Method and system for monitoring and suppressing DC bias current of power transformer based on magnetoelectric coupling sensor

Abstract

The present invention relates to the technical field of power equipment status monitoring, and specifically relates to a method and system for monitoring and suppressing DC bias current of a power transformer based on a magnetoelectric coupling sensor. The method includes: installing magnetoelectric coupling sensors at the monitoring nodes of each power transformer to capture the magnetic field signals during the operation of the power transformer in real time; using an improved information acquisition optimization-based successive variational mode decomposition method to adaptively decompose the preprocessed current signal into a number of intrinsic mode functions, and extracting the DC bias current data of the power transformer from them. A deep recurrent neural network model is used to capture the long-term dependencies provided by historical data and the short-term dynamic changes provided by real-time data to complete the prediction of the DC bias current data at the next moment. The present invention realizes the accurate detection of the DC bias current of the power transformer, effectively suppresses the DC bias component, and improves the ability of the power system to operate safely and stably.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment condition monitoring, and particularly to a method and system for monitoring and suppressing DC bias current of a power transformer based on a magnetoelectric coupling sensor. Background Art

[0002] Under the background of accelerating the construction of a new power system and the grid connection of a high proportion of new energy, as a core device of the power grid, the accurate monitoring of the operating state of a power transformer is crucial for ensuring the safe and stable operation of the system. The bias current is an important parameter of a power transformer. By real-time monitoring the bias current, the abnormal state of the transformer can be detected in time and corresponding measures can be taken to ensure the safe and reliable operation of power equipment.

[0003] Traditional methods for detecting bias current mainly rely on current transformers or Hall sensors for signal acquisition. However, the measurement accuracy of current transformers is poor and they cannot be used for DC detection; Hall sensors are greatly affected by temperature and have low sensitivity. For example, the utility model patent with the publication number CN205450103U uses a current transformer and the invention patent with the publication number CN114678962B uses a Hall sensor. However, the current transformer can only measure AC current, and the accuracy of the Hall sensor is easily affected by temperature drift, and its ability to cope with strong electromagnetic interference is limited and its long-term stability is insufficient.

[0004] Moreover, most of the sensors used in the current on-line monitoring of transformer cores are active sensors, and active sensors have the following disadvantages when measuring the current of a transformer core:

[0005] (1) Dependence on external power supply. Active sensors require an additional power supply to support the operation of their internal circuits, which increases the system complexity and potential failure points. Especially when working in harsh environments such as high voltage or outdoors for a long time, the reliability is affected;

[0006] (2) Sensitive to temperature. The electronic components inside active sensors are easily affected by temperature fluctuations, resulting in measurement drift;

[0007] (3) Insufficient electromagnetic compatibility. Active sensors will be affected by electromagnetic interference under working conditions with strong electromagnetic fields or complex high-frequency harmonics, resulting in signal distortion or measurement errors;

[0008] (4) High cost. The manufacturing cost of active sensors is significantly higher than that of passive sensors, which limits their large-scale application in low-cost monitoring scenarios, and the accuracy of the monitoring and processing methods used in the existing technology is not guaranteed.

[0009] The magnetoelectric coupling sensor, such as the invention patent of JP6131601B2, discloses a magnetic sensor, which discloses the specific principle of converting magnetic flux into electrical signals, and the utility model patent with the publication number of CN218766979U discloses a circuit for signal processing using a magnetoelectric sensor. However, there is still a blank state in the real-time monitoring and processing method of the magnetoelectric coupling sensor combined with the transformer core, and in particular, the specific solution cannot be obtained through the existing technology in terms of monitoring the health status of the transformer core and suppressing the DC bias current. Summary of the Invention

[0010] Object of the Invention: In order to overcome the above-mentioned deficiencies of the prior art, the present invention provides a method for monitoring and suppressing the DC bias current of a power transformer based on a magnetoelectric coupling sensor. This method solves the problems of low accuracy in monitoring the health status of the transformer core caused by DC bias during the operation of the transformer core and the inability to predict and effectively suppress the DC bias in a timely manner. The present invention also provides a system for monitoring and suppressing the DC bias current of a power transformer based on a magnetoelectric coupling sensor.

