A multi-unmanned aerial vehicle power consumption prediction and charging optimization method

Abstract

This invention discloses a method for predicting the power consumption of multiple unmanned aerial vehicles (UAVs) and optimizing charging, comprising: collecting UAV flight data and environmental sensor data in real time, processing the data using a regression model to obtain the power consumption rate; if the trend exceeds a preset threshold, integrating historical flight records through an information processing step to determine the predicted critical point time; obtaining the location and capacity data of available charging stations based on the predicted critical point time, and determining a charging timing sequence; processing multiple UAV requests using a priority sorting algorithm according to the determined charging timing sequence to obtain a resource allocation scheme; if there are conflicts in the resource allocation scheme, adjusting the sequence parameters through an information processing step to obtain an optimized allocation result; and updating the model parameters based on the optimized allocation result using real-time feedback data to determine the power consumption rate prediction for the next cycle.

Description

Technical Field

[0001] This invention relates to the field of power maintenance technology, specifically a method for predicting the power consumption and optimizing the charging of multiple unmanned aerial vehicles (UAVs). Background Technology

[0002] As an important component of modern technology, drones demonstrate irreplaceable value in various fields such as power monitoring and emergency rescue, and their operational efficiency and continuity directly affect the success or failure of missions. However, with the increasing complexity of drone application scenarios, ensuring their power supply and optimizing charging management have become key issues affecting the industry's development. The endurance of drones and the continuity of mission execution urgently require innovative technological means to solve the numerous challenges in the charging process.

[0003] Existing drone charging management methods suffer from the following problems: the complexity of mission types and environmental conditions makes it difficult to accurately predict the rate of power consumption. For example, in high-temperature or strong-wind environments, drone energy consumption increases significantly, and current technologies struggle to capture these changes in real time. This hinders precise control over charging timing and resource allocation. Because the critical point of power depletion cannot be predicted in advance, drones are often forced to abort missions due to insufficient power, or scheduling chaos occurs when charging resources are scarce. Summary of the Invention

[0004] The purpose of this invention is to solve the above problems and provide a method for predicting the power consumption of multiple drones and optimizing charging, which significantly improves the continuity of drone missions and the efficiency of resource utilization, realizes dynamic power management and efficient charging scheduling, reduces the risk of interruption and improves the overall operational efficiency.

[0005] The technical solution adopted by this invention to solve its technical problem is: A method for predicting the power consumption and optimizing the charging of multiple drones includes the following steps: The S101 collects drone flight data and environmental sensor data in real time, processes this data using a regression model, and obtains the power consumption rate. S102 obtains the correlation features between task type and environmental variables based on the obtained power consumption rate, and uses a neural network to analyze these features to determine the changing trend of the consumption rate. If the trend of change exceeds the preset threshold, the information processing stage will integrate historical flight records to determine the predicted time of the critical point. S104 determines the charging timing sequence by judging the predicted time of the critical point; S105 arranges the sequence according to the determined charging time, uses a priority sorting algorithm to process multiple drone requests, and obtains a resource allocation scheme; S106 If there is a conflict in the resource allocation scheme, the sequence parameters are adjusted through the information processing stage to obtain the optimized allocation result; Based on the optimized allocation results, S107 obtains real-time feedback data to update the model parameters and determines the consumption rate prediction for the next cycle.

[0006] Further, step S101 includes: By collecting UAV flight data and environmental sensor data in real time, a data processing flow is constructed to obtain the flight data and environmental data after preliminary processing. Based on the flight data and environmental data, a regression model is used to analyze and determine the predicted value of the power consumption rate.

[0007] Further, step S102 includes: Real-time data on power consumption is acquired, categorized and organized according to task type and environmental variables, to obtain a preliminary structured dataset; Based on the structured dataset, a pre-established neural network model is used to process the correlation features between task types and environmental variables, and to determine the mutual influence relationships between features; By analyzing the aforementioned interrelationships, key corresponding points between power consumption and speed changes are extracted to determine the pattern of power consumption speed changes. Based on the changing pattern of the consumption rate, and considering the influence of environmental variables, the contribution of task type to the rate change is analyzed to obtain specific trend influencing factors; Based on the aforementioned trend influencing factors, a predictive framework for consumption trends is constructed to obtain the future direction of speed changes and determine the prediction results.

[0008] Further, step S102 includes: Based on the prediction results, combined with the influence of variables and the characteristics of task types, a dynamic adjustment strategy for power consumption is generated, and the accuracy of the optimized trend prediction is judged. Based on the optimized trend prediction accuracy, the parameters of the neural network model are continuously updated to obtain a consumption trend output that conforms to the actual scenario.

