Transformer thermal circuit model parameter correction method based on extended Kalman filtering

Abstract

The invention discloses a transformer thermal circuit model parameter correction method based on extended Kalman filtering, and belongs to the technical field of transformer temperature rise tests, and the method comprises the steps: S1, defining a state equation representing the change of dynamic parameters along with time in a transformer thermal circuit model according to a state transition matrix; s2, on the basis of the physical relation between the observation parameters and the dynamic parameters in the transformer thermal circuit model, defining a nonlinear equation representing the nonlinear relation between the observation parameters and the dynamic parameters; s3, linearizing the nonlinear equation based on extended Kalman filtering to obtain a linear equation; and S4, obtaining a first predicted value of the dynamic parameter based on the state equation, obtaining a second predicted value of the observation parameter based on the linear equation, and correcting the first predicted value of the dynamic parameter based on the second predicted value and the actual value of the observation parameter to obtain a final value of the dynamic parameter. The technical problem of low transformer temperature prediction accuracy caused by the fact that the temperature dependence of thermal resistance is not considered in the prior art is solved.

Description

Technical Field

[0001] The invention relates to the field of transformer temperature rise test, and in particular to a transformer thermal circuit model parameter correction method based on extended Kalman filtering. Background Art

[0002] With the rapid development and complexity of the power system, the transformer, as a key equipment in the power system, has a direct impact on the stability and reliability of the entire power grid. The transformer generates heat during operation. If the temperature is too high or the temperature changes abnormally, it may cause aging of the insulation material, degradation of equipment performance, and even serious failures. Therefore, accurate prediction of transformer temperature is not only an important means to ensure the safe operation of the transformer, but also a key measure to improve the stability of the power system, extend the life of the equipment, and reduce operation and maintenance costs. The prior art mainly predicts the transformer temperature through the transformer thermal circuit model, such as a transformer winding hot spot temperature prediction method under low temperature conditions with patent number CN106595884A, first establishes an improved three-heat circuit model; modifies the improved three-heat circuit model according to different oil flow cooling methods; compares the temperature rise value of the transformer oil detected by the temperature sensor with the temperature value calculated by the modified three-heat circuit model, and modifies the transformer thermal circuit model close to the actual one; based on the ambient temperature, calculates the average oil temperature of the transformer through the transformer thermal circuit model close to the actual one, and then uses the average oil temperature of the transformer as the reference temperature of the middle layer oil temperature of the improved three-heat circuit model, calculates the top layer oil temperature in the model, and finally calculates the hot spot temperature value of the transformer winding. The accuracy of the hot spot temperature value of the transformer winding is improved, but the thermal resistance parameter in the transformer thermal circuit model of the above scheme is constant, ignoring the nonlinear characteristics of the thermal resistance changing with temperature during the temperature rise process, especially in oil-immersed transformers, the transformer thermal resistance will change significantly due to the increase in temperature. If constant thermal resistance parameters are still used for temperature prediction, the transformer thermal circuit model will not be able to accurately reflect the actual heat conduction process, thus introducing a large prediction error. Summary of the invention

[0003] In view of the technical problem that the temperature dependence of thermal resistance is not considered in the prior art, resulting in low transformer temperature prediction accuracy, the present invention provides a transformer thermal circuit model parameter correction method based on extended Kalman filtering, which defines a nonlinear equation through the physical relationship between the observed parameters and the dynamic parameters in the transformer thermal circuit model, accurately reflects the nonlinear characteristics of thermal resistance changing with temperature, and then linearizes the nonlinear equation to obtain a linear equation through the local linearization characteristics of the extended Kalman filter, further obtains the predicted value of the observed parameter based on the linear equation, and corrects the predicted value of the dynamic parameter in combination with the actual value of the observed parameter to obtain the final value of the dynamic parameter. The present invention overcomes the technical problem that the temperature dependence of thermal resistance is not considered in the prior art, resulting in low transformer temperature prediction accuracy, and significantly improves the transformer temperature prediction accuracy.