[0011] Technical Solution: According to the first aspect of the present invention, there is provided a method for monitoring and suppressing the DC bias current of a power transformer based on a magnetoelectric coupling sensor, the method comprising:

[0012] Install magnetoelectric coupling sensors at the monitoring nodes of each power transformer to capture the magnetic field signals during the operation of the power transformer in real time, thereby obtaining the original current signals, and preprocess the collected original current signals;

[0013] Adopt a successive variational mode decomposition method based on improved information acquisition optimization to adaptively decompose the preprocessed current signal into several intrinsic mode functions, and extract the DC bias current data of the power transformer caused by stray current therefrom. The improved information acquisition optimization successive variational mode decomposition method seeks the optimal penalty factor through an improved information acquisition optimization algorithm IAO, and applies the optimal penalty factor to the successive variational mode decomposition method, and analyzes and extracts the direction and amplitude characteristics of the DC bias current according to all the decomposed mode components;

[0014] Combine the historical DC bias current data of the power transformer, the real-time monitored DC bias current data and external variables as the input of the deep recurrent neural network model to capture the long-term dependencies provided by the historical data and the short-term dynamic changes provided by the real-time data, and complete the prediction of the DC bias current data at the next moment. The external variables include: global magnetic field index, grounding electrode current and core temperature.

[0015] Furthermore, the method further comprises:

[0016] According to the augmented objective function of the unconstrained programming set, an improved external penalty function method is adopted to solve the corresponding nonlinear programming problem to find the global optimal solution, and the series resistance of the neutral point of each power transformer is obtained; the augmented objective function of the unconstrained programming is obtained based on the nonlinear programming minimization objective function and penalty term of the bias current suppression parameter;

[0017] According to the obtained series resistance of the neutral point, the DC bias current suppression strategy is adjusted in real time to reduce the DC bias component.

[0018] Furthermore, it includes:

[0019] In the improved information acquisition optimization successive variational mode decomposition method, the optimal penalty factor is sought through the improved information acquisition optimization algorithm IAO, specifically including:

[0020] The Circle chaotic map is used to initialize the population position, the fitness value is calculated according to the objective function, and the current optimal solution is sought according to the current fitness value;

[0021] The exploration stage and exploitation stage of the IAO algorithm are used to optimize the individual positions of the population, and the fitness value is updated during the optimization process, and the current optimal solution is sought according to the calculated fitness value;

[0022] According to the given maximum number of iterations, the loop iteration is performed, and the global optimal solution, that is, the optimal penalty factor, is output.

[0023] Furthermore, it includes:

[0024] The initialization of the population position using the Circle chaotic map is expressed as: ;

[0025] where, represents the chaotic value corresponding to the i th population individual; , represent constant factors; represents the modulo operator, which is used to limit the value range of the variable; represents the th initial position of the population individual; , represent the upper and lower limits of the search space. Furthermore, it includes:

[0026] In the calculation of the fitness value according to the objective function and the search for the current optimal solution according to the current fitness value, the objective function is a multi-objective fitness function combined by the envelope entropy and the energy difference , expressed as: ; wherein, and are weight coefficients, which are adjusted as needed. Further, it includes:

[0027] Applying the optimal penalty factor to the successive variational mode decomposition method, and analyzing and extracting the direction and amplitude characteristics of the DC bias current according to all the mode components obtained by decomposition, including:

[0028] Input the preprocessed current signal and set relevant parameters. The relevant parameters include the penalty factor, discrimination accuracy, discrimination accuracy, noise variance, and initialize the mode components, Lagrange multipliers, and center frequencies;

[0029] Iteratively update the parameter mode components, Lagrange multipliers, and center frequencies until the convergence condition is reached, and output the mode components and their center frequencies corresponding to the current iteration;

[0030] Judge whether all the obtained modes meet the constraint conditions. If they do not meet the constraint conditions, re-initialize the mode components, Lagrange multipliers, and center frequencies. Otherwise, the iteration ends, and all the mode components and their center frequencies are output;

[0031] When DC bias occurs in the power transformer, there is a DC component in the excitation current, which also includes fundamental wave, second and higher harmonic components. Among them, the characteristics of the DC component can effectively characterize whether the transformer has DC bias and its severity, and analyze and extract the direction and amplitude characteristics of the DC bias current from all the mode components obtained by decomposition.