[0009] Further, step S103 includes: The trend detection module monitors the changing trend data in real time. If the changing trend data exceeds a preset threshold, the information processing stage is triggered to obtain a preliminary abnormal signal. Based on the initial abnormal signal, the data integration function is invoked to extract relevant data from historical records and determine the feature content that matches the trend data. For the aforementioned features, the correlation between historical records and current trend detection results is analyzed through information processing to determine the critical prediction time range; Key time nodes are obtained from the time range, and preset logical rules are used. If the key time node is consistent with the abnormal pattern in the historical record, the time window for critical prediction is further narrowed. By using the reduced time window and combining the data integration results, the persistence of the changing trend data is analyzed, and the predicted time point of the critical point is determined.

[0010] Further, step S104 includes: The data acquisition module obtains critical point data and prediction time data from a preset database, and then cleans and formats the data to obtain a standardized time prediction dataset. Based on the time prediction dataset, a preset threshold is used to determine the time. If the predicted time reaches the critical point, the charging station information acquisition process is triggered to determine the list of charging stations that meet the conditions. The location information and capacity data of each charging station are obtained from the charging station list. The location information is then geocoded to obtain the corresponding coordinate dataset. Based on the coordinate dataset and capacity data, the availability of charging stations is analyzed. If the capacity data is lower than a preset threshold, the charging station is removed, and an optimized set of available charging stations is obtained. Using the available charging station set, combined with predicted time and location information, the matching degree of each charging station is calculated, and the K-nearest neighbor algorithm is used to sort the matching degrees to determine the charging station with the highest priority. Based on the highest priority charging station, and combined with predicted time and capacity data, a charging timing sequence is generated. The sequence is then validated using a time window to obtain the final charging arrangement result.

[0011] Further, step S105 includes: Acquire charging demand data for multiple drones, including the request time and battery status of each drone, and determine the initial request sequence; Based on the initial request sequence, a priority sorting method is used to evaluate the charging timing of each drone. If the battery level of a drone is lower than a preset threshold, its priority is increased, resulting in an adjusted task priority list. By analyzing the number and location distribution of currently available charging resources using the task priority list and the resource scheduling status of charging stations, a preliminary resource allocation plan is generated. For the preliminary resource allocation scheme, real-time charging station load data is obtained. If the load of a charging station exceeds the preset limit, the resource scheduling strategy is adjusted to determine the final allocation scheme. Further, step S105 includes: Based on the final allocation scheme, the charging requests of multiple drones are processed, and the charging timing arrangement of each drone is updated in real time to obtain a dynamically adjusted timing sequence. By using the dynamically adjusted timing sequence, the resource scheduling and task priority execution are monitored. If the charging needs of a certain drone are not met, the priority sorting process is re-triggered to update the resource allocation scheme and charging timing arrangement, and generate an optimized execution plan.

[0012] Further, step S106 includes: Conflict data is obtained from resource allocation schemes, and the existence of scheme conflicts is determined through information processing to obtain preliminary conflict detection results; If a conflict is detected, the impact of the conflict data analysis sequence parameters is determined to identify the range of parameters that need to be adjusted. Based on the determined parameter range, the sequence parameters are adjusted using preset logical rules to obtain the adjusted parameter configuration; The allocation scheme is regenerated using the adjusted parameter configuration, and it is determined whether there are still conflicts to obtain the updated scheme data. If the updated scheme data still has conflicts, the cause of the conflicts will be analyzed through the information processing stage to determine the direction of the second adjustment; Based on the direction of the secondary adjustment, the resource allocation scheme is optimized using the support vector machine algorithm to obtain the final optimization result.

[0013] Further, step S107 includes: Obtain the optimized resource allocation results, extract key fields from the allocation results, and construct a structured dataset; Based on the structured dataset, obtain real-time feedback data, and perform integrity checks on the timestamps and event records in the feedback data; If there are missing fields in the feedback data, the data is filled in using a preset interpolation method to obtain a complete feedback dataset; Using the complete feedback dataset, the consumption rate change trend is analyzed, and the support vector machine algorithm is used to model the change trend to determine the key factors affecting the rate change. Based on the key factors, the model parameters are updated, and the updated parameters are standardized to obtain an adjusted parameter set. By combining the set of adjustment parameters and historical cycle feedback data, the consumption rate of the next cycle is predicted, and it is determined whether the predicted value exceeds the preset threshold range. If the predicted value exceeds the preset threshold range, an anomaly marker is triggered to identify potential risk points; Based on the potential risk points and the allocation results, an adjustment strategy is generated, and the adjustment strategy is logically verified to obtain the final allocation adjustment scheme. Based on the final allocation adjustment scheme, the resource allocation plan for the next cycle is updated, and the priority field in the allocation plan is sorted to determine the execution order for the next cycle.

