A method for predicting future traffic volume and energy consumption of drones in the transportation industry

Abstract

The present invention discloses a method for predicting the future traffic volume and energy consumption of unmanned aerial vehicles in the transportation industry, and belongs to the technical field of air traffic planning. It solves the problem that the traditional unmanned aerial vehicle traffic monitoring method in the prior art does not involve considering multiple influencing factors to predict the traffic volume of unmanned aerial vehicles; the present invention constructs an index system for influencing the future traffic volume of unmanned aerial vehicles in the transportation industry, obtains an element matrix corresponding to the influencing index, inputs it into the future traffic volume prediction model of unmanned aerial vehicles for training, and outputs the estimated traffic volume of unmanned aerial vehicles; establishes an element matrix corresponding to the influencing factors of unmanned aerial vehicle energy consumption, and obtains the element matrix of unmanned aerial vehicle flight energy consumption based on the power data collected by the power sensor; constructs an unmanned aerial vehicle energy consumption calculation model, outputs the estimated energy consumption per unit time of the unmanned aerial vehicle, and combines the estimated unmanned aerial vehicle traffic volume to obtain the energy consumption under the future unmanned aerial vehicle traffic volume. The present invention effectively improves air traffic safety and can be applied to predicting the traffic volume of unmanned aerial vehicles.

Description

Technical Field

[0001] The invention relates to a method for predicting future traffic volume and energy consumption of unmanned aerial vehicles in the transportation industry, and belongs to the technical field of air traffic planning. Background Art

[0002] With the development of drone technology and the expansion of commercial scope, air traffic will become increasingly busy. Effective traffic forecasting can help relevant agencies to make reasonable airspace planning, avoid air traffic congestion, and ensure the safe operation of drones. The widespread use of drones requires a lot of energy and resource support, including power supply, charging facilities, and maintenance bases. By predicting future traffic volume, the layout and scale of these resources can be planned more effectively, energy utilization can be optimized, and operating costs can be reduced. Accurately predicting the future traffic volume and flow of drones can enhance safety management and emergency response capabilities.

[0003] On the other hand, in the transportation industry, drones are an important means of transportation, and the prediction of their energy consumption is of great significance to the operational efficiency and environmental impact of the entire industry. By accurately predicting the energy consumption of future drone traffic, multi-faceted optimization and cost control can be achieved. First, energy suppliers can more reasonably plan the production and distribution of electricity or other energy based on forecast data, ensure the timeliness and adequacy of energy supply, and avoid waste of resources. This not only reduces operating costs, but also mitigates the impact of drones on the environment, such as reducing carbon emissions and energy consumption. This prediction and optimization mechanism promotes the efficient use of energy, enhances the industry's energy management capabilities, and supports sustainable development goals.

[0004] In the prior art 1, the patent document with publication number CN102208132A discloses a traffic volume prediction device and a traffic volume prediction method. The device stores the existing vehicle speed records and the traffic volume change pattern. When new vehicle speed data is input, the device will update the pattern data. When predicting the traffic volume, the device will check the historical pattern most similar to the current data, and then predict the next traffic situation based on the most matching pattern; in the prior art 2, the patent document with publication number CN116386312A discloses a method and system for constructing a traffic volume prediction model. By preprocessing the traffic data and dividing it into a training set, a validation set and a test set, the system analyzes the spatial relationship between the road section nodes by constructing an adjacency matrix, and extracts time series data to understand the dependency between different time points. The spatial and temporal dependency is processed through a neural network structure of a fully connected layer to construct a traffic volume prediction model; however, the above method does not involve the prediction of drone traffic volume, and does not consider the impact of different factors on drone traffic volume.

[0005] Therefore, a method for predicting future traffic volume and energy consumption of drones in the transportation industry is needed. Summary of the invention

[0006] A brief overview of the present invention is provided below in order to provide a basic understanding of certain aspects of the present invention. It should be understood that this overview is not an exhaustive overview of the present invention. It is not intended to identify key or important parts of the present invention, nor is it intended to limit the scope of the present invention. Its purpose is merely to present certain concepts in a simplified form as a prelude to a more detailed description discussed later.

[0007] In view of this, in order to solve the problem that the traditional UAV traffic monitoring method in the prior art does not involve considering multiple influencing factors to predict the UAV traffic volume, the present invention provides a method for predicting the future traffic volume and energy consumption of UAVs in the transportation industry.

