Self-adaptive planning aviation fuel cell range extending power system evaluation method and self-adaptive planning aviation fuel cell range extending power system evaluation system

Abstract

The invention provides a self-adaptive planning aviation fuel cell extended range power system evaluation method and system, and belongs to the technical field of power system evaluation, and the method comprises the steps: constructing a state space model of an aviation fuel cell extended range power system, building a multi-stage decision model, solving a control strategy of each stage through a self-adaptive dynamic planning algorithm, and calculating the state space model of the aviation fuel cell extended range power system. Performing difference mining on each state difference matrix, and determining an incidence relation graph of state variables and control variables in the corresponding flight stage; acquiring current state data of the aviation fuel cell extended range power system in real time, adjusting optimization parameters of a control strategy in a corresponding stage to obtain an optimal strategy in combination with a state space model and an incidence relation graph, and dynamically adjusting output power of a fuel cell and a super capacitor in combination with a power distribution strategy; and when the power output margin is lower than a safety preset threshold value, an alarm signal is triggered. And the requirements of efficient and safe operation of the aviation fuel cell range extending system are met.

Description

Technical Field

[0001] The present invention relates to the technical field of power system evaluation, and in particular to an adaptive planning aviation fuel cell range-extending power system evaluation method and system. Background Art

[0002] As the aviation industry transitions toward green and low-carbon development, hydrogen fuel cell range-extended propulsion systems have become a core power solution for low-altitude aircraft such as drones and eVTOLs due to their advantages such as zero emissions and high energy density. However, the complex operating conditions of aviation scenarios (such as the instantaneous high power demand during takeoff, the need for efficient endurance during level flight, and the need for energy recovery during landing) pose severe challenges to the dynamic power control, energy storage coordination, and safety redundancy of the power system. Existing technologies have the following key issues:

[0003] Existing aviation fuel cell systems mostly use fixed power allocation strategies, making it difficult to adapt to the dramatic power fluctuations during flight. For example, although a certain company's fuel cell drone achieved a 120-minute flight endurance, its power regulation relied on simple charge and discharge threshold control, without considering the fuel cell stack temperature (T), output voltage (V), supercapacitor state of charge (SOC), and payload (Gp). Moreover, although the supercapacitor had a power density of 25,000W / kg and low-temperature performance of -45°C, the lack of adaptive coordination with the fuel cell led to delayed power response during takeoff and excessive energy efficiency loss during level flight, failing to meet the requirements for efficient and safe operation of aviation fuel cell range-extending systems.

[0004] Therefore, the present invention proposes an adaptive planning aviation fuel cell range-extending power system evaluation method and system. Summary of the Invention

[0005] The present invention provides an adaptive planning aviation fuel cell range-extending power system evaluation method and system to solve the above-mentioned technical problems.

[0006] The present invention provides an adaptive planning aviation fuel cell range-extending power system evaluation method, comprising:

[0007] Constructing a state space model of an aviation fuel cell range-extending power system, wherein the state space model includes: state variables, control variables, and output variables of the system;

[0008] A multi-stage decision model is established to divide the aviation flight process into three stages: takeoff, level flight, and landing. Each stage corresponds to different power requirements and constraints. The control strategy for each stage is solved using an adaptive dynamic programming algorithm. The state transition equation of the adaptive dynamic programming algorithm is:

[0009] s(k+1)=f(s(k),u(k))+w(k)

[0010] Wherein, s(k+1) is the predicted state vector at the k+1th flight time, f() is the state transfer function, s(k) is the state vector at the kth flight time, u(k) is the control vector at the kth flight time, and w(k) is the noise vector of the aviation fuel cell range-extended power system;

[0011] Obtain the actual state vector y(k+1) at the k+1th flight time, and use the state difference vector c(k+1) between the actual state vector y(k+1) and the predicted state vector s(k+1) to construct a state difference matrix for each flight phase;

[0012] Perform difference mining on each state difference matrix to determine the correlation graph between state variables and control variables in the corresponding flight phase;

[0013] Real-time collection of current state data of the aviation fuel cell range-extending power system, and in combination with the state space model and the association relationship map, adjustment of optimization parameters of the control strategy for the corresponding stage determined by the dynamic programming algorithm based on the current state data, to obtain the optimal strategy for the corresponding stage, and dynamic adjustment of the output power of the fuel cell and the ultracapacitor in combination with the power allocation strategy;

[0014] The power output margin of the aviation fuel cell range-extending power system is dynamically adjusted according to the current state data, and an alarm signal is triggered when the power output margin is lower than a safety preset threshold.

[0015] Preferably, constructing a state space model of an aviation fuel cell range-extending power system includes:

[0016] Extracting a phase vector set of the fuel cell from the start of takeoff to the end of landing from a historical database, wherein the phase vector set includes a phase vector of each historical flight moment under different flight phases;

[0017] Determine the quantitative relationship between the state variables and the output variables in the stage vector:

[0018] When the number of the state variables is less than the number of the output variables, determining the input controllable quantity that affects the control variable and the output controllable quantity that has a potential effect on the output variable;

[0019] Mapping the input controllable quantities and the output controllable quantities into a variable space coordinate system, first retaining controllable quantities in the output controllable quantities that have a unidirectional dependency relationship with the output controllable quantities, and performing state analysis on the first retained controllable quantities until a first number of retained input controllable quantities is greater than a second number of retained output controllable quantities, and a sum of the number of state variables and the first number is greater than the sum of the number of output variables and the second number;

[0020] Among them, status analysis includes:

[0021] Extracting the first amplification vector at the first flight moment and the second amplification vector at the next flight moment in each flight phase, determining the effective amplification coefficient of each amplification variable, and obtaining the coefficient vector of the corresponding amplification variable in the flight phase;

[0022] When the coefficient vector tends to a stable state, it is preferentially retained;

[0023] Otherwise, keep the original phase vector;

[0024] A state space model is constructed based on the retained vector and combined with the least squares method.

[0025] Preferably, the multi-objective function of the power allocation strategy is:

[0026]

[0027] Among them, w1, w2, w3, and w4 represent weight coefficients, which are dynamically adjusted according to the flight stage, equipment health status, and load; Preq represents the required power during the flight stage; Topt represents the optimal operating temperature of the fuel cell in the corresponding stage; H 安全阈值 represents the hydrogen safety threshold of the corresponding stage; Prated represents the rated power of the fuel cell; Trange represents the allowable temperature; Hreated represents the total hydrogen storage; Is represents the instantaneous discharge current of the supercapacitor; Ir represents the rated current of the supercapacitor, minJ represents the multi-objective function; k represents the slope of the Sigmoid function, with a value of 10 to 20; M represents the membrane dryness index; ∝ represents the wetness of the proton exchange membrane of the fuel cell, with a value of 0 to 1; T represents the temperature of the fuel cell stack; H represents the remaining hydrogen amount; Pfc represents the actual output power of the fuel cell; Pcap represents the actual output power of the supercapacitor.

