A coastal airport intelligent ecological regulation and control system based on digital twinning and multi-objective optimization

Abstract

The application discloses a coastal airport intelligent ecological regulation system based on digital twinning and multi-objective optimization, belongs to the technical field of intelligent airport operation management and environmental safety control, and comprises a state reconstruction module, an original real-time data set of a physical sensor and an environmental prediction state generated by a digital twinning model are acquired, and a fused environmental state is generated by performing weighted fusion based on a deviation measurement between the data set and the environmental prediction state; a risk assessment module, based on the fused environmental state, quantifies a predicted flight safety risk and an ecological resilience risk, and determines dynamic weights corresponding to the two risks respectively according to a preset rule to obtain a weighted risk assessment result; and a collaborative decision optimization module, based on the weighted risk assessment result, solves a multi-objective optimization problem with a risk weighted sum as an objective function to generate optimal regulation instructions, and the application ensures all-weather and high-reliability environmental state acquisition.

Description

Technical Field

[0001] This invention relates to the field of smart airport operation management and environmental safety control, specifically a smart ecological regulation system for coastal airports based on digital twins and multi-objective optimization. Background Technology

[0002] In the operation and management of coastal airports, traditional technological approaches mainly rely on physical sensor networks deployed throughout the airport area, such as meteorological and bird activity monitoring equipment, and established emergency response rules to ensure flight safety and fulfill environmental responsibilities. However, this model has many limitations when dealing with the unique, complex, and dynamic environment of coastal areas, and it is difficult to meet the requirements of modern airports for precision, intelligence, and sustainable development.

[0003] The aforementioned situation and shortcomings mainly stem from the following core challenges:

[0004] The unreliability of data perception is a problem. Traditional management systems rely heavily on direct readings from physical sensors. However, in coastal environments, equipment is subjected to harsh conditions such as salt spray corrosion and strong electromagnetic interference, which often leads to a decline in sensor data quality, instability, or even failure. This results in serious reliability issues with the data sources upon which the system relies for decision-making, making it impossible to form a comprehensive and reliable environmental situational awareness.

[0005] The inherent conflict of decision-making objectives means that airport operations naturally need to balance two conflicting goals: immediate flight safety and long-term ecological protection. For example, bird control operations ensure flight safety but may have a negative impact on ecosystems such as bird habitats. Traditional decision-making logic based on static rules is difficult to find the optimal balance between these two objectives in real time and dynamically in complex and ever-changing scenarios.

[0006] The static rigidity of system models presents a common technical challenge for advanced systems that incorporate digital models: model drift. That is, the physical world environment is constantly changing, and if the digital twin model does not have adaptive capabilities, its internal parameters and prediction logic will gradually become detached from reality, leading to model rigidity and failure, a continuous decline in prediction accuracy, and ultimately affecting the effectiveness of the entire system.

[0007] When faced with emergencies, such as flocks of birds gathering or severe weather, managers not only struggle to obtain stable and reliable environmental information, but also cannot balance the conflict between safety and ecology based on scientific optimization strategies. Their decisions are often passive and suboptimal. More importantly, the system lacks a mechanism for continuous learning and self-improvement during operation, which makes it impossible to guarantee the effectiveness and accuracy of its long-term operation, ultimately hindering the transformation of coastal airport operations towards intelligence, refinement, and sustainability.

[0008] The information disclosed in the background section above is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0009] The purpose of this invention is to provide a smart ecological control system for coastal airports based on digital twins and multi-objective optimization, so as to solve the problems mentioned in the background art.

[0010] The technical solution of the present invention includes:

[0011] The state reconstruction module acquires the raw real-time dataset from the physical sensors and the predicted environmental state generated by the digital twin model. It then performs weighted fusion based on the deviation metric between the dataset and the predicted environmental state to generate a fused environmental state.

[0012] The risk assessment module, based on the integrated environmental conditions, quantifies and predicts flight safety risks and ecological resilience risks, and determines the dynamic weights corresponding to the two risks according to preset rules to obtain the weighted risk assessment results.

[0013] The collaborative decision-making optimization module, based on the weighted risk assessment results, solves a multi-objective optimization problem with the risk weighted sum as the objective function to generate the optimal control instruction;

[0014] The model evolution module updates the internal parameter vector of the digital twin model using an online gradient descent algorithm based on the deviation metric between the fused environmental state and the predicted environmental state, thus obtaining the evolved model.

[0015] Preferably, the step by which the state reconstruction module generates the fusion environment state is as follows:

[0016] Calculate the deviation measure between the original real-time dataset and the predicted environmental state;

[0017] Based on the deviation metric, the dynamic fusion coefficient is determined using a preset exponential decay rule;

[0018] A weighted average of the original real-time dataset and the predicted environmental state is generated by using dynamic fusion coefficients.

