A multi-source energy optimal configuration method for a hypersonic vehicle

Abstract

The application discloses a multi-source energy optimal configuration method of a hypersonic aircraft, and energy configuration with the lowest fuel consumption and the largest thrust is obtained by using a multi-objective adaptive covariance matrix and a chaotic search group algorithm (MOACCGA), so that the improved method of the adaptive covariance matrix is adapted to strong nonlinear changes of a system, the improved method of the chaotic search is used to avoid falling into local optimization, accurate distribution of multi-source electric energy of the hypersonic aircraft in a dynamic environment is realized, and the technical problem of energy optimal configuration of the hypersonic aircraft is solved.

Description

TECHNICAL FIELD

[0001] The application relates to an aero-engine, in particular to a multi-source energy optimal configuration method of a hypersonic aircraft, and belongs to the technical field of aerospace. BACKGROUND

[0002] The hypersonic aircraft is called the third revolutionary achievement in the history of aviation after the propeller and the jet propeller by military experts due to the characteristics of good stealth, fast flight speed, wide striking range and large effective payload. Under the support of new combined cycle power technology, in recent years, the hypersonic aircraft technology has developed by leaps and bounds. At present, the domestic research focuses on the key technology of the main power core machine, and the research on the effective electric energy generation method and efficient energy configuration of the hypersonic aircraft is relatively weak.

[0003] With the development of the aircraft in the direction of multi-electricity / total electricity, the demand for electricity of the hypersonic aircraft is also increasing, which puts forward higher requirements for the electric energy generation system on the aircraft. The multi-source energy extraction under the turbine-based combined power is a core technology to be solved. The turbine-based combined power applied to the hypersonic aircraft mainly extracts the main power intake / oil-gas total enthalpy, shaft power, heat energy and fuel cold The multi-source electric energy is generated by using the methods, which will affect the performance of the main power unit fuel consumption rate, the turbine power compressor stability margin, the ram power intake stability margin, the combustion stability and the like. The influence degree and mechanism of different energy extraction methods and parameters on the performance of the main power are different. Therefore, strengthening the research on the influence mechanism of the multi-source energy extraction on the performance of the main power and the multi-source energy optimal configuration method is beneficial to promoting the balanced development of the hypersonic aircraft system technology and meeting the requirements of the comprehensive optimization design method of the higher thrust-to-weight ratio and the lower fuel consumption rate. SUMMARY

[0004] The application aims at the deficiencies of the prior art, and provides a multi-source energy optimal configuration method of a hypersonic aircraft. The method adopts a multi-target adaptive covariance matrix and a chaotic search group algorithm MOACCGA to obtain an energy configuration scheme with the lowest fuel consumption rate and the maximum thrust, an improved method of the adaptive covariance matrix is adopted to adapt to the strong nonlinear change of the system, and an improved method of the chaotic search is adopted to avoid falling into local optimization, so that the multi-source electric energy of the hypersonic aircraft can be accurately distributed in a dynamic environment, and the technical problem of the energy optimal configuration of the hypersonic aircraft is solved.

[0005] In order to achieve the above application purpose, the application adopts the following technical scheme:

[0006] A multi-source energy optimization configuration method of a hypersonic vehicle, based on a parallel turbo-based combined cycle (TBCC) engine and a multi-source energy extraction system, contains two working modes of a turbofan engine and a ramjet engine;

[0007] The method is characterized in that: the power generation of each energy extraction mode is used as an optimization design variable, the fuel consumption rate and the thrust are used as optimization objective functions, the mechanical shaft speed, the turbine inlet temperature, the stability margin and the like are used as constraint conditions, a multi-objective adaptive covariance matrix and a chaos search group algorithm (MOACCGA) is used, an improved method of the adaptive covariance matrix is used to adapt to the strong nonlinear change of the system, and an improved method of the chaos search is used to avoid falling into a local optimum, so that the multi-source electric energy of the hypersonic vehicle is accurately distributed in a dynamic environment, the energy user configuration with the lowest fuel consumption rate and the maximum thrust is obtained, and the multi-source electric energy distribution with the Pareto optimal fuel consumption rate and thrust in the entire flight envelope is finally obtained; the method comprises the following steps:

[0008] S1, determining the working mode under the current flight state and selecting the corresponding multi-source energy extraction mode

[0009] According to the current flight height H and the flight Mach number Ma of the aircraft, the working mode of the current engine is confirmed and the corresponding multi-source energy extraction mode is selected; in the turbofan engine mode, there are three energy extraction modes: (1) shaft power extraction power generation, extraction from the high-pressure shaft and the low-pressure shaft; (2) bleed air power generation, extraction from the high-pressure stage, the intermediate stage and the outer bypass; (3) gas power generation, extraction from the high-pressure shaft and the low-pressure shaft; shaft power extraction power generation, that is, the high / low-pressure shaft of the turbofan engine is adjusted by the gear box to drive the motor to generate electricity; bleed air power generation, that is, the bleed air of the high-pressure stage / intermediate stage / outer bypass of the turbofan engine drives the motor to generate electricity; gas power generation, that is, the bleed air is combusted in the combustion chamber and then drives the motor to generate electricity; in the ramjet engine mode, there are four energy extraction modes: (1) bleed air power generation; (2) semiconductor temperature difference power generation; (3) oil-gas turbine power generation; (4) battery; bleed air power generation, that is, the bleed air of the ramjet engine drives the motor; semiconductor temperature difference power generation, that is, the heat from the combustion chamber wall surface forms a temperature difference, and the P and N type thermocouples generate a potential difference to supply power to the load; oil-gas turbine power generation, that is, the carbon-hydrogen fuel is cracked into mixed oil-gas by the combustion chamber cooling channel, and then the oil-gas turbine is used for power generation; the battery does not need to extract energy from the engine, and only the weight causes fuel compensation;