[0014] The beneficial effects of this invention are: 1. This invention addresses the unique business scenario problem faced by multiple drones in complex mission environments, where real-time prediction of power consumption rates is difficult and conflicts in charging schedules can easily lead to mission interruptions. This problem integrates logically related factors such as the influence of environmental variables on consumption trends, historical data for predicting critical points, and priority conflicts in resource allocation. This invention collects flight and sensor data in real time, uses a regression model to calculate the power consumption rate, and utilizes neural networks to analyze mission type and environmental characteristics to determine the changing trend. If the trend exceeds a threshold, historical records are used to determine the predicted critical point time. Based on this, charging station data is obtained to generate a timing sequence. A priority ranking algorithm is then used to process requests from multiple drones. In case of conflicts, sequence parameters are adjusted to optimize allocation, and model parameters are updated based on real-time feedback to predict the consumption of the next cycle. This method significantly improves the continuity of drone missions and resource utilization efficiency, achieves dynamic power management and efficient charging scheduling, reduces the risk of interruption, and improves overall operational efficiency. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the structure of the present invention; Figure 2 This is a flowchart for generating the resource allocation scheme of the present invention. Detailed Implementation

[0016] like Figure 1 As shown, a method for predicting the power consumption and optimizing the charging of multiple drones includes the following steps: The S101 collects drone flight data and environmental sensor data in real time, processes this data using a regression model, and obtains the power consumption rate. S102 obtains the correlation features between task type and environmental variables based on the obtained power consumption rate, and uses a neural network to analyze these features to determine the changing trend of the consumption rate. If the trend of change exceeds the preset threshold, the information processing stage will integrate historical flight records to determine the predicted time of the critical point. S104 determines the charging timing sequence by judging the predicted time of the critical point; S105 arranges the sequence according to the determined charging time, uses a priority sorting algorithm to process multiple drone requests, and obtains a resource allocation scheme; S106 If there is a conflict in the resource allocation scheme, the sequence parameters are adjusted through the information processing stage to obtain the optimized allocation result; Based on the optimized allocation results, S107 obtains real-time feedback data to update the model parameters and determines the consumption rate prediction for the next cycle.

[0017] Step S101 includes: constructing a data processing flow by collecting UAV flight data and environmental sensor data in real time to obtain the preliminarily processed flight data and environmental data; and using a regression model to analyze the flight data and environmental data to determine the predicted value of the power consumption rate.

[0018] In the process of processing drone flight data and environmental data, the timeliness and accuracy of the data can be ensured by constructing a real-time data acquisition and processing workflow. Drone flight data includes flight speed, altitude, attitude angles, etc., while environmental data covers information such as wind speed, temperature, and humidity.

[0019] After initial data processing, a regression model can be used to predict the rate of energy consumption. The regression model is trained using historical flight data and environmental data to analyze the relationship between energy consumption and flight parameters and environmental factors.

[0020] Step S102 includes: acquiring real-time power consumption data, classifying and organizing it according to task type and environmental variables to obtain a preliminary structured dataset; based on the structured dataset, using a pre-established neural network model to process the correlation features between task type and environmental variables, and determining the mutual influence relationship between features; through the mutual influence relationship, extracting key corresponding points between power consumption and speed change, and judging the change pattern of consumption speed; based on the change pattern of consumption speed, combined with the influence of environmental variables, analyzing the contribution of task type to speed change, and obtaining specific trend influencing factors; based on the trend influencing factors, constructing a prediction framework for consumption trend, obtaining the future trend of speed change, and determining the prediction result.

[0021] In real-time data processing of drone power consumption, categorizing and organizing data based on task type and environmental variables can be achieved by constructing a data acquisition framework. Flight missions can be divided into three categories: inspection, transportation, and aerial photography, while environmental variables such as wind speed, temperature, and humidity are recorded. For example, assuming an inspection mission has a wind speed of 5 m / s, a temperature of 20 degrees Celsius, and humidity of 60%, preliminary data processing yields a structured dataset containing the correspondence between task type and environmental variables. This classification method facilitates subsequent analysis of power consumption differences under different mission scenarios.

