Unmanned aerial vehicle hangar environment monitoring method and system based on deep learning
Through deep learning, analyzing the temperature and humidity changes in the drone hangar environment, and combining with the fresh air system to carry out intelligent regulation, it solves the problems of difficulty in handling outliers, alarm lag and low regulation efficiency in traditional methods, achieving more efficient and more Flexible environmental monitoring and regulation.
Patent Information
- Application Number
- CN202410977083.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-19
- Publication Date
- 2024-11-12
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional drone hangar environmental monitoring methods are difficult to deal with the problems of outliers, alarm lag and inefficient regulation.
Using a deep learning-based method, we analyze the data change trends by collecting and preprocessing temperature and humidity data, building a trend classification model, combining the fresh air system for environmental adjustment, and achieving adaptive thresholds and intelligent regulation.
It effectively avoids alarm lag, reduces false alarm conditions, improves environmental regulation efficiency, reduces energy consumption, and ensures long-term storage and stable performance of drones and equipment.
Smart Images

Figure CN118939965A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of unmanned aerial vehicles, and in particular, relates to a method and system for monitoring the hangar environment of an unmanned aerial vehicle based on deep learning. Background Art
[0002] The purpose of drone hangar environmental monitoring is to prevent the performance degradation or damage of drones due to hangar temperature and humidity conditions, and to ensure the stability of drones when performing tasks. Traditional drone hangar temperature and humidity monitoring methods are mainly based on fixed thresholds. However, due to the complex working environment of drones, the detection method based on fixed thresholds will produce false alarms when the system produces abnormal values. In addition, when the temperature and humidity exceed the standard operating environment of the hangar, the alarm and control will be issued, which has caused the hangar to work in an unsuitable environment for a period of time. This alarm method has a certain lag. At the same time, traditional control methods often ignore the influence of external temperature and humidity, and rely solely on the hangar's own equipment for control. Not only is the energy consumption high, but the control efficiency is low, and it is difficult to adapt to different climatic conditions, which is not conducive to the long-term preservation and performance stability of drones and hangar equipment. Summary of the invention
[0003] In view of this, the present invention aims to provide a UAV hangar environment monitoring method and system based on deep learning, aiming to solve the problems of traditional UAV hangar monitoring methods being difficult to handle abnormal values, alarm lag and low control efficiency.
[0004] A method for monitoring the hangar environment of a UAV based on deep learning, comprising the following steps:
[0005] S1: Collect historical internal temperature data, internal humidity data, external temperature data and external humidity data of the drone hangar, and preprocess them to obtain preprocessed internal temperature data, internal humidity data, external temperature data and external humidity data;
[0006] S2: Analyze the change trends of the pre-processed internal temperature data and internal humidity data to obtain the temperature and humidity change trend characteristics;
[0007] S3: construct a trend classification model, classify the temperature and humidity change trends according to the temperature and humidity change trend characteristics, and obtain the temperature and humidity change trend categories;
[0008] S4: Analyze the external temperature data and external humidity data in combination with the standard temperature and humidity range for drone storage to obtain a judgment result on whether the fresh air system is activated, so as to decide whether to use the fresh air system to adjust the drone hangar environment;
[0009] S5: When the temperature and humidity change trend category is detected as "about to exceed the threshold", combined with the judgment result of whether the fresh air system is started, the control device is controlled to adjust the temperature and humidity.
[0010] Furthermore, the step S1 collects historical internal temperature data, internal humidity data, external temperature data and external humidity data of the drone hangar, and performs preprocessing to obtain the preprocessed internal temperature data, internal humidity data, external temperature data and external humidity data, including:
[0011] S11: Collect internal temperature data, internal humidity data, external temperature data and external humidity data of the drone hangar through temperature and humidity sensors;
[0012] S12: Clean the internal temperature data, internal humidity data, external temperature data, and external humidity data of the drone hangar, delete null values and abnormal values, and obtain the cleaned internal temperature data, internal humidity data, external temperature data, and external humidity data;
[0013] S13: standardizing the internal temperature data, internal humidity data, external temperature data and external humidity data after cleaning, unifying the dimensions, and obtaining the pre-processed internal temperature data, internal humidity data, external temperature data and external humidity data.
[0014] Furthermore, the step S2 analyzes the change trends of the pre-processed internal temperature data and internal humidity data to obtain the temperature and humidity change trend characteristics, including:
[0015] S21: Use the time series convolutional network to extract the trend of the preprocessed internal temperature data to obtain the temperature change trend characteristics. The calculation method is:
[0016] F1=TCN(x 1,t )
[0017] Among them, F1 is the temperature change trend feature, TCN() is the time series convolution network operation, x 1,t is the preprocessed internal temperature data at time t, where t is the time index;
[0018] S22: Generate an internal humidity data difference according to the pre-processed internal humidity data, and the calculation method is:
[0019] Δx 2,t =x 2,t -x 2,t-1
[0020] Where Δx 2,t is the internal humidity data difference, x 2,t is the preprocessed internal humidity data at time t, x 2,t-1 is the pre-processed internal humidity data at time t-1;
[0021] S23: The internal humidity data difference Δx2,t Perform sequence marking. When there are three consecutive positive or negative numbers, mark the first number of the three numbers as 1 and the remaining numbers as 0 to obtain a trend marking sequence, which is recorded as L t ;
[0022] S24: After the internal humidity data difference, the trend annotation sequence and the internal humidity data are spliced, they are input into the gated recurrent unit to obtain the humidity change trend characteristics. The calculation method is:
[0023] Z t =concat(Δx 2,t ,L t ,x 2,t )
[0024] F2=GRU(Z t )
[0025] Among them, Z t is the concatenation vector, concat() is the concatenation operation, F2 is the humidity change trend feature, GRU() is the gated recurrent unit operation;
[0026] Humidity data often shows a small change range in drone hangar environment monitoring, which makes it difficult to directly capture its characteristics. To solve this problem, step S22 adopts the difference data method, which highlights the change trend of humidity data by performing difference calculation on the pre-processed internal humidity data, making it easier for the subsequent gated cycle unit to capture these change characteristics.
