Heat energy gradient utilization method integrating working condition estimation and fractional derivative

Abstract

The invention discloses a heat energy gradient utilization method integrating working condition estimation and fractional derivative, which comprises the following steps: firstly, establishing a waste heat estimation model for a dynamic heat source, and calculating expected heat energy data of a future time window; secondly, the expected data are input into a fractional order fuzzy control model, the model processes the time sequence characteristics of the data by applying fractional order derivatives, and an optimal regulation and control instruction is solved in combination with a fuzzy rule base generated offline; and finally, a flux control valve group is controlled through the instruction, heat source gas is sequentially distributed into the high-temperature reaction module, the medium-temperature reaction module and the low-temperature reaction module, and gradient recovery of heat energy is achieved. According to the method, passive response is replaced by active pre-judgment, the problem of control lag when the dynamic heat source working condition changes drastically is effectively solved by fusing the prediction model and fractional order control, and the heat recovery efficiency and the system robustness are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of engine waste heat recovery, in particular to a heat energy gradient utilization method combining working condition estimation and fractional derivative. BACKGROUND

[0002] During operation, an aero-engine consumes a large amount of chemical fuel to generate thrust, but a considerable portion of the energy is directly discharged into the atmosphere in the form of high-temperature exhaust gas, causing huge energy waste and unnecessary thermal pollution. How to efficiently recover and reuse this part of high-temperature waste heat has become one of the key technical directions for improving the overall energy efficiency of aircraft, reducing operating costs, and reducing carbon footprint.

[0003] Among the many heat energy recovery schemes, thermo-chemical heat storage technology has attracted much attention due to its high energy density, long-term stable storage, and flexibility in releasing energy on demand. This technology converts the heat energy in high-temperature exhaust gas into chemical energy for storage through reversible chemical reactions.

[0004] However, the application of thermo-chemical heat storage technology to aero-engines faces enormous challenges. First, the working condition of an aero-engine is highly dynamic. During different stages of flight, such as takeoff, climb, cruise, and descent, the exhaust gas discharged by the engine will experience dramatic and rapid fluctuations in temperature, pressure, flow rate, and composition.

[0005] Second, to maximize energy recovery efficiency, a single thermo-chemical reaction often cannot cover the wide temperature range of the engine exhaust gas. Therefore, an efficient heat recovery system may need to integrate multiple reaction units with different chemical systems for different temperature ranges (e.g., high, medium, and low).

[0006] This poses a highly complex control problem: the system must be able to distribute the dynamically fluctuating "heat source" (high-temperature exhaust gas) to multiple parallel or series reaction units in real time and accurately, while ensuring that each reaction unit can maintain near its optimal reaction conditions.

[0007] Existing conventional control strategies, such as fixed-parameter-based controllers or simple logic threshold controls, are difficult to cope with such complex working conditions with multiple variables, strong coupling, and fast time variation. These methods often have a lag in response and cannot adapt to rapid switching of engine working conditions in time, resulting in low heat energy recovery efficiency; or they can only achieve good control at a single, stable operating point (such as cruise state), but perform poorly at other flight stages, and may even cause the failure of heat storage materials or damage to the system due to improper control (such as temperature exceeding limits or uneven flow distribution).

[0008] Therefore, there is an urgent need in the field for a new control method that adapts to the continuous dynamic changes of the aero-engine throughout the flight mission profile, improving the overall efficiency and system stability of waste heat recovery.

[0009] To this end, a kind of heat energy cascade utilization method is proposed, which combines working condition estimation and fractional derivative. SUMMARY

[0010] The purpose of the present application is to provide a kind of heat energy cascade utilization method that combines working condition estimation and fractional derivative, which replaces "passive response" with "active prediction", effectively overcomes the control lag problem when the working condition of dynamic heat source changes dramatically, and significantly improves the efficiency and system robustness of heat recovery. It includes: first, a waste heat estimation model is established for dynamic heat source, to calculate the expected heat energy data in the future time window;Then, input the expected data into the fractional fuzzy control model, the model applies fractional derivative to process the time series characteristics of data, and combines the offline generated fuzzy rule base to solve the optimal control instruction;Finally, control a group of flux control valves through the instruction, and distribute the heat source gas to the high-temperature, medium-temperature and low-temperature reaction modules in turn to realize the cascade recovery of heat energy.

[0011] To achieve the above purpose, the present application provides the following technical solutions: A kind of heat energy cascade utilization method that combines working condition estimation and fractional derivative, comprising: Monitoring the specified aircraft performing flight tasks, collecting engine dynamic operating parameters of the aircraft, and calling flight state data from the flight control system; Comprehensive flight task, flight state data and engine dynamic operating parameters, calculate the expected heat energy data; The expected heat energy data is used for pre-control of the thermochemical conversion and storage device, which is mounted on the specified aircraft and communicates with the combustion chamber, and is used to process high-temperature gas discharged by the exhaust system;The gas guide pipe is provided inside the thermochemical conversion and storage device, one end of the gas guide pipe is connected with the high-temperature exhaust port, and the other end is connected with the inlet of the first reaction module;The flow paths of the first reaction module, the second reaction module and the third reaction module are connected in series;The flux control valve group is used to control the gas flowing into each reaction module; The expected heat energy data is used as an input variable to input into a fractional fuzzy control model, the fuzzy control model uses historical thermochemical reaction experimental data to solve the optimal control instruction, and adjusts the opening and closing parameters of the flux control valve group using the optimal control instruction.

[0012] Preferably, the flight state data is divided into: The parameters characterizing the operating conditions of the aircraft include flight altitude, flight speed, climb rate and body attitude angle; the parameters characterizing the external environmental conditions include ambient atmospheric temperature, atmospheric pressure and wind speed. The engine dynamic operating parameters include engine fan speed, core engine speed, fuel flow and turbine inlet temperature.

[0013] Preferably, the step of calculating the expected thermal energy data specifically includes: According to the flight mission and the parameters characterizing the operating conditions of the aircraft, the current flight mission phase is identified; According to the flight mission phase, a corresponding residual heat estimation model is selected, and the engine dynamic operating parameters and the parameters characterizing the external environmental conditions are input into the model to obtain the expected thermal energy data; The residual heat estimation model is trained using historical data matching the flight mission phase.

[0014] Preferably, the training process of the residual heat estimation model includes: Historical engine operating data, historical external environmental parameters and historical heat data are collected to form an original database; data cleaning, standardization and missing value interpolation are performed on the original database to generate a purified feature set; the feature set is classified based on historical flight mission phases, and independent data subsets are created for each phase; time window technology is applied to each data subset to extract input-output sample pairs for supervised learning; the sample pairs are divided into training data and test data; a gradient boosting decision tree algorithm is used to construct an initial regression model using the training data; the test data is used to verify and optimize the hyperparameters of the initial regression model, and the final residual heat estimation model is obtained.

