NO2 sensor and real-time monitoring method for air discharge of air switch cabinet

Abstract

The invention discloses an NO2 sensor and an air switch cabinet air discharge real-time monitoring method, and the construction of a sensor array of the NO2 sensor specifically comprises the following steps: depositing a heating electrode on a silicon substrate as a heating layer; and a plurality of interdigital electrodes are arranged at the top of the sensor array and are used for measuring the resistance of various materials. An interdigital electrode layer is arranged on the heating layer, specifically, the arrangement mode between electrodes is similar to a finger staggered structure, and the four interdigital electrodes in the interdigital electrode layer are used for measuring the resistance of different materials. Si3N4 layers are arranged between the silicon substrate and the heating layer and between the heating layer and the interdigital electrode layer to serve as insulating layers. The sensor array of the NO2 sensor has good selectivity, and can be used for detecting 250ppb of NO2; the sensor shows a good response recovery characteristic on NO2; and estimation is carried out through the CNN-ALSTM model, the concentration is predicted, and the prediction result is relatively accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of air quality detection, and in particular to an NO2 sensor and a method for real-time monitoring of air discharge in an air switch cabinet based on the NO2 sensor. Background Art

[0002] In the field of electric power, air-insulated power equipment is the development direction of environmentally friendly switchgear. When local discharge or local overheating occurs in the air switch cabinet, the air and solid insulating materials inside the equipment decompose, resulting in the production of various characteristic gases, such as O3, NO, NO2, and CO. Studies have shown that there is a linear relationship between the concentration of NO2 and CO gases and the severity of the fault. Therefore, the type and severity of air discharge faults can be determined by NO2 detection. In previous studies, metal oxide semiconductor sensors have the advantages of small size, low cost, and fast response speed, and are widely used for gas detection. For example, In2O3 has received considerable attention in the field of gas sensing due to its high sensitivity to NO2, and noble metal doping and heterostructure manufacturing are usually used to improve the sensing performance of pure oxide materials. Some researchers have constructed MoS2 / In2O3 heterostructures for sensing improvement of pure oxide materials, and the modified materials show high response and excellent selectivity for 5ppm NO2.

[0003] The metal oxide semiconductor sensors currently used for NO2 gas detection have the advantages of small size, low cost, and fast response speed. However, when the sensor is exposed to interfering gases (such as CO), the sensor also exhibits cross-sensitivity defects. Although the performance of the sensor can be improved by adopting methods of modified materials, NO2 detection relying on a single sensor cannot provide a multi-dimensional sensing signal. When using a sensor array for continuous exposure to the detected gas for online monitoring, in this case, the sensor is susceptible to gas adsorption poisoning effects, temperature and humidity drift, resulting in baseline drift and significant gas identification errors.

[0004] Existing metal oxide semiconductor sensors for NO2 gas detection have weak anti-interference capabilities when exposed to interfering gases, and the detection effect of the sensor will deteriorate when interfering gases are present. In addition, NO2 detection that relies on a single sensor cannot provide multi-dimensional sensing signals. When using a sensor array, it is easily affected by gas adsorption poisoning effects, temperature and humidity drift, which affects the detection performance.

[0005] Chinese patent document CN117191888 A discloses a gas sensor array and detection method for indoor air quality detection. First, a gas sensor array capable of detecting indoor pollutants such as formaldehyde, toluene, methane and NO2 is prepared. Concentration gradient tests are performed on formaldehyde, toluene, methane and NO2 at different concentrations at the optimal working temperature to obtain the response curve of each sensor in the test gas, and the maximum response value of the sensor array to the test gas is obtained. Then, two three-layer reverse neural network models are established to respectively identify the gas type and predict the gas concentration. The obtained responses are input into the model to achieve qualitative and quantitative measurement of indoor air pollutants. The classification accuracy can reach 100%, and the mean square error between the predicted concentration and the true value is 0.025. The sensor array has the characteristics of high sensitivity, low detection limit, small size, easy to carry, low cost, simple application and easy integration, and has broad application prospects.

