Aircraft agile protection guarantee and energy supply system

Abstract

The invention provides an airplane agile protection guarantee and energy supply system, which belongs to the technical field of airplane protection guarantee pressure regulation and energy supply, and comprises a protection air bag body, an energy supply unit, a pressure regulation unit and a core control unit of an integrated neural network prediction module, the core control unit is connected with the energy supply unit and the pressure regulation and control unit through communication modules, and the core control unit adopts an S7-1500 series PLC (carrying a neural network prediction module) to serve as a system center to coordinate energy distribution and pressure regulation and control. The pressure regulation and control unit is driven at night to maintain airbag pressure, and a generator energy supplement strategy is dynamically planned by means of the neural network prediction module. The dual purposes of pressure stability and energy efficiency are achieved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of agile protection and guarantee of aircraft and collaborative energy supply of multiple equipment, and particularly relates to an agile protection and guarantee system of aircraft and an energy supply system. BACKGROUND

[0002] The agile protection and guarantee of aircraft (realized by rapid forming relying on inflatable technology) is widely applied to scenes such as field operation due to advantages such as rapid forming, low cost and strong mobility, and its core relies on the internal pressure of the air bag to maintain the stability of the structure. The significant diurnal temperature difference in the field leads to thermal expansion and contraction of the air in the air bag: the temperature of the air bag rises during the day, and the internal pressure exceeds the safety threshold, so the air needs to be discharged to protect the air bag material; at night, the temperature drops, and the internal pressure drops, so the air needs to be inflated to ensure the carrying capacity. Therefore, the energy supply of the existing inflatable hangar has the following disadvantages:

[0003] 1. The existing energy supply system is designed unreasonably: the traditional method is to directly power the inflatable system (such as a 4 kW low-pressure fan and a 4.3 kW high-pressure fan) by a generator. Since the load needs to be started, the power of the generator is usually more than 3 times the power of the load (such as a 12 kW or more generator to drive a 4 kW fan), which leads to a large volume, heavy weight, high fuel consumption of the energy supply system, and it is difficult to meet the needs of mobile deployment in the field.

[0004] 2. Energy supply and demand do not match: the energy demand for diurnal pressure regulation is significantly different (discharge is the main during the day, and energy consumption is low; inflation is the main at night, and energy consumption is high), but the existing system mostly uses a generator to provide continuous power or a battery to be fully charged, and does not adjust dynamically in combination with the demand, resulting in low utilization of clean energy such as solar energy and serious waste of fuel.

[0005] Therefore, there is an urgent need for a comprehensive system that can integrate agile protection and efficient energy supply. SUMMARY

[0006] To solve the above technical problems, the application provides an agile protection and guarantee system of aircraft and an energy supply system, which realizes pressure stabilization of the inflatable guarantee cabin and efficient energy supply of multiple equipment.

[0007] To achieve the above purpose, the application adopts the following technical solutions:

[0008] An agile aircraft protection and energy supply system includes a protective airbag body, an energy supply unit, a pressure regulation unit, and a core control unit with an integrated neural network prediction module. Communication modules and data exchange are achieved via shielded cables. The protective airbag is sealed to the inflation and deflation modules of the pressure regulation unit via air ducts. A pressure detection module is fixed inside the protective airbag and connected to the communication port of the core control unit via a shielded cable. The solar power module and small-power generator module of the energy supply unit are connected in parallel with an energy storage module via a common DC bus. This DC bus is connected to the working power supply terminals of the core control unit and the pressure regulation unit via shielded cables through the AC output terminal of an inverter. The core control unit is equipped with a neural network prediction module, and its communication port forms a closed-loop data link with the battery management system of the energy supply unit, the variable frequency fan control terminal of the pressure regulation unit, the electric exhaust valve drive terminal, and the status monitoring module via shielded cables, enabling the output of generator start / stop commands only to meet nighttime power demands.

[0009] Furthermore, the solar power supply module consists of a solar panel with a conversion efficiency of not less than 22% and a maximum power point tracking controller. The output terminal of the maximum power point tracking controller is connected in parallel with the positive and negative terminals of the battery pack via a DC bus. During the day, the charging current is locked by the core control unit at the optimal power of 6.336 kW.

[0010] Furthermore, the low-power generator module is a rated 6-8 kW diesel generator, whose output is connected to the DC bus via a rectifier unit, and the generator start-stop control line is directly connected to the digital output port of the core control unit, and only accepts start commands when the battery energy is predicted to be insufficient.

[0011] Furthermore, the energy storage module consists of a 19.2 kWh / 96 V battery pack and a battery management system. The SOC signal line of the battery management system is connected to the core control unit via the controller local area network bus. The inverter input terminal is fixedly connected to the positive and negative terminals of the battery and outputs 380 V AC to power the pressure regulation unit. SOC refers to the state of charge.

[0012] Furthermore, the pressure detection module consists of at least 13 pressure sensors, of which 11 are fixed in the middle of the 11 air ribs of the protective airbag body, and the remaining 2 are fixed in the top of the inner side of the two end doors. All sensors are connected in series via RS-485 bus to the AI ​​module of the core control unit.

