An online prediction method, device, storage medium and fault diagnosis method for the grounding current of a transformer core

Abstract

The present invention discloses an online prediction method, device, storage medium and fault diagnosis method for the grounding current of a transformer core, including: collecting real-time data of characteristic quantities affecting the grounding current of the transformer core and performing preprocessing; inputting the preprocessed real-time data of characteristic quantities into a pre-constructed and trained SSA-optimized BP neural network model to obtain an online prediction result of the grounding current of the transformer core; the training method of the SSA-optimized BP neural network model includes: collecting historical data of characteristic quantities affecting the grounding current of the transformer core; building a BP neural network model according to the dimension of the input matrix, training the built BP neural network model based on the historical data of characteristic quantities, and optimizing the parameters of the BP neural network model by using the SSA algorithm during the training process to obtain the SSA-optimized BP neural network model. The present invention can predict future data of the grounding current of the transformer core so as to prevent and control the occurrence of multi-point grounding faults of the transformer core in a timely manner.

Description

Technical Field

[0001] The present invention relates to the technical field of power transformers, and particularly relates to an online prediction method and device for the grounding current of a transformer core, a storage medium, and a fault diagnosis method. Background Art

[0002] A transformer is an essential device for maintaining the stable operation of a power system among grid equipment. As an important part of the normal operation of the power system, the normal operation of the transformer determines whether the power system can supply power stably, safely, and reliably. As the core part of the power transformation equipment, the internal iron core of the transformer must be reliably grounded at one point; however, when the iron core of the transformer forms multiple-point grounding due to floating dust or contact with other metal structures, short circuits will occur, causing faults such as heat and discharge in the transformer core, seriously threatening the safety of the power grid.

[0003] With the development of online monitoring technology, the grounding current of the transformer core can also be monitored online, and when a fault occurs in the transformer core, its grounding current will change significantly. Attention should be paid when the grounding current exceeds the normal range. However, due to many influencing factors of the grounding current of the transformer core, as well as the limitations of its own complexity and non-linear factors, the existing online monitoring system can only detect the real-time data of the grounding current operation of the transformer core, and it is difficult to predict the future data of the grounding current of the iron core by general statistical methods. Therefore, potential faults of the equipment cannot be predicted, and guiding information such as fault diagnosis and corresponding maintenance, repair, and emergency treatment measures cannot be well provided. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and provide an online prediction method and device for the grounding current of a transformer core, a storage medium, which can predict the future data of the grounding current of the transformer core, and also provide a fault diagnosis method and device for the transformer core to prevent and control the occurrence of multi-point grounding faults of the transformer core in a timely manner.

[0005] To achieve the above purpose, the present invention is implemented by the following technical solutions:

[0006] In a first aspect, the present invention provides an online prediction method for the grounding current of a transformer core, and the method includes:

[0007] Collect real-time data of characteristic quantities affecting the grounding current of the transformer core and perform preprocessing;

[0008] Input the preprocessed real-time data of characteristic quantities into a pre-constructed and trained BP neural network model to obtain an online prediction result of the grounding current of the transformer core;

[0009] Among them, the dimension of the input matrix of the BP neural network model is determined according to the characteristic quantities with a correlation degree higher than the set threshold that affect the grounding current of the transformer core; the BP neural network model is trained by the historical data of the characteristic quantities that affect the grounding current of the transformer core and optimized by the Sparrow Search Algorithm (SSA).

[0010] Combined with the first aspect, further, the correlation degree is calculated by using the grey relational analysis method.

[0011] Combined with the first aspect, further, it also includes: updating the weights of the BP neural network model according to the real-time data of the input characteristic quantities and the online prediction results of the corresponding grounding current of the transformer core.

[0012] Combined with the first aspect, preferably, the characteristic quantities that affect the grounding current of the transformer core include the grounding current data of the core at the previous moment, the micro water content in the transformer oil, the H2 concentration, the C2H2 concentration, the CH4 concentration, the C2H4 concentration, the C2H6 concentration, the carbon monoxide and carbon dioxide concentrations, the winding temperature, and the oil temperature in the oil tank.

