Fuel cell maximum power point tracking method for unmanned aerial vehicle applications

Abstract

The application relates to a fuel cell maximum power point tracking method for unmanned aerial vehicle application and belongs to the field of unmanned aerial vehicle energy management.The method is realized as follows: a fuel cell and a DC / DC converter coupling model is established, and fuel cell state information is collected in real time.A fuel cell maximum power point tracking controller is constructed based on a deep deterministic policy gradient (DDPG) algorithm, a DC / DC converter control signal is output through a deep reinforcement learning strategy network, and online dynamic adjustment of a fuel cell working point is realized.A cooperative power distribution mechanism between a fuel cell and a lithium battery is constructed in combination with unmanned aerial vehicle flight power and lithium battery SOC state, so that efficient operation of the fuel cell and safety of the lithium battery are guaranteed.The fuel cell state information includes output voltage and current.The application can realize rapid and self-adaptive tracking of the fuel cell maximum power point under complex flight conditions, improve fuel cell energy utilization efficiency, reduce hydrogen consumption, and enhance the energy supply stability of the unmanned aerial vehicle energy system.

Description

Technical Field

[0001] This invention relates to a maximum power point tracking method for fuel cells for unmanned aerial vehicle (UAV) applications, belonging to the field of UAV energy management. Background Technology

[0002] In recent years, with the widespread application of drones in logistics, disaster monitoring, aerial inspection, and long-endurance missions, the insufficient endurance of traditional lithium-ion battery drones has become increasingly prominent. Hydrogen fuel cells, due to their high energy density, long endurance, and environmental friendliness, are gradually becoming an important energy source for long-endurance drones. However, fuel cells exhibit significant nonlinear output characteristics, with their maximum power point dynamically drifting with changes in load, hydrogen supply pressure, operating temperature, and flight conditions. If the fuel cell cannot operate stably near its maximum power point for extended periods, it will lead to decreased energy efficiency, increased hydrogen consumption, and reduced power supply stability. Existing maximum power point tracking methods for fuel cells mainly include the perturbation-observation method, the incremental conductance method, and the model prediction method. The traditional perturbation-observation method is prone to steady-state oscillations and has a slow tracking speed under dynamic conditions; the incremental conductance method is sensitive to measurement noise; while the model prediction method offers high control accuracy, it is highly dependent on the system model and has high real-time computational complexity. In recent years, deep reinforcement learning, with its strong nonlinear approximation and online decision-making capabilities, has been increasingly applied to the field of energy optimization control. However, most existing research focuses on offline optimization or static condition control, with limited research on real-time adaptive tracking of the fuel cell's maximum power point under dynamic flight conditions of UAVs. Therefore, there is an urgent need to propose a method that can adapt to dynamic load changes in UAVs and achieve online adaptive tracking of the fuel cell's maximum power point to improve fuel cell energy utilization efficiency and UAV endurance. Summary of the Invention

[0003] To address the issues of low tracking accuracy, slow response speed, and insufficient robustness of existing fuel cell maximum power point tracking methods under dynamic conditions of unmanned aerial vehicles (UAVs), this invention aims to provide a fuel cell maximum power point tracking method for UAV applications. This method can achieve rapid and adaptive tracking of the fuel cell's maximum power point under complex flight conditions, thereby improving fuel cell energy utilization efficiency, reducing hydrogen consumption, and enhancing the power supply stability of the UAV's energy system.

[0004] The objective of this invention is achieved through the following technical solution:

[0005] This invention discloses a fuel cell maximum power point tracking (MPPT) method for UAV applications. It establishes a coupling model between the fuel cell and a DC / DC converter, and collects fuel cell state information in real time. A fuel cell MPPT controller is constructed based on the Deep Deterministic Policy Gradient (DDPG) algorithm. A deep reinforcement learning policy network outputs control signals to the DC / DC converter, enabling online dynamic adjustment of the fuel cell operating point. Combining the UAV's flight power and the lithium battery's state of charge (SOC), a collaborative power allocation mechanism between the fuel cell and the lithium battery is constructed to ensure efficient fuel cell operation and lithium battery safety. The fuel cell state information includes output voltage and current.

[0006] This invention discloses a fuel cell maximum power point tracking method for unmanned aerial vehicle (UAV) applications, comprising the following steps:

[0007] Step 1: Status Information Acquisition; Real-time acquisition of fuel cell output voltage Output current .