[0010] S2: determining multiple optimization objectives of the multi-source energy optimization configuration of the hypersonic vehicle

[0011] The fuel consumption rate Wf and the thrust FN of the engine are used as optimization objectives, so that the fuel economy and the thrust of the system are considered, the endurance of the hypersonic vehicle is improved, and the flight quality is also improved;

[0012] S3: Determine the constraint conditions and working requirements of the multi-source energy optimization configuration of the hypersonic vehicle

[0013] The constraint conditions of the turbofan engine mode are: turbine pre-temperature T, rotation speed N of the high-pressure shaft LPS , rotation speed N of the low-pressure shaft HPS , stability margin SM of the high-pressure compressor HPC , and stability margin SM of the low-pressure compressor LPS ; and the working requirements of the turbofan engine mode are: electric energy requirement P w ;

[0014] The constraint condition of the ramjet engine mode is: combustion stability SM of the combustion chamber COV ; and the working requirement of the ramjet engine mode is: electric energy requirement P c ;

[0015] S4: Optimize the energy distribution scheme of the multi-source energy extraction

[0016] Within a flight envelope, for flight height H, flight Mach number Ma, electric energy requirement P w and P c that dynamically change with the engine working mode, the output power of the multi-source energy extraction system is distributed by using a multi-objective adaptive covariance matrix and chaotic search group algorithm MOACCGA.

[0017] Further, since the fuel compensation loss and the thrust loss of the engine are caused by the energy extraction from the engine. In the turbofan engine mode, the multi-source energy optimization configuration of the hypersonic vehicle with the fuel consumption rate Wf and the thrust FN as the optimization objectives can be described as:

[0018]

[0019] In formula (1), P z,1 , P z,2 are the output powers of the high-pressure shaft and the low-pressure shaft of the turbofan engine mode for the shaft power extraction and power generation; P y,1 , P y,2 , P y,3 are the output powers of the high-pressure stage, the intermediate stage, and the outer duct bleed air of the turbofan engine mode for power generation; P r,1 , P r,2 , P r,1 are the output powers of the high-pressure stage, the intermediate stage, and the outer duct bleed gas of the turbofan engine mode for power generation;

[0020] In the ramjet engine mode, the multi-source energy optimization configuration of the hypersonic vehicle can be described as:

[0021]

[0022] P yq , P wc , P wl , P dc are output power of ramjet mode bleed air power generation, semiconductor thermoelectric power generation, oil vapor turbine power generation and battery respectively;

[0023] Reasonable distribution of output power of multi-source energy extraction system is carried out for multi-source energy optimal configuration, that is, in the environment of flight state change, the optimal energy distribution mode is obtained through optimization calculation based on multi-objective adaptive covariance matrix and chaos search group algorithm MOACCGA, and under the constraints of no over-temperature, no over-speed and no surge of the engine, the Pareto optimal solution of fuel consumption and thrust is obtained, which not only meets the power demand of the hypersonic aircraft, but also realizes the minimum fuel consumption and the maximum thrust of the system; The population in MOACCGA is divided into three types: discoverer, follower and patrolman, which behaves like the search behavior of animals. The discoverer is characterized by the optimization object, and seeks the optimal fitness value through fast non-dominated sorting; the follower adopts the evolutionary strategy of adaptive covariance matrix to obtain the reliable estimate of the path; in addition, the patrolman explores the huge search space by using chaos search; in the MOACCGA algorithm, the Pareto optimal front is obtained through fast non-dominated sorting; the evolutionary strategy of adaptive covariance matrix improves the response speed of the algorithm to adapt to the strong nonlinear change in the high dynamic environment; the high ergodicity of chaos search enhances the local search ability of the algorithm and overcomes the problem of falling into local optimum in the multi-peak solution space; the specific steps include the following:

[0024] S2-1, setting initialization parameters: setting the population size m=100; setting the maximum number of iterations T max =100 under the same environment; setting the target function dimension d1=2; in the turbofan engine mode, setting the population dimension d2=7, setting the upper limit of the variable X max =P w ; in the turbofan engine mode, setting the population dimension d2=4, setting the upper limit of the variable X max =P c ; setting the lower limit of the variable X min =0; setting the chaos variable u=4. In the actual working condition collection process, each individual in the population is a kind of energy distribution mode of multi-source energy optimal configuration;

[0025] S2-2, initializing the population: initializing the initial population and initializing the covariance matrix, in order to make the position distribution of the initial population more uniform, 70% of the individuals in the initial population are patrolmen for chaos search, after chaos search, the whole population is fast non-dominated sorted, and the mean value of the fitness values of all individuals in the population is obtained

[0026] S2-3, generating subgroup: after fast non-dominated sorting, selecting the optimal individual in the population as the discoverer, and selecting the top 50% individuals in the population as the followers, and the followers are evolved by adaptive covariance matrix;

[0027] S2-4, forming a new population: merging the parent population and the subgroup evolved by adaptive covariance to form a new population, and selecting 70% individuals in the new population as the scouts for chaotic search. After chaotic search, the population is fast non-dominated sorted, and new discoverers and followers are selected;

[0028] S2-5, iteration: iteratively repeating steps S2-1 to S2-4 until the number of iterations in the iteration formula reaches the maximum number of iterations T preset by the current environment max , obtaining the optimal power generation allocation mode under the current environment;

[0029] S2-6, record: recording the optimal multi-source power allocation scheme of each environment, and forming the dynamic multi-source power allocation data of the Pareto optimal fuel consumption rate and thrust in the entire flight envelope according to the obtained optimal multi-source power allocation scheme.