[0019] Preferably, the logic for determining the dynamic fusion coefficient is as follows:

[0020] As the bias metric increases, the dynamic fusion coefficient decays exponentially and approaches 0, thereby increasing the confidence weight for the predicted state of the environment.

[0021] As the bias metric decreases and approaches 0, the dynamic fusion coefficient approaches 1 to increase the trust weight of the original real-time dataset.

[0022] Preferably, the steps for determining the dynamic weights are as follows:

[0023] The dynamic weights of flight safety risks are correlated with key flight safety indicators;

[0024] When key flight safety indicators fall below the preset safety threshold, the dynamic weight of flight safety risks is increased.

[0025] When key flight safety indicators are better than or equal to the safety threshold, the dynamic weight of reducing flight safety risks is applied.

[0026] Preferably, the key flight safety indicator is the expected collision time, and the safety threshold is a safe time threshold determined according to civil aviation safety regulations and aircraft performance.

[0027] Preferably, the steps for generating the optimal control command are as follows:

[0028] An optimization objective function is constructed with the weighted sum of flight safety risks and ecological resilience risks as its content, and the weights are dynamic weights determined by the risk assessment module.

[0029] The optimal control command is obtained by solving for the minimum value of the objective function.

[0030] Preferably, the steps for obtaining the evolved model are as follows:

[0031] The deviation between the fused environmental state and the predicted environmental state is defined as the loss function.

[0032] Calculate the gradient of the loss function with respect to the internal parameter vector of the digital twin model;

[0033] The internal parameter vector is adjusted based on the gradient and evolution rate coefficients to generate the evolved model.

[0034] Preferably, the adjustment logic for the evolution rate coefficient is as follows:

[0035] When the continuous-time change rate of the fused environment state exceeds the preset environmental change threshold, the evolution rate coefficient is increased to accelerate model adaptation.

[0036] When the continuous-time rate of change of the fused environment state is less than or equal to the environmental change threshold, the evolution rate coefficient maintains the baseline value.

[0037] Preferably, the optimal control command is a decision vector, and each component of the decision vector corresponds to a parameter of a controllable physical device at an airport.

[0038] This invention provides an improved intelligent ecological control system for coastal airports based on digital twins and multi-objective optimization, which has the following improvements and advantages compared with the prior art:

[0039] 1. This invention enables the system to accurately capture instantaneous changes when the sensor is functioning normally, and to smoothly transition to model-based prediction when the sensor malfunctions, thereby ensuring all-weather, highly reliable acquisition of environmental conditions;

[0040] 2. The collaborative decision optimization module of this invention transforms this problem into a solvable multi-objective optimization problem, making the system decision no longer rigid, but possessing the ability to dynamically adjust the strategy focus according to the urgency of the situation, similar to human experts. Moreover, its decision-making process is mathematically provable to be optimal. This self-learning and adaptive closed-loop characteristic enables the evolved model to persistently and accurately map the dynamics of the physical world, ensuring that the foundation of the entire intelligent control system's decision-making quality is solid and reliable. Attached Figure Description

[0041] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0042] Figure 1 This is a flowchart of a smart ecological control system for coastal airports based on digital twins and multi-objective optimization, according to the present invention. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0044] Example 1:

[0045] Please see Figure 1 This invention provides a smart ecological control system for coastal airports based on digital twins and multi-objective optimization, comprising:

[0046] The state reconstruction module acquires the raw real-time dataset from the physical sensors and the predicted environmental state generated by the digital twin model. It then performs weighted fusion based on the deviation metric between the dataset and the predicted environmental state to generate a fused environmental state.

[0047] The risk assessment module, based on the integrated environmental conditions, quantifies and predicts flight safety risks and ecological resilience risks, and determines the dynamic weights corresponding to the two risks according to preset rules to obtain the weighted risk assessment results.

[0048] The collaborative decision-making optimization module, based on the weighted risk assessment results, solves a multi-objective optimization problem with the risk weighted sum as the objective function to generate the optimal control instruction;

[0049] The model evolution module updates the internal parameter vector of the digital twin model using an online gradient descent algorithm based on the deviation measure between the fused environmental state and the predicted environmental state, thus obtaining the evolved model.

[0050] This embodiment provides a smart ecological control system for coastal airports based on digital twins and multi-objective optimization;

[0051] During system operation, the state reconstruction module aims to accurately generate the environmental state. Its specific purpose is to generate a high-confidence fused environmental state when the data quality of physical sensors deployed at the coastal airport deteriorates due to salt spray corrosion or signal interference. This is achieved by fusing sparse physical sensor data with prior prediction data from the digital twin model. In this embodiment, the module acquires two types of data: one is the raw real-time dataset collected by physical sensors such as airport meteorological and bird activity monitoring systems, which is represented as a multi-dimensional vector; the other is the predicted environmental state derived by the system's built-in digital twin model based on the previous state, also represented as a vector with the same dimensions as the aforementioned dataset. The module performs weighted fusion based on a deviation metric between these two types of data. This deviation metric is a scalar value that quantifies the degree of inconsistency between physical observations and model predictions, thereby generating the fused environmental state. The fused environmental state is a multi-dimensional state vector, providing a unified and reliable data foundation for subsequent risk assessment and decision-making.

