Hydrogen energy unmanned aerial vehicle control system and method based on model prediction algorithm

Abstract

The application relates to the technical field of unmanned aerial vehicle control, and discloses a hydrogen energy unmanned aerial vehicle control system and method based on a model prediction algorithm, which comprises the following steps: an airborne controller acquires task flight tracks, unmanned aerial vehicle flight state data and power supply data, generates a reference flight state vector and a comprehensive state vector; current power gaps, power gap increments, bus voltage drops and smooth bus voltage drop sensitivities are determined according to actual motor group power, actual fuel cell power and direct current bus voltages; the allowed power gap and the thrust increment reduction ratio are determined in combination with the bus minimum voltage and the bus safety voltage margin, and are written into the dynamic power boundary of the model prediction control; after the controller solves the control sequence, the control instruction is executed, and the total thrust instruction increment and the power battery power instruction are limited when the instruction corresponding power gap exceeds the allowed power gap, and are used for the track, height and attitude control of the hybrid energy unmanned aerial vehicle.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) control technology, and more specifically, to a hydrogen-powered UAV control system and method based on model prediction algorithms. Background Technology

[0002] Multirotor UAVs powered by a hybrid of hydrogen fuel cells and batteries have a good foundation for energy utilization in long-endurance missions. Fuel cells are suitable for providing relatively stable and continuous power, while batteries are suitable for handling transient power demands during climbs, turns, and wind corrections. Flight control systems typically need to simultaneously handle trajectory tracking, attitude stabilization, propulsion power distribution, DC bus voltage maintenance, and energy storage safety constraints.

[0003] In actual flight, the power output of the motor group changes rapidly with the total thrust command. The fuel cell output exhibits a response lag, and the power battery, bidirectional converter, DC bus, and cables share the burden of short-term power shortages. When the power battery capacity decreases, the bidirectional converter approaches its current limit, or the bus load becomes heavy, the same power shortage may result in different bus voltage drops. If the flight control system only uses a fixed minimum bus voltage, a fixed maximum power battery capacity, or a fixed thrust increment limit, the controller will struggle to promptly identify the current power supply chain stress state.

[0004] For hydrogen-powered drones, undervoltage on the power bus directly affects ESC input, motor speed, and propulsion output, further impacting altitude hold, attitude response, and trajectory tracking. In engineering, without altering the existing power supply chain connections, it's necessary to incorporate the power supply chain's voltage response to power deficit changes into model predictive control constraints and execution limits, utilizing existing sampled data such as DC bus voltage, actual fuel cell power, and actual motor group power. This allows flight control and power management to work collaboratively within the same control cycle. Summary of the Invention

[0005] This invention provides a hydrogen-powered unmanned aerial vehicle (UAV) control system and method based on model prediction algorithms, which solves the technical problems mentioned in the background art.

[0006] To achieve the above-mentioned objectives, the present invention provides the following technical solution.

[0007] A model-predictive algorithm-based control method for hydrogen-powered drones is applied to hydrogen-powered drones equipped with fuel cells and power batteries. The onboard controller executes the following steps in each control cycle: Acquire mission trajectory, UAV flight status data and power supply data; generate reference flight status vector based on mission trajectory; and generate comprehensive status vector based on UAV flight status data and power supply data. Power supply data includes DC bus voltage, fuel cell actual power and motor group actual power. The current power gap is determined based on the actual power of the motor group and the actual power of the fuel cell, and the power gap increment and bus voltage drop are also determined. The bus voltage drop sensitivity is determined based on the power gap increment and the bus voltage drop, and the bus voltage drop sensitivity is smoothed to obtain the smoothed bus voltage drop sensitivity. The allowable power gap and thrust increment reduction ratio are determined based on the DC bus voltage, the minimum bus voltage, the bus safety voltage margin, and the smooth bus voltage drop sensitivity. Establish a dynamic power boundary to ensure that the difference between the predicted power of the motor group and the predicted power of the fuel cell does not exceed the allowable power gap, and limit the total thrust command increment to the thrust increment reduction ratio. The control sequence is solved based on the integrated state vector, reference flight state vector, prediction model and dynamic power boundary; The control command is executed, and the difference between the predicted power of the motor group and the predicted power of the fuel cell corresponding to the control command is determined as the power gap corresponding to the command. When the power gap corresponding to the command exceeds the allowable power gap, the total thrust command increment and the power battery power command are restricted.

[0008] Compared with the prior art, the present invention has the following substantial features and significant progress: This invention introduces a real-time response of the DC bus to changes in power supply and demand in the closed-loop flight control of an unmanned aerial vehicle (UAV). The power supply withstand state formed by the actual power of the motor group, the actual power of the fuel cell, and the DC bus voltage is used for model predictive control constraints and execution limits. This enables the controller to adjust the future power gap and the increment of the total thrust command according to the power supply capacity during climb, turn, and gust correction. This reduces the motor speed drop, thrust loss, altitude error, attitude error, and trajectory deviation caused by undervoltage of the bus, while maintaining trajectory tracking and safety degradation control under the coordinated power supply of the fuel cell and the power battery. Attached Figure Description

[0009] Figure 1 This is a system architecture and control mechanism diagram of the hydrogen-powered drone control method and system based on model prediction algorithm of the present invention. Detailed Implementation

[0010] The following description is provided in conjunction with the accompanying drawings. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the invention; those skilled in the art can make equivalent substitutions or combinations for the specific implementations.

[0011] Example 1: This implementation applies to hydrogen-powered unmanned aerial vehicles (UAVs) equipped with fuel cells and power batteries. The hydrogen-powered UAV can be a multi-rotor UAV, equipped with a fuel cell power supply branch, a power battery power supply branch, a DC bus, and a motor array. The fuel cell is suitable for providing relatively stable continuous power, while the power battery is suitable for supplementing transient power during rapid climbs, turns, wind corrections, and altitude maintenance. The motor array obtains propulsion power from the DC bus. The onboard controller simultaneously processes mission trajectory, UAV flight status data, and power supply data in each control cycle, ensuring that trajectory tracking, attitude control, propulsion power requirements, and power supply safety constraints are considered uniformly within the same control cycle.

[0012] like Figure 1 As shown, the hydrogen-powered drone 100 includes a reference sequence generation module 110, a data acquisition and state estimation module 120, a bus voltage drop sensitivity calculation module 130, a dynamic power boundary generation module 140, a model prediction control module 150, a command execution and degradation control module 160, and a hybrid propulsion system 170.

[0013] Specifically, the onboard controller does not only determine whether the DC bus voltage is below a fixed threshold, nor does it simply determine whether the power battery power has reached a fixed upper limit. Instead, within each control cycle, it determines the current power gap, the power gap increment, and the bus voltage drop based on the actual power of the motor group, the actual power of the fuel cell, and the DC bus voltage. Furthermore, it determines the bus voltage drop sensitivity and the smoothing bus voltage drop sensitivity. The smoothing bus voltage drop sensitivity represents the degree of bus voltage drop caused by a unit increase in power gap. It reflects the power supply withstand status after the combined effects of the current fuel cell response, the power battery's power supply capacity, the bidirectional converter's current limiting status, cable losses, and the bus capacitor's support capacity.

