Dry-type transformer abnormal working condition monitoring method based on multi-source information fusion technology

Abstract

The invention provides a dry-type transformer abnormal working condition monitoring method based on a multi-source information fusion technology, which comprises the following steps: firstly, determining various indexes capable of intuitively reflecting the operation condition of a dry-type transformer, and installing a multi-feature acquisition sensor of the dry-type transformer according to the indexes; then determining the operation health state grade of the dry-type transformer; preprocessing multi-feature information collected by a sensor by adopting a technical means, and obtaining the weight of each feature vector by utilizing an analytic hierarchy process; and finally, training and learning the training set feature information acquired by the dry-type transformer by using a set pair analysis method, mastering the rule of internal change, calculating the correlation degree of different evaluation grades, verifying the acquired test set data, obtaining the state evaluation result of the dry-type transformer, performing timeliness processing on different evaluation results, and obtaining the state evaluation result of the dry-type transformer. The abnormal working condition of the dry-type transformer is monitored.

Description

Technical Field

[0001] The invention relates to the technical field of transformer monitoring, and in particular to a method for monitoring abnormal working conditions of dry-type transformers based on multi-source information fusion technology. Background Art

[0002] As one of the important components in the power system network, the dry-type transformer has the functions of voltage conversion, power transmission, power control, and safety isolation. Its normal and stable operation plays an important role in the safety of the power grid. Common types of dry-type transformer failures include: abnormal temperature rise, abnormal internal noise, sudden change in insulation value, and odor. The traditional method of judging its operating conditions requires a lot of manpower to analyze the monitored U, I, and temperature data, or conduct on-site inspections and observations to judge its operating conditions. Such methods cannot judge the operating conditions of dry-type transformers in real time and comprehensively, and there are safety hazards. Therefore, it is of great significance to ensure the healthy operation of dry-type transformers and to promptly detect and eliminate their abnormal operating conditions.

[0003] At present, the methods for judging the operating condition of dry-type transformers include: installing voltage transformers, current transformers, and temperature sensors to monitor the corresponding voltage, current, and temperature change patterns. If the monitored values ​​change significantly or exceed the set values, an alarm is issued; or on-site inspections are conducted to judge the operating conditions.

[0004] In the prior art, the method for judging the operating condition of the dry-type transformer is too simple and cannot judge the operating condition of the dry-type transformer in real time and comprehensively, which has potential safety hazards and is not conducive to the safe and stable operation of the equipment. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a dry-type transformer abnormal condition monitoring method based on multi-source information fusion technology, collect all-round real-time operation data information of the dry-type transformer, use Mel frequency cepstrum coefficients, deep learning network models, hierarchical analysis method and other technologies to perform voiceprint data and signal data fusion analysis and deep mining, grasp the change trend of each characteristic item parameter sequence of the dry-type transformer in normal operation, realize rapid detection of abnormal state of the dry-type transformer, comprehensively, timely and accurately grasp the health status of key components, and provide strong support for equipment fault handling and safe operation.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for monitoring abnormal working conditions of dry-type transformers based on multi-source information fusion technology, comprising the following steps: Step 1: Determine the key parameter indicators of the dry-type transformer under normal operating conditions; Step 2: According to the key parameter indicators of the dry-type transformer under normal operating conditions determined in step 1, corresponding sensors are installed inside the dry-type transformer to extract the characteristics of the key parameter indicators; Step 3, determine the operating health status level of the dry-type transformer; Step 4: Under normal operating conditions of the dry-type transformer, the data information collected by the voiceprint sensor and the vibration sensor in step 2 is processed using the Mel frequency cepstrum coefficient technology to obtain the corresponding feature vector, and the training set and the test set are divided according to 8:2; Step 5: Under the normal operating conditions of the dry-type transformer, the computer monitoring system host computer is used to perform feature processing on the data information collected by the dry-type transformer operating voltage U, current I, temperature sensor, temperature sensor, and gas monitoring sensor to obtain the corresponding feature vector, and divide the training set and the test set into 8:2; Step 6: Determine the weight of each feature vector obtained in step 4 and step 5 by using the analytic hierarchy process; Step 7: Take the feature vectors obtained in steps 4 and 5 as input, use set pair analysis to train and learn the training set feature information collected from the dry-type transformer, master the rules of its internal changes, calculate the correlation between different evaluation levels, and finally verify the collected test set data, and obtain the dry-type transformer status evaluation results, perform timeliness processing on different evaluation results, and realize the monitoring of abnormal working conditions of dry-type transformers.

