Air cargo load planning optimization system based on flight scheduling data

Abstract

The application relates to the field of aviation logistics data processing, and discloses an aviation freight booking plan optimization system based on flight scheduling data, which comprises a demand data analysis unit, a transport capacity weight calculation unit, an association matching logic unit and a scheduling disturbance filtering unit. The demand data analysis unit extracts a demand vector, the transport capacity weight calculation unit determines a transport capacity node, the association matching logic unit establishes a mapping relationship, and the scheduling disturbance filtering unit calculates the attenuation of the matching association weight according to a time sequence drift amount and compares a judgment threshold value. The application converts the rigid binding of a time stamp into dynamic association based on a logic weight by constructing a mapping architecture of the demand vector and the transport capacity node, absorbing scheduling fluctuations through a weight attenuation mechanism, avoiding cascading loss of computing resources, and ensuring the system operation stability.

Description

Technical Field

[0001] This invention belongs to the field of air logistics data processing technology, specifically relating to an adaptive optimization system for cargo booking plans based on dynamic flight time-series scheduling. Background Technology

[0002] The current modern air logistics and ocean shipping system relies on data processing systems to arrange booking plans. The mainstream technical approach uses static constraint logic based on the estimated departure time stamp of the carrier to establish a corresponding binding relationship between booking requests and physical voyages. The physical navigation environment is affected by weather, port conditions and geopolitical conditions, which causes the scheduling data to objectively exhibit high-frequency nonlinear time-series drift characteristics. The existing static binding architecture implicitly relies on the absolute rigidity of the scheduling on the time axis. When the physical timestamp shifts slightly, the underlying data mapping chain breaks. In order to maintain normal operation, the system is usually forced to reserve a large amount of physical redundancy buffer, resulting in implicit loss of capacity resources.

[0003] In logistics hub scheduling, the underlying control algorithm logic architecture determines the system's ability to withstand disturbances. For example, Chinese invention patent application CN120471542A discloses a method, system, and equipment for optimizing cargo flight scheduling. It constructs a mixed integer linear programming model that minimizes cost and timeliness, and uses an exact algorithm to seek the globally optimal scheduling scheme. This approach relies on the strong temporal coupling between capacity timestamps and order demand, and belongs to static optimization under ideal working conditions. Faced with nonlinear scheduling drift, the algorithm lacks an endogenous absorption mechanism for temporal fluctuations. Small input deviations break the original logical constraints, and the system frequently triggers global rescheduling operations, causing disordered cascading losses of computing resources and causing oscillations in loading decisions at the physical execution end. For the mapping failure caused by scheduling drift, conventional improvement paths usually involve shortening the data synchronization cycle or increasing the frequency of reassignment operations. However, analysis shows that attempts to linearly increase the computational density cannot eliminate the essential misalignment between data logic and physical reality. Instead, they are prone to triggering disordered cascading consumption of computing resources during scheduling fluctuations, leading to transient empty space mismatches at the physical execution end, constituting an underlying contradiction that existing methods cannot resolve.

[0004] Therefore, the technical problem to be solved by this invention is how to break the rigid constraints of physical timestamps and construct a dynamic mapping mechanism that can endogenously absorb data fluctuations and maintain logical continuity. Summary of the Invention

[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: An air cargo booking plan optimization system based on flight scheduling data, comprising: The demand data parsing unit is used to acquire booking request data and parse the booking request data into a demand vector containing cargo tolerance time window parameters. The capacity weight calculation unit is used to obtain route and flight schedule data and parse the route and flight schedule data into capacity nodes that include the expected timestamp and historical fluctuation penalty factor; The association matching logic unit is connected to the demand data parsing unit and the capacity weight calculation unit. It is used to calculate the matching association weight based on the demand vector and the capacity node, and to establish the mapping relationship between the booking request data and the capacity node. The scheduling disturbance filtering unit is used to monitor the time-series drift of the estimated timestamp relative to the initial physical schedule in the capacity nodes in real time. The scheduling disturbance filtering unit attenuates the matching association weight according to the time-series drift and compares the attenuated matching association weight with a preset judgment threshold. When the attenuated matching association weight is lower than the judgment threshold, the scheduling disturbance filtering unit issues a rescheduling instruction to adjust the frequency of rescheduling instructions within a preset statistical period, offsetting the disturbance caused by the offset of the estimated timestamp to the system, and keeping the computing resource load within a safe redundancy range.