[0014] Based on this, the airborne controller determines the allowable power gap according to the DC bus voltage, the minimum bus voltage, the bus safety voltage margin, and the smoothing bus voltage drop sensitivity, and determines the thrust increment reduction ratio according to the relationship between the allowable power gap and the reference power gap. The allowable power gap and the thrust increment reduction ratio together form the dynamic power boundary, which is used for subsequent model predictive control solutions and pre-control command limiting.

[0015] Example 2: At the start of each control cycle, the onboard controller first acquires the mission track. The mission track can be provided by a pre-set list of waypoints, ground station upload commands, or the onboard mission management program. The mission track may include mission waypoints, target altitude, target heading, target speed, and permitted flight area. The mission track describes the spatial path that the hydrogen-powered UAV 100 should follow during the current mission phase, as well as the altitude, speed, and heading requirements that must be met along that spatial path.

[0016] The reference sequence generation module 110 generates a reference flight state vector based on the mission trajectory. The reference flight state vector can cover the finite prediction time domain corresponding to the prediction steps, and progressively provides the target flight state for each prediction step within the prediction time domain according to the sampling time interval. For multi-rotor UAVs, the reference flight state vector includes at least three-dimensional position, three-dimensional velocity, and yaw direction; in engineering implementation, the roll angle and pitch angle can be calculated based on the desired horizontal acceleration output by the position controller, a common implementation method in multi-rotor flight control.

[0017] The mathematical representation of the reference flight state vector is as follows ,in Indicates the current sampling sequence number. This indicates the prediction step number. The current sampling number is used to identify the current control cycle, and the prediction step number is used to identify the number of steps relative to the current control cycle within the prediction time domain. When generating the reference flight state vector, the reference sequence generation module 110 can generate the reference sequence according to the sampling time interval. The mission trajectory is discretized so that each prediction step corresponds to a target flight state that can be solved by the model prediction control module 150.

[0018] When the target altitude or velocity in the mission trajectory undergoes a step change, the reference sequence generation module 110 can generate a reference flight state vector using a conventional trajectory smoothing method, ensuring the reference flight state vector remains continuous in the prediction time domain. This trajectory smoothing method can be velocity limiting, acceleration limiting, or waypoint interpolation, its purpose being solely to avoid unnecessary abrupt changes in control commands caused by the reference flight state vector itself.

[0019] The reference flight state vector is output to the model predictive control module 150. When solving the control sequence in the subsequent process, the model predictive control module 150 uses the reference flight state vector as the target term for the flight tracking error, so that the control sequence tracks the mission trajectory as closely as possible while satisfying the dynamic power boundary, fixed equipment boundary, and safety constraints.

[0020] Example 3: The data acquisition module simultaneously collects UAV flight status data and power supply data. UAV flight status data can include UAV position, speed, attitude quaternions, and angular velocity, sourced from inertial measurement units, satellite positioning receivers, and barometric altimeters. Power supply data includes DC bus voltage, actual fuel cell power, and actual motor group power, and may also include actual battery power, remaining battery charge percentage, fuel cell temperature, and remaining hydrogen mass.

[0021] The data acquisition module communicates with the inertial measurement unit, satellite positioning receiver, barometric altimeter, electronic speed controller (ESC), DC bus sensor, fuel cell controller, power battery management system, and hydrogen storage measurement device. The DC bus sensor collects the DC bus voltage; the fuel cell controller outputs the actual fuel cell power, or calculates the actual fuel cell power from the fuel cell side voltage and current; the ESC outputs the input power of each ESC, and the data acquisition module sums these inputs to obtain the actual power of the motor group; the power battery management system outputs the actual power of the power battery and the remaining battery charge percentage; and the hydrogen storage measurement device collects the remaining hydrogen mass.

[0022] To ensure that various measured values ​​can be used in calculations within the same control cycle, the data acquisition module aligns the UAV flight status data and power supply data according to a unified clock. When different sensors have inconsistent sampling frequencies, the data acquisition module can use methods such as maintaining the previous effective value, linear interpolation, or nearest-neighbor synchronization to form a data set for the same control cycle. The aforementioned time alignment method is part of existing data synchronization processing, and its function is to ensure that the DC bus voltage, actual fuel cell power, and actual motor group power correspond to the same current sampling sequence number when calculating the current power gap, power gap increment, and bus voltage drop.

[0023] The onboard controller stores the DC bus voltage, actual fuel cell power, and actual motor group power for the current and previous control cycles. The DC bus voltage for the current control cycle is expressed as follows: The DC bus voltage of the previous control cycle is expressed as: The actual power of the fuel cell in this control cycle is expressed as: The actual power of the fuel cell in the previous control cycle is expressed as: The actual power of the motor group in this control cycle is expressed as follows: The actual power of the motor group in the previous control cycle is expressed as: After saving the above historical values, the bus voltage drop sensitivity calculation module 130 can determine the actual response of the power supply chain to the increase in power shortage based on the changes in adjacent control cycles.

[0024] The state estimation module generates a comprehensive state vector based on the UAV's flight state data and power supply data. The comprehensive state vector is represented as follows: It can include flight state vectors DC bus voltage Remaining battery charge percentage Fuel cell temperature Actual power of fuel cells Actual power of the motor group Remaining hydrogen mass And the sensitivity of the smoothing bus voltage drop in the previous control cycle .

[0025] The state estimation module can perform conventional low-pass filtering on inertial measurement, position, altitude, and voltage / current data, and use an extended Kalman filter or complementary filter to obtain the UAV position, UAV velocity, UAV attitude quaternion, and UAV angular velocity. The purpose of filtering and state estimation is to reduce the impact of measurement noise on the model predictive control solution, so that the synthesized state vector can be used as the initial state input to the model predictive control module 150 for the current control cycle.

[0026] The power supply data and integrated state vector output by the data acquisition and state estimation module 120 are derived from the actual operation of the hydrogen-powered drone 100, rather than from offline fixed parameters. Therefore, the bus voltage drop sensitivity calculation and model predictive control solution can utilize real-world state information within the same control cycle.

[0027] Example 4: The bus voltage drop sensitivity calculation module 130 determines the current power gap based on the actual power of the motor group and the actual power of the fuel cell. The current power gap represents the difference between the actual power of the motor group and the actual power of the fuel cell, which is the power supply and demand difference that needs to be shared by the power battery, bus capacitor and other power supply branches in the current control cycle. The larger the current power gap, the greater the instantaneous difference between the actual power of the fuel cell and the actual power of the motor group. ; in, This indicates the current power deficit, expressed in watts (W). This indicates the actual power of the motor group, in watts (W). This represents the actual power of the fuel cell, measured in watts (W). The actual power of the motor group is obtained by summing the input power of the ESC, while the actual power of the fuel cell is obtained from the output of the fuel cell controller or calculated from the fuel cell-side voltage and current.

[0028] The bus voltage drop sensitivity calculation module 130 determines the power gap increment based on the current power gap. The power gap increment represents the change in the current power gap between two adjacent control cycles, and is particularly used to identify whether the actual power of the motor group is increasing relative to the actual power of the fuel cell. When the actual power of the motor group rises rapidly while the actual power of the fuel cell cannot rise synchronously due to the fuel cell response time limitation, the power gap increment is usually positive. ; in, This represents the power deficit increment, in W. This indicates the current power gap corresponding to the current sampling sequence number; This indicates the current power deficit corresponding to the previous control cycle. The onboard controller stores the DC bus voltage, actual fuel cell power, and actual motor group power from the previous control cycle, thus enabling it to obtain... .