[0008] The technical solution is as follows: A method for predicting future traffic volume and energy consumption of drones in the transportation industry, comprising the following steps:

[0009] S1. Based on the data of the influencing indicators, a system of influencing indicators of future UAV traffic volume in the transportation industry is constructed to obtain the element matrix corresponding to the influencing indicators;

[0010] S2. Construct a future UAV traffic volume prediction model, input the element matrix corresponding to the influencing index to train the model, and output the estimated UAV traffic volume;

[0011] S3. Collect data based on the factors affecting the energy consumption of the UAV, establish an element matrix corresponding to the factors affecting the energy consumption of the UAV, and obtain the element matrix of the UAV flight energy consumption based on the power data collected by the power sensor;

[0012] S4. Construct a UAV energy consumption calculation model, input the element matrix corresponding to the factors affecting UAV energy consumption to train the model, output the estimated energy consumption per unit time of the UAV, and combine it with the estimated UAV traffic volume to obtain the energy consumption under future UAV traffic volume.

[0013] Furthermore, in said S1, the influencing indicators include policy restriction index, technology maturity index, market demand index, weather condition index, airspace utilization rate, social acceptance index and accident rate;

[0014] The element matrix of the policy restriction index is , , an is the total number of drone traffic surveys, is the policy restriction index corresponding to the an-th acquisition of drone traffic volume;

[0015] The technology maturity index element matrix is: , , is the technology maturity index corresponding to the an-th acquisition of drone traffic volume;

[0016] The market demand index element matrix is , , is the market demand index corresponding to the an-th acquisition of drone traffic volume;

[0017] The weather condition index element matrix is , , is the weather condition index corresponding to the an-th acquisition of drone traffic;

[0018] The airspace utilization element matrix is: , , is the airspace utilization rate corresponding to the an-th acquisition of drone traffic volume;

[0019] The social acceptance index element matrix is , , is the social acceptance index corresponding to the an-th acquisition of drone traffic volume;

[0020] The accident rate element matrix is , , is the accident rate corresponding to the an-th acquisition of drone traffic.

[0021] Further, in said S2, the future traffic volume prediction model of drones includes an input layer, 8 hidden layers, 2 self-attention layers, 2 residual layers and an output layer;

[0022] The neural network structure of the established drone future traffic volume prediction model is connected in the following order:

[0023] Input layer: the policy restriction index element matrix , Technology Maturity Index Element Matrix , Market demand index element matrix , Weather Condition Index Element Matrix , airspace utilization element matrix , Social Acceptance Index Element Matrix and the accident rate element matrix As input;

[0024] Hidden layer 1: Number of neurons: 1024, activation function: ReLU, batch normalization;

[0025] Hidden layer 2: number of neurons: 512, activation function: ReLU, batch normalization;

[0026] Self-attention layer 1: The output matrix of hidden layer 2 is ; Initialize the weight matrix and establish the first initial weight query matrix , the first initial weight key matrix and the first initial weight matrix , the output matrix of hidden layer 2 Transformed into the first query matrix , first bond matrix , the first value matrix , , , ; Calculate the first query matrix With the first key matrix The similarity of , get the first attention score matrix , ,in, is the first attention score matrix Corresponding scaling factor; using the softmax function to the first attention score matrix Perform normalization so that the sum of the elements in each row is 1, and obtain the first normalized score matrix , ; First normalized score matrix Capture the output matrix of hidden layer 2 The internal relations of the layer and serve as the input of the next layer;

[0027] Hidden layer 3: number of neurons: 256, activation function: ReLU, batch normalization;

[0028] Hidden layer 4: number of neurons: 128, activation function: ReLU, batch normalization;

[0029] Residual layer 1: The output of hidden layer 1 is connected to the output of hidden layer 3 through residual connection.

[0030] Hidden layer 5: Number of neurons: 64, activation function: ReLU, batch normalization;

[0031] Hidden layer 6: number of neurons: 32, activation function: ReLU, batch normalization;

[0032] Self-attention layer 2: The output matrix of hidden layer 6 is ; Initialize the weight matrix and establish the second initial weight query matrix , the second initial weight key matrix and the second initial weight matrix , the output matrix of hidden layer 6 Transformed into the second query matrix , the second key matrix and the second value matrix , , , ; Calculate the second query matrix With the second key matrix The similarity of , get the second attention score matrix , ,in, is the second attention score matrix Corresponding scaling factor; using softmax function for the second attention score matrix Perform normalization so that the sum of the elements in each row is 1, and obtain the second normalized score matrix , ; The second normalized score matrix Capture the output matrix of hidden layer 6 The internal relations of the layer and serve as the input of the next layer;

[0033] Hidden layer 7: Number of neurons: 16, activation function: ReLU, batch normalization;

[0034] Residual layer 2: The output of hidden layer 4 is connected to the output of hidden layer 6 through residual connection;

[0035] Hidden layer 8: number of neurons: 8, activation function: ReLU, batch normalization;

[0036] Output layer: Number of neurons: 1, the output layer is the drone traffic element matrix , , is the drone traffic volume obtained for the anth time. After the model training is completed, the output layer outputs the estimated drone traffic volume .

