Multi-target dynamic optimization hydrogen energy unmanned aerial vehicle energy management method and system under MPC framework

Abstract

The invention discloses a multi-target dynamic optimization hydrogen energy unmanned aerial vehicle energy management method and system under an MPC framework, belongs to the technical field of energy management, comprehensively considers factors such as energy consumption, battery degradation and output fluctuation, and realizes safe and efficient operation of a hybrid power system based on a predictive control idea. The management system comprises a sensing subsystem, a control subsystem, an interaction subsystem and a power supply subsystem. The management method comprises the following specific steps: adjusting a predictive control time domain according to a historical demand power sequence; updating the aging parameter of the hydrogen fuel cell, and fixing the weight coefficient of the cost function; through a multi-target online optimization algorithm, obtaining an optimal SOC change curve under a prediction time domain; and outputting the reference power of the fuel cell through linear MPC control according to the reference change curve.

Description

Technical Field

[0001] The present invention belongs to the technical field of energy management, and particularly relates to a hydrogen energy unmanned aerial vehicle energy management method and system for multi-objective dynamic optimization under an MPC framework. Background Art

[0002] In recent years, with the continuous development of aviation technology, the development of unmanned aerial vehicles (UAVs) has received great attention. As an important core component of UAVs, the UAV power system determines the mobility and endurance of UAVs. Hydrogen energy UAVs use hydrogen fuel cells as power sources, which have high energy density, can meet the long endurance requirements, and can minimize the fuel carrying mass. However, their dynamic response is poor, and an auxiliary power supply system needs to be added, such as hanging a power battery, and a reasonable hybrid power topology needs to be designed to give full play to the characteristics of various types of power sources. At the same time, a reasonable energy management strategy needs to be designed to allocate power to each energy source, which has a positive effect on maintaining the stable operation state of the UAV, reducing energy consumption, and improving navigation safety. However, traditional multi-energy hybrid power energy management methods have defects such as parameter design relying on experience, non-optimal allocation, single optimization target, and insufficient real-time performance.

[0003] As an emerging hybrid power solution, hydrogen-lithium hybrid UAVs have significant advantages in endurance and energy efficiency, but still face some technical and application problems, such as energy distribution and dynamic response, system reliability design, etc. Based on the above problems, the present invention proposes a hydrogen energy UAV energy management method and system for multi-objective dynamic optimization under an MPC framework, taking model predictive control as the basic idea, providing a multi-objective real-time energy management method considering energy consumption, battery degradation, and output fluctuation, as well as a supporting hydrogen-lithium hybrid UAV energy management implementation system, to improve the real-time performance, stability, and safety of energy output under complex flight conditions of the UAV. Summary of the Invention

[0004] In order to make up for the deficiencies of the prior art, the purpose of the present invention is to provide a hydrogen energy UAV energy management method and system for multi-objective dynamic optimization under an MPC framework, comprehensively considering factors such as energy consumption, battery degradation, and output fluctuation, and based on the idea of predictive control, realizing the safe and efficient operation of the hybrid power system.

[0005] The present invention can be realized by the following technical solutions:

[0006] On the one hand, the present invention provides a hydrogen energy UAV energy management method for multi-objective dynamic optimization under an MPC framework, including the following steps:

[0007] Step 1: Adjust the prediction control time domain according to the historical demand power sequence;

[0008] Step 2: Update the hydrogen fuel cell aging parameters and fix the weight coefficients of the cost function;

[0009] Step 3: Obtain the optimal SOC change curve in the prediction time domain through the multi-objective online optimization algorithm;

[0010] Step 4: Output the reference power of the fuel cell through linear MPC control according to the reference change curve.

