Partial discharge intensive multi-mode equipment and data detection and diagnosis method based on Internet of Things

Abstract

The invention relates to the field of partial discharge detection, and provides a partial discharge intensive data and multi-mode equipment detection and diagnosis method based on the Internet of Things, and the method comprises the steps: collecting multi-mode detection data collected by a multi-source sensor disposed in equipment in the operation process of the equipment; carrying out preliminary data cleaning on the collected data in a local edge computing server of the equipment, and eliminating errors generated in data collection; transmitting the processed data to a central analysis platform, and dynamically distributing detection data weights according to data types and environmental parameters; and analyzing the processed data through a comprehensive analysis model, and judging whether partial discharge exists or not and the discharge intensity. On the basis of an end-to-end network information transmission mode, intensive management of partial discharge data is achieved, multi-mode detection data in equipment is collected, cleaning and data noise reduction are carried out on multi-source data, and meanwhile, the multi-source data weight is set, so that the partial discharge detection comprehensiveness and accuracy are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of equipment detection and diagnosis, and in particular to a method for local discharge intensive data and multi-mode detection and diagnosis equipment based on the Internet of Things. Background Art

[0002] With the development of high-voltage transmission technology, high-voltage electrical equipment plays an increasingly important role in power systems. However, as voltage increases, the electric field strength experienced by local insulation in electrical equipment also increases. At the same time, the unevenness of the electric field strength causes the insulation material to be exposed to uneven high-voltage conditions. Small air inclusions within the insulation of conductors can cause partial discharge in weak areas of the insulation material. Under long-term high-voltage operation, local insulation breakdown may occur, leading to damage to the electrical equipment. According to statistics, more than half of high-voltage electrical equipment failures are caused by gradual aging and damage to the insulation material due to partial discharge. However, this partial discharge defect is currently difficult to detect during operation.

[0003] The existing technical solutions have the following problems:

[0004] 1. Most current testing methods involve performing a withstand voltage test before applying the equipment. However, this method makes it difficult to determine partial discharge breakdown in advance. Premature testing can damage the insulation layer of electrical equipment, further increasing the equipment failure rate. Furthermore, existing partial discharge testing and data are subject to significant errors due to the influence of the testing environment.

[0005] 2. In existing detection methods, different sensors produce diverse detection results, making it difficult to conduct comprehensive analysis of multi-source data. Different forms of detection results are scattered, making it difficult to establish correlation analysis. Sensor result analysis is often relatively independent, and detection result analysis often requires manual judgment, resulting in a low degree of automation. Summary of the Invention

[0006] In order to solve the above technical problems, the present invention provides a method for centralized partial discharge data and multi-mode detection and diagnosis equipment based on the Internet of Things to solve the problems in the existing technology of difficult partial discharge detection and low comprehensiveness of detection result analysis.

[0007] A method for centralized partial discharge data and multi-mode detection and diagnosis equipment based on the Internet of Things, comprising the following steps:

[0008] Step 1: Collect multimodal detection data collected by multi-source sensors deployed in the equipment during the equipment operation process;

[0009] Step 2: Perform preliminary data cleaning on the local edge computing server of the equipment to eliminate errors generated during data collection.

[0010] Step 3: The cleaned data is transmitted to the central analysis platform, and the detection data weight is dynamically assigned according to the data type and environmental parameters;

[0011] Step 4: By establishing a diagnostic model and analyzing the data cleaned in step 2, determine whether there is partial discharge and the discharge intensity.

[0012] Preferably, the multimodal detection data in step 1 includes electromagnetic noise intensity, ultra-high frequency electromagnetic pulse signal and metal shell instantaneous voltage signal to ground, and the signal includes ultra-high frequency electromagnetic pulse signal frequency, ultra-high frequency electromagnetic pulse time domain voltage signal, data acquisition time and data change rate.