[0015] Preferably, the thermochemical conversion and storage device is configured with three reaction modules according to the heat gradient: The first reaction module is used to carry out the reduction reaction of the metal as a high-temperature stage; the second reaction module is used to carry out the decomposition reaction of the metal hydroxide as a medium-temperature stage; the third reaction module is used to carry out the dehydration reaction of the inorganic salt as a low-temperature stage; and each reaction module is attached with a tail gas disposal unit to capture and safely dispose of the by-product gas generated in the reaction.

[0016] Preferably, the configuration of the flux control valve group includes: a first flux valve arranged between the gas flow guide main pipe and the inlet of the first reaction module; a second flux valve arranged between the outlet of the first reaction module and the inlet of the second reaction module; a third flux valve arranged between the outlet of the second reaction module and the inlet of the third reaction module; each flux valve is used to independently regulate the high-temperature gas flowing into the corresponding reaction module.

[0017] Preferably, the historical thermo-chemical reaction experiment data is obtained through simulation of the experiment process, including: building an experiment platform, the experiment platform is equipped with a second device which is identical in structure to the thermo-chemical conversion and storage device; controlling the second device to reproduce a plurality of operating conditions, and recording experiment data under different temperature, gas flow, pressure and component combinations, the experiment data including control instruction records, gas state parameters and thermal conversion efficiency.

[0018] Preferably, the step of building a fuzzy rule base of the fractional order fuzzy control model includes: defining the standardized gas state data obtained through the experiment data as input and the control instruction data as output; fuzzifying the input and output, defining membership functions and fuzzy sets; applying a K-center clustering algorithm to divide the experiment data into a plurality of data clusters, and generating initial fuzzy rules for each cluster, and collecting into a candidate rule base; taking thermal conversion rate data as an evaluation index, refining the candidate rule base, removing redundant and conflicting rules, and forming the final fuzzy rule base.

[0019] Preferably, the operation steps of the fractional order fuzzy control model in solving the optimal control instruction are: According to the heat absorption characteristics of each reaction module, the order of the fractional derivative is optimized and determined in the offline stage, and the Greenwald-Letnikov fractional derivative is applied to process the expected thermal energy data to obtain a fractional derivative value; the fractional derivative value is fuzzified to convert into a fuzzy membership value; the fuzzy membership value is submitted to a fuzzy reasoning engine, the fuzzy reasoning engine performs reasoning based on the fuzzy rule base to output fuzzy results; and the fuzzy results are de-fuzzified to restore the optimal control instruction.

[0020] Further, the fractional order adaptive fuzzy control model is also integrated with an online learning mechanism, which can fine-tune the fuzzy rule base according to the real-time collected thermal conversion rate data, so as to adapt to the dynamic changes of the engine operating state, and improve the robustness and adaptability of the thermal energy recovery system. This mechanism continuously optimizes the control parameters to ensure that the system can still maintain high-efficiency thermal energy recovery performance in complex flight environments.

[0021] The beneficial effects of the present application are: 1. By introducing a "predict first, then control" control mode, the application realizes the control upgrade from "passive response" to "active prediction". This method establishes a residual heat estimation model that can predict the future thermal state by analyzing flight tasks, flight state data and engine dynamic operating parameters. This mode enables the control system to learn about the upcoming thermal fluctuations in advance, thereby adjusting the flux control valve group of the thermal chemical conversion and storage device in advance. This overcomes the delay and overshoot problem of traditional control strategies when facing rapid switching of engine operating conditions (such as take-off and climb), significantly improving the dynamic response speed and control accuracy of the system.

[0022] 2. By using an advanced fractional order fuzzy control model, the application provides a control logic with high robustness and adaptability, effectively solving the complex control problem of multivariable and nonlinear residual heat system of aero-engine. The advantage of this model is that its fuzzy logic (rule base constructed based on historical experimental data and clustering algorithm) is good at handling and describing the uncertainty and fuzzy relationship in this complex system. At the same time, the application of fractional derivative (such as the Grunwald-Letnikov definition) introduces the "memory" feature to the controller, making it have stronger representation ability and flexibility to the dynamic characteristics of the system, thereby greatly improving the stability and robustness of the system under complex disturbances.

[0023] 3. By combining the above prediction ability and robust control logic, the application realizes the "full operating condition" collaborative optimization of multi-stage reaction modules, maximizing the total efficiency of residual heat recovery. The method in the background art cannot optimize multiple reaction units in different temperature ranges at the same time. This application can intelligently and real-time accurately distribute high-temperature gas between the first, second and third reaction modules (corresponding to high, medium and low temperature zones) according to the temperature gradient of the expected thermal energy data. This ensures that high-temperature thermal energy is efficiently utilized by high-temperature modules, while medium and low-temperature residual heat is also fully absorbed by subsequent modules, realizing the cascade utilization of waste gas thermal energy and collaborative optimization under full flight profile, thereby significantly improving the overall energy utilization efficiency of the aero-engine. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 A thermal energy cascade utilization method flowchart combining operating condition estimation and fractional derivative is provided for the embodiments of the application; Figure 2 A thermal chemical reactor device structure schematic diagram is provided for the embodiments of the application; Figure 3 A flowchart of the simulation test process is provided for the embodiments of the application. DETAILED DESCRIPTION

[0025] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be clearly and completely described below, obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work belong to the protection scope of the present application.

[0026] Embodiment one As Figure 1 The figure shows a process flow chart of a thermal energy cascade utilization method combining working condition estimation and fractional derivative according to an embodiment of the present application, as Figure 2 The figure shows a structural schematic diagram of a thermochemical reactor device according to an embodiment of the present application.

[0027] A thermal energy cascade utilization method combining working condition estimation and fractional derivative, comprising: Monitoring a specified aircraft performing a flight task, collecting engine dynamic operation parameters of the aircraft, and calling flight state data from a flight control system; Integrating the flight task, the flight state data and the engine dynamic operation parameters to calculate expected thermal energy data; The expected thermal energy data is used for pre-regulation of a thermochemical conversion and storage device, the thermochemical conversion and storage device is carried on the specified aircraft and is in communication with a combustion chamber, and is used for processing high-temperature gas discharged by an exhaust gas discharge system; the thermochemical conversion and storage device is internally provided with a gas flow guide main pipe, one end of the gas flow guide main pipe is connected with a high-temperature exhaust gas discharge port, and the other end is connected with an inlet of a first reaction module; flow paths of the first reaction module, a second reaction module and a third reaction module are sequentially connected in series; a flux control valve group is used for regulating gas flowing into each reaction module; The expected thermal energy data is input as an input variable into a fractional order fuzzy control model, the fuzzy control model uses historical thermochemical reaction experimental data to solve optimal regulation instructions, and the optimal regulation instructions are used to adjust opening and closing parameters of the flux control valve group.