[0027] Step S24 uses sequence labeling to represent the continuous humidity change trend through a series of label values (1 and 0), which provides the model with more data prompts and helps the model to capture the humidity change trend more accurately. Compared with the traditional GRU input, the concatenated vector constructed by the present invention has stronger data representation ability, and the model can learn the general pattern of humidity change trend instead of relying on specific humidity data values.
[0028] Furthermore, the step S3 constructs a trend classification model, classifies the temperature and humidity change trends according to the temperature and humidity change trend characteristics, and obtains the temperature and humidity change trend categories, including:
[0029] S31: Construct a temperature classifier, input the preprocessed temperature data and temperature change characteristics into the temperature classifier, and obtain the temperature change trend category. The calculation method is:
[0030] h1=sigmoid(W1·concat(Δx 1,t ,F1)+b1)
[0031] y1=softmax(W′1h1+b′1)
[0032] Among them, h1 is the temperature change trend feature, sigmoid() is the sigmoid function, W1 is the first parameter matrix, concat() is the concatenation operation, Δx 1,t is the internal humidity data difference, F1 is the temperature change trend feature, b1 is the first bias vector; y1 is the temperature change trend category probability, softmax() is the softmax function, W1′ is the first temperature parameter matrix, and b′1 is the first temperature bias vector;
[0033] After calculating the temperature change trend category probability y1, select the category with the largest probability as the temperature change trend category; the temperature change trend category includes: temperature rise, temperature drop, and temperature stability;
[0034] S32: Build a humidity classifier, input the preprocessed humidity data and humidity change characteristics into the humidity classifier, and obtain the humidity change trend category. The calculation method is:
[0035] h2=sigmoid(W2·concat(Δx 2,t ,F2)+b2)
[0036] y2=softmax(W′2h2+b′2)
[0037] Among them, h2 is the humidity change trend feature, W2 is the second parameter matrix, b2 is the second bias vector; y2 is the humidity change trend category probability, W′2 is the second humidity parameter matrix, and b′2 is the second humidity bias vector;
[0038] After calculating the humidity change trend category probability y2, select the category with the largest probability as the humidity change trend category; the humidity change trend category includes: humidity increase, humidity decrease, and humidity stability;
[0039] S33: Calculate the mean and standard deviation of the temperature data collected for the last 10 times, where the temperature mean is denoted as μ1 and the temperature standard deviation is denoted as σ1; calculate the mean and standard deviation of the humidity data collected for the last 10 times, where the humidity mean is denoted as μ2 and the humidity standard deviation is denoted as σ2;
[0040] S34: Calculate the temperature over-limit threshold value based on the temperature mean and temperature standard deviation collected in the last 10 times; calculate the humidity over-limit threshold value based on the humidity mean and humidity standard deviation collected in the last 10 times. The calculation method is:
[0041]
[0042] Among them, threshold1 is the temperature overrun threshold; for the "temperature rising" category, limit1 is the upper limit of the standard temperature range for storing the drone, and for the "temperature falling" category, limit1 is the lower limit of the standard temperature range for storing the drone; k1 is the temperature standard deviation coefficient, and || is the absolute value operation;
[0043] threshold2 is the humidity overrun threshold; for the "humidity rising" category, limit2 is the upper limit of the standard humidity range for storing the drone, and for the "humidity falling" category, limit2 is the lower limit of the standard humidity range for storing the drone; k2 is the humidity standard deviation coefficient;
[0044] S35: For the categories of "temperature rising" and "temperature falling" in terms of temperature change trend, combined with the temperature overrun threshold, calculate whether the current temperature is about to exceed the threshold and update the temperature change trend category. The calculation method is as follows:
[0045] If |x 1,t - limit1| < threshold1, then update the temperature change trend category to "temperature rising, about to exceed the threshold" or "temperature falling, about to exceed the threshold";
[0046] If |x 1,t - limit1| > threshold1, then do not update the temperature change trend category;
[0047] For the categories of "humidity rising" and "humidity falling" in terms of humidity change trend, combined with the humidity overrun threshold, calculate whether the current humidity is about to exceed the threshold and update the humidity change trend category. The calculation method is as follows:
[0048] If |x 2,t - limit2| < threshold2, then update the humidity change trend category to "humidity rising, about to exceed the threshold" or "humidity falling, about to exceed the threshold";
[0049] If |x 2,t - limit2| > threshold2, then do not update the humidity change trend category;
[0050] Traditional methods mostly use the temperature and humidity values at a certain moment to issue an alarm, while the present invention combines the pre-processed temperature and humidity data with the temperature and humidity change characteristics to make trend judgments and issue warning information; this method of issuing an early warning when an abnormal value is reached can prevent the drone hangar from working in an unsuitable environment in advance, reduce the working time of the drone hangar in an unsuitable environment, and thus reduce potential damage to the drone and its equipment; in addition, simply issuing an alarm based on the temperature and humidity values at a certain moment is prone to false alarms due to instantaneous fluctuations or sensor errors, while the present invention issues an early warning by analyzing the temperature and humidity change trend. Even if an abnormal value appears at a certain moment, the overall temperature and humidity change trend is still in a stable state, and the early warning will not be triggered, thus reducing false alarms;
[0051] The threshold is dynamically calculated based on recent historical temperature data, so the system can automatically adapt to changing temperature trends. If the temperature data drops overall, the threshold will be adjusted accordingly to ensure that the system does not frequently trigger over-limit alarms due to small temperature fluctuations. Through adaptive thresholds, the system can issue an alarm in advance before the temperature approaches the upper limit, allowing operators to take preventive measures to avoid potential damage caused by over-limit temperatures.