[0007] Preferably, in step 1, determining the key parameter indicators of the dry-type transformer under normal operating conditions specifically includes: the voltage U and current I of the dry-type transformer during normal operation, the temperature variation range of each component of the dry-type transformer, the internal humidity variation range of the dry-type transformer, the operating sound print range of each component of the dry-type transformer, and the type and content range of the gas generated during normal operation of the dry-type transformer.

[0008] Preferably, in step 2, the installed sensors include: a cooling fan voiceprint sensor, a transformer group voiceprint sensor, a base vibration sensor, a support frame vibration sensor, a transformer group temperature sensor, a transformer group humidity sensor, and a transformer group gas monitoring sensor.

[0009] Preferably, in step 3, the operating health status level of the dry-type transformer is determined, and is represented by a membership function combining a semi-trapezoidal and a semi-ridged form, which is defined as shown in the following formula: ; In the formula g 1( x m ), g 2( xm ), g 3( x m ), g 4( x m ) represent the four levels corresponding to dry-type transformers: normal, attention, abnormal, and severe. Early warning prompts are given for attention levels, and fault alarms are given for abnormal and severe levels. g i The four intervals are 0~0.3, 0.1~0.6, 0.4~0.9, 0.7~1, b i The five cut-off values ​​are 0.1, 0.3, 0.4, 0.6, 0.7, and 0.9.

[0010] Preferably, in step 4, under the normal operating condition of the dry-type transformer, the Mel frequency cepstrum coefficient technology is used to process the data information collected by the voiceprint sensor and the vibration sensor, and the specific process is as follows: 1) The original data information collected by the voiceprint sensor and the vibration sensor is processed by pre-emphasis, framing and windowing in sequence. The pre-emphasis processing reduces the impact of the environment and the transmission system itself on the signal itself. The framing operation divides the long speech into several segments. The windowing operation makes a smooth transition of the speech signal at the truncated position after framing. 2) After pre-emphasis, framing and windowing, the data information collected by the sensor is subjected to Fast Fourier Transform (FFT) to convert the transmission signal into energy distribution for observation. Different energy distributions represent different speech characteristics. 3) Further modulus processing, through n The Mel filter bank is transformed into the Mel frequency domain, and finally logarithmically transformed to obtain a set of n The coefficient c( i ) constitutes the MFCC feature vector of the frame signal. A set of MFCC feature vectors is taken for each sensor position, and its change formula is as follows: ; Where: c ( l ) is the l dimensional cepstral coefficient value; s ( m ) is the first in the Mel filter bank m The output of a triangular filter; N for s ( m )’s length; p is the serial number of the discrete cosine transform component; Mis the number of triangular filters in the Mel filter bank; L is the order of MFCC.

[0011] Preferably, in step 5, under the normal operating condition of the dry-type transformer, the computer monitoring system host computer is used to perform feature processing on the data information collected by the dry-type transformer operating voltage U, current I, temperature sensor, temperature sensor, and gas monitoring sensor to obtain the corresponding feature vector, and the specific steps are: 1) Accurately extract the data characteristic information collected by the dry-type transformer operating voltage U, current I, temperature sensor, temperature sensor, and gas monitoring sensor from the background database of the computer monitoring system; 2) Use the deep learning model to clean the extracted feature data, analyze and find abnormal point data, reconstruct and repair the data, and obtain the corresponding feature vector group.

[0012] Preferably, the specific steps of step 6 are as follows: 1) A hierarchical structure constructed based on the target problem, with the first level being the overall target, the second level being the criteria and sub-criteria targets, and the third level being the decision options; 2) Obtaining relative importance from a standard matrix; 3) Aggregate the weights of the entire hierarchy to determine the combined weight.

[0013] Preferably, in step 7, the set pair analysis method is specifically as follows: The set pair analysis method quantitatively describes the correlation of a pair of sets from three perspectives: similarity, opposite, and difference. It has two important indicators: set pair and correlation, which are specifically expressed as follows: ; ; Where: n, l, v, u They represent the total number of attributes, the number of identical attributes, the number of opposites, and the number of different attributes of the set pair, respectively. p It represents the difference coefficient with a value range of [-1, 1]. q Indicates the opposite coefficient, the value is -1; The correlation between the two sets is further expanded from the ternary correlation to the s-element correlation, s ≥ 3, as shown in the following formula: ; e s For different levels of difference, p s are coefficients at different levels; The health status of the dry-type transformer involved in the present invention is divided into four levels: normal, concerned, abnormal, and severe. The corresponding four-element correlation degree is shown in the following formula: ; In the formula y i , t i Status indicator x i The corresponding weight size and the degree of membership of the normal state, e i,1 , e i,2 Status indicator x i The membership degrees of abnormal state and attention state respectively, p i,1 , p i,2 Status indicator x i The coefficients belonging to abnormal state and attention state respectively, m i Status indicator x i The degree of membership in a serious state; The correlation between the status indicator and the operating status level is shown in the following formula: .