[0028] Preferably, the aviation flight process is divided into three stages: takeoff, level flight, and landing, including:

[0029] Based on the improved K-means clustering and altitude change rate cross-validation method, the power demand slope is extracted and the flight phases are intelligently divided into three phases: takeoff, level flight, and landing.

[0030] Among them, the energy is provided by super capacitor during the take-off phase;

[0031] The fuel cell provides energy during the level flight phase, and the excess energy is used to charge the super capacitor.

[0032] The landing phase is powered by both supercapacitors and fuel cells.

[0033] Preferably, the hard constraint condition for the output power of the super capacitor is:

[0034]

[0035] Among them, SOC represents the percentage of charge of the super capacitor; Gmax is the maximum load threshold; and Gp is ​​the current load.

[0036] Preferably, dynamically adjusting the power output margin of the aviation fuel cell range-extending power system according to the current state data includes:

[0037] Determining, based on the correlation relationship map corresponding to the flight phase, a correction factor f(T, Gp) of the rated output power of the fuel cell in the current state data based on temperature and load;

[0038] Correct Pfc+Pcap-Preq based on the correction factor to obtain the power output margin Margin;

[0039] Margin=Prated·(1-0.0005·N1)·f(T,Gp)·+Pcap,e·(1-0.0001·N2)

[0040] -Preq

[0041] Wherein, N1 represents the number of cycles of the fuel cell; N2 represents the number of cycles of the supercapacitor; and Pcap,e represents the rated output power of the supercapacitor.

[0042] Preferably, determining the temperature-based correction factor f(T,Gp) of the rated output power of the fuel cell in the current state data based on the correlation relationship map corresponding to the flight phase includes:

[0043] Based on the correlation map of each flight phase, the key impact characteristics of temperature on the rated output power of the fuel cell are extracted. The key impact characteristics include the temperature sensitivity threshold, load sensitivity threshold, and the coupling coefficient between the power attenuation coefficient and the temperature change rate and load change rate at different stages. The correlation map contains at least the nonlinear correlation path and path influence weight between the temperature variable, load variable and the rated output power of the fuel cell;

[0044] A dynamic weighting factor for each flight phase is introduced to weightedly integrate the key influencing features of each phase and establish a temperature-load-power correction model. The dynamic weighting factor is adaptively adjusted according to the power demand intensity, temperature fluctuation amplitude, and load size of the corresponding phase.

[0045] Real-time collection of temperature data, load data, and rated output power reference value for the current flight phase, inputting the data into the temperature-load-power correction model, and calculating an initial temperature-load correction factor;

[0046] The initial temperature-load correction factor is calibrated online by feedback of the deviation between the actual output power and the corrected rated power until the deviation meets the preset accuracy requirement, and the final correction factor f(T,Gp) is obtained.

[0047] The present invention provides an adaptive planning aviation fuel cell range-extending power system evaluation system, comprising:

[0048] A model building module is used to build a state space model of the aviation fuel cell range-extending power system, wherein the state space model includes: state variables, control variables and output variables of the system;

[0049] The multi-stage decision module is used to establish a multi-stage decision model, dividing the aviation flight process into three stages: takeoff, level flight, and landing. Each stage corresponds to different power requirements and constraints. The control strategy for each stage is solved by an adaptive dynamic programming algorithm. The state transition equation of the adaptive dynamic programming algorithm is:

[0050] s(k+1)=f(s(k),u(k))+w(k)

[0051] Wherein, s(k+1) is the predicted state vector at the k+1th flight time, f() is the state transfer function, s(k) is the state vector at the kth flight time, u(k) is the control vector at the kth flight time, and w(k) is the noise vector of the aviation fuel cell range-extended power system;

[0052] A matrix construction module is used to obtain the actual state vector y(k+1) at the k+1th flight time, and use the state difference vector c(k+1) between the actual state vector y(k+1) and the predicted state vector s(k+1) to construct a state difference matrix for each flight phase;

[0053] The graph determination module is used to perform difference mining on each state difference matrix and determine the correlation graph between the state variables and the control variables in the corresponding flight phase;

[0054] a dynamic adjustment module for collecting current state data of the aviation fuel cell range-extending power system in real time, and adjusting the optimization parameters of the control strategy for the corresponding stage determined by the dynamic programming algorithm based on the current state data in combination with the state space model and the association relationship map, to obtain the optimal strategy for the corresponding stage, and dynamically adjusting the output power of the fuel cell and the ultracapacitor in combination with the power allocation strategy;

[0055] An alarm module is used to dynamically adjust the power output margin of the aviation fuel cell range-extending power system according to the current status data, and trigger an alarm signal when the power output margin is lower than a safety preset threshold.

[0056] Compared with the prior art, the present invention has the following advantages:

[0057] Through the full process of state modeling → multi-stage prediction → error closed loop → correlation mining → strategy optimization → safety guarantee, the static limitations of traditional control are broken through: it not only accurately matches the multi-stage dynamic power requirements of aviation, but also quantifies safety redundancy through health decay and environmental correction, realizes deep synergy between fuel cells and ultra-energy capacitors, reduces energy loss, and meets the needs of efficient and safe operation of aviation fuel cell range extension systems.

[0058] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0059] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0061] Figure 1 This is a flow chart of an adaptive planning aviation fuel cell range-extending power system evaluation method according to an embodiment of the present invention;

[0062] Figure 2 This is a structural diagram of an adaptive planning aviation fuel cell range-extending power system evaluation system according to an embodiment of the present invention;

[0063] Figure 3 This is a diagram of the linkage alarm interaction between the dynamic adjustment module and the alarm module in an embodiment of the present invention. DETAILED DESCRIPTION

[0064] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0065] The present invention provides an adaptive planning aviation fuel cell range extension power system evaluation method, such as Figure 1 As shown, including:

[0066] Step 1: Construct a state space model of the aviation fuel cell range-extending power system, wherein the state space model includes: state variables, control variables, and output variables of the system;

[0067] Step 2: Establish a multi-stage decision model, dividing the aviation flight process into three stages: takeoff, level flight, and landing. Each stage corresponds to different power requirements and constraints. Use an adaptive dynamic programming algorithm to solve the control strategy for each stage. The state transition equation of the adaptive dynamic programming algorithm is:

[0068] s(k+1)=f(s(k),u(k))+w(k)

[0069] Wherein, s(k+1) is the predicted state vector at the k+1th flight time, f() is the state transfer function, s(k) is the state vector at the kth flight time, u(k) is the control vector at the kth flight time, and w(k) is the noise vector of the aviation fuel cell range-extended power system;

[0070] Step 3: Obtain the actual state vector y(k+1) at the k+1th flight time, and use the state difference vector c(k+1) between the actual state vector y(k+1) and the predicted state vector s(k+1) to construct a state difference matrix for each flight phase;

[0071] Step 4: Perform difference mining on each state difference matrix to determine the correlation graph between state variables and control variables in the corresponding flight phase;

[0072] Step 5: Real-time collection of current state data of the aviation fuel cell range-extending power system, and in combination with the state space model and the correlation relationship map, adjustment of optimization parameters of the control strategy for the corresponding stage determined by the dynamic programming algorithm based on the current state data, obtaining the optimal strategy for the corresponding stage, and dynamically adjusting the output power of the fuel cell and the ultracapacitor in combination with the power allocation strategy;

[0073] Step 6: Dynamically adjust the power output margin of the aviation fuel cell range-extending power system according to the current state data, and trigger an alarm signal when the power output margin is lower than a safety preset threshold.

