Transformer fault diagnosis method and system based on multi-parameter fusion

Abstract

The invention relates to a transformer fault diagnosis method and system based on multi-parameter fusion. The method comprises the following steps: collecting multi-parameter data of a local end transformer; encrypting the multi-parameter data by using an encryption algorithm, and transmitting the multi-parameter data to an edge end; the edge end decrypts the encrypted multi-parameter data and carries out data processing on the decrypted multi-parameter data; the edge end fuses the processed multi-parameter data by using a data fusion algorithm based on a D-S evidence theory to obtain multi-parameter fusion data; and constructing a fault diagnosis model based on deep learning to perform fault diagnosis on the multi-parameter fusion data, obtaining a fault diagnosis result, and uploading the fault diagnosis result to the cloud.

Description

Technical Field

[0001] The present application relates to the field of transformer fault diagnosis, and mainly to a transformer fault diagnosis method and system based on multi-parameter fusion. Background Art

[0002] As core equipment in power systems, the reliability of transformer operation directly impacts the safety and stability of power supply. Transformer failures can lead to regional power outages, equipment damage, and even cascading safety incidents. Therefore, real-time, accurate fault diagnosis is crucial for power system operation and maintenance. With the development of smart grid and IoT technologies, transformer monitoring has evolved from single-parameter detection (such as gas concentration) to multi-source data fusion. Current monitoring data covers multiple dimensions, including gas composition, winding temperature, oil level, current and voltage, providing a rich information foundation for fault diagnosis.

[0003] However, the fusion diagnosis of multi-parameter data faces two major challenges: one is the uncertainty and conflict of different parameters (such as fault judgment when the gas concentration is abnormal but the temperature is normal), and the other is the need for real-time processing of massive data (traditional cloud-based diagnosis models have transmission delays and bandwidth pressure).

[0004] Traditional fault diagnosis methods (such as the IEC three-ratio method) rely primarily on gas composition analysis, determining fault types solely through the concentration ratios of gases like H and CH2. This method has significant drawbacks: First, it fails to integrate electrical parameters such as temperature and current, making it prone to missed faults when gas characteristics are atypical (e.g., an early inter-turn short circuit does not trigger significant gas evolution). Second, gas detection is affected by factors such as oil temperature and load fluctuations, making it difficult for a single parameter to distinguish between true faults and interference signals. Existing fusion algorithms (such as simple weighted averaging) are prone to making erroneous decisions. Furthermore, fixed-weight fusion strategies struggle to cope with the changing characteristics of transformers under varying loads and operating conditions.

[0005] In addition, uploading massive monitoring data to the cloud is subject to network transmission delays and cannot meet the "second-level warning" requirements for transformer failures; the entire cloud transmission of sensitive multi-parameter data increases the risk of data leakage and tampering.

[0006] In summary, there is an urgent need for a method that can effectively integrate multi-source heterogeneous data, handle evidence uncertainty, and realize real-time fault diagnosis at the edge. Summary of the Invention

[0007] In order to solve the above-mentioned problems existing in the prior art, the present application provides a transformer fault diagnosis method and system based on multi-parameter fusion.

[0008] The technical solution of this application is as follows:

[0009] On the one hand, the present invention proposes a transformer fault diagnosis method based on multi-parameter fusion, the method comprising:

[0010] Collect multi-parameter data of local transformer;

[0011] The multi-parameter data is encrypted using an encryption algorithm and transmitted to the edge; the edge decrypts the encrypted multi-parameter data and processes the decrypted multi-parameter data;

[0012] The edge uses the data fusion algorithm based on DS evidence theory to fuse the processed multi-parameter data to obtain multi-parameter fusion data;

[0013] A fault diagnosis model is built based on deep learning to perform fault diagnosis on multi-parameter fusion data, obtain fault diagnosis results and upload them to the cloud.

