A method of energy management for a distributed hybrid electric propulsion aircraft

Abstract

The application provides an energy management method of a distributed hybrid electric propulsion aircraft, which comprises the following steps: before the execution of a flight task, performing global optimization on hyperparameters of a DDPG algorithm by using an SSA algorithm; training and outputting an optimal hyperparameter combination of the DDPG algorithm under the current flight task by using the SSA algorithm; initializing the DDPG algorithm by using the optimal hyperparameters and performing policy learning and optimization by using an Actor-Critic network structure; during the execution of the flight task, the DDPG algorithm receives real-time environmental information and dynamically outputs an optimal energy management strategy with the objective of minimizing equivalent fuel consumption. By using the SSA-DDPG collaborative optimization framework, the application effectively solves the unstable training problem of a traditional DDPG algorithm caused by hyperparameter sensitivity, and can significantly improve the convergence speed, stability, safety and real-time performance of the energy management strategy.

Description

Technical Field

[0001] This invention relates to the field of energy management technology for hybrid electric propulsion aircraft, and more specifically to an energy management method for distributed hybrid electric propulsion aircraft. Background Technology

[0002] Distributed hybrid electric propulsion systems are power systems that use engines to drive generators to produce electricity, which, combined with energy storage devices such as batteries, directly converts electrical energy into mechanical energy to drive multiple propellers or ducted fans, providing propulsion for the aircraft. The hybrid electric propulsion characteristics of distributed hybrid electric propulsion systems offer advantages such as improved propulsion efficiency, reduced fuel consumption, reduced emissions, and reduced noise. Simultaneously, the distributed configuration design gives the aircraft high reliability and safety. However, current distributed hybrid electric propulsion aircraft still face challenges such as immature battery technology and low battery density, resulting in short range. Therefore, research on energy management for distributed hybrid electric propulsion aircraft has become a key technology for solving this problem.

[0003] Energy management strategies for distributed hybrid electric propulsion aircraft have high requirements for safety, flexibility, and real-time performance. Currently, mainstream rule-based and optimization-based energy management methods are unable to simultaneously meet these requirements.

[0004] In recent years, learning-based energy management methods have shown great potential, especially the Deep Deterministic Policy Gradient (DDPG) algorithm, which can learn policies in a continuous action space. It exhibits strong learning ability and adaptability under various operating conditions, ensuring policy safety and flexibility while also consuming relatively few computational resources and offering a degree of real-time performance. However, the training process of DDPG is highly sensitive to hyperparameters, including the learning rate, discount factor, noise figure, soft update coefficient, and experience pool sampling size. Improper hyperparameter settings can lead to training instability, slow convergence, and large fluctuations in policy performance, making it difficult to guarantee a stable and efficient energy management policy in actual flight missions.

[0005] Sparrow Search Algorithm (SSA) is a class of intelligent swarm optimization algorithms that have been developed in recent years. It has the characteristics of strong global optimization ability, few parameters, simple structure, and fast convergence speed, and is suitable for high-dimensional nonlinear optimization problems.

[0006] Therefore, there is an urgent need for an energy management method for distributed hybrid electric propulsion aircraft that combines the advantages of SSA and DDPG to overcome the shortcomings of existing technologies, ensure the safety and flexibility of the strategy, and improve the stability and overall energy efficiency of the distributed hybrid electric propulsion system. Summary of the Invention

[0007] In view of this, in order to solve or at least partially solve the above-mentioned technical problems, the present invention provides an energy management method for a distributed hybrid electric propulsion aircraft, which can ensure a stable and efficient energy management strategy in actual flight missions and help improve the stability and overall energy efficiency of the distributed hybrid electric propulsion system.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] In a first aspect, the present invention provides an energy management method for a distributed hybrid electric propulsion aircraft, the method comprising the following steps:

[0010] S1. Obtain flight mission profile information and environmental simulation information through the distributed hybrid electric propulsion system module;

[0011] S2. Before the flight mission is executed, the flight mission profile information is used as input in the offline optimization module to initialize the SSA algorithm.