[0011] Further, in the said Step 1, the specific calculation process is as follows:

[0012] ① Based on the bus voltage V at the current sampling moment k bus , the current I of the DCDC converter DCDC , and the current I of the lithium battery bat , obtain the required power p at moment k k :

[0013] p k = V bus · (I DCDC + I bat ),

[0014] ② Update the sliding window sequence P of the historical load demand power with a length of N load = [p1, p2, p3... p k 1×N , calculate the standard deviation σ of the demand power sequence pload (k) and compare it with the historical maximum fluctuation standard deviation σ max and update σ max , where is the mean value of the historical load demand power sliding window sequence:

[0015]

[0016] ③ Calculate the prediction control time domain N at moment k pre , where the maximum prediction control time domain

[0017]

[0018] ④ According to the obtained historical demand power vector P load , the prediction control time domain N at moment k pre , substitute them into the first-order single-variable grey system model GM(1,1), P load as the historical sample input, N pre is the number of data predicted by the GM(1,1) model system, set the upper and lower limits of the predicted demand power, and update to obtain the predicted load demand power vector P with a length of N pre : pre :

[0019] ​

[0020] Furthermore, the calculation process of step 2 is as follows:

[0021] ① Based on the fuel cell semi-empirical voltage output model considering the leakage current and concentration polarization empirical formula, collect the fuel cell output voltage V fc (k), fuel cell output current I fc (k), fuel cell operating temperature T,

[0022]

[0023] where N cell is the number of fuel cell stack cells, E oc is the open circuit voltage of the fuel cell, T is the fuel cell operating temperature, i0 is the exchange current density, i n is the leakage current density, i is the fuel cell operating current density, A fc is the cross-sectional area of the fuel cell proton exchange membrane, R ohm is the ohmic internal resistance of the fuel cell, and a, b, c, n are empirical constants;

[0024] ② Use the nonlinear system parameter identification method to online identify the parameters [E oc , a, b, c, n, i0, i n , R ohm , so as to obtain the fuel cell output power-output current P-I curve equation;

[0025] P fc = I fc ·V fc ,

[0026] Assume that when the average current of the fuel cell is I ave , the standard output power is P iave , then the fuel cell degradation rate A:

[0027]

[0028] Let the maximum degradation rate of the fuel cell be ε, and calculate the fuel cell degradation cost weight coefficient β:

[0029]

[0030] ③ Design fuzzy controllers FLC1 and FLC2. The fuzzy controller FLC1 takes the current demand power P req at time k and the power battery SOC as inputs, and outputs the hydrogen consumption cost weight coefficient α; the fuzzy controller FLC2 is based on the change rate of the demand power per unit sampling time and the change rate of the power battery SOC Output the weight coefficient γ of the SOC fluctuation cost as the input.

[0031] Further, the calculation process of step 3 is as follows:

[0032] ① Based on the obtained predicted demand power sequence P pre , perform dynamic programming within the prediction time domain N pre . Select the control variable as the fuel cell output power P fc , and select the state variable as the power battery SOC;

[0033]

[0034] P pre (i) = P bat (i) + P fc (i),

[0035] In the formula, P bat is the battery output power, I bat is the battery output current, U oc is the battery open-circuit voltage, R bat is the battery internal resistance, P pre (i) represents the demand power at the i-th moment within the prediction time domain, i ∈ [k, k + N pre ;

[0036]

[0037] In the formula, ΔT is the prediction time step, and Q bat is the total battery charge;

[0038] When performing reverse inference of dynamic programming, considering the moment of P pre (i), the cost function:

[0039] COSTFCN (i) (m,n) = [α, β, γ] · [J1, J2, J3] Τ ,

[0040] In the formula, m represents the m-th point after discretizing the SOC value into M points, n represents the n-th point after discretizing the fuel cell output power value into N points, [α, β, γ] is the weight coefficient vector, which is fixed before running the energy management method at each sampling moment, and [J1, J2, J3] is the cost function vector;

[0041] ② For the equivalent hydrogen consumption cost J1:

[0042]

[0043] In the formula, η fcFor fuel cell efficiency, LHV is the lower heating value of H2, beq is the equivalent consumption coefficient of the battery, MP is the equivalent hydrogen consumption coefficient, m ave is the hydrogen consumption at the standard output power P iave of the fuel cell;

[0044]

[0045] In the formula, η fcave is the average efficiency of the fuel cell, and mu is the equivalent factor;

[0046] ③ For the fuel cell aging cost J2, consider the fuel cell output power P fc compared with the high output power threshold P fcH and the low output power threshold P fcL for interval division:

[0047]

[0048] ④ For the SOC fluctuation cost J3, consider the numerical fluctuation of SOC(i) at time i and SOC(i + 1) at the next moment:

[0049]

[0050] ⑤ Consider the optimization problem considered in the forward optimization of dynamic programming:

[0051]

[0052] In the formula, SOC k represents the battery SOC value at the current k moment, serving as the initial value SOC0 for forward optimization, and P' fc and P' bat and SOC' respectively represent the fuel cell power, power battery power, and battery SOC change rate per unit time. Based on the SOC k at the k moment, search for the optimal control quantity P fc level by level from the cost function table, as well as the next-level SOC value, to obtain the optimal reference sequence SOC pre under the prediction horizon N ref :

[0053]

[0054] Furthermore, the calculation process of step 4 is as follows:

[0055] ① Establish a state transition equation, with the power battery SOC as the state variable X k , and the output powers P fc and P bat of the fuel cell and the power battery as the control quantities U k ,

[0056] x(k + 1) = A·x(k) + B·u(k),

[0057] u(k) = [P fc , P bat Τ ,

[0058] U k = [u(k|k) Τ , u(k|k + 1) Τ ...u(k|k + N pre ) Τ Τ ,

[0059] X k = [SOC(k|k), SOC(k|k + 1)...SOC(k|k + N pre )] Τ ,

[0060] In the formula, the state transition matrix control matrix E bat is the total energy of the power battery;

[0061] ② Establish the optimization objective function, consider the cumulative deviation between the predicted state vector and the reference value, and the constraint on the output power of the fuel cell as the control variable:

[0062] min J(U k ) = min((X k - R k ) Τ Q(X k - R k ) + U Τ k W U k ),

[0063] s.t. u min < u(k + i|k) < u max , i = 0, 1...N pre ,

[0064] In the formula, R k is the SOC reference sequence SOC ref , Q is the state deviation weight coefficient matrix, and W is the control variable constraint coefficient matrix:

[0065]

[0066] ③ Convert the optimization problem into a quadratic optimization function, and call the linear quadratic programming function to solve the optimal control sequence U k , select the first element p of the control sequence​​fc As the fuel cell reference output, consider the maximum and minimum value ranges of the control variables and the equality constraints:

[0067] U min ≤U k ≤U max ,

[0068] A eq ·X k =B eq ,

[0069] wherein, A eq 、B eq are linear equality constraints, namely:

[0070] P pre (i)=P fc (i)+P bat (i), i = 0, 1... N pre ,

[0071]

[0072] B eq =[1, 2, p3... Npre Τ 。

[0073] On the other hand, the present invention provides a hydrogen - energy UAV energy management system for multi - objective dynamic optimization under an MPC framework, which executes the above - mentioned multi - objective dynamic optimization hydrogen - energy UAV energy management method under an MPC framework, and includes a sensing subsystem, a control subsystem, an interaction subsystem, and a power supply subsystem. Among them,

[0074] the sensing subsystem is used to obtain the fuel cell side voltage, current, working temperature, hydrogen consumption rate, the output current of the fuel cell power converter, and the power battery SOC, current, and bus voltage;

[0075] the control subsystem includes a real - time calculation unit and a communication control unit. The real - time calculation unit runs the above - mentioned multi - objective dynamic optimization hydrogen - energy UAV energy management method based on the data collected by the sensing subsystem, and outputs the fuel cell reference output power. The communication control unit interacts with the fuel cell power converter for commands;

[0076] ​The control subsystem includes an upper - layer energy management controller and a lower - layer power controller. The upper - layer energy management controller calculates and outputs the allocated power according to the data obtained by the sensing subsystem through the energy management method of a hydrogen - powered UAV with multi - objective dynamic optimization under the MPC framework, and transmits it to the lower - layer power controller through internal serial communication. The lower - layer power tracking controller performs feedback control based on the power allocated by the fuel cell and the DCDC output current, so that the output power of the fuel cell is maintained at the expected value;

[0077] The interaction subsystem is used to display the energy state of the hydrogen - lithium hybrid system, display system fault information such as abnormal output of the fuel cell, communication failure, and abnormal battery power in real - time, and allows adjustment of the parameters of the energy management method;

[0078] The power supply subsystem is used to convert the voltage of the electrical energy output by the fuel cell to supply power to the sensing subsystem, the control subsystem, the interaction subsystem, and the fuel - cell power converter.

