Stratosphere airship regional residence control method, device, equipment and medium
By using wind speed prediction models and model predictive control algorithms, the propulsion and propeller speed of the airship are adjusted in real time, solving the problems of stratospheric airship control lag and high energy consumption, and achieving stable dwell and endurance in complex wind fields.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- LINZHOU (NINGBO) TECH CO LTD
- Filing Date
- 2026-02-04
- Publication Date
- 2026-06-26
Smart Images
Figure CN122284283A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of airship control technology, specifically to a method, device, control equipment, and storage medium for controlling the regional stay of a stratospheric airship. Background Technology
[0002] As a near-space platform that relies on buoyancy and propulsion systems to achieve dwell and maneuverability, stratospheric airships have significant advantages such as long endurance, large coverage radius, and regional dwelling capability. They have irreplaceable application value in areas such as continuous regional surveillance, meteorological detection, and communication relay, and their operational efficiency directly depends on the success rate of regional dwelling.
[0003] The core difference between stratospheric airships and traditional weather balloons lies in their active control capabilities. Traditional weather balloons have no propulsion system and drift entirely with the wind, making stationary deployment impossible. Stratospheric airships, on the other hand, use propulsion systems (such as propeller speed adjustment) to counteract the effects of wind, but their flight systems exhibit strong nonlinear characteristics: their large size leads to significant inertial lag, aerodynamic parameters dynamically change with altitude and wind speed, and propulsion efficiency is severely affected by airflow disturbances. Existing stratospheric airship control technologies largely rely on real-time wind speed measurement and PID feedback regulation, failing to fully consider the spatiotemporal correlation and predictability of stratospheric wind speed, resulting in control actions lagging behind wind field changes. When encountering unpredictable wind speed increases, propulsion power must be significantly increased to offset the wind load, not only increasing energy consumption but also potentially leading to loss of control due to power saturation, making it difficult to meet the requirements for regional stationary deployment. Therefore, there is an urgent need for a technical solution that can overcome the above-mentioned defects in order to improve the success rate of stratospheric airships in regional stay, reduce energy consumption, and ensure the continuity of long-endurance missions. Summary of the Invention
[0004] The purpose of this application is to provide a method, device, control equipment and storage medium for controlling the regional stay of stratospheric airships, so as to solve the problems of low airship stay success rate and control lag caused by stratospheric wind speed disturbance, and realize the early response of airships to wind field changes and regional stay.
[0005] In a first aspect, embodiments of this application provide a method for controlling the regional stay of a stratospheric airship, including: Based on real-time wind field data of the current location of the airship in the target area, a wind speed prediction model is used to predict the wind speed changes of the current location and a certain range around it within a preset time period, and outputs a wind speed prediction sequence. The wind speed prediction model includes an Empirical Mode Decomposition (EEMD) module, a Convolutional Neural Network (CNN) module, a Long Short-Term Memory (LSTM) network module, and a location encoding module. The wind speed prediction sequence includes eastward, northward, and vertical wind speed components. The current state vector of the airship is generated in real time based on its current position, speed, attitude and wind load. Based on the wind speed prediction sequence and the current state vector, a preset model predictive control algorithm is used to calculate the expected thrust and propeller speed adjustment required to offset future wind loads, with the goal of minimizing wind load offsetting error and optimizing energy consumption, under the condition of meeting the saturation constraints of the propulsion system. Based on the airship dynamics model, the desired thrust and propeller speed adjustment are corrected in real time to generate control commands. The control commands are sent to the propulsion system.
[0006] In one possible implementation, the training process of the wind speed prediction model includes: The historical wind field data of the target area within a historical time period is obtained. The historical wind field data includes basic information corresponding to each spatial location point in the target area and wind speed components in the east, north and vertical directions. The basic information includes longitude, latitude, altitude and recording time. Obtain wind field data for a spatial location point and its eight surrounding neighboring points from the historical wind field data; For each acquired spatial location point, the EEMD module is used to decompose a wind speed component at that spatial location point to obtain the intrinsic mode function (IMF) of the wind speed component. The intrinsic mode function (IMF) of the wind speed component is extracted and concatenated using the CNN module to obtain the IMF features of the wind speed component. The location encoding module is used to add the corresponding temporal location information to the IMF feature of the wind speed component to obtain the location-enhanced IMF feature of the wind speed component. The input and output of the LSTM module are as follows: the time sequence of the enhanced IMF feature at a single location is extracted by a sliding window as the input, and the original wind speed component value at the first time after the sliding window is output as a single value. The wind speed prediction model is trained using the steps described above.
[0007] In one possible implementation, the wind speed prediction model includes: The CNN module consists of two 1D convolutional layers, the LSTM module consists of two LSTM layers, and the position encoding module uses trigonometric function position encoding.