[0052] Based on the integrated environmental conditions, the risk assessment module performs quantitative risk prediction. The specific purpose of the risk assessment module is to transform complex environmental conditions into an intuitive measure of their impact on core business, flight safety, long-term goals, and ecological sustainability. In this embodiment, the module has two built-in risk prediction models, which are used to quantify flight safety risks and ecological resilience risks, respectively. Flight safety risks refer to the possibility of direct threats to aircraft operation safety, such as bird strikes and severe weather. Ecological resilience risks refer to the potential negative impact of airport operations on the long-term health of surrounding wetlands, bird habitats, and other ecosystems. The module not only calculates risk values ​​but also determines dynamic weights corresponding to the two types of risks according to a set of preset rules. The weights reflect the system's emphasis on safety and ecological protection in a specific environment and output a weighted risk assessment result.

[0053] To generate the optimal control strategy, the collaborative decision optimization module generates the optimal control command based on the weighted risk assessment results. The specific purpose of the collaborative decision optimization module is to automatically generate an optimal balance strategy that can simultaneously take into account flight safety and ecological resilience within a very short decision window. In this embodiment, the module constructs the problem as a multi-objective optimization problem. The objective function to be solved is based on the weighted sum of two risks. The goal is to find a set of control actions that minimizes the weighted sum of risks. After solving the optimization problem, the module generates the optimal control command. The command is a specific decision vector, and its components directly correspond to the parameters of controllable physical equipment in the airport, such as the power of bird deterrent equipment or the frequency of navigation lights.

[0054] To ensure the long-term effectiveness of the system model, the model evolution module works in parallel to achieve adaptive updates of the digital twin model. The specific purpose of the model evolution module is to ensure that the digital twin model can continuously track the dynamic changes of the real physical environment and avoid model stagnation and failure. In this embodiment, the module uses the deviation metric between the fused environment state and the predicted environment state calculated during the state reconstruction process as the core input, and uses an online gradient descent algorithm to iteratively update the internal parameter vector of the digital twin model. In this way, the model can learn from its own prediction error, gradually improve the prediction accuracy, and finally obtain an evolved model. The evolved model will be used in the next time step to generate a more accurate predicted environment state.

[0055] This system, through the collaborative work of four modules—state reconstruction, risk assessment, collaborative decision-making, and model evolution—forms a complete technological closed loop of perception, cognition, decision-making, and learning. It not only overcomes the challenge of unstable physical sensor data quality in coastal environments, generating reliable environmental situational awareness, but also dynamically balances the conflicting goals of flight safety and ecological protection based on real-time risk, automatically solving for the optimal control commands. More importantly, through the model evolution mechanism, the system possesses continuous learning and adaptive capabilities, ensuring its long-term operational effectiveness and accuracy, ultimately achieving intelligent, refined, and sustainable operation of coastal airports.

[0056] The steps used by the state reconstruction module to generate the state of the fusion environment are as follows:

[0057] Calculate the deviation measure between the original real-time dataset and the predicted environmental state;

[0058] Based on the deviation metric, the dynamic fusion coefficient is determined using a preset exponential decay rule;

[0059] A weighted average of the original real-time dataset and the predicted environmental state is generated by using dynamic fusion coefficients.

[0060] The logic for determining the dynamic fusion coefficient is as follows:

[0061] As the bias metric increases, the dynamic fusion coefficient decays exponentially and approaches 0, thereby increasing the confidence weight for the predicted state of the environment.

[0062] As the bias metric decreases and approaches 0, the dynamic fusion coefficient approaches 1 to increase the trust weight of the original real-time dataset.

[0063] In this embodiment, the steps for the state reconstruction module to generate the fusion environment state are defined;

[0064] Within the state reconstruction module, the underlying logic lies in calculating the deviation metric between the original real-time dataset and the predicted environmental state; the original real-time dataset... This refers to a vector, where each component represents the observation value of various physical sensors within the airport at the current moment, provided by the sensor subsystem; environmental prediction state. It refers to and Vectors with the same physical dimensions and data structure, predicted by the digital twin model based on the state at the previous time step for the current time step, are provided by the model evolution module; deviation metric. It is a scalar function used to quantify the difference between the two vectors, and its role is to evaluate the reliability of physical observations; in this embodiment, the Mahalanobis distance can be used as the metric function because it can effectively take into account the correlation between the state variables.