[0013] Combined with the first aspect, further, the preprocessing includes: filling the error data and missing data in the real-time data of the characteristic quantities.

[0014] Combined with the first aspect, further, the method for filling the error data and missing data in the real-time data of the characteristic quantities includes: using the average value of the predicted value of the grounding current data of the core at this moment by the BP prediction model and the real value or filled value of the grounding current data obtained at the previous moment as the filled value at this moment for filling.

[0015] Combined with the first aspect, further, before training the BP neural network model with the historical data of the characteristic quantities that affect the grounding current of the transformer core, the K-Nearest Neighbor Algorithm (KNN) is used to fill the error data and missing data in the historical data of the characteristic quantities.

[0016] Combined with the first aspect, further, the method for optimizing the BP neural network model by using SSA includes the following steps:

[0017] Initialize the model parameters of the built BP neural network model;

[0018] Construct the input matrix for training the BP neural network model according to the input matrix dimension and the historical data of the characteristic quantities;

[0019] Input the constructed input matrix into the BP neural network model for model training, and calculate the initial fitness value of each sparrow by using SSA, and update and calculate the warning value according to the fitness value;

[0020] Update the positions of the followers and the positions of the sparrows aware of danger based on the fitness value and the warning value;

[0021] Calculate the fitness value after the position update until the current best fitness value and the optimal position are updated and then assigned to the BP neural network model as the weight and threshold respectively;

[0022] Repeat the above steps for model training until the preset model training stop condition is reached, and obtain the BP neural network model optimized by SSA.

[0023] Combined with the first aspect, further, the method for updating the weights of the BP neural network model includes the following steps:

[0024] According to the loaded input matrix, obtain the true value of the iron core grounding current at the current moment t and the predicted value of the iron core grounding current at the previous moment for the current moment Calculate the error e according to formula (1) t as:

[0025]

[0026] Based on the error function e t , calculate the updated weight T(i) update:

[0027]

[0028] In the formula, T(i) represents the weight updated for the i-th time within the number of epochs, and its initial value T(0) is the weight matrix when the BP neural network model was trained and completed at the previous moment; i = 1, 2... N, N is the number of epochs; w ih represents the weight from the input layer to the hidden layer; w ho represents the weight from the hidden layer to the output layer; ∈ is the learning rate; δ is the preset constant value; the cumulant r(i) is the value updated for the i-th time during the online update process, and r(i) is obtained through formula (3):

[0029]

[0030] In the formula, ρ is the attenuation rate; r is the introduced cumulative variable, and the initial value of r is r(0) = 0; r(i - 1) is the value of the (i - 1)-th iteration;

[0031] Repeat the above steps to update the error function e t , until the preset termination condition is reached, and finally obtain the updated weight matrix at the current moment.

[0032] Combined with the first aspect, preferably, the sampling time interval of the historical data of the characteristic quantity affecting the iron core grounding current of the transformer is 30 minutes.

[0033] In a second aspect, the present invention provides an on-line prediction device for the grounding current of a transformer core, characterized in that the device comprises:

[0034] A data preprocessing module, configured to collect real-time data of characteristic quantities affecting the grounding current of the transformer core and perform preprocessing;

[0035] A prediction module, configured to input the preprocessed real-time data of characteristic quantities into a pre-constructed and trained BP neural network model to obtain an on-line prediction result of the grounding current of the transformer core;

[0036] Wherein, the input matrix dimension of the BP neural network model is determined according to the characteristic quantities with a correlation degree higher than a set threshold that affect the grounding current of the transformer core; the BP neural network model is trained by historical data of characteristic quantities affecting the grounding current of the transformer core and optimized by the Sparrow Search Algorithm (SSA).

[0037] In combination with the second aspect, further, the device further comprises:

[0038] A weight update module, configured to update the weights of the BP neural network model according to the input real-time data of characteristic quantities and the corresponding on-line prediction result of the grounding current of the transformer core.

