Switch cabinet temperature and humidity monitoring method, label and system based on conformal sensing label

Abstract

The invention discloses a switch cabinet temperature and humidity monitoring method based on a conformal sensing label, and the method comprises the steps: S1, carrying out the data collection and storage through an automatic encoder, and compressing the temperature and humidity time sequence data; s2, reading the compressed data through Kalman filtering, decoding and recovering the compressed data into time sequence data, and performing denoising; s3, a Q-Learning algorithm is adopted to carry out dynamic optimization of a data threshold value; s4, predicting a data exception probability through a logistic regression algorithm, and performing data exception segmentation and exception marking according to the data exception probability; s5, calculating an abnormal score through an isolated forest model by using the data marked as abnormal in the step S4, and carrying out advanced detection; and S6, obtaining the data anomaly probability obtained in the step S4 and the data anomaly score obtained in the step S5, setting an alarm decision according to the data anomaly probability and the data anomaly score falling value interval, and outputting an alarm result. The temperature and humidity data are analyzed and subjected to anomaly detection by using a multi-modal algorithm, so that the reliability and the accuracy of the system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of radio frequency identification, and in particular to a method, a tag and a system for monitoring the temperature and humidity of a switchgear cabinet based on a conformal sensing tag. Background Art

[0002] In the power system, the switchgear cabinet is an important power equipment, which is often used to control, protect and switch circuits. Due to the particularity of the internal environment of the switchgear, abnormal changes in temperature and humidity may cause overheating, damage or failure of the equipment. Therefore, real-time monitoring of the temperature and humidity conditions inside the switchgear cabinet is of great significance for ensuring the normal operation of the equipment.

[0003] The traditional methods for processing temperature and humidity monitoring data are single, and it is difficult to dynamically adapt to the threshold optimization requirements under complex working conditions.

[0004] When traditional RFID tags are used inside the switchgear cabinet, they will be interfered by metal, and due to electromagnetic shielding, the signal strength for receiving the tags outside the switchgear cabinet will be greatly limited.

[0005] Traditional temperature and humidity monitoring systems often rely on active sensors or devices that require frequent maintenance, and they are inconvenient to install and have poor anti-interference ability. Summary of the Invention

[0006] The present invention aims to at least solve one of the technical problems in the related art to some extent. For this reason, an object of the present invention is to propose a method, a tag and a system for monitoring the temperature and humidity of a switchgear cabinet based on a conformal sensing tag, which ensure the accurate transmission of data and the realization of temperature and humidity threshold classification through a multi-modal data processing algorithm, and use passive conformal sensing RFID tags to break through the electromagnetic shielding limitation of the switchgear cabinet.

[0007] In a first aspect, a method for monitoring the temperature and humidity of a switchgear cabinet based on a conformal sensing tag proposed by the present invention comprises the following steps:

[0008] S1: Collect and store data through an autoencoder, and compress the temperature and humidity time-series data;

[0009] S2: Read the compressed data in step S1 through a Kalman filter and decode and restore it to time-series data for denoising;

[0010] S3: Apply the data processed in step S2, and perform dynamic optimization of the data threshold by using the Q-Learning algorithm;

[0011] S4: Apply the data processed in step S3, predict the data anomaly probability through a logistic regression algorithm, and perform data anomaly segmentation and anomaly marking according to the data anomaly probability;

[0012] S5: Apply the data marked as abnormal in step S4, calculate the anomaly score through the Isolation Forest model, and perform advanced detection;

[0013] S6: Obtain the data anomaly probability obtained in step S4 and the data anomaly score obtained in step S5, set the alarm decision based on the numerical intervals into which the data anomaly probability and data anomaly score fall, and output the alarm result;

[0014] S7: Use the PCA algorithm to continuously and dynamically monitor the data, and implement feedback regulation based on the monitoring results.