[0014] The present invention proposes a dry-type transformer abnormal condition monitoring method based on multi-source information fusion technology, which collects all-round real-time operation data information of the dry-type transformer, and uses Mel frequency cepstrum coefficients, deep learning network models, hierarchical analysis method and other technologies to perform voiceprint data and signal data fusion analysis and deep mining, so as to grasp the change trend of each characteristic parameter sequence of the dry-type transformer in normal operation, realize rapid detection of abnormal state of the dry-type transformer, comprehensively, timely and accurately grasp the health status of key performance, provide a theoretical basis for fault handling, and provide strong support for ensuring the safe operation of equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 is a flow chart of the method of the present invention; Figure 2 It is a membership function distribution diagram of the present invention; Figure 3 This is a diagram of the data preprocessing process using the Mel-frequency cepstral coefficient technology of the present invention. DETAILED DESCRIPTION

[0016] like Figure 1As shown, a method for monitoring abnormal working conditions of dry-type transformers based on multi-source information fusion technology includes the following steps: Step 1: Determine the key parameter indicators of the dry-type transformer under normal operating conditions; Step 2: According to the key parameter indicators of the dry-type transformer under normal operating conditions determined in step 1, corresponding sensors are installed inside the dry-type transformer to extract the characteristics of the key parameter indicators; Step 3, determine the operating health status level of the dry-type transformer; Step 4: Under normal operating conditions of the dry-type transformer, the data information collected by the voiceprint sensor and the vibration sensor in step 2 is processed using the Mel frequency cepstrum coefficient technology to obtain the corresponding feature vector, and the training set and the test set are divided according to 8:2; Step 5: Under the normal operating conditions of the dry-type transformer, the computer monitoring system host computer is used to perform feature processing on the data information collected by the dry-type transformer operating voltage U, current I, temperature sensor, temperature sensor, and gas monitoring sensor to obtain the corresponding feature vector, and divide the training set and the test set into 8:2; Step 6: Determine the weight of each feature vector obtained in step 4 and step 5 by using the analytic hierarchy process; Step 7: Take the feature vectors obtained in steps 4 and 5 as input, use set pair analysis to train and learn the training set feature information collected from the dry-type transformer, master the rules of its internal changes, calculate the correlation between different evaluation levels, and finally verify the collected test set data, and obtain the dry-type transformer status evaluation results, perform timeliness processing on different evaluation results, and realize the monitoring of abnormal working conditions of dry-type transformers.

[0017] Preferably, in step 1, determining the key parameter indicators of the dry-type transformer under normal operating conditions specifically includes: the voltage U and current I of the dry-type transformer during normal operation, the temperature variation range of each component of the dry-type transformer, the internal humidity variation range of the dry-type transformer, the operating sound print range of each component of the dry-type transformer, and the type and content range of the gas generated during normal operation of the dry-type transformer.

[0018] Preferably, in step 2, the installed sensors include: a cooling fan voiceprint sensor, a transformer group voiceprint sensor, a base vibration sensor, a support frame vibration sensor, a transformer group temperature sensor, a transformer group humidity sensor, and a transformer group gas monitoring sensor.

[0019] Preferably, in step 3, the operating health status level of the dry-type transformer is determined, and is represented by a membership function combining a semi-trapezoidal and a semi-ridged form, which is defined as shown in the following formula: ; In the formula g 1( x m ), g 2( x m ), g 3( x m ), g 4( x m ) represent the four levels corresponding to dry-type transformers: normal, attention, abnormal, and severe. Early warning prompts are given for attention levels, and fault alarms are given for abnormal and severe levels. g i The four intervals are 0~0.3, 0.1~0.6, 0.4~0.9, 0.7~1, b i The five cutoff values ​​are 0.1, 0.3, 0.4, 0.6, 0.7, and 0.9. The membership function combining the semi-trapezoidal and semi-ridge shapes is shown in the attached figure. Figure 2 shown.