[0022] When using neural network models to process the correlation features between task type and environmental variables, a multi-layer network structure can be designed. The input layer receives task type and environmental variable data, while the hidden layer extracts the potential relationships between features. For example, in high wind speeds, inspection tasks may require more power to maintain stable flight, while transportation tasks may be less affected.

[0023] When extracting key correlations between power consumption and speed changes, historical data analysis can reveal a pattern where power consumption increases by approximately 20% as speed increases from 10 m / s to 15 m / s. Environmental variables, such as reduced motor efficiency in high humidity environments, may further increase the consumption rate.

[0024] When analyzing the contribution of task type to speed changes, it can be found that aerial photography tasks, due to frequent adjustments in speed and altitude, have large fluctuations in power consumption, while inspection tasks are relatively stable. Combined with environmental variables such as rising temperature leading to decreased battery efficiency, the influencing factors of the trend can be further clarified.

[0025] When constructing a consumption trend prediction framework, historical data and current environmental variables can be used to predict the changing trend of power consumption rate over the next hour. For example, with continuously increasing wind speed, the predicted consumption rate may rise from 2% to 3% per minute, providing a reference for flight planning.

[0026] Step S102 includes: generating a dynamic adjustment strategy for power consumption based on the prediction results, combined with the influence of variables and the characteristics of the task type, and judging the accuracy of the optimized trend prediction; continuously updating the parameters of the neural network model based on the optimized trend prediction accuracy to obtain a consumption trend output that conforms to the actual scenario.

[0027] When generating a dynamic power consumption adjustment strategy, the flight speed can be reduced from 15 m / s to 12 m / s in high wind conditions based on forecasts, thereby reducing power consumption. This strategy can effectively extend flight time and, combined with mission type characteristics, ensure mission completion quality.

[0028] When continuously updating the parameters of the neural network model, the model weights can be adjusted using real-time flight data to better reflect the actual scenario. For example, after multiple inspection missions, the model gradually adapts to the wind speed variation patterns in a specific area, improving the prediction accuracy from 80% to 85%, thus outputting a more reliable consumption trend.

[0029] Step S103 includes: monitoring the changing trend data in real time through the trend detection module; if the changing trend data exceeds a preset threshold, triggering the information processing stage to obtain a preliminary abnormal signal; based on the preliminary abnormal signal, calling the data integration function to extract relevant data from historical records and determine the feature content matching the changing trend data; for the feature content, analyzing the correlation between historical records and the current trend detection results through the information processing stage to determine the time range for critical prediction; obtaining key time nodes from the time range, and using preset logical rules, if the key time nodes are consistent with the abnormal patterns in historical records, further narrowing the time window for critical prediction; through the narrowed time window, combined with the data integration results, analyzing the persistence of the changing trend data, and determining the time point for predicting the critical point.

[0030] The trend detection module can collect data during device operation in real time. If the power consumption rate of a device suddenly rises from the normal value of 5% per hour to 8% per hour in a short period of time, exceeding the preset threshold of 6%, the system will immediately trigger the information processing stage and generate a preliminary abnormal signal.

[0031] For initial anomaly signals, the data integration function extracts power consumption data from similar time periods over the past 30 days from historical records, filtering out features similar to the current trend, such as records under conditions of high ambient temperature and heavy workload. The system uses these features as the basis for analysis to determine whether the current anomaly is consistent with historical patterns.

[0032] The information processing stage involves in-depth analysis of the correlation between historical records and current trends to determine the time range for critical predictions. For example, if comparisons reveal that similar anomalies typically peak after 3 hours, the system will initially set the time range to within the next 3 hours. Through preset logical rules, key time nodes are further filtered. For instance, if historical anomaly peaks frequently occur in the second hour, the system will prioritize this node, narrowing the critical prediction time window to between 1.5 and 2.5 hours to more accurately pinpoint potential problem moments.

[0033] Step S104 includes: acquiring critical point data and predicted time data from a preset database through a data acquisition module; cleaning and formatting the data to obtain a standardized time prediction dataset; judging the time based on the time prediction dataset using a preset threshold; if the predicted time reaches a critical point, triggering the charging station information acquisition process to determine a list of charging stations that meet the conditions; acquiring the location information and capacity data of each charging station through the charging station list; performing geocoding on the location information to obtain a corresponding coordinate dataset; analyzing the availability of charging stations based on the coordinate dataset and capacity data; if the capacity data is lower than a preset threshold, removing the charging station to obtain an optimized set of available charging stations; calculating the matching degree of each charging station using the set of available charging stations, combined with the predicted time and location information; sorting the matching degrees using the K-nearest neighbor algorithm to determine the charging station with the highest priority; generating a charging timing arrangement sequence based on the highest priority charging station, combined with the predicted time and capacity data; performing time window verification on the sequence to obtain the final charging arrangement result.