[0030] Further, in S2-2, the specific steps of initializing the population are as follows:

[0031] The position of the initial population is generated by formula (3):

[0032] X=X min +r·(X min +X max ) (3)

[0033] In formula (3), r represents a random number in 0 to 1;

[0034] Initialize the covariance matrix C. Let the matrix C have the standard orthogonal basis B=[b1,b2,...,b n ] of the eigenvector, and have the corresponding eigenvalues λ1 2 , λ2 2 ,..., λ n 2 , and the eigenvalue matrix D=diag(λ1,λ2,...,λ n ). The covariance matrix is represented as follows:

[0035]

[0036] In order to make the position distribution of the initial population more uniform, 70% of the individuals in the initial population can be selected as the scouts for chaotic search, and the process of chaotic search is as follows:

[0037] Step 1: generate chaotic variables

[0038] In the nth search, the chaotic variable cxof D different trajectories is randomly generated n d (d = 1, 2,..., D), not containing the five fixed points of the chaotic iteration equation

[0039] Step 2: Linear mapping

[0040] cx n d is linearly mapped to the optimization variable interval [X min , X max ], to obtain the position rx n d . The process of linear mapping is represented by the following formula:

[0041] rx n d = X min + cx n d (X max -X min ) (5)

[0042] Step 3: Chaos search

[0043] The target object x n d is subjected to chaos search to obtain a new value x n+1 d . The process of chaos search is represented by the following formula:

[0044] x n+1 d = x n d + β · rx n d (6)

[0045] Step 4: Update

[0046] The updated cx n d value, the update process is represented by the following formula:

[0047] cx n d = u · cx n d (1-cx n d ) (7)

[0048] In formula (7), u is a chaotic variable;

[0049] After chaos search, the whole population is fast non-dominated sorted, and the mean of all individual fitness values in the population is obtained by formula (8)

[0050]

[0051] Further, the specific steps of generating a subgroup in S2-3 are as follows:

[0052] After fast non-dominated sorting, the optimal individual in the population is selected as the discoverer, and the top 50% individuals in the population are selected as the followers. The followers perform adaptive covariance matrix evolution. The evolution process of the adaptive covariance matrix is as follows:

[0053] Step 1: Update the mean

[0054] The adaptive covariance matrix evolution strategy uses a learning rate α μ to control the update speed of the mean. Its update process is represented by the following formula:

[0055]

[0056] In the formula, x (t) is the sample of the t th generation.

[0057] Step 2: Control the step size

[0058] During the evolution process, the update speed of the standard deviation σ is largely dependent on the step size. The adaptive covariance matrix evolution strategy constructs an evolution path P σ by comparing the length of the evolution path P σ with the expected length under random selection, and adjusting the standard deviation σ accordingly;

[0059] The update process of the evolution path is represented by the following formula:

[0060]

[0061] In the formula, α σ is the update learning rate;

[0062] Under random distribution, the expected length of the evolution path is According to the proportion , the step size is adjusted. Therefore, the update process of the standard deviation σ is represented by the following formula:

[0063]

[0064] In the formula, d σ is the updated damping factor;

[0065] Step 3: Adaptive covariance matrix

[0066] with average step size y i Re-evaluate the covariance matrix C:

[0067]

[0068] The sample size of the population determines the accuracy and iteration speed of the algorithm, and the accuracy and iteration speed of the algorithm are two contradictory indicators. In order to solve this problem, two independent paths are used to adaptively update the covariance matrix C:

[0069] 3) Rank-μ update

[0070] In order to make the algorithm converge faster, the population information of the previous generations is used to make up for the small population size, and the mean of the covariance matrix C of each generation is used for updating:

[0071] C (t+1) ≈ avg(C λ (i) ; i = 1, 2,... t) (13)

[0072] The learning rate a cμ is introduced into the population information, and the updating process of C can be represented as:

[0073]

[0074] In equation (14), n is the dimension of the search space.

[0075] 4) Rank-1 update

[0076] In order to increase the probability of sampling the optimal sample in the sampling process, an exponential smoothing method is used to construct the evolutionary path P c , and the updating process of P c can be represented as:

[0077]

[0078] In equation (15), a cp is the learning rate of P c ;

[0079] According to the evolutionary path P c , the covariance matrix C is updated:

[0080]

[0081] In summary, Rank-μ and Rank-1 are combined, and the updating process of the covariance matrix C is represented as:

[0082]

[0083] The advantages and remarkable effects of the present application are as follows:

[0084] (1) The present application is directed to multi-source energy extraction of a hypersonic aircraft, considers multiple evaluation targets such as fuel consumption rate and thrust, determines optimization constraint conditions, reduces the fuel consumption rate and improves the thrust while meeting the power demand of the hypersonic aircraft. The power generation of each energy extraction mode is used as an optimization design variable, and a multi-objective adaptive covariance matrix and chaos search group algorithm MOACCGA is used to solve the variable, and finally a multi-source electric energy distribution scheme with Pareto optimal fuel consumption rate and thrust in the entire flight envelope is obtained, which provides necessary technical support for energy configuration of the hypersonic aircraft, and can reduce the fuel consumption rate and improve the thrust on the basis of ensuring that the engine does not overheat, does not surge and does not overheat.