[0013] Furthermore, the inflation module includes two 4 kW low-pressure fans and two 4.3 kW high-pressure fans. The air outlets of the low-pressure fans are connected in parallel to the main inflation interface via a tee. The air outlets of the high-pressure fans are connected to the end door area and the middle air rib via branch interfaces. The frequency conversion control terminals of the fans are connected to the analog output port of the core control unit via shielded cables.

[0014] Furthermore, the exhaust module consists of four electric exhaust valves, two of which are basic exhaust valves symmetrically installed on the middle air passage branches on both sides of the airbag, and two spare exhaust valves are installed at the lower corners of the two end doors. The drive coils of each valve are connected in parallel to the digital output port of the core control unit via shielded cables.

[0015] Furthermore, the core control unit adopts an S7-1500 series PLC and inserts a neural network prediction module on its guide rail. The neural network prediction module exchanges data with the CPU in real time through the internal backplane bus. The CPU communication port is connected to the 4G data transmission unit via Ethernet to realize remote data uploading and command downloading.

[0016] Furthermore, the status monitoring module includes a data acquisition submodule and a 4G data transmission unit. The RS-485 port of the data acquisition submodule is connected in parallel to a pressure detection module, a battery management system, an ambient temperature and humidity sensor, and a solar radiation intensity sensor. The 4G data transmission unit is connected to the data acquisition submodule via a serial port and sends real-time data to the IoT server.

[0017] Furthermore, the core control unit has a built-in load priority control program, and its digital output port implements graded power supply in the order of pressure regulation unit, heater, winch, lighting and auxiliary equipment. When the battery SOC is ≤20%, the program shields the output of non-core loads to ensure that the pressure regulation unit is continuously powered; SOC refers to the state of charge.

[0018] Beneficial effects:

[0019] 1. In this invention, photovoltaic charging is prioritized during the day and battery discharge is performed at night. The neural network starts and stops the low-power diesel engine as needed, which greatly improves the utilization rate of solar energy and saves fuel, thus achieving efficient energy utilization.

[0020] 2. This invention utilizes an energy storage-inverter to buffer shocks, eliminating the need for diesel engines to directly handle peak loads during wind turbine startup, significantly reducing system size and weight, and making mobile deployment more convenient.

[0021] 3. This invention employs 13 pressure sensors, graded charging and discharging valves, and load priority management to automatically stabilize the airbag pressure within a safe range day and night, and prevent overcharging and over-discharging of the battery, thereby extending the cycle life and ensuring the structural stability of the protection system and the long lifespan of the equipment. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of an aircraft agile protection and energy supply system according to the present invention;

[0023] Figure 2 This is a perspective view of the protective airbag body of the present invention;

[0024] The attached diagram is labeled as follows: 1. Protective airbag body; 2. Energy supply unit; 21. Solar power supply module; 22. Small power generator module. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0026] like Figure 1 , Figure 2 As shown, the present invention provides an agile protection and energy supply system for aircraft, including a protective airbag body 1, an energy supply unit 2, a pressure regulation unit, and a core control unit with an integrated neural network prediction module, and realizes communication module connection and data interaction through a shielded cable.

[0027] The core control unit uses an S7-1500 series PLC (equipped with a neural network prediction module) as the central hub of the system to coordinate energy distribution and pressure regulation. During the day, it prioritizes the storage of solar energy through the energy supply unit 2, and at night it drives the pressure regulation unit to maintain the airbag pressure. At the same time, it uses the neural network prediction module to dynamically plan the generator energy replenishment strategy to achieve the dual goals of pressure stability and efficient energy utilization.

[0028] In other words, during the day, the battery is charged primarily by solar power modules (with a conversion rate of around 20%), reducing reliance on diesel generators and increasing solar energy utilization to over 60%. By using LSTM neural networks to predict nighttime energy consumption, the generator only replenishes the energy needed to meet nighttime demands (rather than fully charging the battery). Combined with a small-power generator (significantly reduced size and fuel consumption), fuel waste is reduced by over 40%. Through load priority control (pressure regulation unit takes priority) and BMS (battery management system) to prevent overcharging and over-discharging (shutdown when SOC (state of charge) reaches the standard), unnecessary power consumption by non-core equipment is avoided, while extending battery life (2500-2800 cycles) and reducing indirect energy losses caused by equipment replacement.

[0029] The end of the protective airbag body 1 is provided with an energy supply unit 2, a pressure regulation unit, an energy storage unit, and a core control unit with an integrated neural network prediction module; the core control unit is connected to the communication modules of the energy supply unit and the pressure regulation unit respectively, and is used to coordinate energy distribution and pressure regulation to achieve stable airbag pressure and efficient energy utilization.

[0030] like Figure 2As shown, the energy supply unit 2 includes a solar power module 21, a small-power generator module 22, and an energy storage module, which are used to provide power support for the system.

[0031] The energy storage module of energy supply unit 2 outputs 380 V AC power through a 30 KVA inverter. Part of the power is transmitted to the core control unit through a shielded cable to provide it with working power. The other part is also transmitted to the inflation module (2 low-pressure fans of 9-19 and 2 high-pressure fans of GHBH5D736AR8) and the exhaust module (electric exhaust valve) of the pressure regulation unit through a shielded cable to provide power to these actuators that directly act on the protective airbag body 1.