[0014] Preferably, the multi-parameter data is encrypted using an encryption algorithm, specifically:

[0015] The local processing center performs preliminary processing on the multi-parameter data and uses the hash function to calculate the unique identifier of the transformer and the hash value of the corresponding multi-parameter data;

[0016] The transformer unique identifier, multi-parameter data and hash value are encrypted to obtain encrypted data.

[0017] Preferably, the edge end decrypts the encrypted multi-parameter data, specifically:

[0018] The edge end calculates the received unique identifier of the transformer and the hash value of the corresponding multi-parameter data to obtain a verification value; the verification value is compared with the received hash value. If they are consistent, it means that the encrypted data has not been tampered with and data processing is performed; otherwise, it means that the encrypted data has been tampered with and a request failure signal is returned to the local segment.

[0019] Preferably, the data fusion algorithm based on DS evidence theory integrates historical fault data and expert experience, calculates the ratio of the current multi-parameter data to the amount of historical fault data based on the historical fault data, and obtains the first probability;

[0020] Calculate the ratio of current multi-parameter data to the number of historical fault data based on historical fault data to obtain a first probability;

[0021] Assign a value to the probability of each fault type occurring in the current multi-parameter data based on expert experience to obtain a second probability;

[0022] Weighted calculation of the first probability and the second probability to obtain a basic probability distribution;

[0023] Based on Dempster's synthesis rule, the processed multi-parameter data are randomly combined in pairs, and the basic probability distributions corresponding to any two combinations are synthesized;

[0024] The synthesis process is iterated to obtain multi-parameter fusion data.

[0025] On the other hand, the present invention also proposes a transformer fault diagnosis system based on multi-parameter fusion, which includes a data acquisition module, an encryption and decryption module, a fusion module and a fault diagnosis module, wherein:

[0026] The data acquisition module is used to collect multi-parameter data of the local transformer; and transmit the multi-parameter data to the encryption and decryption module;

[0027] The encryption and decryption module is used to encrypt the multi-parameter data using an encryption algorithm and transmit it to the edge end; the edge end decrypts the encrypted multi-parameter data and performs data processing on the decrypted multi-parameter data;

[0028] The fusion module is used at the edge end to fuse the processed multi-parameter data using a data fusion algorithm based on DS evidence theory to obtain multi-parameter fusion data;

[0029] The fault diagnosis module constructs a fault diagnosis model based on deep learning to perform fault diagnosis on multi-parameter fusion data, obtains fault diagnosis results and uploads them to the cloud.

[0030] Preferably, the system further includes a communication feedback module for debugging the transformer fault diagnosis system, monitoring multi-parameter data communication in real time, and providing feedback; and optimizing the fusion module and the fault diagnosis module based on the feedback content.

[0031] On the other hand, the present invention also proposes an electronic device, comprising a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the program, a transformer fault diagnosis method based on multi-parameter fusion as described in any embodiment of the present invention is implemented.

[0032] On the other hand, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a transformer fault diagnosis method based on multi-parameter fusion as described in any embodiment of the present invention.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] 1) The present invention provides a transformer fault diagnosis method and system based on multi-parameter fusion. By integrating multi-dimensional parameters such as gas concentration, winding temperature, and current, it breaks the one-sidedness of single-parameter diagnosis, improves the accuracy of fault identification, enhances the ability to resist environmental interference, and avoids false alarms caused by factors such as oil temperature and load fluctuations in a single parameter.

[0035] 2) The present invention provides a transformer fault diagnosis method and system based on multi-parameter fusion, which utilizes DS evidence theory to enhance diagnostic robustness. It uses basic probability distribution to quantify the uncertainty of multi-source data and fuses conflicting evidence through Dempster synthesis rule, thus solving the decision-making problem when multi-parameter data conflicts and improving the diagnostic reliability in evidence conflict scenarios.