[0037] Further, in S3, the power sensors installed in the WN drones are used to collect relevant data on factors affecting the energy consumption of the drones, and the factors affecting the energy consumption of the drones include motor efficiency, propeller diameter, fuselage mass, load, flight speed, flight mode, wind speed, wind direction, temperature, humidity and flight altitude;

[0038] The motor efficiency element matrix corresponding to the motor efficiency is: , , For the Motor efficiency corresponding to each drone;

[0039] The propeller diameter element matrix corresponding to the propeller diameter is , , For the Propeller diameter corresponding to each drone;

[0040] The fuselage mass element matrix corresponding to the fuselage mass is: , , For the The body mass corresponding to each drone;

[0041] Load element matrix corresponding to load , , For the The corresponding load of each drone;

[0042] The flight speed element matrix corresponding to the flight speed is: , , For the The flight speed of each drone;

[0043] The flight mode element matrix corresponding to the flight mode is , , For the The flight modes corresponding to each drone;

[0044] The wind speed element matrix corresponding to the wind speed is , , For the Wind speed corresponding to each drone;

[0045] The wind direction element matrix corresponding to the wind direction is , , For the The wind direction corresponding to each drone;

[0046] The temperature element matrix corresponding to the temperature is , , For the The temperature corresponding to each drone;

[0047] The humidity element matrix corresponding to humidity is , , For the Humidity corresponding to each drone;

[0048] The flight height element matrix corresponding to the flight height, , For the The flight altitude corresponding to each drone;

[0049] According to the element matrix corresponding to the factors affecting the energy consumption of UAV, the power data collected based on the power sensor , calculate the energy consumption of the drone during flight ;

[0050] Energy consumption during drone flight It is expressed as:

[0051] ;

[0052] Among them, t is the flight time of the UAV, is the total flight time of the drone;

[0053] Calculate the drone flight energy consumption per unit time corresponding to all drones, and establish the drone flight energy consumption element matrix AE. ,in, is the energy consumption of the drone flight per unit time, .

[0054] Further, in said S4, the drone energy consumption calculation model includes an input layer, 7 hidden layers, a self-attention layer and an output layer;

[0055] The neural network structure of the established UAV energy consumption calculation model is connected in the following order:

[0056] Input layer: The motor efficiency element matrix , propeller diameter element matrix , fuselage mass element matrix , load element matrix , flight speed element matrix , flight mode element matrix , wind speed element matrix , Wind direction element matrix , temperature element matrix , humidity element matrix and the flight height element matrix As input;

[0057] Hidden layer 1: Number of neurons: 512, activation function: ReLU, batch normalization;

[0058] Hidden layer 2: number of neurons: 256, activation function: ReLU, batch normalization;

[0059] Self-attention layer: The output matrix of hidden layer 2 is ; Initialize the weight matrix, where the third initial weight query matrix is ​​established in turn , the third initial weight key matrix and the third initial weight matrix , the output matrix of hidden layer 2 Transformed into the third query matrix , the third key matrix and the third value matrix , , , ; Get the third query matrix With the third key matrix The similarity of , and the third attention score matrix is ​​obtained , ,in, is the third attention score matrix Corresponding scaling factor; using softmax function to the third attention score matrix Perform normalization so that the sum of the elements in each row is 1, and obtain the third normalized score matrix , ; The third normalized score matrix Capture the output matrix of hidden layer 2 The internal relations of the layer and serve as the input of the next layer;

[0060] Hidden layer 3: Number of neurons: 128, activation function: ReLU, batch normalization;

[0061] Residual layer 1: The output of hidden layer 1 is connected to the output of hidden layer 3 through residual connection.

[0062] Hidden layer 4: number of neurons: 64, activation function: ReLU, batch normalization;

[0063] Hidden layer 5: number of neurons: 32, activation function: ReLU, batch normalization;

[0064] Hidden layer 6: number of neurons: 16, activation function: ReLU, batch normalization;

[0065] Residual layer 2: The output of hidden layer 4 is connected to the output of hidden layer 6 through residual connection;

[0066] Hidden layer 7: Number of neurons: 8, activation function: ReLU, batch normalization;

[0067] Output layer: Number of neurons: 1, i.e., the drone flight energy consumption element matrix AE. After the model training is completed, the output layer outputs the estimated energy consumption per unit time of the drone. ;

[0068] Based on estimated drone traffic and the estimated energy consumption per unit time of the drone , the energy consumption ZAE under future UAV traffic volume is calculated;

[0069] The energy consumption ZAE under future drone traffic is expressed as:

[0070] .

