A control method for energy efficiency optimization of an electric aircraft environmental control system

Abstract

The application discloses a kind of aircraft electrically powered environmental control system (EECS) energy efficiency optimization control method, change traditional area control strategy into the combination of three-variable optimization and double-target control, with the coefficient of performance (COP) of system as index, by solving the optimal solution set of optimization variable under different working conditions to maximize COP.On this basis, the value of two control quantities is calculated by the controller to meet the performance requirements of the system.First, a GRNN model is established, which is trained by a large amount of offline data under various conditions, so that it can accurately predict the steady-state COP of the system under different conditions.Then, the DE algorithm searches for the optimal solution set of optimization variables corresponding to the optimal COP within the solution space that satisfies the performance constraints of the components.Finally, the corresponding control quantities are solved at the optimal operating point by the controller, thereby achieving energy efficiency optimization control.

Description

Technical Field

[0001] This invention relates to aircraft cabin environmental control, and more particularly to a control method for optimizing the energy efficiency of an aircraft electric environmental control system (EECS), belonging to the field of energy efficiency management technology for aircraft thermal management systems. Background Technology

[0002] Aircraft environmental control systems (EHS) provide ventilation and temperature control by supplying air to the cabin, maintaining a comfortable environment for passengers. Currently, most EHS systems utilize engine bleed air, employing a series of control and power optimization techniques to adjust the air circulation components to appropriate temperature and pressure for cabin supply. However, exhaust air from the engine negatively impacts fuel consumption. To ensure engine operation in economic mode, the bleed air temperature and pressure are typically set to constant values, often exceeding the actual needs of the EHS, resulting in significant power waste. With increasing flight mission complexity and higher overall aircraft performance requirements, this approach faces growing limitations. The Electric Environmental Control System (EECS) eliminates traditional pneumatic systems, converting engine bleed air into ambient air compressed by an electric compressor. This system employs a two-stage compression and two-stage expansion cooling method, with multiple valves coordinating internal circulation. The electric compressor and electric fan can freely adjust their speeds according to different operating conditions. For the same flight mission, the same control objective can be achieved through different flow distributions between different channels, thus enabling energy efficiency optimization. However, the entire system is highly nonlinear, has many actuators, and there is strong coupling between control targets. Traditional control strategies cannot take into account the synergistic effect between various control channels, making it difficult to achieve optimal system energy efficiency.

[0003] From the current research status both domestically and internationally, in terms of control strategies, the commonly used regional control strategy involves dividing the EECS (Extended Takeoff Control System) into several functionally independent regions to achieve order reduction, and then designing independent control algorithms for each region. Optimization technology research mainly focuses on component-level parameter optimization and system architecture optimization. For example, some scholars have modeled refrigeration systems using the heat flow method and parameter matching method, respectively, and provided optimal structural and dimensional parameters for the system from the perspectives of minimum takeoff weight and fuel loss. However, for specific systems and specific mission requirements, how to combine optimization methods with control methods to achieve optimization by adjusting the system's operating state still lacks a mature theoretical framework and evaluation methods.

[0004] In recent years, the rise of artificial neural networks has attracted widespread attention from scholars. The unique advantage of neural networks in control and optimization lies in their ability to approximate nonlinear functions with high precision. In optimization problems with insufficient information about the object, neural networks are often used as substitutes. By establishing a neural network with a suitable structure based on the specific problem, the object can be approximated with high precision using limited data. Combining numerical optimization algorithms to solve the trained neural network model can yield an approximate optimal solution to the original problem. Currently, there are few reports on the application of neural networks in the design of controllers and optimizers in the field of environmental control systems. However, neural networks have been widely used in the optimization and control of complex systems and processes, and have certain reference value. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to address the shortcomings of the above-mentioned background technology by providing a control method for optimizing the energy efficiency of an aircraft electric environmental control system (EECS), which adopts the following technical solution: A control method for optimizing the energy efficiency of an aircraft electric environmental control system, wherein the EECS uses an additional electric motor-driven compressor to compress ambient air as an air source, directly introducing air from the ambient atmosphere, the electric motor-driven compressor compresses ram air, which is cooled by a primary radiator and then pressurized by the compressor, and then enters a secondary heat exchanger for further cooling, the cooled air is cooled and dehumidified by a condenser, and then enters a first-stage and a second-stage turbine cooler for further cooling, and finally the cooled air is supplied to the cabin to balance the heat load;