[0006] Preferably, when parsing the demand vector, the demand data parsing unit decomposes the data structure of the booking request data and extracts feature parameters. Specifically, this includes: extracting the cargo delivery time limit and loading preparation time from the booking request data; identifying cargo priority tags in the booking request data; mapping the cargo priority tags to weight enhancement factors in the demand vector; calculating the upper and lower boundary values ​​of the cargo tolerance time window parameters based on the cargo delivery time limit and loading preparation time; and mapping the cargo tolerance time window parameters to a multi-dimensional vector space to generate a demand vector characterizing the spatiotemporal distribution characteristics of booking demand. Each dimension in the demand vector corresponds to a discretized time slice of the cargo delivery time limit, providing a high-dimensional input matrix for subsequent calculation of matching association weights, ensuring that the sensitivity of the matching process to different time windows is quantitatively represented.

[0007] Preferably, the capacity weight calculation unit determines the dispersion of the historical fluctuation penalty factor based on the mean square error algorithm of the historical delay distribution data of the route, so as to use the historical fluctuation penalty factor to characterize the stability coefficient value of the capacity node on the time axis.

[0008] Preferably, the scheduling disturbance filtering unit is equipped with a smoothing filtering algorithm, which is used to lock the current value of the matching association weight when the time drift is less than a preset 5% disturbance interval.

[0009] Preferably, when the attenuated matching association weight is not lower than the judgment threshold, the scheduling disturbance filtering unit sends a locking signal to maintain the current matching link, so that the scheduling disturbance filtering unit stops sending rescheduling instructions to the system's underlying scheduling interface within the disturbance interval.

[0010] Preferably, the system also includes a capacity pool storage unit, which is connected to the capacity weight calculation unit. The capacity pool storage unit is used to store capacity nodes marked with historical fluctuation penalty factors and to provide a structured query index for the association matching logic unit.

[0011] Preferably, before issuing the rescheduling instruction, the scheduling disturbance filtering unit also calls the path cost evaluation module; the path cost evaluation module calculates the increase in computing resource load caused by the rescheduling instruction, and when the increase in computing resource load exceeds 80% of the preset load threshold, the scheduling disturbance filtering unit suspends the rescheduling instruction.

[0012] Preferably, the system also includes a feedback correction module, which is connected to the associated matching logic unit and is used to correct the deviation of the historical fluctuation penalty factor in the capacity node based on the operational feedback response data of the booking plan.

[0013] Compared with the prior art, the beneficial effects of the present invention are: 1. In optimizing booking plans for flight scheduling data, a decoupled mapping architecture between order demand vectors and capacity potential nodes is constructed in the data processing flow. This changes the rigid binding mode between orders and physical flights based on timestamps in traditional logistics scheduling. The static data chain, which is easily broken by external environmental interference, is transformed into a dynamic association based on logical weights. When nonlinear time-series drift occurs in the flight conditions at the physical level, the resulting fluctuations are absorbed by the weight decay mechanism inside the topological gravity mapping layer. This eliminates the need for the system to immediately trigger global-level redistribution calculations every time the scheduling is fine-tuned, effectively avoiding the cascading loss of computing resources caused by high-frequency data oscillations and ensuring the operational stability of the data processing system under logistics delay surges.

[0014] 2. Through the synergistic effect of the demand analysis unit and the scheduling potential field construction unit, the timeliness tolerance parameter of the booking request and the penalty factor of historical route fluctuations are woven into the calculation process of the matching confidence weight. This deep coupling of multi-dimensional features enables the system to have the ability to predict potential scheduling risks when establishing the initial matching matrix. This transforms the generated booking plan from simply meeting current time requirements into a logical distribution with probabilistic fault tolerance space. This technical approach reduces the need for subsequent local or global reconstruction actions forced by scheduling delays by quantifying uncertainties in advance, thereby improving the output certainty of logistics resource scheduling instructions from a mechanism level. Attached Figure Description

[0015] Figure 1 This is a diagram of the system architecture for optimizing booking plans based on dynamic weight association, as described in this invention. Figure 2 This is a flowchart of the scheduling disturbance filtering and computing resource load balancing control process of the present invention.

