A method for improving external heat dissipation of aircraft electric propulsion system based on deep reinforcement learning

Abstract

The present invention discloses a method for improving the external heat dissipation of an aircraft electric propulsion system based on deep reinforcement learning, comprising step 1: establishing a heat dissipation structure library with the position and size of the heat dissipation air inlet of the electric propulsion system on the aircraft fairing as variables according to the aerodynamic shape of the outer fairing of the electric propulsion system of a manned electric aircraft; step 2: setting the maximum allowable operating temperature of the motor and the controller according to the allowable operating temperature range of the electric propulsion system of the manned electric aircraft. The present invention applies artificial intelligence methods to the turbulence calculation process, vividly and intuitively explains the influence of the shear turbulence generated by the rotation of the propeller of the aircraft under different flight profiles on the external flow characteristics of the aircraft, optimizes the numerical simulation model using the turbulence experiment results, and improves the external heat dissipation conditions of the aircraft electric propulsion system.

Description

Technical Field

[0001] The present invention relates to the technical field of optimization of electric propulsion systems for manned electric aircraft, and in particular to a method for improving external heat dissipation of aircraft electric propulsion systems based on deep reinforcement learning. Background Art

[0002] As the electric propulsion system of manned electric aircraft develops towards high power, integration and miniaturization, the motor and controller are the main high-power power electronic equipment in the aircraft electric propulsion system. The high heat flux density of the motor and controller causes the aircraft electric propulsion system to work in a high temperature state. If the heat cannot be dissipated in time and effectively, it is easy to cause the aircraft electric propulsion system to "stop" due to overheating, which directly endangers the operation safety of the aircraft. According to the installation position of the propeller on the manned electric aircraft, it can be divided into two types: pull-in type and push-in type. The propeller of the former is installed in front of the motor, "pulling" the motor forward; the propeller of the latter is installed after the motor, "pushing" the motor forward. Regardless of whether the propeller adopts the pull-in type or the push-in type, in order to reduce the take-off weight of the aircraft and increase the endurance time, the motor direct-drive propeller structure is usually adopted. Therefore, the aircraft electric propulsion system composed of the motor and controller is placed in the engine compartment at the rear end of the propeller. The high heat flux density of the electric propulsion system always endangers the operation safety of the aircraft. In order to improve the heat dissipation conditions of the aircraft electric propulsion system, usually on the aircraft engine compartment fairing, the heat dissipation air intake holes are opened at the corresponding position of the heat dissipation structure of the electric propulsion system, and the air intake holes are opened. The position and size of the hole directly affect the heat dissipation effect of the aircraft electric propulsion system. Due to the rotation of the propeller, the airflow at the heat dissipation air inlet of the aircraft electric propulsion system presents a variable turbulent state. The randomness and complexity of turbulent motion have led to the fact that there is no mature theory that can accurately describe the turbulent state. With the rapid development of computer technology, numerical simulation has gradually become a major means of studying turbulent physical phenomena. How to accurately simulate multi-scale turbulent flow, reasonably determine the boundary conditions of irregular turbulent flow, and accurately count the irregular turbulent flow momentum have become hot topics in turbulent numerical simulation research.

[0003] At present, there are three main methods for numerical simulation of turbulence, namely: direct numerical simulation (DNS), Reynolds average simulation (RANS) and large eddy simulation (LES). DNS is a method for accurate numerical simulation of turbulence. Accurate numerical simulation of turbulence must meet the calculation requirements of large-scale turbulent flow and accurately calculate the movement of small-scale turbulence. As an irregular multi-scale movement, turbulence has a very broad spectrum in both space and time. Therefore, the calculation amount of turbulence calculated by DNS is relatively large. However, the main disadvantages of DNS are: DNS has high requirements on computer performance and takes a long time to calculate. For example, when using DNS to simulate slot flow with a Reynolds number of only about 3,000, the number of grid points is as high as 2×106, and the calculation time is about 250 hours. Limited by the calculation amount of the computer, there is currently no computer required for direct numerical simulation of complex engineering turbulence. Turbulence DNS method can only be used as a research tool for simple turbulence with low Reynolds number.