[0008] Step 2: Establish the power relationship between fuel cells and hybrid electric drones;

[0009]

[0010] in, For the power required by the drone, For fuel cell output power, For DC / DC converter efficiency, This refers to the output power of the lithium battery.

[0011] Step 3: Establish a deep reinforcement learning controller; the controller uses the fuel cell output voltage, output current, and operating point deviation as state inputs, and the state space representation of the state inputs is as follows:

[0012]

[0013] in, This is the current state. for The deviation of the current operating point of the fuel cell from its maximum power point at any given time is expressed as:

[0014]

[0015] in, and These represent the changes in fuel cell output voltage and output current within adjacent sampling periods; when The deep reinforcement learning controller adjusts the control current of the DC / DC converter, increasing the fuel cell output voltage and thus driving the operating point toward the maximum power point; when The controller adjusts the control current of the DC / DC converter, reducing the fuel cell output voltage and driving the operating point towards the maximum power point; when If the fuel cell operating point is determined to be at the maximum power point, the controller will maintain the current control strategy unchanged.

[0016] The action output of the deep reinforcement learning controller is the control current of the DC / DC converter, and its action space... Represented as:

[0017]

[0018] in, for The DC / DC converter controls the current at all times; by adjusting... This changes the operating point of the fuel cell, causing it to move towards its maximum power point.

[0019] Step 4: Execute maximum power point tracking control for the fuel cell; within each sampling period, record the current state. The input is a deep reinforcement learning controller, which outputs a DC / DC converter control current command from the policy network. The output current of the fuel cell is:

[0020]

[0021] in, This is the DC bus voltage;

[0022] Calculate the current fuel cell output power based on the adjusted fuel cell output voltage and output current:

[0023]

[0024] Step 5: Design a collaborative power allocation mechanism to allocate power in the hybrid energy system based on the power demand of the UAV, the maximum power point of the fuel cell, and the SOC state of the lithium battery, thereby achieving power allocation between the fuel cell and the lithium battery.

[0025] Step Six: Update the control strategy online; update the status information and reward function in real time during each sampling period, continuously optimize the maximum power point tracking strategy through online learning, and realize the maximum power point tracking of the fuel cell based on the optimized maximum power point tracking strategy.

[0026] Furthermore, the power allocation criteria described in step five are as follows:

[0027] ① When SOC > SOC H At that time, the output power of the fuel cell is controlled to equal the demand power. Control the lithium battery output power to zero. And prohibit lithium battery charging;

[0028] ②When SOC L <SOC<SOC H At that time, the fuel cell is controlled to perform maximum power point tracking to output maximum power. And control the output power of the lithium battery to supplement the remaining power demand. ;

[0029] ③ When SOC < SOC L When the energy is depleted, the drone will perform an emergency landing.

[0030] Where SOC represents the remaining capacity of the lithium battery. H This refers to the upper limit threshold of the State of Charge (SOC) of a lithium battery. L This is the lower limit threshold of the SOC of lithium batteries. This represents the output power corresponding to the maximum power point of the fuel cell.

[0031] Furthermore, the reward function is:

[0032]

[0033] in, For the total reward, As a power ratio bonus, Rewards are given for trends in power changes. As a reward weighting coefficient, This represents the change in fuel cell output power within adjacent sampling periods. This is the ratio of the current output power to the maximum power.

[0034] Beneficial effects

[0035] 1. This invention discloses a fuel cell maximum power point tracking method for unmanned aerial vehicle (UAV) applications. Addressing the dynamic drift of the fuel cell's maximum power point caused by changes in load power and operating conditions during UAV flight, this method enables online adaptive tracking and adjustment of the fuel cell's operating point. This invention provides a control basis for the efficient operation, power distribution, and energy management of fuel cells in UAV hybrid energy systems, thereby improving energy utilization efficiency, power supply stability, and flight endurance.

[0036] 2. This invention discloses a fuel cell maximum power point tracking method for UAV applications. Addressing the problem of difficulty in real-time and stable tracking of the fuel cell's maximum power point under complex operating conditions, a maximum power point tracking controller is constructed. This controller dynamically adjusts the fuel cell's output operating point through a DC / DC converter, achieving online adaptive tracking of the fuel cell's maximum power point, reducing reliance on an accurate fuel cell model and online computational complexity.