[0079] Compared with the prior art, the present invention has the following advantages:

[0080] (1) Aiming at the working characteristics of the power system of the hydrogen - lithium hybrid UAV, considering multiple optimization objectives such as energy consumption, battery degradation, and output fluctuation, the present invention proposes a multi - objective dynamic optimization method, which can flexibly self - adjust the optimization objectives during operation and improve the optimization efficiency of the energy management method;

[0081] (2) Based on the predictive control idea, the present invention proposes an energy management method for a hydrogen - powered UAV with multi - objective dynamic optimization under the MPC framework. Compared with the existing MPC - based energy management methods, the present application uses a dynamic programming method with self - adjusting predictive control time domain to obtain the reference optimal state variable trajectory curve, which is an improvement to the prediction and optimization links in MPC control, improving the accuracy and timeliness of the fuel - cell power output;

[0082] (3) The management system proposed by the present invention embodies the hierarchical control idea. For the upper - layer control module, it undertakes two tasks: energy - state assessment and power distribution. It collects the state parameters on the energy output side during operation in real - time, takes them into account in the optimization objectives, and outputs the reference power of the fuel cell. The lower - layer power tracking module controls the power converter to accurately control the current output by the fuel cell to the bus, improving the control real - time performance and the stability of the output electrical energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 It is the flow chart of the management method of the present invention;

[0084] Figure 2 It is the multi - objective dynamic optimization method diagram under the MPC framework of the present invention;

[0085] Figure 3It is the three-dimensional diagram of the fuzzy control of the weight coefficient in the energy management method of the embodiment of the present invention;

[0086] Figure 4 It is the schematic diagram of the management system framework of the present invention;

[0087] Figure 5 It is the simulation diagram of the management system of the present invention. Specific embodiments

[0088] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0089] The hydrogen energy unmanned aerial vehicle involved in the present invention uses a composite energy source mainly based on a hydrogen fuel cell and supplemented by a lithium battery as power. The hydrogen fuel cell is connected to a unidirectional power converter, and the power battery is connected to a battery management system and incorporated into the DC bus.

[0090] Embodiment 1

[0091] As Figure 1 and Figure 2 shown, a multi-objective dynamic optimization energy management method for a hydrogen energy unmanned aerial vehicle under an MPC framework includes the following steps:

[0092] (1) Step 1: Adjust the prediction control time domain according to the historical demand power sequence.

[0093] Based on the bus voltage V at the current sampling moment k bus , the DCDC converter current I DCDC , and the lithium battery current I bat , the demand power p at moment k k is obtained:

[0094] p k = V bus ·(I DCDC + I bat ),

[0095] Update the sliding window sequence P of the historical load demand power with a length of N load = [p1, p2, p3... p k 1×N , calculate the standard deviation σ of the demand power sequence pload (k) and compare it with the historical maximum fluctuation standard deviation σ max and update σ max , where is the mean value of the historical load demand power sliding window sequence:

[0096] ​

[0097] Calculate the predictive control time domain N at time k pre , where the maximum predictive control time domain

[0098]

[0099] According to the obtained historical demand power vector P load and the predictive control time domain N at time k pre , substitute them into the first-order single-variable grey system model GM(1,1), with P load as the historical sample input and N pre being the number of data predicted by the GM(1,1) model system. Set the upper and lower limits of the predicted demand power, and update to obtain the predicted load demand power vector P pre with a length of N pre :

[0100]

[0101] (2) Step 2: Update the aging parameters of the hydrogen fuel cell and fix the weight coefficient of the cost function.

[0102] Based on the semi-empirical voltage output model of the fuel cell considering the leakage current and concentration polarization experience formula, collect the fuel cell output voltage V fc (k), the fuel cell output current I fc (k), and the fuel cell operating temperature T

[0103]

[0104] where N cell is the number of fuel cell stack cells, E oc is the open circuit voltage of the fuel cell, T is the fuel cell operating temperature, i0 is the exchange current density, i n is the leakage current density, i is the fuel cell operating current density, A fc is the cross-sectional area of the fuel cell proton exchange membrane, R ohm is the ohmic internal resistance of the fuel cell, and a, b, c, n are empirical constants.