[0039] In combination with the second aspect, further, the weight update module comprises:

[0040] A first calculation unit, configured to obtain the true value of the grounding current of the transformer core at the current moment t according to the loaded input matrix and the predicted value of the grounding current of the transformer core at the current moment at the previous moment Calculate the error e according to formula (4) t as:

[0041]

[0042] A second calculation unit, configured to calculate the updated weight T(i) update based on the error function e t :

[0043]

[0044] In the formula, T(i) represents the weight updated for the i-th time within the number of epochs, and its initial value T(0) is the weight matrix when the BP neural network model was trained at the previous moment; i = 1, 2... N, where N is the number of epochs; w ih represents the weight from the input layer to the hidden layer; w ho represents the weight from the hidden layer to the output layer; ∈ is the learning rate; δ is a preset constant value; the cumulative quantity r(i) is the value updated for the i-th time during the on-line update process, and r(i) is obtained through formula (6):

[0045]

[0046] Wherein, ρ is the attenuation rate; r is the introduced cumulative variable, and the initial value of r, r(0)=0; r(i - 1) is the value of the (i - 1)-th iteration;

[0047] Repeat the above calculation to update the error function e t , until a preset termination condition is reached, and finally obtain the updated weight matrix at the current moment.

[0048] In a third aspect, the present invention provides an on-line prediction device for the grounding current of a transformer core, including a processor and a storage medium;

[0049] The storage medium is used to store instructions;

[0050] The processor is used to operate according to the instructions to execute the steps of the on-line prediction method for the grounding current of the transformer core according to any one of the first aspect.

[0051] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the on-line prediction method for the grounding current of the transformer core according to any one of the first aspect are realized.

[0052] In a fifth aspect, the present invention provides a method for diagnosing transformer core faults, including:

[0053] Predict the grounding current of the transformer core according to the on-line prediction method for the grounding current of the transformer core according to any one of the first aspect;

[0054] If the predicted grounding current of the transformer core is within a preset current range, it is determined that the transformer core has a fault.

[0055] In combination with the fifth aspect, preferably, the preset current range is: 90 - 110 mA.

[0056] In a sixth aspect, the present invention provides a device for diagnosing transformer core faults, characterized in that it includes:

[0057] A current prediction module, which is used to predict the grounding current of the transformer core according to the on-line prediction method for the grounding current of the transformer core according to any one of the first aspect;

[0058] A fault judgment module, which is used to determine that the transformer core has a fault if the grounding current of the transformer core predicted by the current prediction module is within a preset current range.

[0059] In combination with the sixth aspect, preferably, the fault judgment module includes:

[0060] A preset unit for presetting the current range to 90 - 110 mA.

[0061] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0062] The on-line prediction method for the grounding current of the transformer core provided by the present invention predicts the grounding current of the transformer core through a BP neural network model, and uses the SSA algorithm to optimize the parameters of the BP neural network model. Compared with the traditional optimization algorithm, its algorithm is easier to implement and the prediction result is more accurate.

[0063] The grey relational analysis method is used to screen each characteristic quantity affecting the grounding current of the transformer core, and the characteristic quantities with low correlation are eliminated, so as to reduce the dimension of the input matrix of the BP neural network model, reduce overfitting, and effectively improve the prediction accuracy.

[0064] According to the real-time data of the newly input characteristic quantity and the corresponding on-line prediction result of the grounding current of the transformer core, the weights of the BP neural network model are updated, and it is not necessary to calculate the error of the overall data, which effectively saves time and memory while improving the prediction accuracy.

[0065] The transformer core fault diagnosis method provided by the present invention can accurately predict the future data of the grounding current of the transformer core, so that it can better diagnose whether there are potential faults in the equipment according to the preset current range, so as to take corresponding measures as early as possible and improve the operation safety of the equipment. Description of the Drawings

[0066] Figure 1 is a flowchart of an on-line prediction method for the grounding current of a transformer core provided by an embodiment of the present invention;

[0067] Figure 2 is a structural principle block diagram of an on-line prediction device for the grounding current of a transformer core provided by an embodiment of the present invention. Detailed Embodiments

[0068] The technical solutions of the present invention will be described in detail below through the drawings and specific embodiments. It should be understood that the specific features in the embodiments of the present application and the embodiments are detailed descriptions of the technical solutions of the present application, rather than limitations on the technical solutions of the present application. Without conflict, the technical features in the embodiments of the present application and the embodiments can be combined with each other.