[0036] 3) The present invention provides a transformer fault diagnosis method and system based on multi-parameter fusion. Based on localized calculation, it improves real-time performance and security, completes data encryption transmission at the local end, decryption preprocessing and evidence fusion at the edge end, and only uploads diagnostic results to the cloud instead of original data, achieving fault warning in seconds; reducing data transmission volume, reducing communication bandwidth pressure, and preventing tampering and leakage during transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flow chart of a method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The specific embodiments of the present invention are described below to facilitate understanding of the present invention by those skilled in the art. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations utilizing the concepts of the present invention are protected.

[0039] The present invention provides the following technical solution: a transformer fault diagnosis method and system based on multi-parameter fusion.

[0040] Example 1:

[0041] See Figure 1 This embodiment provides a transformer fault diagnosis method based on multi-parameter fusion, and the specific steps include:

[0042] S1. Collect multi-parameter data of the local transformer;

[0043] The multi-parameter data includes transformer gas, winding temperature, oil level, short-circuit current, overcurrent data, overvoltage data, overload data, grounding data, etc.;

[0044] By combining simulation with actual testing, the sensor installation locations are optimized and the installation points that can fully and accurately capture the transformer's multi-parameter data are determined, ensuring that parameter collection is complete and highly accurate.

[0045] S2. Encrypt the multi-parameter data using an encryption algorithm and transmit it to the edge;

[0046] The local processing center performs preliminary processing on the multi-parameter data and uses the hash function to calculate the unique identifier of the transformer and the hash value of the corresponding multi-parameter data;

[0047] Encrypt the transformer unique identifier, multi-parameter data and hash value to obtain encrypted data;

[0048] In this embodiment, the unique identifier of the transformer is the transformer model;

[0049] S3. The edge end decrypts the encrypted multi-parameter data and processes the decrypted multi-parameter data.

[0050] The edge end calculates the received unique identifier of the transformer and the hash value of the corresponding multi-parameter data to obtain a verification value; the verification value is compared with the received hash value. If they are consistent, it indicates that the encrypted data has not been tampered with and data processing is performed; otherwise, it indicates that the encrypted data has been tampered with and a request failure signal is returned to the local segment;

[0051] The data processing includes specifically data cleaning, which includes processing missing values, outliers and standardizing data formats;

[0052] S4. The edge uses a data fusion algorithm based on DS evidence theory to fuse the processed multi-parameter data to obtain multi-parameter fused data;

[0053] Define an identification framework θ={θ1,...,θ n ,...,θ N}, where θ n represents the nth state, θ N Represents the Nth state, n represents the nth state; and generates the 2nd power of the recognition frame 2 θ , which includes all subsets of the identification framework;

[0054] Based on historical failure data and expert experience or machine learning, θ Each subset in the distribution of basic probability mass function m(x i ), x i ∈2 θ , where m() represents the basic probability mass function, x i Represents the i-th multi-parameter data;

[0055] S41. The data fusion algorithm based on DS evidence theory integrates historical fault data and expert experience, calculates the ratio of current multi-parameter data to the number of historical fault data based on the historical fault data, and obtains the first probability;

[0056] According to the historical fault data, the proportion of the current multi-parameter data to the number of historical fault data is calculated to obtain the first probability, which is expressed as:

[0057]

[0058] Where p i,1 (θ n ) represents the first probability that the i-th multi-parameter data belongs to the n-th fault type; N represents the number of fault types; D i,n Indicates the number of i-th multi-parameter data belonging to n-th fault type;

[0059] Assign a value to the probability of each fault type occurring in the current multi-parameter data based on expert experience to obtain a second probability;

[0060] The first probability and the second probability are weighted to obtain the basic probability distribution, which can be expressed as:

[0061] m i (θ n )=α·p i,1 (θ n )+β·p i,2 (θ n );

[0062] α+β=1;

[0063] Where m i (θ n ) represents the basic probability distribution of the i-th multi-parameter data belonging to the n-th fault type; p i,2 (θ n ) represents the second probability that the i-th multi-parameter data belongs to the n-th fault type; α represents the first probability weight; β represents the second probability weight;