[0028] Further, the flight state data is divided into: Parameters representing aircraft operation working conditions, including flight altitude, flight speed, climb rate and body attitude angle; parameters representing external environmental conditions, including external atmospheric temperature, atmospheric pressure and wind speed; The engine dynamic operation parameters include: engine fan speed, core engine speed, fuel flow and turbine inlet temperature.

[0029] Specifically, the method of the present application continuously monitors the aircraft and engine states during the entire process of the designated aircraft performing a flight task (e.g. a route flight including take-off, climb, cruise, descent and landing).

[0030] The system real-time retrieves detailed flight state data through the onboard flight control system (FCS) and the air data computer (ADC). In the present embodiment, these data are specifically divided into: Parameters representing the operating conditions of the aircraft: These mainly include the real-time flight altitude (e.g. 11000 meters in the cruise phase), flight speed (e.g. 0.82 Mach), climb rate (e.g. 0 m / s in the cruise phase) and body attitude angle (e.g. pitch angle and roll angle) measured by the inertial navigation system (INS) and the global positioning system (GPS).

[0031] Parameters representing the external environmental conditions: These mainly include the external atmospheric temperature (e.g. -55°C), atmospheric pressure (e.g. 22.6 kPa) and wind speed and direction data measured by the outboard sensors (such as the air data probe).

[0032] At the same time, the method real-time collects the engine dynamic operating parameters of the double-sided engines through the data bus of the full authority digital electronic control system (FADEC) of the engine. In the present embodiment, these parameters specifically include: engine fan speed (N1); core engine speed (N2); fuel flow (WF); turbine inlet temperature (T4.5 or calculated value EGT).

[0033] All the collected flight state data and engine dynamic operating parameters are summarized to form a high-dimensional real-time data stream, which will serve as the original input for the subsequent step (i.e. calculating the expected thermal energy data).

[0034] By specifically defining the flight state data as operating condition parameters such as flight altitude, speed, attitude angle, and the engine dynamic operating parameters as fan speed, fuel flow and turbine temperature, the present application provides accurate, complete and high-dimensional input features for the subsequent residual heat estimation model. This detailed data definition ensures that the prediction model can fully capture all the key factors affecting the exhaust heat energy, thereby significantly improving the accuracy and reliability of the expected thermal energy data and laying a solid data foundation for the subsequent precise control.

[0035] Further, the step of calculating the expected thermal energy data specifically includes: According to the flight task and the parameter representing the operating condition of the aircraft, a current flight task phase is identified; according to the flight task phase, a corresponding residual heat estimation model is selected, and the engine dynamic operating parameter and the parameter representing the external environmental condition are input into the model to obtain the expected heat energy data; the residual heat estimation model is trained by using historical data matched with the flight task phase.

[0036] Specifically, the system first runs a flight phase identification module, which is a rule-based decision tree in this embodiment. The decision tree identifies the current flight task phase according to the flight task (for example, the preset route point height) and the real-time parameter representing the operating condition of the aircraft.

[0037] For example: IF (flight height < 3000 meters AND climb rate > 5 meters / second) THEN phase = “take-off and climb phase”.

[0038] For example: IF (flight height > 10000 meters AND climb rate ≈ 0 meters / second AND flight speed > 0.75 Mach) THEN phase = “cruise phase”.

[0039] At the current time in this embodiment, the system identifies that the current phase is “cruise phase”, and the system maintains a model library (for example, a hash table with the enumeration value of the flight task phase as the key), and according to the identified “cruise phase”, the system dynamically loads and selects a residual heat estimation model (for example, model_cruise.h5) optimized for the “cruise phase” from the model library.

[0040] Subsequently, the system applies the real-time collected engine dynamic operating parameters (N1, N2, WF, T4.5) and the parameters representing the external environmental conditions (ambient air temperature, atmospheric pressure) to the same time window technology (such as extracting statistical features in the past 5 minutes) as the training phase, and constructs a feature vector consistent with the training data format as input to the selected residual heat estimation model.

[0041] The residual heat estimation model performs high-speed operation on the input feature vector to obtain the expected heat energy data, which is a time series of predicted heat energy states in a specified time window (for example, 10 minutes in the future, one point every 30 seconds) in the future, and the output data format is as follows: [(T_gas[t+1], P_gas[t+1], F_gas[t+1]), (T_gas[t+2], P_gas[t+2], F_gas[t+2]),...], which details the dynamic change trend of the gas temperature, gas pressure and gas flow of the high-temperature gas to be discharged by the exhaust gas emission system.

[0042] The step of calculating the expected thermal energy data further comprises: A confidence score of the expected thermal energy data is calculated in parallel, which is generated by comparing the current engine dynamic operating parameters and the flight state data with the data distribution of the historical data used to train the residual heat estimation model; the confidence score quantifies the prediction uncertainty of the residual heat estimation model under the current working condition, wherein a low score indicates that the current working condition is beyond the reliable prediction range of the model; the fractional order fuzzy control model also receives the confidence score as an input when calculating the optimal control instruction; when the confidence score is lower than a preset safety threshold, the model is forced to switch to a predefined conservative control mode, which prioritizes the operational stability of the thermal chemical conversion and storage device.

[0043] Specifically, to ensure that the residual heat estimation model of the present application can still guarantee the operational safety of the entire thermal chemical conversion and storage device when it faces extreme working conditions that have never been seen or rarely seen in its training data, a conservative control mechanism based on a confidence score is introduced in a preferred embodiment of the present application.

[0044] The core of this mechanism is to quantify the similarity between the "current working condition" and the "historical training data". During the offline training process of the residual heat estimation model (such as the gradient boosting decision tree), the system not only uses the purified feature set to train the prediction model, but also trains one or more "novelty point detection" models in parallel. In this embodiment, the system trains an Isolation Forest algorithm model for each independent data subset of a flight mission phase (such as the "cruise phase"). This model is specifically designed to learn the data distribution of "normal" working conditions in this phase.

[0045] During real-time flight of the aircraft, when the flight mission phase recognition module determines that it is currently in the "cruise phase", the system not only calls the residual heat estimation model to calculate the expected thermal energy data, but also inputs the current engine dynamic operating parameters and flight state data (such as the feature vector composed of flight altitude, N1, WF, etc.) into the corresponding "cruise phase" Isolation Forest model in parallel. The Isolation Forest model outputs an anomaly score, which is converted into a standardized confidence score (for example, between 0.0 and 1.0). A high score (for example, 0.95) indicates that the current working condition (for example, N1 = 85%, WF = 1.0) is highly consistent with the historical training data, and the output of the residual heat estimation model (i.e. expected thermal energy data) is highly reliable. Conversely, a low score (for example, 0.3) indicates that the current working condition (for example, the aircraft encounters strong vertical wind shear, causing N1 and WF to exhibit high-frequency and severe fluctuations that have never been seen in the training set) is an anomaly point, and the output of the residual heat estimation model is highly unreliable.