[0012] S3. Train the DDPG algorithm using the SSA algorithm, and find and output the optimal hyperparameters of the DDPG algorithm for the current flight mission;

[0013] S4. Initialize the DDPG algorithm based on the optimal hyperparameters;

[0014] S5. During flight mission execution, the online optimization module receives real-time dynamic environmental information. The DDPG algorithm outputs an energy management strategy through the Actor network and evaluates, optimizes, and updates the output energy management strategy through the Critic network. When the iteration termination condition is met, the optimal energy management strategy with the goal of minimizing equivalent fuel consumption is output.

[0015] Furthermore, in S1, the distributed hybrid electric propulsion power system module includes: a battery, an engine, a generator, an energy management controller, an electric propulsion device, and an aircraft environmental observation module. The acquired flight mission profile information includes: flight phase structure information, flight condition information, environmental and meteorological information, energy system load and constraint information, and mission and time structure information.

[0016] Furthermore, in S2, the fitness function of the initialized SSA algorithm is... for:

[0017]

[0018] in, , , These are the weighting coefficients for fuel, battery SOC, and power error, respectively. The battery is in its state of charge. For the desired SOC; For engine power, For actual output propulsion power, Battery charging and discharging power; For engine efficiency; For random disturbance terms; This indicates the low calorific value of the fuel; k represents the kth segment of the flight mission, and K represents the complete flight mission. The time index representing the flight mission profile. Indicates the total flight mission time. This represents the time interval of the k-th flight mission. This represents the required propulsion power for the k-th segment of the flight mission.

[0019] Furthermore, in S3, the hyperparameters of the DDPG algorithm include the Actor network learning rate. Critic network learning rate Explore noise intensity Explore noise attenuation Discount Factor Soft update coefficient The number of samples B is drawn for each training session.

[0020] Furthermore, the constraints for the number of samples B drawn in each training iteration include:

[0021] in, , , , , , , Let A represent the learning rate of the Actor network, the learning rate of the Critic network, the exploration noise intensity, the exploration noise attenuation, the discount factor, the soft update coefficient, and the number of samples drawn in each training iteration for the i-th sparrow individual, respectively. and These are the minimum and maximum learning rates of the Actor network, respectively. and These are the minimum and maximum learning rates of the Critic network, respectively. and These represent the minimum and maximum values ​​of the noise intensity, respectively. and These are the minimum and maximum values ​​for exploring noise attenuation, respectively. and These are the minimum and maximum values ​​of the discount factor, respectively. and These are the minimum and maximum values ​​of the soft update coefficient, respectively. and These represent the minimum and maximum number of samples drawn in each training session, respectively.

[0022] Furthermore, in S3, the SSA algorithm in the offline optimization module applies each randomly generated set of DDPG algorithm hyperparameters to the DDPG algorithm under simulated flight conditions, and determines whether to terminate the loop and output the optimal solution based on fitness evaluation, including the following steps:

[0023] (1) Initialize the sparrow population individuals, and the Actor network learning rate. Critic network learning rate Explore noise intensity Explore noise attenuation Discount Factor Soft update coefficient Each training session extracts a sample size B as the state input to form a single sparrow.

[0024] (2) Map the sparrow population to the DDPG algorithm;

[0025] (3) Design a fitness function to classify the roles of discoverer, joiner, and vigilant;

[0026] (4) Update the location and fitness of the discoverer, joiner, and vigilant;

[0027] (5) Perform fitness assessment and update the best and worst positions;

[0028] (6) Determine whether the termination condition is met. If the termination condition is met, stop the search and output the result; otherwise, repeat steps (1) to (5).

[0029] Furthermore, the termination condition of the SSA algorithm is:

[0030]

[0031] in, This indicates the termination condition of the SSA algorithm. For the first The global optimal fitness in the next iteration. For the first The global optimal fitness in the next iteration. This represents the rate of change in fitness.

[0032] Furthermore, in S5, the reward function of the DDPG algorithm is:

[0033]

[0034] in, Let be the reward function, representing the th The reward obtained in the next iteration , , These are the weighting coefficients for fuel, battery SOC, and power error, respectively. Indicates the first The actual output propulsion power of the next iteration Indicates the first The demand for power in the next iteration Indicates the first Battery state of charge in the next iteration Indicates the first The expected SOC of the next iteration Indicates the first Equivalent fuel consumption in the next iteration. The weighting factor for power smoothness. Indicates the first The actual output propulsion power of the next iteration.