[0069] The term "and / or" in this article is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this article generally represents an "or" relationship between the associated objects before and after.

[0070] Embodiment 1:

[0071] As Figure 1 shown, this embodiment introduces an online prediction method for the grounding current of a transformer core, which specifically includes the following steps:

[0072] Step 1: Collect real-time data of characteristic quantities affecting the grounding current of the transformer core and perform preprocessing;

[0073] Step 2: Input the preprocessed real-time data of characteristic quantities into a pre-constructed and trained BP neural network model to obtain the online prediction result of the grounding current of the transformer core.

[0074] As an embodiment of the present invention, the training method of the BP neural network model includes:

[0075] Step 2.1: Collect historical data of characteristic quantities affecting the grounding current of the transformer core;

[0076] Step 2.2: Use the KNN algorithm to fill in the incorrect data and missing data in the historical data of characteristic quantities; specifically, by calculating the distances between the points of the missing data and incorrect data and other data characteristics, select K feature points with the smallest distances, and apply the K data characteristics to perform mean processing at this point, and the obtained result is the filled data;

[0077] Step 2.3: Analyze the relationship of characteristic quantities based on the filled historical data of characteristic quantities, and calculate the correlation degrees of each characteristic quantity affecting the grounding current of the transformer core respectively;

[0078] Step 2.4: Screen out the characteristic quantities with correlation degrees higher than the set threshold to confirm the input matrix dimension of the BP neural network model;

[0079] Step 2.5: Build a BP neural network model according to the input matrix dimension, train the built BP neural network model based on the historical data of characteristic quantities, and use SSA to optimize the parameters of the BP neural network model during the training process to obtain the SSA-optimized BP neural network model.

[0080] In step 2 of the online prediction method for the grounding current of the transformer core provided by the embodiment of the present invention, the weights of the BP neural network model are updated according to the real-time data of the input characteristic quantities and the corresponding online prediction results of the grounding current of the transformer core each time after model training, which specifically includes the following steps:

[0081] According to the loaded input matrix, obtain the true value of the grounding current of the iron core at the current moment t and the predicted value of the grounding current of the iron core at the current moment at the previous moment Calculate the error e according to formula (1) t as:

[0082]

[0083] Based on the error function e t , calculate the updated weight T(i) update:

[0084]

[0085] Wherein, T(i) represents the weight updated for the i-th time within the number of epochs, and its initial value T(0) is the weight matrix when the previous moment of the BP neural network model is trained and completed; i = 1, 2... N, N is the number of epochs; w ih represents the weight from the input layer to the hidden layer; w ho represents the weight from the hidden layer to the output layer; ∈ is the learning rate; δ is a preset constant value; the cumulant r(i) is the value updated for the i-th time during the online update process, and r(i) is obtained through formula (3):

[0086]

[0087] Wherein, ρ is the attenuation rate; r is the introduced cumulative variable, and the initial value of r, r(0)=0; r(i - 1) is the value of the (i - 1)-th iteration; in the embodiment of the present invention, the value of the attenuation rate ρ is 0.9, and the value of the learning rate ∈ is 0.1;

[0088] Repeat the above steps to update the error function e t , until a preset termination condition is reached, and finally obtain the weight matrix updated at the current moment.

[0089] In the embodiment of the present invention, the weight of the BP neural network model is updated according to the real-time data of the newly input feature quantity and the corresponding online prediction result of the transformer core grounding current, without calculating the error for the overall data, saving time and memory and being beneficial to improving the prediction accuracy.

[0090] As an embodiment of the present invention, the feature quantities affecting the transformer core grounding current include the core grounding current data at the previous moment, the micro water content in the transformer oil, the H2 concentration, the C2H2 concentration, the CH4 concentration, the C2H4 concentration, the C2H6 concentration, the carbon monoxide and carbon dioxide concentrations, the winding temperature, and the oil temperature in the oil tank.