[0015] Preferably, in step S1:

[0016] Input: Temperature and humidity data within the time window

[0017] Encoder: z = f e (x i ) = σ(W e x i + b e )

[0018] Decoder:

[0019] Loss function (mean squared error):

[0020] Among them, T t-k ,..., T t represents the temperature sequence from time t - n to the current time t, H t-k ,..., H t represents the humidity sequence from time t - n to the current time t, k is the window length, W e , b e are the encoder weights and biases; W d , b d are the decoder weights and biases; σ is the activation function.

[0021] Preferably, in step S2:

[0022] Use Kalman filtering for data cleaning and preprocessing:

[0023] The state equation is:

[0024] The observation equation is:

[0025] Among them, ΔT t , ΔH t are the temperature and humidity change rates; Δt is the time difference between two adjacent time steps, that is, the time length from time t - 1 to t; is the process noise; T t obs is the temperature value actually measured by the sensor at time t, is the humidity value actually measured by the sensor at time t, is the observation noise.

[0026] Preferably, in step S3:

[0027] The Q-Learning algorithm is used to dynamically optimize the data threshold:

[0028] State:

[0029] where T t k is the temperature data after Kalman filter output, is the humidity data after Kalman filter output;

[0030] Action: Adjust the threshold a ∈ {ΔT max , ΔH max};

[0031] Q-value update:

[0032] Reward function:

[0033] where α is the learning rate and γ is the learning factor.

[0034] Preferably, in step S4:

[0035] The data anomaly probability is predicted by the logistic regression algorithm, and data anomaly segmentation and anomaly marking are performed based on the data anomaly probability:

[0036] Input features:

[0037] where ΔT t k is the temperature change rate after Kalman filter output, is the humidity change rate after Kalman filter output;

[0038] Predicted probability:

[0039] Classification rule:

[0040] where β0, β1, …, β4 are model coefficients.

[0041] Preferably, in step S5:

[0042] The anomaly score is calculated by the Isolation Forest model for advanced detection:

[0043] Abnormal score:

[0044] where h(x) is the path length and E is the expected value, is the harmonic number;

[0045] If Score(x) > 0.6, it is confirmed as a real anomaly;

[0046] Splitting rule: Randomly select feature T t k or Randomly select a separation value.

[0047] Preferably, in step S6:

[0048] Set the alarm decision according to the data anomaly probability and the numerical interval where the data anomaly score falls, and output the alarm result:

[0049] Alarm rule logic: If the anomaly probability P > 0.9 and Score(x) > 0.6, a first-level alarm is triggered;

[0050] If the anomaly probability P > 0.7 and the first-level alarm condition is not met, a second-level alarm is executed.

[0051] Preferably, in step S7:

[0052] Use the PCA algorithm to continuously and dynamically monitor the data, and implement feedback adjustment according to the monitoring results:

[0053] Covariance matrix:

[0054]

[0055] where Var is the variance and Cov is the covariance;

[0056] KL divergence detects drift:

[0057]

[0058] where U is the historical distribution and V is the current distribution;

[0059] If D KL > 0.1, it is determined that data drift has occurred, then feedback adjustment is triggered, Q-Learning training is restarted, the threshold is dynamically adjusted; and model retraining is performed to update the parameters of the logistic regression and isolation forest models.

[0060] In a second aspect, a conformal sensing tag for a switchgear cabinet proposed by the present invention can be applied to any one of the above-mentioned switchgear cabinet temperature and humidity monitoring methods based on conformal sensing tags. The conformal sensing tag includes: a temperature and humidity sensing chip and an antenna radiation structure. The antenna radiation structure includes a pair of radiation conductors symmetrically distributed along the axis of the humidity sensing chip. Each radiation conductor includes an inner longitudinal radiation arm, a horizontal S-shaped radiation arm, a horizontal L-shaped radiation arm, and a rectangular radiation arm. The humidity sensing chip is sequentially connected through an L-shaped feeder line, a straight connecting arm, and one end of the inner longitudinal radiation arm. The humidity sensing chip and the inner longitudinal radiation arm are arranged inside the switchgear cabinet. The other end of the inner longitudinal radiation arm is provided with a U-shaped bent radiation arm adapted to the slot on the side wall of the switchgear cabinet. One end of the U-shaped bent radiation arm is connected to the inner longitudinal radiation arm. The other end of the U-shaped bent radiation arm is sequentially connected to the rectangular radiation arm through an outer longitudinal radiation arm, an arc-shaped bent connecting arm, a horizontal L-shaped radiation arm, and a horizontal S-shaped radiation arm. The horizontal L-shaped radiation arm, the horizontal S-shaped radiation arm, and the rectangular radiation arm are all arranged outside the switchgear cabinet.