[0046] The confidence score (e.g. 0.3) is taken as an additional input variable of the fractional order fuzzy control model, in which a preset safety threshold (e.g. 0.5) is set. The fuzzy knowledge base contains rules specifically designed to handle low confidence, which have the highest priority, such as the rule IF (confidence score IS "high" OR confidence score IS "medium") THEN (control mode IS "regular optimal"), and the rule IF (confidence score IS "low") THEN (control mode IS "conservative").

[0047] When the confidence score (0.3) is lower than the safety threshold (0.5), the conservative rule is activated, and the specific measures are as follows: the controller ignores the optimal control command calculated from the untrusted expected thermal energy data (e.g. an erroneous command that requires a drastic action of the valve). Instead, the controller outputs a predefined conservative control command. The selection of this command depends on the severity of the confidence score: Strategy one (medium untrustworthy): if the confidence score is lower than the safety threshold (e.g. 0.5) but still higher than a critical threshold (e.g. 0.2), the system determines that the predicted data is unreliable, but the working condition is not yet out of control. At this time, the "command freezing" strategy is adopted, that is, the last safe opening of the flux control valve group before triggering the conservative mode is kept unchanged to avoid false adjustment.

[0048] Strategy two (highly untrustworthy): if the confidence score is lower than the critical threshold (e.g. 0.2), the system determines that the working condition has completely exceeded the understandable range. At this time, in order to ensure absolute safety, the "slowly returning to the middle" strategy is adopted, that is, regardless of the current opening, all flux control valve groups are adjusted to a preset safe "middle state" (e.g. the first flux valve 30%, the second flux valve 30%, and the third flux valve 10%) at an extremely slow change rate (e.g. 0.5% per second).

[0049] This conservative mode will continue until the real-time data acquisition working condition returns to normal, and the confidence score is higher than the safety threshold (0.5) again.

[0050] By identifying the flight task phase according to the flight working condition parameters in real time, and dynamically selecting a residual heat estimation model from the model library that matches it, the present application realizes an adaptive prediction strategy. This "expert model" method, in which each model performs its own function, avoids the performance compromise of a single model when dealing with take-off, cruising, and descending, which are completely different working conditions. It ensures that the system can call the optimal model for calculation at any flight stage, thereby greatly improving the prediction accuracy and robustness of the expected thermal energy data within the entire flight envelope.

[0051] Further, the training process of the residual heat estimation model comprises: The historical engine operation data, historical external environment parameters and historical heat data are collected to form an original database; data cleaning, standardization and missing value interpolation are performed on the original database to generate a purified feature set; the feature set is classified based on historical flight task stages, and an independent data subset is created for each stage; a time window technique is applied to each data subset to extract input-output sample pairs for supervised learning; the sample pairs are divided into training data and test data; a gradient boosting decision tree algorithm is used to construct an initial regression model using the training data; the test data is used to verify and optimize the hyperparameters of the initial regression model, and a final residual heat prediction model is obtained.

[0052] Specifically, to realize the function of calculating the expected heat energy data, the training of the residual heat prediction model must be completed in advance, which is trained using historical data matching the flight task stage. The training process (in this embodiment, taking the training of the "cruise phase" model as an example) is as follows: A large amount of (for example, 1000 flight cycles) historical flight data is collected to form an original database. These data include all parameters defined in real-time data collection (flight conditions, environmental parameters, engine parameters), as well as real historical heat data (i.e., gas temperature, gas pressure, gas flow) measured by a high-temperature sensor array installed at the high-temperature exhaust outlet of the engine.

[0053] Data cleaning (e.g., removing abnormal values outside the physical range or 3-sigma range of the sensor) is performed on the original database, and interpolation is used to handle missing values in the data (e.g., linear interpolation for short-term missing values). Subsequently, standardization processing (e.g., Z-score standardization) is performed on all numerical features to generate a purified feature set.

[0054] According to the historical parameters representing the operating conditions of the aircraft, the purified feature set is classified by flight task stage (takeoff, climb, cruise, etc.), and an independent data subset is created for each stage.

[0055] On the data subset of the "cruise phase", a time window technique is applied. Specifically, a sliding window method is used to flatten or extract statistical features (such as mean, variance) of the past 5 minutes of engine dynamic operation parameters and parameters representing external environmental conditions, and construct a high-dimensional feature vector as the input (X) in the input-output sample pair. The real historical heat data at a specific time point (e.g., 1st, 5th, 10th minute) within the next 10 minutes is used as the multi-output prediction label (Y) of the sample pair.

[0056] All input-output sample pairs of the "cruise phase" are divided into training data and test data in a ratio of 80% / 20% and a gradient boosting decision tree algorithm is adopted. In this embodiment, a multi-output initial regression model is constructed using the LightGBM (LGBM) framework, and the training data are used for training.

[0057] After the training is completed, the prediction accuracy of the initial regression model is verified using the test data. The key hyperparameters (e.g., the number of trees, the number of leaf nodes, and the learning rate) of the model are optimized through grid search, with the root mean square error being minimized as the target, to obtain a final residual heat estimation model. The above process is repeated for other flight phases to generate respective models, which are collectively stored in a model library for real-time calling for calculation of expected thermal energy data.

[0058] Further, an online calibration mechanism is also included, which comprises: At time T, the expected thermal energy data output by the residual heat estimation model is recorded as a predicted value; at time T+N, the actual thermal energy parameter in the corresponding engine dynamic operating parameter is collected as a measured value; the predicted value and the measured value are compared to generate a deviation signal; when the deviation signal is determined to be a systematic error within a preset time window, a calibration coefficient is automatically generated; and the calibration coefficient is applied to the input end of the fractional order fuzzy control model.

[0059] Specifically, to solve the problem that the thermal chemical conversion and storage device (e.g., the catalyst in the reaction module) naturally ages with the increase of flight hours, causing the historical thermal chemical reaction experimental data (measured in a brand-new state) to gradually lose accuracy, in a preferred embodiment of the present application, the fractional order fuzzy control model integrates an online calibration mechanism.

[0060] The mechanism works in real time during flight. At time T, the residual heat estimation model calculates and outputs the expected thermal energy data (predicted value) at time T+N (N is a specified time delay, for example, T+5 minutes), for example, the predicted turbine outlet temperature is 800°C. When the flight time reaches T+N minutes, the real-time data acquisition module obtains the "actual thermal energy parameter" (current measured value) in the engine dynamic operating parameters, for example, the measured turbine outlet temperature is 815°C. At this time, the online calibration mechanism captures the +15°C deviation signal. The mechanism does not immediately react to a single deviation signal to prevent false positives caused by transient disturbances (such as turbulence) in flight. Instead, the mechanism accumulates and counts the deviation signal within a preset time window (for example, the past 1 hour), for example, through a low-pass filter or calculating its moving average. When the moving average of the deviation signal consistently deviates from zero within the preset time window (for example, the average deviation of the past 1 hour is +14.8°C), the mechanism determines that it is a systematic error caused by model drift or hardware aging. Once the systematic error is determined, the online calibration mechanism automatically generates a calibration coefficient or offset (in this case, a temperature offset of +14.8°C).