[0052] Furthermore, the step S4 combines the standard temperature and humidity ranges for storing drones, analyzes the external temperature data and the external humidity data, and obtains a judgment result on whether the fresh air system is started, so as to decide whether to combine the fresh air system to adjust the drone hangar environment, including:
[0053] S41: Calculate the difference between the pre-processed external temperature data and the standard temperature range in which the drone is stored, and obtain temperature difference data; calculate the difference between the pre-processed external humidity data and the standard humidity range in which the drone is stored, and obtain humidity difference data;
[0054] S42: construct a fresh air start discriminator, and input the pre-processed external temperature data, humidity data, temperature difference data, and humidity difference data into the fresh air start discriminator to obtain a judgment result of whether the fresh air system is started. The calculation method is:
[0055] pro1 = softmax(MLP1(x′ 1,t ,x′ 2,t ,Δx′ 1,t ,Δx′ 2,t ))
[0056] Among them, pro1 is the probability of fresh air starting, MLP1() is the first fully connected layer operation, and x′ 1,t is the preprocessed external temperature data at time t, x′ 2,t is the preprocessed external humidity data at time t, Δx′ 1,tis the external temperature difference data, Δx′ 2,t It is the external humidity difference data;
[0057] After calculating the fresh air startup probability pro1, if pro1>0.5, the judgment result of whether the fresh air system is started is "yes"; if pro1≤0.5, the judgment result of whether the fresh air system is started is "no".
[0058] Furthermore, in step S5, when the temperature and humidity change trend category is detected as "imminently critical", combined with the judgment result of whether the fresh air system is started, the control device is controlled to adjust the temperature and humidity, including:
[0059] S51: If step S43 determines that the fresh air system needs to be started, the fresh air system is started to ventilate the drone hangar, and the internal temperature data and internal humidity data of the drone hangar are monitored at the same time;
[0060] S52: Construct a fresh air stop discriminator, and input the internal temperature data, internal humidity data, external temperature data, and external humidity data of the drone hangar into the fresh air stop discriminator to obtain a judgment result of whether the fresh air system is stopped. The calculation method is:
[0061] pro2 = softmax(MLP2(x 1,t ,x 2,t ,x′ 1,t ,x′ 2,t ))
[0062] Among them, pro2 is the probability of fresh air stopping, and MLP2() is the second fully connected layer operation;
[0063] S53: If step S42 determines that the fresh air system does not need to be started, or step S52 determines that the fresh air system should be stopped, the fresh air system is turned off, and a dehumidifier, a humidifier and an air conditioner are used to continue to control the temperature and humidity of the internal environment of the drone hangar;
[0064] Temperature and humidity control in combination with external temperature and humidity conditions can improve control efficiency. When the external temperature and humidity environment is more suitable, natural wind can be used to quickly improve the temperature and humidity conditions in the hangar. After control to a certain extent, combined with the control of the hangar's own equipment, the operating time and energy consumption of equipment such as air conditioners, humidifiers / dehumidifiers can be reduced while ensuring the control effect.
[0065] Compared with the prior art, the advantages of the present invention are:
[0066] (1) The present invention uses deep learning technology to analyze the changing trends of temperature and humidity data, predict potential environmental problems in advance, and effectively avoid alarm lags. At the same time, because the present invention combines data at each moment for early warning, it has the ability to resist single abnormal values, reducing false alarms caused by abnormal values. At the same time, the present invention combines the fresh air system to control the temperature and humidity of the drone hangar, thereby improving the efficiency of environmental control and reducing energy consumption.
[0067] (2) In terms of algorithm improvement, the present invention performs difference and sequence labeling processing on humidity data with unclear fluctuations, and then combines it with a gated recurrent unit to extract trends, so as to more accurately capture the temporal characteristics of humidity changes. In terms of temperature and humidity exceeding limit thresholds, the present invention proposes an adaptive threshold method, in which the threshold is adaptively adjusted according to the changes in historical data, making the alarm more flexible and adaptable.
[0068] (3) In terms of energy conservation and environmental protection, the present invention constructs a fresh air start discriminator and a stop discriminator to comprehensively analyze the internal and external temperature and humidity of the drone hangar. When the external temperature and humidity environment is more suitable, natural wind is used to quickly improve the temperature and humidity conditions in the hangar; after the control reaches a certain level, the dehumidifier, humidifier and air conditioner are combined for control, and the operating time and energy consumption of equipment such as air conditioners, humidifiers / dehumidifiers are reduced while ensuring the control effect. BRIEF DESCRIPTION OF THE DRAWINGS
[0069] Figure 1 A schematic diagram of the process of UAV hangar environment monitoring based on deep learning provided by the present invention;
[0070] Figure 2 This is a schematic diagram of the algorithm flow for extracting temperature and humidity change trend features provided by the present invention.