[0006] Therefore, it is necessary to design an intelligent NO2 sensor and a real-time monitoring method for air discharge in an air switch cabinet based on the intelligent NO2 sensor. The NO2 sensor constructs a sensor array of various sensor materials, has good selectivity, is not easily interfered by other gases, and can be used to monitor NO2 gas in air switch cabinets with good accuracy, sensitivity, stability and selectivity. Summary of the invention

[0007] The purpose of the present invention is to provide a NO2 sensor to solve all or part of the above problems, construct a sensor array of multiple sensor materials, have good selectivity, not easily interfered by other gases, and can be used for monitoring NO2 gas in air switch cabinets, with good accuracy, sensitivity, stability and selectivity.

[0008] In order to solve the above technical problems, the technical solution adopted by the present invention is that the construction of the sensor array of the NO2 sensor specifically includes: depositing a heating electrode on a silicon substrate as a heating layer; and arranging a plurality of interdigitated electrodes on the top of the sensor array for measuring the resistance of a variety of materials.

[0009] By adopting the above technical scheme, the NO2 sensor constructs a sensor array of various sensor materials and introduces the CNN-ALSTM model, which improves the accuracy of the model in identifying NO2. It has good selectivity and is not easily interfered by other gases. It can be used to monitor NO2 gas in air switch cabinets with good accuracy, sensitivity, stability and selectivity.

[0010] Preferably, the interdigital electrodes are arranged in a staggered manner to form an interdigital electrode layer, and the interdigital electrodes in the interdigital electrode layer are used to measure the resistance of different materials.

[0011] Preferably, an insulating layer is provided between the silicon substrate and the heating layer, and between the heating layer and the interdigital electrode layer, respectively.

[0012] Preferably, four interdigital electrodes are arranged on the top of the sensor array to measure the resistance of four materials, specifically: 10-15wt% WO3-10-15% SnCl2-In2O3, 5-10wt% NiO-10-15% SnCl2-In2O3, 10-15wt% TiO2-10-15% SnCl2-In2O3, 5-10wt% SnO2-10-15% SnCl2-In2O3; the four interdigital electrodes are respectively welded on the four corner pins of the interdigital electrode layer. The sensor array based on four materials is adopted, and the CNN-ALSTM model is introduced to improve the accuracy of the model in identifying NO2, and it has good accuracy, sensitivity, stability and selectivity, and is more suitable for gas monitoring in air switch cabinets.

[0013] Preferably, the sensor array is packaged on a small circuit board; the pads of the sensor array are connected to the circuit board, and a metal shell is used to package the NO2 sensor.

[0014] Preferably, the size of the circuit board is 8×12 mm, and the length, width and height of the metal shell are 6-7 cm, 7.5-8 cm and 5-6 cm respectively. Preferably, the size of the circuit board is 8×12 mm, and the length, width and height of the stainless steel metal shell are 7 cm, 7.5 cm and 5 cm respectively.

[0015] Preferably, the interdigital electrodes are a gold-plated layer with a thickness of 250 to 255 nm.

[0016] The purpose of the present invention is to provide a real-time monitoring method for air discharge of an air switch cabinet based on a NO2 sensor, which has good accuracy, sensitivity, stability and selectivity.

[0017] In order to solve the above technical problems, the present invention also provides a real-time monitoring method for air discharge of an air switch cabinet based on the above NO2 sensor, which specifically includes the following steps:

[0018] S1. Design of gas-sensitive measurement system: Place the NO2 sensor in an air chamber, the air chamber is provided with an air inlet and an air outlet, apply a reference voltage and a heating voltage to the NO2 sensor, and connect the client; the gas enters the air chamber through the air inlet;

[0019] S2. Design of detection device: The NO2 detection device consists of a gas-sensitive measurement system, a temperature and humidity sensor (SHT30) and a microcontroller (STM32 F103C8T6); the gas-sensitive measurement system and the temperature and humidity sensor (SHT30) are connected to the microcontroller (STM32F103C8T6) respectively; the microcontroller (STM32F103C8T6) is connected to the client; and real-time detection is performed to continuously collect and further transmit the gas sensor signals collected by the gas-sensitive measurement system and the ambient temperature and humidity signals collected by the temperature and humidity sensor to the computer;

[0020] S3, gas testing: adjust through the client and control the intake air type and concentration through the microcontroller, collect the resistance changes of the gas sensitive measurement system and the temperature and humidity sensor through the data acquisition card; repeat the collection multiple times to obtain the data set and divide the data set;

[0021] S4. Training CNN-ALSTM model: construct and train the CNN-ALSTM model, input the data obtained in step S3 into the CNN-ALSTM model, and predict the concentration of the gas to be tested.

