Method for multi-source energy optimization of hypersonic vehicles based on proxy model

Abstract

The application discloses a hyperplane vehicle multi-source energy optimization method based on an agent model, and comprises the following steps: S1, a parallel TBCC engine and a multi-source energy extraction system mechanism model is established; S2, the mechanism model is simulated to generate sample library data; S3, the operation rules of the mechanism model are generalized by using a GA-BP neural network, and a real-time simulation agent model is established to solve the problem of insufficient real-time performance of the mechanism model; S4, a hyperplane vehicle multi-source energy optimization design problem based on the neural network agent model is constructed; S5, an optimization algorithm NSGA-II is used to solve the optimal multi-source energy configuration mode; and S6, whether the current engine parameters are within the safety constraints is judged, and if not, the step S5 is returned. The application can reduce the calculation cost of the hyperplane vehicle multi-source energy optimization and improve the optimization efficiency.

Description

Technical Field

[0001] This invention discloses a multi-source energy optimization method for hypersonic aircraft based on a proxy model, belonging to the field of integrated energy optimization for aircraft. Background Technology

[0002] Hypersonic vehicles are ideal platforms for future fighter jets, placing higher demands on maneuverability, flight time, and range. Furthermore, the trend of electromechanical equipment replacing traditional pneumatic and hydraulic systems will increase the power consumption of these vehicles by several orders of magnitude. Existing airborne power generation systems primarily extract energy from fixed parts of the engine, through bleed air, combustion, or shaft power extraction. This single energy extraction method cannot meet the power demands of future hypersonic vehicles, nor can it adaptively optimize energy efficiency under high maneuverability. This single method of energy extraction from fixed parts significantly impacts engine performance, severely limiting the aircraft's cruise time and flight radius. For hypersonic vehicles with high power, long range, and high thrust-to-weight ratio requirements, a multi-source, intelligently managed energy extraction method needs to be designed to meet the power needs of hypersonic vehicles while minimizing the impact on engine performance.

[0003] Current domestic and international research largely focuses on key technologies for the main propulsion of hypersonic vehicles. Novel combined cycle propulsion technology has seen significant advancements. Turbine-based combined cycle (TBCC) engines, composed of turbine and ramjet engines connected in parallel or series, possess high specific impulse and a wide flight envelope, along with advantages such as good low-speed fuel economy, conventional takeoff and landing capabilities, and reusability. Hypersonic vehicles powered by TBCC engines can achieve full-envelope operation from ground takeoff to high Mach number cruise. However, research on effective power generation methods and high-efficiency energy optimization for hypersonic vehicles remains relatively weak. Hypersonic vehicles utilize methods such as main propulsion intake / fuel-gas total enthalpy extraction, shaft power extraction, thermal energy extraction, and fuel cooling... The generation of multi-source electrical energy using various methods all affect engine performance indicators such as fuel consumption, thrust, and stability margin. Furthermore, different energy extraction methods and parameter configurations of multi-source energy have different mechanisms of influence on engine performance. Therefore, strengthening research on multi-source energy optimization methods is beneficial for promoting the balanced development of hypersonic vehicle system technology and meeting the comprehensive optimization design requirements for higher thrust-to-weight ratios and lower fuel consumption rates. Simultaneously, reducing operating costs and improving combat performance are the key contradictions in the future development of hypersonic fighter jets. Novel multi-source electrical energy generation systems and efficient multi-source energy optimization methods are among the technical fields for resolving this contradiction. Therefore, the research on the wide-area multi-source energy optimization method for hypersonic vehicles proposed in this invention is of great significance for the design of future military aircraft. Summary of the Invention

[0004] The technical problem to be solved by this invention is to address the shortcomings of the aforementioned background technology by providing a multi-source energy optimization method for hypersonic aircraft based on a surrogate model. This method employs a real-time simulateable surrogate model to solve the problem of insufficient real-time performance of the mechanistic model under complex and variable flight conditions, and uses a multi-objective optimization algorithm to achieve real-time multi-source energy optimization.