[0032] A pressure detection module (at least 13 pressure sensors) deployed on the surface of the protective airbag body 1 transmits the collected real-time pressure signals of the airbag to the core control unit via a shielded cable. The core control unit generates control commands based on the pressure signals (e.g., ≥ 6 kPa during the day, ≤ 4.2 kPa at night), and then sends them to the execution components of the pressure regulation unit via the shielded cable (e.g., activating the exhaust valve to release pressure, activating the fan to inflate). The fan outlet of the pressure regulation unit is sealed to the inflation port of the protective airbag body 1 via a dedicated air pipeline, and the exhaust valve is sealed to the exhaust port of the protective airbag body 1 via an air pipeline. Ultimately, the power from the energy supply unit 2 is converted into the inflation and deflation actions of the protective airbag body 1 through the execution components, thus completing the pressure regulation.

[0033] The specific connection between the protective airbag body 1 and the energy supply unit 2 is mainly through cables and air pipelines respectively. The power line is used to supply power to the actuators such as the fan, while the air system realizes the delivery and discharge of gas. The two are physically independent and are linked with the core control unit and pressure control unit through standardized interfaces to ensure that the internal pressure of the airbag is dynamically adjusted according to the preset threshold and maintain the stability of the hangar environment.

[0034] The pressure control unit includes an inflation module, an exhaust module, and a pressure detection module, used to respond to changes in airbag pressure caused by diurnal temperature variations.

[0035] The neural network prediction module is used to predict energy demand and guide generators to replenish energy as needed.

[0036] The energy storage unit includes an inverter and a battery; it is mainly used for energy storage and supply requirements.

[0037] The working process of this invention is as follows:

[0038] 1. System initialization and daily monitoring startup: After the system is powered on, the core control unit (S7-1500 PLC) first establishes communication connections with the pressure detection module (13 pressure sensors) of the protective airbag body 1, the energy supply unit 2 (solar power supply module 21, small power generator module 22), the energy storage unit (inverter + battery) and the pressure control unit. It collects data such as the current airbag pressure, ambient temperature and humidity, battery SOC, and solar irradiance in real time through the shielded cable to complete the initialization self-test.

[0039] 2. Daytime workflow (energy storage + pressure relief):

[0040] If solar irradiance ≥ 300W / m 2 The core control unit commands the solar power module 21 to start, and tracks the maximum power point through the MPPT (Maximum Power Point Tracking Controller) controller, converting solar energy into a 96V adaptive voltage, and prioritizing charging the battery of the energy storage unit (matching the optimal charging power of 6.336 kW).

[0041] Meanwhile, the pressure detection module provides real-time feedback on the airbag pressure. If the pressure is ≥6 kPa (the upper limit of the daytime safety threshold), the core control unit sends a command to the pressure control unit to activate the electric exhaust valve, which releases the airbag pressure to the 5.5-6 kPa range before closing the exhaust valve to prevent damage to the airbag material due to overpressure.

[0042] 3. The nighttime workflow (energy release + pressure replenishment) is as follows:

[0043] When the ambient light intensity is <100 W / m 2 (At night) the core control unit switches the energy supply logic and instructs the inverter of the energy storage unit to start, converting the 96V DC power from the battery into 380V AC power to power the pressure regulation unit.

[0044] If the pressure detection module reports that the airbag pressure is ≤4.2 kPa (lower limit of the nighttime safety threshold), the core control unit first starts the two 9-19 type low-pressure fans of the pressure regulation unit to quickly inflate to 4.5 kPa and then shuts down the low-pressure fans. Then, it starts the two GHBH5D736AR8 type high-pressure fans to precisely replenish the pressure until the airbag pressure stabilizes in the 4.2-6 kPa range, and then stops the high-pressure fans.

[0045] 4. Workflow during cloudy / rainy days / when solar energy is insufficient (predictive energy replenishment + energy security):

[0046] If the daytime solar irradiance remains below 200 W / m 2The core control unit calls the LSTM neural network prediction module, which combines the meteorological data (temperature, sunshine duration) for the next 24 hours with the current battery SOC to calculate the energy consumption required for nighttime inflation and the energy that the battery needs to replenish.

[0047] If the battery energy is predicted to be insufficient, the core control unit instructs the small power generator module 22 to start (in an environment of -10℃, the diesel heating device should be started first for 5-10 minutes). The generator directly replenishes the battery of the energy storage unit until it meets the nighttime needs (no need to fully charge). Then, the core control unit instructs the generator to stop.

[0048] At night, it continues to operate according to the logic of battery power supply and fan replenishment. If the battery SOC is unexpectedly ≤20%, the core control unit will trigger the generator for emergency power replenishment to ensure that pressure regulation is not interrupted.

[0049] 5. Anomaly Handling and Linkage Protection:

[0050] If the pressure detection module reports that the airbag pressure exceeds 6.2 kPa or is lower than 3.8 kPa, the core control unit will not only control the pressure regulation unit but also link the alarm module to send an SMS alarm.