[0022] By adopting the above technical solution, a NO2 sensor with a sensing array of four sensor materials is used, and a CNN-ALSTM model is introduced at the same time, which improves the accuracy of the model in identifying NO2, has good selectivity, is not easily interfered by other gases, and can be used for monitoring NO2 gas in air switch cabinets.

[0023] Preferably, the specific steps of step S4 are:

[0024] S41, data preprocessing: First, before applying the model, the data set is normalized and preprocessed; normalizing the data is helpful to accelerate the convergence speed of the model and improve the accuracy of gas concentration estimation;

[0025] S42. Build and train the model: The gas concentration estimation model CNN-ALSTM model consists of four parts: CNN layer, attention layer, RNN layer and fully connected layer. The training set is used for training to obtain the CNN-ALSTM model.

[0026] S43, indication evaluation: using root mean square error RMSE and determination coefficient R 2 The trained CNN-ALSTM model is evaluated by indicators, where the root mean square error RMSE is calculated as follows:

[0027]

[0028] In the formula, c pre and c are the predicted gas concentration and the actual gas concentration respectively; T Sis the total time step; it can be seen that the smaller the RMSE, the closer the predicted value is to the true value, indicating that the deviation of the model prediction is smaller;

[0029] Coefficient of determination R 2 The calculation formula is:

[0030]

[0031] In the formula, Represents the average value of the real gas concentration; compared with RMSE, a coefficient of determination close to 1 indicates that the regression model is better;

[0032] S44, gas concentration estimation: Considering the continuity of the time series, a sliding window is used to select input data with a window size of t; the 4×t sensor signals are input into the CNN-ALSTM model to estimate the gas concentration at the current moment.

[0033] Preferably, the gas-sensitive measurement system in step S1 further includes a power supply; and the top of the NO2 detection device in step S2 is detachably connected to a metal dust screen for protecting the sensor and providing a channel for the diffusion of surrounding gas molecules.

[0034] Compared with the prior art, the present invention has the following beneficial effects: the sensor array of the NO2 sensor has good selectivity and can detect 250 ppb of NO2; and the sensor exhibits good response recovery characteristics to NO2; and the concentration is estimated and predicted through the CNN-ALSTM model, and the NO2 concentration estimated by the model is relatively accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The present invention will now be described by way of example with reference to the accompanying drawings, in which:

[0036] Figure 1 A schematic diagram of a sensor array of the intelligent NO2 sensor of the present invention;

[0037] Figure 2 A schematic diagram of a gas-sensitive measurement system of an intelligent NO2 sensor of the present invention;

[0038] Figure 3 : is a response recovery characteristic diagram of the intelligent NO2 sensor of the present invention; wherein Figure 3 (a) to (d) are the responses of 10wt% WO3-10% SnCl2-In2O3 (W-Sn-In) to the detection of 5ppm, 10ppm, 30ppm, and 50ppm NO2; Figure 3 (e) is the response recovery of 10wt% WO3-10% SnCl2-In2O3 (W-Sn-In) detecting NO2; Figure 3(f) is the response recovery of 5wt% NiO-10% SnCl2-In2O3 (Ni-Sn-In) detecting NO2; Figure 3 (g) in the figure is the response recovery of 10wt% TiO2-10% SnCl2-In2O3 (Ti-Sn-In) detecting NO2; Figure 3 (h) is the response recovery of 5wt% SnO2-10% SnCl2-In2O3 (Sn-Sn-In) detecting NO2;

[0039] Figure 4 It is a schematic diagram for comparing the estimated value and the actual situation of NO2 concentration at different humidity in the real-time monitoring method of air discharge of air switch cabinet based on NO2 sensor of the present invention;