[0032] Further, it includes:

[0033] The specific content of the augmented objective function set without constraints is: adding the non-linear programming minimization objective function of the bias current suppression parameter and the penalty function corresponding to the non-linear programming constraint condition of the bias current suppression parameter as the augmented objective function;

[0034] The non-linear programming minimization objective function of the bias current suppression parameter is expressed as: ;

[0035] Wherein, is the decision variable; is the series resistance of the th monitoring node of the power transformer; is the DC bias current of the th monitoring node of the power transformer; is the number of monitoring nodes of the power transformer; and are weight coefficients, which are adjusted as needed;

[0036] The non - linear programming constraint conditions of the bias current suppression parameters are expressed as: ; where, are the upper and lower limit constraints of the DC bias current of each power transformer node;

[0037] The augmented objective function is expressed as: ;

[0038] where, is the penalty parameter; . Further, it includes:

[0039] The improved outer penalty function method is adopted to solve the corresponding non - linear programming problem to find the global optimal solution, and the series resistance of the neutral point of each power transformer is obtained, including:

[0040] Step 1: Set the initial penalty parameter , allowable error , any initial point , ;

[0041] Step 2: Taking as the initial point, the gradient descent method is used to calculate the optimal solution of the unconstrained problem of minimizing the augmented objective function ;

[0042] Step 3: If , then end and output the optimal solution of the original problem, that is, the series resistance of the neutral point; otherwise, execute Step 4;

[0043] Step 4: Update the penalty parameter using the penalty parameter update formula, , and execute Step 2;

[0044] The penalty parameter update formula is: ;

[0045] where, is the gain coefficient; is the smoothing term;

[0046] When the constraint violation degree is large, the penalty parameter increases faster, at this time, the influence of the penalty term is increased faster, and the infeasible solution is quickly iterated into a feasible solution; when the constraint violation degree is small, the penalty parameter increases slower, at this time, it is avoided that the penalty parameter increases too fast resulting in difficult solution, and the feasible solution is gradually transitioned to the optimal solution.

[0047] Further, it includes:

[0048] The DC bias current suppression strategy is adjusted in real - time according to the obtained series resistance of the neutral point to reduce the DC bias component, including:

[0049] Connect a small resistor in series with the neutral point of the power transformer. When the DC bias current flows into the transformer, the series resistor limits the magnitude of the DC bias current, and a protection gap is used as a protection device for the small resistor. The expression for the series small resistor is: ;

[0050] In the formula, is the set value of the adjustable resistor at the th power transformer node; is the series resistor at the th power transformer node obtained by solving the non-linear programming problem. On the other hand, the present invention also provides a DC bias current monitoring and suppression system for a power transformer based on a magnetoelectric coupling sensor. The system includes:

[0051] A measurement module for installing magnetoelectric coupling sensors at the monitoring nodes of each power transformer to capture the magnetic field signals during the operation of the power transformer in real time, and then obtain the original current signals;

[0052] A signal processing module for preprocessing the collected original current signals;

[0053] A data processing module for adaptively decomposing the preprocessed current signal into a number of intrinsic mode functions by using the successive variational mode decomposition method based on improved information acquisition optimization, and extracting the DC bias current data of the power transformer caused by stray current therefrom. The improved information acquisition optimization successive variational mode decomposition method seeks the optimal penalty factor through the improved information acquisition optimization algorithm IAO, and applies the optimal penalty factor to the successive variational mode decomposition method, and analyzes and extracts the direction and amplitude characteristics of the DC bias current according to all the decomposed mode components;

[0054] A data prediction module for taking the historical DC bias current data of the power transformer, the real-time monitored DC bias current data and external variables as the input of a deep recurrent neural network model, capturing the long-term dependencies provided by the historical data and the short-term dynamic changes provided by the real-time data, and completing the prediction of the DC bias current data at the next moment. The external variables include: the global magnetic field index, the ground electrode current, and the core temperature.