[0064] Based on Dempster's synthesis rule, the processed multi-parameter data are randomly combined in pairs, and the basic probability distribution corresponding to any two combinations is synthesized, which can be expressed as follows:

[0065]

[0066] Where m i,j(A) represents the basic probability distribution of the i-th multi-parameter data and the j-th multi-parameter data pair A; K represents the conflict coefficient; A represents the intersection of the first hypothesis subset B and the second hypothesis subset C; B represents the first hypothesis subset; C represents the second hypothesis subset; m i (B) represents the basic probability distribution of the i-th multi-parameter data pair B; m j (C) represents the basic probability distribution of the j-th multi-parameter data pair C;

[0067] Among them, if the conflict coefficient K is closer to 1, it means that there are more "completely incompatible" parts between the multi-parameter data and the conflict is more serious; set a conflict discount factor, multiply the basic probability distribution corresponding to the conflicting multi-parameter data by the conflict discount factor to generate a new basic probability distribution for the multi-parameter data;

[0068] The iterative synthesis process obtains multi-parameter fusion data, which can be expressed as follows:

[0069] X'=m 1,2 ⊕...⊕m i,j ⊕...⊕m (N-1),N ;

[0070] Where X' represents multi-parameter fusion data; m i,j represents the basic probability distribution of the i-th multi-parameter data and the j-th multi-parameter data; m (N-1),N Represents the basic probability distribution of the N-1th multi-parameter data and the Nth multi-parameter data;

[0071] The method for generating basic probability distribution further includes:

[0072] The processed multi-parameter data is input into the machine learning model to calculate the possibility that each multi-parameter data belongs to each fault category and obtain the assignment probability;

[0073] Normalizing the distribution probability to obtain the basic probability distribution of the multi-parameter data for different fault types;

[0074] The machine learning models include support vector machines, random forests, Gaussian mixture models, variational autoencoders, etc.

[0075] In this embodiment, the basic probability distribution of different fault types based on the multi-parameter data after processing is calculated based on the support vector machine. Specifically, the processed multi-parameter data is input into the support vector machine, and the possibility of each multi-parameter data belonging to each fault category is calculated to obtain the distribution probability;

[0076] Normalizing the distribution probability to obtain the basic probability distribution of the multi-parameter data for different fault types;

[0077] In another embodiment, the basic probability distribution of the processed multi-parameter data to different fault types is calculated based on the Gaussian mixture model, specifically: initializing N Gaussian distributions, each Gaussian distribution corresponds to a fault type, and calculating the probability that each multi-parameter data belongs to each Gaussian distribution, that is, the basic probability distribution of the current multi-parameter data belonging to each fault type;

[0078] S5. Build a fault diagnosis model based on deep learning to perform fault diagnosis on multi-parameter fusion data, obtain fault diagnosis results and upload them to the cloud;

[0079] In this embodiment, fault diagnosis is performed on the S9-1000kVA transformer;

[0080] The gas concentrations in the collected gas are H = 150ppm, CH = 5ppm, CH = 20ppm; the winding temperature is 75°C; the oil level is 0.4; and the short-circuit current is 1.2A.

[0081] The hash function is used to calculate the unique identification of the transformer and the hash value of the corresponding multi-parameter data, which is expressed as hash = SHA256 (str (X) + BIO), X = [150, 5, 20, 75, 0.4, 1.2], where SHA256 represents the hash function, X represents the multi-parameter data, hash represents the hash value, and BIO is S9-1000kVA, which represents the unique identification of the transformer;

[0082] Encrypt [BIO, X, hash] using AES-256 with key s and transmit it to the edge.

[0083] The edge end uses the same key s to decrypt, recalculate the hash value of the transformer unique identifier and the corresponding multi-parameter data, and obtain the verification value represented as hash';

[0084] Compare hash' with hash. If they are consistent, it means that the encrypted data has not been tampered with and the data is processed. Otherwise, it means that the encrypted data has been tampered with and a request failure signal is returned to the local segment.