[0013] Preferably, the data cleaning in step 2 is to process the UHF electromagnetic pulse signal and the instantaneous voltage signal of the metal shell of the equipment to the ground collected in step 1;

[0014] First, remove the missing values ​​of the collected signal, determine the signal outliers according to the 3σ principle, discard the data signal outliers, and use the average value of the adjacent data of the missing value to supplement the missing value;

[0015] Secondly, the NTP protocol is used to perform time synchronization calibration on multi-source data to ensure that the timestamps of each data collection source are aligned with high precision, thereby effectively reducing data synchronization errors and further improving the consistency of data collection time.

[0016] Preferably, in step 3, data transmission first packages the data and transmits the data packets to the central analysis platform, then performs weight analysis on the data source and calculates the corresponding weights of the corresponding multimodal data;

[0017] The data packet processing and transmission realizes efficient transmission in different network environments by dynamically selecting the transmission protocol, compressing and encapsulating the data, and embedding metadata on the received original data, thereby ensuring the integrity and real-time performance of the data.

[0018] The weight analysis is used to receive structured data after data packet processing and transmission, calculate the multimodal data fusion weight based on device type, environmental parameters and signal-to-noise ratio, and support accurate analysis of the diagnostic model.

[0019] Preferably, the data packet processing and transmission in step 3 detects network bandwidth and latency in real time, dynamically selects MQTT or CoAP protocol for data packet transmission, and uses LZ4 lossless compression algorithm to compress large-volume ultrasonic waveform data. At the same time, the device coding and sensor health are embedded in the data packet to achieve a balance between transmission efficiency and data integrity, ensuring that the equipment maintains end-to-end low-latency data packet transmission even when operating in a complex network environment.

[0020] Preferably, the weight analysis in step 3 determines the basic weights of the multi-source sensors based on the equipment type and performs a comprehensive calculation in combination with the detection environment data. The calculation process is as follows:

[0021] First, a sensor base weight is preset based on the device type to reflect the inherent reliability of different sensors in a specific device. The sensor base weight is determined based on the device code in the data packet. The sensor base weight determination rule is set based on the partial discharge detection standard.

[0022] Secondly, the wavelet threshold denoising method is used to reduce the noise of the UHF electromagnetic pulse signal. The noise reduction process is as follows:

[0023] Assume that the original signal of the ultra-high frequency electromagnetic pulse is x (t) , the effective signal is s (t) , the noise signal is n (t) ,x (t) =s (t) +n (t) , t=1,2,...,N;

[0024] Where t represents the signal acquisition time point, and N is the total number of signal acquisition time points;

[0025] By applying the wavelet threshold denoising method, from x (t) Restore s (t) ;

[0026] First, Daubechies4 is selected as the wavelet basis function of wavelet decomposition. (t) Perform L-layer decomposition to obtain the wavelet coefficients of each layer and the approximate coefficient α low Indicates low-frequency components, which are effective signals when the device is running, detail coefficient Represents high-frequency components, including noise and instantaneous impact data of equipment operation;

[0027]

[0028] Among them, L is the number of wave decomposition layers, β is the number of detail coefficients, α low and is calculated by wavelet transform decomposition. Represents the high-frequency component detail coefficient, and the noise energy is distributed in the highest frequency detail coefficient Its standard deviation σ nois The calculation formula is:

[0029]

[0030] Among them, 0.6745 is the conversion factor between the median and standard deviation in Gaussian distribution, Indicates the absolute value of the detail coefficient The median of , N leng is the data length;

[0031] Secondly, for each layer detail coefficient Set the threshold λ L ,

[0032] Detail coefficient The soft threshold function is used point by point to suppress noise and retain valid signals by setting detail coefficients above the threshold to zero, thus obtaining the processed coefficients. Reconstruct the signal by inverse wavelet transform:

[0033]

[0034] in, represents the effective signal after processing, IDWT represents inverse discrete wavelet transform;

[0035] Next, calculate the weight w corresponding to the UHF electromagnetic pulse signal f , the calculation process is as follows:

[0036] First, calculate the time domain energy density G(f) of the UHF electromagnetic pulse signal:

[0037]

[0038] Wherein, T is the UHF electromagnetic pulse signal acquisition time period, t1 is the start time point of the acquisition time period, t2 is the end time point of the acquisition time period, g(t) is the UHF electromagnetic pulse time domain voltage signal, which is acquired by the UHF electromagnetic pulse;