[0034] When making threshold judgments based on the time prediction dataset, assuming the preset critical point time is 2 hours in advance for warning, if the predicted time shows that there are only 1.5 hours left until the critical point, the system will automatically trigger the charging station information acquisition process.

[0035] For the processing of location information and capacity data of the charging station list, geocoding can convert the address into latitude and longitude coordinates, and if the capacity data is lower than the preset threshold, such as the minimum capacity requirement being 30% and a station having only 20% capacity remaining, it will be removed.

[0036] After analyzing charging station availability and generating an optimized set, the matching degree can be calculated by combining predicted time and location information. The K-nearest neighbor algorithm can be used to comprehensively rank the stations based on distance and capacity. For example, if a user's current location is 5 kilometers from a charging station with 80% capacity, while another station is 10 kilometers away with 90% capacity, the system will prioritize recommending the closer station based on a weighted rule.

[0037] When generating a charging schedule and verifying the time window, the system may plan for the user to reach the highest priority charging station within the next hour based on the predicted time, and verify whether the charging station is still available within that time window. If a time conflict is found, the schedule will be adjusted, recommending the next best station. This flexible adjustment mechanism can effectively cope with dynamic changes and ensure the feasibility of the schedule.

[0038] From an overall process perspective, the aforementioned stages are interconnected, forming a closed loop from data collection to final scheduling. This enables rapid response to user needs based on critical point prediction, improving resource utilization efficiency. This approach is particularly suitable for resource scheduling in highly dynamic scenarios, providing users with timely and reliable support. like Figure 2 As shown, step S105 includes: acquiring charging demand data for multiple drones, the data including the request time and battery status of each drone, and determining an initial request sequence; evaluating the charging timing of each drone using a priority sorting method based on the initial request sequence, and increasing the priority of a drone if its battery level is below a preset threshold, thus obtaining an adjusted task priority list; analyzing the quantity and location distribution of currently available charging resources using the task priority list and the resource scheduling status of charging stations, and generating a preliminary resource allocation plan; acquiring real-time charging station load data for the preliminary resource allocation plan, and adjusting the resource scheduling strategy if the load of a charging station exceeds a preset limit, thus determining the final allocation plan.

[0039] In managing the charging needs of multiple drones, obtaining the request time and battery status of each drone is the primary step. Suppose there are five drones that initiate charging requests at 8:00 AM, 8:30 AM, 9:00 AM, 9:30 AM, and 10:00 AM, with battery statuses of 20%, 35%, 15%, 40%, and 25%, respectively. This data can be used to create an initial request sequence, reflecting the drones' demand timeline and urgency.

[0040] Regarding the priority ranking of the initial request sequence, assuming a preset battery threshold of 20%, drones with 15% and 20% battery will be prioritized. In the adjusted task priority list, these two drones will be at the top, with the rest arranged in order of request time. This ranking method ensures that drones with extremely low battery receive resources first, preventing task interruption due to battery depletion.

[0041] When analyzing the resource scheduling of charging stations, assume there are three charging stations located in areas A, B, and C, with 2, 1, and 3 charging piles currently available, respectively. A preliminary resource allocation plan is obtained by combining a task priority list and the location distribution of the charging stations.

[0042] In response to adjustments to real-time charging station load data, assuming that the load of charging station A in region A has reached 90%, exceeding the preset upper limit of 80%, the resource scheduling strategy needs to be adjusted to reassign drones originally assigned to region A to charging station C in region A, thus ensuring charging efficiency.

[0043] Further, step S105 includes: processing the charging requests of multiple drones according to the final allocation scheme, updating the charging timing arrangement of each drone in real time, and obtaining a dynamically adjusted timing sequence; monitoring resource scheduling and task priority execution through the dynamically adjusted timing sequence; if the charging demand of a certain drone is not met, re-triggering the priority sorting process, updating the resource allocation scheme and charging timing arrangement, and generating an optimized execution plan.

[0044] When handling multiple drone charging requests and dynamically adjusting the timing sequence, assuming a drone with 15% battery has an estimated charging time of 30 minutes at charging station C, its charging time will be scheduled to begin at 9:15. The schedules for other drones will also be dynamically updated based on real-time conditions, forming a flexible timing sequence. This dynamic adjustment can effectively cope with sudden changes in demand.