[0105] In this embodiment, the nonlinear least squares parameter identification method with a forgetting factor is used to perform online identification to obtain the parameters [E oc , a, b, c, n, i0, i n , R ohm , and substitute them into the semi-empirical voltage output model of the fuel cell considering the leakage current and concentration polarization experience formula to update the fuel cell output power-output current P-I curve function:

[0106] P fc = Ifc ·V fc ,

[0107] Let the average current of the fuel cell be I ave At this time, the standard output power is P iave , then the fuel cell degradation rate A is:

[0108]

[0109] Let the maximum degradation rate of the fuel cell be ε, and calculate the fuel cell degradation cost weight coefficient β:

[0110]

[0111] As Figure 3 shown, design the fuzzy controller FLC1 and the fuzzy controller FLC2. The fuzzy controller FLC1 takes the required power P req at the current k moment and the power battery SOC as inputs, and outputs the hydrogen consumption cost weight coefficient α; the fuzzy controller FLC2 takes the change rate of the required power per unit sampling time and the change rate of the power battery SOC as inputs, and outputs the SOC fluctuation cost weight coefficient γ.

[0112] (3) Step 3: Obtain the optimal SOC change curve in the prediction time domain through the multi-objective online optimization algorithm.

[0113] ① Based on the obtained predicted demand power sequence P pre , perform dynamic programming within the prediction time domain N pre . The control variable is selected as the fuel cell output power P fc , and the state variable is selected as the power battery SOC;

[0114]

[0115] P pre (i) = P bat (i) + P fc (i),

[0116] In the formula, P bat is the battery output power, I bat is the battery output current, U oc is the battery open-circuit voltage, R bat is the battery internal resistance, P pre (i) represents the required power at the i-th moment within the prediction time domain, i ∈ [k, k + N pre ;[[ID= / / ]] [[ID= / / ]]

[0117] [[ID= / / ]] [[ID= / / ]] [[ID= / / ]]

[0118] where ΔT is the predicted time step and Q bat is the total battery power.

[0119] When performing backward inverse dynamic programming, consider P pre at time (i), the cost function:

[0120] COSTFCN (i) (m,n) = [α, β, γ] · [J1, J2, J3] Τ ,

[0121] where m represents the m-th point after discretizing the SOC value into M points, n represents the n-th point after discretizing the fuel cell output power value into N points, [α, β, γ] is the weight coefficient vector, which is fixed before running the energy management method at each sampling moment, and [J1, J2, J3] is the cost function vector.

[0122] For the equivalent hydrogen consumption cost J1:

[0123]

[0124] [[ID=Z7]]where η fc is the fuel cell efficiency, LHV is the lower heating value of H2, beq is the equivalent consumption coefficient of the battery, MP is the equivalent hydrogen consumption coefficient, and m ave is the hydrogen consumption when the fuel cell standard output power is P iave ;

[0125]

[0126] where η fcave is the average fuel cell efficiency and mu is the equivalent factor.

[0127] For the fuel cell aging cost J2, consider the fuel cell output power P fc compared with the high output power threshold P fcH and the low output power threshold P fcL to perform interval division, where P fcmax and P fcmin represent the maximum and minimum values of the fuel cell output:

[0128] [[ID=5Y]]

[0129] For the SOC fluctuation cost J3, consider the numerical fluctuation of SOC(i) at time i and SOC(i + 1) at the next time:

[0130]

[0131] Consider the optimization problem considered in the forward optimization of dynamic programming:

[0132]

[0133] In the formula, SOC k represents the battery SOC value at the current k-th moment, which is used as the initial value SOC0 for forward optimization. P' fc , P' bat , and SOC' respectively represent the fuel cell power, power battery power, and battery SOC change rate per unit time. Based on the SOC at the k-th moment k , the optimal control quantity P fc , and the next-level SOC value are searched step by step from the cost function table to obtain the optimal reference sequence SOC pre under the prediction horizon N ref :

[0134]

[0135] (4) Step 4: According to the reference change curve, the reference power of the fuel cell is output through linear MPC control.