[0029] The bus voltage drop sensitivity calculation module 130 determines the bus voltage drop based on the DC bus voltage. The bus voltage drop represents the degree of decrease in DC bus voltage between two adjacent control cycles. If the DC bus voltage in the current control cycle is lower than the DC bus voltage in the previous control cycle, the bus voltage drop is positive; if the DC bus voltage in the current control cycle is not lower than the DC bus voltage in the previous control cycle, this change does not indicate a voltage drop. ; in, This indicates the voltage drop at the busbar, expressed in volts (V). This represents the DC bus voltage of the previous control cycle; This represents the DC bus voltage for the current control cycle. The voltage drop is directly derived from voltage sampling of the same DC bus, thus reflecting the instantaneous voltage performance of the hybrid propulsion system 170 under real load changes.

[0030] When determining the current power deficit, power deficit increment, and bus voltage drop, the onboard controller can discard measurements that significantly exceed the sensor's range. If the DC bus voltage, actual fuel cell power, or actual motor group power is determined to be invalid within a certain control cycle, the onboard controller can use the valid sampled value from the previous control cycle as a temporary input and write this data missing status into the fault status. This processing is a standard engineering measure for sensor anomaly rejection and data retention, aiming to avoid sudden changes in bus voltage drop sensitivity caused by invalid sampling.

[0031] The current power gap indicates the difference between the actual power of the fuel cell and the actual power of the motor group. The increase in the power gap indicates whether the difference is widening. The decrease in the bus voltage indicates whether the DC bus voltage is decreasing as the difference widens.

[0032] Example 5: The bus voltage drop sensitivity calculation module 130 determines the bus voltage drop sensitivity based on the degree of decrease in bus voltage relative to the power gap increment when the power gap increment is positive. Bus voltage drop sensitivity represents the degree of DC bus voltage drop when the power gap increases by one unit, and the unit can be V / W. A higher bus voltage drop sensitivity indicates that the current power supply chain is more prone to DC bus voltage drops when the actual power of the motor group continues to increase while the actual power of the fuel cell has not yet increased synchronously; a lower bus voltage drop sensitivity indicates that the power battery, bidirectional converter, cables, and bus capacitors have a stronger ability to support the increase in power gap. ; in, This indicates the bus voltage drop sensitivity, expressed in V / W. Indicates the amount of voltage drop at the busbar; Indicates the power deficit increment; This represents the power calculation protection value, expressed in watts (W). The power calculation protection value is used to prevent the numerical value from being amplified due to an excessively small denominator when the power deficit increment approaches zero. This is used to prevent changes in the DC bus voltage that are not decreasing from being amplified as a decrease in bus voltage. This is used to prevent negative changes from being considered as effective excitation when the power gap does not increase.

[0033] When the power deficit increment is not positive, the bus voltage drop sensitivity calculation module 130 maintains the smoothed bus voltage drop sensitivity of the previous control cycle, without updating the smoothed bus voltage drop sensitivity using the change in this control cycle. This is because when the power deficit does not increase, the change in DC bus voltage may come from fuel cell power recovery, power battery charging and discharging switching, bus load reduction, or measurement noise, and cannot directly represent the degree of bus voltage drop caused by the increase in power deficit.

[0034] To reduce the impact of sensor noise and disturbances in a single control cycle on the bus voltage drop sensitivity, the bus voltage drop sensitivity calculation module 130 smooths the bus voltage drop sensitivity to obtain a smoothed bus voltage drop sensitivity. The smoothing process can employ first-order time smoothing, combining the new bus voltage drop sensitivity with the smoothed bus voltage drop sensitivity from the previous control cycle according to a sensitivity smoothing ratio. ; in, This indicates the sensitivity to smoothing bus voltage drop, expressed in V / W. This represents the sensitivity smoothing ratio and is a dimensionless quantity. This indicates the bus voltage drop sensitivity determined in this control cycle; This indicates the sensitivity of the smoothing bus voltage drop in the previous control cycle. The sensitivity smoothing ratio can be calibrated between 0 and 1. The larger the sensitivity smoothing ratio, the greater the impact of the new measurement on the smoothing result; the smaller the sensitivity smoothing ratio, the greater the impact of the smoothing result of the previous control cycle on the current result.

[0035] The sensitivity smoothing ratio, power calculation protection value, and preset effective excitation range can be determined based on the power supply sampling noise of a similar hydrogen-powered UAV 100. Specifically, the airborne controller performs bench tests or hover tests when the fuel cell power supply branch, power battery power supply branch, DC bus, and motor group are all in a normal connection state. It records the DC bus voltage, actual fuel cell power, and actual motor group power over several consecutive control cycles, and calculates the power fluctuation amplitude and voltage fluctuation amplitude corresponding to the measured noise. The power calculation protection value is taken as a conventional value that is greater than the power fluctuation amplitude but less than the normal maneuver power gap increment. The preset effective excitation range is taken as the power increment range that can distinguish between the actual power increase and noise disturbance. The sensitivity smoothing ratio is taken as a dimensionless value that allows the smoothing bus voltage drop sensitivity to change with continuous maneuvering without excessive jumps at a single noise peak. After the fuel cell, power battery, ESC, thruster, or DC bus sensor is replaced, the airborne controller re-performs the above calibration.

[0036] When the power gap increment is less than the preset effective excitation range, the DC bus voltage data is judged as abnormal, the actual power data of the fuel cell is judged as abnormal, or the actual power data of the motor group is judged as abnormal, the bus voltage drop sensitivity calculation module 130 can skip the bus voltage drop sensitivity update of the current control cycle and use the most recent effective smoothed bus voltage drop sensitivity.

[0037] The smoothing bus voltage drop sensitivity is output to the dynamic power boundary generation module 140, so that the dynamic power boundary is not determined by a fixed calibration value alone, but is continuously updated by the DC bus voltage response obtained from actual sampling. This process illustrates that this embodiment differs from model predictive control based on simple trajectory tracking, as it incorporates the power supply withstand state of the hybrid propulsion system 170 into the model predictive control.

[0038] Example 6: The dynamic power boundary generation module 140 determines the allowable power gap based on the DC bus voltage, the minimum bus voltage, the bus safety voltage margin, and the sensitivity to smoothing the bus voltage drop. The allowable power gap represents the maximum supply-demand power difference allowed under the current bus condition, which is the upper limit that the difference between the predicted power of the motor group and the predicted power of the fuel cell can be allowed to reach while maintaining the DC bus voltage and preserving the bus safety voltage margin. ; in, Indicates the allowable power deficit, in watts (W). This represents the DC bus voltage, in units of V; This indicates the minimum bus voltage, in units of V; This indicates the busbar safety voltage margin, expressed in volts (V). This indicates the sensitivity to smoothing bus voltage drop, expressed in V / W. This represents the sensitivity calculation protection value, expressed in V / W. The sensitivity calculation protection value is used to prevent numerical amplification when the smoothing bus voltage drop sensitivity approaches zero.