[0071] The beneficial effects of the present invention are as follows: the present invention constructs an index system for the impact of future drone traffic volume in the transportation industry, providing quantitative analysis indicators for the prediction of future drone traffic volume; the present invention predicts the future drone traffic volume by constructing a drone future traffic volume prediction model, which helps to improve air traffic safety and efficiency; the present invention performs a method for calculating energy consumption under future drone traffic volume based on the estimated future drone traffic volume, which can improve the environmental friendliness of the development process of the drone industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0073] Figure 1 The figure is a flowchart of a method for predicting future traffic volume and energy consumption of drones in the transportation industry. DETAILED DESCRIPTION

[0074] In order to make the technical solutions and advantages of the embodiments of the present invention more clearly understood, the exemplary embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than an exhaustive list of all the embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0075] refer to Figure 1 The present embodiment is described in detail, a method for predicting future traffic volume and energy consumption of drones in the transportation industry, specifically comprising the following steps:

[0076] S1. Based on the data of the influencing indicators, a system of influencing indicators of future UAV traffic volume in the transportation industry is constructed to obtain the element matrix corresponding to the influencing indicators;

[0077] S2. Construct a future UAV traffic volume prediction model, input the element matrix corresponding to the influencing index to train the model, and output the estimated UAV traffic volume;

[0078] S3. Collect data based on the factors affecting the energy consumption of the UAV, establish an element matrix corresponding to the factors affecting the energy consumption of the UAV, and obtain the element matrix of the UAV flight energy consumption based on the power data collected by the power sensor;

[0079] S4. Construct a UAV energy consumption calculation model, input the element matrix corresponding to the factors affecting UAV energy consumption to train the model, output the estimated energy consumption per unit time of the UAV, and combine it with the estimated UAV traffic volume to obtain the energy consumption under future UAV traffic volume.

[0080] Furthermore, in said S1, the influencing indicators include policy restriction index, technology maturity index, market demand index, weather condition index, airspace utilization rate, social acceptance index and accident rate;

[0081] (1) Regulations and policies: Regulations and policies related to drones stipulate the operating rules of drones, including flight altitude, no-fly zones, flight time, etc. These regulations directly restrict the use of drones in specific areas and times, thereby affecting traffic volume. The policy restriction index is used to characterize the strictness of drone operation regulations and policies. The larger the policy restriction index, the stricter the restrictions on drones. The policy restriction index is obtained by expert scoring and is 1, 2, 3, 4 or 5;

[0082] The element matrix of the policy restriction index is , , an is the total number of drone traffic surveys, is the policy restriction index corresponding to the an-th acquisition of drone traffic volume;

[0083] (2) Technology development level: The technological progress of UAVs (such as navigation technology, automatic obstacle avoidance system, and communication technology) can improve the operating efficiency and safety of UAVs, allowing more UAVs to operate safely in the same airspace, which will have an impact on traffic volume. The technology maturity index is used to score the technology development level of UAVs through expert scoring. The score range is 1-10. The smaller the score, the lower the technology maturity.

[0084] The technology maturity index element matrix is: , , is the technology maturity index corresponding to the an-th acquisition of drone traffic volume;

[0085] (3) Economic demand: With the continuous development of the economy and the transportation industry, the demand for drones will change, and the volume of drone traffic will also change. The market demand index is used through expert scoring to represent the demand for drones in the economic situation. The score range is 1-10. The smaller the score, the lower the market demand.

[0086] The market demand index element matrix is , , is the market demand index corresponding to the an-th acquisition of drone traffic volume;

[0087] (4) Weather conditions: Good weather conditions can increase the number of drone flights, while unfavorable weather conditions (such as strong winds, rain, snow, low visibility, etc.) will limit drone flights and reduce traffic. The weather conditions are characterized by a weather condition index through expert scoring, with a score range of 1-10. The smaller the score, the less suitable the weather conditions are for drone flights.

[0088] The weather condition index element matrix is , , is the weather condition index corresponding to the an-th acquisition of drone traffic;

[0089] (5) Airspace management level: A higher airspace management level will maximize the operating efficiency of drones, reasonably allocate the flight plans and routes of drones, avoid congestion, and thus increase the volume of drone traffic in the airspace. Based on data obtained from public channels, the airspace utilization rate is used to characterize the airspace management level. A high airspace utilization rate indicates a high airspace management level.

[0090] The airspace utilization element matrix is: , , is the airspace utilization rate corresponding to the an-th acquisition of drone traffic volume;

[0091] (6) Social acceptance: The public's concerns about drone privacy infringement, noise interference and other issues will limit the widespread use of drones and affect traffic volume. The higher the society's acceptance of drone technology, the wider its application scope and the higher its frequency of use in the transportation industry. The social acceptance index is used to represent the social acceptance of drones through expert scoring. The score range is 1-10. The lower the score, the lower the social acceptance of drones.