[0008] In one possible implementation, the current state vector is 12-dimensional, including longitude, latitude, altitude, eastward velocity, northward velocity, vertical velocity, roll angle, pitch angle, yaw angle, eastward wind load, northward wind load, and vertical wind load.
[0009] In one possible implementation, the propulsion system saturation constraints include: Thrust constraint: The thrust is between the maximum and minimum thrust specified by the propulsion system; Speed constraint: The maximum and minimum speed values are obtained based on the linear relationship between speed and propeller thrust, and the speed is between the maximum and minimum values; Control increment constraint: The speed adjustment at adjacent time points is between the maximum and minimum allowable speed change.
[0010] In one possible implementation, the objective function is constructed with the goals of minimizing wind load offsetting error and optimizing energy consumption as follows: ; in, The time domain length for wind field prediction. To extend the time domain length of the system control; To compensate for wind load error, This represents the predicted wind load. This represents the reaction force generated by the airship's propulsion system. This is the error weight matrix; This represents the planned force output of the airship propulsion system at time t+i. This is the energy consumption weighting coefficient.
[0011] Secondly, embodiments of this application provide a stratospheric airship regional stay control device, comprising: The wind speed prediction module is used to predict wind speed changes within a preset time period and its surrounding area within the target region based on real-time wind field data of the airship's current location, and outputs a wind speed prediction sequence. The wind speed prediction model includes an Empirical Mode Decomposition (EEMD) module, a Convolutional Neural Network (CNN) module, a Long Short-Term Memory (LSTM) network module, and a location encoding module. The wind speed prediction sequence includes eastward, northward, and vertical wind speed components. The state estimation module is used to acquire and generate the current state vector of the airship in real time based on its current position, speed, attitude and wind load. The control decision module is used to calculate the expected thrust and propeller speed adjustment required to offset future wind loads based on the wind speed prediction sequence and the current state vector, using a preset model predictive control algorithm, under the condition of meeting the saturation constraints of the propulsion system, with the goal of minimizing wind load offsetting error and optimizing energy consumption. The nonlinear compensation module is used to correct the desired thrust and propeller speed adjustment in real time based on the airship dynamics model, and generate control commands. An execution module is used to send the control commands to the propulsion system.
[0012] In one possible implementation, the propulsion system saturation constraints include: Thrust constraint: The thrust is between the maximum and minimum thrust specified by the propeller; Speed constraint: The maximum and minimum speed values are obtained based on the linear relationship between speed and propeller thrust, and the speed is between the maximum and minimum values; Control increment constraint: The speed adjustment at adjacent time points is between the maximum and minimum allowable speed change.
[0013] Thirdly, embodiments of this application provide a control device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor runs the computer program, it implements the stratospheric airship regional dwelling control method described in the first aspect of this application.
[0014] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-readable instructions thereon, which can be executed by a processor to implement the stratospheric airship regional dwelling control method described in the first aspect of this application.
[0015] The stratospheric airship regional stationing control method, device, control equipment, and storage medium provided in this application firstly achieves proactive adjustment of control actions through wind speed prediction, effectively solving the lag problem of traditional real-time feedback control. Simultaneously, its nonlinear compensation mechanism can accurately adapt to the airship's highly nonlinear dynamic characteristics, further enhancing control robustness under complex wind fields. This synergistic effect of the mechanism improves the airship's regional stationing success rate in complex stratospheric wind fields. Furthermore, prediction-based control decisions can optimize propulsion system energy consumption allocation in advance, avoiding redundant power consumption caused by sudden wind speed changes, thus contributing to endurance. In addition, it enables the airship to adapt to complex meteorological conditions such as sudden wind speed changes, ensuring the continuity and stability of stationing missions. Attached Figure Description
[0016] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart of a stratospheric airship regional dwell control method provided in this application is shown; Figure 2 This paper presents a schematic diagram of the overall architecture of the stratospheric airship regional dwelling control method provided in this application; Figure 3A schematic diagram of the architecture of the EEMD-CNN-LSTM model provided in this application is shown; Figure 4 A flowchart illustrating the training process of the wind speed prediction model provided in this application is shown; Figure 5 A schematic diagram of the structure of a stratospheric airship regional stay control device provided in this application is shown. Detailed Implementation
[0017] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
[0018] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains.
[0019] Furthermore, the terms "first" and "second," etc., are used to distinguish different objects, not to describe a specific order. Additionally, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to those processes, methods, products, or apparatuses.
[0020] This application provides a method and apparatus for controlling the regional stay of a stratospheric airship, a control device, and a computer-readable storage medium, which will be described below with reference to the accompanying drawings.