[0051] The BMS of the energy storage unit can monitor the voltage, current, temperature and remaining capacity (SOC) of the battery in real time, and link with the core control unit to monitor the battery voltage and temperature in real time. If overcharging (SOC≥80%) or over-discharging (SOC≤20%) occurs, it will immediately feed back to the core control unit to suspend charging or trigger the generator to supplement power, thus extending the battery life.

[0052] The specific configuration and working logic of the energy supply unit are as follows:

[0053] The solar power module 21 includes a high-efficiency solar panel and an MPPT controller. The solar panel has a conversion efficiency of ≥22%, and the MPPT controller is used to track the maximum power point.

[0054] The low-power generator module 22 is a diesel generator with a rated power of 6-8 KW and an output voltage of It is also equipped with a diesel heating device, which can be started after heating for 5-10 minutes in an environment of -10℃;

[0055] The energy storage module is a 19.2 kWh / 96 V battery pack and BMS, with a maximum discharge power of 57.6 KW, an optimal charge and discharge power of 6.336 KW, and a cycle life of 2500-2800 cycles.

[0056] The working logic of the energy supply is as follows: during the day, the solar power module charges the battery pack; at night, the battery pack supplies power to the pressure regulation unit through a 30KVA inverter (input 84-104V, output 380V); when solar energy is insufficient on cloudy or rainy days, the small-power generator module supplements the battery pack with power.

[0057] When operating, when solar irradiance is ≥300 W / m 2 At this time, the MPPT controller converts the DC power output from the photovoltaic array into a 96V adaptive voltage to charge the battery pack, with the charging current controlled at 65A (matching the optimal charging power of 6.336 KW).

[0058] In night mode (19:00-07:00 the next day), the core control unit switches the inverter working mode, and the battery pack outputs 380 V AC power through the inverter to power the pressure control unit and core equipment.

[0059] In the cloudy / rainy day mode, when the solar irradiance is <200 W / m 2 Furthermore, when the battery charging current is less than 10 A, the core control unit triggers the generator to start, and the generator directly supplies power to the battery pack, with the supply voltage stabilizing at 102 V.

[0060] Preferably, the maximum starting impact power of the inflation module during operation is 12.9 KW; traditional solutions require a generator with a power of ≥38.7 KW and a volume of approximately 1.2 m³. 3 Weighing approximately 500 kg and with a rated fuel consumption of approximately 8 L / h, the 8 KW generator of this invention is directly connected to the battery pack and only undertakes the task of supplementing power, without the need for matching load to start up.

[0061] The low-power generator module 21 is directly connected to the battery pack of the energy storage module and is only used to replenish the battery. It does not need to match the starting impact power of the charging module. Compared with the traditional generator that needs to be 3 times the power of the charging motor, the volume is reduced by ≥40%, the weight is reduced by ≥35%, and the fuel consumption is reduced by ≥50%.

[0062] The specific configuration and control logic of the pressure control unit are as follows:

[0063] The inflation module includes two 4 kW low-pressure blowers (model 9-19, flow rate 1535-2062 m³ / h). 3 (4.5-4.6 KPa total pressure, with frequency converter) and two 4.3 KW high-pressure blowers (700 m³ / h, total pressure 4.5-4.6 KPa, with frequency converter) and two GHBH5D736AR8 models. 3 / h, total pressure 14KPa);

[0064] The exhaust module includes an electric exhaust valve with a power ≤0.1 KW;

[0065] The pressure detection module includes at least 13 pressure sensors for detecting the pressure of the airbag module and the end door;

[0066] The control logic is as follows: when the internal pressure of the airbag is ≥ 6 kPa during the day, the core control unit starts the exhaust valve to release the pressure to 5.5-6 kPa; when the internal pressure is ≤ 4.2 kPa at night, the low-pressure fan is first started to quickly inflate to 4.5 kPa, and then the high-pressure fan is switched to accurately replenish the pressure to 4.2-6 kPa.

[0067] Preferably, the daytime pressure relief control includes: when the average pressure of the 13 pressure sensors is ≥ 6 kPa, the core control unit activates 2 exhaust valves and closes them when the pressure drops to 5.8 kPa; if the pressure continues to rise to 6.2 kPa, all 4 exhaust valves are activated until the pressure drops back to 5.5 kPa.

[0068] Preferably, the nighttime inflation control includes: when the average pressure is ≤ 4.2 kPa, first start two low-pressure blowers for rapid inflation, and after 15 minutes the pressure rises to 4.5 kPa, shut down the low-pressure blowers and start one high-pressure blower for precise inflation, and stop the machine when the pressure reaches 5.0 kPa; if the pressure drops to 4.3 kPa again within 30 minutes, start two high-pressure blowers for coordinated inflation.