[0055] Furthermore, the system further includes:

[0056] A data analysis module for solving the corresponding non-linear programming problem by using the improved external penalty function method according to the augmented objective function of the unconstrained programming to find the global optimal solution and obtain the series resistors at the neutral points of each power transformer; the augmented objective function of the unconstrained programming is obtained based on the non-linear programming minimization objective function and the penalty term of the DC bias current suppression parameter;

[0057] A dynamic governance module, which is used to adjust the DC bias current suppression strategy in real time according to the obtained neutral point series resistance, and reduce the DC bias component.

[0058] Finally, the present invention also provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, it implements the method for monitoring and suppressing the DC bias current of a power transformer based on a magnetoelectric coupling sensor as described above.

[0059] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0060] (1) The present invention adopts a composite sensing material combination and an array deployment of sensors for data acquisition sensors, which can improve the detection accuracy and anti-interference ability;

[0061] (2) The present invention uses an improved IAO-SVMD algorithm to perform better modal decomposition on complex data and extract the characteristics of the bias current. The DRNN network model is used to introduce external variables and combine historical data with real-time data to achieve accurate prediction. The nonlinear programming between multiple transformer nodes is used to seek the optimal suppression parameters of the system. At the same time, considering the DC bias current suppression technology, dynamic governance is emphasized. The present invention realizes the accurate detection of the DC bias current of the power transformer, effectively suppresses the DC bias component, and improves the ability of the power system to operate safely and stably. This not only helps to improve the health monitoring of the power transformer and improve the operating state of the power transformer, but also helps to provide data support for the fault warning and emergency response of the power system, thereby enhancing the security and reliability of the power grid.

[0062] (3) The present invention uses the Circle chaotic mapping to initialize the population position, replacing the random initialization of the population position in the original IAO algorithm. This improvement can effectively cover each area and will not cause repeated initial states. Moreover, when the algorithm is run multiple times, it avoids falling into the same local optimum and improves the optimization efficiency of the algorithm.

[0063] (4) The present invention combines the envelope entropy and the energy difference as the objective function in the IAO algorithm. This can not only measure the complexity of each modal component after signal decomposition, but also ensure that the total energy of each modal component after decomposition is consistent with the original signal, avoiding energy loss. The optimal penalty factor is obtained by improving the IAO algorithm, thereby improving the decomposition effect of the successive variational mode decomposition method SVMD.

[0064] (5) According to the obtained optimal penalty factor, the present invention performs adaptive modal decomposition on the output complex signal through the improved IAO-SVMD algorithm. Since the decomposition result of the successive variational mode decomposition method SVMD is mainly affected by the penalty factor the optimal penalty factor Since it has a great impact on the output accuracy of this algorithm, on this basis, this application adaptively realizes the Intrinsic Mode Function (IMF) decomposition by introducing a constraint criterion without the need to set the decomposition layer number. .

[0065] (6) The present invention adopts a Deep Recurrent Neural Network (DRNN) and introduces external variables, fuses the decomposed real-time monitoring signals and historical operation data, combines the historical operation bias magnetic current data, current real-time monitoring data of the power transformer and external variables as the input of the DRNN model, captures the long-term dependencies provided by historical data and the short-term dynamic changes provided by real-time data, and accurately predicts the DC bias magnetic current data at the next moment. The present invention uses a Deep Recurrent Neural Network (DRNN) to realize the fusion prediction of the historical operation data and real-time monitoring data of the bias magnetic current, and on this basis, solves the optimal bias magnetic current suppression parameters through nonlinear programming from the system perspective to achieve global co-optimization, and then achieves the transformation from passive response to active suppression.