[0016] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0018] An air cargo booking optimization system based on flight scheduling data includes: The demand data parsing unit is used to acquire booking request data and parse the booking request data into a demand vector containing cargo tolerance time window parameters. The capacity weight calculation unit is used to obtain route and flight schedule data and parse the route and flight schedule data into capacity nodes that include the expected timestamp and historical fluctuation penalty factor; The association matching logic unit is connected to the demand data parsing unit and the capacity weight calculation unit. It is used to calculate the matching association weight based on the demand vector and the capacity node, and to establish the mapping relationship between the booking request data and the capacity node. The scheduling disturbance filtering unit is used to monitor the time-series drift of the estimated timestamp relative to the initial physical schedule in the capacity nodes in real time. The scheduling disturbance filtering unit attenuates the matching association weight according to the time-series drift and compares the attenuated matching association weight with a preset judgment threshold. When the attenuated matching association weight is lower than the judgment threshold, the scheduling disturbance filtering unit issues a rescheduling instruction to adjust the frequency of rescheduling instructions within a preset statistical period, offsetting the disturbance caused by the offset of the estimated timestamp to the system, and keeping the computing resource load within a safe redundancy range.

[0019] The demand data parsing unit described in this invention, when parsing the demand vector, decomposes the data structure of the booking request data and extracts feature parameters. Specifically, this includes: extracting the cargo delivery time limit and loading preparation time from the booking request data; identifying cargo priority tags in the booking request data; mapping the cargo priority tags to weight enhancement factors in the demand vector; calculating the upper and lower boundary values ​​of the cargo tolerance time window parameters based on the cargo delivery time limit and loading preparation time; and mapping the cargo tolerance time window parameters to a multi-dimensional vector space to generate a demand vector characterizing the spatiotemporal distribution characteristics of booking demand. Each dimension in the demand vector corresponds to a discretized time slice of the cargo delivery time limit, providing a high-dimensional input matrix for subsequent calculation of matching association weights, ensuring that the sensitivity of the matching process to different time windows is quantitatively represented.

[0020] The capacity weight calculation unit described in this invention determines the dispersion of the historical fluctuation penalty factor based on the mean square error algorithm of the historical delay distribution data of the route, so as to use the historical fluctuation penalty factor to characterize the stability coefficient value of the capacity node on the time axis.

[0021] The scheduling disturbance filtering unit described in this invention is equipped with a smoothing filtering algorithm. The smoothing filtering algorithm is used to lock the current value of the matching association weight when the time drift is less than a preset 5% disturbance interval.

[0022] According to the present invention, when the attenuated matching association weight is not lower than the judgment threshold, the scheduling disturbance filtering unit sends a locking signal to maintain the current matching link, so that the scheduling disturbance filtering unit stops sending rescheduling instructions to the underlying scheduling interface of the system within the disturbance interval.

[0023] The system described in this invention also includes a capacity pool storage unit, which is connected to the capacity weight calculation unit. The capacity pool storage unit is used to store capacity nodes marked with historical fluctuation penalty factors and to provide a structured query index for the association matching logic unit.

[0024] The scheduling disturbance filtering unit described in this invention is also used to call the path cost evaluation module before issuing the rescheduling instruction; the path cost evaluation module calculates the increase in computing resource load caused by the rescheduling instruction, and when the increase in computing resource load exceeds 80% of the preset load threshold, the scheduling disturbance filtering unit suspends the rescheduling instruction.

[0025] The present invention describes identifying cargo priority tags in booking request data and mapping these cargo priority tags to weight enhancement factors in the demand vector.

[0026] The system described in this invention also includes a feedback correction module, which is connected to the associated matching logic unit and is used to correct the deviation of the historical fluctuation penalty factor in the capacity node based on the operational feedback response data of the booking plan.

