed cruising, following or congestion slow driving conditions in urban roads; WP2 (low-medium load working point): the reference power is preset to P ref2 kW, corresponding to normal acceleration, deceleration and medium-speed cruising conditions in urban roads; WP3 (medium-high load working point): the reference power is preset to P ref3 kW, corresponding to higher-speed cruising and slow acceleration conditions in urban and suburban roads; WP4 (high load working point): the reference power is preset to P ref4 kW, corresponding to high-power demand conditions such as overtaking, high-speed cruising or climbing on expressways.

[0075] wherein, P ref1 P ref2 P ref3 P ref4 These working points and their mapping relationship with driving conditions are shown in FIG. 3. Figure 4

[0076] Finally, it should be pointed out that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the purpose and scope of the technical solutions, and they should all be covered in the scope of the claims of the present application.​​​​

Claims

1. An energy management method for multi-stack fuel cell vehicles based on hierarchical distributed model predictive control, characterized in that: Includes the following steps: S1: Configure the power system of the fuel cell vehicle and establish its mathematical model; S2: Based on the mathematical model established in S1, construct a centralized model predictive control (Cen-MPC) optimization problem, and transform the Cen-MPC control architecture into a gain-scheduled decentralized model predictive control (Dec-MPC) architecture to achieve distributed collaborative management of each energy unit.

2. The energy management method for multi-stack fuel cell vehicles based on hierarchical distributed model predictive control according to claim 1, characterized in that: In S1, the configuration and modeling of the fuel cell vehicle powertrain system specifically includes: S11: Establish the power balance equations for the fuel cell (FC) module and battery cells; S12: Based on the power balance equation established in S11, limit the instantaneous power upper limit, lower limit and conversion rate of each FC module; S13: Establish a first-order RC model of the battery pack and analyze the power of the battery cells. P B,k Apply power and conversion rate limits; S14: Calculate the battery state of charge (SOC) level and apply constraints to it; S15: Model the two converters that connect the fuel cell module and the DC bus.

3. The energy management method for multi-stack fuel cell vehicles based on hierarchical distributed model predictive control according to claim 2, characterized in that: In S11, the power balance equation is: in, Represents each module power, This indicates the power provided by the battery cells. This indicates the requested power from the propulsion system.

4. The energy management method for multi-stack fuel cell vehicles based on hierarchical distributed model predictive control according to claim 3, characterized in that: In S12, the power limit for each FC module is expressed by the following formula: in, and yes The minimum and maximum values, and It is the slew rate boundary. Indicates the time step.

5. The energy management method for multi-stack fuel cell vehicles based on hierarchical distributed model predictive control according to claim 2, characterized in that: In S13, the first-order RC model of the battery pack is represented by the following formula: in Indicates the battery pack current. Indicates open-circuit voltage. Indicates a series ohmic resistor. Indicates terminal voltage. Indicates polarization resistance. Indicates a polarized capacitor; Battery cell power P B,k The power and conversion rate limitations are as follows: in and They are The minimum and maximum limits, and yes The slew rate boundary.

6. The energy management method for multi-stack fuel cell vehicles based on hierarchical distributed model predictive control according to claim 2, characterized in that: In S14, the formula for calculating the battery's state of charge (SoC) and the constraints on the battery SoC level are as follows: in, and These represent the minimum and maximum limits of the SoC, respectively, and the initial SoC level. yes , Indicates battery capacity.

7. The energy management method for multi-stack fuel cell vehicles based on hierarchical distributed model predictive control according to claim 2, characterized in that: In step S15, the modeling of the two converters is represented by the following formula: in, and They are The current and voltage, Indicates the inductance of a smooth inductor. Indicates the resistance of the smoothing inductor. This represents average efficiency. This indicates the modulation ratio of the converter.

8. The energy management method for multi-stack fuel cell vehicles based on hierarchical distributed model predictive control according to claim 1, characterized in that: In step S2, converting Cen-MPC to gain-scheduled Dec-MPC specifically includes: S21: Based on the model established in S1, a centralized multi-objective energy management optimization problem based on convex optimization is constructed; S22: By introducing a virtual power vector and adding consistency constraints, the Cen-MPC optimization problem in S21 is decomposed and assigned to each FC module control unit, thereby realizing the transformation to Dec-MPC control; S23: Develop a gain scheduling strategy to schedule multiple preset MPC working points based on real-time driving cycle information in order to coordinate the control of the distributed controller.

9. The energy management method for multi-stack fuel cell vehicles based on hierarchical distributed model predictive control according to claim 8, characterized in that: In S21, the centralized energy management optimization problem is constructed as follows: in, express The cost of hydrogen, This indicates the SOC penalty cost for the battery cell. Representation module m power, c applies powertrain and coupling constraints to the module, respectively; Depend on Calculation, where It is a quadratic approximation function used to calculate the cost of hydrogen consumption. Indicates the price of hydrogen. Indicates the time step. This represents the penalty term used to measure changes in SOC levels, defined as follows: in, Indicates the initial SOC. This indicates a significant positive coefficient.

10. The energy management method for multi-stack fuel cell vehicles based on hierarchical distributed model predictive control according to claim 8, characterized in that: In step S22, a virtual power vector is introduced. And add the following constraints to ensure that the decomposed subproblems converge to the optimal solution of the original problem: in, Indicates the number of iterations n At that time, the predicted power sequence of the first fuel cell module in the future prediction time domain {i+1,...,i+K}; Indicates the number of iterations n At that time, the controller of the first module copies and stores the predicted power sequence of the second module; Indicates the number of iterations n At that time, the predicted power sequence of the second fuel cell module in the future prediction time domain {i+1,..., i+K}; Indicates the number of iterations n At that time, the controller of the second module copies and stores the predicted power sequence of the first module; Indicates the number of iterations n At that time, it corresponds to the global consensus power vector of the first module; Indicates the number of iterations n At that time, it corresponds to the global consistency power vector of the second module; The gain scheduling strategy in S23 is as follows: when the required reference power is between two preset operating points, the control outputs of the two operating points are linearly weighted and mixed based on percentage error.

Images