[0085] (2) In terms of algorithm, according to the characteristics of the hypersonic aircraft, the fuel consumption and thrust of the hypersonic aircraft under the condition of change of the optimization design variable are considered to present strong nonlinear change and multi-peak distribution in the solution space, and under the premise of analyzing the energy distribution trend of the hypersonic aircraft, the thrust and fuel consumption rate are used as the target, and the safety is used as the constraint, the improved method of adaptive covariance matrix evolution is used to adapt to the strong nonlinear change of the system, the improved method of chaos search is used to overcome the problem of falling into local optimum, and a multi-objective adaptive covariance matrix and chaos search group algorithm MOACCGA is proposed to solve the energy configuration problem of the hypersonic aircraft. BRIEF DESCRIPTION OF DRAWINGS

[0086] Figure 1 It is a schematic diagram of a TBCC engine and a multi-source energy extraction system;

[0087] Figure 2 It is a whole optimization flowchart of multi-source energy configuration of a hypersonic aircraft;

[0088] Figure 3 It is a distribution diagram of thrust in the solution space;

[0089] Figure 4 It is a distribution diagram of fuel consumption rate in the solution space;

[0090] Figure 5 It is a flowchart of a multi-objective adaptive covariance matrix and chaos search group algorithm MOACCGA. DETAILED DESCRIPTION

[0091] The technical solutions of the present application will be described in detail below with reference to the drawings.

[0092] The multi-source energy optimization configuration method of the hypersonic aircraft provided by the present application is realized based on a parallel turbine-based combined cycle (TBCC) engine and a multi-source energy extraction system. Figure 1As shown in the figure, the two working modes contained in the parallel TBCC engine are: turbofan engine mode, ramjet engine mode, and the multiple energy extraction modes contained in the multi-source energy extraction system are: shaft power extraction power generation, bleed air power generation, gas power generation, semiconductor temperature difference power generation, oil gas turbine power generation, and battery. In the turbofan engine mode, three energy extraction modes are contained: (1) shaft power extraction power generation (extracted from the high-pressure shaft and the low-pressure shaft); (2) bleed air power generation (extracted from the high-pressure stage, the intermediate stage and the external duct); (3) gas power generation (extracted from the high-pressure shaft and the low-pressure shaft). Shaft power extraction power generation, that is, the high / low-pressure shaft of the turbofan engine drives the motor to generate electricity after gear box adjustment; bleed air power generation, that is, the motor is driven to generate electricity by the bleed air from the high-pressure stage / intermediate stage / external duct of the turbofan engine; gas power generation, that is, the motor is driven to generate electricity by the bleed air after combustion in the combustion chamber. In the ramjet engine mode, four energy extraction modes are contained: (1) bleed air power generation; (2) semiconductor temperature difference power generation; (3) oil gas turbine power generation; (4) battery. Bleed air power generation, that is, the motor is driven by the bleed air from the isolation section of the ramjet engine; semiconductor temperature difference power generation, that is, the heat from the wall surface of the combustion chamber forms a temperature difference, which causes the P and N type thermocouples to generate a potential difference, thereby supplying power to the load; oil gas turbine power generation, that is, the hydrocarbon fuel is cracked into mixed oil gas by heat absorption in the combustion chamber cooling channel, and then the mixed oil gas is introduced into the oil gas turbine to generate electricity; the battery does not need to extract energy from the engine, and only the weight causes fuel compensation.

[0093] As shown in the figure, the overall optimization process of the multi-source energy configuration of the hypersonic vehicle includes four steps of S1-S4. Figure 2

[0094] S1: Determine the working mode and the multi-source energy extraction mode under the current flight state

[0095] According to the current flight height H and the flight Mach number Ma of the aircraft, the working mode of the current engine is confirmed and the corresponding multi-source energy extraction mode is selected.

[0096] S2: Determine the multiple optimization objectives of the multi-source energy optimization configuration of the hypersonic vehicle

[0097] The multi-source energy extraction will affect the fuel consumption rate Wf and the thrust FN of the engine, and the fuel consumption rate and the thrust as the main indicators can directly reflect the economy and performance of the engine. Therefore, the fuel consumption rate and the thrust of the main engine are taken as the optimization objectives, so that the system has good fuel economy and sufficient thrust, which can improve the endurance of the supersonic vehicle and make it have good flight quality.

[0098] S3: Determine the constraint conditions and working requirements of the multi-source energy optimization configuration of the hypersonic vehicle

[0099] The constraint conditions of the turbofan engine mode are: turbine inlet temperature T, high-pressure shaft speed N​LPS , the rotation speed of the low-pressure shaft N HPS , the stability margin of the high-pressure compressor SM HPC , and the stability margin of the low-pressure compressor SM LPS ; the working requirement of the turbofan engine mode is the electric power requirement P w .

