Dynamic analysis method for anti-short-circuit capability of in-operation transformer in consideration of health state

Abstract

The invention belongs to the technical field of power equipment monitoring, and relates to an in-operation transformer anti-short-circuit capability dynamic analysis method considering a health state, and the method comprises the steps: obtaining the health state perception data of a transformer at a historical moment; calculating an anti-short-circuit capability equivalent loss coefficient according to the dynamic equivalent short-circuit current; performing iterative training on a preset neural network by taking the health state sensing data as input and the anti-short-circuit capability equivalent loss coefficient as output to obtain an anti-short-circuit capability dynamic analysis model; and inputting real-time health state sensing data into the anti-short-circuit capability dynamic analysis model, and outputting through the anti-short-circuit capability dynamic analysis model to obtain a real-time anti-short-circuit capability equivalent loss coefficient, namely, obtaining a transformer anti-short-circuit capability evaluation result. According to the method, the health state and the anti-short-circuit capability equivalent loss coefficient are mapped through the dynamic equivalent short-circuit current and the neural network, and the defect that the real-time influence of insulation aging cannot be quantified through a traditional method is overcome.

Description

Technical Field

[0001] The present invention belongs to the technical field of power equipment monitoring, and particularly relates to a dynamic analysis method for the short-circuit withstand capacity of in-service transformers considering the health state. Background Art

[0002] As a core device in the power system, the reliability and safety of transformers are directly related to the stable operation of the power grid. In actual operation, transformers often face the risk of short-circuit impact, and their short-circuit withstand capacity is an important indicator to measure whether the equipment can operate safely. However, with the long-term operation of transformers, the influence of insulation aging, mechanical structure deterioration, and external environmental factors will cause their short-circuit withstand capacity to gradually decline. This change in the health state makes it difficult for traditional static evaluation methods to accurately reflect the actual short-circuit withstand capacity of transformers. The prior art of Chinese Patent No. 202210057776.1, "A Dynamic Evaluation Method for the Short-Circuit Resistance of Power Transformers Based on Multiple Parameters", usually evaluates the short-circuit withstand capacity of transformers based on design parameters and initial states. However, this method relies on historical data segmentation and cannot dynamically track the capacity loss caused by the deterioration of the health state during the operation of transformers. It has poor real-time performance, and this method only uses normalization weighting and does not have the ability to non-linearly fit multi-source heterogeneous data, resulting in large evaluation errors and potential safety hazards. In addition, although existing monitoring means can collect multi-source data such as the vibration, noise, and dissolved gases in oil of transformers, there is a lack of an analysis model that dynamically correlates these data with the short-circuit withstand capacity, resulting in the inability to achieve real-time evaluation of the short-circuit withstand capacity. Summary of the Invention

[0003] In order to solve the problem that the in-service transformers cannot dynamically calculate in real time the loss of short-circuit withstand capacity caused by insulation aging and health deterioration after the operation of the transformers, the present invention proposes a dynamic analysis method for the short-circuit withstand capacity of in-service transformers considering the health state. Through the combination of historical health state perception data, the calculation of dynamic equivalent short-circuit current and static equivalent short-circuit current, the training of neural network models, and real-time evaluation, the accurate dynamic analysis of the short-circuit withstand capacity of in-service transformers is realized.

[0004] The present invention provides a dynamic analysis method for the short-circuit withstand capacity of in-service transformers considering the health state, including: Step 1: Obtain the health state perception data of the transformer at historical moments; Step 2: Iteratively invert and calculate the static equivalent short-circuit current of the transformer's short-circuit withstand capacity according to a preset coupling model under a preset vibration; Step 3: Obtain the current peak value in an actual short-circuit impact event and correct it to obtain the dynamic equivalent short-circuit current, and calculate the equivalent loss coefficient of the short-circuit withstand capacity according to the dynamic equivalent short-circuit current; Step 4: Correlate the health state perception data of the transformer at the short-circuit impact moment with the equivalent loss coefficient of the short-circuit resistance ability; use the health state perception data as the input and the equivalent loss coefficient of the short-circuit resistance ability as the output to iteratively train a preset neural network to obtain a dynamic analysis model of the short-circuit resistance ability. Step 5: Input the real-time health state perception data into the dynamic analysis model of the short-circuit resistance ability, and obtain the real-time equivalent loss coefficient of the short-circuit resistance ability through the output of the dynamic analysis model of the short-circuit resistance ability. Calculate the actual short-circuit current that the in-service transformer can currently withstand according to the real-time equivalent loss coefficient of the short-circuit resistance ability and the static equivalent short-circuit current, that is, obtain the evaluation result of the transformer's short-circuit resistance ability.