[0039] Secondly, calculate the UHF electromagnetic pulse signal signal-to-noise ratio (SNR):

[0040]

[0041] in, is the time domain energy density of the original signal of the UHF electromagnetic pulse, is the time domain energy density of the effective signal of the UHF electromagnetic pulse;

[0042] Then, calculate the weight w corresponding to the UHF electromagnetic pulse signal f , the calculation formula is as follows:

[0043]

[0044] in, is the basic weight of the UHF electromagnetic pulse sensor, β is the UHF electromagnetic pulse weight adjustment coefficient, which is determined based on the UHF electromagnetic pulse signal signal-to-noise ratio (SNR). The UHF electromagnetic pulse weight adjustment coefficient determination rule is based on the partial discharge detection standard setting. v It is the basic weight of the instantaneous voltage signal of the metal casing of the equipment to the ground.

[0045] Preferably, the multi-source data analysis in step 4 is obtained by fusing and calculating the received ultra-high frequency electromagnetic pulse signal and the metal shell instantaneous voltage signal to the ground, and the calculation process is as follows:

[0046] First, the signal is normalized, and the calculation formula is as follows:

[0047]

[0048] Among them, H t and are the original sequence value and the normalized sequence value at time t, respectively. t is the time point of original data collection. is the mean value of the original data, ξ is the standard deviation of the original data;

[0049] Next, the comprehensive partial discharge index is calculated based on the normalized values ​​of the UHF electromagnetic pulse signal and the metal shell instantaneous voltage signal to ground and the corresponding weights. The calculation process is as follows:

[0050]

[0051] Where SSE is the comprehensive index of partial discharge, i is the sensor number, θ is the total number of sensors, W i is the sensor weight corresponding to sensor i, H i Detects data feature values ​​for sensors.

[0052] Preferably, the partial discharge judgment and discharge intensity judgment process in step 4 is as follows:

[0053] First, determine whether the characteristic value of the UHF electromagnetic pulse signal exceeds the threshold described Determined by the partial discharge detection standard, if the characteristic value exceeds It can be determined that there is partial discharge in the equipment, and the partial discharge intensity in the equipment is determined by comparing the partial discharge comprehensive index with the partial discharge detection standard. If the characteristic values ​​of the ultra-high frequency electromagnetic pulse signal in the equipment do not exceed It can be determined that there is no partial discharge in the equipment.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] 1. The present invention collects multimodal detection data from the equipment, cleans and denoises the multi-source data, and reduces the impact of equipment operation noise on the detection results. At the same time, it sets multi-source data weights to enhance the comprehensiveness of partial discharge detection.

[0056] 2. The present invention realizes the centralized management of partial discharge data through preliminary data cleaning on the edge computing platform and end-to-end network information transmission. It comprehensively analyzes multi-source data in the equipment based on the central analysis platform, effectively improving the accuracy of partial discharge detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a step diagram of the working method of the present invention.

[0058] Figure 2 It is a calculation flow chart of the present invention. DETAILED DESCRIPTION

[0059] The following embodiments of the present invention are described in further detail with reference to the accompanying drawings and examples. The following examples are used to illustrate the present invention but are not intended to limit the scope of the present invention.

[0060] The present invention provides a method for centralized partial discharge data and multi-mode detection and diagnosis equipment based on the Internet of Things, comprising the following steps:

[0061] Step 1: Collect multimodal detection data collected by multi-source sensors deployed in the equipment during the equipment operation process;

[0062] Step 2: Perform preliminary data cleaning on the local edge computing server of the equipment to eliminate errors generated during data collection.

[0063] Step 3: The cleaned data is transmitted to the central analysis platform, and the detection data weight is dynamically assigned according to the data type and environmental parameters;

[0064] Step 4: By establishing a diagnostic model and analyzing the data cleaned in step 2, determine whether there is partial discharge and the discharge intensity.