[0065] Based on the deviation metric, a pre-defined exponential decay rule is used to determine the dynamic fusion coefficients. The principle behind this rule is that for sensor data with a greater deviation from the model's prediction, the confidence level should decrease exponentially. The dynamic fusion coefficients... It is a dimensionless scalar with a value range between [0,1], and its function is to dynamically allocate the proportion of physical observation data and model prediction data in the final fused state; its determination logic is realized through the following formula:

[0066] Dynamic fusion coefficients, original real-time dataset, environment prediction state deviation metric function, sensitivity parameters:

[0067]

[0068] in, Dynamic fusion coefficient Raw real-time dataset, Environmental prediction status Deviation measurement function Sensitivity parameters; Exponential function;

[0069] Sensitivity parameters here It is a preset normal number; the method of determination is: based on the validation set containing historical fault data, offline optimization is performed with the goal of minimizing the fusion error, and thus it is calibrated.

[0070] For example, the validation set could contain 10,000 data points, each of which includes a timestamp and a raw, real-time dataset collected by physical sensors. An environmental prediction state generated by a digital twin model And a real-world environmental state obtained through high-precision, interference-resistant reference-level sensors or post-hoc manual verification. The validation set of fault data should cover at least three typical sensor fault modes: signal drift, complete data loss, and impulse noise exceeding 5 standard deviations; the optimization objective is to find a... Values ​​such that, over the entire validation set, the formula is:

[0071]

[0072] Calculated fusion environment state Compared to the real environment The mean square error between them is minimized; Dynamic fusion coefficient; : The original real-time dataset; Environmental prediction status;

[0073] Physical quantities are deviation measurement functions The reciprocal of the dimensions, to ensure that the exponent term is dimensionless; the specific manifestation of the logic is: when the deviation is measured... An increase in the dynamic fusion coefficient indicates a significant discrepancy between the physical sensor data and the model predictions, potentially indicating sensor malfunction or severe interference. The system exhibits exponential decay and approaches zero, increasing the system's ability to predict environmental conditions. Trust weights; conversely, when bias measures When the value decreases and approaches 0, it indicates a high degree of consistency between the physical observations and model predictions, and the dynamic fusion coefficient... As the value approaches 1, the system increases its input to the original real-time dataset. Trust weight;

[0074] A weighted average of the original real-time dataset and the predicted environmental state is generated using dynamic fusion coefficients. The process is accomplished using the following formula based on the complementary filtering principle:

[0075] Fusion Environment State Dynamic Fusion Coefficient Original Real-Time Dataset Environment Prediction State:

[0076]

[0077] in, Dynamic fusion coefficient Raw real-time dataset, Environmental prediction status;

[0078] The fusion environment state calculated by the formula It combines real-time observations of the physical world with prior knowledge from digital models, providing optimal estimates for subsequent processing;

[0079] Through the above implementation methods, this system realizes an intelligent state fusion mechanism; instead of using fixed weights for data fusion, it dynamically adjusts the fusion strategy based on the real-time reliability of sensor data; this exponentially decaying reliability allocation logic can effectively suppress the impact of abnormal data caused by sensor momentary failures or strong environmental interference on the overall situational awareness of the system, while making full use of the real information provided by the sensors when they are working normally, thereby significantly improving the accuracy, robustness and reliability of the fused environmental state.

[0080] The system exhibits unprecedented robustness and high fidelity at the environmental perception level. Traditional airport monitoring systems heavily rely on direct readings from physical sensors; in harsh environments such as coastal salt spray corrosion and strong electromagnetic interference, data quality degradation can lead to severe system performance degradation or even failure. The state reconstruction module of this invention solves this problem; instead of simply filtering or averaging data, it establishes a dynamic fusion mechanism based on confidence assessment. The core of this mechanism can be derived from the state fusion formula. The physical meaning described here refers to the generated fusion environment state. It is a raw, real-time dataset full of uncertainty. Predicting the state of the environment with prior knowledge The optimal weighting; the determination of the weighting coefficients, i.e., the dynamic fusion coefficients. The underlying logic is: when the deviation between physical observations and model predictions is measured... When it approaches zero, the coefficient When the deviation approaches 1, the system trusts physical measurements; as the deviation increases, The degradation is exponential, and the system relies more on the inference of the internal model. This design enables the system to accurately capture instantaneous changes when the sensor is normal, and to smoothly transition to relying on model prediction when the sensor is abnormal, thereby ensuring all-weather, high-reliability acquisition of environmental conditions.

[0081] Example 2

[0082] The steps for determining dynamic weights are as follows:

[0083] The dynamic weights of flight safety risks are correlated with key flight safety indicators;

[0084] When key flight safety indicators fall below the preset safety threshold, the dynamic weight of flight safety risks is increased.