[0069] Specifically, all 13 pressure sensors are fixed to the inside of the protective airbag body 1 through sealed interfaces, and are distributed as follows:

[0070] Eleven pressure sensors correspond to eleven independent air rib modules (the core load-bearing units of the hangar inflation structure) of the protective airbag body. Each sensor is fixed at the middle height of the corresponding air rib module (avoiding areas where airbag wrinkles are likely to occur) to ensure the collection of the average pressure of the module. The remaining two pressure sensors are installed in the sealing areas of the first and second end doors of the protective airbag body (the connection between the end door and the protective airbag body), respectively, and are fixed on the top of the inner side of the end door to monitor the pressure of the end door sealing area (to avoid pressure leakage at the end door that could lead to overall pressure imbalance).

[0071] All four electric exhaust valves are sealed to the air passage pipes of the protective airbag body 1 via flange interfaces and are symmetrically distributed: two basic exhaust valves (the two that are activated first) are installed on the air passage branch pipes in the middle of both sides of the protective airbag body, near the opposite side of the low-pressure blower inflation port, to ensure that they are staggered from the inflation path when depressurizing and to avoid airflow interference; the other two standby exhaust valves (the two that are activated when the pressure exceeds 6.2 kPa) are installed at the corners of both ends of the protective airbag body (near the bottom of the end door), forming a symmetrical layout with the basic exhaust valves on both sides and at both ends, to ensure that the pressure in all areas of the airbag drops evenly when depressurizing and to prevent local indentation.

[0072] Two 9-19 type low-pressure blowers (4 KW, high flow rate) are sealed to the main inflation port of the protective airbag body 1 via pressure-resistant rubber hoses (100 mm inner diameter). The main inflation port is located on the main air passage pipe in the middle of one side of the protective airbag body (diagonally distributed with the basic exhaust valve on that side). The air outlets of the two low-pressure blowers are connected in parallel to the main inflation port via tee connectors to ensure stable airflow into the main air passage of the airbag during high-flow inflation, achieving rapid pressurization. Two GHBH5D736AR8 type high-pressure blowers (4.3 KW, high precision) are sealed to the branch inflation ports of the protective airbag body 1 via pressure-resistant rubber hoses (50 mm inner diameter). One high-pressure blower is prioritized and connected to the airway branch pipe near the end door of the protective airbag body to accurately replenish air to the sealing area of ​​the end door (the end door is prone to local pressure loss due to opening and closing); another high-pressure blower is connected to the branch interface of the air rib module in the middle of the protective airbag body; the two blowers form an air replenishment layout in the middle of the end door to ensure that the pressure in all areas of the airbag reaches the standard simultaneously when accurately replenishing pressure, and avoid local pressure deviation.

[0073] The neural network prediction module of the core control unit adopts a customized LSTM architecture. The input layer receives a 55×8 time series matrix (including meteorological data for the next 24 hours, historical solar power, and real-time status parameters), and extracts features through a double-layer bidirectional LSTM hidden layer (32 neurons in the first layer and 16 neurons in the second layer). The output layer outputs the total energy consumption for nighttime refilling and the amount of energy that the battery needs to be replenished through a dual-channel fully connected layer.

[0074] The customized LSTM architecture achieves a prediction error of ≤5% through a weighted mean squared error loss function, an adaptive learning rate optimizer, and an online incremental fine-tuning mechanism.

[0075] In the dual-layer bidirectional LSTM hidden layer, the first LSTM extracts short-term local features (including the weight of temperature changes in the next 6 hours), and the second bidirectional LSTM integrates global correlation features from the past to the present and then to the future. Each layer is connected in series with an adaptive random deactivation layer (i.e., an adaptive Dropout layer) (random deactivation rate 0.2-0.3) and a batch normalization layer.

[0076] The adaptive random deactivation layer is an optimization layer that can dynamically adjust the proportion of neurons temporarily deactivated based on the complexity of the input data features. The core principle is to randomly stop 20%-30% of the neurons in the layer during model training (i.e., not participate in the calculation and parameter update of the current round), and the deactivation ratio will be adaptively adjusted. For example, when the temperature change in the next 6 hours is large and the feature complexity is high, the deactivation rate is appropriately increased (approaching 0.3) to avoid the model from over-relying on local abnormal features; when the input battery SOC changes steadily and the features are simple, the deactivation rate is appropriately reduced (approaching 0.2) to ensure sufficient learning of effective features.

[0077] This layer, in conjunction with the batch normalization layer, can further reduce the risk of model overfitting (avoiding the model performing well only on training data but with increased errors in actual predictions), ensuring that the LSTM module's predictions of nighttime gas charging energy consumption and battery replenishment power are more stable, with errors controlled within ≤5%, providing a reliable basis for generators to replenish energy on demand.

[0078] The neural network prediction module forms a closed loop with the core control unit. The prediction results are directly used to calculate the supplementary energy of the small power generator. The supplementary energy target is dynamically corrected every 30 minutes through short-term prediction to achieve on-demand supplementary energy and avoid over-generation.