[0071] Figure 3 This is a schematic diagram of the algorithm flow for determining the temperature and humidity change trend category provided by the present invention. DETAILED DESCRIPTION
[0072] The present invention is further described below in conjunction with the accompanying drawings, but the present invention is not limited in any way. Any changes or substitutions made based on the teachings of the present invention belong to the protection scope of the present invention.
[0073] Example 1: A method for monitoring the environment of a drone hangar based on deep learning, such as Figure 1 As shown, the following steps are included:
[0074] S1: Collect historical internal temperature data, internal humidity data, external temperature data and external humidity data of the drone hangar, and preprocess them to obtain preprocessed internal temperature data, internal humidity data, external temperature data and external humidity data;
[0075] S11: Collect internal temperature data, internal humidity data, external temperature data and external humidity data of the drone hangar through temperature and humidity sensors;
[0076] S12: Clean the internal temperature data, internal humidity data, external temperature data, and external humidity data of the drone hangar, delete null values and abnormal values, and obtain the cleaned internal temperature data, internal humidity data, external temperature data, and external humidity data;
[0077] S13: standardizing the internal temperature data, internal humidity data, external temperature data and external humidity data after cleaning, unifying the dimensions, and obtaining the pre-processed internal temperature data, internal humidity data, external temperature data and external humidity data.
[0078] S2: Analyze the change trends of the pre-processed internal temperature data and internal humidity data to obtain the temperature and humidity change trend characteristics;
[0079] S21: Use the time series convolutional network to extract the trend of the preprocessed internal temperature data to obtain the temperature change trend characteristics. The calculation method is:
[0080] F1=TCN(x 1,t )
[0081] Among them, F1 is the temperature change trend feature, TCN() is the time series convolution network operation, x 1,t is the preprocessed internal temperature data at time t, where t is the time index;
[0082] S22: Generate an internal humidity data difference according to the pre-processed internal humidity data, and the calculation method is:
[0083] Δx 2,t =x 2,t -x 2,t-1
[0084] Where Δx 2,t is the internal humidity data difference, x 2,t is the preprocessed internal humidity data at time t, x 2,t-1 is the pre-processed internal humidity data at time t-1;
[0085] S23: The internal humidity data difference Δx 2,t Perform sequence marking. When there are three consecutive positive or negative numbers, mark the first number of the three numbers as 1 and the remaining numbers as 0 to obtain a trend marking sequence, which is recorded as L t ;
[0086] S24: After the internal humidity data difference, the trend annotation sequence and the internal humidity data are spliced, they are input into the gated recurrent unit to obtain the humidity change trend characteristics. The calculation method is:
[0087] Z t =concat(Δx 2,t ,L t ,x 2,t )
[0088] F2=GRU(Z t )
[0089] Among them, Z t is the concatenation vector, concat() is the concatenation operation, F2 is the humidity change trend feature, GRU() is the gated recurrent unit operation;
[0090] Humidity data often shows a small change range in drone hangar environment monitoring, which makes it difficult to directly capture its characteristics. To solve this problem, step S22 adopts the difference data method, which highlights the change trend of humidity data by performing difference calculation on the pre-processed internal humidity data, making it easier for the subsequent gated cycle unit to capture these change characteristics.
[0091] Step S24 uses sequence annotation to represent the continuous humidity change trend through a series of annotation values (1 and 0), which gives the model more data prompts and helps the model to capture the humidity change trend more accurately. Compared with the traditional GRU input, the concatenated vector constructed by the present invention has a stronger data representation ability, and the model can learn the general pattern of humidity change trend instead of relying on specific humidity data values.
[0092] For example, there are 10 moments of humidity data:
[0093] [45.2,45.3,45.5,45.8,46.0,46.1,46.2,45.7,45.9,45.9]
[0094] The corresponding internal humidity data difference is:
[0095] [0.1,0.2,0.3,0.2,0.1,0.1,-0.5,0.2,0.0]
[0096] The corresponding trend annotation sequence is:
[0097] [1,0,0,1,0,0,0,0,0,0].