[0035] Furthermore, the termination condition Φ` of the DDPG algorithm is:

[0036]

[0037] Where Φ` represents the termination condition of the DDPG algorithm. For the first The reward obtained in the next iteration Let T represent the average rate of change of return, and T represent the total number of iterations.

[0038] Secondly, the present invention also provides an energy management system for a distributed hybrid electric propulsion aircraft. The system includes a distributed hybrid electric propulsion power system module, an offline optimization module, and an online optimization module, and uses the above-described method to obtain the optimal energy management strategy.

[0039] As can be seen from the above technical solution, compared with the prior art, the present invention provides an energy management method and system for a distributed hybrid electric propulsion aircraft. This method employs an SSA-DDPG algorithm collaborative framework, mainly including: before the flight mission is executed, using the SSA algorithm to globally optimize the hyperparameters of the DDPG algorithm; training and outputting the optimal hyperparameter combination of the DDPG algorithm under the current flight mission using the SSA algorithm; initializing the DDPG algorithm using the optimal hyperparameters, and performing policy learning and optimization through an Actor-Critic network structure; during the flight mission, the DDPG algorithm receives real-time environmental information and dynamically outputs the optimal energy management strategy aimed at minimizing equivalent fuel consumption. Its technical effects include:

[0040] 1) The technical solution of this invention adopts the SSA algorithm to search for the optimal combination of hyperparameters of the DDPG algorithm during the offline training stage, so that the DDPG algorithm can converge to the high-performance strategy faster under complex flight mission profiles and multi-source energy coupling conditions, thereby improving the training efficiency and stability of the overall energy management system.

[0041] 2) The technical solution of this invention adopts the DDPG algorithm as the upper-level control algorithm for energy management in the online optimization module. Through the optimized network structure, it can directly process the continuous action variables of engine power and battery power to obtain smooth and high-precision power allocation results. At the same time, it has end-to-end adaptive learning capability, and can obtain near-optimal strategies under complex flight conditions without relying on an accurate system model, thereby effectively reducing fuel consumption. In addition, the DDPG algorithm can adjust the strategy in real time according to state changes, improving the robustness and overall energy efficiency of the distributed hybrid electric propulsion system.

[0042] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0043] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0046] Figure 1 This is a schematic diagram of an energy management method for a distributed hybrid electric propulsion aircraft provided in an embodiment of the present invention.

[0047] Figure 2 This is a schematic diagram of the solution process of the SSA algorithm for the offline optimization module provided in this embodiment of the invention.

[0048] Figure 3 This is a schematic diagram of the solution process of the DDPG algorithm for the online optimization module provided in this embodiment of the invention.

[0049] Figure 4 This is a schematic diagram of the overall structure of the energy management system for a distributed hybrid electric propulsion aircraft, provided as an embodiment of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0051] In the description of this invention, it should be noted that some processes described in this application specification and drawings include multiple operations that appear in a specific order. However, it should be clearly understood that these operations may be performed in any order or in parallel. Furthermore, various numbers are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0052] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0053] like Figure 1 As shown, this invention proposes an energy management method for distributed hybrid electric propulsion aircraft based on the SSA-DDPG algorithm. This method mainly includes the following steps:

[0054] S1. Establish a distributed hybrid electric propulsion power system module, which includes the dynamics and energy model of the aircraft, batteries, engines, generators, energy management controllers, electric propulsion devices, and an aircraft environmental observation module;

[0055] S2. Establish an offline optimization module. Before the flight mission is executed, input the distributed hybrid electric propulsion system module information (such as flight mission profile information) into the offline optimization module and initialize the SSA algorithm. The SSA algorithm randomly generates a group of sparrows, each of which corresponds to a set of DDPG algorithm hyperparameters. The individual position is updated and iterated according to the discoverer-joiner-watcher mechanism. When the iteration termination condition is met, the globally optimal parameters are output.

[0056] S3. Train the DDPG algorithm using the SSA algorithm, and find and output the optimal hyperparameters of the DDPG algorithm for the current flight mission;

[0057] S4. Establish an online optimization module, which takes the optimal hyperparameters of the DDPG algorithm output by the offline optimization module as input to initialize the DDPG algorithm; the online optimization module is the DDPG algorithm initialized based on the optimal hyperparameters output by the offline optimization, which outputs an energy management strategy through an Actor network and evaluates, optimizes and updates the output energy management strategy through a Critic network. When the iteration termination condition is met, the optimal energy management strategy is output.