[0006] Its key features include: transforming the traditional regional control strategy into a combination of three-variable optimization and dual-objective control; using the coefficient of performance (COP) of the energy efficiency control system (EECS) as an indicator; maximizing the COP by solving for the optimal solution set of the optimization variables under different operating conditions; and then, using a controller to calculate the values ​​of two control variables to meet the system's performance requirements. First, a generalized regression neural network model is established and trained using a large amount of offline data under various operating conditions to predict the steady-state COP of the EECS under different conditions. Subsequently, a differential evolution algorithm searches for the optimal COP-corresponding optimization variable solution set within the solution space that satisfies component performance constraints. Finally, the controller solves for the corresponding control variables at the optimal operating point, thereby achieving energy efficiency optimization control.

[0007] Includes the following steps:

[0008] S1, Design the control scheme for EECS and build the EECS model.

[0009] Based on meeting the cabin's cooling and air supply requirements, and according to the component configuration, control functions, and architecture requirements of the EECS (Electrical Air Supply Control System), its optimization variables, control objects, and control quantities are designed. The control objects of the EECS are the air supply volume that meets the EECS performance requirements and the air supply temperature calculated from the cooling and air supply volumes. The actuators include an electric compressor, a cold air duct valve, a bypass valve, an economic cooling valve, and a low-limit valve. Among them, the bypass valve, economic cooling valve, and low-limit valve all change the heat flow rate, which can adjust the temperature of the corresponding node and ultimately affect the air supply temperature. The cold air duct valve changes the cold flow rate. To maximize the COP (Coefficient of Performance) of the EECS, the opening degree of the bypass valve, economic cooling valve, and low-limit valve are selected as optimization variables. Based on this, the cold air duct valve opening degree is used as a control quantity to meet the air supply temperature requirement, and the electric compressor speed is used as a control quantity to meet the air supply volume requirement.

[0010] S2, Obtain training samples for the COP prediction model.

[0011] The steady-state COP of the EECS under different environmental and operational parameters is predicted by a generalized regression neural network (GRNN). The input of the GRNN includes three environmental parameters: cruise altitude, flight Mach number, and cabin thermal load, and three operational parameters: bypass valve opening, economic cooling valve opening, and lower limit valve opening. The output of the GRNN is the outlet temperature of the electric compressor and the cabin air supply temperature. A large number of simulation data is collected to collect learning samples. The minimum number of samples is about 2400 sets to better describe the characteristics of the EECS. Specifically, the three environmental parameters of cruise altitude, flight Mach number, and cabin thermal load are set according to a gradient. Under one set of environmental parameters, the three operational parameters of bypass valve opening, economic cooling valve opening, and lower limit valve opening are set according to a gradient. Two PID controllers are used to control the electric compressor speed and the cooling duct valve opening to meet the cabin air supply and cooling requirements, respectively. The outlet temperature of the electric compressor and the cabin air supply temperature under steady state are recorded as learning samples.

[0012] S3, Construct the topology of the generalized regressive neural network GRNN.

[0013] GRNN is a four-layer network, including an input layer, a pattern layer, a summation layer, and an output layer. The input layer has six neurons, corresponding to the current cruise altitude, flight Mach number, cabin thermal load, bypass valve opening, economy cooling valve opening, and lower limit valve opening, respectively. The input layer neurons directly pass the input variables to the pattern layer. The number of neurons in the pattern layer is equal to the number of training samples. The output of a single neuron is:

[0014]

[0015] Where X = [X1, X2, ..., X6] are the six input vectors of the network, X iσ is the learning sample corresponding to the i-th neuron, and σ is the neuron width.

[0016] The summation layer uses two types of neurons for summation calculations, with the total number being the output vector dimension plus one; the first type directly adds all the outputs of the pattern layer, and its output S D It can be described as

[0017]

[0018] The second method involves weighted summation of the outputs of the mode layer, resulting in the output S. N It can be described as

[0019]

[0020] Among them, y ij It is the j-th element in the i-th output sample.