[0040] Figure 5 The SEM images of four materials of the sensor array of the smart NO2 sensor of the present invention; Figure 5 (a) is 10wt% WO3-10% SnCl2-In2O3 (W-Sn-In); Figure 5 (b) is 5wt% NiO-10% SnCl2-In2O3 (Ni-Sn-In); Figure 5 (c) is 10wt% TiO2-10% SnCl2-In2O3 (Ti-Sn-In); Figure 5 (d) is 5wt% SnO2-10% SnCl2-In2O3 (Sn-Sn-In);

[0041] Figure 6 The present invention is a flow chart of a gas concentration estimation method based on a continuous response mode in a real-time monitoring method for air discharge in an air switch cabinet based on a NO2 sensor. DETAILED DESCRIPTION

[0042] All features disclosed in this specification, or steps in all methods or processes disclosed, except mutually exclusive features and / or steps, can be combined in any manner.

[0043] Any feature disclosed in this specification (including any additional claims and abstract), unless otherwise stated, may be replaced by other equivalent or alternative features having similar purposes. That is, unless otherwise stated, each feature is only an example of a series of equivalent or similar features.

[0044] Example: Figure 1The figure shows a schematic diagram of a sensor array; the construction of the sensor array of the NO2 sensor specifically includes: depositing a 200nmPt heating electrode on a 500nm silicon substrate as a heating layer; arranging a plurality of interdigital electrodes on the top of the sensor array for measuring the resistance of a variety of materials; a plurality of interdigital electrodes are arranged in a staggered manner to form an interdigital electrode layer, and the arrangement of the interdigital electrodes is similar to the structure of interdigital fingers, and each interdigital electrode in the interdigital electrode layer is used to measure the resistance of different materials; and a 460nm added Pt heating electrode is arranged between the silicon substrate and the heating layer and between the heating layer and the interdigital electrode layer. The Si3N4 layer is used as an insulating layer; four interdigital electrodes are arranged on the top of the sensor array to measure the resistance of four materials, specifically: 10-15wt% WO3-10-15% SnCl2-In2O3, 5-10wt% NiO-10-15% SnCl2-In2O3, 10-15wt% TiO2-10-15% SnCl2-In2O3, 5-10wt% SnO2-10-15% SnCl2-In2O3; the four interdigital electrodes are respectively welded to the four corner pins of the interdigital electrode layer through platinum wire leads. The sensor array is packaged on a small circuit board; the pads of the sensor array are connected to the circuit board by an aluminum wire welding machine, and the NO2 sensor is packaged in a stainless steel metal shell. The size of the circuit board is 8×12mm, and the length, width and height of the stainless steel metal shell are 6-7cm, 7.5-8cm and 5-6cm respectively. The interdigital electrodes are a 250-255nm gold-plated layer.

[0045] Specific embodiment: The construction of the sensor array of the NO2 sensor specifically includes: depositing a 200nmPt heating electrode on a 500nm silicon substrate as a heating layer; arranging a plurality of interdigitated electrodes on the top of the sensor array for measuring the resistance of a plurality of materials; arranging an interdigitated electrode layer on the heating layer, specifically, the arrangement of the electrodes is similar to the structure of interlaced fingers, and the four interdigitated electrodes in the interdigitated electrode layer are used to measure the resistance of different materials; and respectively arranging a plurality of interdigitated electrodes between the silicon substrate and the heating layer, and between the heating layer and the interdigitated electrode layer. A 460nm Si3N4 layer is added as an insulating layer; four interdigital electrodes are set on the top of the sensor array to measure the resistance of four materials, specifically: 10wt% WO3-10% SnCl2-In2O3, 5wt% NiO-10% SnCl2-In2O3, 10wt% TiO2-10% SnCl2-In2O3, 5wt% SnO2-10% SnCl2-In2O3; the four interdigital electrodes are welded to the four corner pins of the interdigital electrode layer through platinum wire leads. The SEM images of the four materials of the sensor array are shown in Figure 5; Figure 5(a) is 10wt% WO3-10% SnCl2-In2O3 (W-Sn-In); Figure 5 (b) is 5wt%NiO-10%SnCl2-In2O3(Ni-Sn-In); Figure 5 (c) is 10wt% TiO2-10% SnCl2-In2O3 (Ti-Sn-In); Figure 5 (d) is 5wt% SnO2-10% SnCl2-In2O3 (Sn-Sn-In); Figure 5 As can be seen from (a) to (d), the particles of all four materials (W-Sn-In, Ni-Sn-In, Ti-Sn-In and Sn-Sn-In) show an irregular polyhedral structure with a diameter of about 200nm. The sensor array is packaged on a small circuit board; the pads of the sensor array are connected to the circuit board by an aluminum wire welding machine, and the NO2 sensor is packaged in a stainless steel metal shell. The size of the circuit board is 8×12mm, and the length, width and height of the stainless steel metal shell are 7cm, 7.5cm and 5cm respectively. The interdigital electrode is a 250nm gold-plated layer.