[0058] The S5 and DDPG algorithms receive real-time dynamic environmental information during flight mission execution and update and output the optimal energy management strategy with the goal of minimizing equivalent fuel consumption, which is suitable for the current flight state of the aircraft.

[0059] The specific embodiments of the present invention will be described in detail below:

[0060] 1. Establish a distributed hybrid electric propulsion aircraft power system module:

[0061] 1.1 This invention obtains flight phase structure information, flight condition information, environmental and meteorological information, energy system load and constraint information, and mission and time structure information from flight mission profile information and uploads them to the offline optimization module.

[0062] 1.2 Upload the dynamic model of the distributed hybrid electric propulsion aircraft to the offline optimization module. The dynamic model of the distributed hybrid electric propulsion system module is as follows:

[0063]

[0064] in, Power is driven by demand. V is the air density; V is the vacuum velocity; Wing area; e is the zero-lift drag coefficient; e is the induced drag coefficient; W is the aircraft weight; m is the climb rate; m is the mass of the aircraft. It is acceleration.

[0065] 1.3 Upload the energy model of the distributed hybrid electric propulsion aircraft to the offline optimization module. The energy model of the distributed hybrid electric propulsion power system module is as follows:

[0066] The power relationships of the aircraft are as follows:

[0067]

[0068] in, To generate electricity for the engine, Battery charging and discharging power; To improve the efficiency of converting the electrical input power of the motor into output power;

[0069] Engine power generation for:

[0070]

[0071] in, Engine power; For engine efficiency, The overall efficiency of the generator;

[0072] The equivalent fuel consumption relationship for a distributed hybrid electric propulsion aircraft is as follows:

[0073]

[0074] in, For equivalent fuel consumption, This refers to the equivalent fuel consumption on the engine side. Battery-side equivalent fuel consumption; LHV is the lower heating value of fuel.

[0075] 1.4 Establish the objective function for energy management of a distributed hybrid electric propulsion aircraft with the goal of minimizing equivalent fuel consumption. :

[0076]

[0077] Where t represents the t-th iteration, and T represents the total number of iterations.

[0078] 2. Establish an offline optimization module, using the information of the distributed hybrid electric propulsion aircraft power system module as input, to initialize the SSA algorithm.

[0079] 2.1 Establishing the fitness function of the SSA algorithm for:

[0080]

[0081] in, , , These are the weighting coefficients for fuel, battery SOC, and power error, respectively. The battery is in its state of charge. For the desired SOC; For engine power, For actual output propulsion power, Battery charging and discharging power; For engine efficiency; To prevent SSA from generating random trajectories and maintain population diversity, small random perturbation terms are used. ; This indicates the low calorific value of the fuel; k represents the kth segment of the flight mission, and K represents the complete flight mission. The time index representing the flight mission profile. Indicates the total flight mission time. This represents the time interval of the k-th flight mission. This represents the required propulsion power for the k-th segment of the flight mission.

[0082] 2.2 Constraining the state input of the SSA algorithm (Actor network learning rate) Critic network learning rate Explore noise intensity Explore noise attenuation Discount Factor Soft update coefficient The range of the number of samples B drawn for each training iteration includes:

[0083] in, , , , , , , Let A represent the learning rate of the Actor network, the learning rate of the Critic network, the exploration noise intensity, the exploration noise attenuation, the discount factor, the soft update coefficient, and the number of samples drawn in each training iteration for the i-th sparrow individual, respectively. and These are the minimum and maximum learning rates of the Actor network, respectively. and These are the minimum and maximum learning rates of the Critic network, respectively. and These represent the minimum and maximum values ​​of the noise intensity, respectively. and These are the minimum and maximum values ​​for exploring noise attenuation, respectively. and These are the minimum and maximum values ​​of the discount factor, respectively. and These are the minimum and maximum values ​​of the soft update coefficient, respectively. and These represent the minimum and maximum number of samples drawn in each training session, respectively.