[0136] ① Establish the state transition equation, with the power battery SOC as the state variable X k , and the output powers P fc , P bat as the control quantity U k ,

[0137] x(k + 1) = A·x(k) + B·u(k),

[0138] u(k) = [P fc , P bat Τ ,

[0139] U k = [u(k|k) Τ , u(k|k + 1) Τ ...u(k|k + N pre ) Τ Τ ,

[0140] X k = [SOC(k|k), SOC(k|k + 1)...SOC(k|k + N pre )] Τ ,

[0141] In the formula, the state transition matrix the control matrix E bat is the total energy of the power battery.

[0142] ​​Establish an optimization objective function, considering the cumulative deviation between the predicted state vector and the reference value, and the constraint on the output power of the fuel cell as the control variable:

[0143] minJ(U k )=min((X k -R k ) Τ Q(X k -R k )+U Τ k WU k ),

[0144] s.t.u min <u(k+i|k)<u max ,i=0,1...N pre ,

[0145] where, R k is the SOC reference sequence SOC ref , Q is the state deviation weight coefficient matrix, and W is the control variable constraint coefficient matrix:

[0146]

[0147] Transform the optimization problem into a quadratic optimization function, and call the linear quadratic programming function to solve the optimal control sequence U k , select the first element p fc of the control sequence as the reference output of the fuel cell, considering the maximum and minimum value intervals of the control variable and the equality constraint:

[0148] U min ≤U k ≤U max ,

[0149] A eq ·X k =B eq ,

[0150] where, A eq and B eq are the linear equality constraints, that is:

[0151] P pre (i)=P fc (i)+P bat (i), i=0,1...N pre ,

[0152]

[0153] B eq =[1,2,p3... Npre Τ .​

[0154] The fuel cell power converter adopts a constant power output. The converter control module receives the output command p of the upper-layer energy management controller through the internal serial communication line fc , adjusts the power value output to the bus, and stabilizes it through current loop control.

[0155] Embodiment 2

[0156] As Figure 4 shown, a hydrogen energy UAV energy management system for multi-objective dynamic optimization under the MPC framework includes a sensing subsystem, a control subsystem, an interaction subsystem, and a power supply subsystem.

[0157] Specifically, the sensing subsystem includes voltage and current sensors. The external CAN communication line obtains data from the fuel cell controller and the power battery management system, and collects the voltage, current, working temperature of the fuel cell side, the output current of the DCDC module, the SOC of the power battery, the current, and the bus voltage at the load end. The control subsystem includes an upper-layer energy management controller and a lower-layer power controller. The upper-layer energy management controller calculates and outputs the allocated power according to the data obtained by the sensing subsystem through the multi-objective dynamic optimization method of the hydrogen energy UAV energy management method under the MPC framework, and transmits it to the lower layer through internal serial communication. The lower-layer power tracking controller performs feedback control according to the fuel cell allocated power and the DCDC output current to keep the fuel cell output power at the expected value. The interaction subsystem includes an information display and an alarm, which are used to display the energy state of the hydrogen-lithium hybrid system and real-time display system fault information such as abnormal output of the fuel cell, communication failure, and abnormal battery power of the lithium battery; interactive buttons and a serial communication module are used to send and receive commands to the controller of the control subsystem and adjust the parameters of the energy management method. The power supply subsystem includes a DC voltage conversion module. One output of the hydrogen fuel cell is used as the input of the power supply subsystem, and is converted into three outputs with different voltage levels to supply power to the sensing subsystem, the control subsystem, and the interaction subsystem.

[0158] As Figure 5 shown is the simulation diagram of the energy management system. Under the simulated real flight environment of the load, the demand power changes in different working conditions such as takeoff, high and low speed cruising, climbing and falling during speed change, and landing. The energy management system dynamically adjusts the optimization strategy and distributes the output power of the hydrogen-lithium hybrid system based on the real-time collected state parameters of the hydrogen fuel cell and the lithium battery and historical load data.

[0159] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A hydrogen energy UAV energy management method for multi-objective dynamic optimization under the MPC framework, characterized in that, It includes the following steps: Step 1: Adjust the prediction control time domain according to the historical demand power sequence; Step 2: Update the aging parameters of the hydrogen fuel cell and fix the weight coefficient of the cost function; Step 3: Obtain the optimal SOC change curve in the prediction time domain through a multi-objective online optimization algorithm; Step 4: According to the reference change curve, output the reference power of the fuel cell through linear MPC control.