[0005] Further preferably, in Step 1, the health state perception data includes: short-circuit impact cracking working conditions, short-circuit resistance sensitive monitoring data, winding maintenance test data, and product design data. Short-circuit impact cracking working condition data: historical short-circuit current peak , short-circuit impact duration and cumulative impact times ; Short-circuit resistance sensitive monitoring data: winding vibration acceleration spectrum , noise sound pressure level , concentration of dissolved gases in oil (H2, C2H2) ; Winding maintenance test data: winding deformation amount , insulation resistance R ins ; Product design data: short-circuit impedance , winding elastic modulus E, and insulation spacer spacing d.

[0006] Further preferably, Step 2 includes: Using the winding elastic modulus E and insulation spacer spacing d in the product design data, establish a winding displacement equation: , where is the winding displacement, is the winding material density, is the core magnetic permeability, is the time, is the static equivalent characteristic current, is the unit distance in the vertical force direction of the insulation spacer; In events where the historical short-circuit impact current ratio exceeds 50%, extract the vibration acceleration peak A peak ; When terminate the iteration, is the critical winding displacement, and the output static equivalent characteristic current is used as the static equivalent short-circuit current.

[0007] Further preferably, in step three, obtain the current peak value in the actual short-circuit impact event and correct it to the dynamic equivalent short-circuit current according to the following rules : Historical short-circuit current peak value Limitation: If , then ; Short-circuit impact duration Attenuation: If is greater than 0.5 seconds, attenuate according to ; Cumulative impact times Correction: If > 10, add a correction coefficient , that is .

[0008] Further preferably, in step three, the calculation process of the equivalent loss coefficient of short-circuit resistance is as follows: Based on finite element simulation, calculate the theoretical vibration and theoretical leakage magnetic flux under the dynamic equivalent short-circuit current ; in the short-circuit impact event, synchronously record the actual vibration and actual leakage magnetic flux ; The equivalent loss coefficient of short-circuit resistance is expressed as: , wherein , respectively represent the weights of vibration and leakage magnetic flux on the attenuation of short-circuit resistance, is the vibration frequency.

[0009] Further preferably, in step four, the dynamic analysis model of short-circuit resistance is established by mapping the health state and the equivalent loss coefficient of short-circuit resistance, specifically including: Step S41, input feature construction: Correlate the health state data at the short-circuit impact moment with the equivalent loss coefficient of short-circuit resistance , and the input features include: vibration deviation , leakage magnetic deviation , winding deformation , concentration of dissolved gas in oil , insulation resistance R ins , short-circuit impedance , winding elastic modulus E and insulation pad spacing d; Step S42: Training of neural network model: The neural network structure includes an input layer: receiving feature vectors ; Hidden layer: two fully connected layers; Output layer: constrained by the Sigmoid function ∈(0, 1], directly mapping the equivalent loss coefficient of short - circuit resistance ability; The training and optimization data set is historical short - circuit event data with labels, divided into a training set, a validation set, and a test set according to 7:2:1; optimized by the Adam algorithm.

[0010] Further preferably, in step five, after pre - processing the real - time health status perception data, a real - time feature vector X real is constructed; X real is input into the trained neural network model to output the current real - time equivalent loss coefficient of the short - circuit resistance ability of the in - service transformer .

[0011] Further preferably, according to the real - time equivalent loss coefficient of the short - circuit resistance ability, calculate the current actual short - circuit current that the in - service transformer can withstand : , where is the static equivalent characteristic current, is the real - time equivalent loss coefficient of the short - circuit resistance ability.