[0021] Please refer to Figure 1 It shows a flowchart of a stratospheric airship regional stay control method provided in this application. Figure 1 The flowchart shown includes steps S101-S105. For ease of understanding, this application provides the following... Figure 2 The diagram shows the overall architecture of the stratospheric airship regional stay control method. Figure 2 The medium architecture includes wind speed prediction (corresponding to S101), airship state estimation (corresponding to S102), control decision (corresponding to S103), nonlinear compensation (corresponding to S104), and execution feedback (corresponding to S105).
[0022] like Figure 1 , 2 As shown, the method includes the following steps: S101. Based on the real-time wind field data of the current location of the airship in the target area, a wind speed prediction model is used to predict the wind speed changes of the current location and a certain range around it within a future preset time period, and outputs a wind speed prediction sequence. Specifically, the preset time period can be the next 18 hours, 24 hours, etc. The time step of the wind speed prediction sequence can be 1 hour, including eastward, northward, and vertical wind speed components.
[0023] The process by which the wind speed prediction model determines the current position of the airship and its surrounding area is as follows: the spatial coordinates of the current position of the airship are converted into latitude and longitude grid indices, and boundary clipping is used to ensure that the index is within the target area; the grid point search function is called to generate an index list of the central grid point (the position of the airship) and the surrounding 8 neighboring grid points (a total of 9 points), and the problem of index out-of-bounds at boundary points is handled (points outside the boundary are replaced by the central point).
[0024] The wind speed prediction model includes an ensemble empirical mode decomposition (EEMD) module, a convolutional neural network (CNN) module, a long short-term memory (LSTM) network module, and a position encoding module. The wind speed prediction sequence includes eastward, northward, and vertical wind speed components. The position encoding module can employ the Transformer's "Sinusoidal Positional Encoding" module.
[0025] The wind speed prediction model is actually an EEMD-CNN-LSTM model composed of an EEMD module, a CNN module, and an LSTM module (e.g., ...). Figure 3 As shown in the figure, the wind speed prediction model, combined with a location encoding module, is trained using a large amount of original real datasets of the target area and predicts the changes in the three-dimensional wind speed components of the airship's current position and its surrounding area within a certain range over the next 24 hours, outputting a wind speed prediction sequence. The core objective of the wind speed prediction model in this application is to enable the model to accurately learn the temporal patterns of the wind field and achieve three-dimensional wind speed prediction for the next 24 hours.
[0026] like Figure 4 As shown, the training process of the wind speed prediction model includes: S201. Obtain historical wind field data of the target area within a historical time period. The historical wind field data includes basic information corresponding to each spatial location point in the target area and wind speed components in the east, north, and vertical directions. The basic information includes longitude, latitude, altitude, and recording time. S202. Obtain wind field data of a spatial location point and its eight surrounding neighboring points from the historical wind field data; The raw data acquisition and preprocessing are performed through the above steps S201 and S202, specifically including: 1. Read NetCDF format wind field data and extract dimensions such as longitude (lon), latitude (lat), time (time), and pressure layer (level), as well as the three-dimensional components of easterly wind (u), northerly wind (v), and vertical wind (w) (original data dimensions: time×level×lat×lon).
[0027] 2. Perform latitude and longitude alignment, dimension reconstruction, and barometric pressure-altitude conversion on the original data to lock the current position of the airship and 9 surrounding grid points (center + 8 neighbors).
[0028] Among them, latitude and longitude alignment processing: design dimension matching logic, take the minimum length of the longitude and latitude sequences as a unified scale, truncate the latitude and longitude data, reverse the latitude sequence to ascending order (south→north) and reverse the longitude sequence to ascending order (west→east) to ensure the consistency of the spatial grid; output the range of extreme values of latitude and longitude after alignment to realize the traceability of preprocessing results.
[0029] Among them, the dimension reconstruction process involves constructing a dimension exchange algorithm to reconstruct the wind field data from (time×level×lat×lon) to (lon×lat×level×time), adapting to the subsequent spatial grid slicing and time series extraction requirements, and outputting the dimension information of the reconstructed wind field data.
[0030] Among them, the pressure-to-altitude conversion process converts the pressure layer data (hPa) into altitude (km) using the relevant pressure-altitude formula.
[0031] 3. Extract local wind field data, perform time-height two-dimensional interpolation on the u, v, and w components of each grid point, filter out missing values (NaN), and ensure that the effective time step of each grid point is ≥20, finally obtaining effective time series data of u, v, and w for 9 grid points.
[0032] S203. For each acquired spatial location point, the EEMD module is used to decompose a wind speed component of the spatial location point to obtain the intrinsic mode function (IMF) of the wind speed component. Specifically, the EEMD batch decomposition process is as follows: 1. Configure EEMD decomposition parameters: Gaussian white noise intensity 0.1~0.4, ensemble decomposition times 100~200, suppress mode aliasing through multiple ensemble decompositions.