[0091] In step 1 of the on-line prediction method for the grounding current of the transformer core provided by the embodiment of the present invention, the preprocessing includes: filling the error data and missing data in the real-time data of the characteristic quantities, and the filling method includes: using the predicted value of the BP prediction model for the grounding current data of the core at this moment and the average value of the real value or the filled value of the grounding current data of the core obtained at the previous moment as the filled value at this moment for filling; the filled value in the embodiment of the present invention is obtained based on the prediction result of the previous moment, so the interference to the prediction accuracy of the next moment is relatively small.

[0092] As an embodiment of the present invention, in the historical data of the characteristic quantities affecting the grounding current of the transformer core, the time interval for collecting data is set to 30 minutes, and it can also be set to other durations according to actual needs.

[0093] As an embodiment of the present invention, the grey relational analysis method is used to screen the characteristic quantities affecting the grounding current of the transformer core, and the characteristic quantities with low correlation are removed, so as to reduce the dimension of the input matrix of the BP neural network model, reduce overfitting, and effectively improve the prediction accuracy; specifically, it includes:

[0094] The correlation coefficients of the characteristic quantities affecting the grounding current of the transformer core are calculated respectively through formula (4):

[0095]

[0096] In the formula: θ ji (k) represents the correlation coefficient between the i-th reference sequence and the k-th sample of the j-th comparison sequence; i = 1, 2...; j = 1, 2...; k = 1, 2...; ρ is the resolution, and its value is between [0, 1];

[0097] Y i (k) is the data of the grounding current of the core in the characteristic quantities affecting the grounding current of the transformer core, that is, the i-th reference sequence; Y j (k) is the data of other characteristic quantities affecting the grounding current of the transformer core, that is, the j-th comparison sequence; minmin|Y i (k)-Y j (k)| and max max|Y i (k)-Y j (k)| respectively represent the minimum value and the maximum value after the numerical difference between the i-th reference sequence matrix and the j-th comparison sequence matrix;

[0098] Combined with the correlation coefficient θ ji (k), the correlation degrees of the characteristic quantities affecting the grounding current of the transformer core are calculated through formula (5):

[0099]

[0100] where: r ij is the correlation degree between the i-th reference sequence and the j-th comparison sequence; n = 1, 2…;

[0101] The characteristic quantities with correlation degrees higher than the set threshold are screened out. In the embodiment of the present invention, the value of the set threshold is 0.7, that is, the characteristic quantities with correlation degrees lower than 0.7 are excluded to obtain the input matrix dimension of the BP neural network model;

[0102] In step 2.5 of the on-line prediction method for the grounding current of the transformer core provided by the embodiment of the present invention, the method for optimizing the parameters of the BP neural network model by using the SSA algorithm includes the following steps:

[0103] Step a: Initialize the model parameters of the built BP neural network model;

[0104] Step b: Construct an input matrix for training the BP neural network model according to the input matrix dimension and the historical data of the characteristic quantities;

[0105] Step c: Input the constructed input matrix into the BP neural network model for model training, and calculate the initial fitness value of each sparrow by using the SSA algorithm, and update and calculate the warning value according to the fitness value;

[0106] Step d: Update the positions of the followers and the positions of the sparrows aware of danger based on the fitness value and the warning value;

[0107] Step e: Calculate the fitness value after position update until the current best fitness value and the optimal position are updated and then assigned to the BP neural network model as the weight and the threshold respectively;

[0108] Step f: Repeat the above steps for model training until the preset model training stop condition is reached to obtain the SSA-optimized BP neural network model.

[0109] It should be noted that compared with traditional optimization algorithms such as the particle swarm optimization algorithm PSO, the sparrow search algorithm SSA has fewer parameters and stronger local search ability. In terms of the performance of benchmark functions, it has stronger stability and convergence speed; by optimizing the parameters of the BP neural network model through SSA, its algorithm is easier to implement and the prediction results are more accurate.

[0110] The on-line prediction method for the grounding current of the transformer core provided by the embodiment of the present invention can accurately predict the future data of the grounding current of the transformer core, provide a reference for maintenance personnel for the fault diagnosis of the transformer core, so as to prevent and control the occurrence of multi-point grounding faults in time, improve the operation reliability of the transformer, extend the overhaul period and the service life of the transformer, and has irreplaceable advantages and extremely broad application prospects compared with traditional means.