[0065] Example 1:

[0066] like Figure 1 - Figure 2 As shown, in this embodiment, a company needs to detect the partial discharge of electrical equipment in operation. By applying this method to detect the partial discharge of equipment in operation, the following steps are included:

[0067] Step 1: Collect multimodal detection data collected by multi-source sensors deployed in the equipment during the operation of the equipment, including electromagnetic noise intensity, ultra-high frequency electromagnetic pulse signals, and metal shell instantaneous voltage signals to the ground. The signals include ultra-high frequency electromagnetic pulse signal frequency, ultra-high frequency electromagnetic pulse time domain voltage signal, data acquisition time, and data change rate.

[0068] Step 2: Perform preliminary data cleaning on the local edge computing server of the equipment to eliminate errors generated during data collection;

[0069] The data cleaning is to process the UHF electromagnetic pulse signal and the instantaneous voltage signal of the metal shell of the equipment to the ground collected in step 1;

[0070] First, remove the missing values ​​of the collected signal, determine the signal outliers according to the 3σ principle, discard the data signal outliers, and use the average value of the adjacent data of the missing value to supplement the missing value;

[0071] Secondly, the NTP protocol is used to perform time synchronization calibration on multi-source data to ensure that the timestamps of each data collection source are aligned with high precision, thereby effectively reducing data synchronization errors and further improving the consistency of data collection time.

[0072] The above data cleaning and data integration process can effectively remove erroneous values ​​in the detection data and data detection errors caused by data asynchrony during the data detection process.

[0073] Step 3: The processed data is transmitted to the central analysis platform, and the detection data weight is dynamically assigned according to the data type and environmental parameters;

[0074] First, the data is packaged and transmitted to the central analysis platform. Then, the data source is weighted and the corresponding weights of the multimodal data are calculated.

[0075] The data packet processing and transmission realizes efficient transmission in different network environments by dynamically selecting the transmission protocol, compressing and encapsulating the data, and embedding metadata on the received original data, thereby ensuring the integrity and real-time performance of the data.

[0076] The data packet processing and transmission detects network bandwidth and latency in real time, dynamically selects MQTT or CoAP protocols for data packet transmission, and uses the LZ4 lossless compression algorithm to compress large-volume ultrasonic waveform data. At the same time, the device code and sensor health are embedded in the data packet to achieve a balance between transmission efficiency and data integrity, ensuring that the equipment maintains end-to-end low-latency data packet transmission even when operating in a complex network environment.

[0077] The weight analysis is used to receive structured data after data packet processing and transmission, calculate the multimodal data fusion weight based on device type, environmental parameters and signal-to-noise ratio, and support accurate analysis of the diagnostic model.

[0078] The weight analysis determines the basic weights of multi-source sensors based on the equipment type and performs a comprehensive calculation based on the detection environment data. The calculation process is as follows:

[0079] First, a sensor base weight is preset based on the device type to reflect the inherent reliability of different sensors in a specific device. The sensor base weight is determined based on the device code in the data packet. The sensor base weight determination rule is set based on the partial discharge detection standard.

[0080] Secondly, the wavelet threshold denoising method is used to reduce the noise of the UHF electromagnetic pulse signal. The noise reduction process is as follows:

[0081] Assume that the original signal of the ultra-high frequency electromagnetic pulse is x (t) , the effective signal is s (t) , the noise signal is n (t) ,x (t) =s (t) +n (t) , t=1,2,...,N;

[0082] Where t represents the signal acquisition time point, and N is the total number of signal acquisition time points;

[0083] By applying the wavelet threshold denoising method, from x (t) Restore s (t) ;

[0084] First, Daubechies4 is selected as the wavelet basis function of wavelet decomposition. (t) Perform L-layer decomposition to obtain the wavelet coefficients of each layer and the approximate coefficient α low Indicates low-frequency components, which are effective signals when the device is running, detail coefficient Represents high-frequency components, including noise and instantaneous impact data of equipment operation;

[0085]

[0086] Among them, L is the number of wave decomposition layers, β is the number of detail coefficients, α low and is calculated by wavelet transform decomposition. Represents the high-frequency component detail coefficient, and the noise energy is distributed in the highest frequency detail coefficient Its standard deviation σ nois The calculation formula is:

[0087]

[0088] Among them, 0.6745 is the conversion factor between the median and standard deviation in Gaussian distribution, Indicates the absolute value of the detail coefficient The median of , N leng is the data length;

[0089] Secondly, for each layer detail coefficient Set the threshold λ L ,

[0090] Detail coefficient The soft threshold function is used point by point to suppress noise and retain valid signals by setting detail coefficients above the threshold to zero, thus obtaining the processed coefficients. Reconstruct the signal by inverse wavelet transform:

[0091]

[0092] in, represents the effective signal after processing, IDWT represents inverse discrete wavelet transform;

[0093] Next, calculate the weight w corresponding to the UHF electromagnetic pulse signal f , the calculation process is as follows:

[0094] First, calculate the time domain energy density G(f) of the UHF electromagnetic pulse signal:

[0095]

[0096] Wherein, T is the UHF electromagnetic pulse signal acquisition time period, t1 is the start time point of the acquisition time period, t2 is the end time point of the acquisition time period, g(t) is the UHF electromagnetic pulse time domain voltage signal, which is acquired by the UHF electromagnetic pulse;

[0097] Secondly, calculate the UHF electromagnetic pulse signal signal-to-noise ratio (SNR):

[0098]

[0099] Among them, G(x (t) ) is the time domain energy density of the original signal of the UHF electromagnetic pulse, is the time domain energy density of the effective signal of the UHF electromagnetic pulse;

[0100] Then calculate the weight w corresponding to the ultra-high frequency electromagnetic pulse signal f , the calculation formula is as follows:

[0101]

[0102] in, is the basic weight of the UHF electromagnetic pulse sensor, β is the UHF electromagnetic pulse weight adjustment coefficient, which is determined based on the UHF electromagnetic pulse signal signal-to-noise ratio (SNR). The UHF electromagnetic pulse weight adjustment coefficient determination rule is based on the partial discharge detection standard setting. v It is the basic weight of the instantaneous voltage signal of the metal casing of the equipment to the ground.

[0103] The above-mentioned weight calculation method can effectively remove data errors caused by noise in data detection, and effectively calculate the signal-to-noise ratio of the ultra-high frequency electromagnetic pulse signal through wavelet noise reduction, thereby determining the degree to which the ultra-high frequency electromagnetic pulse is affected by the environment, thereby quantifying the weight of the ultra-high frequency electromagnetic pulse in partial discharge. By comprehensively calculating the partial discharge characteristic value of the detection device in the equipment, the comprehensiveness of partial discharge detection is effectively improved, and the detection precision and accuracy are improved.

[0104] Step 4: By establishing a diagnostic model and analyzing the data cleaned in step 2, it is determined whether there is partial discharge and the discharge intensity. The analysis is obtained by fusing the received UHF electromagnetic pulse signal and the metal shell instantaneous voltage signal to the ground. The calculation process is as follows:

[0105] First, the signal is normalized, and the calculation formula is as follows:

[0106]

[0107] Among them, H t and are the original sequence value and the normalized sequence value at time t, respectively. t is the time point of original data collection. is the mean value of the original data, ξ is the standard deviation of the original data;

[0108] Next, the comprehensive partial discharge index is calculated based on the normalized values ​​of the UHF electromagnetic pulse signal and the metal shell instantaneous voltage signal to ground and the corresponding weights. The calculation process is as follows:

[0109]

[0110] Where SSE is the comprehensive index of partial discharge, i is the sensor number, θ is the total number of sensors, W i is the sensor weight corresponding to sensor i, H i Detects data feature values ​​for sensors.

[0111] Furthermore, the partial discharge judgment and discharge intensity judgment process are as follows:

[0112] First, determine whether the characteristic value of the UHF electromagnetic pulse signal exceeds the threshold described Determined by the partial discharge detection standard, if the characteristic value exceeds It can be determined that there is partial discharge in the equipment, and the partial discharge intensity in the equipment is determined by comparing the partial discharge comprehensive index with the partial discharge detection standard. If the characteristic values ​​of the ultra-high frequency electromagnetic pulse signal in the equipment do not exceed It can be determined that there is no partial discharge in the equipment.