[0005] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0006] S1. Establish a mechanism model for a parallel TBCC engine and its multi-source energy extraction system. The parallel TBCC engine is composed of a turbofan engine and a ramjet engine connected in parallel, with their parallel dual-channel configuration being relatively independent. The main components of the turbofan engine channel are: low-speed intake, low-pressure compressor, high-pressure compressor, main combustion chamber, high-pressure turbine, low-pressure turbine, bypass duct, mixing chamber, afterburner, and low-speed nozzle. The main components of the ramjet engine channel are: high-speed intake, isolator, combustion chamber, and high-speed nozzle. The multi-source energy extraction system model includes shaft power generation, bleed air power generation, gas turbine power generation, oil-gas turbine power generation, and a battery. When the engine operates in turbofan engine mode, the following energy extraction methods are used:

[0007] 1) Shaft power extraction for power generation

[0008] 2) Exhaust gas power generation

[0009] 3) Gas turbine power generation, extracting energy from the high-pressure stage / intermediate stage / bypass duct.

[0010] When the engine is operating in ramjet engine mode, the following energy extraction methods are used:

[0011] 1) Exhaust gas power generation

[0012] 2) Gas turbine power generation

[0013] 3) Oil and gas turbine power generation

[0014] 4) Storage battery;

[0015] S2, in both turbofan and ramjet engine modes, selected multiple typical flight conditions according to gradients, and simulated the mechanistic model with different power generation configuration parameters for the multi-source energy extraction system. The engine fuel consumption rate, thrust, speed, and turbine inlet temperature under steady-state conditions of the mechanistic model were recorded as sample data. A normalization method was used to process the simulation sample data and construct a sample library for the surrogate model. The normalized data a i It can be represented as

[0016]

[0017] In the formula, a represents the original sample data, a max a is the maximum value of the corresponding variable in the sample data. min This represents the minimum value of the variable corresponding to the sample data.

[0018] S3. Utilizing the operating rules of the GA-BP neural network generalization mechanism model, a real-time simulation surrogate model is established. Flight altitude, flight Mach number, and power generation from the multi-source energy extraction system are used as inputs to the surrogate model, while engine fuel consumption rate, thrust, speed, turbine inlet temperature, and stability margin are used as outputs. The GA-BP neural network leverages the global optimization capability of the genetic algorithm, inputting optimized weights and thresholds into the BP neural network training dataset. In this invention, the output variables of the real-time simulation surrogate model are optimized using the GA algorithm to obtain optimal or near-optimal weights and thresholds. These optimized weights and thresholds are then substituted into the BP neural network to train the sample data, thereby generating the real-time simulation surrogate model. The neural network settings are as follows:

[0019] 1) Input layer parameters

[0020] The input parameters are flight conditions and the power generation of the multi-source energy extraction system. Therefore, five input neurons are set in the turbofan engine mode, representing flight altitude, flight Mach number, power generation from shaft power extraction, power generation from bleed air, and power generation from gas turbine. In the ramjet engine mode, six input neurons are set, representing flight altitude, flight Mach number, power generation from bleed air, power generation from gas turbine, power generation from oil and gas turbine, and energy storage of the battery.

[0021] 2) Hidden layer parameters

[0022] The number of neurons in the hidden layer has a significant impact on the convergence speed and generalization ability of a neural network. The number of neurons in the hidden layer can be calculated using empirical formulas:

[0023]

[0024] In the formula, N hid denoted as the number of hidden layer neurons, nn as the number of input layer neurons, mm as the number of output layer neurons, and a as an adjustable constant in the range of 1 to 10. The final number of hidden layers is obtained by repeatedly trying different values ​​of a.

[0025] 3) Output layer parameters

[0026] The number of neurons in the output layer of the neural network is determined by the target requirements. In this invention, the aim is to predict the engine's performance parameters under the current flight conditions and multi-source energy parameter configuration. Therefore, six output neurons are set in turbofan engine mode.