[0033] 2. Perform EEMD decomposition on all valid u, v, and w components of the 9 grid points one by one. Each component is decomposed into "n IMFs (Intrinsic Mode Functions) + 1 residual" (n is the number of IMFs, which is determined by the signal complexity).
[0034] 3. Remove residuals (which are not used in training but only for final result reconstruction), retaining only IMFs as model input features. Each IMF corresponds to a wind field fluctuation of a certain frequency, such as short-term gusts or medium-term trends.
[0035] After batch decomposition of EEMD, training samples are constructed: the window length lookback=15 is set, the time series sequence of a certain IMF in the previous 15 hours is used as input (X), and the original wind speed component value of the first hour after the window is used as output (y), that is, the past 15 hours are used to predict the next hour. The samples of each IMF are divided into training set and test set in an 8:2 ratio.
[0036] Constructing an EEMD-CNN-LSTM model with positional encoding: 1. Location Encoding Generation: Construct a sine-cosine location encoding function to generate a location encoding matrix that matches the input time series length (lookback=15). The encoding dimension is consistent with the dimension of the subsequent LSTM input features. Location encoding is used to mark the temporal position of each time step (e.g., hour 1, hour 2), avoiding model confusion about the time order and improving the accuracy of time series prediction.
[0037] 2. Model Structure Design: Design the EEMD-CNN-LSTM_with_PosEnc model, including: ①CNN part: It consists of two 1D convolutional layers (output channels 16 and 32), which extract local features for each IMF component and enhance nonlinear expression through ReLU activation function to extract local features of each IMF (such as wind speed change patterns and continuous fluctuation patterns within a 3-hour period).
[0038] ② Location Encoding Injection: After concatenating the CNN output features of all IMFs, the result is added to the location encoding matrix broadcast to enhance temporal location information. By adding time labels to local features, the model's perception of temporal location is strengthened.
[0039] ③LSTM time series modeling: It consists of two LSTM layers (with hidden layer dimensions of 64 and 128 respectively) to process the spliced time series features and suppress overfitting by using a dropout rate of 0.2; the first LSTM layer captures short-term time series correlations (such as the fluctuations in the last 5 hours) and the second LSTM layer captures long-term correlations (such as the trend of the previous day).
[0040] ④ Output layer: Outputs single-step prediction values through a fully connected layer.
[0041] This application also includes the following design for the aforementioned wind speed prediction model: User interaction and parameter input design: Develop input functions with range validation to support interactive input of the latitude and longitude of the lower left vertex of the task area, map size type (5°×5°, 4°×4°, 3°×3°, 2°×2°), and task time interval (start time, current time). Each type of parameter has upper and lower limit validation. If the range is exceeded, a prompt will be triggered and re-entry will be required.
[0042] S204. The intrinsic mode function (IMF) of the wind speed component is extracted and concatenated using the CNN module to obtain the IMF features of the wind speed component. S205. Using the location encoding module, the corresponding temporal location information is added to the IMF feature of the wind speed component to obtain the location-enhanced IMF feature of the wind speed component. S206. Input and output of the LSTM module: The input is a time sequence of the location-enhanced IMF feature extracted by a sliding window, and the output is the original wind speed component value at the first moment after the sliding window. S207. The wind speed prediction model is trained through the above steps.
[0043] Batch training of multi-grid point-multi-component models: Iterate through all valid grid points and the u, v, and w components of each point, and perform single-component model training: ① Initialize the model, MSE (mean squared error) loss function, and Adam optimizer (learning rate 2e-4); ② Build a data loader (batch size 32), perform 150 training rounds, and print the training / test loss every 30 rounds; ③ After training, save the optimal parameters for each model.
[0044] In summary, the core logic of training the EEMD-CNN-LSTM model is as follows: first, simplify the complex signal (EEMD decomposition), then extract features in layers (CNN + positional encoding), and finally model temporal dependencies (LSTM). Through batch training and precise monitoring, the model can learn the multi-scale wind field patterns of different grid points and different wind speed components, ultimately achieving high-precision, long-term wind speed prediction.
[0045] S102. Real-time acquisition and generation of the current state vector of the airship based on its current position, speed, attitude and wind load; Specifically, by integrating GPS positioning, inertial navigation, and wind speed sensor data from the stratospheric airship, its current position, velocity, attitude, and wind load are calculated in real time to generate a current state vector. This current state vector is 12-dimensional, including longitude, latitude, altitude, eastward velocity, northward velocity, vertical velocity, roll angle, pitch angle, yaw angle, eastward wind load, northward wind load, and vertical wind load.