[0004] The RANS method is a method for numerical simulation of turbulence that uses the theory of turbulence patterns to average the non-steady-state NS equations over time to solve the time-averaged quantities required in engineering. The main disadvantages of the RANS method are: (1) The Reynolds stress model is not universal. Even for solving the same type of problem, when the boundary conditions change, the Reynolds stress model needs to be rebuilt, the calculation process is relatively complicated, and the accuracy of the Reynolds stress model has a great impact on the calculation results; (2) The size and directionality of the vortex are ignored, and the movement and dynamics of the vortex are not considered enough, which cannot accurately describe the characteristics of the fluid flow; (3) Due to the limitations of the turbulence model conditions, the accuracy of the calculation results obtained for solving the numerical simulation problems of complex flow turbulence such as unsteady and large separation is low; (4) The calculation relies heavily on the experience of the given flow field shape and boundary conditions, and has poor general adaptability.

[0005] The LES method is a turbulence numerical simulation method different from the DNS method and the RANS method. Turbulence can be regarded as composed of vortices of different scales, but it is unrealistic to use the DNS method to fully simulate vortices of various scales. Since macroscopic parameters such as turbulent diffusion and momentum transport are reflected through large-scale vortex motion, it is of great significance to numerically simulate large-scale vortices. People design a mathematical method to "filter out" small-scale vortices smaller than the grid and obtain the local average value of the grid points. The LES method uses a sub-grid scale model to simulate the influence of small-scale turbulent motion on large-scale turbulent motion, that is, directly numerically simulate large-scale turbulent motion, average the NS equations in a small spatial domain (or filter), so as to remove small-scale vortices from the turbulent field and derive the equations satisfied by large vortices. The main disadvantages of the LES method are: (1) In order to ensure the calculation accuracy, the grid division of the small vortex model is required to be denser, which requires a high storage capacity of the computer; (2) In the LES method simulation calculation process, a large number of nonlinear partial differential equations need to be solved, which requires a large amount of calculation and a high processing power of the computer; (3) Due to the complexity of actual turbulence, the LES method still requires a lot of time and experimental costs.

[0006] The core problem of LES simulation calculation is how to construct a sub-grid model through large-scale flow field information. Traditional sub-grid models include Smagorinsky model, similarity model, gradient model, optimization model, one-equation model and second-order moment model. Traditional sub-grid models only use single-point flow field information and simple functional relationships, so they have large relative errors in prior verification and too strong dissipation in a posteriori verification. Deep reinforcement learning (DRL) method is a method that combines the perception ability of deep learning and the decision-making ability of reinforcement learning. It can be directly controlled according to the input data information, and can provide clear solutions to the perception and decision-making problems of complex systems. It is an artificial intelligence method that is closer to human thinking.

[0007] The policy gradient method has become one of the most competitive among all kinds of DRL methods due to its huge parallelization potential and ability to solve continuous problems. The policy gradient method can directly optimize the objective function by using optimization algorithms such as neural networks, and explicitly express it when the problem is established, ensuring the real-time and visibility of problem processing. Therefore, the policy gradient method can be widely used in solving turbulence modeling problems. At present, the application of DRL method in turbulence numerical simulation is still relatively small, but DRL method has been widely used in many fields such as robots, microgrids and intelligent transportation.

[0008] Therefore, it is very meaningful to provide a method for calculating the heat dissipation turbulence field of a manned electric aircraft based on deep reinforcement learning to improve the external heat dissipation conditions of the aircraft electric propulsion system. Summary of the invention

[0009] In view of this, the present invention discloses a method for improving the external heat dissipation of an aircraft electric propulsion system based on deep reinforcement learning; by optimizing the position and size of the air inlet holes of the heat dissipation structure of the aircraft electric propulsion system, the air flow rate of the air inlet holes is increased, and the external heat dissipation conditions of the aircraft electric propulsion system are improved, so as to effectively solve the heat dissipation problem of the aircraft electric propulsion system.