[0085] Define an identification frame θ = {θ1 (normal), θ2 (turn-to-turn short circuit), θ3 (core grounding), θ4 (oil quality deterioration)}; and generate all subsets of the identification frame 2 θ =A, where A represents {(normal), (normal, turn-to-turn short circuit), (normal, turn-to-turn short circuit, core grounded), ...};

[0086] According to historical fault data, when the gas concentration H>100ppm and CH>5ppm, 80% is due to inter-turn short circuit, 15% is due to oil deterioration, and 5% is normal. Calculate the first probability, for example, p 1,1 (θ2)=0.8,p1,1 (θ4)=0.15,p 1,1 (θ1) = 0.05;

[0087] According to expert experience, the probability of the fault type occurring in the current multi-parameter data is assigned to obtain a second probability, such as p 1,2 (θ2)=0.7,p 1,2 (θ2,θ3)=0.12,p 1,2 (θ1) = 0.08;

[0088] Calculate the basic probability distribution, for example, m1(θ4) = 0.6 × 0.15 + 0.4 × 0 = 0.09, m1(θ2) = 0.6 × 0.8 + 0.4 × 0.7 = 0.76, m1(θ1) = 0.6 × 0.05 + 0.4 × 0.08 = 0.062, m1(θ2, θ3) = 0.4 × 0.12 = 0.48;

[0089] The six parameters are divided into three groups, represented as [H, CH], [CH, winding temperature] and [oil level, short-circuit current];

[0090] Calculate the conflict coefficient between the first and second groups, for example, expressed as K = m1(B) × m2(C) + m1(C) × m2(B) = 0.152, where B is θ2 and C is all subsets containing θ2, such as {θ2}, {θ2,θ3}, {θ2,θ4}, etc.; synthesize the basic probability distribution of the first and second groups into:

[0091] After iterating the synthesis process, the multi-parameter fusion data is expressed as X'=[0,0.91,0,0,0.09];

[0092] Input the multi-parameter fusion data into the fault diagnosis model to obtain the probability of belonging to each fault type H(θ2) = 0.95, H(θ1) = 0.02, H(θ3) = 0.01, H(θ4) = 0.02;

[0093] The fault type with the highest probability is selected as the fault diagnosis result. Therefore, in this embodiment, the fault diagnosis result of the S9-1000kVA transformer is turn-to-turn short circuit.

[0094] Example 2:

[0095] This embodiment provides a transformer fault diagnosis system based on multi-parameter fusion, which includes a data acquisition module, an encryption and decryption module, a fusion module, and a fault diagnosis module, wherein:

[0096] The data acquisition module is used to collect multi-parameter data of the local transformer; and transmit the multi-parameter data to the encryption and decryption module;

[0097] The encryption and decryption module is used to encrypt the multi-parameter data using an encryption algorithm and transmit it to the edge end; the edge end decrypts the encrypted multi-parameter data and performs data processing on the decrypted multi-parameter data;

[0098] The fusion module is used at the edge end to fuse the processed multi-parameter data using a data fusion algorithm based on DS evidence theory to obtain multi-parameter fusion data;

[0099] The fault diagnosis module builds a fault diagnosis model based on deep learning to perform fault diagnosis on multi-parameter fusion data, obtains fault diagnosis results and uploads them to the cloud;

[0100] The system also includes a communication feedback module for debugging the transformer fault diagnosis system, monitoring multi-parameter data communication in real time, and providing feedback; and optimizing the fusion module and the fault diagnosis module based on the feedback content.

[0101] Example 3:

[0102] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, a transformer fault diagnosis method based on multi-parameter fusion as described in any embodiment of the present invention is implemented.

[0103] Example 4:

[0104] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, a transformer fault diagnosis method based on multi-parameter fusion as described in any embodiment of the present invention is implemented.