[0045] When monitoring resource scheduling and task execution, if it is found that the charging needs of a certain drone are not being met, such as a drone with 25% battery being delayed due to insufficient charging station resources, the priority sorting process will be re-triggered to re-analyze resource distribution and update the allocation plan.

[0046] Step S106 includes: obtaining conflict data from the resource allocation scheme; determining whether there is a scheme conflict through an information processing step to obtain a preliminary conflict detection result; if a scheme conflict is detected, analyzing the impact of the sequence parameters on the conflict data to determine the parameter range that needs to be adjusted; adjusting the sequence parameters according to the determined parameter range using preset logical rules to obtain the adjusted parameter configuration; regenerating the allocation scheme through the adjusted parameter configuration, determining whether there is still a conflict, and obtaining updated scheme data; if there is still a conflict in the updated scheme data, analyzing the cause of the conflict through an information processing step to determine the direction of the secondary adjustment; and optimizing the resource allocation scheme using a support vector machine algorithm according to the direction of the secondary adjustment to obtain the final optimization result.

[0047] In a scenario involving the allocation of charging resources for multiple drones, assuming five drones request charging simultaneously, but only three charging stations are currently available, the system, through information processing, detects that two drones cannot be immediately allocated resources, forming a preliminary conflict detection result. The system records the specific parameters of these conflicts, such as the drone's battery percentage and the geographical distribution of the charging stations.

[0048] To address the impact of conflict data analysis sequence parameters, the priority weights of battery status and request time can be considered. The system may identify drones with less than 30% battery and prioritize resource allocation. Therefore, drones with 20% battery should be prioritized. The parameter range to be adjusted is to increase the priority weight from the default 1.0 to 1.5, while reducing the weight of drones with later request times to 0.8.

[0049] When adjusting sequence parameters using preset logical rules, a simple rule base can be used, such as doubling the priority when the battery level is below 30%. After adjustment, drones with 20% battery are ranked first.

[0050] After regenerating the allocation plan, the system may still find conflicts, such as charging stations being too far away, preventing some drones from reaching them in time. In this case, by analyzing the causes of the conflicts through information processing, it is found that the geographical location parameters were not fully considered. The direction of the secondary adjustment is determined to be to optimize the path planning and prioritize the allocation of charging stations that are closer.

[0051] When using the Support Vector Machine (SVM) algorithm for optimization, the model can be trained using historical data, taking factors such as the drone's battery level, distance, and request time as input features to output the optimal resource allocation combination. Assuming that after optimization, the system will allocate the nearest charging station to the drone with the lowest battery level, ensuring improved resource utilization while reducing drone power consumption during flight.

[0052] Step S107 includes: obtaining the optimized resource allocation result, extracting key fields from the allocation result, and constructing a structured dataset; obtaining real-time feedback data based on the structured dataset, and performing a completeness judgment on the timestamps and event records in the feedback data; if there are missing fields in the feedback data, completing the data using a preset interpolation method to obtain a complete feedback dataset; analyzing the consumption rate change trend using the complete feedback dataset, modeling the change trend using a support vector machine algorithm, and determining the key factors affecting the rate change; updating the model parameters based on the key factors, standardizing the updated parameters to obtain an adjustment parameter set; combining the adjustment parameter set and historical periodic feedback data to predict the consumption rate of the next period, and determining whether the predicted value exceeds a preset threshold range; if the predicted value exceeds the preset threshold range, triggering an anomaly marker to determine potential risk points; generating an adjustment strategy based on the potential risk points and the allocation result, performing logical verification on the adjustment strategy to obtain a final allocation adjustment scheme; updating the resource configuration plan for the next period using the final allocation adjustment scheme, sorting the priority fields in the configuration plan, and determining the execution order for the next period.

[0053] In resource allocation scenarios, after obtaining the optimized allocation results, it is necessary to extract key fields from the results to build a structured dataset. Building a structured dataset helps transform complex allocation information into a clear data view, providing a foundation for real-time feedback.

[0054] To ensure the completeness of real-time feedback data, timestamps and event logs can be checked. Suppose that the feedback data from a resource allocation event records the usage of a device from 8:00 AM to 10:00 AM, but data from 10:00 AM to 12:00 PM is missing. In this case, the system will identify the discontinuity in timestamps and mark it as a missing field. For the missing data, a preset interpolation method can be used to complete the data based on the average usage rate of the preceding and following time periods.