[0021] The output layer has two neurons, each corresponding to the outlet temperature T of the electric compressor. ec and cabin air supply temperature T cab The output layer divides the output of the summation layer, that is:

[0022]

[0023] Assuming the cabin air recirculation flow rate is half the supply air volume, and taking the temperature rise of the recirculated air after passing through the fan as 4°C, the COP of the EECS is obtained by the following formula:

[0024]

[0025] Where Q is the cabin cooling capacity, W is the system power consumption of EECS, and C p T is the specific heat capacity of air at constant pressure. sa For gas supply temperature, T amb For ambient temperature, This refers to the gas supply flow rate.

[0026] S4, train the generalized regression neural network (GRNN) model;

[0027] The learning samples obtained in S2 were divided into two groups. One group was used as the training set to train the generalized regression neural network, and the other group was used as the test set to verify the accuracy of GRNN in predicting COP of systems under different operating conditions.

[0028] S5, Define the individual and fitness functions for the differential evolution algorithm;

[0029] In the differential evolution algorithm, each individual represents a set of valve openings. The individual value after each iteration is substituted into the prediction model to obtain the corresponding prediction value, which is then converted to obtain the COP value as the fitness value of the differential evolution algorithm.

[0030] S6, Set the parameters of the differential evolution algorithm

[0031] The population size of the differential evolution algorithm is 30-100, the maximum number of iterations is 100-200, and the values ​​of the mutation factor and crossover factor are selected according to the quality of the algorithm's convergence results.

[0032] S7, impose constraints on the differential evolution algorithm based on EECS performance requirements and component operating characteristics;

[0033] The solution set obtained by the differential evolution algorithm needs to meet the following conditions: First type of condition: each set of solutions needs to be generated under the same cooling capacity and gas supply conditions; 2) each set of solutions should take into account the EECS operating status, the valve opening should be within 0-90 degrees, and the compressor outlet temperature and condenser inlet temperature should be within the normal operating range.

[0034] For the first type of condition, since the GRNN's learning samples have already been processed by the PID controller, only constraints need to be applied to the second type of condition: the compressor outlet temperature should be below 210 degrees Celsius; the condenser inlet temperature should be above the dew point; and the valve opening should be within 0-90 degrees.

[0035] S8 performs iterative calculations using the differential evolution algorithm to solve for the optimal COP corresponding to the operation parameters.

[0036] (1) Initialization

[0037] Each individual in the differential evolution algorithm is a three-dimensional vector representing a set of valve openings, described as follows: The initial population can be randomly generated from the feasible region.

[0038] X0 = X L +(X U -X L rand(1,NP) (6)

[0039] Among them, X L X is the lower bound of the individual values. U NP represents the upper bound of the individual's value and the population size.

[0040] (2) Variation

[0041] The mutation vector is generated by weighting two independent individuals onto another individual, laying the foundation for the offspring population. A mutated individual is described as follows:

[0042] V(i)=X(r1)+F·(X(r2)-X(r3)); i=1,2,...,NP; r1,r2,r3∈(1,NP),r1≠r2≠r3 (7)

[0043] Where F is the variation factor;

[0044] (3) Cross

[0045] To enhance population diversity, some mutated individuals are introduced into the population:

[0046]

[0047] Where CR is the crossover factor, d rand It is a random integer between 1 and 3, ensuring that at least one dimension of the variable comes from the mutation operation;

[0048] (4) Selection

[0049] After each iteration, the fitness values ​​of the individuals are calculated and compared to select the better individual for the next iteration. (2) to (4) are repeated until the preset maximum number of iterations is reached, ultimately obtaining the valve opening corresponding to the optimal COP:

[0050]

[0051] Where g is the number of iterations and f is the fitness function, i.e., COP.

[0052] Preferably, in step S2, the number of learning samples collected through simulation data is 4200 sets.

[0053] Furthermore, in S6, the method for selecting the values ​​of the mutation factor and crossover factor based on the quality of the algorithm's convergence results is as follows: the algorithm undergoes ten independent random initialization experiments. If the algorithm converges within the maximum number of iterations in each experiment, and the deviation between the maximum and minimum values ​​of COP in all convergence results is less than 5%, the values ​​of the mutation factor and crossover factor are considered appropriate. Otherwise, the mutation factor and crossover factor are scaled, and ten independent random initialization experiments are repeated until the convergence results meet the evaluation criteria.

[0054] Preferably, in step S6, the differential evolution algorithm uses a population size of 50 and a maximum number of iterations of 200. The mutation factor is 0.7, and the crossover factor is 0.5.