[0066] (7) The present invention adjusts the DC bias magnetic current suppression strategy in real time according to the optimized parameters solved by nonlinear programming, and cuts the DC bias magnetic component in real time. Brief Description of the Drawings

[0067] Figure 1 is the specific flowchart of the method for monitoring and suppressing the DC bias magnetic current of a power transformer based on a magnetoelectric coupling sensor according to Embodiment 1 of the present invention;

[0068] Figure 2 is the parameter optimization flowchart of the improved IAO algorithm according to Embodiment 1 of the present invention;

[0069] Figure 3 is the schematic structural diagram of the system for monitoring and suppressing the DC bias magnetic current of a power transformer based on a magnetoelectric coupling sensor according to Embodiment 2 of the present invention. Detailed Embodiment

[0070] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0071] Embodiment 1

[0072] The present invention provides a method for monitoring and suppressing the DC bias magnetic current of a power transformer based on a magnetoelectric coupling sensor. As Figure 1 shown, the method includes the following steps:

[0073] S1 Install magnetoelectric coupling sensors at the monitoring nodes of each power transformer to capture the magnetic field signals during the operation of the power transformer in real time, and then obtain the original current signals;

[0074] In this embodiment, the magnetoelectric coupling sensors of the measurement module adopt a multi-channel array layout, which are distributed on the surface of the iron core of the power transformer to realize the conversion from magnetic field signals to mechanical signals and then to electrical signals.

[0075] Optionally, the magnetoelectric coupling sensor is composed of magnetostrictive material and piezoelectric material. Under the magnetization of the external magnetic field, the size of the magnetostrictive material changes in one or some directions to generate force, specifically one of Fe-Ga, Terfenol-D, Metglas, and NiFe2O4.

[0076] Optionally, an external force is applied to the piezoelectric material in a specific direction, and the material deforms and generates charges, specifically one of quartz, LiNbO3, BaTiO3, PZT, PVDF, and PMN-PT.

[0077] Optionally, the structure of the magnetoelectric coupling sensor is determined by the composite form of the two materials, specifically one of 0-3 type granular composite materials, 2-2 type laminated composite materials, and 1-3 type columnar composite materials.

[0078] S2 And preprocess the collected original current signals;

[0079] The preprocessing in this embodiment includes: filtering, gain amplification, synchronous phase, and digitization processing of the original current signals to eliminate noise interference and improve the signal-to-noise ratio. Specifically, it includes:

[0080] By injecting a known current into the transformer winding using a standard current source and recording the output current of the magnetoelectric coupling sensor, establish the relationship curve between the two. And according to the relationship curve between the actual current of the transformer winding and the output current of the magnetoelectric coupling sensor, convert the current signal output by the magnetoelectric coupling sensor into the actual current signal of the transformer. On this basis, perform filtering, gain amplification, synchronous phase, and digitization processing on the signal.

[0081] (a) A filter circuit, whose input is the acquisition signal of the magnetoelectric coupling sensor, adopts a low-pass filter with a cut-off frequency set to 2 kHz to eliminate high-frequency noise in the acquisition signal, such as switching noise, etc.;

[0082] (b) A gain amplifier, whose input is the noise-reduced signal passing through the filter circuit, is used to amplify the weak current signal in multiple stages, dynamically adjust the gain according to the signal amplitude to ensure that the input signal of the analog-to-digital converter (ADC) is within the optimal range, and its gain is set to 10 dB - 40 dB;

[0083] (c) A signal synchronization unit, whose input is the amplified signal that has passed through the gain amplifier. It uses a phase-locked detection circuit. By detecting the phase information with the signal source and comparing and analyzing the phase difference between the two signals, it adjusts the signal phase to achieve signal synchronization;

[0084] (d) An analog-to-digital conversion unit, whose input is the synchronized signal that has passed through the signal synchronization unit. It is used to convert the analog signal into a digital signal. It uses the AD7606 chip, which is of the SAR type, has 16-bit synchronous sampling, is powered by a single 5V power supply, and has a throughput rate of 200kSPS for all channels.

[0085] S3 adopts a successive variational mode decomposition method based on improved information acquisition optimization to adaptively decompose the preprocessed current signal into several intrinsic mode functions, and extracts the DC bias current data of the power transformer caused by stray current from them. The successive variational mode decomposition method based on improved information acquisition optimization seeks the optimal penalty factor through the improved information acquisition optimization algorithm IAO, and applies the optimal penalty factor to the successive variational mode decomposition method. The direction and amplitude characteristics of the DC bias current are analyzed and extracted according to all the decomposed mode components.