[0012] The beneficial effects of the present invention are as follows: By combining the calculation of historical health status perception data, dynamic equivalent short - circuit current and static equivalent short - circuit current, the training of the neural network model and real - time evaluation, the problem that the traditional method cannot calculate the short - circuit resistance ability of the transformer in real - time and dynamically is solved, and real - time and dynamic evaluation can be realized, providing reliable technical support for the safe operation of the transformer. The present invention corrects the dynamic equivalent short - circuit current rule and maps the health status and loss coefficient through the neural network, solving the defect that the traditional method cannot quantify the real - time impact of insulation aging. Brief Description of the Drawings

[0013] Figure 1 is the flow chart of the present invention. Detailed Embodiment

[0014] The present invention will be further described in detail below with reference to the drawings.

[0015] Referring to Figure 1 , a dynamic analysis method for the short - circuit resistance ability of an in - service transformer considering the health status includes: Step 1. Obtain the health status perception data of the transformer at historical moments; the health status perception data includes: short-circuit impact cracking working conditions, short-circuit sensitive monitoring data, winding maintenance test data, and product design data; Short-circuit impact cracking working condition data: historical short-circuit current peak , short-circuit impact duration and cumulative impact times ; Short-circuit sensitive monitoring data: winding vibration acceleration spectrum , noise sound pressure level , concentration of dissolved gases in oil (H2, C2H2) ; Winding maintenance test data: winding deformation amount , insulation resistance R ins ; Product design data: short-circuit impedance , winding elastic modulus E and insulation pad spacing d.

[0016] After obtaining the health status perception data of the transformer at historical moments, construct an adaptive sliding window based on the multimodal joint distribution, and perform collaborative detection on the outliers in the health status perception data through a dynamic probability graph model to obtain the preprocessed health status perception data.

[0017] Step 2. Iteratively invert and calculate the static equivalent short-circuit current of the transformer's short-circuit resistance ability according to the preset coupling model under the preset vibration; Using the winding elastic modulus E and insulation pad spacing d in the product design data, establish a winding displacement equation: , where is the winding displacement amount, is the winding material density, is the core magnetic permeability, is the time, is the static equivalent characteristic current, is the unit distance in the vertical force direction of the insulation pad; In the event that the historical short-circuit impact current ratio exceeds 50%, extract the vibration acceleration peak A peak ; Through the winding displacement equation, when the iteration is terminated (in this embodiment, = 2mm), is the critical winding displacement amount, and the output static equivalent characteristic current is used as the static equivalent short-circuit current.

[0018] Step 3: Obtain the peak current in the actual short - circuit impact event, correct it to obtain the dynamic equivalent short - circuit current, and calculate the equivalent reduction factor of the short - circuit withstand ability based on the dynamic equivalent short - circuit current; Step 4: Correlate the health - state perception data of the transformer at the short - circuit impact moment with the equivalent reduction factor of the short - circuit withstand ability; use the health - state perception data as the input and the equivalent reduction factor of the short - circuit withstand ability as the output, and perform iterative training on the preset neural network to obtain a dynamic analysis model of the short - circuit withstand ability; Step 5: Input the real - time health - state perception data into the dynamic analysis model of the short - circuit withstand ability, output the real - time equivalent reduction factor of the short - circuit withstand ability through the dynamic analysis model of the short - circuit withstand ability, and calculate the current that the in - service transformer can actually withstand at present according to the real - time equivalent reduction factor of the short - circuit withstand ability and the static equivalent short - circuit current, that is, obtain the evaluation result of the transformer's short - circuit withstand ability.

[0019] In Step 3, obtain the peak current in the actual short - circuit impact event and correct it to the dynamic equivalent short - circuit current according to the following rules : Historical short - circuit current peak Limitation: If , then ; Short - circuit impact duration Attenuation: If is greater than 0.5 seconds, attenuate according to ; Cumulative impact times Correction: If > 10, add a correction factor , that is .