[0092] The social acceptance index element matrix is , , is the social acceptance index corresponding to the an-th acquisition of drone traffic volume;

[0093] (7) Driving safety: The safety accident problem of drones has increased the supervision of drone operations and has a certain restrictive effect on drone traffic. Based on data obtained from public channels, the accident rate is used to represent the driving safety of drones. The higher the accident rate, the worse the flight safety of drones.

[0094] Construct the accident rate element matrix as , , is the accident rate corresponding to the an-th acquisition of drone traffic.

[0095] Further, in said S2, the future traffic volume prediction model of drones includes an input layer, 8 hidden layers, 2 self-attention layers, 2 residual layers and an output layer;

[0096] The neural network structure of the established drone future traffic volume prediction model is connected in the following order:

[0097] Input layer: the policy restriction index element matrix , Technology Maturity Index Element Matrix , Market demand index element matrix , Weather Condition Index Element Matrix , airspace utilization element matrix , Social Acceptance Index Element Matrix and the accident rate element matrix As input;

[0098] Hidden layer 1: Number of neurons: 1024, activation function: ReLU, batch normalization;

[0099] Hidden layer 2: number of neurons: 512, activation function: ReLU, batch normalization;

[0100] Self-attention layer 1: The output matrix of hidden layer 2 is ; Initialize the weight matrix and establish the first initial weight query matrix , the first initial weight key matrix and the first initial weight matrix , the output matrix of hidden layer 2 Transformed into the first query matrix , first bond matrix , the first value matrix , , , ; Calculate the first query matrix With the first key matrix The similarity of , get the first attention score matrix , ,in, is the first attention score matrix Corresponding scaling factor; using the softmax function to the first attention score matrix Perform normalization so that the sum of the elements in each row is 1, and obtain the first normalized score matrix , ; First normalized score matrix Capture the output matrix of hidden layer 2 The internal relations of the layer and serve as the input of the next layer;

[0101] Hidden layer 3: number of neurons: 256, activation function: ReLU, batch normalization;

[0102] Hidden layer 4: number of neurons: 128, activation function: ReLU, batch normalization;

[0103] Residual layer 1: The output of hidden layer 1 is connected to the output of hidden layer 3 through residual connection.

[0104] Hidden layer 5: Number of neurons: 64, activation function: ReLU, batch normalization;

[0105] Hidden layer 6: number of neurons: 32, activation function: ReLU, batch normalization;

[0106] Self-attention layer 2: The output matrix of hidden layer 6 is ; Initialize the weight matrix and establish the second initial weight query matrix , the second initial weight key matrix and the second initial weight matrix , the output matrix of hidden layer 6 Transformed into the second query matrix , the second key matrix and the second value matrix , , , ; Calculate the second query matrix With the second key matrix The similarity of , get the second attention score matrix , ,in, is the second attention score matrix Corresponding scaling factor; using softmax function for the second attention score matrix Perform normalization so that the sum of the elements in each row is 1, and obtain the second normalized score matrix , ; The second normalized score matrix Capture the output matrix of hidden layer 6 The internal relations of the layer and serve as the input of the next layer;

[0107] Hidden layer 7: Number of neurons: 16, activation function: ReLU, batch normalization;

[0108] Residual layer 2: The output of hidden layer 4 is connected to the output of hidden layer 6 through residual connection;

[0109] Hidden layer 8: number of neurons: 8, activation function: ReLU, batch normalization;

[0110] Output layer: Number of neurons: 1, the output layer is the drone traffic element matrix JL, , is the drone traffic volume obtained for the anth time. After the model training is completed, the output layer outputs the estimated drone traffic volume .

[0111] Further, in S3, based on The power sensors installed in each drone collect data related to factors affecting drone energy consumption, including motor efficiency, propeller diameter, fuselage mass, load, flight speed, flight mode, wind speed, wind direction, temperature, humidity and flight altitude.

[0112] (1) Motor efficiency: Motor efficiency refers to the efficiency of the motor in converting electrical energy into mechanical energy. The motor efficiency element matrix corresponding to the motor efficiency is: , , For the Motor efficiency corresponding to each drone;

[0113] (2) Propeller diameter: The propeller size affects the thrust and flight efficiency of the UAV. The propeller diameter element matrix corresponding to the propeller diameter is: , , For the Propeller diameter corresponding to each drone;

[0114] (3) Fuselage mass: The fuselage mass affects the take-off and flight stability of the UAV. The fuselage mass element matrix corresponding to the fuselage mass is: , , For the The body mass corresponding to each drone;

[0115] (4) Load: Load refers to the weight that a drone can carry in addition to its own weight. The load element matrix corresponding to the load is , , For the The corresponding load of each drone;

[0116] (5) Flight speed: An increase in flight speed usually leads to an increase in wind resistance, which requires more energy to maintain flight. The flight speed element matrix corresponding to the flight speed is: , , For the The flight speed of each drone;