[0084] 2.3. The training termination condition Φ for the SSA sparrow search algorithm is:

[0085]

[0086] in, This indicates the termination condition of the SSA algorithm. For the first The global optimal fitness in the next iteration. For the first The global optimal fitness in the next iteration. This represents the rate of change in fitness.

[0087] 3. The offline optimization module trains the DDPG algorithm using the SSA algorithm to find and output the optimal hyperparameters for the DDPG algorithm in the current flight mission:

[0088] 3.1 The SSA algorithm applies each randomly generated set of DDPG algorithm hyperparameters to the DDPG algorithm under simulated flight conditions. Based on fitness evaluation, it determines whether to terminate the loop and output the optimal solution. The solution process is as follows: Figure 2 As shown, the technical steps include the following:

[0089] 3.11. Initialize the sparrow population and the Actor network learning rate. Critic network learning rate Explore noise intensity Explore noise attenuation Discount Factor Soft update coefficient Each training session extracts a sample size B as the state input to form a population of sparrows. =[ , , , , , , ];

[0090] 3.12 Mapping the sparrow population to the DDPG algorithm:

[0091]

[0092] For Actor networks, For Critic network; Indicates the first The i-th sparrow in the next iteration.

[0093] 3.13. Design a fitness function to classify the roles of discoverer, joiner, and vigilant:

[0094] The designed fitness function is:

[0095]

[0096] By calculating the fitness of each initial sparrow individual, the individuals are arranged in ascending order of fitness. The first ρN individuals are selected as discoverers, and the remaining (1-ρN) individuals are selected as joiners, randomly chosen from among them. Each individual is a watcher, N is the initial total number of sparrow individuals, and ρ is the proportion of discoverers among all individuals. The proportion of vigilants among all individuals;

[0097] 3.14. Update the positions and fitness of Discoverers, Joiners, and Watchers:

[0098] Discoverer Update:

[0099] in, For the first The position information of the i-th sparrow in the j-th dimension of the next iteration; Indicates the first The position information of the i-th sparrow in the j-th dimension of the next iteration.

[0100] This represents the maximum number of iterations. ` is a random number that follows a uniform distribution and takes values ​​in the range [0,1]. This is the alarm value; This is a safety threshold; Q is a random step size vector; L is a direction factor.

[0101] Joiner Update:

[0102] in, For the first The latest position of the newest member in the next iteration; For the first The worst position in the next iteration of the population; is a random perturbation matrix, where each element has a random magnitude of 1 or -1.

[0103] Vigilant Update:

[0104] in, For the first The optimal position in the population during the next iteration; This is the step size parameter; Direction factor; For the first The fitness of the i-th sparrow in the next iteration; For the first The global optimal fitness in the next iteration; For the first The worst global fitness in the next iteration; As an intermediate variable; Indicates the first The worst position in the group during the next iteration.

[0105] 3.15. Perform fitness assessment and update the best and worst positions:

[0106] Optimal position update: ; Indicates the first The fitness of the i-th sparrow in the next iteration;

[0107] Worst position update: .

[0108] 3.16. Determine if the termination condition is met. If the termination condition is met, stop the search and output the globally optimal parameters. If not, repeat steps 3.11 to 3.15.

[0109] 4. Establish an online optimization module, using the output of the offline optimization module as input, to initialize the DDPG algorithm:

[0110] 4.1 Establish the reward function of the DDPG algorithm in the online optimization module. for:

[0111]

[0112] in, Let be the reward function, representing the th The reward obtained in the next iteration , , These are the weighting coefficients for fuel, battery SOC, and power error, respectively. Indicates the first The actual output propulsion power of the next iteration Indicates the first The demand for power in the next iteration Indicates the first Battery state of charge in the next iteration Indicates the first The expected SOC of the next iteration Indicates the first Equivalent fuel consumption in the next iteration. The weighting factor for power smoothness. Indicates the first The actual output propulsion power of the next iteration.

[0113] 4.2 The training termination condition Φ` of the DDPG algorithm in the online optimization module is:

[0114]

[0115] Where Φ` represents the termination condition of the DDPG algorithm. For the first The reward obtained in the next iteration This represents the average rate of change in returns.