[0098] S3: construct a trend classification model, classify the temperature and humidity change trends according to the temperature and humidity change trend characteristics, and obtain the temperature and humidity change trend categories;
[0099] S31: Construct a temperature classifier, input the preprocessed temperature data and temperature change characteristics into the temperature classifier, and obtain the temperature change trend category. The calculation method is:
[0100] h1=sigmoid(W1·concat(Δx 1,t ,F1)+b1)
[0101] y1=softmax(W′1h1+b′1)
[0102] Among them, h1 is the temperature change trend feature, sigmoid() is the sigmoid function, W1 is the first parameter matrix, concat() is the concatenation operation, Δx 1,t is the internal humidity data difference, F1 is the temperature change trend feature, b1 is the first bias vector; y1 is the temperature change trend category probability, softmax() is the softmax function, W1′ is the first temperature parameter matrix, and b′1 is the first temperature bias vector;
[0103] After calculating the temperature change trend category probability y1, select the category with the largest probability as the temperature change trend category; the temperature change trend category includes: temperature rise, temperature drop, and temperature stability;
[0104] S32: Build a humidity classifier, input the preprocessed humidity data and humidity change characteristics into the humidity classifier, and obtain the humidity change trend category. The calculation method is:
[0105] h2=sigmoid(W2·concat(Δx 2,t ,F2)+b2)
[0106] y2=softmax(W′2h2+b′2)
[0107] Among them, h2 is the humidity change trend feature, W2 is the second parameter matrix, b2 is the second bias vector; y2 is the humidity change trend category probability, W′2 is the second humidity parameter matrix, and b′2 is the second humidity bias vector;
[0108] After calculating the humidity change trend category probability y2, select the category with the largest probability as the humidity change trend category; the humidity change trend category includes: humidity increase, humidity decrease, and humidity stability;
[0109] S33: Calculate the mean and standard deviation of the temperature data collected for the last 10 times, where the temperature mean is denoted as μ1 and the temperature standard deviation is denoted as σ1; calculate the mean and standard deviation of the humidity data collected for the last 10 times, where the humidity mean is denoted as μ2 and the humidity standard deviation is denoted as σ2;
[0110] S34: Calculate the temperature overrun threshold based on the average value and standard deviation of the temperatures collected in the most recent 10 times; calculate the humidity overrun threshold based on the average value and standard deviation of the humidities collected in the most recent 10 times. The calculation method is as follows:
[0111]
[0112] Among them, threshold1 is the temperature overrun threshold; for the "temperature rising" category, limit1 is the upper limit of the standard temperature range for storing the drone, and for the "temperature falling" category, limit1 is the lower limit of the standard temperature range for storing the drone; k1 is the temperature standard deviation coefficient, and || represents the absolute value operation;
[0113] threshold2 is the humidity overrun threshold; for the "humidity rising" category, limit2 is the upper limit of the standard humidity range for storing the drone, and for the "humidity falling" category, limit2 is the lower limit of the standard humidity range for storing the drone; k2 is the humidity standard deviation coefficient;
[0114] S35: For the categories of "temperature rising" and "temperature falling" in terms of the temperature change trend category, combined with the temperature overrun threshold, calculate whether the current temperature is about to exceed the threshold, and update the temperature change trend category. The calculation method is as follows:
[0115] If |x 1,t - limit1| < threshold1, then update the temperature change trend category to "temperature rising, about to exceed the threshold" or "temperature falling, about to exceed the threshold";
[0116] If |x 1,t - limit1| > threshold1, then do not update the temperature change trend category;
[0117] For the categories of "humidity rising" and "humidity falling" in terms of the humidity change trend category, combined with the humidity overrun threshold, calculate whether the current humidity is about to exceed the threshold, and update the humidity change trend category. The calculation method is as follows:
[0118] If |x 2,t - limit2| < threshold2, then update the humidity change trend category to "humidity rising, about to exceed the threshold" or "humidity falling, about to exceed the threshold";
[0119] If |x 2,t - limit2| > threshold2, then do not update the humidity change trend category;
[0120] Traditional methods mostly use the temperature and humidity values at a certain moment to issue an alarm, while the present invention combines the pre-processed temperature and humidity data with the temperature and humidity change characteristics to make trend judgments and issue warning information; this method of issuing an early warning when an abnormal value is reached can prevent the drone hangar from working in an unsuitable environment in advance, reduce the working time of the drone hangar in an unsuitable environment, and thus reduce potential damage to the drone and its equipment; in addition, simply issuing an alarm based on the temperature and humidity values at a certain moment is prone to false alarms due to instantaneous fluctuations or sensor errors, while the present invention issues an early warning by analyzing the temperature and humidity change trend. Even if an abnormal value appears at a certain moment, the overall temperature and humidity change trend is still in a stable state, and the early warning will not be triggered, thus reducing false alarms;
[0121] The threshold is dynamically calculated based on recent historical temperature data, so the system can automatically adapt to the changing trend of temperature. If the temperature data drops overall, the threshold will be adjusted accordingly to ensure that the system does not frequently trigger over-limit alarms due to small temperature fluctuations. With adaptive thresholds, the system can issue an alarm in advance before the temperature approaches the upper limit, allowing operators to take preventive measures to avoid potential damage caused by over-limit temperature.
[0122] For example, if the upper temperature limit is 25°C, the temperature data collected in the last 10 times are 23.0, 23.2, 23.4, 23.1, 23.3, 23.5, 23.7, 23.6, 23.2, 23.4 (unit: Celsius), and the temperature standard deviation coefficient is 15, then:
[0123] The recent historical temperature average is 23.3℃, and the recent historical temperature standard deviation is 0.13
[0124]
[0125] Calculation shows that the temperature over-limit threshold at this time is 1.74℃;
[0126] Assume that the new temperature data collected in the last 10 times are: 18.0, 18.2, 18.4, 18.1, 18.3, 18.5, 18.7, 18.6, 18.2, 18.4 (unit: Celsius), the upper limit of temperature is still 25℃, and the temperature standard deviation coefficient k is still 15, then:
[0127] The recent historical temperature average is 18.3℃, and the recent historical temperature standard deviation is 0.12
[0128]
[0129] It can be seen that as the temperature decreases, the temperature over-limit threshold becomes smaller, that is, the temperature can have a larger fluctuation space without triggering an over-limit.