[0100] The constraint condition of the ramjet engine mode is the combustion stability of the combustion chamber SM COV ; the working requirement of the ramjet engine mode is the electric power requirement P c .

[0101] S4: optimizing the energy distribution scheme of the multi-source energy extraction

[0102] Within a flight envelope, for the dynamic environment of the change of the flight height, the flight Mach number, and the working mode, the output power of the multi-source energy extraction system is distributed by using the multi-objective adaptive covariance matrix and the chaotic search group algorithm MOACCGA.

[0103] Specifically, the fuel compensation loss and the thrust loss of the engine are both caused by the energy extraction from the engine. In the turbofan engine mode, for the multi-source energy optimization configuration of the hypersonic vehicle with the specific fuel consumption Wf and the thrust FN as the optimization objectives, it can be described as:

[0104]

[0105] In formula (1), P z,1 , P z,2 are the output powers of the high-pressure shaft and the low-pressure shaft of the turbo engine mode for the shaft power extraction and power generation; P y,1 , P y,2 , P y,3 are the output powers of the high-pressure stage, the intermediate stage, and the outer duct bleed air of the turbo engine mode for power generation; P r,1 , P r,2 , P r,1 are the output powers of the high-pressure stage, the intermediate stage, and the outer duct bleed air of the turbo engine mode for power generation.

[0106] In the ramjet engine mode, the multi-source energy optimization configuration of the hypersonic vehicle can be described as:

[0107]

[0108] In formula (2), P yq , P wc , P wl , P dc are the output powers of the bleed air power generation, the semiconductor thermoelectric power generation, the oil turbine power generation, and the battery respectively.

[0109] The multi-source energy extraction system's output power is rationally allocated to optimize multi-source energy configuration. Specifically, in a changing flight environment, the optimal energy allocation is calculated using a multi-objective adaptive covariance matrix and a chaotic search cluster algorithm (MOACCGA). Under the constraints of engine overheating, overspeed, and surge, the Pareto optimal solution for fuel consumption and thrust is achieved. This solution not only meets the hypersonic vehicle's power requirements but also minimizes fuel consumption and maximizes thrust. The population in the MOACCGA can be divided into three types: discoverers, followers, and patrollers, each exhibiting animal-like search behavior. Discoverers, characterized by optimization objects, seek the optimal fitness value through a fast non-dominated sorting algorithm. Followers employ an evolutionary strategy based on an adaptive covariance matrix to obtain reliable path estimates. Furthermore, patrollers utilize chaotic search to explore the vast search space. In the MOACCGA algorithm, the Pareto optimal frontier is obtained through fast non-dominated sorting; the evolutionary strategy of the adaptive covariance matrix can improve the response speed of the algorithm to adapt to strong nonlinear changes in highly dynamic environments; and the highly ergodic chaotic search enhances the local search capability of the algorithm and overcomes the problem of the algorithm falling into local optimality in the multi-peak solution space.

[0110] like Figure 3 As shown in , the thrust changes strongly nonlinearly in the solution space and has multiple maximum points; Figure 4 As shown in Figure 2, the fuel consumption rate varies strongly nonlinearly in the solution space and has multiple minimum points. Therefore, it can be concluded that the changes in thrust and fuel consumption with power generation have the following characteristics:

[0111] 1) Thrust and fuel consumption changes are highly nonlinear and unpredictable. Multi-source energy extraction can cause sudden changes in compressor efficiency, altering the engine's internal thermal cycle and further impacting engine performance. When power generation capacity reaches a certain value, it can suddenly rise or fall.

[0112] 2) Changing the power generation from multiple energy sources causes a sudden change in fuel consumption and thrust, resulting in a multi-peak distribution and multiple extreme values ​​in the solution space. The dotted lines represent the maximum and minimum values ​​in the solution space.

[0113] like Figure 5 As shown in Figure 2, the optimization process based on the multi-objective adaptive covariance matrix and chaotic search group algorithm MOACCGA includes the following steps:

[0114] S4-1: Setting initialization parameters

[0115] Set the population size m = 100; set the 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 = 7, set the variable upper bound X max= P w ; In turbofan engine mode, set population dimension d2 = 4, set variable upper limit X max = P c ; Set variable lower limit X min = 0; Set chaos variable u = 4. In the actual working condition acquisition process, each individual in the population is an energy allocation method of multi-source energy optimal configuration.

[0116] S4-2: Initialize population

[0117] Generate the position of the initial population by formula (3):

[0118] X = X min + r·(X min + X max ) (3)

[0119] In formula (3), r represents a random number within 0 to 1.

[0120] Initialize the covariance matrix C, set the matrix C to have the standard orthogonal basis B = [b1, b2,..., b n ] of the eigenvector, and have corresponding eigenvalues λ1 2 , λ2 2 ,..., λ n 2 , the eigenvalue matrix D = diag(λ1, λ2,..., λ n ). The covariance matrix is expressed as follows:

[0121]

[0122] In order to make the position distribution of the initial population more uniform, 70% of the individuals in the initial population are selected as scouts for chaos search, and the process of chaos search is as follows:

[0123] Step 1: Generate chaos variable

[0124] In the nth search, generate D different chaotic variables cx n d (d = 1, 2,..., D) of different trajectories at random, which do not contain the five fixed points of the chaotic iteration equation

[0125] Step 2: Linear mapping

[0126] Map cx n d to the optimization variable interval [X min , X max ] linearly to obtain the position rx n dThe process of linear mapping is represented by the following equation:

[0127] rx n d = X min + cx n d (X max - X min ) (5)

[0128] Step 3: Chaos search

[0129] Chaotic search is performed on the target object x n d to obtain a new value x n+1 d . The process of chaotic search is represented by the following equation:

[0130] x n+1 d = x n d + β · rx n d (6)

[0131] Step 4: Update

[0132] The updated value of cx n d is represented by the following equation:

[0133] cx n d = u · cx n d (1 - cx n d ) (7)

[0134] In equation (7), u is a chaotic variable.