[0010] The technical solution provided by the present invention is specifically a method for improving external heat dissipation of an aircraft electric propulsion system based on deep reinforcement learning, which is characterized by comprising:

[0011] Step 1: According to the aerodynamic shape of the outer fairing of the electric propulsion system of the manned electric aircraft, a heat dissipation structure library is established with the position and size of the heat dissipation air inlet of the electric propulsion system on the aircraft fairing as variables;

[0012] Step 2: According to the allowable operating temperature range of the electric propulsion system of the manned electric aircraft, set the maximum allowable operating temperature of the motor and controller;

[0013] Step 3: Determine the reward value of deep reinforcement learning based on the difference between the operating temperature of the motor and controller in the electric propulsion system and the allowable operating temperature. If the operating temperature is equal to or lower than the allowable operating temperature value, the reward value is 1; otherwise, it is -1;

[0014] Step 4: Propose the position and size of the heat dissipation air inlet of the electric propulsion system on the aircraft fairing, and give the initial value of the outer dimensions of the heat dissipation air inlet of the electric propulsion system;

[0015] Step 5: Perform deep reinforcement learning to obtain the optimized position and size of the cooling air inlet of the electric propulsion system of the manned electric aircraft;

[0016] Step 6: Based on the position, size and sub-grid stress parameters of the air inlet of the aircraft electric propulsion system heat dissipation structure obtained by deep reinforcement learning, combined with the actual operating conditions of the aircraft, the three-dimensional fluid field finite element calculation boundary conditions are given;

[0017] Step 7: Build a three-dimensional fluid field finite element simulation model to perform three-dimensional simulation calculations on the heat dissipation structure of the aircraft electric propulsion system;

[0018] Step 8: Obtain the maximum operating temperature of the motor and controller through three-dimensional fluid field finite element simulation calculation; determine whether the calculated value exceeds the maximum allowable set temperature value. If it exceeds the maximum allowable operating temperature setting value, recalculation is required; if it does not exceed, test samples are trial-produced according to the position and size of the heat dissipation air inlet of the aircraft electric propulsion system, and the test environment is built to conduct prototype tests;

[0019] Step 9: Obtain the test value of the maximum operating temperature of the motor and the controller according to the test results of the prototype, and determine whether the test value of the maximum operating temperature of the motor and the controller exceeds the maximum allowable setting value. If it exceeds, re-set the boundary conditions of the three-dimensional fluid field finite element calculation, and perform the three-dimensional simulation calculation of the heat dissipation structure of the aircraft electric propulsion system;

[0020] Step 10: If it does not exceed, the design optimization of the position and size of the heat dissipation air inlet holes of the electric propulsion system on the aircraft fairing is completed.

[0021] Preferably, the deep reinforcement learning in step 5 comprises the following steps:

[0022] Step 51: taking the actual operating conditions of the aircraft propeller and the compressible turbulent large eddy simulation control equation as environmental variables;

[0023] Step 52: according to the given strategy of the position and size of the heat dissipation air inlet of the aircraft electric propulsion system, a certain position and size are determined, and the Agent, as an algorithm model responsible for deciding the position and size of the heat dissipation air inlet of the aircraft electric propulsion system, interacts with the environment to achieve high-dimensional observation, and uses the deep reinforcement learning method of the recurrent neural network to perceive the maximum allowable operating temperature of the motor and controller with high heat flux density in the aircraft electric propulsion system;

[0024] Step 53: Determine a reward strategy based on the maximum allowable operating temperature of the motor and the controller, and give a reward based on the strategy. If the operating temperature is equal to or lower than the allowable operating temperature value, the reward value is 1; otherwise, it is -1.

[0025] Step 54: Evaluate the value function of each action based on the expected return, map the current state to the corresponding action; the environment responds to this action and obtains the next observation value and reward value;

[0026] Step 55: Repeat the loop to obtain the optimized position and size of the air inlet of the heat dissipation structure of the aircraft electric propulsion system, and complete deep reinforcement learning.