[0004] In order to solve the above technical problems, the present invention provides a transformer thermal circuit model parameter correction method based on extended Kalman filtering, comprising the following steps: S1: Define the state equations that characterize the time-varying dynamic parameters in the transformer thermal circuit model based on the state transfer matrix; S2: Based on the physical relationship between the observed parameters and the dynamic parameters in the transformer thermal circuit model, a nonlinear equation is defined to characterize the nonlinear relationship between the observed parameters and the dynamic parameters; S3: linearizing the nonlinear equation based on extended Kalman filtering to obtain a linear equation; S4: Obtain a first predicted value of the dynamic parameter based on the state equation, obtain a second predicted value of the observed parameter based on the linear equation, and correct the first predicted value of the dynamic parameter based on the second predicted value and the actual value of the observed parameter to obtain a final value of the dynamic parameter.

[0005] After adopting the above technical solution, the present invention has the following advantages: By defining a nonlinear equation through the physical relationship between the observed parameters and the dynamic parameters in the transformer thermal circuit model, the nonlinear characteristics of the thermal resistance changing with temperature are accurately reflected. Then, by extending the local linearization characteristics of the Kalman filter, the nonlinear equation is linearized to obtain a linear equation, so that the predicted value of the observed parameter is obtained according to the predicted value of the dynamic parameter, and the predicted value of the observed parameter is combined with the actual value for analysis, and then the predicted value of the dynamic parameter is corrected through the linear equation and the combined analysis result to obtain the final value of the dynamic parameter. The obtained linear equation establishes a bridge between the predicted value of the dynamic parameter, the predicted value of the observed parameter and the actual value of the observed parameter, so that the predicted value of the thermal resistance and other parameters can be dynamically corrected by comparing the predicted value of the observed parameter with the actual value, and finally an accurate estimate of the dynamic parameter is obtained. This overcomes the technical problem that the temperature dependence of the thermal resistance is not considered in the prior art, resulting in low accuracy in transformer temperature prediction, and significantly improves the accuracy of transformer temperature prediction and the adaptability of the transformer thermal circuit model; At the same time, by linearizing the nonlinear equations, that is, converting complex nonlinear problems into linear problems that are easy to solve, the prediction efficiency of the transformer temperature is also improved.

[0006] Preferably, the S1 comprises: S11: based on the characteristics of the random distribution variables in the transformer thermal circuit model, the cumulative distribution function of the random distribution variables is obtained, and then the uniform distribution variables are obtained, and the uniform distribution variables are mapped to Gaussian distribution variables by inverse Gaussian; S12: Obtain the state equation based on Gaussian distribution variables and the state transfer matrix.

[0007] In this scheme, by mapping the randomly distributed variables into Gaussian distributed variables suitable for extended Kalman filtering processing, while retaining the influence of external interference data on transformer temperature prediction, it is also convenient to convert it into linear influence through extended Kalman filtering, thereby improving the efficiency and accuracy of transformer temperature prediction.

[0008] Preferably, in S12, the state equation is: ; In the formula, x k+1 represents the value of the dynamic parameter at time k+1, A represents the state transfer matrix, x k represents the value of the dynamic parameter at time k, w k Represents the Gaussian distribution variable at time k.

[0009] Preferably, in S2, the observed parameters include at least winding temperature, top oil temperature, oil tank temperature and ambient temperature; The dynamic parameters at least include the thermal resistance between the winding and the oil, the thermal resistance between the oil and the oil tank, and the thermal resistance between the oil tank and the ring.

[0010] Preferably, in S2, the nonlinear relationship includes: ; ; ; In the formula, is the winding temperature, is the transformer load loss, is the winding heat capacity, is the top oil temperature, is the time step, is the thermal resistance of the winding to the oil, is the heat capacity of oil, is the no-load loss, is the fuel tank temperature, is the thermal resistance between the oil and the tank, is the heat capacity of the fuel tank, is the ambient temperature, is the thermal resistance between the tank and the ring.

[0011] Preferably, in S2, the nonlinear equation is: ; In the formula, z k is the value of the observed parameter at time k, h represents the nonlinear relationship function between the dynamic parameter and the observed parameter, v k represents the observed disturbance variable at time k.

[0012] Preferably, S3 includes: S31: Deriving the nonlinear equation to obtain an initial linear equation, inputting historical data corresponding to the observed parameters and the dynamic parameters into the initial linear equation to obtain a linearization degree of the initial linear equation; S32: When the linearization degree meets the preset condition, the initial linear equation is a linear equation; otherwise, the initial linear equation is regarded as a nonlinear equation and S31 is executed.