[0037] 3. The present invention discloses a fuel cell maximum power point tracking method for UAV applications. It addresses the problem of unreasonable power distribution between fuel cells and lithium batteries under dynamic loads of UAVs and the difficulty in fully utilizing fuel cells. It constructs a collaborative power distribution mechanism that combines the fuel cell maximum power point tracking results with the lithium battery SOC state, which can reduce fuel cell hydrogen consumption, ensure lithium battery safety, and improve the power supply stability and endurance of UAVs. Attached Figure Description

[0038] Figure 1 This is a flowchart of a fuel cell maximum power point tracking method for drone applications.

[0039] Figure 2 This is a comparison chart of the maximum power point tracking results for fuel cells. Figure 2 (a) represents the changes in power demand and temperature; Figure 2 (b) represents the output power of the fuel cell; Figure 2 (c) represents the output power of the lithium battery; Figure 2 (d) shows the change in SOC of the lithium battery. Detailed Implementation

[0040] To better illustrate the purpose and advantages of this invention, the following description, in conjunction with the accompanying drawings and simulation examples, further explains the content of this invention.

[0041] The sampling time of the fuel cell UAV energy management method based on deep reinforcement learning maximum power point tracking is 0.1 s. A deep deterministic policy gradient algorithm was developed based on MATLAB, and related experiments were conducted on a hardware-in-the-loop test platform. This test platform consists of a 500 W fuel cell system and a 6s2p lithium battery pack, with an ITECH electronic load used to replace the power demand of the UAV. The main parameters of the fuel cell and lithium battery are shown in Table 1.

[0042] Table 1. Main parameters of lithium batteries and fuel cells

[0043]

[0044] To verify the feasibility and beneficial effects of the fuel cell maximum power point tracking method for UAV applications disclosed in this invention, the technical solution of this invention is clearly and thoroughly described below in a case study. The flowchart of the method is as follows: Figure 1 As shown.

[0045] like Figure 1 As shown in the figure, this example discloses a fuel cell maximum power point tracking method for UAV applications. The specific implementation steps are as follows:

[0046] Step 1: Status Information Acquisition; Real-time acquisition of fuel cell output voltage Output current Bus voltage Drone payload power and DC / DC converter status information.

[0047] Step 2: Establish a fuel cell system model and a hybrid electric drone model; the fuel cell output voltage is affected by Nernst voltage, activation loss, ohmic loss, and concentration loss. Therefore, the fuel cell model output voltage can be expressed as:

[0048]

[0049] in, The number of fuel cell units. For Nernst voltage, To reduce activation loss, For Ohm's loss, For concentration loss;

[0050] A three-dimensional point mass model is used to describe the energy consumption of a hybrid-powered UAV:

[0051]

[0052] in, For horizontal coordinates and height, For flight speed, For the track angle, For heading angle, and These are the vertical load factor and the horizontal load factor, respectively. For the thrust of the propeller, For aerodynamic drag, For the quality of drones, It is the acceleration due to gravity;

[0053] During the drone's flight, the propulsion system provides corresponding thrust according to the flight status to overcome flight drag and maintain stable flight; the drone's power requirements are shared by the fuel cell and lithium battery, and their power relationship is expressed as follows:

[0054]

[0055] in, For the efficiency of the propeller, The power consumed by airborne equipment. For fuel cell power, For lithium battery power, For DC / DC converter efficiency.

[0056] Step 3: Establish a deep reinforcement learning controller; the controller uses the fuel cell output voltage, output current, and operating point deviation as state inputs, and its state space is represented as follows:

[0057]

[0058] in, This is the current state. for The deviation of the current operating point of a fuel cell from its maximum power point at any given time can be expressed as:

[0059]

[0060] in, and These represent the changes in fuel cell output voltage and output current within adjacent sampling periods; when The controller adjusts the control current of the DC / DC converter to increase the output voltage of the fuel cell, thereby driving the operating point towards the maximum power point; when The controller adjusts the control current of the DC / DC converter, reducing the fuel cell output voltage and thus driving the operating point towards the maximum power point; when If the fuel cell operating point is determined to be at the maximum power point, the controller will maintain the current control strategy unchanged.

[0061] The action output of the deep reinforcement learning controller is the control current of the DC / DC converter, and its action space is... Represented as:

[0062]

[0063] in, for The DC / DC converter controls the current at all times; by adjusting... This changes the operating point of the fuel cell, causing it to move towards its maximum power point.