[0027] Example 1: In international logistics scheduling involving cross-border air cargo transportation on multiple routes, the booking data processing system handles flight status timeline drift caused by air hub occupancy. This environment includes received cargo booking requests and real-time updated carrier departure timestamps. When the average offset time of the carrier exceeds 12 hours and the fluctuation range covers multiple voyages, the static timestamp binding logic causes a break in the association between orders and physical voyages, leading to cascading losses of computing resources in global reallocation retrieval. The demand data parsing unit extracts timeliness elements from the booking data and calculates the upper and lower boundary values ​​of the cargo tolerance time window parameters based on cargo delivery time limits and loading preparation time. Then, the cargo tolerance time window parameters are converted into a demand vector, and simultaneously... The force weight calculation unit receives the scheduling data sequence of physical carrier nodes and converts each independent physical flight into a capacity potential node to generate an initial capacity map. The process of converting independent physical flights into capacity potential nodes is achieved by constructing a binary continuous probability density function based on remaining available cabin space and time drift. The system maps the discrete scheduling database into a topological space composed of a continuous time coordinate system, and performs exponential weight decay calculations on both sides with the expected departure time stamp as the peak center. This transforms the traditional rigid matching operation of time stamps into a continuous potential field distribution that characterizes the proximity tolerance. Based on this, the association matching logic unit establishes weighted connections between the demand vector and the capacity potential nodes, and defines the calculation formula for the connection weight as follows: ,in, Represents the matching and association weights between nodes. The remaining available cabin space represents the capacity potential energy node. This represents the time difference between the cargo tolerance time window parameter and the flight schedule data. The historical volatility penalty factor represents the scheduling data series. and These represent the system coefficients of the corresponding parameters. Before applying this calculation formula, the association matching logic unit needs to perform deterministic dimensionality reduction mapping on the generated high-dimensional demand vector. The system uses principal component extraction logic to extract the cargo tolerance time slice coordinates representing the core constraints of timeliness in the multidimensional demand vector and orthogonally project them onto the one-dimensional time axis. This extracts scalar parameters with the same mathematical dimension as the route and flight scheduling data, which are directly used as the time difference Δt in the formula. This gives the original abstract multidimensional feature matrix the basis for substituting into elementary algebraic equations for weight allocation. When performing numerical substitution, based on the difference in physical dimensions and orders of magnitude between the remaining available cabin space base and the time difference, a range standardization procedure is introduced to extract the extreme value range of available cabin space and the time deviation limit span in the historical operation dataset of the relevant route. The actual sampled variables are then used as the basis for weight allocation. With variables The linear proportional relationship is used to uniformly map the data to the dimensionless reference interval of 0 to 1; after pre-normalization, the output is... The physical confidence criterion is transformed into a weighted average of dimensionless components of the same dimension, thus avoiding the phenomenon of multiple variables overwhelming the underlying data.

[0028] When the monitoring signal indicates that the estimated departure time of a specific capacity node has drifted, the scheduling disturbance filtering unit maintains the original image address and updates the historical fluctuation penalty factor of the corresponding capacity potential node in real time. Driving matching association weights The system performs synchronous attenuation calculation; the associated matching logic unit evaluates the difference in topology change between the attenuated matching matrix and the previous matching matrix; when the difference in topology change is less than the preset reconstruction energy threshold, the system determines that the current scheduling fluctuation is within the tolerance time window parameter of the cargo; the system issues a locking signal to maintain the current matching link to block the issuance of repeated rescheduling instructions; only when matching the associated weight... When the data falls below a preset threshold, the system clears the memory address of that data and performs a local mapping reconstruction. This local mapping reconstruction includes: re-traversing the system's capacity pool for any remaining candidate capacity nodes based on weight calculation logic, calculating the newly associated weight values ​​relative to the current demand vector for each node; generating a local priority queue based on the calculated values ​​from largest to smallest; extracting the first target capacity node in the queue and comparing its estimated departure time with the boundary value of its booking request tolerance time window; if it falls within the tolerance boundary, overwriting the underlying address to establish a new associated matching order and exiting the reconstruction loop; if it exceeds the tolerance boundary, sequentially extracting the next node in the queue and repeating the conditional judgment until a compliant node is found or the queue is cleared. This method eliminates the disturbance of scheduling fluctuations within a specific range to the overall shipping plan, enabling the logistics system to maintain the stability of loading decisions under delay conditions and improving the full load rate of core capacity resources.