[0061] In subsequent operations, before the residual heat estimation model outputs the expected thermal energy data (for example, the new predicted value is 790°C) and sends it to the input end of the fractional order fuzzy control model, the online calibration mechanism will first apply the offset (+14.8°C) to the data, so that the controller actually receives an input value of 804.8°C. In this way, the model realizes online self-calibration without relying on additional hardware, ensuring the accuracy of control instructions throughout the life cycle of the device.

[0062] By adopting a systematic model training process, including data cleaning, standardization processing, and applying time window technology to extract sample pairs, the invention ensures the construction quality of the residual heat estimation model, especially using the gradient boosting decision tree, a high-efficiency regression algorithm, which enables it to fully learn the complex nonlinear relationships in historical data. This refined offline training process ensures that the online deployed model has high prediction accuracy and good generalization ability, providing reliable future operating input for the controller.

[0063] Further, the thermo-chemical conversion and storage device is configured with three reaction modules according to a heat gradient: The first reaction module, as a high-temperature stage, is used to carry out the reduction reaction of metal; the second reaction module, as a medium-temperature stage, is used to carry out the decomposition reaction of metal hydroxide; the third reaction module, as a low-temperature stage, is used to carry out the dehydration reaction of inorganic salt; and each reaction module is attached with a tail gas disposal unit to capture and safely dispose of the by-product gas generated in the reaction.

[0064] In this embodiment, the thermochemical conversion and storage device for carrying out heat recovery is mounted on the specified aircraft, and its physical structure (see Figure 2 ) is connected to the tail gas discharge system downstream of the combustion chamber.

[0065] The device is internally provided with a gas flow guide main pipe, the inlet end of which is connected to the high-temperature tail gas discharge port. Downstream of the gas flow guide main pipe, a group of flux control valves is connected in series to three reaction units designed for different temperature ranges in order: the first reaction module, the second reaction module, and the third reaction module. The flux control valve group (for example, containing three independent high-speed pneumatic regulating valves) is responsible for accurately controlling the flow of high-temperature gas into each reaction module.

[0066] To realize the step-by-step utilization of high-temperature tail gas, the three reaction modules are respectively filled with different thermochemical energy storage materials, which are specifically defined as follows: The first reaction module (high-temperature zone): as a high-temperature reaction unit, it is internally filled with, for example, iron trioxide (Fe2O3). Its design goal is to carry out the reduction reaction of metal, which is carried out at high temperature (for example, >800°C) and can efficiently absorb the high-grade heat energy in the tail gas.

[0067] The second reaction module (medium-temperature zone): as a medium-temperature reaction unit, it is internally filled with, for example, calcium hydroxide (Ca(OH)2). Its design goal is to carry out the decomposition reaction of metal hydroxide, which is carried out at medium temperature (for example, 400°C-600°C) and is used to absorb the medium-temperature waste heat discharged by the first reaction module.

[0068] The third reaction module (low-temperature zone): as a low-temperature reaction unit, it is internally filled with, for example, magnesium sulfate heptahydrate (MgSO4·7H2O). Its design goal is to carry out the dehydration reaction of inorganic salt, which is carried out at a lower temperature (for example, <200°C) and is used to capture the low-grade heat energy after flowing through the second reaction module.

[0069] In addition, each reaction module is equipped with a gas product collection and processing system for safely and harmlessly collecting and processing the gas generated in the reaction process (for example, dehydration), preventing it from entering the aircraft interior.

[0070] By designing the thermo-chemical conversion and storage device as three functionally distinct reaction modules, i.e. units for performing high-temperature metal reduction, medium-temperature metal hydroxide decomposition and low-temperature inorganic salt dehydration respectively, the present application realizes a highly efficient thermal energy "cascade utilization" architecture. This design enables the high-grade thermal energy of high-temperature tail gas, medium-temperature waste heat and low-grade thermal energy to be captured and stored by the most suitable chemical reactions. It avoids the stringent temperature requirements of a single reactor, greatly broadens the working interval of thermal energy recovery, and significantly improves the overall conversion efficiency of tail gas thermal energy.

[0071] Further, the configuration of the flux control valve group includes: a first flux valve arranged between the gas flow guide manifold and the inlet of the first reaction module; a second flux valve arranged between the outlet of the first reaction module and the inlet of the second reaction module; a third flux valve arranged between the outlet of the second reaction module and the inlet of the third reaction module; each flux valve is used to independently regulate the high-temperature gas flowing into the corresponding reaction module.

[0072] Specifically, to realize precise control of the thermo-chemical conversion and storage device, the group of flux control valves in this embodiment is further specified, and the valve group is composed of three independent flux valves: The first flux valve: the valve is arranged between the gas flow guide manifold and the inlet of the first reaction module (high-temperature zone).

[0073] The second flux valve: the valve is arranged between the outlet of the first reaction module and the inlet of the second reaction module (medium-temperature zone).

[0074] The third flux valve: the valve is arranged between the outlet of the second reaction module and the inlet of the third reaction module (low-temperature zone).

[0075] In this embodiment, the three flux valves are high-speed response electric or pneumatic regulating valves, which can receive optimal control instructions calculated from a multi-stage fuzzy control model, and each flux valve is used to independently regulate the mass flow of high-temperature gas flowing into the corresponding reaction module, thereby ensuring that each module operates within its optimal reaction temperature interval.

[0076] By configuring a first flux valve, a second flux valve and a third flux valve for the three reaction modules respectively, and physically arranging them in sequence on the gas flow guide manifold, the present application provides precise and independent physical execution mechanisms for optimal control instructions. This "one-to-one" valve configuration enables the controller to flexibly and proportionally distribute the gas flow into each reaction module. It ensures that high-temperature, medium-temperature and low-temperature gases can be accurately directed to their respective reaction units, and is the key hardware foundation for realizing multi-module collaborative work and thermal energy cascade utilization.

[0077] As Figure 3 Fig. 1 shows a flowchart of a simulation process according to an embodiment of the present application.

[0078] Further, the historical thermo-chemical reaction experiment data is obtained through a simulation process, including: building a test platform, the test platform is equipped with a second device which is identical in structure to the thermo-chemical conversion and storage device; operating the second device to reproduce a plurality of operating conditions, and recording test data under different temperature, gas flow, pressure and composition combinations, the test data including control instruction records, gas state parameters and thermal conversion efficiency.