[0074] In this embodiment, the state variables include at least: the temperature T of the fuel cell stack, the output voltage V, the remaining hydrogen amount H, the state of charge SOC of the super capacitor, and the load Gp; the control variables include the actual output power Pfc of the fuel cell, the actual output power Pcap of the super capacitor, and the flow rate Q of the air compressor; the output variables include at least: the total output power Ptotal of the system (positively correlated with the load).

[0075] In this embodiment, the state variables are system intrinsic characteristic parameters (reflecting the operating status), such as the fuel cell stack temperature T (stack thermal state, rising from 25°C to 70°C during takeoff, the greater the load, the faster the temperature rises, monitored by distributed optical fiber sensors (5 cm spacing, accuracy ±0.3°C)), the output voltage V (stack electrochemical characteristics, stable at 650V±5V during level flight, the voltage drops slightly when the load increases, measured by a Hall voltage sensor (0.1 level accuracy)), the remaining hydrogen volume H (based on endurance, H=100% before takeoff, the greater the load, the faster the consumption, landing requires ≥15%, estimated by flow sensor integration + electrochemical model), the supercapacitor state of charge SOC (power level, maintained at 30% to 70% during level flight, the greater the load, the faster the discharge, calculated by current integration + OCV-SOC curve (cubic polynomial fitting)), and the load Gp (real-time cargo / personnel weight, collected by the landing gear pressure sensor).

[0076] Control variables: Actively adjustable inputs, such as the fuel cell's actual output power Pfc (gradually increases from 50% of the rated power at takeoff; the greater the payload, the greater the increase, controlled by hydrogen valve opening + air compressor flow), the supercapacitor's actual output power Pcap (short-term 12C discharge at takeoff; the greater the payload, the higher the discharge rate, controlled by a bidirectional DC / DC converter), and the air compressor flow Q (increased by 20% during takeoff to enrich oxygen; the greater the payload, the greater the flow, achieved by variable frequency compressor + PID control).

[0077] Output variables: System output characteristics, such as the system's total output power Ptotal (Ptotal = Pfc + Pcap, matching the eVTOL lift system requirements and collected in real time by a power analyzer).

[0078] In this embodiment, multiple stages are divided: according to the operating conditions, take-off (power slope <0.05MW / s, altitude change rate <0.5m / s, the greater the load, the higher the power slope), level flight (power slope <0.05MW / s, altitude change rate <0.5m / s, the power is stable when the load is stable), landing (power slope <-0.5MW / s, altitude change rate <-4m / s, the greater the load, the more power is recovered), and the stage boundaries are identified by improved K-means clustering + altitude change rate cross-validation (clustering is implemented by Python's scikit-learn library, and IMU altitude data is integrated).

[0079] In this embodiment, the state difference matrix is ​​classified by flight phase, and the difference vectors of consecutive moments within the phase are stacked (for example, c(k) at the 100 moments before takeoff is composed of a 100×4 matrix, constructed using Python's numpy array). Difference mining: The state difference matrix is ​​analyzed to locate the influence of the control variable on the state (for example, a 10% increase in Pfc and a 5% increase in Gp → a 3°C increase in T, trained using the random forest algorithm (Python's scikit-learn library) to output variable association weights);

[0080] Correlation diagram: Visualize the causal relationship between states (T, SOC, Gp) and control variables (Pfc, Q) (such as the chain association of "Gp↑→Pfc↑→T↑→V↓", modeled using Bayesian networks (Python's pgmpy library) and drawn using the Graphviz tool).

[0081] Current state data: The current state (e.g., T = 60°C, SOC = 50%, Gp = 2000 kg, transmitted 100 times per second via the CAN bus) is collected in real time. Optimization parameter adjustment is based on the correlation map and dynamically corrects the control strategy (e.g., when Gpp is too high during level flight, the optimization weight of Pfc is increased to reduce capacitor discharge). The adaptive dynamic programming (ADP) controller (Actor-Critic architecture, TensorFlow training) updates the policy network parameters in real time.

[0082] The power allocation strategy is implemented in stages (takeoff: Level flight: extreme value search optimizes efficiency, the greater the payload, the higher the proportion of fuel cell output). The stage logic is implemented through embedded C++ code, and power commands are sent via the CAN bus.

[0083] In this embodiment, the power output margin is calculated through the stage correlation map and online calibration. The greater the load, the higher the power demand, and the margin calculation needs to be corrected synchronously.

[0084] The beneficial effects of the above technical solution are: through the full process of state modeling → multi-stage prediction → error closed loop → correlation mining → strategy optimization → safety guarantee, it breaks through the static limitations of traditional control: it not only accurately matches the multi-stage dynamic power requirements of aviation, but also quantifies safety redundancy through health decay and environmental correction, realizes deep coordination between fuel cells and ultra-energy capacitors, reduces energy loss, and meets the needs of efficient and safe operation of aviation fuel cell range extension systems.

[0085] The present invention provides an adaptive planning aviation fuel cell range-extending power system evaluation method, which constructs a state space model of the aviation fuel cell range-extending power system, including:

[0086] Extracting a phase vector set of the fuel cell from the start of takeoff to the end of landing from a historical database, wherein the phase vector set includes a phase vector of each historical flight moment under different flight phases;

[0087] Determine the quantitative relationship between the state variables and the output variables in the stage vector:

[0088] When the number of the state variables is less than the number of the output variables, determining the input controllable quantity that affects the control variable and the output controllable quantity that has a potential effect on the output variable;

[0089] Mapping the input controllable quantities and the output controllable quantities into a variable space coordinate system, first retaining controllable quantities in the output controllable quantities that have a unidirectional dependency relationship with the output controllable quantities, and performing state analysis on the first retained controllable quantities until a first number of retained input controllable quantities is greater than a second number of retained output controllable quantities, and a sum of the number of state variables and the first number is greater than the sum of the number of output variables and the second number;

[0090] Among them, status analysis includes:

[0091] Extracting the first amplification vector at the first flight moment and the second amplification vector at the next flight moment in each flight phase, determining the effective amplification coefficient of each amplification variable, and obtaining the coefficient vector of the corresponding amplification variable in the flight phase;

[0092] When the coefficient vector tends to a stable state, it is preferentially retained;

[0093] Otherwise, keep the original phase vector;

[0094] A state space model is constructed based on the retained vector and combined with the least squares method.