[0029] Example 2: This example serves as a verification of the control effect and parameter boundary of the system described in Example 1. Its dynamic mapping mechanism and adaptive optimization logic are completely consistent with those in Example 1. The difference lies in that this example focuses on stress testing the system's load-planning stability using statistical samples in a verification system simulating multi-node intermodal air cargo transport. Specifically, the system processes 10,000 sets of booking flow data with time-series characteristics from a publicly available civil aviation flight dynamics dataset. Furthermore, the system uses the Monte Carlo method to superimpose a delay jitter noise with a signal-to-noise ratio of 20.0 dB onto the departure timestamps of the carriers at the input end, to verify the system's load-planning stability under non-ideal scheduling data disturbances. The association matching logic unit employs a matching association weight calculation model, the calculation formula of which is... ,in, This represents the matching association weight between nodes, i.e., a dimensionless association strength index. This represents the system weight coefficient, with a value of 0.65. The remaining available cabin space represents the capacity potential energy node. This represents the time difference between the cargo tolerance time window parameter and the flight schedule data. This represents the system adjustment coefficient, with a value of 0.35. The historical volatility penalty factor represents the historical volatility penalty factor of the scheduling data series. When the statistical mean increases from 0.15 to 0.85, the system reduces the matching association weight. The initial allocation value tilts load planning resources toward high-confidence voyages.

[0030] When faced with an average scheduling drift of 12.5 hours, the sample group of this invention maintained a full load rate of 89.4% for its capacity resources. Control group one, which uses a matching method based on departure timestamps, saw its full load rate drop to 68.2%, and its CPU utilization surged to 92.0% due to global reallocation calculations. Control group two removed the historical fluctuation penalty factor. Subsequently, when handling more than 15.0% of random temporal disturbances, 42.0% of invalid mapping reconstruction actions were generated. For 2500 sets of demand vectors containing cargo tolerance time window parameters, when the parameter boundaries covered a 24.0-hour drift, the frequency of system rescheduling commands decreased by more than 75.0%. The observed performance indicators, showing the trend of parameter changes, indicate that when the reconstruction energy threshold is in the range of 0.15 to 0.45, the system's absorption rate of fluctuations and response delay are in balance. The extreme boundary parameters used to determine whether the energy threshold is in the optimal balance range, and the weight coefficient α and adjustment coefficient β in the matching weight model, are set to 0.65 and 0.3 respectively. 5. All of these are stable extreme values ​​with engineering guidance obtained by reading the actual scheduling delay ledger data of the air logistics network over the past 36 months and using a Monte Carlo simulation model with the objective function of maximizing the full load rate of the system's capacity pool. After multiple rounds of parameter traversal evolution fitting, the values ​​are obtained. When the reconstruction energy threshold exceeds the degradation point of 0.75, the filtering effect of the scheduling disturbance filtering unit on the deviation data weakens, resulting in a non-linear decrease in the loading success rate. The experimental data confirms the effectiveness of the scheme of absorbing scheduling fluctuations through the weight decay mechanism in suppressing the cascading loss of computing resources and verifies the performance of the dynamic correlation architecture between demand vector and capacity nodes in a complex logistics environment.

[0031] The scheduling disturbance filtering unit controls the rescheduling instructions within a preset statistical period through an instruction frequency adjustment loop. The total output within the loop is calculated by setting the instruction counter at the beginning of each preset statistical cycle. Perform a zeroing and reset; when the associated matching logic unit determines the attenuated associated weight... Below the judgment threshold At that time, the system retrieves the instruction counter. The real-time value, if the real-time value does not reach the preset frequency limit. The scheduling disturbance filtering unit then outputs a rescheduling instruction for the current booking request and executes it. The system performs an accumulation operation; otherwise, it writes the current mapping reconstruction request into a delayed processing queue. This represents the preset statistical period. Represents the instruction counter. This represents the decayed matching association weight. This represents the threshold for judgment. Representing the upper limit of frequency, this process ensures that the system's computing resource load remains within a safe redundancy range under conditions of frequent flight delays by truncating the pulse width of the control flow.