[0020] The calculation process of the equivalent reduction factor of the short - circuit withstand ability is as follows: Based on finite - element simulation, calculate the theoretical vibration and theoretical leakage magnetic flux under the dynamic equivalent short - circuit current ; in the short - circuit impact event, synchronously record the actual vibration and actual leakage magnetic flux ; The equivalent reduction factor of the short - circuit withstand ability is expressed as: , In the formula , respectively represent the weights of vibration and leakage magnetic flux on the attenuation of the short - circuit withstand ability, is the vibration frequency.

[0021] In step 4, the dynamic analysis model of short - circuit resistance is modeled by mapping the health state and the equivalent loss coefficient of short - circuit resistance, specifically including: Step S41, input feature construction: Associate the health state data at the moment of short - circuit impact with the equivalent loss coefficient of short - circuit resistance The input features include: (1) Mechanical state: vibration deviation , magnetic flux leakage deviation , winding deformation amount ; (2) Insulation state: concentration of dissolved gases in oil , insulation resistance R ins ; (3) Product design parameters: short - circuit impedance , winding elastic modulus E and insulation spacer spacing d; Step S42, neural network model training: The network structure includes an input layer: receiving feature vectors Hidden layer: 2 fully - connected layers (64 neurons, ReLU activation); Output layer: constrained by the Sigmoid function ∈(0,1], directly mapping the equivalent loss coefficient of short - circuit resistance; The training and optimization data set is historical short - circuit event data with labels, divided into a training set, a validation set, and a test set in a ratio of 7:2:1; optimized by the Adam algorithm, and early stopping is used to prevent overfitting.

[0022] In step 5, after pre - processing the real - time health state perception data, construct the real - time feature vector X real ; Input X real into the trained neural network model, and output the current real - time equivalent loss coefficient of short - circuit resistance of the in - operation transformer . The calculation method of short - circuit current is as follows: , where is the static equivalent characteristic current, is the real - time equivalent loss coefficient of short - circuit resistance.

[0023] The technical solution of the present invention solves the problem that the traditional method cannot calculate the short-circuit withstand ability of transformers in real time and dynamically by combining historical health state perception data, the calculation of static and dynamic equivalent short-circuit current values, the training of neural network models, and real-time evaluation. In actual application scenarios, for example, transformers operating in a large substation will experience short-circuit impact events at regular intervals. Through the method of the present invention, the health state of the transformer can be monitored in real time, its short-circuit withstand ability can be dynamically evaluated, and a maintenance plan can be formulated or the operation strategy can be adjusted according to the evaluation results. Compared with the traditional method, the present invention not only improves the evaluation accuracy, but also significantly reduces the failure risk caused by insufficient short-circuit withstand ability of transformers, providing a reliable guarantee for the safe and stable operation of the power system.

Claims

1. A dynamic analysis method for the short-circuit withstand capacity of an in-service transformer considering the health status, characterized in that, Including: Step 1, obtaining the health state perception data of the transformer at historical moments; Step 2, iteratively inversely calculating the static equivalent short-circuit current of the short-circuit withstand capacity of the transformer according to a preset coupling model under a preset vibration; Step 3, obtaining the current peak value in an actual short-circuit impact event and correcting it to obtain the dynamic equivalent short-circuit current, and calculating the equivalent loss coefficient of the short-circuit withstand capacity according to the dynamic equivalent short-circuit current; Step 4, correlating the health state perception data of the transformer at the short-circuit impact moment with the equivalent loss coefficient of the short-circuit withstand capacity; Using the health state perception data as the input and the equivalent loss coefficient of the short-circuit withstand capacity as the output, iteratively training a preset neural network to obtain a dynamic analysis model of the short-circuit withstand capacity; Step 5, inputting the real-time health state perception data into the dynamic analysis model of the short-circuit withstand capacity, outputting the real-time equivalent loss coefficient of the short-circuit withstand capacity through the dynamic analysis model of the short-circuit withstand capacity, and calculating the currently actual withstandable short-circuit current of the in-service transformer according to the real-time equivalent loss coefficient of the short-circuit withstand capacity and the static equivalent short-circuit current, that is, obtaining the evaluation result of the short-circuit withstand capacity of the transformer.