[0013] In this scheme, the linearization degree of the initial linear equation is verified by historical data. When the conditions are not met, the linearization degree of the initial linear equation is re-acquired until the preset conditions are met. This overcomes the prediction error caused by insufficient linearization of the nonlinear equation and further improves the accuracy of transformer temperature prediction.

[0014] Preferably, in S4, the step of correcting the first predicted value of the dynamic parameter based on the second predicted value and the actual value of the observed parameter to obtain the final value of the dynamic parameter includes: The inverse matrix to be obtained is obtained through the measurement matrix corresponding to the linear equation, and the Kalman gain is obtained based on the inverse matrix to be obtained; The state estimation value is obtained based on the Kalman gain, the second predicted value and the actual value, and the state covariance matrix is ​​obtained. When the state covariance matrix meets the preset requirements, the first predicted value is the final value. Otherwise, the state estimation value is corrected based on the state covariance matrix to obtain the final value.

[0015] Preferably, before obtaining the Kalman gain based on the inverse matrix to be determined, the method further includes: The desingularization parameters are obtained through the cross-validation method, and the inverse matrix to be determined is updated based on the desingularization parameters.

[0016] In this scheme, the inverse matrix to be inverted is updated by removing singular parameters, which overcomes the problem of high error rate in Kalman gain calculation caused by the singularity of the inverse matrix to be inverted. While improving the accuracy of the final value, it also improves the adaptability and accuracy of the transformer thermal circuit model.

[0017] The beneficial effects of this program: By defining a nonlinear equation through the physical relationship between the observed parameters and the dynamic parameters in the transformer thermal circuit model, the nonlinear characteristics of the thermal resistance changing with temperature are accurately reflected. Then, by extending the local linearization characteristics of the Kalman filter, the nonlinear equation is linearized to obtain a linear equation, so that the predicted value of the observed parameter is obtained according to the predicted value of the dynamic parameter, and the predicted value of the observed parameter is combined with the actual value for analysis, and then the predicted value of the dynamic parameter is corrected through the linear equation and the combined analysis result to obtain the final value of the dynamic parameter. The obtained linear equation establishes a bridge between the predicted value of the dynamic parameter, the predicted value of the observed parameter and the actual value of the observed parameter, so that the predicted value of the thermal resistance and other parameters can be dynamically corrected by comparing the predicted value of the observed parameter with the actual value, and finally an accurate estimate of the dynamic parameter is obtained. This overcomes the technical problem that the temperature dependence of the thermal resistance is not considered in the prior art, resulting in low accuracy in transformer temperature prediction, and significantly improves the accuracy of transformer temperature prediction and the adaptability of the transformer thermal circuit model; By mapping the random distribution variables into Gaussian distribution variables suitable for extended Kalman filtering, the influence of external interference data on transformer temperature prediction is retained, and it is also convenient to convert it into linear influence through extended Kalman filtering. Furthermore, the linearization degree of the initial linear equation is verified by historical data. When the conditions are not met, the linearization degree of the initial linear equation is re-acquired until the preset conditions are met. The prediction error caused by insufficient linearization of the nonlinear equation is overcome, and the accuracy of transformer temperature prediction is further improved. By removing singular parameters and updating the inverse matrix to be inverted, the problem of high error rate in Kalman gain calculation caused by the singularity of the inverse matrix to be inverted is overcome. While improving the accuracy of the final value, it also significantly improves the adaptability and accuracy of the transformer thermal circuit model.

[0018] The present invention also provides a storage medium, in which computer executable instructions are stored. When the computer executable instructions are loaded and executed by a processor, the steps of the transformer thermal circuit model parameter correction method based on extended Kalman filtering are implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Other features, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings. The drawings are only for the purpose of illustrating preferred embodiments and are not to be considered as limiting the present invention. Also, the same reference symbols are used throughout the drawings to represent the same parts.

[0020] Figure 1 The present invention is a flow chart of a transformer thermal circuit model parameter correction method based on extended Kalman filtering. Detailed Implementation Modes

[0021] To make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific implementation modes described herein are only the best embodiments of the present invention, which are only used to explain the present invention and do not limit the protection scope of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0022] Before discussing the exemplary embodiments in more detail, it should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the operations (or steps) as sequential processes, many of the operations (or steps) can be implemented in parallel, concurrently or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operations are completed, but it can also have additional steps not included in the drawings; the process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0023] Embodiment 1: As Figure 1 shown, a method for correcting parameters of a transformer thermal circuit model based on extended Kalman filtering includes the following steps: S1: Define a state equation characterizing the variation of dynamic parameters in the transformer thermal circuit model with time according to the state transition matrix.