[0046] S103. Based on the wind speed prediction sequence and the current state vector, a preset model predictive control algorithm is used to calculate the expected propulsion force and propeller speed adjustment required to offset future wind loads, with the goal of minimizing wind load offsetting error and optimizing energy consumption, under the condition of meeting the saturation constraints of the propulsion system. Model Predictive Control (MPC) is a model-based predictive control method that establishes a dynamic mathematical model of the system to predict the system state and output over a future period. It then optimizes the control variables in the current state to achieve optimal control of the control system. The MPC algorithm is the core of control decision-making; based on future wind field predictions and the real-time state of the airship, it plans the optimal control scheme in advance, solving the lag problem of traditional control while balancing dwell stability, energy consumption optimization, and system safety.
[0047] In this step, based on the wind speed prediction sequence and the current state vector, the MPC algorithm is used to calculate the expected thrust and propeller speed adjustment required to offset future wind loads, taking into account propulsion system saturation constraints and optimal energy consumption targets; a 12-dimensional real-time state vector is obtained (longitude, latitude, altitude, eastward speed, northward speed, vertical speed, roll angle, pitch angle, yaw angle, eastward wind load, northward wind load, and vertical wind load); the wind speed prediction sequence for the next 24 hours (time step of 1 hour, including eastward, northward, and vertical wind speed components) output by the wind speed prediction model is received, requiring the hardware to synchronize the data update frequency with the state vector.
[0048] Based on the Newton-Euler equations, a mathematical model for the airship's state changes is established, with the following formula logic: ; Meaning of each parameter: The current moment; This is the current state vector; For a moment Predicted The state at any given moment; For a moment Predicted The amount of control at any given moment; For prediction Wind load at any given moment; The state transition matrix describes the kinematic characteristics of the airship itself and is determined by physical parameters such as mass and moment of inertia. The control input matrix represents the influence coefficient of propulsion on the state. Let be the wind load disturbance matrix, which describes the relationship between wind load and state; Wind load Calculations based on wind speed prediction sequences: ; This formula serves as a classic aerodynamic drag formula. A variation of this formula is used to calculate the predicted wind loads on a stratospheric airship at future times. It calculates the wind loads at the current time. For the future The aerodynamic forces (wind loads) experienced by the airship at a given moment. It is calculated based on the airship's velocity relative to the air (relative airspeed).
[0049] It is the predicted wind load, that is, the wind load in the future. At any given moment, the aerodynamic forces (mainly drag) acting on the airship are the primary "disturbing forces" that the MPC controller attempts to counteract. The coefficient 0.5 in the equation is the standard constant coefficient in the hydrodynamic dynamic pressure formula. This refers to the air density at the airship's altitude. In the stratosphere (approximately 20 km altitude), the air density is extremely low (about 7% of that at sea level). This parameter is usually associated with altitude. The function, i.e. ; This refers to the characteristic area of an airship, typically its maximum windward cross-sectional area or aerodynamic reference surface area. The larger the airship, the larger its windward area, and the greater the wind load it experiences. It is the aerodynamic drag coefficient, a dimensionless coefficient characterizing the aerodynamic characteristics of an airship's shape. The better the streamlined the airship, the lower its drag coefficient. The smaller the value, the less wind resistance is experienced. This coefficient is usually determined by wind tunnel experiments or CFD simulations, and may vary with the angle of attack; It is the predicted airship velocity vector, representing the velocity at time [time]. The predicted future moment for the airship The ground velocity is taken from the state vector. The velocity component in the image. This is the airship's velocity programmed by the controller; It is the output of the wind speed prediction model, at a future time. The ambient wind speed is the only external input variable (disturbance term) in the formula. This refers to the airship's speed relative to the air. Aerodynamic forces depend on the airship's relative motion with the air, not its absolute speed relative to the ground. If the airship flies with the wind, its relative speed is low, and the force is small. If the airship flies against the wind (which is usually the case when hovering), its relative speed is high, and the force is large. This means that the aerodynamic force is proportional to the square of the relative velocity, which is a typical nonlinear term. This implies that a small increase in wind speed will lead to a sharp (square) increase in wind load, which is why stratospheric airship control requires "nonlinear compensation".
[0050] The saturation constraints of the propulsion system include: 1. Thrust constraint: The thrust is between the maximum and minimum thrust specified by the propulsion system. ; in, To determine the maximum thrust limit for the propulsion system.
[0051] 2. Rotational speed constraint: The maximum and minimum rotational speeds are obtained based on the linear relationship between rotational speed and propeller thrust, and the rotational speed is between the maximum and minimum values; The propeller speed and thrust are linearly related. ( This is a proportionality coefficient matrix. (where the rotational speed is the vector), therefore the rotational speed constraint is: ; in, , This is the maximum speed limit of the propeller.