[0079] Specifically, before the deployment of the controller, historical thermo-chemical reaction experiment data and fuzzy knowledge base must be prepared for it, the historical thermo-chemical reaction experiment data is obtained through a simulation process, including: building a test platform, the test platform is equipped with a second device which is identical in structure to the thermo-chemical conversion and storage device; operating the second device to reproduce a plurality of operating conditions, and recording test data under different temperature, gas flow, pressure and composition combinations, the test data including control instruction records, gas state parameters and thermal conversion efficiency.

[0080] The plurality of operating conditions reproduced specifically include the following types of tests: Simulate engine exhaust conditions (input simulation): (1) Steady state condition: for example, simulate the cruise phase, continuously provide 850°C, 0.5 MPa stable high temperature gas; or simulate the medium temperature condition, continuously provide 600°C, 0.4 MPa gas.

[0081] (2) Transient condition: for example, simulate take-off climb, the gas temperature linearly climbs from 500°C to 800°C in 5 minutes; or simulate cruise descent, the temperature and pressure rapidly step down from 800°C to 400°C in 30 seconds.

[0082] Traverse valve control strategy (adjustment simulation): Under each of the above simulation conditions (for example, fixed at 850°C steady state condition), the test platform will actively traverse different flux control valve group opening and closing parameter combinations and record the results. For example: Test A (high temperature main first module): input 850°C gas, control instruction record set to {first flux valve: 100%, second flux valve: 0%, third flux valve: 0%}. The system records the total thermal conversion efficiency at this time (for example, 85%).

[0083] Test B (high temperature split flow): input is 850°C gas, control instruction record is set to {first flux valve: 80%, second flux valve: 20%, third flux valve: 0%}, the system records the total heat conversion efficiency at this time (for example, 88%).

[0084] Test C (medium temperature main attack second module): input is 600°C gas (working condition 2), control instruction record is set to {first flux valve: 10%, second flux valve: 90%, third flux valve: 0%}. The system records the heat conversion efficiency at this time (for example, 90%).

[0085] By performing hundreds of such permutation and combination tests, the present application obtains data on which valve combination can bring the highest heat conversion efficiency under any input working condition, which forms the basis for building a subsequent fuzzy knowledge base.

[0086] By building a test platform identical in structure to the actual thermochemical conversion and storage device and reproducing various steady-state and transient operating conditions on the platform, the present application solves a key "data acquisition" problem. This simulation test method enables safe and efficient collection of heat conversion efficiency data corresponding to different control instructions under various working conditions during the research and development phase. It provides essential and high-confidence historical thermochemical reaction experimental data for subsequent construction of a fuzzy knowledge base, avoiding expensive and dangerous real machine flight tests.

[0087] Further, the step of constructing a fuzzy rule base for the fractional order fuzzy control model comprises: The standardized gas state data obtained through the test data is defined as input, and the control instruction data is defined as output. The input and output are fuzzified, and the membership functions and fuzzy sets are defined. A K-center point clustering algorithm is applied to divide the test data into multiple data clusters, and initial fuzzy rules are generated for each cluster to form a candidate rule base. The heat conversion rate data is used as an evaluation index to refine the candidate rule base, eliminate redundant and conflicting rules, and form the final fuzzy rule base.

[0088] Specifically, the fuzzy rule base is the core of the controller for decision-making, and its construction process is disclosed in this embodiment as the following steps: Firstly, the normalized gas state data (such as gas temperature, gas pressure, gas flow) obtained from the test data is defined as input, and the regulation instruction data (such as the opening percentage of valves 1, 2, and 3) is defined as output; secondly, the continuous value range of the input and output is fuzzified, which is divided into several fuzzy sets such as {too low, low, appropriate, high, too high} or {closed, slightly open, small open, medium open, large open}, and a membership function (such as Gaussian function or triangular function) is defined for each set; then, a K-center clustering algorithm is applied to divide the historical thermochemical reaction test data (i.e. a large amount of test data points) into multiple (for example, 100) typical working condition clusters; subsequently, by analyzing the center points of each working condition cluster and fuzzifying them, an initial "IF-THEN" fuzzy rule is generated for each cluster (for example, the rule IF (gas temperature IS "high" AND gas pressure IS "medium") THEN (the first flux valve IS "large open" AND the second flux valve IS "slightly open" AND the third flux valve IS "closed")), and all these rules constitute a candidate rule base; finally, taking the heat conversion rate data in the historical thermochemical reaction test data as an evaluation index, the candidate rule base is tested and refined, rules that result in a heat conversion efficiency lower than a preset threshold are removed, and redundant and conflicting rules are merged, thereby forming a final fuzzy rule base that is simplified and efficient.

[0089] By adopting a fuzzy knowledge base construction method combined with a K-center clustering algorithm, the application successfully converts a large amount of historical thermochemical reaction test data into a set of simplified and efficient "IF-THEN" control rules. The method first fuzzifies the input and output variables, then automatically generates working condition clusters through a clustering algorithm, and then optimizes and filters the rules using heat conversion efficiency data. This systematic process solidifies the experience and test data of experts into machine-executable logic, significantly reducing the design complexity of the controller and ensuring the intelligence and optimality of its decision-making.

[0090] Further, the running steps of the fractional order fuzzy control model in solving the optimal regulation instruction are: According to the heat absorption characteristics of each reaction module, a fractional order derivative order is optimized and determined in the offline stage, and a fractional order derivative value is obtained by applying a Greenwald-Leontief fractional order derivative to process the expected heat energy data; the fractional order derivative value is fuzzified and converted into a fuzzy membership value; the fuzzy membership value is submitted to a fuzzy reasoning engine, which performs reasoning based on the fuzzy rule base to output a fuzzy result; and the fuzzy result is de-fuzzified to restore the optimal regulation instruction.

[0091] Specifically, during the flight of the aircraft, the controller performs real-time calculation in the online running stage: the controller receives the expected heat energy data time series generated by the residual heat estimation model at each control cycle (e.g., 100 ms).