[0116] 5. After receiving real-time dynamic environmental information, the DDPG algorithm outputs the optimal energy management strategy:

[0117] 5.1 After receiving real-time dynamic environmental information, the DDPG algorithm outputs an energy management strategy aimed at minimizing equivalent fuel consumption. The solution process is as follows: Figure 3 As shown, the technical steps include the following:

[0118] 5.11 Input the globally optimal parameters, i.e., the optimized network hyperparameters, Actor network learning rate. 、 Critic Network Learning Rate Explore noise intensity Explore noise attenuation Discount Factor Soft update coefficient The number of samples B drawn in each training session;

[0119] 5.12. Initialize network parameters and generate the DDPG algorithm;

[0120] 5.13 Input status, i.e., real-time environmental information, and Actor network output actions;

[0121] 5.14. Critic network evaluation action;

[0122] 5.15. Experience pool for storing state and action data;

[0123] 5.16. Sample the data set and iteratively train the DDPG algorithm.

[0124] 5.17, Actor network and Critic network updated;

[0125] 5.18, Target network soft update;

[0126] 5.19. Determine whether the termination condition is met. If the termination condition is met, stop the search and output the energy management strategy (power allocation of propulsion power, engine power, and battery charging and discharging power based on the goal of minimizing equivalent fuel consumption in different flight phases of the entire flight mission profile); if not, repeat steps 5.13 to 5.18.

[0127] In one embodiment, the overall structure of the energy management system established by the present invention is as follows: Figure 4As shown, the system comprises: a distributed hybrid electric propulsion system module, used to input flight mission profile and environmental simulation information to the offline optimization module before flight mission execution, and to input real-time dynamic environmental information to the online optimization module during flight mission execution, and to receive and execute energy management strategy instructions from the online optimization module; an offline optimization module, used to train the DDPG algorithm of the online optimization module using the SSA algorithm as input before flight mission execution, and to find and output the optimal hyperparameter set of the DDPG algorithm; and an online optimization module, used to update and output the optimal energy management strategy with the goal of minimizing equivalent fuel consumption in real time after receiving real-time dynamic environmental information during flight mission execution, through the DDPG algorithm. Utilizing this system can improve the robustness of the distributed hybrid electric propulsion system and reduce its fuel consumption.

[0128] From the description of the above embodiments, those skilled in the art will understand that the present invention provides an energy management method and system for a distributed hybrid electric propulsion aircraft, which has the following advantages:

[0129] 1) The technical solution of this invention adopts the SSA algorithm to search for the optimal combination of hyperparameters of the DDPG algorithm during the offline training stage, so that the DDPG algorithm can converge to the high-performance strategy faster under complex flight mission profiles and multi-source energy coupling conditions, thereby improving the training efficiency and stability of the overall energy management system.

[0130] 2) The technical solution of this invention adopts the DDPG algorithm as the upper-level control algorithm for energy management in the online optimization module. Through the optimized network structure, it can directly process the continuous action variables of engine power and battery power to obtain smooth and high-precision power allocation results. At the same time, it has end-to-end adaptive learning capability, and can obtain near-optimal strategies under complex flight conditions without relying on an accurate system model, thereby effectively reducing fuel consumption. In addition, the DDPG algorithm can adjust the strategy in real time according to state changes, improving the robustness and overall energy efficiency of the distributed hybrid electric propulsion system.

[0131] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0132] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An energy management method for a distributed hybrid electric propulsion aircraft, characterized in that, The method includes the following steps: S1. Obtain flight mission profile information and environmental simulation information through the distributed hybrid electric propulsion system module; S2. Before the flight mission is executed, the flight mission profile information is used as input to initialize the SSA algorithm in the offline optimization module; the fitness function of the initialized SSA algorithm is... for: in, , , These are the weighting coefficients for fuel, battery SOC, and power error, respectively. The battery is in its state of charge. For the desired SOC; For engine power, For actual output propulsion power, Battery charging and discharging power; For engine efficiency; For random disturbance terms; This indicates the low calorific value of the fuel; k represents the kth segment of the flight mission, and K represents the complete flight mission. The time index representing the flight mission profile. Indicates the total flight mission time. This represents the time interval of the k-th flight mission. This represents the required propulsion power for the k-th segment of the flight mission; S3. Train the DDPG algorithm using the SSA algorithm, and find and output the optimal hyperparameters of the DDPG algorithm for the current flight mission; S4. Initialize the DDPG algorithm based on the optimal hyperparameters; S5. During flight mission execution, the online optimization module receives real-time dynamic environmental information. The DDPG algorithm outputs an energy management strategy through the Actor network and evaluates, optimizes, and updates the output energy management strategy through the Critic network. When the iteration termination condition is met, the optimal energy management strategy with the goal of minimizing equivalent fuel consumption is output.