[0039] when When the voltage is relatively large, the DC bus voltage has a significant margin relative to the minimum bus voltage and the bus safety voltage margin, allowing for a corresponding increase in the allowable power deficit; when When the voltage is low, the DC bus voltage is close to the reserve limit, and the allowable power gap decreases accordingly. When the sensitivity of smoothing the bus voltage drop increases, the bus voltage drop caused by the increase in the unit power gap is greater, and the allowable power gap decreases accordingly. Thus, the allowable power gap reflects both the current DC bus voltage level and the actual response of the power supply chain to the increase in the power gap.

[0040] When the allowable power gap obtained according to the above relationship is less than zero, the dynamic power boundary generation module 140 can limit the allowable power gap to zero, or provide this state to the command execution and degradation control module 160 as one of the bases for entering the power limiting, return, or landing process. This processing is a safety boundary processing of the calculation results, so that the control command will not continue to increase the difference between the predicted power of the motor group and the predicted power of the fuel cell when the DC bus voltage already lacks the bus safety voltage margin.

[0041] The dynamic power boundary generation module 140 determines the thrust increment reduction ratio based on the relationship between the allowable power gap and the pre-calibrated reference power gap. The reference power gap represents the power gap used for comparison under normal operating conditions and can be determined in calibration flights, bench tests, or simulation tests of the same type of hydrogen-powered UAV 100. The thrust increment reduction ratio is used to reduce the total thrust command increment, so that the total thrust command increment is reduced when the difference between the DC bus voltage and the minimum bus voltage decreases. ; in, It represents the percentage reduction in thrust increment and is a dimensionless quantity. Indicates the allowable power gap; This indicates the reference power gap, in watts (W). This indicates the power calculation protection amount. Because the thrust increment reduction ratio is... The limit is that its value does not exceed 1. When the allowable power gap is not less than the reference power gap, the thrust increment reduction ratio is close to 1, and the model predictive control module 150 can use the conventional total thrust command increment; when the allowable power gap is less than the reference power gap, the thrust increment reduction ratio is less than 1, and the total thrust command increment is reduced.

[0042] The baseline power gap and baseline thrust increment can be determined using data from normal cruise, regular climb, regular turns, and regular wind resistance corrections. During calibration, the onboard controller records the current power gap, power gap increment, total thrust command increment, and DC bus voltage in test segments where the low-voltage safety threshold, fuel cell high-temperature safety threshold, low battery charge safety threshold, and low hydrogen reserve safety threshold are not triggered. The power gap that does not cause a continuous drop in DC bus voltage and does not trigger the power boundary of the power battery is selected as a candidate value for the baseline power gap, and the total thrust command increment that can be stably executed in the corresponding test segment is selected as a candidate value for the baseline thrust increment. The final baseline power gap and baseline thrust increment are then determined by combining the rated thrust of the aircraft and the fuel cell response time.

[0043] The allowable power gap and thrust increment reduction ratio have a dual function: on the one hand, when solving the control sequence, it limits the difference between the predicted power of the motor group and the predicted power of the fuel cell in the prediction time domain, as well as the total thrust command increment; on the other hand, it checks the power gap corresponding to the command again and limits it before executing the control command.

[0044] Example 7: The model predictive control module 150 solves for the control sequence based on the integrated state vector, reference flight state vector, predictive model, allowable power gap, and thrust increment reduction ratio. The predictive model may include a multi-rotor rigid body dynamics model, a fuel cell power response model, a motor power model, a battery remaining charge ratio update model, and a DC bus voltage prediction model. These models are conventional calibrable models in model predictive control, and their function is to map the integrated state vector and candidate control commands to the flight and dynamic states in the prediction time domain. ; ; in, Indicates the prediction step number The corresponding predicted state vector includes at least the flight state vector, DC bus voltage, fuel cell predicted power, power battery predicted power, power battery remaining charge percentage, and remaining hydrogen mass. Indicates by sampling time interval Discretized prediction state update relationship; Indicates the prediction step number The corresponding control command vector; This represents the predicted output vector; This indicates the predicted output relationship. The predicted output vector outputs at least the predicted power of the motor group, the predicted power of the fuel cell, the DC bus voltage, and the total thrust command, enabling the dynamic power boundary and the stationary equipment boundary to be written into the solver in each prediction step.

[0045] In a minimum feasible discrete prediction model, the model prediction control module 150 decomposes the prediction state vector into the flight state vector, fuel cell prediction power, motor group prediction power, power battery prediction power, power battery remaining charge ratio, remaining hydrogen mass and DC bus voltage, and updates them in the following order at each prediction step. ; in, Indicates the prediction step number The corresponding flight state vector; Represents the flight state update matrix. Represents the flight control input matrix. Represents the flight state constant term; This represents the flight control sub-vector. The flight state update matrix and the flight control input matrix are obtained by linearizing and discretizing the multi-rotor rigid body dynamics model near the current integrated state vector. The coefficients in each row are calibrated according to the units of the corresponding flight state components, so that the two sides of the equal sign are flight state vectors with the same structure. ; in, Indicates the prediction step number The corresponding predicted power of the fuel cell; Indicates fuel cell power command; This indicates the response ratio of the fuel cell, and can be calculated according to... Sure; This represents the fuel cell response time. The proportional response of a fuel cell is dimensionless; therefore, both sides of the equality in this discrete relation are W. ; in, Indicates the prediction step number The corresponding predicted power of the motor group; This represents the thrust power coefficient, with units of W / N; This represents the torque power coefficient, with units of W / (N·m). This represents the three-axis torque command vector; This indicates the motor power offset, measured in watts (W). The thrust power coefficient, torque power coefficient, and motor power offset can be obtained by fitting data from motor bench tests, hovering tests, and conventional turning tests. ; ; in, Indicates the prediction step number The corresponding predicted power of the power battery; Indicates the prediction step number The corresponding percentage of remaining power battery capacity; This indicates the nominal energy of the power battery, measured in J. This represents the power battery efficiency, which is a dimensionless quantity. When the predicted power of the power battery is positive, the proportion of remaining power battery charge decreases with each prediction step; when the predicted power of the power battery is negative, the proportion of remaining power battery charge increases with each prediction step. ; ; in, Indicates the prediction step number The corresponding remaining hydrogen mass; Indicates the prediction step number The corresponding hydrogen consumption rate; Indicates the prediction step number The corresponding DC bus voltage; This indicates the smoothed bus voltage drop sensitivity determined in the current control cycle. The DC bus voltage prediction model uses the smoothed bus voltage drop sensitivity of the current control cycle to convert the predicted power gap into a bus voltage drop estimate, corresponding to the arrow output from the dynamic power boundary generation module 140 to the model prediction control module 150.

[0046] The calibration of the aforementioned discrete prediction model can utilize bench data, hovering data, normal cruise data, regular climb data, regular turning data, and regular wind-resistant correction data from a similar hydrogen-powered UAV 100. After calibration, in the validation data, the errors in fuel cell predicted power, motor group predicted power, DC bus voltage, and remaining battery charge ratio should not exceed their corresponding normalization scales. If any error exceeds its corresponding normalization scale, the initial values ​​for the fuel cell response ratio, thrust power coefficient, torque power coefficient, motor power bias, or smoothing bus voltage drop sensitivity should be recalibrated.

[0047] The control sequence can be represented as ,in Indicates the prediction step number The corresponding control command vector. The control command vector may include total thrust command, three-axis torque command, fuel cell power command, and power battery power command. The model predictive control module 150 uses the control sequence of the previous control cycle as the initial value and solves for a new control sequence in each control cycle.