[0079] The input layer performs multi-dimensional time-series data fusion processing, including:

[0080] 1. Temporal Reconstruction of Input Parameters: The 8-dimensional input parameters are transformed into a 2D temporal matrix of time steps × features, specifically including: meteorological data for the next 24 hours: sampled at 1-hour granularity to form a 24×3 matrix (3 features: average temperature, sunshine duration, and wind force); historical correlation data: solar power generation in the past 24 hours (1 hour / point, 24×1 matrix), airbag pressure fluctuation value in the past 6 hours (1 hour / point, 6×1 matrix); real-time status parameters: current airbag pressure (1 dimension), target pressure (1 dimension), and current battery SOC (1 dimension); finally forming a 55×8 temporal input matrix (time steps = 24 + 24 + 6 + 1 = 55, feature dimensions = 3 + 1 + 1 + 1 + 1 + 1 = 8), fully preserving the temporal correlation of the parameters;

[0081] 2. Data preprocessing:

[0082] (1) Timing alignment: The parameters of different sampling frequencies are unified to a 1-hour step size by linear interpolation to eliminate the time scale difference;

[0083] (2) Normalization: Improved Min-Max standardization is adopted (dynamically updating the maximum / minimum value), the formula is: ;

[0084] in, This is the normalized result value obtained after modified Min-Max standardization (the value after mapping the original data to a specific range). The original data sample to be normalized (the original values ​​that need to be standardized). It is the minimum value among all data within a dynamically updated sliding window (since the maximum / minimum value is dynamically updated, the extreme value is obtained in real time based on the dataset within the current sliding window). This represents the maximum value among all data within a dynamically updated sliding window (similarly, it is updated in real time based on the dataset within the current sliding window).

[0085] The sliding window is set to 7 days to avoid interference from extreme historical data (such as rare high temperatures) on the current normalization, so that the input value is stable in the [0,1] range and the model convergence speed is improved.

[0086] This invention employs a two-layer structure combining local feature capture with global correlation fusion to enhance the temporal dependency between meteorology and energy consumption.

[0087] The first LSTM layer (32 neurons) of the bidirectional LSTM hidden layer is responsible for extracting short-term local features. Irrelevant information is dynamically filtered through a forget gate (e.g., the influence of wind data from 3 days ago on the current prediction is reduced to 0). The input gate focuses on the temperature change in the next 6 hours (weight coefficient 1.2) and the current SOC (state of charge) value (weight coefficient 1.1). The output gate generates a feature vector related to short-term energy consumption (e.g., the local pattern that for every 1°C drop in temperature, the energy consumption for inflation increases by 3%).

[0088] ;

[0089] in, for The output of the forget gate at each time step (between 0 and 1, where 1 means completely retained and 0 means completely forgotten). The sigmoid activation function (maps values ​​to 0~1); The weight matrix for the forgetting gate (dimension [32, 32+n], where 32 is the hidden layer dimension and n is the number of input features); specifically, the wind data from 3 days ago corresponds to... An element value of 0 (weight decays to 0) indicates that the feature does not affect the forget gate; The hidden state at time t-1 (the output feature of the previous time step); The input feature vector at time t (including wind force, temperature change, SOC, etc.); This is the bias term for the forget gate (dimension [32, 1]).

[0090] The input gate determines which new information is stored in the cell state (with a focus on temperature changes and SOC):

[0091] ;

[0092] ;

[0093] in, for The input gate output at each time step (between 0 and 1, where 1 indicates that the new information is completely preserved); for The state of candidate cells at any given time (new information to be updated); This is the hyperbolic tangent activation function (maps values ​​to -1 to 1); The weight matrix of the input gate (abbreviated as) ); Temperature changes in the next 6 hours The element value is 1.2. This corresponds to the current SOC value. The element value is 1.1 (highest weight); The weight matrix for continuing cell states; This is a bias term for the cell state of the input gate candidate.

[0094] Integrate the results of the forgetting gate and the input gate to update long-term memory:

[0095] ;

[0096] in, for Cellular state at any given moment (long-term memory); for Cellular state at any given moment; This is element-wise multiplication (Hadamard product).

[0097] Determine which information output from the cell state as short-term characteristics (such as energy consumption-related patterns):

[0098] ;

[0099] ;

[0100] in, The output gate at time t is between 0 and 1, controlling the output ratio of the cell state. The hidden state at time t (the output of the first LSTM layer, i.e., the short-term local feature vector, such as the rule that the energy consumption for filling increases by 3% for every 1°C drop in temperature). This is the weight matrix of the output gate; This is the bias term for the output gate.

[0101] The second LSTM layer (16 neurons) of the bidirectional LSTM hidden layer adopts a bidirectional structure (forward + backward propagation). Forward propagation learns the evolution law from historical data to the current state (such as the impact of accumulated sunshine on battery power), and backward propagation obtains the mapping relationship from future weather to future energy consumption (such as the suppression of solar energy supplementation by forecasted rainfall). Finally, they are fused into a global feature vector, which solves the defect of traditional unidirectional LSTM that can only capture past dependencies.

[0102] For the second LSTM layer (16 neurons), the fusion method typically uses feature concatenation. This method can preserve the complete information of the bidirectional branches to the greatest extent. The specific formula and parameter meanings are as follows:

[0103] Forward LSTM: in chronological order Process the sequence and output the forward hidden state. To capture the evolutionary patterns from historical data to the current state (such as the impact of accumulated sunlight on battery power).

[0104] Forward hidden state formula: ,in This is the previous forward-hidden state. This is the current input (the output features of the first LSTM layer).