[0095] In this embodiment, in the aviation scenario, the state variables (T, V, H, SOC, Gp) are often less than the output variables (such as Ptotal, load voltage, fuel cell stack heat flow, hydrogen consumption rate, supercapacitor charge and discharge current, temperature, etc.), resulting in the traditional state space model being "underdetermined" (number of equations < unknown number). By introducing the rate of change through the amplification vector, the state dimension is artificially increased (such as from 5 dimensions to 9 dimensions), so that the number of states + the number of input controllables > the number of outputs + the number of output controllables, satisfying the model solvability (rank condition) and avoiding the dilemma of "no solution" or "multiple solutions".

[0096] Static state-space models (using only the current values ​​of variables, excluding payload) cannot describe the dynamic power changes and payload impacts in aviation. For example, the change in T during takeoff is not only related to the current Pfc, but also strongly correlated with the rate of change of Pfc (heating rate) and the rate of change of Gp. Traditional models underestimate thermal inertia, leading to inaccurate control.

[0097] If you skip "amplification → stability verification" and directly use the original data for fitting: either the model will "fail" due to insufficient dimensions, or "false correlation" will be introduced due to not considering the load, causing the control strategy to oscillate or even fail in actual flight.

[0098] In this embodiment, the historical database stores the full data of past flight missions (including fuel cell temperature, voltage, power, flight altitude, speed, etc. under different loads), which can be accumulated through an onboard data recorder (such as a black box) or a ground simulation platform. For example, 100 flight data of a certain type of eVTOL, including take-off (0-30s), level flight (30-240s), and landing (240-300s) stages.

[0099] The phase vector is a combination of state variables at each moment of flight (for example, at 5 seconds after takeoff, the vector is [T = 35°C, V = 620V, H = 95%, SOC = 85%, Gp = 2000kg, Pfc = 50kW, Pcap = 80kW, Q = 2kg / s]). It covers state (T, V, H, SOC, Gp), control (Pfc, Pcap, Q), and output (Ptotal) variables.

[0100] State variables describe the system's intrinsic characteristics (e.g., T, V, H, SOC, Gp, a total of five). Output variables represent external characteristics (e.g., Ptotal, load voltage, etc.). Here, it is assumed that the number of states exceeds the number of outputs; actual adjustments must be made based on the data scenario. If the number of states is less than the number of outputs, the controllable input variables (e.g., the set values ​​of Q and Pfc, including load compensation) and the controllable output variables (e.g., the proportion of Ptotal contributed by Pfc, which is affected by load) must be decomposed. Specifically, matrix rank analysis is used to determine the variable dimensions, and a cause-and-effect diagram is used to clarify the control-output-load relationship.

[0101] In this embodiment, the variable space coordinate system is to set the input controllable quantities (such as Q and Pfc) as the x and y axes, and the output controllable quantity (such as the Pfc contribution value of Ptotal) as the z axis, to construct a three-dimensional space and visualize the variable association. One-way dependency relationship: If "Gp increases → Preq must increase" (one-way), but "Preq increases → Gp does not necessarily increase", then the controllable quantity of the one-way relationship is retained. Through iterative screening, until the number of input controllable quantities (such as 3: the adjustment amount of Q, Pfc, and Pcap) > the number of output controllable quantities (such as 2: the two-part contribution of Ptotal), and the number of states + input controllable quantity > the number of outputs + output controllable quantity (meeting dimensional solvability), specifically: use a directed acyclic graph (DAG) to model the dependency relationship (such as Python's networkx library), delete the bidirectional edges, retain the unidirectional edges, and count the number of nodes.

[0102] In this embodiment, the amplified vector is obtained by adding the "rate of change" dimension to the basic vector ([T, V, Pfc, Gp]) at the first moment in the stage (e.g., the first second after takeoff) to form the first amplified vector ([T, V, Pfc, Gp, dT / dt, dV / dt, dPfc / dt, dGp / dt]); the second amplified vector is obtained in the same way at the next moment (the second second).

[0103] The effective amplification factor is the ratio of the change in the calculated variable (e.g. It represents the change rate of T for every 100kg increase in Gp. If the standard deviation of the coefficient vector is calculated for 10 consecutive moments and the standard deviation is less than 1% (stable), the expansion mode is retained; otherwise, the original stage vector is rolled back.

[0104] In this embodiment, the retained vectors are selected stable amplification vectors (e.g., the 200 valid time vectors during the takeoff phase), and the state transition relationship is fitted using the least squares method. For example, for T(k+1), the fitting formula is T(k+1) = a·T(k) + b·Pfc(k) + c·Gp(k) + ∈. A, b, and c are solved by minimizing the residual sum of squares (e.g., Matlab's lsqfit function), where ∈ is an error factor with a value of 0.01. Specifically, the retained vectors are arranged in time series, the input matrix (T(k), Pfc(k), Gp(k)) and the output vector (T(k+1)) are constructed, and the numerical fitting tool is used to solve the model parameters.

[0105] The beneficial effects of the above technical solution are: through the whole process of historical data mining → variable dimension adaptation → dependency screening → dynamic stability verification → mathematical model fitting, it breaks through the static assumptions of the traditional state-space model, adapts to the multi-stage dynamic characteristics of aviation flight (differentiated variable correlations of take-off / level flight / landing), and captures the time-varying coupling laws of variables such as power, temperature, and load through amplification vectors and stability analysis. The final constructed model can accurately describe the nonlinear interaction of fuel cell stack-ultracapacitor-load-environment, providing high-precision state prediction and decision-making basis for subsequent adaptive control.

[0106] The present invention provides an adaptive planning aviation fuel cell range-extending power system evaluation method, wherein the multi-objective function of the power allocation strategy is:

[0107]

[0108] Among them, w1, w2, w3, and w4 represent weight coefficients, which are dynamically adjusted according to the flight stage, equipment health status, and load; Preq represents the required power during the flight stage; Topt represents the optimal operating temperature of the fuel cell in the corresponding stage; H 安全阈值represents the hydrogen safety threshold of the corresponding stage; Prated represents the rated power of the fuel cell; Trange represents the allowable temperature; Hreated represents the total hydrogen storage; Is represents the instantaneous discharge current of the supercapacitor; Ir represents the rated current of the supercapacitor, minJ represents the multi-objective function; k represents the slope of the Sigmoid function, with a value of 10 to 20; M represents the membrane dryness index; ∝ represents the wetness of the proton exchange membrane of the fuel cell, with a value of 0 to 1; T represents the temperature of the fuel cell stack; H represents the remaining hydrogen amount; Pfc represents the actual output power of the fuel cell; Pcap represents the actual output power of the supercapacitor.

[0109] In this embodiment, represents the membrane dryness penalty term, The life loss of ultracapacitors due to high-rate discharge is quantified, and the weights are adjusted in real time according to the flight stage (power response is prioritized during takeoff, and life protection is prioritized during level flight) and the health status of the equipment (α is increased when the membrane is dry, and membrane protection is strengthened). This allows the optimization strategy to automatically adapt to the changing operating conditions from power impact to steady-state endurance to safe landing, avoiding the loss of one objective while focusing on another in single-objective optimization.