[0061] In a third aspect, a switchgear cabinet temperature and humidity detection system based on a conformal sensing tag proposed by the present invention includes:

[0062] Conformal sensing tag: used to collect temperature and humidity sensing data inside the switchgear cabinet;

[0063] Reader-writer: deployed outside the switchgear cabinet, used to activate the conformal sensing tag and receive the sensing data returned by it;

[0064] Data processing and observation platform: communicatively connected to the reader-writer, applied to any one of the above-mentioned switchgear cabinet temperature and humidity monitoring methods based on conformal sensing tags, analyzes the sensing data and generates temperature and humidity alarm signals.

[0065] The beneficial effects in the present invention are:

[0066] (1) Compared with the traditional switchgear cabinet temperature and humidity monitoring method, it adopts a passive design and does not require battery power supply, greatly reducing the personnel maintenance cost and equipment failure rate;

[0067] (2) Compared with the traditional RFID tag, by using the externally bent U-shaped radiation arm, the tag antenna can extend out from the slot of the switchgear cabinet, while the temperature and humidity sensing chip monitors the temperature and humidity in the switchgear cabinet, breaking through the limitation of the electromagnetic shielding effect of the switchgear cabinet and realizing efficient signal transmission inside and outside the cabinet;

[0068] (3) Adopting a multi-modal data processing algorithm, integrating an autoencoder, a Kalman filter, Q-Learning dynamic threshold optimization, and an isolation forest model, provides functions of efficient data processing and temperature and humidity threshold classification, improves the accuracy of anomaly detection, and ensures the high reliability of the system;

[0069] (4) The conformal sensing tag has a -3dB bandwidth of 600 - 1200 MHz, providing good coverage of the 840 - 960 MHz UHF band globally; with a center frequency of 925 MHz, it is well within the working frequency band of 920 - 925 MHz specified by the Ministry of Industry and Information Technology for RFID devices, enabling efficient information transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In the drawings:

[0071] Figure 1 is the logic diagram of the switchgear temperature and humidity monitoring method based on the conformal sensing tag proposed by the present invention;

[0072] Figure 2 is the structural schematic diagram of the conformal sensing tag antenna proposed by the present invention;

[0073] Figure 3 is the front view of the conformal sensing tag antenna proposed by the present invention;

[0074] Figure 4 is the schematic diagram of parameter annotation of the conformal sensing tag antenna proposed by the present invention;

[0075] Figure 5 is the architecture diagram of the switchgear temperature and humidity detection system based on the conformal sensing tag proposed by the present invention;

[0076] Figure 6 is the working flow chart of the switchgear temperature and humidity detection system based on the conformal sensing tag proposed by the present invention.

[0077] In the figure: 1 - temperature and humidity sensing chip, 2 - long - arm feeder, 3 - short - arm feeder, 4 - straight - line connecting arm, 5 - inner longitudinal radiation arm, 6 - U - shaped bent radiation arm, 7 - outer longitudinal radiation arm, 8 - arc - shaped bent connecting arm, 9 - first long - horizontal radiation arm, 10 - second long - longitudinal radiation arm, 11 - first connecting arm, 12 - first longitudinal radiation arm, 13 - lower horizontal radiation arm, 14 - second longitudinal radiation arm, 15 - upper horizontal radiation arm, 16 - third longitudinal radiation arm, 17 - second connecting arm, 18 - rectangular radiation arm. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0078] Referring to Figure 1 , a switchgear temperature and humidity monitoring method based on a conformal sensing tag proposed by the present invention has the following method steps:

[0079] S1: Collect and store data through an auto - encoder to compress the temperature and humidity time - series data;

[0080] In this embodiment:

[0081] Input: Temperature and humidity data within a time window

[0082] Encoder: z = f e (x i ) = σ(W e x i + b e )

[0083] Decoder:

[0084] Loss function (mean squared error):

[0085] where T t-k , …, T t represents the temperature sequence from time t - n to the current time t, H t-k , …, H t represents the humidity sequence from time t - n to the current time t, k is the window length, W e , b e are the encoder weights and biases; W d , b d are the decoder weights and biases; σ is the activation function (such as ReLU).