[0048] In a feasible software interface, the control command vector follows... The order is written into the solver, where Indicates the total thrust command. , and These represent the three components of the three-axis torque command. Indicates the fuel cell power command. This represents the power command for the power battery. After the solver outputs the control sequence, the command execution module reads the current control commands in the same order, thereby avoiding interface confusion between the total thrust command, the three-axis torque command, the fuel cell power command, and the power battery power command.

[0049] The control cost includes at least flight tracking error, control command variation, DC bus voltage deviation, hydrogen consumption, and remaining battery charge deviation. Flight tracking error is used to make the predicted flight state close to the reference flight state vector; control command variation is used to suppress excessively rapid changes in control commands; DC bus voltage deviation is used to suppress the DC bus voltage from deviating from the target bus voltage; hydrogen consumption is used to constrain the utilization efficiency of the remaining hydrogen mass; and remaining battery charge deviation is used to make the remaining battery charge percentage close to the target battery charge percentage. ; in, Indicates the value of control; Indicates the number of prediction steps; This represents the normalized flight tracking error; This indicates a change in normalized control commands; This represents the normalized DC bus voltage deviation; This represents the normalized hydrogen consumption rate; This indicates the deviation of the normalized remaining power battery capacity. Indicates the flight error weight; Indicates the weight of the instruction change; Indicates the bus voltage weight; Indicates hydrogen consumption weight; The weight represents the battery charge weight. The flight error weight, command change weight, bus voltage weight, hydrogen consumption weight, and battery charge weight all apply to the normalized dimensionless quantities, thus the control cost value is a dimensionless scalar.

[0050] Normalized flight tracking error can be calculated according to Sure, This represents the normalized scaling matrix for flight states, where its diagonal elements represent the permissible error scales for the corresponding flight state components. Changes in normalized control commands can be made according to... Sure, This represents the normalized scaling matrix for control commands, with its diagonal elements representing the allowable variation scales for total thrust command, triaxial torque command, fuel cell power command, and power battery power command, respectively. The normalized DC bus voltage deviation can be calculated according to... Sure, This represents the normalized scale of the bus voltage; the normalized hydrogen consumption rate can be calculated according to... Sure, This represents the normalized scale for hydrogen consumption rate; the normalized deviation of remaining battery capacity can be calculated according to... Sure.

[0051] The flight status normalization matrix, control command normalization matrix, bus voltage normalization matrix, and hydrogen consumption rate normalization matrix are determined based on the allowable tracking error, rated control command variation, rated DC bus voltage, and rated hydrogen consumption rate of the same type of hydrogen-powered UAV 100. These normalization matrices are used to convert errors of different dimensions into dimensionless quantities.

[0052] The dynamic power boundary is used to limit the difference between the predicted power of the motor cluster and the predicted power of the fuel cell, as well as the increment of the total thrust command, within the prediction time domain. For each prediction step within the prediction time domain, the model prediction control module 150 ensures that the difference between the predicted power of the motor cluster and the predicted power of the fuel cell does not exceed the allowable power gap. ; in, Indicates the prediction step number The corresponding predicted power of the motor group Indicates the prediction step number The corresponding predicted power of the fuel cell, This represents the allowable power gap determined in the current control cycle. This constraint converts the actual voltage response of the current power supply chain into an upper limit for the supply-demand power difference in future forecast steps; ; in, Indicates the prediction step number The corresponding total thrust command, This indicates the total thrust command corresponding to the previous prediction step. This indicates the thrust increment reduction ratio determined in the current control cycle. This represents the baseline thrust increment. This constraint causes the allowable rate of increase in total thrust over the prediction time domain to vary with the power supply conditions.

[0053] The fixed equipment boundaries include the bus voltage boundary, fuel cell power boundary, battery power boundary, and total thrust boundary. The bus voltage boundary limits the DC bus voltage in the prediction time domain to be no lower than the minimum bus voltage and no higher than the maximum bus voltage; the fuel cell power boundary limits the predicted fuel cell power to be between the minimum and maximum fuel cell power; the battery power boundary limits the actual or predicted battery power to be between the minimum and maximum battery power; and the total thrust boundary limits the total thrust command to be between the minimum and maximum total thrust. ; ; ; ; ; in, Indicates the minimum bus voltage. Indicates the highest voltage of the busbar. Indicates the minimum battery capacity percentage. Indicates the highest percentage of battery capacity. Indicates the minimum power of the fuel cell. Indicates the maximum power of the fuel cell. Indicates the minimum power of the power battery. Indicates the maximum power of the power battery. Indicates the minimum total thrust. This represents the maximum total thrust. The above boundaries can be calibrated based on the model of the hydrogen-powered drone 100, the rated parameters of the fuel cell, the rated parameters of the power battery, the rated parameters of the electronic speed controller, and the rated parameters of the thruster. The minimum bus voltage is used to predict the bus voltage boundary in the time domain, and the low bus voltage safety threshold is used for real-time anomaly triggering; the minimum battery charge percentage is used to predict the remaining power battery charge percentage boundary in the time domain, and the low battery charge safety threshold is used for real-time anomaly triggering.

[0054] The model predictive control module 150 can express the linearized prediction model, control cost, dynamic power boundary, and fixed equipment boundary as a quadratic programming problem and input it into the quadratic programming solver; alternatively, in the nonlinear model predictive control implementation, the prediction model and constraint input sequence are used as the quadratic programming solver. In the linearization implementation, the model predictive control module 150 uses the integrated state vector after the execution of the previous control cycle and the control sequence of the previous control cycle as linearization points to obtain the prediction state update matrix, control input matrix, and constant terms, and writes the normalized control cost into the solver objective function; simultaneously, it writes the allowable power gap constraint, thrust increment reduction ratio constraint, bus voltage boundary, fuel cell power boundary, power battery power boundary, and total thrust boundary into the solver constraint matrix. The inputs of the quadratic programming solver or the sequential quadratic programming solver include the integrated state vector, reference flight state vector, prediction model parameters, normalization scale, control weights, dynamic power boundary, and fixed equipment boundary, and the outputs are the total thrust command, three-axis torque command, fuel cell power command, and power battery power command in the prediction time domain. If the dynamic power boundary makes it impossible to fully track the original mission trajectory, the model prediction control module 150 prioritizes satisfying the minimum bus voltage, power battery power boundary, fuel cell power boundary and total thrust boundary, and reduces altitude error, attitude error and trajectory deviation within these boundaries.

[0055] Example 8: The instruction execution module executes the current control instruction in the control sequence. The current control instruction can be the first control instruction in the control sequence, including the total thrust instruction, three-axis torque instruction, fuel cell power instruction, and power battery power instruction for the current control cycle. Before sending the control instruction to the hybrid propulsion system 170, the instruction execution module determines the difference between the predicted power of the motor group and the predicted power of the fuel cell corresponding to the current control instruction as the power gap corresponding to the instruction.

[0056] When the power gap corresponding to the command does not exceed the allowable power gap, the command execution module executes the current control command. At this time, the current control command satisfies both the flight control objective obtained by the model predictive control module 150 and the allowable power gap limit and thrust increment reduction ratio limit provided by the dynamic power boundary generation module 140. Therefore, the total thrust command, three-axis torque command, fuel cell power command, and power battery power command can be directly sent to the hybrid propulsion system 170.