[0046] The real-time monitoring method for air discharge in an air switch cabinet based on a NO2 sensor proposed in the embodiment of the present application specifically includes the following steps:

[0047] S1. Design of gas-sensitive measurement system: Place the NO2 sensor in an air chamber, which is provided with an air inlet and an air outlet. Apply reference voltage and heating voltage to the NO2 sensor and connect it to the client. The gas enters the air chamber through the air inlet. Figure 2 As shown; the gas-sensitive measurement system also includes power supply, signal transmission and other modules; used to test the response characteristics of the sensor array under standard gas;

[0048] S2. Design of the detection device: The NO2 detection device consists of a gas-sensitive measurement system, a temperature and humidity sensor (SHT30) and a microcontroller (STM32 F103C8T6); the gas-sensitive measurement system and the temperature and humidity sensor (SHT30) are connected to the microcontroller (STM32F103C8T6) respectively; the microcontroller (STM32F103C8T6) is connected to the client; and real-time detection is performed, and the gas sensor signals collected by the gas-sensitive measurement system and the ambient temperature and humidity signals collected by the temperature and humidity sensor are continuously collected and further transmitted to the computer; the sensor array is packaged on a small circuit board; the pads of the sensor array are connected to the circuit board by an aluminum wire welding machine, and the NO2 sensor is packaged in a stainless steel metal shell. The size of the circuit board is 8×12mm, and the length, width and height of the stainless steel metal shell are 7cm, 7.5cm and 5cm respectively. The interdigital electrode is a 250nm gold-plated layer; the top of the NO2 detection device in step S2 is detachably connected to a metal dustproof net to protect the sensor and provide a channel for the diffusion of surrounding gas molecules;

[0049] S3, gas testing: adjust through the client and control the intake air type and concentration through the microcontroller, collect the resistance changes of the gas sensitive measurement system and the temperature and humidity sensor through the data acquisition card; repeat the collection multiple times to obtain the data set and divide the data set;

[0050] S4, training CNN-ALSTM model: constructing and training the CNN-ALSTM model, inputting the data obtained in step S3 into the CNN-ALSTM model, and predicting the concentration of the gas to be tested;

[0051] like Figure 6 As shown, in order to predict the NO2 concentration within 1 ppm, a CNN-ALSTM gas concentration estimation model (referred to as CNN-ALSTM model) is proposed, including data preprocessing, model construction and index evaluation; the data set is composed of data detected by the sensor array of four time series in the NO2 detection device; the specific steps of step S4 are:

[0052] S41, data preprocessing: First, the data of the data set is normalized and preprocessed before being applied to the model; normalizing the data is helpful to accelerate the convergence speed of the model and improve the accuracy of gas concentration estimation;

[0053] S42. Build and train the model: The gas concentration estimation model CNN-ALSTM model consists of four parts: CNN layer, attention layer, RNN layer and fully connected layer. The training set is used for training to obtain the CNN-ALSTM model.

[0054] S43, indication evaluation: using root mean square error RMSE and determination coefficient R 2 The trained CNN-ALSTM model is evaluated by indicators, where the root mean square error RMSE is calculated as follows:

[0055]

[0056] In the formula, c pre and c are the predicted gas concentration and the actual gas concentration respectively; T S is the total time step; it can be seen that the smaller the RMSE, the closer the predicted value is to the true value, indicating that the deviation of the model prediction is smaller;

[0057] Coefficient of determination R 2 The calculation formula is:

[0058]

[0059] In the formula, Represents the average value of the real gas concentration. Compared with RMSE, a coefficient of determination close to 1 indicates that the regression model is better.