[0024] The S1 includes: S11: Obtain the cumulative distribution function of the random distribution variable based on the characteristics of the random distribution variable in the transformer thermal circuit model, and then obtain a uniformly distributed variable, and map the uniformly distributed variable to a Gaussian distributed variable through inverse Gaussian; S12: Obtain the state equation based on the Gaussian distributed variable and the state transition matrix.

[0025] In S12, the state equation is: ; In the formula, x k+1 represents the value of the dynamic parameter at the (k + 1)th moment, A represents the state transition matrix, x k represents the value of the dynamic parameter at the kth moment, and w k represents the Gaussian distributed variable at the kth moment.

[0026] In this embodiment, the random distribution variable is the process noise, which indicates the uncertainty of the system state change. The cumulative distribution function of the random distribution variable is obtained based on the characteristics of the random distribution variable in the transformer thermal circuit model, specifically: the cumulative distribution function of the random distribution variable that does not conform to the Gaussian distribution is obtained. The random distribution variables that do not conform to the Gaussian distribution in the random distribution variables are mapped to Gaussian distribution variables through inverse Gaussian, thereby ensuring that the random distribution variables conform to the Gaussian distribution. Under the premise of retaining the influence of external interference data on the transformer temperature prediction, it is also convenient to convert it into a linear influence through the extended Kalman filter, thereby improving the efficiency and accuracy of transformer temperature prediction.

[0027] In this embodiment, taking the ambient temperature as an example, , It represents the value of the thermal resistance of the winding to the oil at time k, It represents the value of the thermal resistance between the oil and the oil tank at time k. The state transfer matrix is ​​used to describe the change of the system state between different time steps. According to the state transfer matrix, the value of the thermal resistance in the next time step can be obtained; since the temperature rise process generally progresses slowly, the oil temperature time constant is large, and the value of the thermal resistance changes little within a small time interval, the state transfer matrix can be set to the unit matrix. It is understandable that in practical applications, the change of the value of the thermal resistance may be affected by factors such as time lag and oil temperature change, so the state transfer matrix and process noise in the state equation will be different in different systems.

[0028] S2: Based on the physical relationship between the observed parameters and the dynamic parameters in the transformer thermal circuit model, a nonlinear equation is defined to characterize the nonlinear relationship between the observed parameters and the dynamic parameters.

[0029] In S2, the observed parameters include at least winding temperature, top oil temperature, oil tank temperature and ambient temperature; The dynamic parameters at least include the thermal resistance between the winding and the oil, the thermal resistance between the oil and the oil tank, and the thermal resistance between the oil tank and the ring.

[0030] In S2, the nonlinear relationship includes: ; ; ; In the formula, is the winding temperature, is the transformer load loss, is the winding heat capacity, is the top oil temperature, is the time step, is the thermal resistance of the winding to the oil, is the heat capacity of oil, is the no-load loss, is the fuel tank temperature, is the thermal resistance between the oil and the tank, is the heat capacity of the fuel tank, is the ambient temperature, is the thermal resistance between the tank and the ring.

[0031] In S2, the nonlinear equation is: ; In the formula, z k is the value of the observed parameter at time k, h represents the nonlinear relationship function between the dynamic parameter and the observed parameter, v k represents the observed disturbance variable at time k.

[0032] In this embodiment, the physical relationship specifically includes: ; ; ; In order to associate the thermal resistance with the actual temperature, a nonlinear equation is introduced. In the transformer thermal circuit model, the winding temperature, top oil temperature and tank temperature, which are closely related to the thermal resistance, are all affected by the thermal resistance. Therefore, a nonlinear equation needs to be established based on the physical relationship between thermal resistance and temperature. In this embodiment, , represents the value of winding temperature at time k, represents the value of the top oil temperature at time k, represents the value of the oil tank temperature at time k, approaches 0, and the observed disturbance variable is the observation noise.

[0033] S3: Linearize the nonlinear equation based on extended Kalman filtering to obtain a linear equation.