[0052] 3. Control increment constraint: The speed adjustment at adjacent time points is between the maximum and minimum allowable speed change.
[0053] To avoid damage to the mechanical structure from sudden changes in rotational speed, the amount of speed adjustment between adjacent moments is limited: ; in, and The minimum and maximum allowable speed variations are determined by the dynamic characteristics of the propulsion system.
[0054] In this step, a weighted quadratic objective function is constructed with the dual objectives of minimizing wind load offsetting error and optimizing energy consumption: ; The objective function J is a quadratic programming problem, which aims to balance the airship's ability to "stay in the region" and "consume energy to offset wind load".
[0055] First item As a wind load offsetting error term, and as a task indicator for "regional residence", among which It is the wind load compensation error, the formula is: This is the difference between the predicted wind load and the reaction force generated by the airship's propulsion system. If this difference is 0, it means the airship has perfectly counteracted the wind force, is in force balance, and can remain stationary (staying put). It is an error weight matrix, representing the degree of importance attached to "position retention accuracy". The larger the airship, the more it "desperately tries to counteract the wind at this moment," even if it requires a huge energy cost. The goal is to minimize force imbalance. The smaller this value, the higher the airship's dwell accuracy and the lower the risk of being blown away by the wind.
[0056] Second item As an energy consumption item for a task, it is also an economic indicator. It is the force that the airship propulsion system will output in the future at the t+i time. This is the energy consumption weighting coefficient, representing the degree of importance attached to "energy saving". The larger the value, the more the airship will tend to output less force to conserve energy, even if this might cause the airship to deviate slightly from its position. The goal of this term is to minimize energy consumption. Since power is proportional to the square of the force (or the cube of the rotational speed), a quadratic form of the control variable is used. It can effectively characterize energy consumption costs.
[0057] The two time-domain parameters in the formula and These represent the time domain length for wind field prediction and the time domain length for planning future propulsion system control, respectively. .
[0058] In the prediction time domain Inside, with the objective function Minimize the objective function and solve the quadratic programming problem under constraints: ; The optimal control sequence is obtained by using the interior point method. (i.e., desired thrust), take the thrust control amount at the current moment. ,pass This is converted into a propeller speed adjustment.
[0059] S104. Based on the airship dynamics model, the desired thrust and propeller speed adjustment are corrected in real time to form control commands. In this step, considering the strong nonlinear characteristics of the airship, the control input is corrected in real time based on the dynamic model to compensate for control deviations caused by volume changes and aerodynamic parameter drift. Compensation is introduced to modify the mathematical model of the airship's state changes and the objective function. ; in, The compensation weighting matrix (balancing compensation intensity and energy consumption). It is the initial gas volume. V (t) is the gas volume at time t; It is the gas density at sea level; ρ air(h) This represents the atmospheric density at the airship's current flight altitude h. g is the acceleration due to gravity; It is the initial drag coefficient. C D (t) is the drag coefficient at time t; This is the initial buoyancy; F air0 Indicates the reference aerodynamic drag / reference aerodynamic load; This is the coefficient of volume change; It is the aerodynamic parameter drift coefficient; Δ f This represents the total nonlinear model deviation compensation amount; formula The first term is the buoyancy deviation caused by volume changes (thermal expansion and contraction / pressure difference), and the second term is the drag deviation caused by aerodynamic parameter drift.
[0060] S105. Send the control command to the propulsion system.
[0061] Specifically, the corrected control commands are sent to the propulsion system. At the same time, the actual output of the propulsion system and the position error of the airship can be collected to form a closed-loop feedback to update the predictive model parameters, ensuring the timeliness and accuracy of the control commands.
[0062] The stratospheric airship regional stationing control method provided in this application firstly achieves proactive adjustment of control actions through wind speed prediction, effectively solving the lag problem of traditional real-time feedback control. Simultaneously, its nonlinear compensation mechanism can accurately adapt to the airship's highly nonlinear dynamic characteristics, further enhancing control robustness under complex wind fields. This synergistic effect of the mechanism improves the airship's regional stationing success rate in complex stratospheric wind fields. Furthermore, prediction-based control decisions can optimize propulsion system energy consumption allocation in advance, avoiding redundant power consumption due to sudden wind speed changes, thus contributing to endurance. In addition, it enables the airship to adapt to complex meteorological conditions such as sudden wind speed changes, ensuring the continuity and stability of stationing missions.
[0063] In the above embodiments, a method for controlling the regional stay of a stratospheric airship is provided. Correspondingly, this application also provides a device for controlling the regional stay of a stratospheric airship, which can be implemented through software, hardware, or a combination of both. For example, the device may include integrated or separate functional modules or units to perform the corresponding steps in the above methods. Please refer to... Figure 5 This illustration shows a schematic diagram of a stratospheric airship regional stay control device provided by some embodiments of this application. Since the device embodiments are basically similar to the method embodiments, the description is relatively simple; relevant details can be found in the description of the method embodiments. The device embodiments described below are merely illustrative.