[0020] Preferably, in step 4, under normal operating conditions of the dry-type transformer, the Mel frequency cepstrum coefficient technology is used to process the data information collected by the voiceprint sensor and the vibration sensor, such as Figure 3 As shown, the specific process is as follows: 1) The original data information collected by the voiceprint sensor and the vibration sensor is processed by pre-emphasis, framing and windowing in sequence. The pre-emphasis processing reduces the impact of the environment and the transmission system itself on the signal itself. The framing operation divides the long speech into several segments. The windowing operation makes a smooth transition of the speech signal at the truncated position after framing. 2) After pre-emphasis, framing and windowing, the data information collected by the sensor is subjected to Fast Fourier Transform (FFT) to convert the transmission signal into energy distribution for observation. Different energy distributions represent different speech characteristics. 3) Further modulus processing, through n The Mel filter bank is transformed into the Mel frequency domain, and finally logarithmically transformed to obtain a set of n The coefficient c( i ) constitutes the MFCC feature vector of the frame signal. A set of MFCC feature vectors is taken for each sensor position, and its change formula is as follows: ; Where: c ( l ) is the ldimensional cepstral coefficient value; s ( m ) is the first in the Mel filter bank m The output of a triangular filter; N for s ( m )’s length; p is the serial number of the discrete cosine transform component; M is the number of triangular filters in the Mel filter bank; L is the order of MFCC.

[0021] Preferably, in step 5, under the normal operating condition of the dry-type transformer, the computer monitoring system host computer is used to perform feature processing on the data information collected by the dry-type transformer operating voltage U, current I, temperature sensor, temperature sensor, and gas monitoring sensor to obtain the corresponding feature vector, and the specific steps are: 1) Accurately extract the data characteristic information collected by the dry-type transformer operating voltage U, current I, temperature sensor, temperature sensor, and gas monitoring sensor from the background database of the computer monitoring system; 2) Use the deep learning model to clean the extracted feature data, analyze and find abnormal point data, reconstruct and repair the data, and obtain the corresponding feature vector group.

[0022] Preferably, the specific steps of step 6 are as follows: 1) A hierarchical structure constructed based on the target problem, with the first level being the overall target, the second level being the criteria and sub-criteria targets, and the third level being the decision options; 2) Obtaining relative importance from a standard matrix; 3) Aggregate the weights of the entire hierarchy to determine the combined weight.

[0023] In step 6, the constructed judgment matrix is ​​shown in Table 1, the consistency ratio of the matrix is ​​represented by CR; the consistency quantitative index of the matrix is ​​represented by CI; CI=( λ max - n ) / ( n -1), λ max is the eigenvalue of the judgment matrix, n is the order of the judgment matrix. RI is the order of the judgment matrix n The specific values ​​are shown in Table 2.

[0024] Table 1 Ratio scaling method Quantization value Factor i is better than factor j 1 Equal importance 3 The influence of factor i is slightly stronger than factor j 5 The influence of i is stronger than factor j 7 The influence of factor i is significantly stronger than factor j 9 The influence of i is definitely stronger than factor j 2、4、6、8 The middle value of the above adjacent judgments Table 2 RI average random consistency value Order n 1 2 3 4 5 6 7 8 9 RI 0 0 0.58 0.90 1.12 1.24 1.32 1.41 1.49 Preferably, in step 7, the set pair analysis method is specifically as follows: The set pair analysis method quantitatively describes the correlation of a pair of sets from three perspectives: similarity, opposite, and difference. It has two important indicators: set pair and correlation, which are specifically expressed as follows: ; ; Where: n, l, v, u They represent the total number of attributes, the number of identical attributes, the number of opposites, and the number of different attributes of the set pair, respectively. p It represents the difference coefficient with a value range of [-1, 1]. q Indicates the opposite coefficient, the value is -1; The correlation between the two sets is further expanded from the ternary correlation to the s-element correlation, s ≥ 3, as shown in the following formula: ; e s For different levels of difference, p s are coefficients at different levels; The health status of the dry-type transformer involved in the present invention is divided into four levels: normal, concerned, abnormal, and severe. The corresponding four-element correlation degree is shown in the following formula: ; In the formula y i , t i Status indicator x i The corresponding weight size and the degree of membership of the normal state, e i,1 , e i,2 Status indicator x i The membership degrees of abnormal state and attention state respectively, p i,1 , p i,2 Status indicator x i The coefficients belonging to abnormal state and attention state respectively, m i Status indicator x i The degree of membership in a serious state; The correlation between the status indicator and the operating status level is shown in the following formula: .