2. The dynamic analysis method for the short-circuit withstand ability of an in-service transformer considering the health status according to claim 1, wherein In Step 1, the health state perception data includes: Short-circuit impact cracking process data: historical peak short-circuit current , short-circuit impact duration and cumulative impact times ; Short-circuit resistance sensitive monitoring data: Winding vibration acceleration spectrum , noise sound pressure level , dissolved gas concentration in oil ; Winding maintenance test data: Winding deformation , Insulation resistance R ins ; Product design data: short-circuit impedance , the elastic modulus E of the winding and the spacing d of the insulating pads.

3. A dynamic analysis method for the short-circuit withstand capacity of an in-service transformer considering the health status according to claim 1, characterized in that The said Step 2 includes: Using the winding elastic modulus E and the insulation pad spacing d in the product design data to establish a winding displacement equation: ， Wherein, is the winding displacement amount, is the density of the winding material, is the magnetic permeability of the iron core, is the time, is the static equivalent characteristic current, is the unit distance in the direction perpendicular to the force on the insulation spacer; Extract the peak vibration acceleration A in events where the historical short-circuit impact current ratio exceeds 50% peak ; When appropriate terminate the iteration is the critical winding displacement, and the output static equivalent characteristic current is used as the static equivalent short-circuit current.

4. A dynamic analysis method for the short-circuit withstand capacity of an in-service transformer considering the health status according to claim 1, characterized in that In the third step, obtain the peak current in the actual short-circuit impact event and correct it to the dynamic equivalent short-circuit current according to the following rules : Historical peak short-circuit current Limitation: If , then ; Short-circuit impact duration Decay: If is greater than 0.5 seconds, follow decay; Cumulative impact times Correction: If > 10, additional correction factor , that is .

5. The dynamic analysis method for the short-circuit withstand capacity of an in-service transformer considering the health status according to claim 1, wherein, In the said Step 3, the calculation process of the equivalent loss coefficient of the short-circuit withstand capacity is as follows: Calculate the dynamic equivalent short-circuit current based on finite element simulation Theoretical vibration under And theoretical leakage magnetic flux ; During the short-circuit impact event, synchronously record the actual vibration And actual leakage magnetic flux ; Equivalent loss coefficient of short-circuit resistance Expressed as: ， where , respectively represent the weights of vibration and leakage magnetic flux against the attenuation of short-circuit resistance, is the vibration frequency.

6. The dynamic analysis method for the short-circuit withstand capacity of an in-service transformer considering the health status according to claim 1, wherein In the said Step 4, the dynamic analysis model of the short-circuit withstand capacity is modeled through the mapping of the health state and the equivalent loss coefficient of the short-circuit withstand capacity, specifically including: Step S41, input feature construction: the health status data at the time of short circuit impact and the equivalent impairment coefficient of short circuit resistance capacity Correlation, input features include: vibration deviation , leakage magnetic deviation , winding deformation , Dissolved gas concentration in oil , Insulation resistance R ins , short circuit impedance , winding elastic modulus E and insulation spacer spacing d; Step S42, neural network model training: The neural network structure includes an input layer: receiving feature vectors Hidden layer: 2 fully connected layers; Output layer: Constrained by the Sigmoid function ∈(0,1], directly map the equivalent loss coefficient of the short-circuit resistance The training optimization dataset is historical short-circuit event data with labels, which are divided into a training set, a validation set, and a test set in a ratio of 7:2:1; optimized by the Adam algorithm.

7. A dynamic analysis method for the short-circuit withstand capacity of an in-service transformer considering the health status according to claim 1, characterized in that, In the fifth step, after preprocessing the real-time health status perception data, a real-time feature vector X is constructed. real ; X real is input into the trained neural network model to output the current equivalent loss coefficient of the short-circuit resistance ability of the in-service transformer. .

8. A dynamic analysis method for the short-circuit withstand capacity of an in-service transformer considering the health status according to claim 1, characterized in that, Short-circuit current Calculate according to the following formula; , where is the static equivalent characteristic current, is the equivalent loss coefficient of real-time short-circuit resistance.

Images