[0055] Advantages and significant effects of the present invention:

[0056] (1) The Electric Environmental Control System (EECS) is a complex five-input four-output system. This energy efficiency optimization control strategy transforms it into a two-input two-output system by introducing optimization variables. Compared with existing technologies, this method can effectively improve the system's COP while reducing the complexity of the control system.

[0057] (2) Taking into account both internal and external factors of EECS, using generalized regression neural networks to predict the steady-state COP of the system can significantly improve the prediction accuracy and is applicable to a variety of complex working conditions.

[0058] (3) The differential evolution algorithm is used to optimize the predicted value of the generalized regression neural network output. Since the generalized regression neural network is trained offline and is only a mathematical model, the solution speed of this method is much faster than the online optimization algorithm, which is more in line with the actual application in engineering.

[0059] (4) The algorithm of the present invention has high implementability and reliability. Simulation results show that, compared with the traditional strategy, the present invention can effectively improve the energy efficiency of EECS. Attached Figure Description

[0060] Figure 1 This is a schematic diagram of an electric environmental control system (EECS).

[0061] Figure 2 This is a model diagram of the Electric Environmental Control System (EECS).

[0062] Figure 3 This is a schematic diagram of a neural network prediction model;

[0063] Figure 4(a) shows the fitting accuracy of the prediction model on the training set;

[0064] Figure 4(b) shows the fitting accuracy of the prediction model on the test set;

[0065] Figure 5 Simulation results for different control strategies within a given envelope. Detailed Implementation

[0066] The technical solution of the invention will be described in detail below with reference to the accompanying drawings:

[0067] The working principle of the electric environmental control system (EECS) is as follows: Figure 1As shown in the diagram. M is the electric motor, C is the compressor, F is the fan, T1 is the first-stage turbine, T2 is the second-stage turbine, PHX is the primary heat exchanger, SHX is the secondary heat exchanger, RHX is the regenerator, CON is the condenser, and WS is the water separator. When an aircraft is flying at high altitudes, the ambient temperature and pressure are very low, making it impossible to directly enter the cabin. First, the ram air needs to be pressurized by an electric compressor. The pressurized gas has increased temperature and pressure, and then it is cooled by the primary radiator. To further reduce the temperature, the gas then enters the air recirculation system for further regulation: first, it is pressurized by the compressor, and then it enters the secondary heat exchanger for further cooling. The cooled air is then cooled and dehumidified by the condenser, and then sequentially enters the first and second-stage turbine coolers for expansion and cooling. Finally, the cooled air is mixed with a portion of the cabin's recirculated air and supplied to the cabin to meet the cabin's cooling and fresh air requirements. The system's operating status is regulated by the electric compressor, electric fan, and multiple sets of control valves. The electric compressor directly affects the intake air condition; adjusting the compressor speed controls the system's air supply. The electric fan and cold air duct valve regulate the cold airflow. As the system's only heat sink, the cold air directly affects the efficiency of the primary and secondary radiators; therefore, adjusting the flow rate indirectly controls the supply air temperature. The turbine bypass valve mixes a hot stream with the cooling air at the second-stage turbine outlet, directly regulating the supply air temperature. The cryogenic limiting valve allows some airflow to bypass the first-stage turbine, limiting the condenser inlet temperature above the dew point and preventing icing and blockage. The economy cooling valve typically opens during high-altitude cruise conditions with very low ambient temperature and humidity. In these conditions, condensation and dehydration of the air are unnecessary, and cryogenic airflow is obtained solely through single-stage expansion of the second-stage turbine. Opening the economy cooling valve switches the system to single-stage expansion, reducing system energy consumption. In addition, the compressor outlet temperature needs to be controlled to prevent compressor overheating and malfunction. However, under the above one-to-one control strategy, the compressor outlet temperature can only be adjusted by all actuators together, which not only increases the complexity of control, but also fails to take into account the system energy efficiency.

[0068] Based on the working principle of EECS (Energy Efficiency Control System), this invention proposes an energy efficiency optimization control strategy for aircraft EECS. This strategy aims to improve EECS energy efficiency by selecting the opening degrees of bypass valves, economic cooling valves, and low-limit valves as optimization variables. Furthermore, the opening degree of the cold air duct valve is used as a control variable to meet the air supply temperature requirement, and the electric compressor speed is used as a control variable to meet the air supply volume requirement. The control of compressor outlet temperature and condenser inlet temperature is considered only as constraints in the optimization process.