[0027] Preferably, in combination with the actual operating conditions of the manned electric aircraft, the shear turbulence motion law generated by the propeller of the aircraft during flight is analyzed, and the compressible turbulent large eddy simulation control equation is established, as shown in equations (1) to (2), and the sub-grid stress is obtained as shown in equation (3);

[0028]

[0029] The subgrid stress is:

[0030]

[0031] The present invention provides a method for improving the external heat dissipation of an aircraft electric propulsion system based on deep reinforcement learning. The powerful data interpolation capability of the DRL method is applied to the large eddy numerical simulation of turbulence, and the construction of a numerical simulation model for the control of large-scale vortex motion of shear turbulence generated by propeller rotation is completed. The outflow characteristics of the radiator of the aircraft electric propulsion system are obtained. On this basis, the position and size of the heat dissipation air inlet holes of the aircraft electric propulsion system on the aircraft engine compartment fairing are optimized, the external heat dissipation conditions of the aircraft electric propulsion system are improved, and it is ensured that the aircraft electric propulsion system will not cause the aircraft to "stop" due to overheating, thereby ensuring the flight safety of the aircraft.

[0032] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0034] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0035] Figure 1 is a flow chart of a method for improving external heat dissipation of an aircraft electric propulsion system based on deep reinforcement learning of the present invention;

[0036] Figure 2 is the deep reinforcement learning flow chart described in the present invention;

[0037] Figure 3 is the spatial streamline diagram described in the specific implementation mode of the present invention;

[0038] Figure 4 It is the airflow characteristic curve of the air inlet hole of the motor controller radiator described in the specific implementation mode of the present invention. DETAILED DESCRIPTION

[0039] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present invention. Instead, they are merely examples of systems consistent with some aspects of the present invention as detailed in the appended claims.

[0040] The motion of shear turbulence generated by the rotation of the propeller of a manned electric aircraft has randomness, uncertainty and complexity. Its calculation process requires large-scale iterative solution of multiple constraints, multiple objectives and multiple variables coupling, and there are problems such as long solution time and poor calculation accuracy. For this reason, this implementation scheme provides a method for calculating the heat dissipation turbulence field of a manned electric aircraft based on deep reinforcement learning, and further provides a method for improving the external heat dissipation of an aircraft electric propulsion system based on deep reinforcement learning. The DRL method, as an artificial intelligence method that is closer to human thinking, has opened up a new idea for the calculation of shear turbulence generated by propeller rotation. With the help of the powerful data interpolation ability of the DRL method, the DRL method is applied to the large eddy numerical simulation of turbulence, and the construction of a numerical simulation model for the control of large-scale vortex motion of shear turbulence generated by propeller rotation is completed, and the outflow characteristics of the radiator of the aircraft electric propulsion system are obtained. On this basis, the position and size of the heat dissipation air inlet of the aircraft electric propulsion system on the aircraft engine compartment fairing are optimized, and the external heat dissipation conditions of the aircraft electric propulsion system are improved to ensure that the aircraft electric propulsion system will not cause the aircraft to "park" due to overheating, thereby ensuring the flight safety of the aircraft.

[0041] Methods based on deep reinforcement learning to improve external heat dissipation of aircraft electric propulsion systems, such as Figure 1 As shown, including:

[0042] Step 1: According to the aerodynamic shape of the outer fairing of the electric propulsion system of the manned electric aircraft, a heat dissipation structure library is established with the position and size of the heat dissipation air inlet of the electric propulsion system on the aircraft fairing as variables;

[0043] Step 2: According to the allowable operating temperature range of the electric propulsion system of the manned electric aircraft, set the maximum allowable operating temperature of the motor and controller;

[0044] Step 3: Based on the heat dissipation structure library, a determination strategy is formulated to give the location and size of the heat dissipation air inlet of the electric propulsion system on the aircraft fairing;

[0045] Step 4: Perform deep reinforcement learning to obtain the optimized position and size of the cooling air inlet of the electric propulsion system of the manned electric aircraft;

[0046] Step 6: Based on the position, size and sub-grid stress parameters of the air inlet of the aircraft electric propulsion system heat dissipation structure obtained by deep reinforcement learning, combined with the actual operating conditions of the aircraft, the three-dimensional fluid field finite element calculation boundary conditions are given;

[0047] Step 7: Based on the actual operating conditions of the aircraft, a three-dimensional fluid field finite element simulation model is built to perform three-dimensional simulation calculations on the heat dissipation structure of the aircraft electric propulsion system;