[0117] (6) Flight mode: The flight mode determines different energy consumption. The flight modes usually include manual mode, stable mode, hovering mode, heading mode, return mode, follow mode, waypoint flight and track flight. For quantitative comparison, the above modes are coded as 1-8 in sequence. The flight mode element matrix corresponding to the flight mode is: , , For the The flight modes corresponding to each drone;

[0118] (7) Wind speed: Wind speed will affect the flight efficiency of the UAV. The wind speed element matrix corresponding to the wind speed is: , , For the Wind speed corresponding to each drone;

[0119] (8) Wind direction: Wind direction can affect the flight path and energy consumption of the UAV. The wind direction angle is used to quantitatively analyze and compare the wind direction. The wind direction element matrix corresponding to the wind direction is: , , For the The wind direction corresponding to each drone;

[0120] (9) Temperature: Temperature affects battery performance and air density, which in turn affects energy consumption. The temperature element matrix corresponding to temperature is: , , For the The temperature corresponding to each drone;

[0121] (10) Humidity: High humidity will affect the efficiency and energy consumption of electronic equipment. The humidity element matrix corresponding to humidity is: , , For the Humidity corresponding to each drone;

[0122] (11) Flight altitude: Flight altitude affects air density and flight efficiency, and thus energy consumption. The flight altitude element matrix corresponding to the flight altitude is: , , For the The flight altitude corresponding to each drone;

[0123] According to the element matrix corresponding to the factors affecting the energy consumption of UAV, the power data collected based on the power sensor , calculate the energy consumption of the drone during flight ;

[0124] Energy consumption during drone flight It is expressed as:

[0125] ;

[0126] Among them, t is the flight time of the UAV, is the total flight time of the drone;

[0127] Calculate the drone flight energy consumption per unit time corresponding to all drones, and establish the drone flight energy consumption element matrix AE. ,in, is the energy consumption of the drone flight per unit time, .

[0128] Further, in said S4, the drone energy consumption calculation model includes an input layer, 7 hidden layers, a self-attention layer and an output layer;

[0129] The neural network structure of the established UAV energy consumption calculation model is connected in the following order:

[0130] Input layer: The motor efficiency element matrix , propeller diameter element matrix , fuselage mass element matrix , load element matrix , flight speed element matrix , flight mode element matrix , wind speed element matrix , Wind direction element matrix , temperature element matrix , humidity element matrix and the flight height element matrix As input;

[0131] Hidden layer 1: Number of neurons: 512, activation function: ReLU, batch normalization;

[0132] Hidden layer 2: number of neurons: 256, activation function: ReLU, batch normalization;

[0133] Self-attention layer: The output matrix of hidden layer 2 is ; Initialize the weight matrix, where the third initial weight query matrix is ​​established in turn , the third initial weight key matrix and the third initial weight matrix , the output matrix of hidden layer 2 Transformed into the third query matrix , the third key matrix and the third value matrix , , , ; Get the third query matrix With the third key matrix The similarity of , and the third attention score matrix is ​​obtained , ,in, is the third attention score matrix Corresponding scaling factor; using softmax function to the third attention score matrix Perform normalization so that the sum of the elements in each row is 1, and obtain the third normalized score matrix , ; The third normalized score matrix Capture the output matrix of hidden layer 2 The internal relations of the layer and serve as the input of the next layer;

[0134] Hidden layer 3: Number of neurons: 128, activation function: ReLU, batch normalization;

[0135] Residual layer 1: The output of hidden layer 1 is connected to the output of hidden layer 3 through residual connection.

[0136] Hidden layer 4: number of neurons: 64, activation function: ReLU, batch normalization;

[0137] Hidden layer 5: number of neurons: 32, activation function: ReLU, batch normalization;

[0138] Hidden layer 6: number of neurons: 16, activation function: ReLU, batch normalization;

[0139] Residual layer 2: The output of hidden layer 4 is connected to the output of hidden layer 6 through residual connection;

[0140] Hidden layer 7: Number of neurons: 8, activation function: ReLU, batch normalization;

[0141] Output layer: Number of neurons: 1, i.e., the drone flight energy consumption element matrix AE. After the model training is completed, the output layer outputs the estimated energy consumption per unit time of the drone. ;

[0142] Based on estimated drone traffic and the estimated energy consumption per unit time of the drone , the energy consumption ZAE under future UAV traffic volume is calculated;

[0143] The energy consumption ZAE under future drone traffic is expressed as:

[0144] .

[0145] Although the present invention has been described according to a limited number of embodiments, it will be apparent to those skilled in the art, with the benefit of the above description, that other embodiments may be envisioned within the scope of the invention thus described. In addition, it should be noted that the language used in this specification is selected primarily for readability and teaching purposes, rather than for explaining or defining the subject matter of the present invention. Therefore, many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the appended claims. The disclosure of the present invention is illustrative, not restrictive, with respect to the scope of the present invention, which is defined by the appended claims.