[0069] The aforementioned energy efficiency optimization control strategy for the aircraft's electric environmental control system (EECS) includes the following steps:

[0070] Step 1: Design the system's energy efficiency optimization and control strategy

[0071] Based on meeting the performance requirements of cabin cooling and air supply, the system's COP is maximized through improved control strategies to reduce energy consumption. According to the component configuration, control functions, and architecture requirements analysis of the electric environmental control system, its optimization variables, control objects, and control quantities are designed. The main function of the electric environmental control system is to provide the cabin with the fresh air necessary to maintain normal passenger life activities, while balancing the cabin's thermal load to control the cabin temperature within a comfortable range. The system's control objects are the air supply volume that meets system performance requirements and the air supply temperature calculated from the cooling and air supply volumes. Available actuators include an electric compressor, a cold air duct valve, a bypass valve, an economic cooling valve, and a low-limit valve. Three of these valves can adjust the temperature of their respective nodes, ultimately affecting the air supply temperature. The cold air duct valve changes the cold air flow rate, while the other valves change the hot air flow rate. To maximize the system COP, the opening degrees of the bypass valve, economic cooling valve, and low-limit valve are selected as optimization variables. Based on this, the opening degree of the cold air duct valve is used as a control variable to meet the air supply temperature requirement, and the speed of the electric compressor is used as a control variable to meet the air supply volume requirement.

[0072] Step 2: Obtain training samples for the COP prediction model

[0073] The steady-state COP of a system depends on both external and internal system conditions. External conditions include the flight environment and cabin thermal loads, while internal conditions refer to the system's operating state, determined by optimizing valve openings. A generalized regression neural network is used to establish the mapping relationship between COP and these parameters and to predict the steady-state COP under different parameters. This invention collected 4200 sets of training samples from a large amount of simulation data. The system simulation model is as follows: Figure 2 As shown, the specific method is as follows: set the cruise altitude, flight Mach number, and cabin thermal load (environmental parameters) according to the gradient. Under a set of environmental parameters, set the bypass valve opening, economic cooling valve opening, and low limit valve opening (operating parameters) according to the gradient. Use two PID controllers to control the electric compressor speed and the cold air duct valve opening to meet the cabin air supply and cooling requirements. Record the electric compressor outlet temperature and cabin air supply temperature under steady state as learning samples.

[0074] Step 3: Construct the topology of the generalized regression neural network

[0075] The topology of the generalized regression neural network established in this invention is as follows: Figure 3As shown. The input layer has six neurons, corresponding to the current cruise altitude, flight Mach number, cockpit thermal load, bypass valve opening, economy cooling valve opening, and minimum valve opening, respectively. The input layer neurons directly pass the input variables to the pattern layer. The number of neurons in the pattern layer is equal to the number of learning samples, and the output of a single neuron can be described as...

[0076]

[0077] Where X = [X1, X2, ..., X6] are the six input vectors of the network, X i σ is the learning sample corresponding to the i-th neuron, and σ is the neuron width.

[0078] The summation layer uses two types of neurons for summation calculations, with the total number being the output vector dimension plus one; the first type directly adds all the outputs of the pattern layer, and its output S D It can be described as

[0079]

[0080] The second method involves weighted summation of the outputs of the mode layer, resulting in the output S. N It can be described as

[0081]

[0082] Among them, y ij It is the j-th element in the i-th output sample.

[0083] The output layer has two neurons, each corresponding to the outlet temperature T of the electric compressor. ec and cabin air supply temperature T cab The output layer divides the output of the summation layer, that is:

[0084]

[0085] Assuming the cabin air recirculation flow rate is half the supply air volume, and taking the temperature rise of the recirculated air after passing through the fan as 4°C, the COP of the EECS is obtained by the following formula:

[0086]

[0087] Where Q is the cabin cooling capacity, W is the system power consumption of EECS, and C p T is the specific heat capacity of air at constant pressure. sa For gas supply temperature, T amb For ambient temperature, This refers to the gas supply flow rate.

[0088] Step 4: Train the generalized regression neural network model

[0089] The 4200 samples were divided into two groups. One group was used as the training set to train the generalized regression neural network, and the other group was used as the test set to verify the accuracy of the generalized regression neural network in predicting the COP of the system under different operating conditions.