[0086] Specifically, in this embodiment, a successive variational mode decomposition based on improved information acquisition optimization (improved IAO-SVMD) is adopted. As Figure 2 shown, the optimal penalty factor is sought through the improved IAO algorithm to improve the decomposition effect of SVMD. It adaptively decomposes the output complex signal of the signal processing module into several intrinsic mode functions, and extracts the DC bias current characteristics of the power transformer caused by stray current from them, that is, the direction and amplitude of the DC bias current.

[0087] The above IAO algorithm is the Information Acquisition Optimizer (abbreviated as IAO). It solves the increasingly complex challenges in the field of continuous non-linear optimization. Inspired by human information acquisition behavior, this algorithm consists of three key strategies: information collection, information filtering and evaluation, and information analysis and organization to adapt to different optimization needs. The results show that IAO has strong competitiveness in terms of convergence rate, solution accuracy and stability. In addition, the time comparison analysis experiment shows its high efficiency.

[0088] Sequential Variational Mode Decomposition (SVMD for short) is a method for signal processing and data analysis. It can decompose complex signals into a series of mode functions, and each mode function represents a specific frequency component in the signal. The main goal of SVMD is to extract different frequency components in the signal and reconstruct them into the original signal.

[0089] The basic principle of SVMD is to decompose the signal into multiple mode functions through variational mode decomposition. In each iteration step, SVMD updates the mode function by minimizing the difference between the signal and the mode function. This process will be repeated continuously until convergence. The finally obtained mode functions can be used to reconstruct the original signal.

[0090] Another key feature of SVMD is sequential decomposition. In each iteration step, SVMD extracts a main frequency component from the signal and removes it from the signal. In this way, each iteration step will extract a frequency component in the signal until all frequency components are extracted. This sequential decomposition method can better capture different frequency components in the signal.

[0091] SVMD has a wide range of applications in signal processing and data analysis. It can be used in multiple fields such as denoising, feature extraction, and spectrum analysis. By decomposing the signal into mode functions, SVMD can better understand and describe the frequency characteristics of the signal. This is very important for signal processing and data analysis.

[0092] The data reconstruction of SVMD is the process of recombining the decomposed mode functions into the original signal. By weighted summing each mode function, the reconstructed signal can be obtained. This process can be used to restore the frequency characteristics of the original signal and can be further analyzed and processed as needed.

[0093] Based on the above existing technologies, this application first uses an improved IAO algorithm based on the Circle chaotic map to obtain the optimal penalty factor required in SVMD. The steps include:

[0094] Step 1: Set the objective function of the IAO algorithm as the fitness function and initialize relevant parameters, including population size, number of iterations, etc.;

[0095] Step 2: Initialize the population position using the Circle chaotic map, replacing the random initialization of the population position in the original IAO algorithm. It effectively covers each area and will not cause repeated initial states. And when the algorithm is run multiple times, it avoids falling into the same local optimum and improves the optimization efficiency of the algorithm, expressed as: In the formula, denotes the chaotic value corresponding to the i th individual of the - population; , ; denotes the modulo operator, which is used to limit the value range of variables; denotes the initial position of the th individual of the - population;

[0096] Step 3: Calculate the fitness value according to the objective function, and seek the current optimal solution based on the calculated fitness value;

[0097] Step 4: Optimize the IAO algorithm in two stages, including the exploration stage and the exploitation stage;

[0098] (a) Take the filtering and evaluation of information as the exploration stage of the IAO algorithm, expressed as: In the formula, denotes the current iteration number; denotes the position of the th individual at the th iteration; denotes the updated position of the th individual; is a random number between [0, 1]; denotes the randomly selected individual position; denotes the error caused by subjective factors in the information filtering and evaluation process, which is defined by the following formula: In the formula, denotes the subjective influence factor, which affects the final acquisition and application of information, and is defined by the following formula: In the formula, , , and are random numbers between [0, 1]; denotes the modulo operator; denotes the reliability factor, which describes the ability of the algorithm to self-adjust according to the information quality at different iteration stages to optimize its behavior, and is defined by the following formula: In the formula, denotes the maximum iteration number; Denote the information quality factor, which is defined by the following formula: wherein, is a random number between [0, 1];

[0099] (b)The development stage of the IAO algorithm through the information analysis and organization process is expressed as: wherein, represents the position of the best population individual generated in the previous iteration process; represents the average value of the positions of the best population individuals generated in the previous iteration process; , , and are random numbers between [0, 1]; represents the control factor when analyzing and organizing information, which is defined by the following formula: Step 5: Update the fitness value through the IAO optimization process, and seek the current optimal solution according to the calculated fitness value;

[0100] Step 6: Determine whether the given maximum number of iterations is reached. If not, go to Step 3; otherwise, output the global optimal solution , that is, the best penalty factor.