[0135] After chaotic search, the entire population is quickly non-dominated sorted, and the mean value of the fitness values of all individuals in the population is obtained by equation (8)

[0136]

[0137] S4-3: Generate sub-population

[0138] After quick non-dominated sorting, the best individual in the population is selected as the discoverer, and the top 50% individuals in the population are selected as the followers. The followers evolve the adaptive covariance matrix, and the evolution process of the adaptive covariance matrix is as follows:

[0139] Step 1: Update the mean value

[0140] Adaptive covariance matrix evolution strategy using learning rate α μ Controls the update speed of the mean. The update process is expressed as follows:

[0141]

[0142] Where x (t) is the sample of generation t.

[0143] Step 2: Control the step length.

[0144] In the evolution process, the update speed of the standard deviation σ depends largely on the step size. Adaptive covariance matrix evolution strategy constructs the evolution path P σ , by comparing the evolutionary path P σ The length of and the expected length under random selection can be adjusted accordingly with the standard deviation σ.

[0145] The updating process of the evolution path is expressed as follows:

[0146]

[0147] Where, α σ is the updated learning rate.

[0148] Under random distribution, the expected length of the evolutionary path is According to the proportion Adjust the step size. Therefore, the update process of the standard deviation σ is expressed as follows:

[0149]

[0150] Where, d σ is the updated damping factor.

[0151] Step 3: Adaptive covariance matrix

[0152] With average step length y i Re-evaluate the covariance matrix C:

[0153]

[0154] When the sample size is large enough, the estimation of the covariance matrix is ​​reliable. However, in each generation, a smaller sample population is required for fast iteration. To solve this problem, two independent paths are used to adaptively update the covariance matrix C.

[0155] 5) Rank-μ Update

[0156] In order to make the algorithm converge faster, the population information of previous generations is used to make up for the disadvantage of small population size, and the mean of the covariance matrix C of each generation is used to update

[0157] C (t+1) ≈avg(C λ (i) ; i = 1, 2,... t) (13)

[0158] The learning rate a cμ is introduced into the population information, and the updating process of C can be expressed as:

[0159]

[0160] In equation (14), n is the dimension of the search space.

[0161] 6) Rank-1 updating

[0162] In order to make the probability of taking the optimal sample larger in the sampling process, the evolutionary path P c is constructed by using the exponential smoothing method. c The updating process of P

[0163]

[0164] In equation (15), a cp is the learning rate of P c .

[0165] According to the evolutionary path P c , the covariance matrix C is updated as:

[0166]

[0167] In summary, the updating process of the covariance matrix C can be expressed as:

[0168]

[0169] S4-4: Form a new population

[0170] The parent population and the offspring population after adaptive covariance evolution are combined to form a new population, and 70% of the individuals in the new population are selected as scouts for chaotic search. After chaotic search, the population is quickly non-dominated sorted, and new discoverers and followers are selected.

[0171] S4-5: Iteration

[0172] Steps S4-3 to S4-4 are repeatedly executed until the number of iterations in the iteration formula reaches the maximum number of iterations T max preset by the current environment, and the optimal power generation allocation mode under the current environment is obtained.

[0173] S4-6: Record

[0174] Record the optimal multi-source electric energy distribution scheme of each environment, and form the dynamic multi-source electric energy distribution data of the Pareto optimal fuel consumption rate and thrust in the entire flight envelope according to the obtained optimal multi-source electric energy distribution scheme.

[0175] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described herein.

[0176] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit the present application; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements should be within the protection scope of the present application.