[0086] Obviously, based on the above: data collection and storage can be achieved through the autoencoder, compressing the temperature and humidity time series data and reducing the storage cost.

[0087] S2: Read the compressed data in step S1 through Kalman filtering and decode it to recover the time series data for denoising;

[0088] In this embodiment:

[0089] Use Kalman filtering for data cleaning and preprocessing:

[0090] The state equation is:

[0091] The observation equation is:

[0092] where ΔT t , ΔH t are the rates of change of temperature and humidity; Δt is the time difference between two adjacent time steps, that is, the time length from time t - 1 to t; is the process noise; T t obs is the actual temperature value measured by the sensor at time t, is the actual humidity value measured by the sensor at time t, is the observation noise.

[0093] Obviously, based on the above: Kalman filtering can smooth the data in real time, eliminate noise, and provide a clean input for subsequent analysis.

[0094] S3: Apply the data processed in step S2 and perform dynamic optimization of the data threshold using the Q-Learning algorithm;

[0095] In this embodiment:

[0096] Perform dynamic optimization of the data threshold using the Q-Learning algorithm:

[0097] State:

[0098] Among them, T t k is the temperature data after Kalman filtering output, is the humidity data after Kalman filtering output;

[0099] Action: Adjust the threshold a ∈ {ΔT max , ΔH max};

[0100] Q-value update:

[0101] Reward function:

[0102] Among them, α is the learning rate and γ is the learning factor.

[0103] Obviously, based on the above: Through the learning of historical data and environmental feedback by the Q-Learning algorithm, the Q-value is continuously updated to find the optimal mapping between the temperature and humidity state and the threshold adjustment action, replacing manual parameter tuning, adapting to complex environmental changes, and improving the intelligent level of power grid equipment operation and maintenance.

[0104] S4: Apply the data processed in step S3, predict the data anomaly probability through the logistic regression algorithm, and perform data anomaly segmentation and anomaly marking based on the data anomaly probability;

[0105] In this embodiment:

[0106] Predict the data anomaly probability through the logistic regression algorithm and perform data anomaly segmentation and anomaly marking based on the data anomaly probability:

[0107] Input features:

[0108] Among them, ΔT t k is the temperature change rate after Kalman filtering output, is the humidity change rate after Kalman filtering output;

[0109] Predicted probability:

[0110] Classification rules:

[0111] where β0, β1, …, β4 are model coefficients.

[0112] Obviously, based on the above: By learning the patterns of temperature and humidity and their changes through the logistic regression algorithm, single-point anomalies are detected in real time, effectively ensuring the instantaneous response ability of the system.

[0113] S5: Apply the data marked as abnormal in step S4, calculate the anomaly score through the Isolation Forest model, and perform advanced detection;

[0114] In this embodiment:

[0115] Calculate the anomaly score through the Isolation Forest model and perform advanced detection:

[0116] Anomaly score:

[0117] where h(x) is the path length and E is the expected value, is the harmonic number;

[0118] If Score(x) > 0.6, it is confirmed as a real anomaly;

[0119] Splitting rule: Randomly select a feature T t k or Randomly select a separation value.

[0120] Obviously, based on the above: By analyzing the isolability of data points through the Isolation Forest, false alarms are effectively filtered and real anomalies are confirmed. Its collaborative work with logistic regression forms a dual guarantee mechanism, significantly improving the accuracy and reliability of the system.