[0111] Example 2:

[0112] like Figure 2 As shown, the embodiment of the present invention provides an online prediction device for transformer core grounding current, which can be used to implement the method described in the first embodiment, specifically including:

[0113] Data preprocessing module, used to collect real-time data of characteristic quantities affecting transformer core grounding current and perform preprocessing;

[0114] The prediction module inputs the pre-processed real-time feature data into the pre-built and trained BP neural network model, obtains the online prediction results of the transformer core grounding current, and outputs the predicted value at the next moment;

[0115] Among them, the input matrix dimension of the BP neural network model is determined based on the characteristic quantities that affect the transformer core grounding current and have a correlation degree higher than a set threshold; the BP neural network model is trained by historical data of the characteristic quantities that affect the transformer core grounding current and is optimized by the sparrow search algorithm SSA.

[0116] As an embodiment of the present invention, the device further includes:

[0117] The weight updating module is used to update the weight of the BP neural network model according to the input real-time data of the characteristic quantity and the corresponding online prediction results of the transformer core grounding current.

[0118] Specifically, the weight updating module includes:

[0119] The first calculation unit is used to obtain the true value of the core grounding current at the current time t according to the loaded input matrix And the predicted value of the core grounding current at the previous moment Calculate the error e according to formula (6) t for:

[0120]

[0121] The second calculation unit is used to calculate the error function e t , calculate the updated weight T(i) update:

[0122]

[0123] Where T(i) represents the weight updated for the i-th time within the epoch number, and its initial value T(0) is the weight matrix when the BP neural network model is trained at the last moment; i = 1, 2…N, N is the epoch number; w ih Represents the weight from the input layer to the hidden layer; w hoRepresent the weights from the hidden layer to the output layer; ∈ is the learning rate; δ is the preset constant value; the cumulant r(i) is the value updated at the i-th update during the online update process, and r(i) is obtained through formula (8):

[0124]

[0125] In the formula, ρ is the attenuation rate; r is the introduced cumulative variable, and the initial value of r, r(0)=0; r(i - 1) is the value at the (i - 1)-th iteration;

[0126] Repeat the above calculation to update the error function e t until the preset termination condition is reached, and finally obtain the updated weight matrix at the current moment.

[0127] The on-line prediction device for the grounding current of the transformer core provided by the embodiment of the present invention and the on-line prediction method for the grounding current of the transformer core provided by Embodiment 1 are based on the same technical concept, can produce the beneficial effects as described in Embodiment 1, and the content not described in detail in this embodiment can be referred to Embodiment 1.

[0128] Embodiment 3:

[0129] The embodiment of the present invention provides an on-line prediction device for the grounding current of a transformer core, including a processor and a storage medium;

[0130] The storage medium is used to store instructions;

[0131] The processor is used to operate according to the instructions to execute the steps of any one of the methods according to Embodiment 1.

[0132] Embodiment 4:

[0133] The embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, characterized in that when the computer program is executed by a processor, the steps of any one of the methods according to Embodiment 1 are implemented.

[0134] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0135] Embodiment 5:

[0136] The embodiment of the present invention provides a method for diagnosing transformer core faults, and the method includes:

[0137] Predict the grounding current of the transformer core according to the on-line prediction method of the grounding current of the transformer core as described in any one of Embodiment 1.

[0138] If the predicted grounding current of the transformer core is within the preset current range, it is determined that a fault has occurred in the transformer core; the preset current range is: 90 - 110 mA.

[0139] The transformer core fault diagnosis method provided by the embodiment of the present invention can enable the monitoring personnel to fully judge whether there is a fault in the core according to the prediction result of the grounding current of the transformer core, and can timely give corresponding guiding information such as maintenance, repair, and emergency treatment measures for transformers with potential faults, ensuring the safe operation of the equipment.

[0140] Embodiment 6:

[0141] The embodiment of the present invention provides a transformer core fault diagnosis device, and the device includes:

[0142] A current prediction module, configured to predict the grounding current of the transformer core according to the on-line prediction method of the grounding current of the transformer core as described in any one of Embodiment 1.