[0060] S44, gas concentration estimation: Considering the continuity of the time series, a sliding window is used to select input data, and the size of the sliding window is t (20s in this embodiment); 4×t (i.e., 4×20s) sensor signals are input into the CNN-ALSTM model to estimate the gas concentration at the current moment; Figure 6As shown in the figure; usually, the convolution layer extracts the features of the raw data by establishing multiple filters; for the sensor response signal, the one-dimensional CNN unit extracts the correlation of the data points and further generates more meaningful features; the length of the convolution kernel is 3, and the number of input and output channels is 4 and 8 respectively; the sliding direction of the convolution kernel is consistent with the time direction of the sensor array response; then, the high-dimensional features mapped from the CNN layer are input to the multi-head attention layer; by simulating the resource allocation of human brain attention, the attention mechanism focuses on the key information in the sequence; the multi-head attention mechanism consists of multiple attentions with different weight factors; each attention module obtains the correlation between various features by calculating the input matrix query, key and value; finally, The different features extracted from each head are connected to generate more global features. In this work, the data is parallelized through 4 scaled dot product attention layers, which is conducive to enhancing the resource allocation of feature values ​​and improving the accuracy of the model. The third part of the gas concentration estimation model uses a recurrent neural network, namely LSTM. The LSTM structure includes a forget gate, an input gate, and an output gate, which can discard worthless historical information and retain meaningful information from the past and current moments. A single-layer LSTM with 64 nodes in the hidden layer is used. In order to integrate the information of the last time step of the LSTM, a fully connected layer is built at the bottom of the module. The fully connected layer consists of two linear layers and an activation function. The rectified linear unit is used for the activation function. Finally, the model outputs the estimated target gas concentration.

[0061] The performance test is carried out using the above-mentioned air switch cabinet air discharge real-time monitoring method based on NO2 sensor, specifically:

[0062] (1) Sensitivity test: The sensor array has good selectivity and can detect 250ppb NO2. And the sensor shows good response recovery characteristics to NO2, such as Figure 3 shown; Figure 3 (a) to (d) are the responses of 10wt% WO3-10% SnCl2-In2O3 (W-Sn-In) to the detection of 5ppm, 10ppm, 30ppm, and 50ppm NO2; Figure 3 (e) is the response recovery of 10wt% WO3-10% SnCl2-In2O3 (W-Sn-In) detecting NO2; Figure 3 (f) is the response recovery of 5wt% NiO-10% SnCl2-In2O3 (Ni-Sn-In) detecting NO2; Figure 3 (g) in the figure is the response recovery of 10wt% TiO2-10% SnCl2-In2O3 (Ti-Sn-In) detecting NO2; Figure 3(h) is the response recovery of 5wt% SnO2-10% SnCl2-In2O3 (Sn-Sn-In) detecting NO2;

[0063] (2) Concentration prediction capability: The NO2 concentration estimated by the CNN-ALSTM model is relatively accurate. Figure 4 As shown in Figure 2, as the humidity increases, the error of the model prediction results will slightly increase; under 45% RH conditions, the RMSE and R 2 They are 107.94ppb and 0.8833 respectively. This may be because as the humidity increases, the baseline resistance of the sensor is larger and the anti-interference ability is stronger, which further leads to a decrease in prediction accuracy;

[0064] in Where c pre and c are the predicted gas concentration and the actual gas concentration, respectively. S is the total time step. It can be seen that the smaller the RMSE, the closer the predicted value is to the true value, indicating that the deviation of the model prediction is smaller;

[0065] In the formula Represents the average value of the real gas concentration; compared with RMSE, a coefficient of determination close to 1 indicates that the regression model is better.

[0066] The present invention is not limited to the above-mentioned specific embodiments, but extends to any new features or any new combination disclosed in this specification, as well as any new method or process steps or any new combination disclosed.

Claims

1. A NO2 sensor, characterized in that: The construction of the sensor array of the NO2 sensor specifically includes: depositing a heating electrode on a silicon substrate as a heating layer; and arranging a plurality of interdigital electrodes on the top of the sensor array for measuring the resistance of a plurality of materials.