[0090] Step 5: Define the individual and fitness functions for the differential evolution algorithm.

[0091] In the differential evolution algorithm, each individual represents a set of valve openings. After each iteration, the individual value is substituted into the prediction model to obtain the corresponding prediction value, which is then converted to obtain the COP value as the fitness value of the differential evolution algorithm.

[0092] Step 6: Set the parameters for the differential evolution algorithm;

[0093] Based on design experience, a population size of 30-100 and a maximum number of iterations of 100-200 are generally preferred, as differential evolution algorithms tend to converge more easily. This scheme uses a population size of 50 and a maximum number of iterations of 200. The selection of the mutation factor and crossover factor depends on experimental experience; this scheme, after multiple adjustments, ultimately selected a mutation factor of 0.7 and a crossover factor of 0.5.

[0094] Step 7: Apply constraints to the algorithm based on system performance requirements and component operating characteristics.

[0095] The solution set obtained by the differential evolution algorithm needs to meet the following two conditions to be meaningful: 1) Each solution set needs to be generated under the same cooling capacity and gas supply conditions; 2) Each solution set needs to take into account the system operating state, the valve opening should be within 0-90 degrees, and the compressor outlet temperature and condenser inlet temperature should be within the normal operating range.

[0096] For the first type of condition, since the training samples of the generalized regression neural network have already been processed by the PID controller, the neural network has generalized this rule accordingly. Therefore, only constraints need to be applied to the second type of condition: the compressor outlet temperature should be below 210 degrees Celsius; the condenser inlet temperature should be above the dew point; and the valve opening should be within 0-90 degrees, i.e., the upper and lower bounds of the variables are 0-90.

[0097] Step 8: Perform iterative calculations using the differential evolution algorithm to solve for the optimal COP corresponding to the operation parameters.

[0098] (1) Initialization

[0099] Each individual in the differential evolution algorithm is a three-dimensional vector representing a set of valve openings, which can be described as: The initial population can be randomly generated from the feasible region.

[0100] X0 = X L +(XU -X L rand(1,NP) (6)

[0101] Among them, X L X is the lower bound of the individual values. U NP represents the upper bound of the individual's value and the population size.

[0102] (2) Variation

[0103] The mutation vector is generated by weighting two independent individuals onto another individual, laying the foundation for the offspring population. A mutated individual can be described as:

[0104] V(i)=X(r1)+F·(X(r2)-X(r3)); i=1,2,...,NP; r1,r2,r3∈(1,NP),r1≠r2≠r3 (7)

[0105] Where F is the variation factor.

[0106] (3) Cross

[0107] To enhance population diversity, some mutated individuals are introduced into the population.

[0108]

[0109] Here, CR is the crossover factor, which determines the probability of the crossover process occurring. rand It is a random integer between 1 and 3, ensuring that at least one dimension of the variable comes from the mutation operation.

[0110] (4) Selection

[0111] After each iteration, the fitness values ​​of the individuals are calculated and compared, and the better individuals are selected for the next iteration. Repeat (2) to (4) until the preset maximum number of iterations is reached, and finally the valve opening corresponding to the optimal COP can be obtained.

[0112]

[0113] Where g is the number of iterations and f is the fitness function, i.e., COP.

[0114] The following provides a more detailed description of an intelligent control method for an aircraft electric environmental control system based on simulation examples:

[0115] This example uses a classic operating condition of 40kW heat load and 6-10km cruise. Figures 4(a) and 4(b) show the fitting accuracy of the generalized regression neural network. Figure 4(a) shows the fitting result of the prediction model on the training set, and the corresponding R... 2The value is 0.9877. Figure 4(b) shows the fitting result of the prediction model on the test set, and the corresponding R value is 0.9877. 2 The value of 0.9910 demonstrates that a well-trained generalized regression neural network has sufficiently high prediction accuracy for the steady-state COP of the system.

[0116] Figure 5 The steady-state COP of different control strategies was compared under the same control objective, namely 40kW cooling capacity and 400g / s gas supply. These included the classic zone control strategy, the cruise economy cooling strategy, and the energy efficiency optimization control strategy used in this invention. It can be seen that the energy efficiency optimization control strategy resulted in the highest system COP across all test conditions.