[0110] In this embodiment, the membrane dryness penalty item and the capacitor life loss item are both affected by the load (the greater the load, the higher the membrane dryness risk and the greater the capacitor discharge current), and the weight is dynamically adjusted with the flight stage and load (power response is prioritized during heavy takeoff and life protection is prioritized during light flight).

[0111] If only power matching is optimized (such as Pfc+Pcap-Preq), risks such as excessive temperature (damaging the fuel cell stack) due to increased load, hydrogen depletion (endangering safety), and accelerated capacitor over-discharge (sudden reduction in life) will be ignored. The multi-objective function, through mathematical coupling, forces the controller to seek the best solution between conflicting objectives, thus filling the gap between engineering safety requirements and single algorithm optimization.

[0112] In this embodiment, Preq includes load-related items, such as, Where Preq(0) is the standard load power requirement of the basic load; KG is the load factor.

[0113] The beneficial effects of the above technical solution are: through the normalized multi-objective function integration of power matching accuracy, temperature stability, hydrogen safety redundancy, ultra-capacitor life protection, and membrane dry state warning, combined with dynamic weight adaptation of flight phase and equipment health, the performance-life-safety coordinated optimization of the aviation fuel cell range extender system under high dynamic conditions is achieved, supporting the efficient and reliable operation of the system.

[0114] The present invention provides an adaptive planning aviation fuel cell range-extending power system evaluation method, which divides the aviation flight process into three stages: takeoff, level flight, and landing, including:

[0115] Based on the improved K-means clustering and altitude change rate cross-validation method, the power demand slope is extracted and the flight phases are intelligently divided into three phases: takeoff, level flight, and landing.

[0116] Among them, the energy is provided by super capacitor during the take-off phase;

[0117] The fuel cell provides energy during the level flight phase, and the excess energy is used to charge the super capacitor.

[0118] The landing phase is powered by both supercapacitors and fuel cells.

[0119] Preferably, the hard constraint condition for the output power of the super capacitor is:

[0120]

[0121] Among them, SOC represents the percentage of charge of the super capacitor; Gmax is the maximum load threshold; and Gp is ​​the current load.

[0122] In this embodiment, the K-means clustering is improved: Traditional K-means: clustering by data distance, the improvement is to integrate the three features of power demand slope, altitude change rate, and load change rate, and assign different weights (power slope weight 0.4, altitude change rate weight 0.3, load change rate 0.3, adapted to aviation conditions). For example, in the take-off phase, the power demand slope is >0.8MW / s (motor rapid acceleration), and the altitude change rate is >6m / s (vertical climb); in the level flight phase, the power slope is <0.05MW / s (smooth cruising), altitude change rate <0.5m / s (level flight), specifically: Python calls scikit-learn.KMeans, customizes the distance function (weighted Euclidean distance: d = 0.4|k1-k2|��+0.3|h1-h2|��+0.3|g01-g02|, k1 and k2 are power slopes, h1 and k2 are altitude change rates, g01 and g02 are payload change rates), and inputs historical flight data (power, altitude, and payload time series) to train the model.

[0123] The altitude change rate is the rate of change of altitude over time (Δh / Δt), which is calculated in real time by the IMU inertial sensor (such as MPU6050, sampling rate 100Hz). The load change rate is the rate of change of load over time (Δg / Δt) (negative when cargo is dropped), which is calculated by the pressure sensor.

[0124] Cross-validation involves clustering and determining the suspected takeoff segment. The conditions for a power slope > 0.8 MW / s, altitude change rate > 6 m / s, and payload change rate ≤ 0 (no takeoff) must all be met. Otherwise, the segment is considered "level flight turbulence." Pcap (rated power) is the theoretical power of the pre-set super capacitor when discharged.

[0125] In this embodiment, 1.2 times the rated power (12C short-time pulse) is allowed, and it is reduced to 1.0 times when the load exceeds the limit, balancing the power demand and the capacitor life (for example, eVTOL vertical takeoff requires 2 times the rated power, the capacitor bears 1.2 times, and the fuel cell supplements 0.8 times).

[0126] The logic of SOC<0.3 (landing phase constraint) is: when the battery power is <30%, it is limited to 0.8 times the rated power (8C) to avoid deep discharge (the capacity decays by 10% after 1000 cycles at 10C, and the cycle life is extended by 50% at 8C). The heavier the load, the stricter the restriction. For example, energy recovery (capacitor charging) is required during landing. If the SOC is already <30%, the discharge power is forced to be reduced and energy replenishment is prioritized.

[0127] The logic behind SOC∈[0.3,0.7] (constraint during level flight) is as follows: When the battery level is moderate, adaptive dynamic programming (ADP) is allowed to freely optimize power distribution, balancing "fuel cell efficiency" and "capacitor life" (prioritizing level flight endurance to avoid frequent high-rate charge and discharge). For example, the ADP controller (deployed in TensorFlow) dynamically outputs Pcap based on the current SOC and power demand, within the rated power range (implicit constraint: |Pcap|≤Pcap,rated). The heavier the payload, the greater the reliance on fuel cell output.

[0128] Pcap, rated, is the rated power of the super capacitor. When the load exceeds the limit, the discharge rate is reduced to avoid overload damage.

[0129] The beneficial effects of the above technical solution are: accurate division of flight stages through dual-feature clustering + cross-validation (solving misjudgment of level flight turbulence and missed judgment of takeoff impact), combined with the hard constraints of super-energy capacitor stage + SOC range + load (appropriate derating when taking off with heavy load, releasing potential when light load), which not only meets the power requirements of multiple stages and load changes in aviation, but also avoids the risk of excessive damage to energy storage components.

[0130] The present invention provides an adaptive planning aviation fuel cell range-extending power system evaluation method, which dynamically adjusts the power output margin of the aviation fuel cell range-extending power system according to the current state data, including:

[0131] Determining, based on the correlation relationship map corresponding to the flight phase, a correction factor f(T, Gp) of the rated output power of the fuel cell in the current state data based on temperature and load;

[0132] Correct Pfc+Pcap-Preq based on the correction factor to obtain the power output margin Margin;

[0133] Margin=Prated·(1-0.0005·N1)·f(T,Gp)+Pcap,e·(1-0.0001·N2)

[0134] -Preq

[0135] Wherein, N1 represents the number of cycles of the fuel cell; N2 represents the number of cycles of the supercapacitor; and Pcap,e represents the rated output power of the supercapacitor.