Claims

1. A method for multi-source energy optimal configuration of a hypersonic aircraft, based on a parallel turbo-based combined cycle (TBCC) engine and a multi-source energy extraction system, comprising a turbofan engine and a ramjet engine in two working modes; characterized in that Taking the power generation of each energy extraction mode as the optimal design variable, taking the fuel consumption and thrust as the optimal objective function, taking the mechanical shaft speed, turbine inlet temperature and stability margin as the constraint condition, using a multi-objective adaptive covariance matrix and chaos search group algorithm (MOACCGA), using the improved method of adaptive covariance matrix to adapt to the strong nonlinear change of the system, and using the improved method of chaos search to avoid falling into local optimum, the method realizes accurate allocation of multi-source electric energy of the hypersonic aircraft in a dynamic environment, obtains the energy user configuration with the lowest fuel consumption and the maximum thrust, and finally obtains the multi-source electric energy distribution with the lowest fuel consumption and the maximum thrust in the entire flight envelope; The method comprises the following steps: S1, determining the working mode under the current flight state and selecting the corresponding multi-source energy extraction mode According to the current flight height H and flight Mach number Ma of the aircraft, the working mode of the current engine is confirmed and the corresponding multi-source energy extraction mode is selected; the turbofan engine mode comprises three energy extraction modes: (1) shaft power extraction power generation, which extracts from two parts of the high-pressure shaft and the low-pressure shaft; (2) bleed air power generation, which extracts from three parts of the high-pressure stage, the intermediate stage and the outer duct; (3) gas power generation, which extracts from two parts of the high-pressure shaft and the low-pressure shaft; shaft power extraction power generation, that is, the high / low-pressure of the turbofan engine drives the motor to generate electricity after gear box adjustment; bleed air power generation, that is, the motor is driven by the bleed air from the high-pressure stage / intermediate stage / outer duct of the turbofan engine to generate electricity; gas power generation, that is, the bleed air is burned in the combustion chamber and then drives the motor to generate electricity; the ramjet engine mode comprises four energy extraction modes: (1) bleed air power generation; (2) semiconductor temperature difference power generation; (3) oil-gas turbine power generation; (4) battery; bleed air power generation, that is, the motor is driven by the bleed air from the isolation section of the ramjet engine; semiconductor temperature difference power generation, that is, heat is introduced from the wall surface of the combustion chamber, a potential difference is generated between the P-type and N-type thermocouples due to the temperature difference, and then power is supplied to the load; oil-gas turbine power generation, that is, the carbon-hydrogen fuel is cracked into mixed oil-gas by absorbing heat in the combustion chamber cooling channel, and then the oil-gas turbine is used for power generation; the battery does not need to extract energy from the engine, and only the weight causes fuel compensation; S2: determining multiple optimization objectives of the multi-source energy optimal configuration of the hypersonic aircraft Taking the fuel consumption Wf and the thrust FN of the engine as the optimization objectives, the fuel economy and the thrust of the system are considered, the endurance of the hypersonic aircraft is improved, and the flight quality is also improved; S3: determining the constraint conditions and working requirements of the multi-source energy optimal configuration of the hypersonic aircraft The constraints of the turbofan engine mode are: the turbine inlet temperature T, the rotation speed N of the high-pressure shaft LPS , the rotation speed N of the low-pressure shaft HPS , the stability margin SM of the high-pressure compressor HPC and the stability margin SM of the low-pressure compressor LPS ; the work requirements of the turbofan engine mode are: the electric power requirement P w ; The constraint of the ramjet mode is the combustion stability SM of the combustion chamber COV The work requirement of the ramjet mode is the electrical power requirement P c ; S4: optimizing the energy distribution scheme of the multi-source energy extraction Within a flight envelope, for flight altitude H, flight Mach number Ma, the electric power demand P dynamically changing with engine operating mode w and P c , the output power of the multi-source energy extraction system is allocated by using a multi-objective adaptive covariance matrix and chaos search group algorithm MOACCGA.

2. The method of claim 1, wherein: Since the fuel compensation loss and the thrust loss of the engine are caused by the extraction of energy from the engine, in the turbofan engine mode, the multi-source energy optimal configuration of the hypersonic aircraft with the fuel consumption Wf and the thrust FN as the optimization objectives is described as follows: In formula (1), P z,1 , P z,2 is the output power of the turbine engine mode high-pressure shaft, low-pressure shaft shaft work extraction power generation; P y,1 , P y,2 , P y,3 is the output power of the turbine engine mode high-pressure stage, intermediate stage, bypass air power generation; P r,1 , P r,2 , P r,3 is the output power of the turbine engine mode high-pressure stage, intermediate stage, bypass gas power generation; In the ramjet engine mode, the multi-source energy optimal configuration of the hypersonic aircraft is described as follows: In formula (2), P yq , P wc , P wl , P dc are output powers of ramjet mode bleed air power generation, semiconductor thermoelectric power generation, oil vapor turbine power generation, and battery, respectively. Reasonable allocation of the output power of the multi-source energy extraction system is performed for multi-source energy optimization configuration, that is, in a changing environment of flight states, an optimal energy allocation mode is obtained through optimization calculation based on a multi-target adaptive covariance matrix and a chaotic search swarm algorithm MOACCGA, and under the constraints of no over-temperature, no over-speed and no surge of the engine, a Pareto optimal solution of fuel consumption and thrust is obtained, which not only meets the power demand of a hypersonic aircraft, but also realizes the lowest fuel consumption and the largest thrust of the system; the population in the MOACCGA algorithm is divided into three types: discoverers, followers and scouts, which behave like the search behavior of animals, the discoverers are characterized by optimization objects, and the optimal fitness value is sought through fast non-dominated sorting; the followers adopt an evolutionary strategy of an adaptive covariance matrix to obtain a reliable estimate of the path; In addition, the scouts explore a huge search space by chaotic search; in the MOACCGA algorithm, the Pareto optimal front is obtained through fast non-dominated sorting; the evolutionary strategy of the adaptive covariance matrix improves the response speed of the algorithm to adapt to strong nonlinear changes in a high dynamic environment; the chaotic search with high ergodicity enhances the local search ability of the algorithm and overcomes the problem of falling into local optimum in the multi-peak solution space; and specifically includes the following steps: S2-1, setting initialization parameters: setting population size m = 100; setting the maximum number of iterations T under the same environment max = 100; setting target function dimension d1 = 2; in the turbofan engine mode, setting population dimension d2 = 7, setting variable upper limit X max = P w ; in the turbofan engine mode, setting population dimension d2 = 4, setting variable upper limit X max = P c ; setting variable lower limit X min = 0; setting chaos variable u = 4, in the actual working condition collection process, each individual in the population is an energy distribution mode of multi-source energy optimization configuration; S2-2, initializing population: initializing the initial population and initializing the covariance matrix, in order to make the position distribution of the initial population more uniform, selecting 70% of the individuals in the initial population as scouts to perform chaotic search, after chaotic search, performing fast non-dominated sorting on the entire population, and obtaining the mean value of the fitness values of all individuals in the population S2-3, generating a sub-population: after fast non-dominated sorting, the optimal individual in the population is selected as a discoverer, and the top 50% individuals in the population are selected as followers, and the followers are evolved by an adaptive covariance matrix; S2-4, forming a new population: the parent population and the sub-population evolved by the adaptive covariance matrix are combined to form a new population, and 70% of the individuals in the new population are selected as scouts for chaotic search, and the population after chaotic search is fast non-dominated sorted to select new discoverers and followers; S2-5, iteration: iteratively performing steps S2-1 to S2-4 until the number of iterations in the iteration formula reaches the maximum number of iterations T preset in the current environment max , obtaining the optimal power generation distribution mode under the current environment; S2-6, recording: record the optimal multi-source power allocation scheme of each environment, and form the dynamic multi-source power allocation data of the Pareto optimal fuel consumption and thrust in the entire flight envelope according to the obtained optimal multi-source power allocation scheme.