[0027] These correspond to fuel consumption rate, thrust, turbine inlet temperature, high-pressure shaft speed, low-pressure shaft speed, and minimum stability margin, respectively; in turbofan engine mode, two parameters are set, corresponding to fuel consumption rate and thrust, respectively.

[0028] S4. Construct a multi-source energy optimization design problem for hypersonic aircraft based on a neural network surrogate model. Under the condition of meeting the electric energy requirements of the aircraft, the fuel consumption rate Wf and thrust FN of the engine are used as optimization objectives, and the engine speed, turbine inlet temperature and stability margin are used as engine safety constraints to balance the fuel economy and flight quality of the aircraft.

[0029] The mathematical description of the specific optimization objective J in the turbofan engine mode is as follows:

[0030]

[0031] In the formula, P y,i P represents the power generation from the high-voltage stage, intermediate stage, and outer bypass duct. r,j P represents the power generation from the high-voltage stage, intermediate stage, and bypass gas turbine. z,k This represents the power generation extracted from the high-pressure and low-pressure shafts. FN0 and Wf0 are the engine's thrust and fuel consumption rate before multi-source energy extraction, respectively.

[0032] The mathematical description of the specific optimization objective in the ramjet engine mode is as follows:

[0033]

[0034] In the formula, P y P is the amount of electricity generated by induced draft gas power generation. r P is the amount of electricity generated by a gas turbine generator. o The power generation of oil and gas turbine generators, P x It stores energy for the battery.

[0035] Considering engine operational safety, it is necessary to ensure that the engine does not overheat, over-rev, or surge while extracting energy from multiple sources. Therefore, safety constraints are established for the turbine inlet temperature, mechanical shaft speed, and compressor stability margin during the multi-source energy optimization process. The mathematical description of the engine's safety constraint C is as follows:

[0036]

[0037] In the formula, T is the turbine inlet temperature, and N is the inlet temperature. HPS The rotational speed of the high-pressure shaft is N. LPS To reduce the speed of the low-pressure shaft, SM HPC For the stability margin of the high-pressure compressor, SM LPC This refers to the stability margin of the low-pressure compressor.

[0038] S5 uses the multi-objective optimization algorithm NSGA-II to solve for the current optimal multi-source energy configuration. Each individual in the algorithm represents a set of power generation configuration parameters of a multi-source energy extraction system. The optimization process can be simply described as finding a set of power generation configuration parameters to optimize the engine's fuel consumption rate and thrust under the conditions of meeting the power demand and engine safety constraints.

[0039] The specific calculation process for NSGA-II to solve for the optimal fuel consumption rate and thrust includes the following steps:

[0040] S5-1, Set initialization parameters, set population size m = 100; set maximum number of iterations T under the same environment. max =100; Set the objective function dimension d1=2; In turbofan engine mode, set the population dimension d2=3; In ramjet engine mode, set the population dimension d2=4; Set the upper bound of the variable X. max =500; Sets the lower bound of the variable X min =0; Set the mutation probability a1 = 0.5, and the mutation probability a2 = 0.8;

[0041] S5-2, the first generation population is sorted non-dominated, and after selection, crossover and mutation operations, the first generation subpopulation is generated;

[0042] S5-3, merges the parent population and the offspring population into a new population;

[0043] S5-4: Perform fast non-dominated sorting on the merged population, calculate the crowding degree and record the frontier solution;

[0044] S5-5: Select elite individuals as the new parent population, and use selection, crossover, and mutation.

[0045] S5-6, Population generation T = T + 1, determine if T has reached the set maximum population generation T. max If T < T max Then return to step S5-3, if T≥T max Then output the leading edge solution;

[0046] S5-7: Record the final Pareto optimal solution and optimal variables. The optimal solution is the one with the minimum fuel consumption rate and the maximum thrust, and the optimal variables are the power generation configuration of the multi-source energy extraction system corresponding to the optimal solution.