[0025] The present invention proposes a dry-type transformer abnormal condition monitoring method based on multi-source information fusion technology. First, various indicators that can intuitively reflect the operating status of the dry-type transformer are determined, and a multi-feature acquisition sensor of the dry-type transformer is installed according to the indicator; then the operating health status level of the dry-type transformer is determined; then technical means are used to pre-process the multi-feature information collected by the sensor, and the weight of each feature vector is obtained by using the hierarchical analysis method; finally, the training set feature information collected by the dry-type transformer is trained and learned by using the set pair analysis method, the law of its internal changes is mastered, the correlation of different evaluation levels is calculated, the collected test set data is verified, and the dry-type transformer status evaluation result is obtained, and the timeliness processing is performed on different evaluation results to realize the monitoring of the abnormal working condition of the dry-type transformer.

[0026] The present invention monitors various parameter indicators of the dry-type transformer during operation, grasps the change trend and internal law of the parameter sequence of each characteristic item of the dry-type transformer during normal operation, realizes the rapid detection of its abnormal state, and comprehensively, timely and accurately grasps the health status of key performance, providing a theoretical basis for fault handling and providing strong support for ensuring the safe operation of equipment.

[0027] The above embodiments are only preferred technical solutions of the present invention and should not be regarded as limiting the present invention. The protection scope of the present invention shall be the technical solutions recorded in the claims, including equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A method for monitoring abnormal working conditions of dry-type transformers based on multi-source information fusion technology, characterized in that: The following steps are involved: Step 1: Determine the key parameter indicators of the dry-type transformer under normal operating conditions; Step 2: According to the key parameter indicators of the dry-type transformer under normal operating conditions determined in step 1, corresponding sensors are installed inside the dry-type transformer to extract the characteristics of the key parameter indicators; Step 3, determine the operating health status level of the dry-type transformer; Step 4: Under normal operating conditions of the dry-type transformer, the data information collected by the voiceprint sensor and the vibration sensor in step 2 is processed using the Mel frequency cepstrum coefficient technology to obtain the corresponding feature vector, and the training set and the test set are divided according to 8:2; Step 5: Under the normal operating conditions of the dry-type transformer, the computer monitoring system host computer is used to perform feature processing on the data information collected by the dry-type transformer operating voltage U, current I, temperature sensor, temperature sensor, and gas monitoring sensor to obtain the corresponding feature vector, and divide the training set and the test set into 8:2; Step 6: Determine the weight of each feature vector obtained in step 4 and step 5 by using the analytic hierarchy process; Step 7: Take the feature vectors obtained in steps 4 and 5 as input, use set pair analysis to train and learn the training set feature information collected from the dry-type transformer, master the rules of its internal changes, calculate the correlation between different evaluation levels, and finally verify the collected test set data, and obtain the dry-type transformer status evaluation results, perform timeliness processing on different evaluation results, and realize the monitoring of abnormal working conditions of dry-type transformers.

2. According to claim 1, a method for monitoring abnormal working conditions of dry-type transformers based on multi-source information fusion technology is characterized in that: In the step 1, the key parameter indicators of the dry-type transformer under normal operating conditions are determined, including: the voltage U and current I of the dry-type transformer during normal operation, the temperature variation range of each component of the dry-type transformer, the internal humidity variation range of the dry-type transformer, the operating sound pattern variation range of each component of the dry-type transformer, and the type and content range of the gas generated during normal operation of the dry-type transformer.

3. According to claim 1, a method for monitoring abnormal working conditions of dry-type transformers based on multi-source information fusion technology is characterized in that: In step 2, the installed sensors include: a cooling fan voiceprint sensor, a transformer group voiceprint sensor, a base vibration sensor, a support frame vibration sensor, a transformer group temperature sensor, a transformer group humidity sensor, and a transformer group gas monitoring sensor.