Claims

1. A method for predicting future traffic volume and energy consumption of UAVs in the transportation industry, characterized in that: The following steps are involved: S1. Based on the data of the influencing indicators, a system of influencing indicators of future UAV traffic volume in the transportation industry is constructed to obtain the element matrix corresponding to the influencing indicators; S2. Construct a future UAV traffic volume prediction model, input the element matrix corresponding to the influencing index to train the model, and output the estimated UAV traffic volume; S3. Collect data based on the factors affecting the energy consumption of the UAV, establish an element matrix corresponding to the factors affecting the energy consumption of the UAV, and obtain the element matrix of the UAV flight energy consumption based on the power data collected by the power sensor; S4. Construct a UAV energy consumption calculation model, input the element matrix corresponding to the factors affecting UAV energy consumption to train the model, output the estimated energy consumption per unit time of the UAV, and combine it with the estimated UAV traffic volume to obtain the energy consumption under future UAV traffic volume.

2. The method for predicting future traffic volume and energy consumption of UAVs in the transportation industry according to claim 1 is characterized in that: In S1, the influencing indicators include policy restriction index, technology maturity index, market demand index, weather condition index, airspace utilization rate, social acceptance index and accident rate; The element matrix of the policy restriction index is , , an is the total number of drone traffic surveys, is the policy restriction index corresponding to the an-th acquisition of drone traffic volume; The technology maturity index element matrix is , , is the technology maturity index corresponding to the an-th acquisition of drone traffic volume; The market demand index element matrix is , , is the market demand index corresponding to the an-th acquisition of drone traffic volume; The weather condition index element matrix is , , is the weather condition index corresponding to the an-th acquisition of drone traffic; The airspace utilization element matrix is: , , is the airspace utilization rate corresponding to the an-th acquisition of drone traffic volume; The social acceptance index element matrix is , , is the social acceptance index corresponding to the an-th acquisition of drone traffic volume; The accident rate element matrix is , , is the accident rate corresponding to the an-th acquisition of drone traffic.

3. The method for predicting future traffic volume and energy consumption of UAVs in the transportation industry according to claim 2 is characterized in that: In S2, the future traffic volume prediction model of drones includes an input layer, 8 hidden layers, 2 self-attention layers, 2 residual layers and an output layer; The neural network structure of the established drone future traffic volume prediction model is connected in the following order: Input layer: the policy restriction index element matrix , Technology Maturity Index Element Matrix , Market demand index element matrix , Weather Condition Index Element Matrix , airspace utilization element matrix , Social Acceptance Index Element Matrix and the accident rate element matrix As input; Hidden layer 1: Number of neurons: 1024, activation function: ReLU, batch normalization; Hidden layer 2: number of neurons: 512, activation function: ReLU, batch normalization; Self-attention layer 1: The output matrix of hidden layer 2 is ; Initialize the weight matrix and establish the first initial weight query matrix , the first initial weight key matrix and the first initial weight matrix , the output matrix of hidden layer 2 Transformed into the first query matrix , first bond matrix , the first value matrix , , , ; Calculate the first query matrix With the first key matrix The similarity of , get the first attention score matrix , ,in, is the first attention score matrix Corresponding scaling factor; using the softmax function to the first attention score matrix Perform normalization so that the sum of the elements in each row is 1, and obtain the first normalized score matrix , ; First normalized score matrix Capture the output matrix of hidden layer 2 The internal relations of the layer and serve as the input of the next layer; Hidden layer 3: number of neurons: 256, activation function: ReLU, batch normalization; Hidden layer 4: number of neurons: 128, activation function: ReLU, batch normalization; Residual layer 1: The output of hidden layer 1 is connected to the output of hidden layer 3 through residual connection. Hidden layer 5: Number of neurons: 64, activation function: ReLU, batch normalization; Hidden layer 6: number of neurons: 32, activation function: ReLU, batch normalization; Self-attention layer 2: The output matrix of hidden layer 6 is ; Initialize the weight matrix and establish the second initial weight query matrix , the second initial weight key matrix and the second initial weight matrix , the output matrix of hidden layer 6 Transformed into the second query matrix , the second key matrix and the second value matrix , , , ; Calculate the second query matrix With the second key matrix The similarity of , get the second attention score matrix , ,in, is the second attention score matrix Corresponding scaling factor; using softmax function for the second attention score matrix Perform normalization so that the sum of the elements in each row is 1, and obtain the second normalized score matrix , ; The second normalized score matrix Capture the output matrix of hidden layer 6 The internal relations of the layer and serve as the input of the next layer; Hidden layer 7: Number of neurons: 16, activation function: ReLU, batch normalization; Residual layer 2: The output of hidden layer 4 is connected to the output of hidden layer 6 through residual connection; Hidden layer 8: number of neurons: 8, activation function: ReLU, batch normalization; Output layer: Number of neurons: 1, the output layer is the drone traffic element matrix , , is the drone traffic volume obtained for the anth time. After the model training is completed, the output layer outputs the estimated drone traffic volume .