[0092] Before the controller performs online calculation, the order of the fractional derivative (v) needs to be determined in the offline stage in advance. The order v is a key control hyperparameter, and its value directly affects the control robustness and response characteristics of the corresponding reaction module. The determination method of the order v is combined with the simulation test process (as claimed in claim 7) of the present application: on the test platform, for the specific reaction type (e.g., reduction reaction of metal) of each reaction module (e.g., the first reaction module), in its typical working temperature range (e.g., > 800°C), the fractional order v is taken as a hyperparameter to be optimized (e.g., grid search or iterative optimization in the range of 0.1 to 0.9), and the "thermal conversion efficiency" in the historical thermochemical reaction experimental data is taken as the evaluation index. Through repeated tests and optimization, an optimal fixed order v value is finally determined for the module, which can make the thermal conversion efficiency highest and the system response most stable. The same experimental optimization method is used to determine the optimal order of the second and third reaction modules. In online running, the controller calls these pre-determined optimal orders v, and applies the Grünwald-Leibniz fractional derivative definition to calculate the trend of the expected heat energy data time series. The advantage of this is that (compared to the traditional integer order derivative) it can take into account both the historical memory and the instantaneous change of the data, making the controller respond more smoothly and not easy to overshoot when facing the dramatic fluctuations of the expected heat energy data (such as flight stage switching), thereby enhancing the robustness of the system; then, the input value after fractional order processing is sent to the offline constructed final fuzzy knowledge base, and a fuzzy inference engine (Mamdani inference engine is used in this embodiment) is activated to match all applicable "IF-THEN" rules and calculate their trigger strengths; then, the fuzzy inference engine merges the fuzzy outputs of all triggered rules (e.g., "large open" and "medium open" are activated at the same time), and performs defuzzification calculation using the center of gravity method to restore the merged fuzzy set to an accurate numerical control signal (e.g., the opening and closing parameters of the first flux valve should be 82.5%). These accurate numerical signals constitute the optimal control instructions; finally, the optimal control instructions are sent to the actuators (e.g., electric motors or pneumatic actuators) of the flux control valve group of the thermochemical conversion and storage device, and the optimal control instructions are used to adjust the opening and closing parameters of the flux control valve group, so that the high-temperature gas flows into the three reaction modules in the optimal proportion, thereby continuously maintaining efficient heat energy recovery under dynamically changing flight conditions.

[0093] The fractional order fuzzy control model after calculating the optimal control instruction further comprises an instruction smoothing processing step; the instruction smoothing processing step receives the optimal control instruction as an original instruction; checks the change rate between the original instruction and the actual opening and closing parameters of the flux control valve group; when the change rate exceeds the preset mechanical and thermal shock safety threshold, the instruction smoothing processing step decomposes the original instruction into an instruction sequence that is smoothly transitioned over time; and the instruction sequence is gradually issued to the flux control valve group.

[0094] Specifically, to protect the flux control valve group (as a precision mechanical component) and the reaction module (as a heat-sensitive component) from the thermal and mechanical shocks caused by the instantaneous jump of the optimal control instruction when switching in the flight phase, the present application integrates a pure software instruction smoothing processing step between the output end of the fractional order fuzzy control model and the execution end of the physical valve. The core of this step is a change rate limiter. In the general running process, when the fractional order fuzzy control model calculates an optimal control instruction (as an original instruction) based on the expected thermal energy data, and the target value of the optimal control instruction is significantly different from the current valve state, the optimal control instruction (original instruction) is sent to the instruction smoothing processing step instead of being directly sent to the physical valve. The instruction smoothing processing step checks the change rate between the target value of the original instruction and the actual opening and closing parameters of the flux control valve group, and then compares this change rate with a preset mechanical or thermal shock safety threshold. The safety threshold is determined according to the mechanical response specification of the valve manufacturer or the thermal capacity experimental data of the reaction module, and in this embodiment, it is set to "the maximum change per second does not exceed 10%". When the calculated change rate (for example, instantaneous change of 75%) exceeds the safety threshold (for example, actual opening and closing parameters, 10% per second), the change rate limiter is activated, and the change rate limiter generates the instruction sequence that is smoothly transitioned over time through a linear interpolation algorithm. Specifically, it obtains the target value of the original instruction (for example, 10%) and (for example, 85%), and calculates a transition time based on the safety threshold (10% per second) (for example, |85%-10%| / 10% / second=7.5 seconds). In the next 7.5 seconds, the instruction smoothing processing step gradually issues a new instruction point in a fixed control clock period (for example, every 100 milliseconds) on the linear path from 85% to 10%. For example, at T+0.1 seconds, the instruction 85% -(10% / second x 0.1 seconds)=84% is issued; at T+0.2 seconds, the instruction 83% is issued, and so on, until at T+7.5 seconds, the instruction reaches the final target value 10% set by the original instruction. In this way, the action of the physical valve is smoothly extended to a controllable time period, the temperature and pressure of the reaction module are slowly changed, and thermal shock and mechanical damage are avoided.

[0095] By applying the Grünwald-Leibnik fractional derivative to process the expected thermal energy data during the online running phase, the present application enables the controller to not only respond to the current value, but also to understand the historical memory and future trend of the data. Compared with the traditional integer order derivative, it responds more smoothly to the dramatic fluctuations of the working condition (such as flight phase switching) and is less likely to overshoot, enhancing the robustness of the system. At the same time, combined with the fuzzy inference machine and the barycentric method to solve the ambiguity, it can quickly and accurately calculate the fuzzy rules into precise optimal control instructions, ensuring the real-time and stability of the control.

[0096] By combining the calculation of expected thermal energy data with the solution of fractional fuzzy control model, the present application constructs a complete closed loop from prediction to control. This method can perceive the upcoming exhaust heat energy changes in advance and use the intelligent model trained based on historical thermal chemical reaction experimental data to calculate the optimal control instructions in real time. This predictive control strategy ensures that the flux control valve group can be accurately adjusted, so that the system can maintain efficient and stable heat energy recovery even when the flight condition changes dramatically.

[0097] Example Two At time T0, the aircraft is at a stable cruising altitude of 11000 meters. The method of the present application continuously monitors its operating condition through the flight control system and obtains a stable flight speed of 0.82 Mach and a climb rate of 0 meters / second. At the same time, it learns from the engine control system that the fuel flow is stable at 1.0 kg / s. The flight mission phase recognition module of the present method confirms that the current task is "cruise phase" according to the height of 11000 meters and the near-zero climb rate.

[0098] The system immediately calls the special "cruise phase" waste heat estimation model from the model library. Based on stable engine fuel flow and other parameters, the model calculates that, within the next 10 minutes, the exhaust gas emission system will continuously discharge high-temperature gas with a stable fluctuation between 840°C and 850°C. This expected thermal energy data is immediately transmitted to the controller, and the fractional fuzzy control model receives this smooth high-temperature prediction sequence, with its Grünwald-Leibnik fractional derivative calculation value close to zero, indicating a "no change in working condition" trend to the fuzzy inference machine. The fuzzy inference machine combines the inputs of "high gas temperature IS" and "trend IS zero" to activate the rules corresponding to the cruise condition in the fuzzy knowledge base. After ambiguity resolution calculation, the controller generates the optimal control instructions and sends them to the actuator. The flux control valve group immediately responds by opening the first flux valve to 85%, slightly opening the second flux valve to 15%, and closing the third flux valve. This adjustment ensures that most of the high-grade heat energy is accurately directed into the first reaction module (high-temperature zone) for the most efficient metal reduction reaction, and the system enters a stable and efficient heat recovery state.