[0136] Preferably, determining the temperature-based correction factor f(T,Gp) of the rated output power of the fuel cell in the current state data based on the correlation relationship map corresponding to the flight phase includes:

[0137] Based on the correlation map of each flight phase, the key impact characteristics of temperature on the rated output power of the fuel cell are extracted. The key impact characteristics include the temperature sensitivity threshold, load sensitivity threshold, and the coupling coefficient between the power attenuation coefficient and the temperature change rate and load change rate at different stages. The correlation map contains at least the nonlinear correlation path and path influence weight between the temperature variable, load variable and the rated output power of the fuel cell;

[0138] A dynamic weighting factor for each flight phase is introduced to weightedly integrate the key influencing features of each phase and establish a temperature-load-power correction model. The dynamic weighting factor is adaptively adjusted according to the power demand intensity, temperature fluctuation amplitude, and load size of the corresponding phase.

[0139] Real-time collection of temperature data, load data, and rated output power reference value for the current flight phase, inputting the data into the temperature-load-power correction model, and calculating an initial temperature-load correction factor;

[0140] The initial temperature-load correction factor is calibrated online by feedback of the deviation between the actual output power and the corrected rated power until the deviation meets the preset accuracy requirement, and the final correction factor f(T,Gp) is obtained.

[0141] In this embodiment, the association relationship map stores the causal relationship between the state (including Gp) within the flight phase and the control variable (such as fuel cell power) (such as the chain path of Gp↑→Preq↑→Pfc↑→T↑→membrane resistance↑→Pfc, rated↓), and quantifies the influence weight of each path through Bayesian network modeling (such as the weight of the Gp→Pfc path in the takeoff phase is 0.6).

[0142] The current status data includes real-time load, stack temperature (distributed fiber optic sensor, 5ms sampling), fuel cell cycle number N1 (flight mission count, each completed "take-off→landing" counted as 1 time), super capacitor 10C cycle number N2 (high rate charge and discharge count, 10C and above charge and discharge counted as 1 time), etc.

[0143] The final correction factor f(T,Gp) describes the attenuation coefficient of load and temperature on the rated power of the fuel cell. It is obtained by dynamic calculation through the stage correlation map + online calibration. Specifically:

[0144] The dynamic weight factor wstage of the flight phase is introduced (takeoff: wstage = 0.7, focusing on response; level flight: wstage = 0.9, focusing on accuracy), and the initial correction factor is calculated by fusion features:

[0145] (1) Basic attenuation term (temperature deviation contribution):

[0146] Assume that the fuel cell temperature deviation ΔT = T - Topt (Topt = 80°C is the optimal temperature), and the load deviation ΔGp = Gp - Gp,opt, where Gp,opt is the optimal load (e.g., the standard load is 2000 kg):

[0147] Attenbase=1-kT·|ΔT|-KG·|ΔGp|, where Attenbase is the basic attenuation coefficient; kT is the power attenuation coefficient (e.g., kT=0.005℃ during takeoff). -1 , for every 1°C deviation, the power attenuation is 0.5%); KG is the power attenuation coefficient associated with the load (calibrated value, such as KG = 0.0001kg -1 , power attenuation is 1% for every 100kg deviation).

[0148] (2) Additional attenuation term (contribution of temperature change rate and load change rate)

[0149] Set the maximum temperature change rate allowed in the stage (Takeoff: 5℃ / s; level flight: 2℃ / s), real-time temperature change rate

[0150] Set the maximum load change rate allowed in the stage (Takeoff: 0.5t / s, level flight: 0.1t / s, wind tunnel calibration), real-time temperature change rate

[0151]

[0152] Where kT1 is the temperature change coupling coefficient (kT1 = 0.1s / °C during takeoff, after the temperature exceeds the threshold, the attenuation is an additional +10%), and kG1 is the load change rate coupling coefficient (calibrated value, such as kG1 = 0.05s / t, after the load change exceeds the threshold, the attenuation is an additional +5%)

[0153] (3) Dynamic weighted fusion:

[0154] Combined with the historical mean correction factor f1(T,Gp) (based on offline data statistics, such as the historical mean of 0.7 during the takeoff phase):

[0155] Online calibration (closed loop correction):

[0156] By feedback of the deviation between the actual output power and the corrected rated power, f(T,Gp) is iteratively optimized:

[0157] (1) Deviation calculation:

[0158] The corrected rated power Prated,x = Prated·f2(T,Gp), and the actual output power Pfc,actual (collected in real time by the power sensor) is: e = Pfc,actual - Prated,x.

[0159] (2) PID controller adjustment:

[0160] Dynamic calibration using PID algorithm (parameters: Kp = 0.01, Ki = 0.001, Kd = 0.005): Continue iterating until the deviation satisfies the accuracy |e| < 1%·Prated, and finally f(T,Gp) = f3(T,Gp).

[0161] The key influencing features extracted are the temperature sensitivity threshold (T<50°C during the takeoff phase is the attenuation critical region), the power attenuation coefficient (for every 1°C drop, Prated attenuates by 0.5%), and the temperature change rate coupling coefficient (when the heating rate is >5°C / s, the attenuation increases by an additional 10%), as shown in Table 1:

[0162] Table 1

[0163]

[0164]

[0165] The dynamic weighting factor is introduced because the power demand is high during takeoff, and the weighting is biased towards fast response (allowing an error of 2% in f(T,Gp)). The focus during level flight is on efficiency and accuracy (error <1%).

[0166] Online calibration is to compare the actual output power with the corrected rated power (Prated·f(T,Gp)) and iteratively adjust f(T,Gp) through PID feedback until the deviation is <1%.

[0167] In this embodiment, the fuel cell is corrected for cycle attenuation (N1 per cycle, 99.95% power remaining) to match its slow attenuation characteristics (5% performance drop after a thousand cycles), and the supercapacitor is corrected for 10C cycle attenuation (N2 per 10C cycle, 99% power remaining) to match its high-rate fragility characteristics (10% performance drop after a hundred 10C cycles).

[0168] The above formula simultaneously integrates temperature-load decay (f(T,Gp)), cycle life (N1 / N2), and real-time demand (Preq), allowing the margin calculation to cover the entire dimension of environment → life → operating conditions, avoiding the one-sidedness of traditional static rated power. Differentiated decay: Distinguishing the life characteristics of fuel cells (slow decay, coefficient 0.0005) and super capacitors (fast decay, coefficient 0.0001), tailoring to the hardware differences between fuel cells (long cycle) and high-rate capacitors (high power but life-sensitive).

[0169] Driven by aviation scenarios, with frequent takeoffs and landings (N1 / N2 accumulates quickly) and a wide temperature range (T fluctuates up to ±60°C), traditional margin formulas will overestimate power and miss risks (e.g., the actual power of the stack at low temperatures is only 50% of the rated power, but the static formula is still calculated as 100%, resulting in an inflated margin). The corrected Margin directly drives a level 3 alarm, ensuring redundancy and controllability under aging conditions and extreme temperatures, and avoiding chain failures caused by a single fault.

[0170] The beneficial effects of the above technical solution are: through dynamic correction of temperature and load + quantification of life attenuation + online closed-loop calibration, it breaks through the limitations of traditional static estimation of power margin and realizes precise control of safety redundancy of aviation hydrogen-electric range extension system under wide temperature range, high cycle and multi-stage working conditions.