2. The method according to claim 1, characterized in that, In S1, the distributed hybrid electric propulsion power system module includes: a battery, an engine, a generator, an energy management controller, an electric propulsion device, and an aircraft environmental observation module. The acquired flight mission profile information includes: flight phase structure information, flight condition information, environmental and meteorological information, energy system load and constraint information, and mission and time structure information.

3. The method according to claim 1, characterized in that, In S3, the hyperparameters of the DDPG algorithm include the Actor network learning rate. Critic network learning rate Explore noise intensity Explore noise attenuation Discount Factor Soft update coefficient ; The number of samples B is drawn for each training session.

4. The method according to claim 3, characterized in that, The constraints for the number of samples B drawn in each training iteration include: in, , , , , , , Let A represent the learning rate of the Actor network, the learning rate of the Critic network, the exploration noise intensity, the exploration noise attenuation, the discount factor, the soft update coefficient, and the number of samples drawn in each training iteration for the i-th sparrow individual, respectively. and These are the minimum and maximum learning rates of the Actor network, respectively. and These are the minimum and maximum learning rates of the Critic network, respectively. and These represent the minimum and maximum values ​​of the noise intensity, respectively. and These represent the minimum and maximum values ​​of noise attenuation, respectively. and These are the minimum and maximum values ​​of the discount factor, respectively. and These are the minimum and maximum values ​​of the soft update coefficient, respectively. and These represent the minimum and maximum number of samples drawn in each training session, respectively.

5. The method according to claim 4, characterized in that, In S3, the SSA algorithm in the offline optimization module applies each randomly generated set of DDPG algorithm hyperparameters to the DDPG algorithm under simulated flight conditions, and determines whether to terminate the loop and output the optimal solution based on fitness evaluation, including the following steps: (1) Initialize the sparrow population individuals, and the Actor network learning rate. Critic network learning rate Explore noise intensity Explore noise attenuation Discount Factor Soft update coefficient Each training session extracts a sample size B as the state input to form a single sparrow. (2) Map the sparrow population to the DDPG algorithm; (3) Design a fitness function to classify the roles of discoverer, joiner, and vigilant; (4) Update the location and fitness of the discoverer, joiner, and vigilant; (5) Perform fitness assessment and update the best and worst positions; (6) Determine whether the termination condition is met. If the termination condition is met, stop the search and output the result; otherwise, repeat steps (1) to (5).

6. The method according to claim 5, characterized in that, The termination condition for the SSA algorithm is: in, This indicates the termination condition of the SSA algorithm. For the first The global optimal fitness in the next iteration. For the first The global optimal fitness in the next iteration. This represents the rate of change in fitness.

7. The method according to claim 1, characterized in that, In S5, the reward function of the DDPG algorithm is: in, Let be the reward function, representing the th The reward obtained in the next iteration , , These are the weighting coefficients for fuel, battery SOC, and power error, respectively. Indicates the first The actual output propulsion power of the next iteration Indicates the first The demand for power in the next iteration Indicates the first Battery state of charge in the next iteration Indicates the first The expected SOC of the next iteration Indicates the first Equivalent fuel consumption in the next iteration. The weighting factor for power smoothness. Indicates the first The actual output propulsion power of the next iteration.

8. The method according to claim 7, characterized in that, The termination condition Φ` of the DDPG algorithm is: Where Φ` represents the termination condition of the DDPG algorithm. For the first The reward obtained in the next iteration Let T represent the average rate of change of return, and T represent the total number of iterations.

9. An energy management system for a distributed hybrid electric propulsion aircraft, characterized in that, The system includes: a distributed hybrid electric propulsion system module, an offline optimization module, and an online optimization module, and uses the method described in any one of claims 1-8 to obtain the optimal energy management strategy.

Images