[0085] When key flight safety indicators are better than or equal to the safety threshold, reduce the dynamic weight of flight safety risks;

[0086] The key flight safety indicator is the expected collision time, and the safety threshold is a safe time threshold determined according to civil aviation safety regulations and aircraft performance.

[0087] In this embodiment, the steps for determining the dynamic weights in the risk assessment module are specified.

[0088] In the risk assessment module, the determination logic of dynamic weights is directly linked to a key flight safety indicator. This key flight safety indicator is a quantitative parameter that most directly and sensitively reflects the immediate risk to flight safety. The role of dynamic weights is to adjust the level of attention to flight safety risks based on the urgency of the safety situation during decision optimization. In this embodiment, the key flight safety indicator is determined as the estimated collision time. This refers to the time required for the spatial positions of birds or drones to overlap with the flight paths of aircraft, based on the current flight paths of birds or drones. The indicator is derived from radar, optical and other sensor data, combined with the aircraft's flight plan, and calculated by a specialized trajectory prediction algorithm.

[0089] The dynamic weight adjustment follows a set of preset rules based on a safety threshold; the core of the rules is to adjust the expected collision time. With safety threshold Comparison; safety threshold This refers to a preset time length, which defines a safe buffer period during which a decision-making system must react; the threshold is determined by the provisions of the Civil Aviation Safety Regulations and the performance parameters of a specific aircraft, such as climb rate and turning radius, to ensure that the system's decision-making is consistent with industry safety standards.

[0090] The steps for weight adjustment are as follows: when the estimated collision time... Inferior, meaning less than the preset safety threshold When the system determines that there is an imminent collision risk, it will proactively and significantly increase the dynamic weight of flight safety risks. In this embodiment, Designed to be a The relevant sigmoid function, when hour, The value will increase rapidly and approach 1;

[0091] The S-shaped function can be calculated using the following logistic function form:

[0092]

[0093] in, To estimate the collision time, For a safe time threshold, It is a positive slope parameter used to control the drasticness of weight changes; The value can be determined based on simulation experiments, for example, set to . To ensure that around Small fluctuations It can quickly jump from a value close to 0 to a value close to 1, thereby enabling a rapid shift in the focus of decision-making; The base of the natural logarithm;

[0094] Conversely, when the estimated collision time Better than or equal to the safety threshold When the system determines that the current safety situation is within an acceptable range, it will reduce the dynamic weight of flight safety risks. This brings it close to 0, thus allowing the decision-making system to allocate more resources and attention to ecological resilience risks;

[0095] The implementation method directly links abstract risk weight adjustments with specific, quantifiable flight safety indicators, making the risk assessment process more objective, transparent, and efficient. By introducing the key indicator of expected collision time and a clear safety threshold, the system can automatically and quickly shift its decision focus to ensuring flight safety when the safety situation deteriorates rapidly, achieving mandatory protection for the highest priority tasks. This mechanism ensures that the system's decision-making behavior complies with the strict requirements of aviation safety regulations, greatly improving the system's response speed and decision-making effectiveness in dealing with sudden safety threats.

[0096] Example 3

[0097] The steps for generating the optimal control command are as follows:

[0098] An optimization objective function is constructed with the weighted sum of flight safety risks and ecological resilience risks as its content, and the weights are dynamic weights determined by the risk assessment module.

[0099] The optimal control command is obtained by solving for the minimum value of the objective function.

[0100] The optimal control command is a decision vector, and each component of the decision vector corresponds to a parameter of a controllable physical device at an airport.

[0101] In this embodiment, the steps of the collaborative decision-making optimization module in generating the optimal control instruction are further explained;

[0102] The process of generating optimal control commands is structured as solving a multi-objective optimization problem; to achieve this objective, a weighted objective function is constructed, which includes the weighted sum of flight safety risk and ecological resilience risk; the objective function... Its function is to provide a quantitative evaluation of each possible regulatory instruction, i.e., the total risk cost; its construction method originates from the multi-objective decision theory in operations research, and is expressed as follows:

[0103] Objective function, decision vector, dynamic weights, risk function, and fusion of environmental states:

[0104]

[0105] in, Objective function : Decision vector, Dynamic weights Risk function :Integration of environmental conditions; : Represents an optimization process, namely, finding the objective function that satisfies the objective function. To achieve the minimum optimal decision vector ; The objective function provides a quantitative evaluation of each possible regulatory instruction, i.e., the total risk cost. The dynamic weights of flight safety risks are determined by the risk assessment module. Flight safety risk function, a predictive model used to predict the execution of control commands. The resulting normalized flight safety risk index; The dynamic weights of ecological resilience risk are determined by the risk assessment module, and their sum is... It is usually constrained to 1; Ecological resilience risk function, a predictive model used to predict the execution of regulatory directives. The resulting normalized ecological resilience risk index;

[0106] In the function, the decision vector It is the variable for optimization, and it is a column vector.