[0130] S4: Analyze the external temperature data and external humidity data in combination with the standard temperature and humidity range for drone storage to obtain a judgment result on whether the fresh air system is activated, so as to decide whether to use the fresh air system to adjust the drone hangar environment;
[0131] S41: Calculate the difference between the pre-processed external temperature data and the standard temperature range in which the drone is stored, and obtain temperature difference data; calculate the difference between the pre-processed external humidity data and the standard humidity range in which the drone is stored, and obtain humidity difference data;
[0132] S42: construct a fresh air start discriminator, and input the pre-processed external temperature data, humidity data, temperature difference data, and humidity difference data into the fresh air start discriminator to obtain a judgment result of whether the fresh air system is started. The calculation method is:
[0133] pro1 = softmax(MLP1(x′ 1,t ,x′ 2,t ,Δx′ 1,t ,Δx′ 2,t ))
[0134] Among them, pro1 is the probability of fresh air starting, MLP1() is the first fully connected layer operation, and x′ 1,t is the preprocessed external temperature data at time t, x′ 2,t is the preprocessed external humidity data at time t, Δx′ 1,t is the external temperature difference data, Δx′ 2,t It is the external humidity difference data;
[0135] After calculating the fresh air startup probability pro1, if pro1>0.5, the judgment result of whether the fresh air system is started is "yes"; if pro1≤0.5, the judgment result of whether the fresh air system is started is "no";
[0136] For example, when the temperature is within a reasonable range, and the humidity is not within a reasonable range but does not deviate much from the standard humidity range for drone storage, the fresh air start discriminator will determine "yes"; on the contrary, if the temperature is within a reasonable range, but the humidity is not only not within a reasonable range but also deviates greatly, the fresh air start discriminator will determine "no".
[0137] S5: When the temperature and humidity change trend category is detected as "about to exceed the threshold", combined with the judgment result of whether the fresh air system is started, the control device is controlled to adjust the temperature and humidity;
[0138] S51: If it is determined in step S43 that the fresh air system needs to be started, the fresh air system is started to ventilate the drone hangar, while monitoring the internal temperature data and internal humidity data of the drone hangar;
[0139] S52: Construct a fresh air stop discriminator, and input the internal temperature data, internal humidity data, external temperature data, and external humidity data of the drone hangar into the fresh air stop discriminator to obtain a judgment result of whether the fresh air system is stopped. The calculation method is:
[0140] pro2 = softmax(MLP2(x 1,t ,x 2,t ,x′ 1,t ,x′ 2,t ))
[0141] Among them, pro2 is the probability of fresh air stopping, and MLP2() is the second fully connected layer operation;
[0142] S53: If step S42 determines that the fresh air system does not need to be started, or step S52 determines that the fresh air system should be stopped, the fresh air system is turned off, and a dehumidifier, a humidifier and an air conditioner are used to continue to control the temperature and humidity of the internal environment of the drone hangar;
[0143] Temperature and humidity control in combination with external temperature and humidity conditions can improve control efficiency. When the external temperature and humidity environment is more suitable, natural wind can be used to quickly improve the temperature and humidity conditions in the hangar. After control to a certain extent, combined with the control of the hangar's own equipment, the operating time and energy consumption of equipment such as air conditioners, humidifiers / dehumidifiers can be reduced while ensuring the control effect.
[0144] The present invention is particularly suitable for drone hangar environment monitoring scenarios where the external temperature and humidity changes are relatively complex; through deep learning technology, the present invention can accurately capture and analyze the changing trends of temperature and humidity inside the hangar, significantly improve the accuracy and timeliness of drone hangar environment monitoring, and reduce drone failures or damages caused by environmental abnormalities, thereby ensuring the safe storage and smooth operation of drones.
[0145] Embodiment 2: The present invention also discloses a drone hangar environment monitoring system based on deep learning, comprising:
[0146] Data collection and preprocessing module: collects historical internal temperature data, internal humidity data, external temperature data and external humidity data of the drone hangar, and performs preprocessing to obtain preprocessed internal temperature data, internal humidity data, external temperature data and external humidity data;
[0147] Change trend feature extraction module: analyze the change trend of the pre-processed internal temperature data and internal humidity data to obtain the temperature and humidity change trend features;
[0148] Change trend classification module: build a trend classification model, classify the temperature and humidity change trends according to the temperature and humidity change trend characteristics, and obtain the temperature and humidity change trend categories;
[0149] Fresh air system identification module: Combined with the standard temperature and humidity range for drone storage, the module analyzes the external temperature data and external humidity data to obtain the identification result of whether the fresh air system is activated, so as to decide whether to use the fresh air system to adjust the drone hangar environment;
[0150] Temperature and humidity control module: When the temperature and humidity change trend category is detected as "about to exceed the threshold", combined with the judgment result of whether the fresh air system is started, the control equipment is controlled to adjust the temperature and humidity.
[0151] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. And the terms "including", "comprising" or any other variants thereof in this article are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "including a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.
[0152] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.
[0153] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.
Claims
1. A UAV hangar environment monitoring method based on deep learning, It is characterized in that The following steps are involved: S1: Collect historical internal temperature data, internal humidity data, external temperature data and external humidity data of the drone hangar, and preprocess them to obtain preprocessed internal temperature data, internal humidity data, external temperature data and external humidity data; S2: Analyze the change trends of the pre-processed internal temperature data and internal humidity data to obtain the temperature and humidity change trend characteristics; S3: construct a trend classification model, classify the temperature and humidity change trends according to the temperature and humidity change trend characteristics, and obtain the temperature and humidity change trend categories; S4: Analyze the external temperature data and external humidity data in combination with the standard temperature and humidity range for drone storage to obtain a judgment result on whether the fresh air system is activated, so as to decide whether to use the fresh air system to adjust the drone hangar environment; S5: When the temperature and humidity change trend category is detected as "about to exceed the threshold", combined with the judgment result of whether the fresh air system is started, the control device is controlled to adjust the temperature and humidity.