[0121] S6: Obtain the data anomaly probability obtained in step S4 and the data anomaly score obtained in step S5, set the alarm decision according to the numerical intervals into which the data anomaly probability and the data anomaly score fall, and output the alarm result;

[0122] In this embodiment:

[0123] Set the alarm decision according to the numerical intervals into which the data anomaly probability and the data anomaly score fall, and output the alarm result:

[0124] Alarm rule logic: If the anomaly probability P > 0.9 and Score(x) > 0.6, then perform a first-level alarm;

[0125] If the anomaly probability P > 0.7 and does not meet the conditions for the first-level alarm, then execute a second-level alarm.

[0126] Obviously, based on the above: through the design of the alarm decision rule, it not only ensures the rapid response to critical anomalies but also avoids missing potential risks, meeting the intelligent requirements of power grid equipment operation and maintenance.

[0127] S7: Continuously and dynamically monitor the data using the PCA algorithm, and implement feedback regulation based on the monitoring results.

[0128] In this embodiment:

[0129] Continuously and dynamically monitor the data using the PCA algorithm, and implement feedback regulation based on the monitoring results:

[0130] Covariance matrix:

[0131]

[0132] Among them, Var is the variance and Cov is the covariance;

[0133] KL divergence detects drift:

[0134]

[0135] Among them, U is the historical distribution and V is the current distribution;

[0136] If D KL > 0.1, it is determined that data drift has occurred, then feedback regulation is triggered, Q-Learning training is restarted, and the threshold is dynamically adjusted; and model retraining is performed to update the parameters of the logistic regression and isolation forest models, and at the same time, system self-optimization signals (such as threshold update logs, model version numbers) are output.

[0137] Obviously, based on the above: through the combined use of the covariance matrix and KL divergence in the PCA algorithm, the system can detect data distribution drift, perceive data distribution changes, and adaptively adjust strategies to improve the intelligent level of power grid equipment operation and maintenance.

[0138] As another embodiment of the present application, referring to Figure 2 and Figure 3, in this embodiment, a conformal sensing tag for switchgear is proposed, which can be applied to any implementation mode of the above-mentioned switchgear temperature and humidity monitoring method based on the conformal sensing tag. The conformal sensing tag includes a temperature and humidity sensing chip 1 and an antenna radiation structure. The antenna radiation structure includes a pair of radiation conductors symmetrically distributed along the axis of the humidity sensing chip 1. Each radiation conductor includes an inner longitudinal radiation arm 5, a horizontal S-shaped radiation arm, a horizontal L-shaped radiation arm, and a rectangular radiation arm 18. The humidity sensing chip 1 is sequentially connected to one end of the inner longitudinal radiation arm 5 through an L-shaped feeder, a straight connecting arm 4. The humidity sensing chip 1 and the inner longitudinal radiation arm 5 are arranged inside the switchgear. The other end of the inner longitudinal radiation arm 5 is provided with a U-shaped bent radiation arm 6 adapted to the slot on the side wall of the switchgear. One end of the U-shaped bent radiation arm 6 is connected to the inner longitudinal radiation arm 5, and the other end of the U-shaped bent radiation arm 6 is sequentially connected to the rectangular radiation arm 18 through an outer longitudinal radiation arm 7, an arc-shaped bent connecting arm 8, a horizontal L-shaped radiation arm, and a horizontal S-shaped radiation arm. The horizontal L-shaped radiation arm, the horizontal S-shaped radiation arm, and the rectangular radiation arm 18 are all arranged outside the switchgear.

[0139] In this embodiment:

[0140] The temperature and humidity sensing chip 1 includes a MEMS temperature-sensitive element and a humidity-sensitive capacitor; it is integrated on a flexible conformal substrate with a curvature radius of 5-20 mm, which is adapted to the surface curvature of the switchgear;

[0141] The L-shaped feeder includes a long-arm feeder 2 and a short-arm feeder 3 connected in sequence;

[0142] The horizontal L-shaped radiation arm includes a long horizontal radiation arm one 9 and a long vertical radiation arm two 10 connected in sequence;

[0143] The horizontal S-shaped radiation arm includes a longitudinal one radiation arm 12, a lower horizontal radiation arm 13, a longitudinal two radiation arm 14, an upper horizontal radiation arm 15, and a longitudinal three radiation arm 16 connected in sequence. The longitudinal one radiation arm 12 is connected to the long vertical radiation arm two 10 through a first connecting arm 11, and the longitudinal three radiation arm 16 is connected to the rectangular radiation arm 18 through a second connecting arm 17.