[0107]

[0108] Each component of the decision vector This corresponds to the specific operating parameters of controllable physical equipment within an airport; for example... It can be the output power of acoustic bird deterrent devices in a specific area. These can be the flashing frequency or duration of airport navigation lights; the range of values ​​for these components is determined by the physical properties of the equipment.

[0109] Risk function in function and These are two specific, computable predictive models, such as neural networks or regression models trained on historical data; they receive decision vectors. and the current fusion environment state vector As input; where Provided by the state reconstruction module, while the decision vector... As independent variables to be optimized, they are generated by the optimization algorithm in each iteration; their role is to predict the execution of control commands. Subsequently, the normalized flight safety risk index and ecological resilience risk index will be generated;

[0110] Flight safety risk function It can be implemented as a three-layer feedforward neural network; the input layer of the network consists of decision vectors. With the fusion environment state vector The network is composed of key safety components, such as bird density, wind shear index, and visibility; it contains two hidden layers, each with 64 neurons, using ReLU as the activation function; the output layer is a single neuron using the Sigmoid activation function to output a normalized risk value between [0,1]; the training dataset for the model is generated based on historical aviation safety reports and aerodynamic simulations, containing at least 100,000 pairs of control instructions-environmental states-risk consequences; and an ecological resilience risk function. A similar structure can be used to construct it, but its input environmental state components focus on ecological indicators, such as noise decibels and water pollutant concentration predictions, while the training data comes from ecological impact assessment reports and long-term environmental monitoring data.

[0111] To ensure dimensional consistency, the outputs of both risk functions are normalized to dimensionless values.

[0112] Dynamic weights in a function and These are dimensionless weights determined by the risk assessment module according to the aforementioned rules, and their sum is usually constrained to 1;

[0113] The optimal control command is generated by solving for the minimum value of the above-mentioned objective function; the collaborative decision-making optimization module uses an efficient optimization algorithm, such as particle swarm optimization or genetic algorithm, to find the objective function. Minimize the optimal decision vector This optimal decision vector The final optimal control command is then issued to the airport's physical equipment for execution.

[0114] The implementation method transforms the complex decision-making problem into a well-structured and solvable mathematical optimization problem. By parameterizing control commands into decision vectors and formalizing the system objective into a risk-weighted objective function, the scientific nature and optimality of the decision-making process are guaranteed. It enables the system to systematically find the optimal operating point that balances flight safety and ecological protection with the minimum comprehensive cost in a high-dimensional decision space, rather than relying on static rules or human experience. This makes the generated control commands not only executable but also theoretically optimal solutions under the current environment, thereby achieving refined and optimized control of airport operations.

[0115] The system achieves a leap from passive response to proactive optimization at the decision-making level. Existing technologies, when dealing with conflicting objectives such as flight safety and ecological protection, often employ static, rule-based logic, making it difficult to find the optimal balance in complex and ever-changing scenarios. The collaborative decision-making optimization module of this invention transforms the problem into a solvable multi-objective optimization problem; its objective function is: This profoundly reflects the scientific nature of its decision-making; the optimal control command here. It is no longer a set of preset actions, but rather actions based on the current environmental state. This can reduce flight safety risks. Ecological resilience risk weighted sum To achieve the minimized solution; particularly crucial is the weighting. and It is a dynamic weight associated with key flight safety indicators, such as the estimated time of collision, when the safety risk exceeds a safety threshold. The impact will increase dramatically, forcing the system to prioritize flight safety. This mechanism makes the system's decision-making no longer rigid, but rather gives it the ability to dynamically adjust its strategic focus based on the urgency of the situation, similar to human experts, and its decision-making process is mathematically provable to be optimal.

[0116] Example 4

[0117] The steps to obtain the evolved model are as follows:

[0118] The deviation between the fused environmental state and the predicted environmental state is defined as the loss function.

[0119] Calculate the gradient of the loss function with respect to the internal parameter vector of the digital twin model;

[0120] The internal parameter vector is adjusted based on the gradient and evolution rate coefficients to generate the evolved model;

[0121] The adjustment logic for the evolution rate coefficient is as follows:

[0122] When the continuous-time change rate of the fused environment state exceeds the preset environmental change threshold, the evolution rate coefficient is increased to accelerate model adaptation.

[0123] When the continuous-time rate of change of the fused environment state is less than or equal to the environmental change threshold, the evolution rate coefficient maintains the baseline value.