[0143] A fault judgment module, configured to determine that a fault has occurred in the transformer core if the grounding current of the transformer core predicted by the current prediction module is within the preset current range.

[0144] As an embodiment of the present invention, the fault judgment module includes: a preset unit, configured to preset the current range as: 90 - 110 mA.

[0145] The on-line prediction device for the grounding current of the transformer core provided by the embodiment of the present invention and the on-line prediction method for the grounding current of the transformer core provided by Embodiment 5 are based on the same technical concept, can produce the beneficial effects as described in Embodiment 5, and the content not described in detail in this embodiment can be referred to Embodiment 5.

[0146] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate for implementing in the process Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks to specify the function of the device.

[0147] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to operate in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction means that implements the functions specified in one or more processes and / or blocks Figure 1 of the one or more processes and / or blocks Figure 1 specified.

[0148] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one or more processes and / or blocks Figure 1 of the one or more processes and / or blocks Figure 1 specified.

[0149] The foregoing is only a preferred embodiment of the present invention, and it should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. An on-line prediction method for the grounding current of a transformer core, characterized in that, The method includes: Collecting real-time data of the characteristic quantities affecting the grounding current of the transformer core and performing preprocessing; Inputting the preprocessed real-time data of the characteristic quantities into a pre-constructed and trained BP neural network model to obtain the online prediction result of the grounding current of the transformer core; Among them, the dimension of the input matrix of the BP neural network model is determined according to the characteristic quantities with a correlation degree higher than the set threshold that affect the grounding current of the transformer core; the BP neural network model is trained by the historical data of the characteristic quantities affecting the grounding current of the transformer core and optimized by the Sparrow Search Algorithm (SSA); It also includes: updating the weights of the BP neural network model, specifically including: Obtain the true value of the iron core grounding current at the current moment \(t\) according to the loaded input matrix and the predicted value of the iron core grounding current at the previous moment for the current moment , calculate the error according to formula (1) as follows: （1） Based on the error function , calculate the updated weights Update: （2） Wherein, represents the weight updated for the i-th time within the number of epochs, and its initial value is the weight matrix when the previous moment of the BP neural network model is trained and completed; i = 1, 2... N, where N is the number of epochs; represents the weight from the input layer to the hidden layer; represents the weight from the hidden layer to the output layer; is the learning rate; is the preset constant value; the cumulant is the value updated for the i-th time during the online update process, which is obtained through formula (3): （3） In the formula, is the attenuation rate; is the introduced cumulative variable, the initial value of ; is the value of the (i - 1)-th iteration; Repeat the above steps to update the error function , until a preset termination condition is reached, and finally obtain the weight matrix updated at the current moment.

2. The on-line prediction method for the grounding current of a transformer core according to claim 1, characterized in that The correlation degree is calculated by using the grey relational analysis method.

3. The on-line prediction method for the grounding current of a transformer core according to claim 1, wherein The characteristic quantities affecting the grounding current of the transformer core include the grounding current data of the core at the previous moment, the micro water content in the transformer oil, the H2 concentration, the C2H2 concentration, the CH4 concentration, the C2H4 concentration, the C2H6 concentration, the carbon monoxide and carbon dioxide concentrations, the winding temperature, and the oil temperature in the oil tank.

4. The on-line prediction method for the grounding current of a transformer core according to claim 3, characterized in that, The preprocessing includes: filling the error data and missing data in the real-time data of the characteristic quantities.

5. The on-line prediction method for the grounding current of a transformer core according to claim 4, characterized in that The method for filling the error data and missing data in the real-time data of the characteristic quantities includes: using the average value of the predicted value of the grounding current data of the core at this moment by the BP prediction model and the true value or filled value of the grounding current data obtained at the previous moment as the filled value at this moment for filling.

6. The on-line prediction method for the grounding current of a transformer core according to claim 1, wherein Before training the BP neural network model with the historical data of the characteristic quantities affecting the grounding current of the transformer core, the K-Nearest Neighbor Algorithm (KNN) is used to fill the error data and missing data in the historical data of the characteristic quantities.