[0144] Specifically:

[0145] The tag is subjected to anti-metallization treatment by using a structure of stacking the chip, the intermediate dielectric layer, and the metal layer, and is encapsulated with high-temperature-resistant ceramics to enhance the mechanical strength.

[0146] When the operating frequency of the temperature and humidity sensing chip 1 is 925 MHz, it has an input impedance and a threshold power of -17.4 dBm.

[0147] Refer to Figure 4 , the copper layer thickness of the antenna radiation structure is 0.2 mm, the line width W1 is 0.6 ± 0.05 mm, and the overall thickness ≤ 0.8 mm;

[0148] The long-arm feeder 2, with a length L1 of 9.8 ± 0.3 mm, is directly connected to the temperature and humidity sensing chip 1;

[0149] The short-arm feeder 3, with a length W2 of 4 ± 0.2 mm, is connected to the straight connecting arm 4;

[0150] The width tolerance of the long-arm feeder 2 and the short-arm feeder 3 is ≤ ±5%; the straight connecting arm 4 is symmetrical with respect to the axis of symmetry of the label.

[0151] The outer-bent U-shaped radiation arm includes:

[0152] The inner longitudinal radiation arm 5, with a length W3 of 3 ± 0.2 mm, is connected to the middle part of the straight connecting arm 4;

[0153] The outer longitudinal radiation arm 7, with a length W4 of 1 ± 0.2 mm, is connected to the arc-shaped bent connecting arm 8;

[0154] The bending radius R1 of the U-shaped bent radiation arm 6 is 0.8 ± 0.1 mm;

[0155] The length ratio of the inner longitudinal radiation arm 5 to the outer longitudinal radiation arm 7 is 3:1, and the width tolerance is ≤ ±5%.

[0156] The horizontal L-shaped radiation arm includes:

[0157] The long longitudinal radiation arm 10, with a length W5 of 12 mm ± 0.3 mm, is connected to the first connecting arm 11;

[0158] The length L3 of the long transverse radiation arm 9 is 9.2 ± 0.3 mm; the width tolerance of the long longitudinal radiation arm 10 and the long transverse radiation arm 9 is ≤ ±5%;

[0159] The bending radius R2 of the arc-shaped bent connecting arm 8 is 0.6 ± 0.1 mm.

[0160] The horizontal S-shaped radiation arm contains 3 periodic bending units, with a bending angle of 170° - 190°, and the adjacent bending spacing is 1.3 ± 0.1 mm, including: the longitudinal first radiation arm 12, the lower transverse radiation arm 13, the longitudinal second radiation arm 14, the upper transverse radiation arm 15, and the longitudinal third radiation arm 16;

[0161] Among them, the lengths W6 of the longitudinal first radiation arm 12, the longitudinal second radiation arm 14, and the longitudinal third radiation arm 16 are all the same, being 5.1 ± 0.2 mm;

[0162] The lengths L4 of the upper transverse radiation arm 15, the lower transverse radiation arm 13, and the first connecting arm 11 are all the same, being 1.3 ± 0.1 mm; the length L5 of the second connecting arm 17 is 1.7 ± 0.1 mm.

[0163] The length L6 of the rectangular radiation arm 18 is 15.6 ± 0.3 mm, and the width W8 is 17 ± 0.3 mm; among which, the distance W7 from the second connecting arm 17 to the top end of the rectangular radiation arm 18 is 12.4 ± 0.5 mm.