[0034] The S3 includes: S31: Deriving the nonlinear equation to obtain an initial linear equation, inputting historical data corresponding to the observed parameters and the dynamic parameters into the initial linear equation to obtain a linearization degree of the initial linear equation; S32: When the linearization degree meets the preset condition, the initial linear equation is a linear equation; otherwise, the initial linear equation is regarded as a nonlinear equation and S31 is executed.

[0035] In this embodiment, the nonlinear equation is derived to obtain the initial linear equation as follows: , the nonlinear relationship function between the dynamic parameters and the observed parameters is differentiated to obtain the Jacobian matrix of the nonlinear equation, and the initial linear equation is obtained through the Jacobian matrix. The Jacobian matrix representing the nonlinear equation, represents the dependent variable in the nonlinear relationship function, specifically the observed parameters, i.e., the temperatures. Represents the independent variable in the nonlinear relationship function, specifically the dynamic parameters, namely the thermal resistances. Specifically, taking the winding temperature and the winding-to-oil thermal resistance as examples: The specific calculation method of the elements in the Jacobian matrix is ​​as follows: , similarly, the partial derivatives of other temperatures with respect to thermal resistance can also be calculated, and the partial derivatives of temperature with respect to thermal resistance constitute the Jacobian matrix. The essence of the nonlinear equation is to calculate the temperature prediction value of each time step through the physical relationship between thermal resistance and the transformer thermal circuit model. In order to use the actual temperature to correct the predicted thermal resistance, the nonlinear equation is linearized by extending the Kalman filter, and the linearization degree of the initial linear equation is further verified by historical data. When the conditions are not met, the linearization degree of the initial linear equation is re-acquired until the preset conditions are met. The acquisition is stopped, which overcomes the prediction error caused by insufficient linearization of the nonlinear equation and further improves the accuracy of transformer temperature prediction. In this embodiment, when the historical data corresponding to the dynamic parameters are input into the initial linear equation, the output data and the historical data corresponding to the observation parameters have a matching degree greater than the preset matching degree, which means that the preset conditions are met. The preset matching degree here is flexibly set according to user needs, and can also be set in combination with the experimental results of historical data, thereby indirectly improving the flexibility of the transformer thermal circuit model.

[0036] S4: Obtain a first predicted value of the dynamic parameter based on the state equation, obtain a second predicted value of the observed parameter based on the linear equation, and correct the first predicted value of the dynamic parameter based on the second predicted value and the actual value of the observed parameter to obtain a final value of the dynamic parameter.

[0037] In S4, the step of correcting the first predicted value of the dynamic parameter based on the second predicted value and the actual value of the observed parameter to obtain the final value of the dynamic parameter includes: The inverse matrix to be obtained is obtained through the measurement matrix corresponding to the linear equation, and the Kalman gain is obtained based on the inverse matrix to be obtained; The state estimation value is obtained based on the Kalman gain, the second predicted value and the actual value, and the state covariance matrix is ​​obtained. When the state covariance matrix meets the preset requirements, the first predicted value is the final value. Otherwise, the state estimation value is corrected based on the state covariance matrix to obtain the final value.

[0038] Before obtaining the Kalman gain based on the inverse matrix to be found, it also includes: The desingularization parameters are obtained through the cross-validation method, and the inverse matrix to be determined is updated based on the desingularization parameters.