[0171] The present invention provides an adaptive planning aviation fuel cell range extension power system evaluation system, such as Figure 2 As shown, including:

[0172] A model building module is used to build a state space model of the aviation fuel cell range-extending power system, wherein the state space model includes: state variables, control variables and output variables of the system;

[0173] The multi-stage decision module is used to establish a multi-stage decision model, dividing the aviation flight process into three stages: takeoff, level flight, and landing. Each stage corresponds to different power requirements and constraints. The control strategy for each stage is solved by an adaptive dynamic programming algorithm. The state transition equation of the adaptive dynamic programming algorithm is:

[0174] s(k+1)=f(s(k),u(k))+w(k)

[0175] Wherein, s(k+1) is the predicted state vector at the k+1th flight time, f() is the state transfer function, s(k) is the state vector at the kth flight time, u(k) is the control vector at the kth flight time, and w(k) is the noise vector of the aviation fuel cell range-extended power system;

[0176] A matrix construction module is used to obtain the actual state vector y(k+1) at the k+1th flight time, and use the state difference vector c(k+1) between the actual state vector y(k+1) and the predicted state vector s(k+1) to construct a state difference matrix for each flight phase;

[0177] The graph determination module is used to perform difference mining on each state difference matrix and determine the correlation graph between the state variables and the control variables in the corresponding flight phase;

[0178] a dynamic adjustment module for collecting current state data of the aviation fuel cell range-extending power system in real time, and adjusting the optimization parameters of the control strategy for the corresponding stage determined by the dynamic programming algorithm based on the current state data in combination with the state space model and the association relationship map, to obtain the optimal strategy for the corresponding stage, and dynamically adjusting the output power of the fuel cell and the ultracapacitor in combination with the power allocation strategy;

[0179] An alarm module is used to dynamically adjust the power output margin of the aviation fuel cell range-extending power system according to the current status data, and trigger an alarm signal when the power output margin is lower than a safety preset threshold.

[0180] In this embodiment, Figure 3 As shown, the preset safety threshold is:

[0181] Warning level: Performance degradation risk;

[0182] Trigger logic: The alarm module detects that the power margin has entered the performance degradation risk range and sends an early warning-level disposal instruction to the dynamic adjustment module.

[0183] Dynamically adjust module response:

[0184] Power distribution adjustment: Increase the fuel cell output power to 90% of the rated power (strengthen the main energy supply and compensate for attenuation), and reduce the supercapacitor discharge rate by 10% (reduce high-rate cycles and delay aging).

[0185] Condition monitoring upgrade: The matrix building module is called to increase the collection frequency of status data (temperature T, SOC, etc.) from 100ms to 50ms, more intensively capturing power attenuation trends. At the same time, the correlation between modules is determined by combining the map, predicting the subsequent impact of temperature, load, and number of cycles on the margin, and optimizing the power strategy in advance.

[0186] Emergency level: Insufficient safety redundancy;

[0187] Trigger logic: The alarm module determines that "safety redundancy is insufficient" and sends an "emergency level disposal instruction" to the dynamic adjustment module. The synchronous marking map determination module focuses on analyzing the "power-life" correlation path.

[0188] Dynamically adjust module response:

[0189] Power hard constraint: The supercapacitor is forced to limit its power to 80% of the rated power (to avoid deep discharge damage), and the fuel cell outputs at the rated power (to fully fill the power gap).

[0190] Load cutoff linkage: Send instructions to the load management unit (system expansion module) through the CAN bus interface (engineering implicit interaction) to cut off non-critical loads (such as avionics entertainment systems and power supply for secondary sensors), retaining only core circuits such as flight control and power systems, giving priority to flight control safety.

[0191] Severity: Risk of forced landing.

[0192] Trigger logic: The alarm module identifies the "forced landing risk" and sends a "severe level disposal instruction" to the dynamic adjustment module, while triggering the multi-stage decision module to switch to "emergency mode".

[0193] Dynamically adjust module response:

[0194] Extreme power output: Activate the emergency power mode, the fuel cell is temporarily overloaded to 110% of the rated power (depending on the temperature correction model of the map determination module to ensure that short-term high temperature does not damage the stack), and the super energy capacitor is discharged at times the rated power (using the remaining life limit to replenish energy).

[0195] Flight control linkage forced landing: Send a forced landing warning signal through the flight control system interface (engineering implicit interaction), triggering the flight control to automatically plan the nearest alternate landing route; at the same time, call the matrix construction module to upload the full status data of the power system (temperature, power, hydrogen volume, load, etc.) to the ground station in real time to support remote emergency decision-making.

[0196] The beneficial effects of the above technical solution are: through the full process of state modeling → multi-stage prediction → error closed loop → correlation mining → strategy optimization → safety guarantee, it breaks through the static limitations of traditional control: it not only accurately matches the multi-stage dynamic power requirements of aviation, but also quantifies safety redundancy through health decay and environmental correction, realizes deep coordination between fuel cells and ultra-energy capacitors, reduces energy loss, and meets the needs of efficient and safe operation of aviation fuel cell range extension systems.

[0197] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. An adaptive programming aviation fuel cell range-extending power system evaluation method, characterized in that: include: Constructing a state space model of an aviation fuel cell range-extending power system, wherein the state space model includes: state variables, control variables, and output variables of the system; A multi-stage decision model is established to divide the aviation flight process into three stages: takeoff, level flight, and landing. Each stage corresponds to different power requirements and constraints. The control strategy for each stage is solved using an adaptive dynamic programming algorithm. The state transition equation of the adaptive dynamic programming algorithm is: s(k+1)=f(s(k),u(k))+w(k) Where s(k+1) is the predicted state vector at the k+1th flight time, f( ) is the state transfer function, s(k) is the state vector at the kth flight time, u(k) is the control vector at the kth flight time, and w(k) is the noise vector of the aviation fuel cell range-extended power system. Obtain the actual state vector y(k+1) at the k+1th flight time, and use the state difference vector c(k+1) between the actual state vector y(k+1) and the predicted state vector s(k+1) to construct a state difference matrix for each flight phase; Perform difference mining on each state difference matrix to determine the correlation graph between state variables and control variables in the corresponding flight phase; Real-time collection of current state data of the aviation fuel cell range-extending power system, and in combination with the state space model and the association relationship map, adjustment of optimization parameters of the control strategy for the corresponding stage determined by the dynamic programming algorithm based on the current state data, to obtain the optimal strategy for the corresponding stage, and dynamic adjustment of the output power of the fuel cell and the ultracapacitor in combination with the power allocation strategy; The power output margin of the aviation fuel cell range-extending power system is dynamically adjusted according to the current state data, and an alarm signal is triggered when the power output margin is lower than a safety preset threshold.