[0032] Example 3: In this example, the architecture of the remaining units, the underlying topology connections, and the basic data processing flow of the air cargo booking plan optimization system are the same as in Example 1. The difference is that this example is mainly used in a logistics collaborative network covering multiple trunk air hubs around the world. The system adaptively calibrates and reconstructs the energy threshold and executes task allocation based on the minimum remaining load priority to handle heterogeneous scheduling data streams caused by differences in loading procedures and customs clearance time fluctuations in different countries. Specifically, when the scheduling disturbance filtering unit faces a deviation signal with time-series deviation characteristics, the preset judgment benchmark cannot perceive the environmental noise intensity of different nodes in real time. The system generates logical oscillations between global rescheduling and local mapping maintenance, causing transient overflow of the processor core load in the computing cluster. The demand data parsing unit extracts the cargo delivery time limit from the booking request data. With loading preparation time And determine the length value of the cargo tolerance time window parameter. The demand data parsing unit calculates the length value as the difference between the cargo delivery time and the loading preparation time. The time axis components of the demand vector are mapped and normalized using cargo volume parameters from booking request data and preset priority coefficients to construct a demand vector in a multi-dimensional space. Based on this, the capacity weight calculation unit determines the value of the reconstructed energy threshold according to the following calibration procedure: During the initial statistical period after system startup, data is continuously collected. Group matching association weights for sample data; calculate the sample variance of the sample data. According to the formula Determine the reconstruction energy threshold, where This represents the reconstruction energy threshold, which is the critical energy value for determining the reconstruction of the topological map. Represents the scaling factor of the system. The standard deviation represents the matching association weight. In this calculation procedure, the scaling factor η of the system is dynamically calculated by cross-multiplying the set trunk air network timeliness priority constant with the variance data of the scheduling offset collected in real time in the current cycle. Its value is adaptively adjusted according to the intensity of each wave of logistics environmental noise. This calibration procedure enables the judgment benchmark to be dynamically locked with the intensity of environmental scheduling fluctuations, eliminating the blindness of parameter selection.

[0033] When the monitoring unit indicates that the departure timestamp of a physical capacity node has shifted, the scheduling disturbance filtering unit retrieves the reconstruction energy threshold. The topology change difference of the matching association weights is quantitatively determined, and when the topology change difference exceeds the reconstruction energy threshold... When the rescheduling trigger condition is met, the task allocation logic component retrieves the real-time load percentage of each processor core in the computing cluster. Before actually sending and allocating the rescheduling task to the underlying layer, the path cost evaluation module pre-extracts the total number of candidate capacity nodes in the system to be traversed and retrieved by the instruction, and multiplies it by the baseline memory cycle constant required for a single association matching operation. Through this linear product operation, the absolute increment parameter of the computing power resources load expected to be additionally occupied by the rescheduling action is output in advance. The resource allocation procedure based on minimum remaining load priority is adopted to map the rematching calculation task for the current order to the physical node with the lowest load percentage for operation. This process evenly distributes the task execution pressure to the cluster nodes, avoids the interference of single node response delay on the global matching logic, maintains the stability of the air cargo system's resource allocation under scheduling jitter conditions, and improves the matching accuracy of physical cabin space and booking requests.