[0064] Step 4: Execute maximum power point tracking control for the fuel cell; within each sampling period, record the current state. The input is a deep reinforcement learning controller, which outputs a DC / DC converter control current command from the policy network. Therefore, the output current of the fuel cell can be obtained as:

[0065]

[0066] Calculate the current fuel cell output power based on the adjusted fuel cell output voltage and output current:

[0067]

[0068] Step 5: Design a collaborative power allocation mechanism to achieve power distribution between the fuel cell and the lithium battery; allocate power to the hybrid energy system based on the UAV's power demand, the fuel cell's maximum power point, and the lithium battery's state of charge (SOC); the power allocation criteria are as follows:

[0069] ① When SOC > SOC H At that time, the output power of the fuel cell is controlled to equal the demand power. Control the lithium battery output power to zero. And prohibit lithium battery charging;

[0070] ②When SOC L <SOC<SOC H At that time, the fuel cell is controlled to perform maximum power point tracking to output maximum power. And control the output power of the lithium battery to supplement the remaining power demand. ;

[0071] ③ When SOC < SOC L When the energy is depleted, the drone will perform an emergency landing.

[0072] Where SOC represents the remaining capacity of the lithium battery. H This refers to the upper limit threshold of the State of Charge (SOC) of a lithium battery. L This is the lower limit threshold of the SOC of lithium batteries. This represents the output power corresponding to the maximum power point of the fuel cell.

[0073] Step Six: Update the control strategy online; update the state information and reward function in real time during each sampling period, and continuously optimize the maximum power point tracking strategy through online learning; output the maximum power point tracking control result based on the optimized maximum power point tracking strategy;

[0074] The reward function is:

[0075]

[0076] in, For the total reward, As a power ratio bonus, Rewards are given for trends in power changes. As a reward weighting coefficient, This represents the change in fuel cell output power within adjacent sampling periods. This is the ratio of the current output power to the maximum power.

[0077]

[0078] The results obtained based on the maximum power point tracking method for fuel cells for UAV applications disclosed in this invention are as follows: Figure 2 As shown in Tables 2 and 3. Figure 2The maximum power point tracking (MPPT) results of the fuel cell obtained by the method of this invention are compared with those of typical energy management strategies such as nonlinear model predictive control (NMPC) and fuzzy logic state machine (FLSM). The results show that, under the same power demand and operating temperature, the method of this invention can maintain the fuel cell output power at a higher and more stable level, thereby improving the utilization efficiency of the fuel cell. Simultaneously, since the fuel cell bears more of the power demand, the compensation power required by the lithium battery is correspondingly reduced, resulting in a smaller decrease in the lithium battery's state of charge (SOC) compared to NMPC and FLSM. This reduces lithium battery energy consumption and improves the range of the hybrid energy system. Tables 2 and 3 compare the hybrid energy system consumption results obtained by the method of this invention with those of NMPC and FLSM. The simulation results in Table 2 show that the fuel cell hydrogen consumption and lithium battery SOC consumption obtained by this invention are lower. The three strategies consume 3.6g, 4.0g, and 4.2g of hydrogen, respectively, and the lithium battery SOC decreases by 50.9%, 52.1%, and 55.9%, respectively. The corresponding total consumption of the hybrid energy system was 8.4g, 9.4g, and 11.0g, respectively. Compared with NMPC and FLSM, under the same conditions, the method of the present invention saved 10.6% and 23.6% of the hybrid energy system consumption, respectively. The experimental results in Table 3 show that the method of the present invention results in lower hydrogen consumption and lithium battery SOC consumption for the fuel cell. The three strategies consumed 4.3g, 4.7g, and 4.8g of hydrogen, respectively, and the lithium battery SOC decreased by 48.6%, 50.3%, and 55.4%, respectively. The corresponding total consumption of the hybrid energy system was 9.7g, 11.2g, and 12.9g, respectively. Compared with NMPC and FLSM, under the same conditions, the method of the present invention saved 13.4% and 24.8% of the hybrid energy system consumption, respectively.

[0079] The maximum power point tracking (MPPT) results obtained by this invention exhibit good dynamic adaptability and power regulation, reflecting the dynamic characteristics of fuel cell output capacity changing with operating conditions during UAV flight. Applying these tracking results to online energy management of UAV hybrid energy systems allows for real-time adjustment of fuel cell output power and provides a basis for power allocation between the fuel cell and lithium battery. Compared to NMPC and FLSM methods, this invention reduces fuel cell output power fluctuations and the risk of lithium battery over-discharge, improves the responsiveness of energy management strategies to dynamic load changes and the rationality of power allocation, thereby enhancing the energy utilization efficiency, power supply stability, and endurance of UAV hybrid energy systems.