[0164] As another embodiment of the present application, referring to Figure 5 and Figure 6 , this embodiment proposes a temperature and humidity detection system for switchgear based on conformal sensing tags, including:

[0165] Conformal sensing tags: used to collect temperature and humidity sensing data inside the switchgear;

[0166] Reader / writer: deployed outside the switchgear, used to activate the conformal sensing tags and receive the sensing data returned by them;

[0167] Data processing and observation platform: communicatively connected to the reader / writer, applied to any implementation manner of the above-mentioned temperature and humidity monitoring method for switchgear based on conformal sensing tags, parsing the sensing data and generating temperature and humidity alarm signals.

[0168] Specifically, the operating frequency band of the reader / writer is 840 - 960 MHz.

[0169] In summary, through the design of passive conformal sensing tags, the present invention breaks through the electromagnetic shielding limit through the external bending type U-shaped radiation antenna structure to achieve efficient signal transmission inside and outside the cabinet; at the same time, it adopts a multi-modal data processing algorithm, integrating an autoencoder, a Kalman filter, Q-Learning dynamic threshold optimization, and an isolation forest model, providing functions of efficient data processing and temperature and humidity threshold classification, improving the accuracy of anomaly detection, and ensuring the high reliability of the system; at the same time, it also provides a temperature and humidity monitoring system for switchgear based on conformal sensing tags, which not only avoids problems such as personnel transfer, sensor maintenance, battery replacement, and equipment repair in the traditional power grid switchgear temperature and humidity monitoring, but also provides stronger anti-interference ability, and is applicable to temperature and humidity monitoring in various power equipment and industrial environments.

Claims

1. A method for monitoring the temperature and humidity of a switchgear cabinet based on a conformal sensing tag, characterized in that, The method steps are as follows: S1: Collect and store data through an autoencoder, and compress the temperature and humidity time-series data; S2: Read the compressed data in step S1 through Kalman filtering and decode it to restore it to time-series data for denoising; S3: Apply the data processed in step S2 and use the Q-Learning algorithm to dynamically optimize the data threshold; S4: Apply the data processed in step S3, predict the data anomaly probability through the logistic regression algorithm, and perform data anomaly segmentation and anomaly marking based on the data anomaly probability; S5: Apply the data marked as abnormal in step S4, calculate the anomaly score through the Isolation Forest model for advanced detection; S6: Obtain the data anomaly probability obtained in step S4 and the data anomaly score obtained in step S5, set the alarm decision based on the numerical intervals into which the data anomaly probability and the data anomaly score fall, and output the alarm result; S7: Use the PCA algorithm to continuously and dynamically monitor the data, and implement feedback regulation based on the monitoring results.

2. The method for monitoring the temperature and humidity of a switchgear cabinet based on a conformal sensing tag according to claim 1, wherein, In step S1: Input: Temperature and humidity data within a time window Encoder: z = f e (x i ) = σ(W e x i + b e ) Decoder: Loss function (mean squared error): Among them, T t-k ,..., T t represents the temperature sequence from time t - n to the current time t, H t-k ,..., H t represents the humidity sequence from time t - n to the current time t, k is the window length, W e , b e are the encoder weights and biases; W d , b d are the decoder weights and biases; σ is the activation function.

3. The method for monitoring the temperature and humidity of a switchgear cabinet based on a conformal sensing tag according to claim 1, wherein In step S2: Use Kalman filtering for data cleaning and preprocessing: The state equation is as follows: The observation equation is as follows: where, ΔT t , ΔH t are the rates of change of temperature and humidity; Δt is the time difference between two adjacent time steps, i.e., the time length from time step t-1 to t; is the process noise; T t obs is the actual temperature value measured by the sensor at time step t, is the actual humidity value measured by the sensor at time step t, is the observation noise.

4. The method for monitoring the temperature and humidity of a switchgear cabinet based on a conformal sensing tag according to claim 3, wherein, In step S3: Use the Q-Learning algorithm to dynamically optimize the data threshold: Status: Among them, T t k is the temperature data after Kalman filtering output, is the humidity data after Kalman filtering output; Action: Adjust the threshold \(a\in\{\Delta T max ,\Delta H max \}\); Q-value update: Reward function: where α is the learning rate and γ is the learning factor.