[0124] In this embodiment, the steps used by the model evolution module to obtain the evolved model are limited;

[0125] The core of the model evolution process is the use of an online gradient descent algorithm, enabling the digital twin model to self-correct based on prediction errors; the first step in the process is to integrate the environmental state. Environmental Prediction Status Deviation measurement between Directly defined as loss function loss function Its purpose is to quantify the predictive accuracy of the current digital twin model; the smaller the value, the more accurate the model's predictions. The value used here for calculation... and All are generated or used by the state reconstruction module in the preceding steps;

[0126] Calculate the loss function Regarding the internal parameter vector of the current digital twin model gradient ; Internal parameter vector This refers to a digital twin model, such as the set of all adjustable parameters in a deep neural network, including weights and biases; gradient. It is a vector pointing to the loss function. The direction of fastest growth serves to indicate the optimal direction for adjusting model parameters;

[0127] The internal parameter vector is updated based on the calculated gradient and evolution rate coefficients to generate the evolved model; the update step is precisely described by the following formula:

[0128] Model parameter vector evolution rate coefficients loss function gradient:

[0129]

[0130] Where M: model parameter vector, Evolution rate coefficient : Loss function gradient; The parameter vector of the evolved next-generation model; The current internal parameter vector of the digital twin model refers to the set of all adjustable parameters in the model, such as weights and biases. Loss function Regarding the current model parameter vector The gradient; The loss function quantifies the prediction accuracy of the current digital twin model and is defined as a measure of the deviation between the fused environmental state and the predicted environmental state. ;

[0131] Here This represents the parameter vector of the evolved next-generation model; the evolution rate coefficient η here is a hyperparameter used to control the step size of each update of the model parameters; evolution rate coefficient The selection of must ensure equality The dimensions at both ends are consistent;

[0132] Evolution rate coefficient The value is dynamically adjustable, and the adjustment logic is designed to be related to the severity of environmental changes; when the continuous-time rate of change of the integrated environmental state, such as the time derivative of wind speed or the rate of change of bird flock density exceeds a preset environmental change threshold. At that time, the system determines that the environment is in a period of rapid change; environmental change threshold. The setting is based on long-term statistical analysis of historical meteorological and ecological data of the airport. For example, the 95th percentile of historical change rate data can be used to identify statistically significant changes. In the event of such rapid change, the system will temporarily increase the evolution rate coefficient. The value of is used to accelerate the model's adaptation to new environments;

[0133] The adjustment logic can be set as follows: when the continuous-time change rate of the fused environment state, for example... The norm exceeds the threshold At that time, the evolution rate coefficient Use a preset acceleration value When the rate of change is less than or equal to the threshold When that happens, the baseline value is used. For example, it can be set The benchmark value It is the empirically optimal value obtained by training the model under stable environmental data, such as This achieves the effect of meticulous refinement in stillness and rapid adjustment in motion;

[0134] When the continuous-time rate of change of the environmental state is less than or equal to a threshold, it indicates that the environment is relatively stable and the evolution rate coefficient is low. Maintain its baseline value to ensure the stability of model convergence; the evolved model It will be used in the next time step (t+1) to generate a new, more accurate environmental prediction state;

[0135] The implementation method endows the digital twin model with true adaptive evolution capabilities. By defining the prediction bias as a loss function and optimizing it using gradient descent, the model can automatically learn from errors and continuously improve prediction accuracy. Furthermore, by introducing a dynamic evolution rate coefficient linked to the rate of environmental change, the system achieves intelligent learning rate adjustment. When the environment changes drastically, the model can accelerate updates to quickly keep up with reality, and when the environment is stable, it can fine-tune to achieve stable convergence. Thus, while ensuring the long-term accuracy of the model, it significantly improves its ability to track dynamic environmental changes and its response sensitivity, ensuring that the basic model of the entire intelligent control system always remains in optimal working condition.

[0136] The system possesses continuous adaptive evolution capabilities, ensuring long-term effectiveness. A core challenge of digital twin technology is that changes in the physical entity can cause the digital model to gradually become ineffective. The model evolution module of this invention introduces an online gradient descent algorithm. This solves the model drift problem; the physical meaning of the formula lies in defining the deviation between model predictions and the fused reality as a loss function. And along the gradient direction that minimizes the loss the most. To correct the model's internal parameter vector This means the model can learn and improve from each of its prediction errors; in addition, the evolution rate coefficient The adjustment logic is that when the rate of environmental change exceeds the environmental change threshold, the adjustment is increased. This value endows the model with a higher level of intelligence; when the environment changes drastically, the model can adjust and adapt at a faster rate, while it can make fine adjustments when the environment is stable; this self-learning and adaptive closed-loop characteristic enables the evolved model to persistently and accurately map the dynamics of the physical world, ensuring that the foundation of the decision-making quality of the entire intelligent control system is solid and reliable.