2. According to the deep learning-based UAV hangar environment monitoring method of claim 1, It is characterized in that The S1 step includes: S11: Collect internal temperature data, internal humidity data, external temperature data and external humidity data of the drone hangar through temperature and humidity sensors; S12: Clean the internal temperature data, internal humidity data, external temperature data, and external humidity data of the drone hangar, delete null values and abnormal values, and obtain the cleaned internal temperature data, internal humidity data, external temperature data, and external humidity data; S13: standardizing the internal temperature data, internal humidity data, external temperature data and external humidity data after cleaning, unifying the dimensions, and obtaining the pre-processed internal temperature data, internal humidity data, external temperature data and external humidity data.
3. According to the deep learning-based UAV hangar environment monitoring method of claim 1, It is characterized in that The S2 step includes: For the preprocessed internal temperature data, a temporal convolutional network is used to extract trends and obtain the temperature change trend characteristics; For the preprocessed internal humidity data, the internal humidity data difference is first calculated, and then the sequence is labeled to obtain a trend labeling sequence; the trend labeling sequence is then concatenated with the preprocessed internal humidity data and input into the gated recurrent unit to obtain the humidity change trend characteristics.
4. According to the deep learning-based UAV hangar environment monitoring method of claim 3, It is characterized in that The S2 step includes: S21: Use the time series convolutional network to extract the trend of the preprocessed internal temperature data to obtain the temperature change trend characteristics. The calculation method is: F 1 =TCN(x 1,t ) Among them, F 1 is the temperature change trend feature, TCN( ) is the temporal convolutional network operation, x 1,t is the preprocessed internal temperature data at time t, where t is the time index; S22: Generate an internal humidity data difference according to the pre-processed internal humidity data, and the calculation method is: Δx 2,t =x 2,t -x 2,t-1 Where Δx 2,t is the internal humidity data difference, x 2,t is the preprocessed internal humidity data at time t, x 2,t-1 is the pre-processed internal humidity data at time t-1; S23: The internal humidity data difference Δx 2,t Perform sequence marking. When there are three consecutive positive or negative numbers, mark the first number of the three numbers as 1 and the remaining numbers as 0 to obtain a trend marking sequence, which is recorded as L t ; S24: After the internal humidity data difference, the trend annotation sequence and the internal humidity data are spliced, they are input into the gated recurrent unit to obtain the humidity change trend characteristics. The calculation method is: Z t =concat(Δx 2 ,t,L t ,x 2 ,t) F 2 =GRU(Z t ) Among them, Z t is the concatenation vector, concat() is the concatenation operation, F 2 is the humidity change trend feature, and GRU() is the gated recurrent unit operation.
5. The method for monitoring the environment of a drone hangar based on deep learning according to claim 1, It is characterized in that The S3 step includes: Construct temperature and humidity classifiers respectively, calculate the temperature and humidity change trend categories; calculate the temperature and humidity over-limit thresholds based on the mean and standard deviation of the temperature and humidity data collected in the last 10 times; update the temperature and humidity change trend categories based on the temperature and humidity over-limit thresholds.
6. The method for monitoring the drone hangar environment based on deep learning according to claim 5, It is characterized in that The S3 step includes: S31: Construct a temperature classifier, input the preprocessed temperature data and temperature change characteristics into the temperature classifier, and obtain the temperature change trend category. The calculation method is: h 1 =sigmoid(W 1 ·concat(Δx 1,t ,F 1 )+b 1 ) y 1 =softmax(W′ 1 h 1 +b′ 1 ) Among them, h 1 is the temperature change trend characteristic, sigmoid( ) is the sigmoid function, W 1 is the first parameter matrix, concat() is the concatenation operation, Ax 1,t is the internal humidity data difference, F 1 is the temperature change trend characteristic, b 1 is the first bias vector; y 1 is the temperature change trend category probability, softmax() is the softmax function, W 1 ′ is the first temperature parameter matrix, b′ 1 is the first temperature bias vector; Calculate the temperature change trend category probability y 1 Then, the category with the largest probability is selected as the temperature change trend category; the temperature change trend category includes: temperature rise, temperature drop, and temperature stability; S32: Build a humidity classifier, input the preprocessed humidity data and humidity change characteristics into the humidity classifier, and obtain the humidity change trend category. The calculation method is: h 2 =sigmoid(W 2 ·concat(Δx 2,t ,F 2 )+b 2 ) y 2 =softmax(W′ 2 h 2 +b′ 2 ) Among them, h 2 is the humidity change trend characteristic, W 2 is the second parameter matrix, b 2 is the second bias vector; y 2 is the probability of humidity change trend category, W′ 2 is the second humidity parameter matrix, b′ 2 is the second humidity bias vector; Calculate the humidity change trend category probability y 2 Then, the category with the largest probability is selected as the humidity change trend category; the humidity change trend category includes: humidity increase, humidity decrease, and humidity stability; S33: Calculate the mean and standard deviation of the temperature data collected for the last 10 times, where the mean temperature is denoted as μ 1 , the temperature standard deviation is denoted as σ 1 ; Calculate the mean and standard deviation of the humidity data collected for the last 10 times, where the humidity mean is denoted as μ 2 , the humidity standard deviation is denoted as σ 2 ; S34: Calculate the temperature over-limit threshold value based on the temperature mean and temperature standard deviation collected in the last 10 times; calculate the humidity over-limit threshold value based on the humidity mean and humidity standard deviation collected in the last 10 times. The calculation method is: Among them, threshold 1 is the