[0048] Step 8: Obtain the maximum operating temperature of the motor and controller through three-dimensional fluid field finite element simulation calculation; determine whether the calculated value exceeds the maximum allowable set temperature value obtained in step 2. If it exceeds the maximum allowable operating temperature setting value, recalculation is required; if it does not exceed, test samples are trial-produced according to the position and size of the heat dissipation air inlet of the aircraft electric propulsion system, and a test environment is built to conduct prototype tests;

[0049] Step 9: Obtain the test value of the maximum operating temperature of the motor and the controller according to the test results of the prototype, and determine whether the test value of the maximum operating temperature of the motor and the controller exceeds the maximum allowable setting value. If it exceeds, re-set the boundary conditions of the three-dimensional fluid field finite element calculation, and perform the three-dimensional simulation calculation of the heat dissipation structure of the aircraft electric propulsion system;

[0050] Step 10: If it does not exceed, the design optimization of the position and size of the heat dissipation air inlet holes of the electric propulsion system on the aircraft fairing is completed.

[0051] In step 5, deep reinforcement learning includes the following steps:

[0052] Step 51: taking the actual operating conditions of the aircraft propeller and the compressible turbulent large eddy simulation control equation as environmental variables;

[0053] Step 52: According to the given strategy of the position and size of the heat dissipation air inlet of the aircraft electric propulsion system, a certain position and size are determined, and the intelligent agent interacts with the environment to achieve high-dimensional observation, and uses the deep reinforcement learning method of the recurrent neural network to perceive the maximum allowable operating temperature of the motor and controller with high heat flux density in the aircraft electric propulsion system;

[0054] Step 53: Determine a reward strategy based on the maximum allowable operating temperature of the motor and the controller, and give a reward based on the strategy. If the operating temperature is equal to or lower than the allowable operating temperature value, the reward value is 1; otherwise, it is -1;

[0055] Step 54: Evaluate the value function of each action based on the expected return, and map the current state to the corresponding action by formulating a certain strategy; the environment responds to this action and obtains the next observation value and reward value;

[0056] Step 55: Repeat the loop to obtain the optimized position and size of the air inlet of the heat dissipation structure of the aircraft electric propulsion system, and complete deep reinforcement learning.

[0057] The above compressible turbulent large eddy simulation control equation and sub-grid stress parameters are established by combining the actual operating conditions of manned electric aircraft and analyzing the shear turbulence motion law generated by the propeller of the aircraft during flight. The compressible turbulent large eddy simulation control equation is shown in equations (1) to (2), and the sub-grid stress is obtained as shown in equation (3);

[0058]

[0059] The subgrid stress is:

[0060]

[0061] This implementation scheme uses the powerful data interpolation capability of the DRL method to apply the DRL method to the large eddy numerical simulation of turbulence, improves the calculation accuracy of the sub-grid model, completes the construction of the mathematical model of shear turbulence under propeller rotation, and obtains the control equation for compressible turbulent large eddy simulation;

[0062] Applying artificial intelligence methods to the turbulence calculation process can vividly and intuitively explain the impact of shear turbulence generated by propeller rotation under different flight profiles on the aircraft's external flow characteristics. The turbulence experimental results are used to optimize the numerical simulation model and improve the external heat dissipation conditions of the aircraft's electric propulsion system.

[0063] The LES method is used to study the shear turbulence phenomenon generated by propeller rotation, and the DRL method is applied to the calculation of turbulence to reveal the characteristics of shear turbulence under specific boundary conditions, which will enrich the relevant system scope of turbulence theory.

[0064] Aiming at the turbulence under the propeller rotation of a certain type of two-seat manned electric aircraft, a method for calculating the heat dissipation turbulence field of a manned electric aircraft based on deep reinforcement learning proposed in this invention is used. The air intake angle and air intake hole area of ​​the aircraft motor controller radiator are taken as optimization variables to optimize the external heat dissipation environment of the aircraft motor controller. The spatial streamline diagram is shown in the figure. Figure 3 As shown in the figure, the airflow characteristic curve of the motor controller radiator air inlet is obtained as Figure 4 As shown, the correctness of the optimization method was finally proved through prototype tests. By applying the calculation method proposed in the present invention, the temperature of the motors and controllers of the main heat-generating components in the electric propulsion system of the aircraft decreased by 4 to 5°C under the same flight conditions and the same external heat dissipation conditions compared with before optimization.