7. The on-line prediction method for the grounding current of a transformer core according to any one of claims 1 to 6, characterized in that, The method for optimizing the BP neural network model by using SSA includes the following steps: Initializing the model parameters of the built BP neural network model; Constructing the input matrix for training the BP neural network model according to the input matrix dimension and the historical data of the characteristic quantities; Inputting the constructed input matrix into the BP neural network model for model training, and calculating the initial fitness value of each sparrow by SSA, and updating and calculating the warning value according to the fitness value; Updating the positions of the followers and the positions of the sparrows aware of danger based on the fitness value and the warning value; Calculating the fitness value after the position update until the current best fitness value and the optimal position are updated and then assigned to the BP neural network model as the weight and threshold respectively; Repeating the above steps for model training until the preset model training stop condition is reached to obtain the BP neural network model optimized by SSA.

8. The on-line prediction method for the grounding current of a transformer core according to claim 1, characterized in that The sampling time interval of the historical data of the characteristic quantities affecting the grounding current of the transformer core is 30 minutes.

9. An on-line prediction device for the grounding current of a transformer core, characterized in that, The device includes: A data preprocessing module, which is used to collect real-time data of the characteristic quantities affecting the grounding current of the transformer core and perform preprocessing; A prediction module, which is used to input the preprocessed real-time data of the characteristic quantities into a pre-constructed and trained BP neural network model to obtain the online prediction result of the grounding current of the transformer core; Among them, the dimension of the input matrix of the BP neural network model is determined according to the characteristic quantities with a correlation degree higher than the set threshold that affect the grounding current of the transformer core; the BP neural network model is trained by the historical data of the characteristic quantities that affect the grounding current of the transformer core and optimized by the Sparrow Search Algorithm (SSA). It further includes: A weight update module, configured to update the weights of the BP neural network model according to the real-time data of the input characteristic quantities and the corresponding online prediction results of the grounding current of the transformer core. The weight update module includes: The first calculation unit is configured to obtain the true value of the iron core grounding current at the current moment t according to the loaded input matrix and the predicted value of the iron core grounding current at the previous moment for the current moment , and calculate the error according to formula (4) as follows: （4） A second computing unit for calculating updated weights based on an error function and Update: （5） Wherein, represents the weight updated for the i-th time within the number of epochs, and its initial value is the weight matrix when the previous moment's training of the BP neural network model is completed; i = 1, 2... N, where N is the number of epochs; represents the weight from the input layer to the hidden layer; represents the weight from the hidden layer to the output layer; is the learning rate; is the preset constant value; the cumulant is the value updated for the i-th time during the online update process, which is obtained through formula (6): （6） In the formula, is the attenuation rate; is the introduced cumulative variable, the initial value of ; is the value of the (i - 1)-th iteration; Repeat the above calculation to update the error function , until a preset termination condition is reached, and finally obtain the weight matrix updated at the current moment.

10. An on-line prediction device for the grounding current of a transformer core, characterized in that, It includes a processor and a storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the online prediction method for the grounding current of the transformer core according to any one of claims 1 to 8.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the online prediction method for the grounding current of the transformer core according to any one of claims 1 to 8.

12. A method for diagnosing transformer core faults, characterized in that, It includes: Predict the grounding current of the transformer core according to the online prediction method for the grounding current of the transformer core according to any one of claims 1 to 8; If the predicted grounding current of the transformer core is within the preset current range, it is determined that the transformer core has a fault.

13. The transformer core fault diagnosis method according to claim 12, wherein The preset current range is: 90~110 mA.

14. A transformer core fault diagnosis device, characterized in that, It includes: A current prediction module, configured to predict the grounding current of the transformer core according to the online prediction method for the grounding current of the transformer core according to any one of claims 1 to 8; A fault judgment module, configured to determine that the transformer core has a fault if the grounding current of the transformer core predicted by the current prediction module is within the preset current range.

15. The transformer core fault diagnosis device according to claim 14, wherein The fault judgment module includes: A preset unit, configured to preset the current range as: 90~110 mA.

Images