[0117] The various embodiments of the present invention have been described above. These descriptions are merely exemplary and not limited to the disclosed embodiments. Furthermore, the energy efficiency optimization control strategy proposed in this invention is applicable not only to electric environmental control systems but also to the optimization control problems of complex systems with multiple control objectives and multiple actuators that have similar properties. Without departing from the scope and spirit of the described embodiments, those skilled in the art can apply it simply by making corresponding parameter modifications.

Claims

1. A control method for optimizing the energy efficiency of an aircraft electric environmental control system (EECS), wherein the EECS uses an additional electric motor-driven compressor to compress ambient air as an air source, directly introducing air from the ambient atmosphere. The electric motor drives the compressor to compress ram air, which is then cooled by a primary radiator and pressurized by the compressor. It then enters a secondary heat exchanger for further cooling. The cooled air is then cooled and dehumidified by a condenser and sequentially enters a first-stage and a second-stage turbine cooler for further cooling. Finally, the cooled air is supplied to the cabin to balance the thermal load. Its features are: The traditional regional control strategy is transformed into a combination of three-variable optimization and dual-objective control. Using the coefficient of performance (COP) of the energy efficiency control system (EECS) as the indicator, the COP is maximized by solving for the optimal solution set of the optimization variables under different operating conditions. Based on this, the controller calculates the values ​​of two control variables to meet the system's performance requirements. First, a generalized regression neural network model is established and trained using a large amount of offline data under various operating conditions to predict the steady-state COP of the EECS under different conditions. Then, a differential evolution algorithm searches for the optimal set of optimization variables corresponding to the optimal COP within the solution space that satisfies component performance constraints. Finally, the controller solves for the corresponding control variables at the optimal operating point, thereby achieving energy efficiency optimization control. Includes the following steps: S1, Design the control scheme for EECS and build the EECS model. Based on meeting the cabin's cooling and air supply requirements, and according to the component configuration, control functions, and architecture requirements of the EECS, its optimization variables, control objects, and control quantities are designed. The control objects of the EECS are the air supply quantity that meets the EECS performance requirements and the air supply temperature calculated from the cooling and air supply quantities. The actuators include an electric compressor, a cold air duct valve, a bypass valve, an economic cooling valve, and a low-limit valve. Among them, the bypass valve, economic cooling valve, and low-limit valve all change the heat flow rate and can adjust the temperature of the corresponding node, ultimately affecting the air supply temperature. The cold air duct valve changes the cold flow rate. To maximize the COP of the EECS, the opening degree of the bypass valve, economic cooling valve, and low-limit valve is selected as the optimization variable. Based on this, the opening degree of the cold air duct valve is used as the control quantity to meet the air supply temperature requirements, and the electric compressor speed is used as the control quantity to meet the air supply quantity requirements. S2, Obtain training samples for the COP prediction model. The steady-state COP of the EECS under different environmental and operational parameters is predicted by a generalized regression neural network (GRNN). The input of the GRNN includes three environmental parameters: cruise altitude, flight Mach number, and cabin thermal load, and three operational parameters: bypass valve opening, economic cooling valve opening, and lower limit valve opening. The output of the GRNN is the outlet temperature of the electric compressor and the cabin air supply temperature. A large number of simulation data are collected to form learning samples, with no less than 2400 sets of samples. Specifically, the three environmental parameters of cruise altitude, flight Mach number, and cabin thermal load are set according to a gradient. Under one set of environmental parameters, the three operational parameters of bypass valve opening, economic cooling valve opening, and lower limit valve opening are set according to a gradient. Two PID controllers are used to control the electric compressor speed and the cold air duct valve opening to meet the cabin air supply and cooling requirements, respectively. The outlet temperature of the electric compressor and the cabin air supply temperature under steady state are recorded as learning samples. S3, Construct the topology of the generalized regressive neural network GRNN. GRNN is a four-layer network, including an input layer, a pattern layer, a summation layer, and an output layer. The input layer has six neurons, corresponding to the current cruise altitude, flight Mach number, cabin thermal load, bypass valve opening, economy cooling valve opening, and lower limit valve opening, respectively. The input layer neurons directly pass the input variables to the pattern layer. The number of neurons in the pattern layer is equal to the number of training samples. The output of a single neuron is: (1) in, These are the six input vectors of the network. It is the first The learning samples corresponding to each neuron The width of the neuron; The summation layer uses two types of neurons for summation