[0057] When the power gap corresponding to the command exceeds the allowable power gap, the command execution module keeps the direction of the three-axis torque command unchanged, reduces the increment of the total thrust command according to the thrust increment reduction ratio, and simultaneously limits the power command of the power battery. Keeping the direction of the three-axis torque command unchanged ensures that the UAV attitude control direction is consistent with the solution result of the model predictive control module 150; reducing the increment of the total thrust command can reduce the rapid increase of the actual power of the motor group; and simultaneously limiting the power command of the power battery can prevent the power battery power supply branch from continuing to bear excessive transient power when the DC bus voltage is close to the safety boundary.

[0058] The adjusted total thrust command and three-axis torque command can be converted into individual motor speed or ESC duty cycle commands using a conventional control allocation matrix. The control allocation matrix can be pre-calibrated based on the number of rotors, rotor layout, rotor direction, and torque coefficient of the multi-rotor UAV. This conversion is a standard engineering implementation of thruster control allocation. In this implementation, the reduction in motor speed caused by DC bus undervoltage is mitigated by limiting the incremental total thrust command and the power battery power command.

[0059] After the rolling update module completes the execution of the control command in the current control cycle, it waits for a sampling time interval. The UAV flight status and power supply data are reacquired. The airborne controller uses the new measurements as the initial values ​​for the integrated state vector in the next control cycle, and saves the DC bus voltage, actual fuel cell power, actual motor group power, and smoothing bus voltage drop sensitivity of the current control cycle as historical values. The prediction time domain is shifted forward by one control cycle, and the model predictive control module 150 resolves the control sequence in the next control cycle.

[0060] The threshold monitoring and degradation control module is used to initiate power limiting, return-to-home, or landing procedures corresponding to the abnormal conditions when: the DC bus voltage is lower than the low-voltage safety threshold; the fuel cell temperature is higher than the high-temperature safety threshold; the remaining battery charge percentage is lower than the low-charge safety threshold; the remaining hydrogen mass is lower than the low-reserve hydrogen safety threshold; or the communication status meets preset abnormal conditions. The low-voltage safety threshold is not lower than the sum of the minimum bus voltage and the bus safety voltage margin; the high-temperature safety threshold is lower than the maximum allowable temperature of the fuel cell while retaining a temperature protection margin; the low-charge safety threshold is not lower than the minimum battery charge percentage; and the low-reserve hydrogen safety threshold is not lower than the hydrogen mass calculated from the return-to-home energy margin. These relationships ensure that the low-voltage safety threshold, high-temperature safety threshold, low-charge safety threshold, and low-reserve hydrogen safety threshold have consistent calibration standards with fixed equipment boundaries and return-to-home procedures. ; ; ; in, This represents the available return energy for the current control cycle, in J. This represents the energy conversion efficiency of a fuel cell, and is a dimensionless quantity. This indicates the lower heating value of hydrogen, expressed in J / kg. This indicates the predicted return energy, expressed in J. Indicates the number of steps predicted for the return journey; Indicates the return prediction step number The corresponding predicted return power, in W; This indicates the return-to-base prediction time interval, in seconds (s). This represents the remaining return-to-home energy, measured in J. When the available return-to-home energy is not less than the sum of the predicted return-to-home energy and the remaining return-to-home energy, the threshold monitoring and degradation control module determines that the return-to-home process can be completed; otherwise, it determines that the return-to-home process cannot be completed and proceeds to the landing process.

[0061] Preset communication anomaly conditions may include continuous No valid communication frames were received from the data acquisition module or ground station during any of the control cycles. Take a value that is not less than 3 and not greater than 3. integers, This indicates the allowed attitude hold time. The preset continuous control cycle can be no less than 3 and no greater than... Integer control cycle; preset recovery time can be set according to Confirmed, among which This indicates the number of recovery judgment cycles, which is not less than the preset continuous control cycle. The above quantitative relationship ensures that communication anomalies, fuel cell temperature recovery, and return-to-home triggers are all judged using the control cycle as the unit.

[0062] Abnormal conditions and power limiting, return-to-home, or landing procedures can be fixedly mapped according to safety priorities. When the DC bus voltage is lower than the low-voltage safety threshold, the threshold monitoring and degradation control module first enters the power limiting procedure, limiting the allowable power gap to zero or reducing it to below the most recent valid allowable power gap, and prohibiting further increases in the total thrust command increment that would expand the corresponding power gap. When the fuel cell temperature is higher than the fuel cell high-temperature safety threshold, the threshold monitoring and degradation control module enters the power limiting procedure and reduces the rate of increase of the fuel cell power command; if the fuel cell temperature does not recover within a preset continuous control cycle, the return-to-home procedure is entered. When the remaining battery charge percentage is lower than the low battery charge safety threshold or the remaining hydrogen mass is lower than the low hydrogen reserve safety threshold, the threshold monitoring and degradation control module determines whether the return-to-home procedure can be completed based on the available return-to-home energy, predicted return-to-home energy, and return-to-home energy reserve; if the return-to-home procedure cannot be completed, the landing procedure is entered. When the communication status meets the preset communication abnormality conditions and cannot be recovered within the preset recovery time, the threshold monitoring and degradation control module enters the return-to-home procedure; if the positioning status or attitude status is abnormal at the same time, the landing procedure is entered. When multiple abnormal conditions exist simultaneously, the procedure with the highest safety priority shall be selected in the order of landing procedure, return procedure, and power limiting procedure.

[0063] When model predictive control fails to solve, the threshold monitoring and degradation control module uses the most recently valid allowable power gap to limit the total thrust command increment. Model predictive control failure can include the solver failing to return a feasible solution within the allowed computation time, missing predictive model input leading to unmet solution conditions, or the combined effect of fixed equipment boundaries and dynamic power boundaries causing the current reference flight state vector to be unreachable. In this case, the airborne controller can use subsequent commands from the feasible control sequence of the previous control cycle, or switch to a conventional proportional-integral-derivative attitude controller, but still use the most recently valid allowable power gap to limit the total thrust command increment.

[0064] The data logging module records the reference trajectory, actual trajectory, attitude, actual power of the motor cluster, actual power of the fuel cell, actual power of the power battery, DC bus voltage, bus voltage drop sensitivity, smoothing bus voltage drop sensitivity, allowable power deficit, thrust increment reduction ratio, remaining power battery charge percentage, remaining hydrogen mass, and fault status. When the UAV reaches its last waypoint, receives a mission termination command, or enters a safe landing state, the onboard controller gradually reduces propulsion power. After landing confirmation, it stops the motors and reduces fuel cell power and shuts off hydrogen supply according to the shutdown sequence specified by the fuel cell controller.

[0065] As an optional implementation, the hydrogen-powered UAV 100 flies smoothly along the mission trajectory, with minimal changes in the actual power of the motor cluster and the actual power of the fuel cell capable of covering a large proportion of the propulsion power demand. At this point, the current power gap is small, and the power gap increment may be close to zero or negative. The bus voltage drop sensitivity calculation module 130 maintains the smooth bus voltage drop sensitivity from the previous control cycle. The allowable power gap determined by the dynamic power boundary generation module 140 is typically not less than the baseline power gap, and the thrust increment reduction ratio is close to 1. The model predictive control module 150 solves the control sequence within the dynamic power boundary using conventional methods, and the instruction execution module typically executes the current control command directly.