2. The adaptive programming aviation fuel cell range-extending power system evaluation method according to claim 1, characterized in that: Construct a state space model of the aviation fuel cell range-extended power system, including: Extracting a phase vector set of the fuel cell from the start of takeoff to the end of landing from a historical database, wherein the phase vector set includes a phase vector of each historical flight moment under different flight phases; Determine the quantitative relationship between the state variables and the output variables in the stage vector: When the number of the state variables is less than the number of the output variables, determining the input controllable quantity that affects the control variable and the output controllable quantity that has a potential effect on the output variable; Mapping the input controllable quantities and the output controllable quantities into a variable space coordinate system, first retaining controllable quantities in the output controllable quantities that have a unidirectional dependency relationship with the output controllable quantities, and performing state analysis on the first retained controllable quantities until a first number of retained input controllable quantities is greater than a second number of retained output controllable quantities, and a sum of the number of state variables and the first number is greater than the sum of the number of output variables and the second number; Among them, status analysis includes: Extracting the first amplification vector at the first flight moment and the second amplification vector at the next flight moment in each flight phase, determining the effective amplification coefficient of each amplification variable, and obtaining the coefficient vector of the corresponding amplification variable in the flight phase; When the coefficient vector tends to a stable state, it is preferentially retained; Otherwise, keep the original phase vector; A state space model is constructed based on the retained vector and combined with the least squares method.

3. The adaptive programming aviation fuel cell range-extending power system evaluation method according to claim 1, characterized in that: The multi-objective function of the power allocation strategy is: Among them, w1, w2, w3, and w4 represent weight coefficients, which are dynamically adjusted according to the flight stage, equipment health status, and load; Preq represents the required power during the flight stage; Topt represents the optimal operating temperature of the fuel cell in the corresponding stage; H 安全阈值 represents the hydrogen safety threshold of the corresponding stage; Prated represents the rated power of the fuel cell; Trange represents the allowable temperature; Hreated represents the total hydrogen storage; Is represents the instantaneous discharge current of the supercapacitor; Ir represents the rated current of the supercapacitor, minJ represents the multi-objective function; k represents the slope of the Sigmoid function, with a value of 10 to 20; M represents the membrane dryness index; ∝ represents the wetness of the proton exchange membrane of the fuel cell, with a value of 0 to 1; T represents the temperature of the fuel cell stack; H represents the remaining hydrogen amount; Pfc represents the actual output power of the fuel cell; Pcap represents the actual output power of the supercapacitor.

4. The adaptive programming aviation fuel cell range-extending power system evaluation method according to claim 1, characterized in that: The aviation flight process is divided into three stages: takeoff, level flight and landing, including: Based on the improved K-means clustering and altitude change rate cross-validation method, the power demand slope is extracted and the flight phases are intelligently divided into three phases: takeoff, level flight, and landing. Among them, the energy is provided by super capacitor during the take-off phase; The fuel cell provides energy during the level flight phase, and the excess energy is used to charge the super capacitor. The landing phase is powered by both supercapacitors and fuel cells.

5. The adaptive programming aviation fuel cell range-extending power system evaluation method according to claim 3, characterized in that: The hard constraints on the output power of the supercapacitor are: Among them, SOC represents the percentage of charge of the super capacitor; Gmax is the maximum load threshold; and Gp is ​​the current load.

6. The adaptive programming aviation fuel cell range-extending power system evaluation method according to claim 5, characterized in that: Dynamically adjusting the power output margin of the aviation fuel cell range-extending power system according to the current state data includes: Determining, based on the correlation relationship map corresponding to the flight phase, a correction factor f(T, Gp) of the rated output power of the fuel cell in the current state data based on temperature and load; Correct Pfc+Pcap-Preq based on the correction factor to obtain the power output margin Margin; Margin=Prated·(1-0.0005·N1)·f(T,Gp)+Pcap,e·(1-0.0001·N2)-Preq Wherein, N1 represents the number of cycles of the fuel cell; N2 represents the number of cycles of the supercapacitor; and Pcap,e represents the rated output power of the supercapacitor.

7. The adaptive programming aviation fuel cell range-extending power system evaluation method according to claim 6, characterized in that: Determining a temperature-based correction factor f(T,Gp) of the rated output power of the fuel cell in the current state data based on a correlation relationship map corresponding to the flight phase includes: Based on the correlation map of each flight phase, the key impact characteristics of temperature on the rated output power of the fuel cell are extracted. The key impact characteristics include the temperature sensitivity threshold, load sensitivity threshold, and the coupling coefficient between the power attenuation coefficient and the temperature change rate and load change rate at different stages. The correlation map contains at least the nonlinear correlation path and path influence weight between the temperature variable, load variable and the rated output power of the fuel cell; A dynamic weighting factor for each flight phase is introduced to weightedly integrate the key influencing features of each phase and establish a temperature-load-power correction model. The dynamic weighting factor is adaptively adjusted according to the power demand intensity, temperature fluctuation amplitude, and load size of the corresponding phase. Real-time collection of temperature data, load data, and rated output power reference value for the current flight phase, inputting the data into the temperature-load-power correction model, and calculating an initial temperature-load correction factor; The initial temperature-load correction factor is calibrated online by feedback of the deviation between the actual output power and the corrected rated power until the deviation meets the preset accuracy requirement, and the final correction factor f(T,Gp) is obtained.

8. An adaptive programming aviation fuel cell range-extending power system evaluation system, characterized in that: include: A model building module is used to build a state space model of the aviation fuel cell range-extending power system, wherein the state space model includes: state variables, control variables and output variables of the system; The multi-stage decision module is used to establish a multi-stage decision model, dividing the aviation flight process into three stages: takeoff, level flight, and landing. Each stage corresponds to different power requirements and constraints. The control strategy for each stage is solved by an adaptive dynamic programming algorithm. The state transition equation of the adaptive dynamic programming algorithm is: s(k+1)=f(s(k),u(k))+w(k) Wherein, s(k+1) is the predicted state vector at the k+1th flight time, f() is the state transfer function, s(k) is the state vector at the kth flight time, u(k) is the control vector at the kth flight time, and w(k) is the noise vector of the aviation fuel cell range-extended power system; A matrix construction module is used to obtain the actual state vector y(k+1) at the k+1th flight time, and use the state difference vector c(k+1) between the actual state vector y(k+1) and the predicted state vector s(k+1) to construct a state difference matrix for each flight phase; The graph determination module is used to perform difference mining on each state difference matrix and determine the correlation graph between the state variables and the control variables in the corresponding flight phase; a dynamic adjustment module for collecting current state data of the aviation fuel cell range-extending power system in real time, and adjusting the optimization parameters of the control strategy for the corresponding stage determined by the dynamic programming algorithm based on the current state data in combination with the state space model and the association relationship map, to obtain the optimal strategy for the corresponding stage, and dynamically adjusting the output power of the fuel cell and the ultracapacitor in combination with the power allocation strategy; An alarm module is used to dynamically adjust the power output margin of the aviation fuel cell range-extending power system according to the current status data, and trigger an alarm signal when the power output margin is lower than a safety preset threshold.

Images