5. A method for monitoring the temperature and humidity of a switchgear cabinet based on a conformal sensing tag according to claim 4, characterized in that, In step S4: Predict the data anomaly probability through the logistic regression algorithm, and perform data anomaly segmentation and anomaly marking based on the data anomaly probability: Input features: Among them, ΔT t k is the temperature change rate after Kalman filtering output, is the humidity change rate after Kalman filtering output; Predicted probability: Classification rules: where β0, β1,..., β4 are the model coefficients.

6. A method for monitoring the temperature and humidity of a switchgear cabinet based on a conformal sensing tag according to claim 5, characterized in that, In step S5: Calculate the anomaly score through the Isolation Forest model for advanced detection: Abnormal score: where h(x) is the path length and E is the expected value, is the harmonic number; If Score(x)>0.6, it is confirmed as a real anomaly; Splitting rule: Randomly select feature T t k or Randomly select a separation value.

7. The method for monitoring the temperature and humidity of a switchgear cabinet based on a conformal sensing tag according to claim 6, characterized in that, In step S6: Set the alarm decision based on the numerical intervals into which the data anomaly probability and the data anomaly score fall, and output the alarm result: Alarm rule logic: If the anomaly probability P>0.9 and Score(x)>0.6, then a first-level alarm is issued; If the anomaly probability P>0.7 and the first-level alarm condition is not met, then a second-level alarm is executed.

8. A method for monitoring the temperature and humidity of a switchgear based on a conformal sensing tag according to claim 1, characterized in that, In step S7: Use the PCA algorithm to continuously and dynamically monitor the data, and implement feedback regulation based on the monitoring results: Covariance matrix: where Var is the variance and Cov is the covariance; KL divergence to detect drift: where U is the historical distribution and V is the current distribution; If D KL > 0.1, it is determined that data drift has occurred, then feedback adjustment is triggered, Q-Learning training is restarted, the threshold is dynamically adjusted; and model retraining is performed to update the parameters of the logistic regression and isolation forest models.

9. A conformal sensing label for a switchgear, characterized in that, It can be applied to the switchgear temperature and humidity monitoring method based on conformal sensing tags described in any one of claims 1-7. The conformal sensing tag includes a temperature and humidity sensing chip (1) and an antenna radiation structure. The antenna radiation structure includes a pair of radiation conductors symmetrically distributed along the axis of the humidity sensing chip (1). Each radiation conductor includes an inner longitudinal radiation arm (5), a horizontal S-shaped radiation arm, a horizontal L-shaped radiation arm, and a rectangular radiation arm (18). The humidity sensing chip (1) is sequentially connected to one end of the inner longitudinal radiation arm (5) through an L-shaped feeder, a straight connecting arm (4). The humidity sensing chip (1) and the inner longitudinal radiation arm (5) are arranged inside the switchgear. The other end of the inner longitudinal radiation arm (5) is provided with a U-shaped bent radiation arm (6) adapted to the slot on the side wall of the switchgear. One end of the U-shaped bent radiation arm (6) is connected to the inner longitudinal radiation arm (5), and the other end of the U-shaped bent radiation arm (6) is sequentially connected to the rectangular radiation arm (18) through an outer longitudinal radiation arm (7), an arc-shaped bent connecting arm (8), a horizontal L-shaped radiation arm, and a horizontal S-shaped radiation arm. The horizontal L-shaped radiation arm, the horizontal S-shaped radiation arm, and the rectangular radiation arm (18) are all arranged outside the switchgear.

10. A temperature and humidity detection system for switchgear based on conformal sensing tags, characterized in that, Comprising: Conformal sensing tag: used to collect temperature and humidity sensing data inside the switchgear; Reader: deployed outside the switchgear, used to activate the conformal sensing tag and receive the sensing data returned by it; Data processing and observation platform: communicatively connected to the reader, applying the switchgear temperature and humidity monitoring method based on conformal sensing tags described in any one of claims 1-7, parsing the sensing data and generating temperature and humidity alarm signals.

Images