[0105] Backward LSTM: Reverse time sequence Process the sequence and output the backward hidden state. To capture the mapping relationship between future weather and future energy consumption (such as the suppression of solar energy replenishment by forecasted rainfall).

[0106] Backward hidden state formula: ,in This is the state that is hidden backwards at the next moment.

[0107] The formula for fusing global feature vectors in the second layer of LSTM. Global feature vector at time step We obtain this by concatenating the forward and backward hidden states: .

[0108] in, The fused global feature vector (32 dimensions, i.e., 16 feedforward neurons + 16 feedback neurons) contains bidirectional dependencies (both the influence of the past on the present and the association of the future on the present). The hidden state of the forward LSTM at time t (dimension 16) encodes local patterns from history to the present (e.g., for every 10 hours of accumulated sunshine in the past 3 days, the battery capacity increases by 5%). The hidden state (dimension 16) of the backward LSTM at time t encodes the correlation between the future and the present (e.g., for every 20% increase in the predicted rainfall probability in the next 24 hours, the solar energy supplementation decreases by 8%). "" indicates a feature concatenation operation (column concatenation), which merges the forward and backward feature vectors into a higher-dimensional global feature.

[0109] The advantages of the fusion method compared to addition Or take the average The advantages of this splicing method are: it fully preserves the unique information of the forward and backward branches (such as the historical cumulative effect and the future forecast impact are different features and will not cancel each other out by addition); and it provides richer feature representation for subsequent tasks (such as energy consumption prediction) through dimensional expansion (16→32), thus solving the limitation of traditional unidirectional LSTM that can only rely on past information.

[0110] Each LSTM layer is followed by an adaptive Dropout layer (the dropout rate is dynamically adjusted with each training round: initially 0.2, increasing to 0.3 as the validation set error rises) and a batch normalization layer. By suppressing the co-adaptation phenomenon of neurons, the model's generalization ability on small sample datasets (such as extreme cold weather samples accounting for only 5%) is improved by 40%.

[0111] The output parameters of the neural network prediction module are: the total energy consumption of the nighttime inflation module and the amount of energy that the battery needs to be replenished; its control logic is: to determine whether to start the small-power generator module based on the output parameters, only replenishing the energy to meet the nighttime demand, without needing to fully charge the battery, with a prediction error ≤5%;

[0112] The output layer adopts a dual-channel fully connected layer structure, simultaneously outputting two core parameters:

[0113] First Channel: Total Energy Consumption of Nighttime Inflation Module (Continuous Value, Unit) The activation function is a linear function (to avoid ReLU truncation of small energy consumption values);

[0114] Second channel: Energy required to replenish the battery (continuous value, unit) The required replenishment is calculated by the difference between the predicted energy consumption and the current SOC (predicted energy consumption - current releaseable energy of the battery).

[0115] An attention mechanism is set between the output layer and the hidden layer to give higher weights to key features such as the temperature in the next 12 hours and the solar power in the next 3 hours (the weights are dynamically updated) to ensure the prediction accuracy of the core parameters.

[0116] The status monitoring module includes a data acquisition submodule and a 4G DTU (4th-Generation Data Transfer Unit) transmission submodule.

[0117] In the status monitoring module, the core function of the 4G DTU transmission submodule is to stably and in real-time transmit the monitoring data (such as equipment operating status, environmental parameters, energy consumption data, etc.) collected by the data acquisition submodule to the back-end monitoring system (such as cloud platform, local server) through the 4G mobile communication network. At the same time, it can also receive control commands (such as parameter adjustment, equipment start and stop signals) issued by the back-end, realize two-way data interaction of data uploading and command downloading, and is a key communication bridge connecting the front-end data acquisition and the back-end system.

[0118] The data acquisition submodule collects the following parameters: airbag pressure, ambient temperature and humidity, solar radiation intensity, battery SOC, and operating status of the fan and generator.

[0119] The 4G DTU transmission submodule transmits the collected data to the IoT server, providing real-time data support for the neural network prediction module.

[0120] The alarm module is linked with the status monitoring module, and its function is as follows:

[0121] Pressure alarm: When the airbag internal pressure is lower than the low pressure alarm setting value, a low pressure alarm is triggered; when it is lower than the emergency low pressure setting value, an emergency low pressure alarm is triggered; when it is higher than the high pressure alarm setting value, a high pressure alarm is triggered.

[0122] Weather warnings are triggered when weather data is captured and strong winds or heavy snow are detected.

[0123] Preferably, the alarm method can be SMS alarm.

[0124] Preferably, the core control unit is also configured with load priority control logic, with the priority from high to low as follows: pressure regulation unit (low-pressure fan, high-pressure fan, exhaust valve) is higher than cabinet heater (0.7KW×2) is higher than winch (1.6KW×4) is higher than lighting (3.2KW) is higher than non-core auxiliary equipment, so as to ensure that the core functions of pressure regulation and energy supply are given priority in power supply.

[0125] The BMS is linked with the core control unit to monitor the battery voltage, current, temperature and SOC in real time. When the SOC is ≤20% and the solar power supply is insufficient, the small power generator module is triggered to start and supplement the power. When the SOC is ≥80% or the predicted supplementary power is reached, the generator is controlled to stop to avoid excessive power generation and waste of energy.