2. The energy management method for a hydrogen - powered unmanned aerial vehicle with multi - objective dynamic optimization under the MPC framework according to claim 1, characterized in that, In the said Step 1, the specific calculation process is as follows: ①Based on the bus voltage V at the current sampling moment k bus , the DCDC converter current I DCDC , and the lithium battery current I bat , the required power p at moment k is obtained k : p k = V bus · (I DCDC + I bat ), ② Update the historical load demand power sliding window sequence \(P\) with length \(N\): load = [p1, p2, p3... p k 1×N , calculate the standard deviation \(\sigma\) of the demand power sequence pload (k) and compare it with the historical maximum fluctuation standard deviation \(\sigma\) max and update \(\sigma\) max , where is the mean of the historical load demand power sliding window sequence:​ ③ Calculate the predictive control horizon N at time k pre , where the maximum predictive control horizon ④According to the obtained historical demand power vector P load , the predictive control time domain N at time k pre , substitute them into the first-order single-variable grey system model GM(1,1), with P load as the historical sample input and N pre being the number of data predicted by the GM(1,1) model system. Set the upper and lower limits of the predicted demand power, and update to obtain the predicted load demand power vector P pre with a length of N pre :

3. A hydrogen energy UAV energy management method for multi-objective dynamic optimization under the MPC framework according to claim 1, characterized in that, The calculation process of the said Step 2 is as follows: ①A semi-empirical voltage output model of a fuel cell based on an empirical formula considering leakage current and concentration polarization, collecting the output voltage V of the fuel cell fc (k), the output current I of the fuel cell fc (k), and the working temperature T of the fuel cell Where N cell is the number of fuel cell stack cells, E oc is the open circuit voltage of the fuel cell, T is the operating temperature of the fuel cell, i0 is the exchange current density, i n is the leakage current density, i is the operating current density of the fuel cell, A fc is the cross-sectional area of the fuel cell proton exchange membrane, R ohm is the ohmic internal resistance of the fuel cell, and a, b, c, n are empirical constants; ②The parameters [E oc , a, b, c, n, i0, i n , R ohm are obtained by online identification using the nonlinear system parameter identification method, so as to obtain the fuel cell output power-output current P-I curve equation; P fc = I fc · V fc , Let the average current of the fuel cell be I ave When the standard output power is P iave , the degradation rate A of the fuel cell is as follows: Let the maximum degradation rate of the fuel cell be ε, and calculate the weight coefficient β of the fuel cell degradation cost: ③Design fuzzy controllers FLC1 and FLC2. Fuzzy controller FLC1 takes the required power P at the current k-th moment req and the state of charge (SOC) of the power battery as inputs, and outputs the weight coefficient α of the hydrogen consumption cost; Fuzzy controller FLC2 takes the change rate of the required power per unit sampling time and the change rate of the SOC of the power battery as inputs, and outputs the weight coefficient γ of the SOC fluctuation cost.