[0034] The scheduling disturbance filtering unit employs a weighted dynamic evolution model to perform attenuation correction on the matching associated weights. This evolution model is based on the exponential degradation physical law in equipment reliability engineering, effectively modeling the loss of fulfillment capacity due to time drift at logistics nodes as a non-linear exponential decay of confidence strength on a continuous time axis. The system calculates in real-time the offset of the estimated timestamp of a specific capacity node relative to the initial physical schedule. And based on the offset duration and historical fluctuation penalty factor The product determines the attenuation operator. The scheduling disturbance filtering unit is based on The formula is used to calculate the decayed matching association weight, where, This represents the decayed matching association weight. Represents the initial matching association weight. This represents an exponential function with the natural constant as its base. Represents the system decay operator. Represents the historical volatility penalty factor. This represents the duration of the offset, due to the historical volatility penalty factor. A nonlinear mapping relationship between offset duration and mapping confidence was established. This procedure enables weight contraction to be triggered when high-delay-fluctuation flights generate a small amount of time-series drift, thereby suppressing the maintenance of invalid associations from the data source.

[0035] Example 4: In this example, the system's component division, basic data interaction process, and dynamic association framework are the same as in the previous examples. The difference is that this example is mainly used in the access phase of a newly established air logistics hub to establish a benchmark data source for the system to determine the reconstruction action, perform parameter initialization calibration, and measure the noise of the operating environment. Specifically, the capacity weight calculation unit selects the flight operation logs of the corresponding air hub for the past 180 natural days, extracts a discrete sample set containing the actual departure time and the planned departure time of the flight, and determines the deviation benchmark of the physical node by calculating the mean absolute value of the offset of each observation value in the discrete sample set relative to the planned departure time. The capacity weight calculation unit counts the ratio of the historical on-time number of each carrier to the total number of takeoffs and landings in the same statistical period, and maps the reciprocal of the ratio to a standardized interval from 0 to 1, thereby determining the historical fluctuation penalty factor used to characterize the certainty of capacity. The initial calibration components.

[0036] When the system performs parameter calibration in a heterogeneous logistics node environment, the scheduling disturbance filtering unit selects the initial statistical period of the system under the current physical hub. It operates on a continuous 24-hour monitoring cycle, collecting 500 sets of matching and correlation weights at high frequency. The system uses a dynamic evolution sample to calculate the noise figure of the current flight route operating environment based on the first derivative volatility of the dynamic evolution sample. Then, it determines the scaling factor based on the product of the noise figure and the preset business timeliness tolerance constant. The range of values ​​for which the energy threshold is reconstructed. Based on scaling factor Sample standard deviation of the matching association weights The result of the product operation is determined, where, Represents the initial statistical period. Represents the matching association weight. Represents the scaling factor of the system. Represents the reconstructed energy threshold. The sample standard deviation representing the matching association weights is used to establish the benchmark data source for determining the reconstruction action, enabling the system to maintain the stability of the load planning decision logic in trunk aviation networks with different time constraints.

[0037] Example 5: In this example, the adaptive optimization logic and basic component connection architecture of the system are the same as those in the previous examples. The difference is that this example is mainly used in the air cargo scheduling and deployment process involving multiple terminals. The system processes the timestamp synchronization deviation from heterogeneous sensor terminals, and the demand data parsing unit obtains the maximum value of the cargo delivery time limit of the current operation node within the preset operating cycle. and minimum value Simultaneously, the demand data parsing unit extracts the length value of the cargo tolerance time window parameter from a specific booking request. The demand data parsing unit uses the formula The time axis component values ​​of the demand vector are calculated, where, The time component of the demand vector represents the numerical value. Represents the length value. Represents the maximum time limit. Representing the minimum time limit, this approach establishes dimensional consistency for demand vectors when deployed across regions.

[0038] The capacity weight calculation unit selects flight operation records of the means of transport, extracts the deviation distribution characteristics between planned departure time and actual departure time, and sets a sliding window of 30 calendar days to collect discrete samples. The capacity weight calculation unit performs exponential decay weighted fitting on the deviation data within the sliding window to calculate the weight decay coefficient. To determine the historical volatility penalty factor The baseline quantization value, where, Represents the historical volatility penalty factor. Represents the weight decay coefficient. Representing the reconstructed energy threshold, this process enables the scheduling disturbance filtering unit to lock the reconstructed energy threshold based on the fluctuation intensity of the current physical hub. The determination boundary ensures that the stability index of the associated matching logic unit is maintained in the preset high range, while eliminating the failure of logic criteria caused by differences in physical environment.