2. The NO2 sensor according to claim 1, characterized in that: The interdigital electrodes are arranged in a staggered manner to form an interdigital electrode layer, and the interdigital electrodes in the interdigital electrode layer are used to measure the resistance of different materials.

3. The NO2 sensor according to claim 2, characterized in that: An insulating layer is disposed between the silicon substrate and the heating layer, and between the heating layer and the interdigital electrode layer.

4. The NO2 sensor according to claim 2, characterized in that: Four interdigital electrodes are arranged on the top of the sensor array for measuring the resistance of four materials, specifically: 10-15wt% WO3-10-15% SnCl2-In2O3, 5-10wt% NiO-10-15% SnCl2-In2O3, 10-15wt% TiO2-10-15% SnCl2-In2O3, 5-10wt% SnO2-10-15% SnCl2-In2O3; the four interdigital electrodes are respectively welded on the four corner pins of the interdigital electrode layer.

5. The NO2 sensor according to claim 3, characterized in that: The sensor array is packaged on a circuit board; the pads of the sensor array are connected to the circuit board, and a metal shell is used to package the NO2 sensor.

6. The NO2 sensor according to claim 5, characterized in that: The size of the circuit board is 8×12 mm, and the length, width and height of the metal shell are 6-7 cm, 7.5-8 cm and 5-6 cm respectively.

7. The NO2 sensor according to claim 4, characterized in that: The interdigital electrodes are a gold-plated layer with a thickness of 250 to 255 nm.

8. A method for real-time monitoring of air discharge in an air switch cabinet based on the NO2 sensor according to any one of claims 1 to 7, characterized in that: The specific steps include: S1. Design of gas-sensitive measurement system: The NO2 sensor is placed in the gas chamber, and the gas chamber is provided with an inlet. An air inlet and an air outlet are provided to apply a reference voltage and a heating voltage to the NO2 sensor and connect the client; the gas enters the air chamber through the air inlet; S2. Design of detection device: The NO2 detection device is composed of a gas-sensitive measurement system, a temperature and humidity sensor and a microcontroller; the gas-sensitive measurement system and the temperature and humidity sensor are respectively connected to the microcontroller; the microcontroller is connected to the client; and real-time detection is performed, and the gas sensor signals collected by the gas-sensitive measurement system and the ambient temperature and humidity signals collected by the temperature and humidity sensor are continuously collected and further transmitted to the computer; S3, gas testing: adjust through the client and control the intake air type and concentration through the microcontroller, collect the resistance changes of the gas sensitive measurement system and the temperature and humidity sensor through the data acquisition card; repeat the collection multiple times to obtain the data set and divide the data set; S4. Training CNN-ALSTM model: construct and train the CNN-ALSTM model, input the data obtained in step S3 into the CNN-ALSTM model, and predict the concentration of the gas to be tested.

9. The real-time monitoring method for air discharge of an air switch cabinet according to claim 8 is characterized in that: The specific steps of step S4 are: S41, data preprocessing: first normalize and preprocess the data of the data set; S42. Build and train the model: The CNN-ALSTM model consists of four parts: CNN layer, attention layer, RNN layer and fully connected layer. The training set is used for training to obtain the CNN-ALSTM model. S43, indication evaluation: using root mean square error RMSE and determination coefficient R 2 The trained CNN-ALSTM model is evaluated by indicators, where the root mean square error RMSE is calculated as follows: In the formula, c pre and c are the predicted gas concentration and the actual gas concentration respectively; T S is the total time step; Coefficient of determination R 2 The calculation formula is: In the formula, Represents the average value of the real gas concentration; S44, gas concentration estimation: A sliding window is used to select input data with a window size of t; 4×t sensor signals are input into the CNN-ALSTM model to estimate the gas concentration at the current moment.

10. The real-time monitoring method for air discharge of an air switch cabinet according to claim 8, characterized in that: The gas-sensitive measurement system in step S1 also includes a power supply; the top of the NO2 detection device in step S2 is connected to a metal dustproof net for protecting the sensor and providing a channel for the diffusion of surrounding gas molecules.

Images