[0099] At T1, 20 minutes after T0, the pilot executes the descent command, reducing the engine thrust, the method of the invention immediately captures from the flight control system that the climb rate has become -10 m / s, the flight altitude starts to decrease, at the same time, the engine control system reports that the fuel flow has been significantly reduced to 0.4 kg / s, the flight mission phase recognition module captures the -10 m / s climb rate, immediately switches the system state from "cruise phase" to "descent phase", this switching triggers the automatic reloading of the estimation model, the system unloads the cruise model and loads the model weight for "descent phase". This new model quickly calculates a new expected heat energy data according to the sudden drop in fuel flow and other parameters, which indicates that the exhaust gas temperature will not be stable in the next 10 minutes, but will quickly cool from 600°C to 450°C. The controller receives this "medium temperature and rapid cooling" prediction sequence, and its Grünwald-Letnikov fractional derivative module immediately calculates a significant negative value, passing the important trend information of "working condition positive rapid cooling" to the fuzzy inference machine. The fuzzy inference machine combines the inputs of "gas temperature IS suitable" and "trend IS negative" to activate a completely different set of rules in the knowledge base, which are optimized for handling medium temperature and cooling conditions. The controller de-fuzzifies to generate a new set of optimal control instructions, and the flux control valve group responds to the new instructions by quickly reducing the opening of the first flux valve from 85% to 20%, while significantly opening the second flux valve from 15% to 75%, and slightly opening the third flux valve to 5%. This "anticipatory" adjustment action has moved the heat energy flow away from the first module that is not suitable for reaction before the high-temperature gas disappears, and instead directed most of the medium-temperature gas into the second reaction module for the most efficient metal hydroxide decomposition reaction, while the third module also begins to capture low-grade heat energy.

[0100] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and variations can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method of thermal energy cascade utilization fusing working condition estimation and fractional derivative, characterized in that, The application relates to a method for monitoring a specified aircraft performing a flight task, collecting engine dynamic operation parameters of the aircraft, and calling flight state data from a flight control system. The flight task, the flight state data and the engine dynamic operation parameters are integrated to calculate expected thermal energy data. The expected thermal energy data are used for pre-control of a thermal chemical conversion and storage device which is mounted on the specified aircraft and is communicated with a combustion chamber and is used for processing high-temperature gas discharged by a tail gas discharge system; the thermal chemical conversion and storage device is internally provided with a gas guide manifold, one end of the gas guide manifold is connected with a high-temperature tail gas discharge port, and the other end is connected with an inlet of a first reaction module; flow paths of the first reaction module, a second reaction module and a third reaction module are sequentially connected in series; and a flux control valve group is used for controlling gas flowing into each reaction module. The expected thermal energy data are input into a fractional order fuzzy control model as input variables, the fuzzy control model uses historical thermal chemical reaction experimental data to solve optimal control instructions, and the optimal control instructions are used for adjusting opening and closing parameters of the flux control valve group. The flight state data are divided into:

2. The method according to claim 1, wherein, Parameters representing aircraft operation conditions, including flight height, flight speed, climbing rate and body attitude angle; and parameters representing external environmental conditions, including external atmospheric temperature, atmospheric pressure and wind speed. The engine dynamic operation parameters include: engine fan rotating speed, core engine rotating speed, fuel flow and turbine inlet temperature. The step of calculating the expected thermal energy data specifically includes:

3. The method according to claim 1, wherein, According to the flight task and the parameters representing aircraft operation conditions, a current flight task stage is identified; According to the flight task stage, a corresponding waste heat estimation model is selected, and the engine dynamic operation parameters and the parameters representing external environmental conditions are input into the model to obtain the expected thermal energy data; The waste heat estimation model is trained by using historical data matched with the flight task stage. The training process of the waste heat estimation model includes:

4. The method according to claim 3, wherein, Historical engine operation data, historical external environmental parameters and historical heat data are collected to form an original database; data cleaning, standardization and missing value interpolation are performed on the original database to generate a purified feature set; the feature set is classified based on historical flight task stages, and independent data subsets are created for each stage; a time window technology is applied to each data subset to extract input-output sample pairs for supervised learning; the sample pairs are divided into training data and test data; a gradient boosting decision tree algorithm is adopted to construct an initial regression model by using the training data; the test data are used to verify and optimize the hyperparameters of the initial regression model, and a final waste heat estimation model is obtained. The thermal chemical conversion and storage device is provided with three reaction modules according to a heat gradient:

5. The method according to claim 1, wherein, The first reaction module is used for carrying a metal reduction reaction in a high-temperature stage; and the second reaction module is used for carrying a metal hydroxide decomposition reaction in a medium-temperature stage. ​ The third reaction module is used for carrying out dehydration reaction of inorganic salt as a low-temperature stage; and each reaction module is provided with a tail gas treatment unit for capturing and safely treating by-product gas generated in the reaction.

6. The method according to claim 1, wherein, The flux control valve group comprises a first flux valve arranged between the gas guide manifold and the inlet of the first reaction module, a second flux valve arranged between the outlet of the first reaction module and the inlet of the second reaction module, and a third flux valve arranged between the outlet of the second reaction module and the inlet of the third reaction module, each of which is used for independently regulating high-temperature gas flowing into the corresponding reaction module.

7. The method according to claim 1, wherein, The historical thermochemical reaction experimental data are obtained through simulation of the test process, including: A test platform is built, which is equipped with a second device identical in structure to the thermochemical conversion and storage device; The second device is controlled to reproduce various operating conditions, and test data under different temperature, gas flow, pressure and composition combinations are recorded, including regulation instruction record, gas state parameter and thermal conversion efficiency.

8. The method according to claim 7, wherein, The steps of constructing the fuzzy rule base of the fractional order fuzzy control model include: The standardized gas state data obtained through the test data are defined as input, and the regulation instruction data are defined as output; the input and output are fuzzified, and the membership function and fuzzy set are defined; a K-center clustering algorithm is applied to divide the test data into multiple data clusters, and initial fuzzy rules are generated for each cluster to form a candidate rule base; the thermal conversion rate data are used as evaluation indexes to refine the candidate rule base, remove redundant and conflicting rules, and form the final fuzzy rule base.

9. The method according to claim 8, wherein, The operation steps of the fractional order fuzzy control model in solving the optimal regulation instruction are as follows: According to the heat absorption characteristics of each reaction module, the order of fractional derivative is optimized and determined in the offline stage, and the Greenwald-Leontief fractional derivative is applied to process the expected heat energy data to obtain a fractional derivative value; the fractional derivative value is fuzzified to convert into a fuzzy membership value; the fuzzy membership value is submitted to a fuzzy reasoning engine, which performs reasoning based on the fuzzy rule base to output fuzzy results; The fuzzy results are de-fuzzied to restore the optimal regulation instruction.

Images