[0126] The coordinated operation of the inverter and the battery serves as the hardware foundation for the LSTM neural network prediction module to achieve on-demand energy replenishment: after the LSTM neural network prediction module predicts the total energy consumption for nighttime refueling, the core control unit calculates the energy replenishment based on the battery's SOC (monitored by the BMS). If energy replenishment is needed, a small-power generator charges the battery. At night, the battery supplies power to the load through the inverter, and the generator does not need to run, achieving the goal of replenishing only the energy required and not wasting fuel.

[0127] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. An aircraft agile protection and energy supply system, characterized in that, The system comprises a protective airbag body, an energy supply unit, a pressure control unit, and a core control unit with an integrated neural network prediction module. Communication modules and data exchange are achieved via shielded cables. The protective airbag is sealed to the inflation and deflation modules of the pressure control unit via air ducts. A pressure detection module is fixed inside the airbag and connected to the communication port of the core control unit via a shielded cable. The solar power module and small-power generator module of the energy supply unit are connected in parallel with the energy storage module via the same DC bus. This DC bus is connected to the working power supply terminals of the core control unit and the pressure control unit via shielded cables through the inverter's AC output terminal. The core control unit is equipped with a neural network prediction module, and its communication port forms a closed-loop data link with the battery management system of the energy supply unit, the variable frequency fan control terminal of the pressure control unit, the electric exhaust valve drive terminal, and the status monitoring module via shielded cables. This enables the output of generator start / stop commands to only supplement the power demand at night.

2. The aircraft agile protection and energy supply system according to claim 1, characterized in that: The solar power module consists of a solar panel with a conversion efficiency of not less than 22% and a maximum power point tracking controller. The output of the maximum power point tracking controller is connected in parallel with the positive and negative terminals of the battery pack via a DC bus. During the day, the charging current is locked by the core control unit at the optimal power of 6.336 kW.

3. The aircraft agile protection and energy supply system according to claim 1, characterized in that: The low-power generator module is a 6-8 kW rated diesel generator. Its output is connected to the DC bus via a rectifier unit, and the generator start-stop control line is directly connected to the digital output port of the core control unit. It only accepts start commands when the battery energy is predicted to be insufficient.

4. The aircraft agile protection and energy supply system according to claim 1, characterized in that: The energy storage module consists of a 19.2 kWh / 96 V battery pack and a battery management system. The SOC signal line of the battery management system is connected to the core control unit via the controller local area network bus. The inverter input is fixedly connected to the positive and negative terminals of the battery and outputs 380 V AC to power the pressure regulation unit. SOC refers to the state of charge.

5. The aircraft agile protection and energy supply system according to claim 1, characterized in that: The pressure detection module consists of at least 13 pressure sensors, of which 11 are fixed in the middle of the 11 air ribs of the protective airbag body, and the remaining 2 are fixed in the top of the inner side of the two end doors. All sensors are connected in series via RS-485 bus to the AI ​​module of the core control unit.

6. The aircraft agile protection and energy supply system according to claim 1, characterized in that: The inflation module includes two 4 kW low-pressure blowers and two 4.3 kW high-pressure blowers. The air outlets of the low-pressure blowers are connected in parallel to the main inflation interface via a tee. The air outlets of the high-pressure blowers are connected to the end door area and the middle air rib via branch interfaces. The frequency converter control terminals of the blowers are connected to the analog output port of the core control unit via shielded cables.

7. The aircraft agile protection and energy supply system according to claim 1, characterized in that: The exhaust module consists of four electric exhaust valves, two of which are basic exhaust valves symmetrically installed on the middle air passage branches on both sides of the airbag, and two spare exhaust valves are installed at the lower corners of the two end doors. The drive coils of each valve are connected in parallel to the digital output port of the core control unit via shielded cables.

8. The aircraft agile protection and energy supply system according to claim 1, characterized in that: The core control unit adopts an S7-1500 series PLC and a neural network prediction module is plugged into its guide rail. The neural network prediction module exchanges data with the CPU in real time through the internal backplane bus. The CPU communication port is connected to the 4G data transmission unit via Ethernet to realize remote data uploading and command downloading.

9. The aircraft agile protection and energy supply system according to claim 1, characterized in that: The status monitoring module includes a data acquisition submodule and a 4G data transmission unit. The RS-485 port of the data acquisition submodule is connected in parallel to a pressure detection module, a battery management system, an ambient temperature and humidity sensor, and a solar radiation intensity sensor. The 4G data transmission unit is connected to the data acquisition submodule via a serial port and sends real-time data to the IoT server.

10. The aircraft agile protection and energy supply system according to claim 1, characterized in that: The core control unit has a built-in load priority control program that provides graded power supply through the digital output port in the order of pressure regulation unit, heater, winch, lighting, and auxiliary equipment. Here, SOC refers to the state of charge. When the battery SOC is ≤20%, the program will shield the power supply output of non-core loads to ensure that the pressure regulation unit is continuously and stably powered.

Images