[0066] As an optional implementation, during rapid climb, the mission trajectory requires an increase in target altitude, and the reference flight state vector generated by the reference sequence generation module 110 includes the climb rate or altitude increase requirement. The model predictive control module 150 tends to increase the total thrust command to reduce altitude error, and the actual power of the motor group increases rapidly with the increment of the total thrust command. Since the actual power of the fuel cell is limited by the fuel cell response time and cannot immediately increase to the same level, the current power gap and power gap increment determined by the bus voltage drop sensitivity calculation module 130 increase. If an increase in the bus voltage drop is detected simultaneously, the smoothing bus voltage drop sensitivity increases, the allowable power gap determined by the dynamic power boundary generation module 140 decreases, and the thrust increment reduction ratio decreases accordingly. During rapid climb, the model predictive control module 150 is limited by the allowable power gap and thrust increment reduction ratio when solving the control sequence; if the pre-execution check finds that the power gap corresponding to the command exceeds the allowable power gap, the command execution module reduces the total thrust command increment according to the aforementioned rules, thereby reducing the motor speed decrease caused by DC bus undervoltage.

[0067] As an optional implementation, the remaining battery charge percentage output by the power battery management system is lower than the normal operating range but not lower than the low battery safety threshold. In this case, the same power gap increment may cause a more significant drop in DC bus voltage. The bus voltage drop sensitivity calculation module 130 obtains a greater smooth bus voltage drop sensitivity through actual sampling, and the dynamic power boundary generation module 140 accordingly reduces the allowable power gap and decreases the thrust increment reduction ratio. This process uses dynamic power boundaries to pre-constrain the predicted power and total thrust command increment of the future motor group.

[0068] As an optional implementation, if the DC bus voltage in a certain control cycle significantly exceeds the sensor's range, or if the actual power data of the motor group is missing, the data acquisition module marks the measurement value as invalid. The bus voltage drop sensitivity calculation module 130 does not use the abnormal measurement value of the control cycle to update the smoothed bus voltage drop sensitivity, but instead maintains the most recently valid smoothed bus voltage drop sensitivity. The model predictive control module 150 can still solve the control sequence based on other valid state variables in the integrated state vector; if the anomaly continues to cause the model predictive control solution to fail, the threshold monitoring and degradation control module uses the most recently valid allowable power gap to limit the total thrust command increment.

[0069] When the DC bus voltage is lower than the low-voltage safety threshold of the bus, the threshold monitoring and degradation control module enters the power limiting process according to the aforementioned abnormal condition mapping, and enters the descent process when the DC bus voltage still cannot be restored within the preset continuous control cycle.

[0070] When the remaining hydrogen mass is below the low hydrogen reserve safety threshold or the remaining battery charge percentage is below the low battery charge safety threshold, the threshold monitoring and degradation control module determines whether the return process can be completed based on the available return energy, predicted return energy, and remaining return energy, and accordingly initiates the return or landing process. When the communication status meets the preset communication anomaly conditions, the threshold monitoring and degradation control module executes the aforementioned degradation control according to the preset recovery time and safety priority.

[0071] This implementation can be executed by a single airborne controller, or by a flight controller, an energy management controller, and a motor controller working together. The airborne controller may include a processor, a memory, a communication interface, and an input / output interface. The memory stores a control program that can be executed by the processor. When the control program is executed, it implements the functions of the reference sequence generation module 110, the data acquisition module, the state estimation module, the bus voltage drop sensitivity calculation module 130, the dynamic power boundary generation module 140, the model prediction control module 150, the instruction execution module, the threshold monitoring and degradation control module, the rolling update module, and the data recording module.

[0072] The fuel cell power supply branch may include a fuel cell, a fuel cell-side DC-DC converter, and a fuel cell controller. The fuel cell controller receives power commands from the fuel cell and adjusts the power supplied by the fuel cell-side DC-DC converter to the DC bus. The power battery power supply branch may include a power battery, a bidirectional DC-DC converter, and a power battery management system. The bidirectional DC-DC converter receives power commands from the power battery and, within the allowable range of the power battery, supplements the transient power required by the DC bus or charges it. The motor group is connected to the DC bus via an electronic speed controller (ESC), which receives motor speed or duty cycle commands calculated from the total thrust command and the three-axis torque command.

[0073] The data acquisition module can transmit UAV flight status data, power supply data, and control commands to various sensors and controllers via CAN, UART, Ethernet, or other airborne communication buses. All communication methods are existing airborne communication methods, and this implementation does not limit the specific communication protocol. Regardless of whether a centralized or distributed airborne controller is used, the DC bus voltage, actual fuel cell power, and actual motor group power must be time-aligned under the same current sampling sequence number.

[0074] The multi-rotor rigid body dynamics model in the prediction model can be calibrated based on the UAV's mass, moment of inertia, rotor layout, and control allocation matrix; the motor power model can be calibrated based on the bench test relationship between motor speed, thruster thrust, and input power; the fuel cell power response model can be calibrated based on fuel cell response time and fuel cell power command following characteristics; and the power battery remaining capacity ratio update model can be updated by conventional integration based on the actual power battery power and sampling time interval. Calibration inputs include the rated mass, moment of inertia, rotor layout, fuel cell rated power, fuel cell response time, power battery capacity, maximum charge / discharge power of the power battery, ESC input power record, DC bus voltage record, and hydrogen storage device capacity record for the same type of hydrogen-powered UAV 100. Calibration outputs include prediction model parameters, flight state normalized scaling matrix, control command normalized scaling matrix, bus voltage normalized scaling matrix, hydrogen consumption rate normalized scaling matrix, reference power gap, reference thrust increment, low-voltage safety threshold of the bus, high-temperature safety threshold of the fuel cell, low battery capacity safety threshold, and low hydrogen reserve safety threshold.

[0075] To ensure real-time operation of the control program, the airborne controller can set an allowable computation time. If the model predictive control module 150 completes the quadratic programming solver or sequential quadratic programming solver solution within the allowable computation time, the instruction execution module executes the current control instruction in the control sequence. If the model predictive control module 150 fails to obtain a feasible control sequence after the allowable computation time, the threshold monitoring and degradation control module triggers model predictive control solution failure handling. This handling ensures that the control cycle is not interrupted due to solver delay.

[0076] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A control method for hydrogen-powered unmanned aerial vehicles (UAVs) based on model prediction algorithms, applied to hydrogen-powered UAVs equipped with fuel cells and power batteries, characterized in that... The airborne controller performs the following steps in each control cycle: Acquire mission trajectory, UAV flight status data and power supply data; generate reference flight status vector based on mission trajectory; and generate comprehensive status vector based on UAV flight status data and power supply data. Power supply data includes DC bus voltage, fuel cell actual power and motor group actual power. The current power gap is determined based on the actual power of the motor group and the actual power of the fuel cell, and the power gap increment and bus voltage drop are also determined. The bus voltage drop sensitivity is determined based on the power gap increment and the bus voltage drop, and then smoothed to obtain the smoothed bus voltage drop sensitivity. Here, the bus voltage drop sensitivity represents the degree of DC bus voltage drop when the power gap increases by one unit, and the smoothed bus voltage drop sensitivity represents the degree of bus voltage drop caused by a unit increase in power gap. The allowable power gap and thrust increment reduction ratio are determined based on the DC bus voltage, the minimum bus voltage, the bus safety voltage margin, and the smoothing bus voltage drop sensitivity. The allowable power gap represents the maximum supply and demand power difference allowed under the current bus condition. The thrust increment reduction ratio is used to reduce the total thrust command increment so that the total thrust command increment is reduced when the difference between the DC bus voltage and the minimum bus voltage decreases. Establish a dynamic power boundary to ensure that the difference between the predicted power of the motor group and the predicted power of the fuel cell does not exceed the allowable power gap, and limit the total thrust command increment to the thrust increment reduction ratio. The control sequence is solved based on the integrated state vector, reference flight state vector, prediction model and dynamic power boundary; The control command is executed, and the difference between the predicted power of the motor group and the predicted power of the fuel cell corresponding to the control command is determined as the power gap corresponding to the command. When the power gap corresponding to the command exceeds the allowable power gap, the total thrust command increment and the power battery power command are restricted.