[0039] In this embodiment, the first predicted value is input into the linear equation to obtain the second predicted value. In addition to being obtained based on the state equation, the first predicted value of the dynamic parameter, namely the thermal resistance, can also be calculated from the state estimate at the previous moment. Another step in the prediction phase is the state covariance matrix prediction to measure the uncertainty of the thermal resistance estimate at the current moment. The prediction of the state covariance matrix is ​​based on the state covariance matrix at the previous moment. and the covariance of the process noise The calculation formula is as follows: , and finally adjust the current state estimate according to the error between the actual value and the second predicted value, including calculating the Kalman gain, updating the state estimate with the actual value, and updating the state covariance matrix. The Kalman gain is used to balance the weight between the prediction error and the observation error, and determines the degree of influence of the second predicted value on the state update. The calculation formula of the Kalman gain is: ,in: is the Kalman gain; is the state covariance matrix; H is the observation matrix, which represents the relationship between the state parameters and the second predicted value; is the observation noise covariance matrix, is the inverse matrix to be calculated. After the Kalman gain is calculated, the state estimate of the thermal resistance is updated according to the difference between the actual value and the second predicted value. The new state estimate is given by the following formula: ,in: is the updated state estimate (i.e., the thermal resistance value of k at the current moment); is the state estimate in the prediction step; is the actual value at the current moment; According to the predicted state The second predicted value is calculated, that is, the value obtained by the observation equation. Then the state covariance matrix is ​​updated. The calculation of the Kalman gain affects not only the update of the state estimate, but also the update of the state estimate uncertainty. The updated state covariance matrix Calculated by the following formula: ,in: is the identity matrix; is the Kalman gain; H is the observation matrix; is the predicted state covariance matrix. For example, the predicted thermal resistance value of k at the current moment is ; Prediction covariance matrix ; The observation matrix is ​​the identity matrix ; Observation noise covariance matrix ; The actual observed value at the current time ; Then the Kalman gain is: ; Update state estimate: , update the covariance matrix: . Correcting the state estimate based on the state covariance matrix to obtain the final value includes: correcting the state estimate by the state covariance matrix to obtain the corrected estimate, and then obtaining the corrected covariance matrix, judging whether the corrected covariance matrix meets the preset requirements, if it meets the preset requirements, the corrected estimate is the final value, otherwise the corrected estimate is corrected by the corrected covariance matrix until the corrected covariance matrix obtained after the correction meets the preset requirements, and then the iteration is stopped. Through the above steps, the change of thermal resistance in the transformer thermal circuit model can be estimated in real time, and the future temperature distribution can be predicted more accurately. The updated covariance matrix reflects the accuracy of the state estimate corrected by the actual value and the second predicted value. If the Kalman gain is large, it means that the dependence on the second predicted value is strong, and the updated state covariance matrix will become smaller, indicating that the uncertainty is reduced; conversely, if the Kalman gain is small, it means that the contribution of the second predicted value is small, and the state covariance matrix is ​​large. The preset requirements are set according to the accuracy of the transformer temperature prediction. If the accuracy of the transformer temperature needs to be maximized, the preset requirements are set relatively small.

[0040] In this embodiment, a desingularization parameter is obtained by a cross-validation method, and the inverse matrix to be determined is updated based on the desingularization parameter as follows: The historical data set is divided into a training set and a validation set through cross-validation to evaluate the performance of the transformer thermal circuit model under different desingularization parameters, and then the optimal desingularization parameters are selected. The optimal desingularization parameters are introduced into the inverse matrix to be determined to update the inverse matrix. represents the updated inverse matrix to be sought, represents the inverse matrix to be sought before the update, represents the optimal desingularization parameter, Represents the unit matrix. By removing singular parameters and updating the inverse matrix to be inverted, the problem of high error rate in Kalman gain calculation caused by the singularity of the inverse matrix to be inverted is overcome. While improving the accuracy of the final value, the adaptability and accuracy of the transformer thermal circuit model are also improved.

[0041] The present invention can dynamically adjust the thermal resistance estimation during the operation of the transformer to adapt to changes in factors such as temperature, oil flow, and environmental conditions. At the same time, by introducing the extended Kalman filter, it can obtain measurement data in real time and adjust the estimation, thereby achieving accurate tracking and adaptation to thermal resistance changes. Furthermore, with the real-time adjustment of thermal resistance, the temperature changes of the windings, oil, and oil tank can be more accurately predicted, thereby achieving precise temperature management. Effective temperature prediction helps to detect abnormal conditions such as transformer overheating early, provide an early warning mechanism, avoid equipment failures, and extend the service life of the transformer.

[0042] Embodiment 2: This embodiment also provides a storage medium, in which computer executable instructions are stored. When the computer executable instructions are loaded and executed by a processor, the steps of the transformer thermal circuit model parameter correction method based on extended Kalman filtering are implemented.

[0043] The specific implementation described above is a preferred implementation of the transformer thermal circuit model parameter correction method based on extended Kalman filtering of the present invention, and is not intended to limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to this specific implementation. All equivalent changes made in accordance with the shape and structure of the present invention are within the protection scope of the present invention.