[0047] S6 determines whether the current engine speed, turbine inlet temperature, and stability margin are within safety constraints; otherwise, it returns to S5. Attached Figure Description

[0048] Figure 1This is a schematic diagram of a multi-source energy optimization method for hypersonic vehicles based on a proxy model.

[0049] Figure 2 This is a schematic diagram of the TBCC engine model.

[0050] Figure 3 This is a schematic diagram of a multi-source energy extraction system model. Detailed Implementation

[0051] The technical solution of the invention will be described in detail below with reference to the accompanying drawings:

[0052] like Figure 1 As shown, this invention proposes a multi-source energy optimization method for hypersonic vehicles based on a surrogate model. This method employs a real-time simulateable surrogate model to address the problem of insufficient real-time performance of the mechanistic model under complex and variable flight conditions, and uses a multi-objective optimization algorithm to achieve real-time multi-source energy optimization.

[0053] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0054] S1, Establish a mechanism model for a parallel TBCC engine and a multi-source energy extraction system. For example... Figure 2 As shown, the parallel TBCC engine is composed of a turbofan engine and a ramjet engine connected in parallel, with its parallel dual-channel configuration being relatively independent. The main components of the turbofan engine channel are: low-speed intake, low-pressure compressor, high-pressure compressor, main combustion chamber, high-pressure turbine, low-pressure turbine, bypass duct, mixing chamber, afterburner, and low-speed nozzle. The main components of the ramjet engine channel are: high-speed intake, isolator, combustion chamber, and high-speed nozzle. Figure 3 As shown, the multi-source energy extraction system model includes shaft power extraction for power generation, bleed air power generation, gas turbine power generation, oil-gas turbine power generation, and a battery. When the engine operates in turbofan engine mode, the following energy extraction methods are used: 1) shaft power extraction for power generation; 2) bleed air power generation; 3) gas turbine power generation, extracting energy from the high-pressure stage / intermediate stage / bypass duct. When the engine operates in ramjet engine mode, the following energy extraction methods are used: 1) bleed air power generation; 2) gas turbine power generation; 3) oil-gas turbine power generation; 4) a battery.

[0055] S2, in both turbofan and ramjet engine modes, selected multiple typical flight conditions according to gradients, and simulated the mechanistic model with different power generation configuration parameters for the multi-source energy extraction system. The engine fuel consumption rate, thrust, speed, and turbine inlet temperature under steady-state conditions of the mechanistic model were recorded as sample data. A normalization method was used to process the simulation sample data and construct a sample library for the surrogate model. The normalized data a i It can be represented as

[0056]

[0057] In the formula, a represents the original sample data, a max a is the maximum value of the corresponding variable in the sample data. min This represents the minimum value of the variable corresponding to the sample data.

[0058] S3. Utilizing the operating rules of the GA-BP neural network generalization mechanism model, a real-time simulation surrogate model is established. Flight altitude, flight Mach number, and power generation from the multi-source energy extraction system are used as inputs to the surrogate model, while engine fuel consumption rate, thrust, speed, turbine inlet temperature, and stability margin are used as outputs. The GA-BP neural network leverages the global optimization capability of the genetic algorithm, inputting optimized weights and thresholds into the BP neural network training dataset. In this invention, the output variables of the real-time simulation surrogate model are optimized using the GA algorithm to obtain optimal or near-optimal weights and thresholds. These optimized weights and thresholds are then substituted into the BP neural network to train the sample data, thereby generating the real-time simulation surrogate model. The neural network settings are as follows:

[0059] 1) Input layer parameters

[0060] The input parameters are flight conditions and the power generation of the multi-source energy extraction system. Therefore, five input neurons are set in the turbofan engine mode, representing flight altitude, flight Mach number, power generation from shaft power extraction, power generation from bleed air, and power generation from gas turbine. In the ramjet engine mode, six input neurons are set, representing flight altitude, flight Mach number, power generation from bleed air, power generation from gas turbine, power generation from oil and gas turbine, and energy storage of the battery.