calculations, with the total number being the output vector dimension plus one; the first type directly adds all the outputs of the pattern layer, and its output... Described as (2) The second method involves weighted summation of the outputs of the pattern layer, and its output... It can be described as (3) in, It is the first The th output sample One element; The output layer has two neurons, each corresponding to the outlet temperature of the electric compressor. and cabin air supply temperature The output layer divides the output of the summation layer, that is: (4) Assuming the cabin air recirculation flow rate is half the supply air volume, and taking the temperature rise of the recirculated air after passing through the fan as 4°C, the COP of the EECS is obtained by the following formula: (5) in, For cabin cooling capacity, The power consumption of the EECS system The specific heat capacity of air at constant pressure. For gas supply temperature, For ambient temperature, This refers to the gas supply flow rate; S4, train the generalized regression neural network GRNN model; The learning samples obtained by S2 are divided into two groups on average. One group is used as the training set to train the generalized regression neural network, and the other group is used as the test set to verify the accuracy of GRNN in predicting COP of systems under different working conditions. S5, Define the individual and fitness functions for the differential evolution algorithm; In the differential evolution algorithm, each individual represents a set of valve openings. The individual value after each iteration is substituted into the prediction model to obtain the corresponding prediction value, which is then converted to obtain the COP value as the fitness value of the differential evolution algorithm. S6, Set the parameters of the differential evolution algorithm The population size of the differential evolution algorithm is 30-100, the maximum number of iterations is 100-200, and the values ​​of the mutation factor and crossover factor are selected according to the quality of the algorithm's convergence results. S7, impose constraints on the differential evolution algorithm based on EECS performance requirements and component operating characteristics; The solution set obtained by the differential evolution algorithm needs to meet the following conditions: First type of condition: each set of solutions needs to be generated under the same cooling capacity and gas supply conditions; 2) each set of solutions should take into account the EECS operating status, the valve opening should be within 0-90 degrees, and the compressor outlet temperature and condenser inlet temperature should be within the normal operating range. For the first type of condition, since the GRNN's learning samples have already been processed by the PID controller, only constraints need to be applied to the second type of condition: the compressor outlet temperature should be below 210 degrees Celsius; the condenser inlet temperature should be above the dew point; and the valve opening should be within 0-90 degrees. S8 performs iterative calculations using the differential evolution algorithm to solve for the optimal COP corresponding to the operation parameters. (1) Initialization Each individual in the differential evolution algorithm is a three-dimensional vector representing a set of valve openings, described as follows: The initial population can be randomly generated from the feasible region. (6) in, This is the lower bound of the individual value. The upper bound of the individual value. Population size; (2) Variation By generating a mutation vector by weighting two independent individuals onto another individual, the foundation for the offspring population is laid. A mutated individual is described as follows: (7) in, It is a variable factor; (3) Cross To enhance population diversity, some mutated individuals are introduced into the population: (8) in, Cross factor It is a random integer between 1 and 3, ensuring that at least one dimension of the variable comes from the mutation operation; (4) Choose After each iteration, the fitness values ​​of the individuals are calculated and compared, and the better individuals are selected for the next iteration. This process (2) to (4) is repeated until the preset maximum number of iterations is reached, and finally the valve opening corresponding to the optimal COP is obtained: (9) in, For the number of iterations, The fitness function is COP.

2. The energy efficiency optimization control method for an aircraft electric environmental control system according to claim 1, characterized in that: In S2, the number of learning samples collected through simulation data is 4200 sets.

3. The energy efficiency optimization control method for an aircraft electric environmental control system according to claim 1 or 2, characterized in that: In step S6, the method for selecting the values ​​of the mutation factor and crossover factor based on the quality of the algorithm's convergence results is as follows: the algorithm undergoes ten independent random initialization experiments. If the algorithm converges within the maximum number of iterations in each experiment, and the deviation between the maximum and minimum values ​​of COP in all convergence results is less than 5%, the values ​​of the mutation factor and crossover factor are considered appropriate. Otherwise, the mutation factor and crossover factor are scaled, and ten independent random initialization experiments are repeated until the convergence results meet the evaluation criteria.

4. The energy efficiency optimization control method for an aircraft electric environmental control system according to claim 3, characterized in that: In S6, the differential evolution algorithm uses a population size of 50, a maximum number of iterations of 200, a mutation factor of 0.7, and a crossover factor of 0.5.

Images