Claims

1. A transformer thermal circuit model parameter correction method based on extended Kalman filtering, characterized in that: The following steps are involved: S1: Define the state equations that characterize the time-varying dynamic parameters in the transformer thermal circuit model based on the state transfer matrix; S2: Based on the physical relationship between the observed parameters and the dynamic parameters in the transformer thermal circuit model, a nonlinear equation is defined to characterize the nonlinear relationship between the observed parameters and the dynamic parameters; S3: linearizing the nonlinear equation based on extended Kalman filtering to obtain a linear equation; S4: Obtain a first predicted value of the dynamic parameter based on the state equation, obtain a second predicted value of the observed parameter based on the linear equation, and correct the first predicted value of the dynamic parameter based on the second predicted value and the actual value of the observed parameter to obtain a final value of the dynamic parameter.

2. The transformer thermal circuit model parameter correction method based on extended Kalman filtering according to claim 1 is characterized in that: The S1 includes: S11: based on the characteristics of the random distribution variables in the transformer thermal circuit model, the cumulative distribution function of the random distribution variables is obtained, and then the uniform distribution variables are obtained, and the uniform distribution variables are mapped to Gaussian distribution variables by inverse Gaussian; S12: Obtain the state equation based on Gaussian distribution variables and the state transfer matrix.

3. The transformer thermal circuit model parameter correction method based on extended Kalman filtering according to claim 2 is characterized in that: In S12, the state equation is: ； In the formula, x k+1 represents the value of the dynamic parameter at time k+1, A represents the state transfer matrix, x k represents the value of the dynamic parameter at time k, w k Represents the Gaussian distribution variable at time k.

4. The transformer thermal circuit model parameter correction method based on extended Kalman filtering according to claim 1 is characterized in that: In S2, the observed parameters include at least winding temperature, top oil temperature, oil tank temperature and ambient temperature; The dynamic parameters at least include the thermal resistance between the winding and the oil, the thermal resistance between the oil and the oil tank, and the thermal resistance between the oil tank and the ring.

5. The transformer thermal circuit model parameter correction method based on extended Kalman filtering according to claim 4 is characterized in that: In S2, the nonlinear relationship includes: ； ； ； In the formula, is the winding temperature, is the transformer load loss, is the winding heat capacity, is the top oil temperature, is the time step, is the thermal resistance of the winding to the oil, is the heat capacity of oil, is the no-load loss, is the fuel tank temperature, is the thermal resistance between the oil and the tank, is the heat capacity of the fuel tank, is the ambient temperature, is the thermal resistance between the tank rings.

6. The transformer thermal circuit model parameter correction method based on extended Kalman filtering according to claim 1 is characterized in that: In S2, the nonlinear equation is: ； In the formula, z k is the value of the observed parameter at time k, h represents the nonlinear relationship function between the dynamic parameter and the observed parameter, v k represents the observed disturbance variable at time k.

7. The transformer thermal circuit model parameter correction method based on extended Kalman filtering according to claim 1 is characterized in that: The S3 includes: S31: Deriving the nonlinear equation to obtain an initial linear equation, inputting historical data corresponding to the observed parameters and the dynamic parameters into the initial linear equation to obtain a linearization degree of the initial linear equation; S32: When the linearization degree meets the preset condition, the initial linear equation is a linear equation; otherwise, the initial linear equation is regarded as a nonlinear equation and S31 is executed.

8. The method for correcting parameters of a transformer thermal circuit model based on extended Kalman filtering according to claim 1, characterized in that: In S4, the step of correcting the first predicted value of the dynamic parameter based on the second predicted value and the actual value of the observed parameter to obtain the final value of the dynamic parameter includes: The inverse matrix to be obtained is obtained through the measurement matrix corresponding to the linear equation, and the Kalman gain is obtained based on the inverse matrix to be obtained; The state estimation value is obtained based on the Kalman gain, the second predicted value and the actual value, and the state covariance matrix is ​​obtained. When the state covariance matrix meets the preset requirements, the first predicted value is the final value. Otherwise, the state estimation value is corrected based on the state covariance matrix to obtain the final value.

9. The method for correcting parameters of a transformer thermal circuit model based on extended Kalman filtering according to claim 8, characterized in that: Before obtaining the Kalman gain based on the inverse matrix to be found, it also includes: The desingularization parameters are obtained through the cross-validation method, and the inverse matrix to be determined is updated based on the desingularization parameters.

10. A storage medium, characterized in that: The storage medium stores computer executable instructions, and when the computer executable instructions are loaded and executed by the processor, the steps of the transformer thermal circuit model parameter correction method based on extended Kalman filtering as described in any one of claims 1 to 9 are implemented.

Images