[0061] 2) Hidden layer parameters

[0062] The number of neurons in the hidden layer has a significant impact on the convergence speed and generalization ability of a neural network. The number of neurons in the hidden layer can be calculated using empirical formulas:

[0063]

[0064] In the formula, N hid denoted as the number of hidden layer neurons, nn as the number of input layer neurons, mm as the number of output layer neurons, and a as an adjustable constant in the range of 1 to 10. The final number of hidden layers is obtained by repeatedly trying different values ​​of a.

[0065] 3) Output layer parameters

[0066] The number of neurons in the output layer of the neural network is determined by the target requirements. In this invention, the aim is to predict the engine's performance parameters under the current flight conditions and multi-source energy parameter configuration. Therefore, six output neurons are set in turbofan engine mode.

[0067] These correspond to fuel consumption rate, thrust, turbine inlet temperature, high-pressure shaft speed, low-pressure shaft speed, and minimum stability margin, respectively; in turbofan engine mode, two parameters are set, corresponding to fuel consumption rate and thrust, respectively.

[0068] S4. Construct a multi-source energy optimization design problem for hypersonic aircraft based on a neural network surrogate model. Under the condition of meeting the electric energy requirements of the aircraft, the fuel consumption rate Wf and thrust FN of the engine are used as optimization objectives, and the engine speed, turbine inlet temperature and stability margin are used as engine safety constraints to balance the fuel economy and flight quality of the aircraft.

[0069] The mathematical description of the specific optimization objective J in the turbofan engine mode is as follows:

[0070]

[0071] In the formula, P y,i P represents the power generation from the high-voltage stage, intermediate stage, and outer bypass duct. r,j P represents the power generation from the high-voltage stage, intermediate stage, and bypass gas turbine. z,k This represents the power generation extracted from the high-pressure and low-pressure shafts. FN0 and Wf0 are the engine's thrust and fuel consumption rate before multi-source energy extraction, respectively.

[0072] The mathematical description of the specific optimization objective in the ramjet engine mode is as follows:

[0073]

[0074] In the formula, P y P is the amount of electricity generated by induced draft gas power generation. r P is the amount of electricity generated by a gas turbine generator. o The power generation of oil and gas turbine generators, P x It stores energy for the battery.

[0075] Considering engine operational safety, it is necessary to ensure that the engine does not overheat, over-rev, or surge while extracting energy from multiple sources. Therefore, safety constraints are established for the turbine inlet temperature, mechanical shaft speed, and compressor stability margin during the multi-source energy optimization process. The mathematical description of the engine's safety constraint C is as follows:

[0076]

[0077] In the formula, T is the turbine inlet temperature, and N is the inlet temperature. HPSThe rotational speed of the high-pressure shaft is N. LPS To reduce the speed of the low-pressure shaft, SM HPC For the stability margin of the high-pressure compressor, SM LPC This refers to the stability margin of the low-pressure compressor.

[0078] S5 uses the multi-objective optimization algorithm NSGA-II to solve for the current optimal multi-source energy configuration. Each individual in the algorithm represents a set of power generation configuration parameters of a multi-source energy extraction system. The optimization process can be simply described as finding a set of power generation configuration parameters to optimize the engine's fuel consumption rate and thrust under the conditions of meeting the power demand and engine safety constraints.