[0064] like Figure 5 As shown, the stratospheric airship regional stay control device 10 provided in this application may include: The wind speed prediction module 101 is used to predict the wind speed changes within a preset time period of the current location and a certain range around the airship in the target area based on real-time wind field data of the current location of the airship, and output a wind speed prediction sequence. The wind speed prediction model includes an Empirical Mode Decomposition (EEMD) module, a Convolutional Neural Network (CNN) module, a Long Short-Term Memory (LSTM) network module, and a location encoding module. The wind speed prediction sequence includes eastward, northward, and vertical wind speed components. The state estimation module 102 is used to acquire and generate the current state vector of the airship in real time based on the airship's current position, speed, attitude and wind load; The control decision module 103 is used to calculate the expected propulsion force and propeller speed adjustment required to offset future wind loads based on the wind speed prediction sequence and the current state vector, using a preset model prediction control algorithm, under the condition of meeting the saturation constraints of the propulsion system, with the goal of minimizing wind load offsetting error and optimizing energy consumption. The nonlinear compensation module 104 is used to correct the desired thrust and propeller speed adjustment in real time based on the airship dynamics model, and to generate control commands. The execution module 105 is used to send the control commands to the propulsion system.
[0065] In one possible implementation, the propulsion system saturation constraints include: Thrust constraint: The thrust is between the maximum and minimum thrust specified by the propeller; Speed constraint: The maximum and minimum speed values are obtained based on the linear relationship between speed and propeller thrust, and the speed is between the maximum and minimum values; Control increment constraint: The speed adjustment at adjacent time points is between the maximum and minimum allowable speed change.
[0066] In one possible implementation, the objective function is constructed with the goals of minimizing wind load offsetting error and optimizing energy consumption as follows: ; in, The time domain length for wind field prediction. To extend the time domain length of the system control; To compensate for wind load error, This represents the predicted wind load. This represents the reaction force generated by the airship's propulsion system. This is the error weight matrix; This represents the planned force output of the airship propulsion system at time t+i. This is the energy consumption weighting coefficient.
[0067] The stratospheric airship regional stationing control device provided in this application firstly achieves proactive adjustment of control actions through wind speed prediction, effectively solving the lag problem of traditional real-time feedback control. Simultaneously, its nonlinear compensation mechanism can accurately adapt to the airship's highly nonlinear dynamic characteristics, further enhancing control robustness under complex wind fields. This synergistic effect of the mechanism improves the airship's regional stationing success rate in complex stratospheric wind fields. Furthermore, prediction-based control decisions can optimize propulsion system energy consumption allocation in advance, avoiding redundant power consumption due to sudden wind speed changes, thus contributing to endurance. In addition, it enables the airship to adapt to complex weather conditions such as sudden wind speed changes, ensuring the continuity and stability of stationing missions.
[0068] This application also provides a control device corresponding to the stratospheric airship regional stay control method provided in the foregoing embodiments. The control device may be a mobile phone, laptop computer, tablet computer, desktop computer, etc., to execute the above-mentioned stratospheric airship regional stay control method.
[0069] The control device provided in this application embodiment and the stratospheric airship regional dwelling control method provided in this application embodiment are based on the same inventive concept and have the same beneficial effects as the methods they adopt, operate or implement.
[0070] This application also provides a computer-readable storage medium corresponding to the stratospheric airship regional dwelling control method provided in the foregoing embodiments, wherein a computer program (i.e., a program product) is stored thereon. When the computer program is run by a processor, it executes the stratospheric airship regional dwelling control method provided in any of the foregoing embodiments.
[0071] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here.
[0072] The computer-readable storage medium provided in the above embodiments of this application and the stratospheric airship regional dwelling control method provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.
[0073] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.
[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and they should all be covered within the scope of the claims and specification of this application.
Claims
1. A method for controlling the regional stay of a stratospheric airship, characterized in that, include: Based on real-time wind field data of the current location of the airship in the target area, a wind speed prediction model is used to predict the wind speed changes of the current location and a certain range around it within a preset time period, and outputs a wind speed prediction sequence. The wind speed prediction model includes an Empirical Mode Decomposition (EEMD) module, a Convolutional Neural Network (CNN) module, a Long Short-Term Memory (LSTM) network module, and a location encoding module. The wind speed prediction sequence includes eastward, northward, and vertical wind speed components. The current state vector of the airship is generated in real time based on its current position, speed, attitude and wind load. Based on the wind speed prediction sequence and the current state vector, a preset model predictive control algorithm is used to calculate the expected thrust and propeller speed adjustment required to offset future wind loads, with the goal of minimizing wind load offsetting error and optimizing energy consumption, under the condition of meeting the saturation constraints of the propulsion system. Based on the airship dynamics model, the desired thrust and propeller speed adjustment are corrected in real time to generate control commands. The control commands are sent to the propulsion system.