[0055] When analyzing trends in consumption rates, the Support Vector Machine (SVM) algorithm can be used for modeling. Suppose a resource's consumption rate over the past week was 10, 12, 15, 14, 18, 20, and 22 units per day, respectively. The system, through modeling, discovers an increasing trend in the rate and identifies that the key factor might be increased usage frequency. It then updates the model parameters and performs standardization to form an adjusted parameter set.

[0056] When predicting the consumption rate for the next cycle by combining the set of adjustment parameters and historical cycle data, assuming that historical data shows the average consumption rate for the previous three cycles was 15 units / day, while the current predicted value is 25 units / day, exceeding the preset threshold of 20 units / day, the system will trigger an anomaly flag. Further analysis reveals potential risk points, such as a shortage risk caused by a surge in resource demand. An adjustment strategy is generated to address this risk, ultimately forming an allocation adjustment plan.

[0057] When updating the resource allocation plan for the next cycle and sorting the priority field, assuming there are three resource types to be allocated, the system prioritizes them as Type A, Type B, and Type C based on the predicted consumption rate and importance, ensuring that critical resources are prioritized. This sorting method optimizes the execution order and improves resource utilization efficiency.

[0058] The aforementioned stages are closely interconnected, forming a complete resource management closed loop from data construction to predictive analysis and strategy adjustment. Each stage is data-driven, ensuring the rationality and foresight of resource allocation and improving overall operational efficiency.

Claims

1. A method for multi-UAV power consumption prediction and charging optimization, characterized in that, Includes the following steps: S101 collects drone flight data and environmental sensor data in real time, and uses a regression model to process these data to obtain the power consumption rate. S102 obtains the correlation features between task type and environmental variables based on the obtained power consumption rate, and uses a neural network to analyze these features to determine the changing trend of the consumption rate; If the trend of change exceeds the preset threshold, the information processing stage will integrate historical flight records to determine the predicted time of the critical point. S104 determines the charging timing sequence by judging the predicted time of the critical point; S105 arranges the sequence according to the determined charging time, uses a priority sorting algorithm to process multiple drone requests, and obtains a resource allocation scheme; S106 If there is a conflict in the resource allocation scheme, the sequence parameters are adjusted through the information processing stage to obtain an optimized allocation result; Based on the optimized allocation results, S107 obtains real-time feedback data to update the model parameters and determines the consumption rate prediction for the next cycle.

2. The multi-UAV power consumption prediction and charging optimization method of claim 1, wherein, Step S101 includes: By collecting UAV flight data and environmental sensor data in real time, a data processing flow is constructed to obtain the flight data and environmental data after preliminary processing. Based on the flight data and environmental data, a regression model is used to analyze and determine the predicted value of the power consumption rate.

3. The multi-UAV power consumption prediction and charging optimization method of claim 1, wherein, Step S102 includes: Real-time data on power consumption is acquired, categorized and organized according to task type and environmental variables, to obtain a preliminary structured dataset; Based on the structured dataset, a pre-established neural network model is used to process the correlation features between task types and environmental variables, and to determine the mutual influence relationships between features; By analyzing the aforementioned interrelationships, key corresponding points between power consumption and speed changes are extracted to determine the pattern of power consumption speed changes. Based on the changing pattern of the consumption rate, and considering the influence of environmental variables, the contribution of task type to the rate change is analyzed to obtain specific trend influencing factors; Based on the aforementioned trend influencing factors, a predictive framework for consumption trends is constructed to obtain the future direction of speed changes and determine the prediction results.

4. The multi-UAV power consumption prediction and charging optimization method of claim 3, wherein, Step S102 includes: Based on the prediction results, combined with the influence of variables and the characteristics of task types, a dynamic adjustment strategy for power consumption is generated, and the accuracy of the optimized trend prediction is judged. Based on the optimized trend prediction accuracy, the parameters of the neural network model are continuously updated to obtain a consumption trend output that conforms to the actual scenario.

5. The multi-UAV power consumption prediction and charging optimization method of claim 1, wherein, Step S103 includes: The trend detection module monitors the changing trend data in real time. If the changing trend data exceeds a preset threshold, the information processing stage is triggered to obtain a preliminary abnormal signal. Based on the initial abnormal signal, the data integration function is invoked to extract relevant data from historical records and determine the feature content that matches the trend data. For the aforementioned features, the correlation between historical records and current trend detection results is analyzed through information processing to determine the critical prediction time range; Key time nodes are obtained from the time range, and preset logical rules are used. If the key time node is consistent with the abnormal pattern in the historical record, the time window for critical prediction is further narrowed. By using the reduced time window and combining the data integration results, the persistence of the changing trend data is analyzed, and the predicted time point of the critical point is determined.