[0079] The specific calculation process for NSGA-II to solve for the optimal fuel consumption rate and thrust includes the following steps:

[0080] S5-1, Set initialization parameters, set population size m = 100; set maximum number of iterations T under the same environment. max =100; Set the objective function dimension d1=2; In turbofan engine mode, set the population dimension d2=3; In ramjet engine mode, set the population dimension d2=4; Set the upper bound of the variable X. max =500; Sets the lower bound of the variable X min =0; Set the mutation probability a1 = 0.5, and the mutation probability a2 = 0.8;

[0081] S5-2, the first generation population is sorted non-dominated, and after selection, crossover and mutation operations, the first generation subpopulation is generated;

[0082] S5-3, merges the parent population and the offspring population into a new population;

[0083] S5-4: Perform fast non-dominated sorting on the merged population, calculate the crowding degree and record the frontier solution;

[0084] S5-5: Select elite individuals as the new parent population, and use selection, crossover, and mutation.

[0085] S5-6, Population generation T = T + 1, determine if T has reached the set maximum population generation T. max If T < T max Then return to step S5-3, if T≥T max Then output the leading edge solution;

[0086] S5-7: Record the final Pareto optimal solution and optimal variables. The optimal solution is the one with the minimum fuel consumption rate and the maximum thrust, and the optimal variables are the power generation configuration of the multi-source energy extraction system corresponding to the optimal solution.

[0087] S6 determines whether the current engine speed, turbine inlet temperature, and stability margin are within safety constraints; otherwise, it returns to S5.

[0088] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style of the specification is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

Claims

1. A method for hypersonic vehicle multi-source energy optimization based on a proxy model, characterized in that, include The following steps are required: S1. Establish a mechanism model for a parallel TBCC engine and a multi-source energy extraction system. The parallel TBCC engine is composed of a turbofan engine and a ramjet engine connected in parallel. Its parallel dual-channel configuration is relatively independent. The main components of the turbofan engine channel are: low-speed intake, low-pressure compressor, high-pressure compressor, main combustion chamber, high-pressure turbine, low-pressure turbine, bypass duct, mixing chamber, afterburner, and low-speed nozzle. The main components of the ramjet engine channel are: high-speed intake, isolator, combustion chamber, and high-speed nozzle. The multi-source energy extraction system model includes shaft power generation, bleed air power generation, gas turbine power generation, oil-gas turbine power generation, and a battery. When the engine operates in turbofan engine mode, the following energy extraction methods are used: 1) Shaft power extraction for power generation 2) Exhaust gas power generation 3) Gas turbine power generation, extracting energy from the high-pressure stage / intermediate stage / bypass duct. When the engine is operating in ramjet engine mode, the following energy extraction methods are used: 1) Exhaust gas power generation 2) Gas turbine power generation 3) Oil and gas turbine power generation 4) Storage battery; S2, in multiple typical flight conditions in turbofan engine mode and ramjet engine mode respectively, selects multiple typical flight conditions according to gradient, sets different power generation configuration parameters of multi-source energy extraction system to simulate the mechanistic model, and uses normalization method to process the sample data obtained from the simulation and build a sample library of proxy models. S3 utilizes the operating rules of the GA-BP neural network generalization mechanism model to establish a proxy model that can be simulated in real time. The flight altitude, flight Mach number, and power generation of the multi-source energy extraction system are used as inputs to the proxy model, while the engine's fuel consumption rate, thrust, speed, turbine inlet temperature, and stability margin are used as outputs to the proxy model. S4, construct a multi-source energy optimization design problem of hypersonic vehicles based on a neural network agent model, in which the specific fuel consumption of the engine is taken as the optimization objective, the engine speed, the turbine inlet temperature, and the stability margin are taken as the engine safety constraint conditions, and the fuel economy and flight quality of the vehicle are considered Wf and thrust FN are balanced. S5. The multi-objective optimization algorithm NSGA-II is used to solve for the current optimal multi-source energy configuration. Each individual in the algorithm represents a set of power generation configuration parameters of a multi-source energy extraction system. The optimization process can be simply described as finding a set of power generation configuration parameters to optimize the engine's fuel consumption rate and thrust under the conditions of meeting the power demand and engine safety constraints. S6 determines whether the current engine speed, turbine inlet temperature, and stability margin are within safety constraints; otherwise, it returns to S5.