4. The method for predicting future traffic volume and energy consumption of UAVs in the transportation industry according to claim 3 is characterized in that: In S3, based on The power sensors installed in each drone collect data related to factors affecting drone energy consumption, including motor efficiency, propeller diameter, fuselage mass, load, flight speed, flight mode, wind speed, wind direction, temperature, humidity and flight altitude. The motor efficiency element matrix corresponding to the motor efficiency is: , , For the Motor efficiency corresponding to each drone; The propeller diameter element matrix corresponding to the propeller diameter is , , For the Propeller diameter corresponding to each drone; The fuselage mass element matrix corresponding to the fuselage mass is: , , For the The body mass corresponding to each drone; Load element matrix corresponding to load , , For the The corresponding load of each drone; The flight speed element matrix corresponding to the flight speed is: , , For the The flight speed of each drone; The flight mode element matrix corresponding to the flight mode is , , For the The flight modes corresponding to each drone; The wind speed element matrix corresponding to the wind speed is , , For the Wind speed corresponding to each drone; The wind direction element matrix corresponding to the wind direction is , , For the The wind direction corresponding to each drone; The temperature element matrix corresponding to the temperature is , , For the The temperature corresponding to each drone; The humidity element matrix corresponding to humidity is , , For the Humidity corresponding to each drone; The flight height element matrix corresponding to the flight height is: , , For the The flight altitude corresponding to each drone; According to the element matrix corresponding to the factors affecting the energy consumption of UAV, the power data collected based on the power sensor , calculate the energy consumption of the drone during flight ; Energy consumption during drone flight It is expressed as: ； Among them, t is the flight time of the UAV, is the total flight time of the drone; Calculate the drone flight energy consumption per unit time corresponding to all drones, and establish the drone flight energy consumption element matrix AE. ,in, is the energy consumption of the drone flight per unit time, .

5. A method for predicting future traffic volume and energy consumption of UAVs in the transportation industry according to claim 4, characterized in that: In said S4, the UAV energy consumption calculation model includes an input layer, 7 hidden layers, a self-attention layer and an output layer; The neural network structure of the established UAV energy consumption calculation model is connected in the following order: Input layer: The motor efficiency element matrix , propeller diameter element matrix , fuselage mass element matrix , load element matrix , flight speed element matrix , flight mode element matrix , wind speed element matrix , Wind direction element matrix , temperature element matrix , humidity element matrix and the flight height element matrix As input; Hidden layer 1: Number of neurons: 512, activation function: ReLU, batch normalization; Hidden layer 2: number of neurons: 256, activation function: ReLU, batch normalization; Self-attention layer: The output matrix of hidden layer 2 is ; Initialize the weight matrix, where the third initial weight query matrix is ​​established in turn , the third initial weight key matrix and the third initial weight matrix , the output matrix of hidden layer 2 Transformed into the third query matrix , the third key matrix and the third value matrix , , , ; Get the third query matrix With the third key matrix The similarity of , and the third attention score matrix is ​​obtained , ,in, is the third attention score matrix Corresponding scaling factor; using softmax function to the third attention score matrix Perform normalization so that the sum of the elements in each row is 1, and obtain the third normalized score matrix , ; The third normalized score matrix Capture the output matrix of hidden layer 2 The internal relations of the layer and serve as the input of the next layer; Hidden layer 3: Number of neurons: 128, activation function: ReLU, batch normalization; Residual layer 1: The output of hidden layer 1 is connected to the output of hidden layer 3 through residual connection. Hidden layer 4: number of neurons: 64, activation function: ReLU, batch normalization; Hidden layer 5: number of neurons: 32, activation function: ReLU, batch normalization; Hidden layer 6: number of neurons: 16, activation function: ReLU, batch normalization; Residual layer 2: The output of hidden layer 4 is connected to the output of hidden layer 6 through residual connection; Hidden layer 7: Number of neurons: 8, activation function: ReLU, batch normalization; Output layer: Number of neurons: 1, i.e., the drone flight energy consumption element matrix AE. After the model training is completed, the output layer outputs the estimated energy consumption per unit time of the drone. ; Based on estimated drone traffic and the estimated energy consumption per unit time of the drone , the energy consumption ZAE under future UAV traffic volume is calculated; The energy consumption ZAE under future drone traffic is expressed as: 。

Images