4. According to claim 1, a method for monitoring abnormal working conditions of dry-type transformers based on multi-source information fusion technology is characterized in that: In step 3, the operating health status level of the dry-type transformer is determined, and is represented by a membership function combining a semi-trapezoidal and a semi-ridged form, which is defined as shown in the following formula: ； In the formula g 1( x m ), g 2( x m ), g 3( x m ), g 4( x m ) represent the four levels corresponding to dry-type transformers: normal, attention, abnormal, and severe. Early warning prompts are given for attention levels, and fault alarms are given for abnormal and severe levels. g i The four intervals are 0~0.3, 0.1~0.6, 0.4~0.9, 0.7~1, b i The five cut-off values ​​are 0.1, 0.3, 0.4, 0.6, 0.7, and 0.

9.

5. According to claim 1, a method for monitoring abnormal working conditions of dry-type transformers based on multi-source information fusion technology is characterized in that: In step 4, under the normal operating condition of the dry-type transformer, the Mel frequency cepstrum coefficient technology is used to process the data information collected by the voiceprint sensor and the vibration sensor. The specific process is as follows: 1) The original data information collected by the voiceprint sensor and the vibration sensor is processed by pre-emphasis, framing and windowing in sequence. The pre-emphasis processing reduces the impact of the environment and the transmission system itself on the signal itself. The framing operation divides the long speech into several segments. The windowing operation makes a smooth transition of the speech signal at the truncated position after framing. 2) After pre-emphasis, framing and windowing, the data information collected by the sensor is subjected to fast Fourier transform, which converts the transmission signal into energy distribution for observation. Different energy distributions represent different speech characteristics. 3) Further modulus processing, through n The Mel filter banks are transformed into the Mel frequency domain, and finally logarithmically transformed to obtain a set of n The coefficient c( i ) constitutes the MFCC feature vector of the frame signal. A set of MFCC feature vectors is taken for each sensor position, and its change formula is as follows: ； Where: c ( l ) is the l dimensional cepstral coefficient value; s ( m ) is the first in the Mel filter bank m The output of a triangular filter; N for s ( m )’s length; p is the serial number of the discrete cosine transform component; M is the number of triangular filters in the Mel filter bank; L is the order of MFCC.

6. According to claim 1, a method for monitoring abnormal working conditions of dry-type transformers based on multi-source information fusion technology is characterized in that: In step 5, under the normal operating condition of the dry-type transformer, the computer monitoring system host computer is used to perform feature processing on the data information collected by the dry-type transformer operating voltage U, current I, temperature sensor, temperature sensor, and gas monitoring sensor to obtain the corresponding feature vector. The specific steps are: 1) Accurately extract the data characteristic information collected by the dry-type transformer operating voltage U, current I, temperature sensor, temperature sensor, and gas monitoring sensor from the background database of the computer monitoring system; 2) Use the deep learning model to clean the extracted feature data, analyze and find abnormal point data, reconstruct and repair the data, and obtain the corresponding feature vector group.

7. According to claim 1, a method for monitoring abnormal working conditions of dry-type transformers based on multi-source information fusion technology is characterized in that: The specific steps of step 6 are as follows: 1) A hierarchical structure constructed based on the target problem, with the first level being the overall target, the second level being the criteria and sub-criteria targets, and the third level being the decision options; 2) Obtaining relative importance from a standard matrix; 3) Aggregate the weights of the entire hierarchy to determine the combined weight.

8. According to claim 1, a method for monitoring abnormal working conditions of dry-type transformers based on multi-source information fusion technology is characterized in that: In step 7, the set pair analysis method is specifically as follows: The set pair analysis method quantitatively describes the correlation of a pair of sets from three perspectives: similarity, opposite, and difference. It has two important indicators: set pair and correlation, which are specifically expressed as follows: ； ； Where: n, l, v, u They represent the total number of attributes, the number of identical attributes, the number of opposites, and the number of different attributes of the set pair, respectively. p It represents the difference coefficient with a value range of [-1, 1]. q Indicates the opposite coefficient, the value is -1; The correlation between the two sets is further expanded from the ternary correlation to the s-element correlation, s ≥ 3, as shown in the following formula: ； e s For different levels of difference, p s are coefficients at different levels; The health status of the dry-type transformer involved in the present invention is divided into four levels: normal, concerned, abnormal, and severe. The corresponding four-element correlation degree is shown in the following formula: ； In the formula y i , t i Status indicator x i The corresponding weight size and the degree of membership of the normal state, e i,1 , e i,2 Status indicator x i The membership degrees of abnormal state and attention state respectively, p i,1 , p i,2 Status indicator x i The coefficients belonging to abnormal state and attention state respectively, m i Status indicator x i The degree of membership in a serious state; The correlation between the status indicator and the operating status level is shown in the following formula: 。

Images