[0137] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A smart ecological control system for coastal airports based on digital twins and multi-objective optimization, characterized in that: include: The state reconstruction module acquires the raw real-time dataset from physical sensors and the predicted environmental state generated by a digital twin model. It then performs weighted fusion based on the deviation metric between the dataset and the predicted environmental state to generate a fused environmental state. The raw real-time dataset from the physical sensors includes the raw real-time dataset collected by physical sensors used for airport meteorological and bird activity monitoring. The predicted environmental state generated by the digital twin model includes the predicted environmental state deduced by the system's built-in digital twin model based on the state at the previous moment. The risk assessment module includes built-in risk prediction models for quantifying flight safety risks and ecological resilience risks. Based on the integrated environmental conditions, it quantifies and predicts flight safety risks and ecological resilience risks, and determines the dynamic weights corresponding to the two risks according to preset rules to obtain a weighted risk assessment result. Among them, flight safety risks refer to the possibility of directly threatening the safety of aircraft operations, including bird strikes and severe weather; ecological resilience risks refer to the potential negative impacts of airport operations on the long-term health of the ecosystem. The collaborative decision-making optimization module, based on the weighted risk assessment results, solves a multi-objective optimization problem with the weighted sum of risks as the objective function to generate the optimal control command; the optimal control command is a decision vector; the objective function of the multi-objective optimization problem is: ,in, The objective function provides a quantitative evaluation of each regulatory instruction, namely the total risk cost. Let be the decision vector. Each component of the decision vector The specific operating parameters of controllable physical equipment within an airport include the output power of acoustic bird deterrent equipment in a specific area and the flashing frequency or duration of airport navigation lights. To integrate with the environmental conditions; This represents an optimization process, namely, finding the objective function that satisfies the objective function. To achieve the minimum optimal decision vector ; The dynamic weights for flight safety risks are determined by the risk assessment module. Given a flight safety risk function, a predictive model is used to predict the execution decision vector. The resulting normalized flight safety risk index; The dynamic weight of ecological resilience risk is determined by the risk assessment module; For the ecological resilience risk function, a predictive model is used to predict the execution decision vector. The resulting normalized ecological resilience risk index; particle swarm optimization or genetic algorithms are used to find the objective function. Minimize the optimal decision vector The optimal control command generated; The model evolution module updates the internal parameter vector of the digital twin model using an online gradient descent algorithm based on the deviation metric between the fused environmental state and the predicted environmental state, thus obtaining the evolved model.

2. The intelligent ecological control system for coastal airports based on digital twins and multi-objective optimization as described in claim 1, characterized in that, The steps for generating the fusion environment state by the state reconstruction module are as follows: Calculate the deviation measure between the original real-time dataset and the predicted environmental state; Based on the deviation metric, the dynamic fusion coefficient is determined using a preset exponential decay rule; A weighted average of the original real-time dataset and the predicted environmental state is generated by using dynamic fusion coefficients.

3. The intelligent ecological control system for coastal airports based on digital twins and multi-objective optimization as described in claim 2, characterized in that, The logic for determining the dynamic fusion coefficient is as follows: As the bias metric increases, the dynamic fusion coefficient decays exponentially and approaches 0, thereby increasing the confidence weight for the predicted state of the environment. As the bias metric decreases and approaches 0, the dynamic fusion coefficient approaches 1 to increase the trust weight of the original real-time dataset.

4. The intelligent ecological control system for coastal airports based on digital twins and multi-objective optimization as described in claim 1, characterized in that, The steps for determining the dynamic weights are as follows: The dynamic weights of flight safety risks are correlated with key flight safety indicators; When key flight safety indicators fall below the preset safety threshold, the dynamic weight of flight safety risks is increased. When key flight safety indicators are better than or equal to the safety threshold, the dynamic weight of reducing flight safety risks is applied.

5. A smart ecological control system for coastal airports based on digital twins and multi-objective optimization as described in claim 4, characterized in that, The key flight safety indicator is the estimated collision time, and the safety threshold is a safe time threshold determined according to civil aviation safety regulations and aircraft performance.

6. The intelligent ecological control system for coastal airports based on digital twins and multi-objective optimization as described in claim 1, characterized in that, The steps for obtaining the evolved model are as follows: The deviation between the fused environmental state and the predicted environmental state is defined as the loss function. Calculate the gradient of the loss function with respect to the internal parameter vector of the digital twin model; The internal parameter vector is adjusted based on the gradient and evolution rate coefficients to generate the evolved model.

7. A smart ecological control system for coastal airports based on digital twins and multi-objective optimization as described in claim 6, characterized in that, The adjustment logic for the evolution rate coefficient is as follows: When the continuous-time change rate of the fused environment state exceeds the preset environmental change threshold, the evolution rate coefficient is increased to accelerate model adaptation. When the continuous-time rate of change of the fused environment state is less than or equal to the environmental change threshold, the evolution rate coefficient maintains the baseline value.

Images