[0113] The embodiments of the present invention are provided for the purpose of illustration and description. Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations of the present invention. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A method for centralized partial discharge data and multi-mode detection and diagnosis equipment based on the Internet of Things, characterized in that: The following steps are involved: Step 1: Collect multimodal detection data collected by multi-source sensors deployed in the equipment during the equipment operation process; Step 2: Perform preliminary data cleaning on the local edge computing server of the equipment to eliminate errors generated during data collection. Step 3: The cleaned data is transmitted to the central analysis platform, and the detection data weight is dynamically assigned according to the data type and environmental parameters; Step 4: By establishing a diagnostic model and analyzing the data cleaned in step 2, determine whether there is partial discharge and the discharge intensity.

2. The method of claim 1, wherein: The multimodal detection data in step 1 includes electromagnetic noise intensity, ultra-high frequency electromagnetic pulse signal and metal shell instantaneous voltage signal to ground, and the signal includes ultra-high frequency electromagnetic pulse signal frequency, ultra-high frequency electromagnetic pulse time domain voltage signal, data acquisition time and data change rate.

3. The method of intensive partial discharge data and multi-mode detection and diagnosis equipment based on the Internet of Things as claimed in claim 2, characterized in that: The data cleaning step 2 is to process the UHF electromagnetic pulse signal and the instantaneous voltage signal of the metal shell of the equipment to the ground collected in step 1; First, remove the missing values ​​of the collected signal, determine the signal outliers according to the 3σ principle, discard the data signal outliers, and use the average value of the adjacent data of the missing value to supplement the missing value; Secondly, the NTP protocol is used to perform time synchronization calibration on multi-source data to ensure that the timestamps of each data collection source are aligned with high precision, thereby effectively reducing data synchronization errors and further improving the consistency of data collection time.

4. The method of claim 1, wherein: In step 3, data transmission first packages the data and transmits the data packets to the central analysis platform, then performs weight analysis on the data source and calculates the corresponding weights of the corresponding multimodal data; The data packet processing and transmission realizes efficient transmission in different network environments by dynamically selecting the transmission protocol, compressing and encapsulating the data, and embedding metadata on the received original data, thereby ensuring the integrity and real-time performance of the data. The weight analysis is used to receive structured data after data packet processing and transmission, calculate the multimodal data fusion weight based on device type, environmental parameters and signal-to-noise ratio, and support accurate analysis of the diagnostic model.

5. The method of intensive partial discharge data and multi-mode detection and diagnosis equipment based on the Internet of Things as claimed in claim 4, characterized in that: The data packet processing and transmission in step 3 detects network bandwidth and latency in real time, dynamically selects MQTT or CoAP protocol for data packet transmission, and uses LZ4 lossless compression algorithm to compress large-volume ultrasonic waveform data. At the same time, the device code and sensor health are embedded in the data packet to achieve a balance between transmission efficiency and data integrity, ensuring that the equipment maintains end-to-end low-latency data packet transmission even when operating in a complex network environment.