[0080] The above detailed description is a further explanation of the purpose, technical solution and beneficial effects of the invention. It should be understood that the above description is only a specific implementation example of the present invention and is only used to explain the present invention. It is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A maximum power point tracking method for fuel cells for unmanned aerial vehicle (UAV) applications, characterized in that, Establish a coupling model between the fuel cell and the DC / DC converter, and collect fuel cell status information in real time; A fuel cell maximum power point tracking controller is constructed based on the deep deterministic policy gradient (DDPG) algorithm. A DC / DC converter control signal is output through a deep reinforcement learning policy network to achieve online dynamic adjustment of the fuel cell operating point. A collaborative power distribution mechanism between the fuel cell and the lithium battery is constructed by combining the UAV flight power and the lithium battery SOC state to ensure the efficient operation of the fuel cell and the safety of the lithium battery. The fuel cell state information includes output voltage and current.

2. The fuel cell maximum power point tracking method for UAV applications as described in claim 1, characterized in that, Includes the following steps: Step 1: Status Information Acquisition; Real-time acquisition of fuel cell output voltage Output current ; Step 2: Establish the power relationship between fuel cells and hybrid electric drones; in, For the power required by the drone, For fuel cell output power, For DC / DC converter efficiency, This refers to the output power of the lithium battery. Step 3: Establish a deep reinforcement learning controller; the controller uses the fuel cell output voltage, output current, and operating point deviation as state inputs, and the state space representation of the state inputs is as follows: in, This is the current state. for The deviation of the current operating point of the fuel cell from its maximum power point at any given time is expressed as: in, and These represent the changes in fuel cell output voltage and output current within adjacent sampling periods; when The deep reinforcement learning controller adjusts the control current of the DC / DC converter, increasing the fuel cell output voltage and thus driving the operating point toward the maximum power point; when The controller adjusts the control current of the DC / DC converter, reducing the fuel cell output voltage and driving the operating point towards the maximum power point; when If the fuel cell operating point is determined to be at the maximum power point, the controller will maintain the current control strategy unchanged. The action output of the deep reinforcement learning controller is the control current of the DC / DC converter, and its action space... Represented as: in, for The DC / DC converter controls the current at all times; by adjusting... This changes the fuel cell's output operating point, causing the fuel cell to move towards its maximum power point. Step 4: Execute maximum power point tracking control for the fuel cell; within each sampling period, record the current state. The input is a deep reinforcement learning controller, which outputs a DC / DC converter control current command from the policy network. The output current of the fuel cell is: in, This is the DC bus voltage; Calculate the current fuel cell output power based on the adjusted fuel cell output voltage and output current: ； Step 5: Design a collaborative power allocation mechanism to allocate power in the hybrid energy system based on the power demand of the UAV, the maximum power point of the fuel cell, and the SOC state of the lithium battery, thereby achieving power allocation between the fuel cell and the lithium battery. Step Six: Update the control strategy online; update the status information and reward function in real time during each sampling period, continuously optimize the maximum power point tracking strategy through online learning, and realize the maximum power point tracking of the fuel cell based on the optimized maximum power point tracking strategy.

3. The fuel cell maximum power point tracking method for UAV applications as described in claim 2, characterized in that, The power allocation criteria described in step five are as follows: ① When SOC > SOC H At that time, the output power of the fuel cell is controlled to equal the demand power. Control the lithium battery output power to zero. And prohibit lithium battery charging; ②When SOC L <SOC<SOC H At that time, the fuel cell is controlled to perform maximum power point tracking to output maximum power. And control the output power of the lithium battery to supplement the remaining power demand. ; ③ When SOC < SOC L When the energy is depleted, the drone will perform an emergency landing. Where SOC represents the remaining capacity of the lithium battery. H This refers to the upper limit threshold of the State of Charge (SOC) of a lithium battery. L This is the lower limit threshold of the SOC of lithium batteries. This represents the output power corresponding to the maximum power point of the fuel cell.

4. The maximum power point tracking method for fuel cells for UAV applications as described in claim 3, characterized in that, The reward function is: in, For the total reward, As a power ratio bonus, Rewards are given for trends in power changes. As a reward weighting coefficient, This represents the change in fuel cell output power within adjacent sampling periods. This is the ratio of the current output power to the maximum power.

Images