2. The surrogate model based hypersonic vehicle multi-source energy optimization method of claim 1, wherein, The establishment of the real-time simulation proxy model of the GA-BP neural network specifically includes the following: The learning samples for this model are obtained through extensive simulations of the mechanistic model. Specifically, the flight altitude, flight Mach number, and power generation configuration parameters of the multi-source energy extraction system are set according to gradients. The engine fuel consumption rate, thrust, speed, and turbine inlet temperature under steady-state conditions of the mechanistic model are recorded as sample data. Based on this, the real-time simulation proxy model is implemented through a GA-BP neural network. The GA-BP neural network utilizes the global optimization capability of the genetic algorithm. The optimized weights and thresholds are input into the BP neural network training dataset. The output variables of the real-time simulation proxy model are optimized through the GA algorithm to obtain the optimal or relatively optimal weights and thresholds. The optimized weights and thresholds are then substituted into the BP neural network to train the sample data, thereby generating the real-time simulation proxy model.

3. The hypersonic vehicle multi-source energy optimization method based on a surrogate model as described in claim 2, characterized in that The structural parameters of the GA-BP neural network include: 1) Input layer parameters The input parameters are flight conditions and the power generation of the multi-source energy extraction system. Therefore, in the turbofan engine mode, five input neurons are set to represent flight altitude, flight Mach number, power generation from shaft power extraction, power generation from bleed air, and power generation from gas turbine; in the ramjet engine mode, six input neurons are set to represent flight altitude, flight Mach number, power generation from bleed air, power generation from gas turbine, power generation from oil and gas turbine, and energy storage of the battery. 2) Hidden layer parameters The number of neurons in the hidden layer has a significant impact on the convergence speed and generalization ability of a neural network. The number of neurons in the hidden layer can be calculated using empirical formulas: (1) wherein, N hid is the number of hidden layer neuron nodes, nn is the number of input layer neuron nodes, mm is the number of output layer neuron nodes, a is an adjustable constant in the interval of 1~10, and the final number of hidden layer is obtained by multiple trial and error of the value of a . 3) Output layer parameters The number of neurons in the output layer of the neural network is determined by the target requirements. It is used to predict the performance parameters of the engine under the current flight conditions and multi-source energy parameter configuration. Therefore, in the turbofan engine mode, 6 output neurons are set, corresponding to fuel consumption rate, thrust, turbine inlet temperature, high-pressure shaft speed, low-pressure shaft speed, and minimum stability margin, respectively; and 2 neurons are set in the turbofan engine mode, corresponding to fuel consumption rate and thrust, respectively.

4. The surrogate model based hypersonic vehicle multi-source energy optimization method of claim 1, wherein, The process of solving for the optimal fuel consumption rate and thrust using the multi-objective optimization algorithm NSGA-II includes the following steps: S1: Set initialization parameters, set population size. m =100; Sets the maximum number of iterations under the same environment. T max = 100; Set the dimension of the objective function. d 1=2; In turbofan engine mode, set the population dimension. d 2=3; In ramjet engine mode, set the population dimension. d 2=4; Set the upper bound of the variable. X max = 500; Sets the lower bound of the variable. X min = 0; sets the mutation probability. Set mutation probability ; S2: Perform non-dominated sorting on the first generation population, and generate the first generation subpopulation after selection, crossover and mutation operations; S3: Merge the parent and offspring populations into a new population; S4: Perform fast non-dominated sorting on the merged population, calculate the crowding degree, and record the frontier solution; S5: Select elite individuals as the new parent population, and perform selection, crossover, and mutation. S6: population number T = T +1, determine T whether the set maximum population number T max is reached, if then return to step S3, if then output the front solution;​ S7: Record the final Pareto optimal solution and optimal variables. The optimal solution is the solution with the minimum fuel consumption rate and the maximum thrust. The optimal variables are the power generation configuration of the multi-source energy extraction system corresponding to the optimal solution.

Images