6. The method of intensive partial discharge data and multi-mode detection and diagnosis equipment based on the Internet of Things as claimed in claim 4, characterized in that: The weight analysis in step 3 determines the basic weights of the multi-source sensors based on the equipment type and performs a comprehensive calculation in combination with the detection environment data. The calculation process is as follows: First, a sensor base weight is preset based on the device type to reflect the inherent reliability of different sensors in a specific device. The sensor base weight is determined based on the device code in the data packet. The sensor base weight determination rule is set based on the partial discharge detection standard. Secondly, the wavelet threshold denoising method is used to reduce the noise of the UHF electromagnetic pulse signal. The noise reduction process is as follows: Assume that the original signal of the ultra-high frequency electromagnetic pulse is x (t) , the effective signal is s (t) , the noise signal is n (t) ,x (t) =s (t) +n (t) , t=1,2,...,N; Where t represents the signal acquisition time point, and N is the total number of signal acquisition time points; By applying the wavelet threshold denoising method, from x (t) Restore s (t) ; First, Daubechies4 is selected as the wavelet basis function of wavelet decomposition. (t) Perform L-layer decomposition to obtain the wavelet coefficients of each layer and the approximate coefficient α low Indicates low-frequency components, which are effective signals when the device is running, detail coefficient Represents high-frequency components, including noise and instantaneous impact data of equipment operation; Among them, L is the number of wave decomposition layers, β is the number of detail coefficients, α low and is calculated by wavelet transform decomposition. Represents the high-frequency component detail coefficient, and the noise energy is distributed in the highest frequency detail coefficient Its standard deviation σ nois The calculation formula is: Among them, 0.6745 is the conversion factor between the median and standard deviation in Gaussian distribution, Indicates the absolute value of the detail coefficient The median of , N leng is the data length; Secondly, for each layer detail coefficient Set the threshold λ L , Detail coefficient The soft threshold function is used point by point to suppress noise and retain valid signals by setting detail coefficients above the threshold to zero, thus obtaining the processed coefficients. Reconstruct the signal by inverse wavelet transform: in, represents the effective signal after processing, IDWT represents inverse discrete wavelet transform; Next, calculate the weight w corresponding to the UHF electromagnetic pulse signal f , the calculation process is as follows: First, calculate the time domain energy density G(f) of the UHF electromagnetic pulse signal: Wherein, T is the UHF electromagnetic pulse signal acquisition time period, t1 is the start time point of the acquisition time period, t2 is the end time point of the acquisition time period, g(t) is the UHF electromagnetic pulse time domain voltage signal, which is acquired by the UHF electromagnetic pulse; Secondly, calculate the UHF electromagnetic pulse signal signal-to-noise ratio (SNR): Among them, G(x (t) ) is the time domain energy density of the original signal of the UHF electromagnetic pulse, is the time domain energy density of the effective signal of the UHF electromagnetic pulse; Then, calculate the weight w corresponding to the UHF electromagnetic pulse signal f , the calculation formula is as follows: in, is the basic weight of the UHF electromagnetic pulse sensor, β is the UHF electromagnetic pulse weight adjustment coefficient, which is determined based on the UHF electromagnetic pulse signal signal-to-noise ratio (SNR). The UHF electromagnetic pulse weight adjustment coefficient determination rule is based on the partial discharge detection standard setting. v It is the basic weight of the instantaneous voltage signal of the metal casing of the equipment to the ground.

7. The method of intensive partial discharge data and multi-mode detection and diagnosis equipment based on the Internet of Things as claimed in claim 6, characterized in that: The multi-source data analysis in step 4 is obtained by fusing and calculating the received UHF electromagnetic pulse signal and the metal shell instantaneous voltage signal to the ground. The calculation process is as follows: First, the signal is normalized, and the calculation formula is as follows: Among them, H t and are the original sequence value and the normalized sequence value at time t, respectively. t is the time point of original data collection. is the mean value of the original data, ξ is the standard deviation of the original data; Next, the comprehensive partial discharge index is calculated based on the normalized values ​​of the UHF electromagnetic pulse signal and the metal shell instantaneous voltage signal to ground and the corresponding weights. The calculation process is as follows: Where SSE is the comprehensive index of partial discharge, i is the sensor number, θ is the total number of sensors, W i is the sensor weight corresponding to sensor i, H i Detects data feature values ​​for sensors.

8. The method of intensive partial discharge data and multi-mode detection and diagnosis equipment based on the Internet of Things as claimed in claim 7, characterized in that: The process of partial discharge judgment and discharge intensity judgment in step 4 is as follows: First, determine whether the characteristic value of the UHF electromagnetic pulse signal exceeds the threshold described Determined by the partial discharge detection standard, if the characteristic value exceeds It can be determined that there is partial discharge in the equipment, and the partial discharge intensity in the equipment is determined by comparing the partial discharge comprehensive index with the partial discharge detection standard. If the characteristic values ​​of the ultra-high frequency electromagnetic pulse signal in the equipment do not exceed It can be determined that there is no partial discharge in the equipment.

Images