4. A hydrogen energy drone energy management method for multi-objective dynamic optimization under the MPC framework according to claim 1, characterized in that The calculation process of the said Step 3 is as follows: ①Based on the obtained predicted demand power sequence P pre , dynamic programming is carried out within the prediction time domain N pre . The control variable is selected as the fuel cell output power P fc , and the state variable is selected as the power battery SOC; P pre (i) = P bat (i) + P fc (i), Wherein, P bat is the battery output power, I bat is the battery output current, U oc is the battery open-circuit voltage, R bat is the battery internal resistance, P pre (i) represents the required power at the i-th moment within the prediction horizon, i ∈ [k, k + N pre ; where ΔT is the predicted time step and Q bat is the total battery charge; When performing reverse inference in dynamic programming, consider P pre (at time (i), the cost function: COSTFCN (i) (m,n) = [α, β, γ] · [J1, J2, J3] Τ , In the formula, m represents the m-th point after discretizing the SOC value into M points, n represents the n-th point after discretizing the fuel cell output power value into N points, [α, β, γ] is the weight coefficient vector, which is fixed before running the energy management method at each sampling moment, and [J1, J2, J3] is the cost function vector; ② For the equivalent hydrogen consumption cost J1: where η fc is the fuel cell efficiency, LHV is the lower heating value of H2, beq is the equivalent consumption coefficient of the cell, MP is the equivalent hydrogen consumption coefficient, and m ave is the hydrogen consumption at the standard output power P iave of the fuel cell; Where, η fcave is the average efficiency of the fuel cell, and mu is the equivalent factor; ③For the fuel cell aging cost J2, consider the fuel cell output power P fc compared with the high output power threshold P fcH , and the low output power threshold P fcL for comparison, and conduct interval division: ④ For the SOC fluctuation cost J3, consider the numerical fluctuation of SOC(i) at the i-th moment and SOC(i + 1) at the next moment: ⑤ Consider the optimization problem considered by the forward optimization of dynamic programming: where SOC k represents the battery SOC value at the current k-th moment, serving as the initial value SOC0 for forward optimization, P' fc and P' bat and SOC' respectively represent the fuel cell power, power battery power, and battery SOC change rate per unit time. Based on the SOC at the k-th moment k , search for the optimal control quantity P fc step by step from the cost function table, as well as the next-level SOC value, to obtain the optimal reference sequence SOC pre under the prediction horizon N ref :

5. A hydrogen energy drone energy management method for multi-objective dynamic optimization under the MPC framework according to claim 1, characterized in that, The calculation process of the said Step 4 is as follows: ① Establish a state transition equation, with the state of charge (SOC) of the power battery as the state variable X k , the output powers P fc , P bat of the fuel cell and the power battery as the control variables U k , x(k + 1) = A·x(k) + B·u(k), u(k) = [P fc , P bat Τ ,​ U k = [u(k|k) Τ , u(k|k + 1) Τ ... u(k|k + N pre ) Τ Τ ,​ X k = [SOC(k|k), SOC(k|k+1)... SOC(k|k+N pre )] Τ , In the formula, the state transition matrix control matrix E bat is the total energy of the power battery; ② Establish an optimization objective function, considering the cumulative deviation between the predicted state vector and the reference value, and the constraint of the output power of the control quantity fuel cell: minJ(U k ) = min((X k -R k ) Τ Q(X k -R k ) + U Τ k WU k ), s.t.u min <u(k + i|k)<u max , i = 0, 1...N pre , where R k is the SOC reference sequence SOC ref , Q is the state deviation weight coefficient matrix, and W is the control quantity constraint coefficient matrix: ③ Transform the optimization problem into a quadratic optimization function and call the linear quadratic programming function to solve the optimal control sequence U k , select the first element p of the control sequence fc as the reference output of the fuel cell, and consider the maximum and minimum value intervals of the control quantity and the equality constraints: U min ≤U k ≤U max , A eq ·X k =B eq , where A eq , B eq are linear equality constraints, i.e.: P pre (i) = P fc (i) + P bat (i), i = 0, 1... N pre , B eq = [1, 2, p3... Npre Τ .​ 6. A hydrogen energy UAV energy management system for multi-objective dynamic optimization under an MPC framework, which executes a method for multi-objective dynamic optimization of a hydrogen energy UAV energy management system under an MPC framework according to any one of claims 1-5, characterized in that, It includes a sensing subsystem, a control subsystem, an interaction subsystem, and a power supply subsystem. Among them, The said sensing subsystem is used to obtain the fuel cell side voltage, current, working temperature, hydrogen consumption rate, the output current of the fuel cell power converter, and the SOC, current, and bus voltage of the power battery; The said control subsystem includes an upper-layer energy management controller and a lower-layer power controller. The upper-layer energy management controller is based on the data obtained by the sensing subsystem, and the lower-layer power tracking controller performs feedback control according to the fuel cell allocated power and the DCDC output current to keep the fuel cell output power at the predicted value; The said interaction subsystem is used to display the energy state of the hydrogen-lithium hybrid system, display system fault information such as abnormal output of the fuel cell, communication failure, and abnormal battery power of the lithium battery in real time, and allow adjustment of the energy management method parameters; The said power supply subsystem is used to convert the voltage of the electric energy output by the fuel cell to supply power to the sensing subsystem, the control subsystem, the interaction subsystem, and the fuel cell power converter.

Images