[0065] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These changes and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for improving external heat dissipation of an aircraft electric propulsion system based on deep reinforcement learning, characterized in that: include Step 1: According to the aerodynamic shape of the outer fairing of the electric propulsion system of the manned electric aircraft, a heat dissipation structure library is established with the position and size of the heat dissipation air inlet of the electric propulsion system on the aircraft fairing as variables; Step 2: According to the allowable operating temperature range of the electric propulsion system of the manned electric aircraft, set the maximum allowable operating temperature of the motor and controller; Step 3: Determine the reward value of deep reinforcement learning based on the difference between the operating temperature of the motor and controller in the electric propulsion system and the allowable operating temperature; if the operating temperature is equal to or lower than the allowable operating temperature value, the reward value is 1; if the higher allowable operating temperature value is -1; Step 4: Propose the position and size of the heat dissipation air inlet of the electric propulsion system on the aircraft fairing, and give the initial value of the outer dimensions of the heat dissipation air inlet of the electric propulsion system; Step 5: Perform deep reinforcement learning to obtain the optimized position and size of the cooling air inlet of the electric propulsion system of the manned electric aircraft; Step 6: Based on the position, size and sub-grid stress parameters of the air inlet of the aircraft electric propulsion system heat dissipation structure obtained by deep reinforcement learning, combined with the actual operating conditions of the aircraft, the three-dimensional fluid field finite element calculation boundary conditions are given; Step 7: Build a three-dimensional fluid field finite element simulation model to perform three-dimensional simulation calculations on the heat dissipation structure of the aircraft electric propulsion system; Step 8: Obtain the maximum operating temperature of the motor and controller through three-dimensional fluid field finite element simulation calculation; determine whether the calculated value exceeds the maximum allowable set temperature value. If it exceeds the maximum allowable operating temperature setting value, recalculation is required; If it does not exceed the limit, test samples shall be manufactured according to the position and size of the heat dissipation air intake of the aircraft electric propulsion system, and the test environment shall be built to conduct prototype tests; Step 9: Obtain the test value of the maximum operating temperature of the motor and the controller according to the test results of the prototype, and determine whether the test value of the maximum operating temperature of the motor and the controller exceeds the maximum allowable setting value. If it exceeds, re-set the boundary conditions of the three-dimensional fluid field finite element calculation, and perform the three-dimensional simulation calculation of the heat dissipation structure of the aircraft electric propulsion system; Step 10: If it does not exceed, the design optimization of the position and size of the heat dissipation air inlet of the electric propulsion system on the aircraft fairing is completed; The deep reinforcement learning described in step 5 includes the following steps: Step 51: taking the actual operating conditions of the aircraft propeller and the compressible turbulent large eddy simulation control equation as environmental variables; Step 52: according to the given strategy of the position and size of the heat dissipation air inlet of the aircraft electric propulsion system, a certain position and size are determined, and the Agent, as an algorithm model responsible for deciding the position and size of the heat dissipation air inlet of the aircraft electric propulsion system, interacts with the environment to achieve high-dimensional observation, and uses the deep reinforcement learning method of the recurrent neural network to perceive the maximum allowable operating temperature of the motor and controller with high heat flux density in the aircraft electric propulsion system; Step 53: Determine a reward strategy based on the maximum allowable operating temperature of the motor and the controller, and give a reward based on the strategy. If the operating temperature is equal to or lower than the allowable operating temperature value, the reward value is 1; otherwise, it is -1; Step 54: Evaluate the value function of each action based on the expected return and map the current state to the corresponding action; the environment responds to this action and obtains the next observation value and reward value; Step 55: Continuously loop to obtain the optimized position and size of the air inlet of the heat dissipation structure of the aircraft electric propulsion system, and complete deep reinforcement learning; Combined with the actual operating conditions of manned electric aircraft, the shear turbulence motion law generated by the propeller of the aircraft during flight is analyzed, and the compressible turbulent large eddy simulation control equation is established, as shown in equations (1) to (2), and the sub-grid stress is obtained as shown in equation (3); The subgrid stress is:

Images