2. The hydrogen-powered unmanned aerial vehicle control method based on model prediction algorithm according to claim 1, characterized in that, The power supply data also includes the actual power of the power battery, the remaining power battery charge percentage, the fuel cell temperature, and the remaining hydrogen mass. The onboard controller stores the DC bus voltage, the actual power of the fuel cell, and the actual power of the motor group for the current control cycle and the previous control cycle.

3. The hydrogen-powered unmanned aerial vehicle control method based on model prediction algorithm according to claim 1, characterized in that, When determining the smoothing bus voltage drop sensitivity, the airborne controller determines the bus voltage drop sensitivity based on the degree of decrease of the bus voltage drop relative to the power gap increment when the power gap increment is positive, and obtains the smoothing bus voltage drop sensitivity through the sensitivity smoothing ratio; when the power gap increment is not positive, the smoothing bus voltage drop sensitivity of the previous control cycle is maintained.

4. The hydrogen-powered unmanned aerial vehicle control method based on model prediction algorithm according to claim 1, characterized in that, When determining the allowable power gap, the airborne controller determines the allowable power gap based on the relationship between the DC bus voltage and the minimum bus voltage and the bus safety voltage margin, as well as the degree of bus voltage drop caused by the increase in unit power gap, as indicated by the smooth bus voltage drop sensitivity, so as to ensure that the DC bus voltage retains the bus safety voltage margin.

5. The hydrogen-powered unmanned aerial vehicle control method based on model prediction algorithm according to claim 1, characterized in that, When generating the thrust increment reduction ratio, the airborne controller determines the thrust increment reduction ratio based on the relationship between the allowable power gap and the pre-calibrated reference power gap, so that the total thrust command increment is reduced when the difference between the DC bus voltage and the minimum bus voltage is reduced.

6. The hydrogen-powered unmanned aerial vehicle control method based on model prediction algorithm according to claim 1, characterized in that, When solving the control sequence, the control cost includes at least flight tracking error, control command change, DC bus voltage deviation, hydrogen consumption, and remaining battery charge deviation. The fixed equipment boundary includes the bus voltage boundary, fuel cell power boundary, battery power boundary, and total thrust boundary. The airborne controller inputs the integrated state vector, reference flight state vector, prediction model, control cost, fixed equipment boundary, allowable power gap, and thrust increment reduction ratio into the quadratic programming solver or sequence quadratic programming solver to obtain the total thrust command, three-axis torque command, fuel cell power command, and battery power command in the prediction time domain.

7. The hydrogen-powered unmanned aerial vehicle control method based on model prediction algorithm according to claim 1, characterized in that, When executing the current control command, the airborne controller executes the current control command if the power gap corresponding to the command does not exceed the allowable power gap; if the power gap corresponding to the command exceeds the allowable power gap, it keeps the direction of the three-axis torque command unchanged, reduces the total thrust command increment according to the thrust increment reduction ratio, and simultaneously limits the power battery power command.

8. A hydrogen-powered drone control system based on model prediction algorithms, applied to hydrogen-powered drones with fuel cell power supply branches, power battery power supply branches, DC buses, and motor groups, characterized in that it includes: The reference sequence generation module is used to read the mission track and generate a reference flight state vector; The data acquisition module is used to simultaneously collect UAV flight status data and power supply data. The power supply data includes DC bus voltage, actual fuel cell power, and actual motor group power. The state estimation module generates a comprehensive state vector based on UAV flight state data and power supply data. The bus voltage drop sensitivity calculation module determines the current power gap based on the actual power of the motor cluster and the actual power of the fuel cell, determines the power gap increment based on the current power gap, determines the bus voltage drop based on the DC bus voltage, determines the bus voltage drop sensitivity based on the power gap increment and the bus voltage drop, and smooths the bus voltage drop sensitivity to obtain the smoothed bus voltage drop sensitivity. Here, the bus voltage drop sensitivity represents the degree of DC bus voltage drop when the power gap increases by one unit, and the smoothed bus voltage drop sensitivity represents the degree of bus voltage drop caused by a unit increase in power gap. The dynamic power boundary generation module is used to determine the allowable power gap based on the DC bus voltage, the minimum bus voltage, the bus safety voltage margin, and the smoothing bus voltage drop sensitivity, and to generate the thrust increment reduction ratio. The dynamic power boundary is used to limit the difference between the predicted power of the motor group and the predicted power of the fuel cell in the prediction time domain, as well as the total thrust command increment. Among them, the allowable power gap represents the maximum supply and demand power difference allowed by the current bus state, and the thrust increment reduction ratio is used to reduce the total thrust command increment so that the total thrust command increment decreases when the difference between the DC bus voltage and the minimum bus voltage decreases. The model predictive control module is used to solve the control sequence based on the integrated state vector, the reference flight state vector, the predictive model, the allowable power gap, and the thrust increment reduction ratio. The instruction execution module is used to execute the current control instruction in the control sequence, determine the difference between the predicted power of the motor group and the predicted power of the fuel cell corresponding to the current control instruction as the power gap corresponding to the instruction, and limit the total thrust instruction increment and the power battery power instruction when the power gap corresponding to the instruction exceeds the allowable power gap.

9. The hydrogen-powered unmanned aerial vehicle control system based on model prediction algorithm according to claim 8, characterized in that, The data acquisition module is communicatively connected to the inertial measurement unit, satellite positioning receiver, barometric altimeter, electronic speed controller, DC bus sensor, fuel cell controller, power battery management system, and hydrogen storage measurement device. The hydrogen storage measurement device is used to collect the remaining hydrogen mass. The data acquisition module performs time alignment of the UAV flight status data and the power supply data according to a unified clock.

10. The hydrogen-powered unmanned aerial vehicle control system based on model prediction algorithm according to claim 9, characterized in that, It also includes a threshold monitoring and degradation control module, which is used to cause the UAV to enter the power limiting, return, or landing process corresponding to the abnormal conditions when the DC bus voltage is lower than the low-voltage safety threshold, the fuel cell temperature is higher than the high-temperature safety threshold, the remaining power battery charge ratio is lower than the low-charge safety threshold, the remaining hydrogen mass is lower than the low-reserve hydrogen safety threshold, or the communication status meets the preset abnormal conditions. In the event that the model predictive control solution fails, the most recent valid allowable power gap is used to limit the increment of the total thrust command.

Images