3. The method of claim 2, wherein: In S2-2, the specific steps of initializing the population are as follows: The positions of the initial population are generated by formula (3): X = X min + r - (X min + X max ) (3) In formula (3), r represents a random number in 0 to 1; Initialize the covariance matrix C, let the matrix C have the standard orthogonal basis B = [b1, b2,..., b n ] of the eigenvectors and have the corresponding eigenvalues λ1 2 , λ2 2 ,..., λ n 2 , the eigenvalue matrix D = diag(λ1, λ2,..., λ n ), and the covariance matrix is expressed as follows: In order to make the position distribution of the initial population more uniform, 70% of the individuals in the initial population are selected as scouts for chaotic search, and the process of chaotic search is as follows: Step 1: generate chaotic variables In the nth search, the chaotic variables cx of D different trajectories are randomly generated n d d = 1, 2,..., D, excluding the 5 fixed points of the chaotic iteration equation Step 2: linear mapping cx n d linearly mapped to the optimization variable interval [X min ,X max ], resulting in a position rx n d , the process of linearly mapping being represented by the following equation: rx n d = x min + cx n d (x max - x min ) (5) Step 3: chaotic search for the target object x n d chaotic search is performed to obtain a new value x n+1 d The process of chaotic search is represented by the following equation: x n+1 d = x n d + β · r x n d (6), β is the chaos search step scaling factor; Step 4: update Updated cx n d The update process is represented by the following equation: cx n d = u - cx n d (1 - cx n d ) (7) In formula (7), u is a chaotic variable; After chaos search, the fast non-dominated sorting is performed on the whole population, and the mean value of the fitness of all individuals in the population is obtained by formula (8) 4. The method of claim 2, wherein: In S2-3, the specific steps of generating a sub-population are as follows: After fast non-dominated sorting, the optimal individual in the population is selected as a discoverer, and the top 50% individuals in the population are selected as followers, and the followers are evolved by an adaptive covariance matrix, and the evolution process of the adaptive covariance matrix is as follows: Step 1: update the mean Adaptive covariance matrix evolution strategy utilizes a learning rate a μ The update speed of the mean value is controlled, and its update process is represented as follows: In the formula, x (t) is the sample of the tth generation, and λ is the number of samples generated and participating in statistical update in each generation. Step 2: control the step size In the evolution process, the updating speed of the standard deviation σ is largely dependent on the step value, and the adaptive covariance matrix evolution strategy constructs the evolution path P σ , adjusts the standard deviation σ by comparing the length of the evolution path P σ with the expected length under random selection. The update process of the evolution path is represented by the following formula: In the formula, α σ is an updated learning rate; The expected length of the evolutionary path under random distribution is According to the proportion The step size is adjusted, so the updating process of the standard deviation σ is represented by the following formula: In the formula, d σ is the updated damping factor; Step 3: adaptive covariance matrix With average step length y i Re-evaluate the covariance matrix C: The sample size of the population determines the accuracy and iteration speed of the algorithm, and the accuracy and iteration speed of the algorithm are two contradictory indicators. In order to solve this problem, two independent paths are used to adaptively update the covariance matrix C; 1) Rank-μ update In order to make the algorithm converge faster, the population information of the previous generation is used to make up for the small population size, and the mean of the covariance matrix C of each generation is used for updating: C (t+1) ≈avg(C λ (i) ; i = 1, 2,... t) (13) The learning rate a cμ The update process of C is represented as: In formula (14), n is the dimension of the search space; 2) Rank-1 update In order to take the optimal sample probability greater in the sampling process, the evolutionary path P is constructed by exponential smoothing method c , the update process of P c is represented as: In formula (15), a cp is a learning rate of P c . According to the evolutionary path P c Update the covariance matrix C: In summary, Rank-μ and Rank-1 are combined, and the update process of the covariance matrix C is represented as: Where, α c1 is the Rank-1 learning rate of the covariance matrix C.

Images