6. The multi-UAV power consumption prediction and charging optimization method of claim 1, wherein, Step S104 includes: The data acquisition module obtains critical point data and prediction time data from a preset database, and then cleans and formats the data to obtain a standardized time prediction dataset. Based on the time prediction dataset, a preset threshold is used to determine the time. If the predicted time reaches the critical point, the charging station information acquisition process is triggered to determine the list of charging stations that meet the conditions. The location information and capacity data of each charging station are obtained from the charging station list. The location information is then geocoded to obtain the corresponding coordinate dataset. Based on the coordinate dataset and capacity data, the availability of charging stations is analyzed. If the capacity data is lower than a preset threshold, the charging station is removed, and an optimized set of available charging stations is obtained. Using the available charging station set, combined with predicted time and location information, the matching degree of each charging station is calculated, and the K-nearest neighbor algorithm is used to sort the matching degrees to determine the charging station with the highest priority. Based on the highest priority charging station, and combined with predicted time and capacity data, a charging timing sequence is generated. The sequence is then validated using a time window to obtain the final charging arrangement result.

7. The multi-UAV power consumption prediction and charging optimization method of claim 1, wherein, Step S105 includes: Acquire charging demand data for multiple drones, including the request time and battery status of each drone, and determine the initial request sequence; Based on the initial request sequence, a priority sorting method is used to evaluate the charging timing of each drone. If the battery level of a drone is lower than a preset threshold, its priority is increased, resulting in an adjusted task priority list. By analyzing the number and location distribution of currently available charging resources using the task priority list and the resource scheduling status of charging stations, a preliminary resource allocation plan is generated. For the preliminary resource allocation scheme, real-time charging station load data is obtained. If the load of a charging station exceeds the preset limit, the resource scheduling strategy is adjusted to determine the final allocation scheme.

8. The multi-UAV power consumption prediction and charging optimization method of claim 7, wherein, Step S105 includes: Based on the final allocation scheme, the charging requests of multiple drones are processed, and the charging timing arrangement of each drone is updated in real time to obtain a dynamically adjusted timing sequence. By using the dynamically adjusted timing sequence, the resource scheduling and task priority execution are monitored. If the charging needs of a certain drone are not met, the priority sorting process is re-triggered to update the resource allocation scheme and charging timing arrangement, and generate an optimized execution plan.

9. The multi-UAV power consumption prediction and charging optimization method of claim 1, wherein, Step S106 includes: Conflict data is obtained from resource allocation schemes, and the existence of scheme conflicts is determined through information processing to obtain preliminary conflict detection results; If a conflict is detected, the impact of the conflict data analysis sequence parameters is determined to identify the range of parameters that need to be adjusted. Based on the determined parameter range, the sequence parameters are adjusted using preset logical rules to obtain the adjusted parameter configuration; The allocation scheme is regenerated using the adjusted parameter configuration, and it is determined whether there are still conflicts to obtain the updated scheme data. If the updated scheme data still has conflicts, the cause of the conflicts will be analyzed through the information processing stage to determine the direction of the second adjustment; Based on the direction of the secondary adjustment, the resource allocation scheme is optimized using the support vector machine algorithm to obtain the final optimization result.

10. The multi-UAV power consumption prediction and charging optimization method of claim 1, wherein, Step S107 includes: Obtain the optimized resource allocation results, extract key fields from the allocation results, and construct a structured dataset; Based on the structured dataset, obtain real-time feedback data, and perform integrity checks on the timestamps and event records in the feedback data; If there are missing fields in the feedback data, the data is filled in using a preset interpolation method to obtain a complete feedback dataset; Using the complete feedback dataset, the consumption rate change trend is analyzed, and the support vector machine algorithm is used to model the change trend to determine the key factors affecting the rate change. Based on the key factors, the model parameters are updated, and the updated parameters are standardized to obtain an adjusted parameter set. By combining the set of adjustment parameters and historical cycle feedback data, the consumption rate of the next cycle is predicted, and it is determined whether the predicted value exceeds the preset threshold range. If the predicted value exceeds the preset threshold range, an anomaly marker is triggered to identify potential risk points; Based on the potential risk points and the allocation results, an adjustment strategy is generated, and the adjustment strategy is logically verified to obtain the final allocation adjustment scheme. Based on the final allocation adjustment scheme, the resource allocation plan for the next cycle is updated, and the priority field in the allocation plan is sorted to determine the execution order for the next cycle.

Images