2. The method for controlling the regional stay of a stratospheric airship according to claim 1, characterized in that, The training process of the wind speed prediction model includes: The historical wind field data of the target area within a historical time period is obtained. The historical wind field data includes basic information corresponding to each spatial location point in the target area and wind speed components in the east, north and vertical directions. The basic information includes longitude, latitude, altitude and recording time. Obtain wind field data for a spatial location point and its eight surrounding neighboring points from the historical wind field data; For each acquired spatial location point, the EEMD module is used to decompose a wind speed component at that spatial location point to obtain the intrinsic mode function (IMF) of the wind speed component. The intrinsic mode function (IMF) of the wind speed component is extracted and concatenated using the CNN module to obtain the IMF features of the wind speed component. The location encoding module is used to add the corresponding temporal location information to the IMF feature of the wind speed component to obtain the location-enhanced IMF feature of the wind speed component. The input and output of the LSTM module are as follows: the time sequence of the enhanced IMF feature at a single location is extracted by a sliding window as the input, and the original wind speed component value at the first time after the sliding window is output as a single value. The wind speed prediction model is trained using the steps described above.
3. The method for controlling the regional stay of a stratospheric airship according to claim 1 or 2, characterized in that, In the wind speed prediction model: The CNN module consists of two 1D convolutional layers, the LSTM module consists of two LSTM layers, and the position encoding module uses trigonometric function position encoding.
4. The method for controlling the regional stay of a stratospheric airship according to claim 1, characterized in that, The current state vector is 12-dimensional, including longitude, latitude, altitude, eastward velocity, northward velocity, vertical velocity, roll angle, pitch angle, yaw angle, eastward wind load, northward wind load, and vertical wind load.
5. The method for controlling the regional stay of a stratospheric airship according to claim 1, characterized in that, The saturation constraints of the propulsion system include: Thrust constraint: The thrust is between the maximum and minimum thrust specified by the propulsion system; Speed constraint: The maximum and minimum speed values are obtained based on the linear relationship between speed and propeller thrust, and the speed is between the maximum and minimum values; Control increment constraint: The speed adjustment at adjacent time points is between the maximum and minimum allowable speed change.
6. The method for controlling the regional stay of a stratospheric airship according to claim 1, characterized in that, The objective function is constructed with the goals of minimizing wind load compensation error and optimizing energy consumption as follows: ； in, The time domain length for wind field prediction. To extend the time domain length of the system control; To compensate for wind load error, This represents the predicted wind load. This represents the reaction force generated by the airship's propulsion system. This is the error weight matrix; This represents the planned force output of the airship propulsion system at time t+i. This is the energy consumption weighting coefficient.
7. A stratospheric airship regional stay control device, characterized in that, include: The wind speed prediction module is used to predict wind speed changes within a preset time period and its surrounding area within the target region based on real-time wind field data of the airship's current location, and outputs a wind speed prediction sequence. The wind speed prediction model includes an Empirical Mode Decomposition (EEMD) module, a Convolutional Neural Network (CNN) module, a Long Short-Term Memory (LSTM) network module, and a location encoding module. The wind speed prediction sequence includes eastward, northward, and vertical wind speed components. The state estimation module is used to acquire and generate the current state vector of the airship in real time based on its current position, speed, attitude and wind load. The control decision module is used to calculate the expected thrust and propeller speed adjustment required to offset future wind loads based on the wind speed prediction sequence and the current state vector, using a preset model predictive control algorithm, under the condition of meeting the saturation constraints of the propulsion system, with the goal of minimizing wind load offsetting error and optimizing energy consumption. The nonlinear compensation module is used to correct the desired thrust and propeller speed adjustment in real time based on the airship dynamics model, and generate control commands. An execution module is used to send the control commands to the propulsion system.
8. The stratospheric airship regional stay control device according to claim 7, characterized in that, The saturation constraints of the propulsion system include: Thrust constraint: The thrust is between the maximum and minimum thrust specified by the propeller; Speed constraint: The maximum and minimum speed values are obtained based on the linear relationship between speed and propeller thrust, and the speed is between the maximum and minimum values; Control increment constraint: The speed adjustment at adjacent time points is between the maximum and minimum allowable speed change.
9. A control device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor, when executing the computer program, implements the method as described in any one of claims 1 to 6.
10. A computer-readable storage medium, characterized in that, It stores computer-readable instructions that can be executed by a processor to implement the method as described in any one of claims 1 to 6.