temperature exceeding limit threshold; for the "temperature rise" category, limit 1 The upper limit of the standard temperature range for drone storage. For the "Temperature Drop" category, limit 1 The lower limit of the standard temperature range for drone storage; k 1 is the temperature standard deviation coefficient, || is the absolute value operation; threshold 2 is the humidity exceeding limit threshold; for the "humidity rising" category, limit 2 The upper limit of the standard humidity range for drone storage. For the "humidity drop" category, limit 2 The lower limit of the standard humidity range for drone storage; k 2 is the humidity standard deviation coefficient; S35: For the temperature change trend categories of "temperature rise" and "temperature drop", combined with the temperature over-limit threshold, calculate whether the current temperature is about to exceed the threshold, and update the temperature change trend category. The calculation method is: If |x 1,t -limit 1 |<threshold 1 , then update the temperature change trend category to "temperature rising, about to exceed the threshold" or "temperature falling, about to exceed the threshold"; If |x 1,t -limit 1 |>threshold 1 , the temperature change trend category will not be updated; For the humidity change trend categories of "humidity increase" and "humidity decrease", combined with the humidity over-limit threshold, calculate whether the current humidity is about to exceed the threshold, and update the humidity change trend category. The calculation method is: If |x 2,t -limit 2 |<threshold 2 , then update the humidity change trend category to "humidity rising, about to exceed the threshold" or "humidity falling, about to exceed the threshold"; If |x 2,t -limit 2 |>threshold 2 , the humidity change trend category will not be updated.
7. The method for monitoring the drone hangar environment based on deep learning according to claim 6, It is characterized in that The S4 step includes: S41: Calculate the difference between the pre-processed external temperature data and the standard temperature range in which the drone is stored, and obtain temperature difference data; calculate the difference between the pre-processed external humidity data and the standard humidity range in which the drone is stored, and obtain humidity difference data; S42: construct a fresh air start discriminator, and input the pre-processed external temperature data, humidity data, temperature difference data, and humidity difference data into the fresh air start discriminator to obtain a judgment result of whether the fresh air system is started. The calculation method is: pro 1 =softmax(MLP 1 (x′ 1 ,t,x′ 2 ,t,Δx′ i ,t,Δx′ 2,t )) Among them, pro 1 is the probability of new wind startup, MLP 1 ( ) is the first fully connected layer operation, x′ 1,t is the preprocessed external temperature data at time t, x′ 2,t is the preprocessed external humidity data at time t, Δx′ 1,t is the external temperature difference data, Δx′ 2,t It is the external humidity difference data; Calculate the probability of fresh air starting 1 After that, if pro 1 >0.5, the result of judging whether the fresh air system is started is "yes"; if pro 1 ≤0.5, the judgment result of whether the fresh air system is started is "No".
8. The method for monitoring the drone hangar environment based on deep learning according to claim 7, It is characterized in that The S5 step includes: S51: If it is determined in step S43 that the fresh air system needs to be started, the fresh air system is started to ventilate the drone hangar, while monitoring the internal temperature data and internal humidity data of the drone hangar; S52: Construct a fresh air stop discriminator, and input the internal temperature data, internal humidity data, external temperature data, and external humidity data of the drone hangar into the fresh air stop discriminator to obtain a judgment result of whether the fresh air system is stopped. The calculation method is: pro 2 =softmax(MLP 2 (x 1,t ,x 2,t ,x′ 1,t ,x′ 2,t )) Among them, pro 2 is the probability of fresh air stopping, MLP 2 ( ) is the second fully connected layer operation; S53: If step S42 determines that the fresh air system does not need to be started, or step S52 determines that the fresh air system should be stopped, the fresh air system is turned off, and a dehumidifier, humidifier and air conditioner are used to continue to control the temperature and humidity of the internal environment of the drone hangar.
9. A UAV hangar environment monitoring system based on deep learning, It is characterized in that include: Data collection and preprocessing module: collects historical internal temperature data, internal humidity data, external temperature data and external humidity data of the drone hangar, and performs preprocessing to obtain preprocessed internal temperature data, internal humidity data, external temperature data and external humidity data; Change trend feature extraction module: analyze the change trend of the pre-processed internal temperature data and internal humidity data to obtain the temperature and humidity change trend features; Change trend classification module: build a trend classification model, classify the temperature and humidity change trends according to the temperature and humidity change trend characteristics, and obtain the temperature and humidity change trend categories; Fresh air system identification module: Combined with the standard temperature and humidity range for drone storage, the module analyzes the external temperature data and external humidity data to obtain the identification result of whether the fresh air system is activated, so as to decide whether to use the fresh air system to adjust the drone hangar environment; Temperature and humidity control module: When the temperature and humidity change trend category is detected as "about to exceed the threshold", combined with the judgment result of whether the fresh air system is started, the control equipment is controlled to adjust the temperature and humidity; To implement a UAV hangar environment monitoring method based on deep learning as described in any one of claims 1-8.
Citation Information
Patent Citations
Equipment control method and device and electronic equipment
CN114706433A
Intelligent equipment monitoring and early warning system
CN117826715A
Greenhouse air conditioner intelligent control method and system based on climate collection
CN118012171A
Constant temperature and humidity control method and system based on far infrared radiation heating principle
CN118259712A