[0105] It is worth noting that the system, electronic device and computer-readable storage medium described in the present invention are all based on the same principles as the method described in Example 1, and will not be repeated here.

[0106] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention's description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A transformer fault diagnosis method based on multi-parameter fusion, characterized in that: The method comprises: Collect multi-parameter data of local transformer; The multi-parameter data is encrypted using an encryption algorithm and transmitted to the edge; the edge decrypts the encrypted multi-parameter data and processes the decrypted multi-parameter data; The edge uses the data fusion algorithm based on DS evidence theory to fuse the processed multi-parameter data to obtain multi-parameter fusion data; A fault diagnosis model is built based on deep learning to perform fault diagnosis on multi-parameter fusion data, obtain fault diagnosis results and upload them to the cloud.

2. A transformer fault diagnosis method based on multi-parameter fusion according to claim 1, characterized in that: The encryption algorithm is used to encrypt the multi-parameter data, specifically: The local processing center performs preliminary processing on the multi-parameter data and uses the hash function to calculate the unique identifier of the transformer and the hash value of the corresponding multi-parameter data; The transformer unique identifier, multi-parameter data and hash value are encrypted to obtain encrypted data.

3. A transformer fault diagnosis method based on multi-parameter fusion according to claim 2, characterized in that: The edge end decrypts the encrypted multi-parameter data, specifically: The edge end calculates the hash value of the received transformer unique identifier and the corresponding multi-parameter data to obtain the verification value; The verification value is compared with the received hash value. If they are consistent, it means that the encrypted data has not been tampered with and data processing is performed; otherwise, it means that the encrypted data has been tampered with and a request failure signal is returned to the local segment.

4. A transformer fault diagnosis method based on multi-parameter fusion according to claim 1, characterized in that: The data fusion algorithm based on DS evidence theory integrates historical fault data and expert experience, calculates the proportion of current multi-parameter data to the number of historical fault data based on historical fault data, and obtains the first probability; Calculate the ratio of current multi-parameter data to the number of historical fault data based on historical fault data to obtain a first probability; Assign a value to the probability of each fault type occurring in the current multi-parameter data based on expert experience to obtain a second probability; Weighted calculation of the first probability and the second probability to obtain a basic probability distribution; Based on Dempster's synthesis rule, the processed multi-parameter data are randomly combined in pairs, and the basic probability distributions corresponding to any two combinations are synthesized; The synthesis process is iterated to obtain multi-parameter fusion data.

5. A transformer fault diagnosis system based on multi-parameter fusion, characterized in that: The system includes a data acquisition module, an encryption and decryption module, a fusion module and a fault diagnosis module, wherein: The data acquisition module is used to collect multi-parameter data of the local transformer; and transmit the multi-parameter data to the encryption and decryption module; The encryption and decryption module is used to encrypt the multi-parameter data using an encryption algorithm and transmit it to the edge end; the edge end decrypts the encrypted multi-parameter data and performs data processing on the decrypted multi-parameter data; The fusion module is used at the edge end to fuse the processed multi-parameter data using a data fusion algorithm based on DS evidence theory to obtain multi-parameter fusion data; The fault diagnosis module constructs a fault diagnosis model based on deep learning to perform fault diagnosis on multi-parameter fusion data, obtains fault diagnosis results and uploads them to the cloud.

6. A transformer fault diagnosis system based on multi-parameter fusion according to claim 5, characterized in that: The system also includes a communication feedback module for debugging the transformer fault diagnosis system, monitoring multi-parameter data communication in real time, and providing feedback; and optimizing the fusion module and the fault diagnosis module based on the feedback content.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the transformer fault diagnosis method based on multi-parameter fusion as described in any one of claims 1 to 4 is implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the transformer fault diagnosis method based on multi-parameter fusion as described in any one of claims 1 to 4 is implemented.

Images