[0039] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0040] Finally, 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.

Claims

1. An air cargo load planning optimization system based on flight scheduling data, characterized by, include: The demand data parsing unit is used to acquire booking request data and parse the booking request data into a demand vector containing cargo tolerance time window parameters. The capacity weight calculation unit is used to obtain route and flight schedule data and parse the route and flight schedule data into capacity nodes that include the expected timestamp and historical fluctuation penalty factor; The association matching logic unit is connected to the demand data parsing unit and the capacity weight calculation unit. It is used to calculate the matching association weight based on the demand vector and the capacity node, and to establish the mapping relationship between the booking request data and the capacity node. The scheduling disturbance filtering unit is used to monitor the time-series drift of the estimated timestamp relative to the initial physical schedule in the capacity nodes in real time. The scheduling disturbance filtering unit attenuates the matching association weight according to the time-series drift and compares the attenuated matching association weight with a preset judgment threshold. When the attenuated matching association weight is lower than the judgment threshold, the scheduling disturbance filtering unit issues a rescheduling instruction to adjust the frequency of rescheduling instructions within a preset statistical period, offsetting the disturbance caused by the offset of the estimated timestamp to the system, and keeping the computing resource load within a safe redundancy range.

2. The system for optimizing the cargo loading plan based on flight scheduling data according to claim 1, wherein, When parsing the demand vector, the demand data parsing unit decomposes the data structure of the booking request data and extracts feature parameters. Specifically, this includes: extracting cargo delivery time limits and loading preparation time from the booking request data; identifying cargo priority tags in the booking request data; mapping cargo priority tags to weight enhancement factors in the demand vector; calculating the upper and lower boundary values ​​of cargo tolerance time window parameters based on cargo delivery time limits and loading preparation time; and mapping the cargo tolerance time window parameters to a multi-dimensional vector space to generate a demand vector representing the spatiotemporal distribution characteristics of booking demand. Each dimension in the demand vector corresponds to a discretized time slice of the cargo delivery time limit, which provides a high-dimensional input matrix for the subsequent calculation of matching association weights, ensuring that the sensitivity of the matching process to different time windows is quantitatively represented.

3. The system of claim 1, wherein, The capacity weight calculation unit determines the dispersion of the historical fluctuation penalty factor based on the mean square error algorithm of the historical delay distribution data of the route, so as to use the historical fluctuation penalty factor to characterize the stability coefficient value of the capacity node on the time axis.

4. The system of claim 1, wherein, The scheduling disturbance filtering unit is equipped with a smoothing filtering algorithm. The smoothing filtering algorithm is used to lock the current value of the matching association weight when the time drift is less than the preset 5% disturbance interval.

5. The system for optimizing the cargo loading plan based on flight scheduling data according to claim 4, wherein, When the attenuated matching association weight is not lower than the judgment threshold, the scheduling disturbance filtering unit sends a locking signal to maintain the current matching link, so that the scheduling disturbance filtering unit stops sending rescheduling instructions to the underlying scheduling interface of the system within the disturbance interval.

6. The system for optimizing the cargo loading plan based on flight scheduling data according to claim 1, wherein, The system also includes a capacity pool storage unit, which is connected to the capacity weight calculation unit. It is used to store capacity nodes marked with historical fluctuation penalty factors and to provide a structured query index for the association matching logic unit.

7. The air cargo booking plan optimization system based on flight scheduling data according to claim 1, characterized in that, Before issuing a rescheduling instruction, the scheduling disturbance filtering unit also calls the path cost evaluation module. The path cost evaluation module calculates the increase in computing resource load caused by the rescheduling instruction. When the increase in computing resource load exceeds 80% of the preset load threshold, the scheduling disturbance filtering unit suspends the rescheduling instruction.

8. The air cargo booking plan optimization system based on flight scheduling data according to claim 1, characterized in that, The system also includes a feedback correction module, which is connected to the associated matching logic unit and is used to correct the deviation of the historical fluctuation penalty factor in the capacity node based on the operational feedback response data of the booking plan.

Images