[0101] Among them, in this embodiment, Step 3 uses a multi-objective fitness function as the objective function, which is expressed as follows:

[0102] Envelope entropy represents the sparsity of the original signal. The smaller the value, the stronger the sparsity of the signal and the less noise. The original signal obtains the envelope signal after Hilbert demodulation , is defined by the following formula: wherein, represents the normalized form of the envelope signal; Energy difference represents the extraction accuracy of the main resonance component. The closer the energy difference is to zero, the better the decomposition effect. Assume is the original signal, is the modal components after decomposition, is defined by the following formula: wherein, is the energy of the original signal; is the sum of the energies of the modal components after decomposition;

[0103] Combine the envelope entropy and the energy difference into a multi-objective fitness function , which can not only measure the complexity of each modal component after signal decomposition, but also ensure that the total energy of each modal component after decomposition is consistent with the original signal, avoiding energy loss. It is defined by the following formula: In the formula, , are weight coefficients, which are adjusted according to needs.

[0104] Secondly, after obtaining the optimal penalty factor, an improved IAO-SVMD is used to adaptively decompose the complex signal output by the signal processing module. The steps are as follows:

[0105] Step 1: Input the preprocessed current signal and set relevant parameters, including the penalty factor , discrimination accuracy , discrimination accuracy , noise variance , etc.;

[0106] Step 2: Initialize the modal components , Lagrange multipliers , center frequencies ;

[0107] Step 3: Update the parameters , , respectively through the following formula; In the formula, represents the noise tolerance; is defined by the following formula: Step 4: Repeat Step 3 until the convergence condition is met, and output the th modal component and its center frequency. The convergence condition is defined by the following formula: Step 5: Determine whether all the obtained modes meet the constraint conditions. If they do not meet the constraint conditions, return to Step 2; otherwise, the iteration ends and all the modal components and their center frequencies are output. The constraint conditions are defined by the following formula: In the formula, represents the sampling length of;

[0108] Step 6: When DC bias occurs in the power transformer, there is a DC component in the excitation current, which also includes fundamental, second - order and higher - order harmonic components. Among them, the characteristics of the DC component can effectively characterize whether the transformer has DC bias and its severity. Analyze all the decomposed modal components and extract the direction and amplitude characteristics of the DC bias current.

[0109] Therefore, in this step, SVMD is used to decompose the complex signal. Regarding the parameter selection of SVMD, the discrimination accuracy 、the discrimination accuracy 、the noise tolerance have relatively limited influence on the decomposition results. Select the commonly used discrimination accuracy 、 values, that is , and set the noise tolerance to 0. The decomposition result of SVMD is mainly affected by the penalty factor . By introducing a constraint criterion, the Intrinsic Mode Function (IMF) decomposition is adaptively realized without the need to set the decomposition layer number , and the penalty factor is optimized and selected by the improved IAO algorithm.

[0110] S4 uses the historical DC bias current data of the power transformer, the real - time monitored DC bias current data and external variables as the input of the deep recurrent neural network model, captures the long - term dependencies provided by the historical data and the short - term dynamic changes provided by the real - time data, and completes the prediction of the DC bias current data at the next moment. The external variables include: the global magnetic field index, the grounding electrode current and the core temperature;

[0111] In this embodiment, a deep recurrent neural network (DRNN) is used to realize the fusion prediction of the historical operation data and real - time monitored data of the bias current, and based on the system perspective, the optimal bias current suppression parameters are solved through nonlinear programming to achieve global co - optimization, and then the transformation from passive response to active suppression is achieved.

[0112] This embodiment does not specifically describe the training and testing processes of the deep recurrent neural network, including but not limited to steps such as dataset collection, pre - processing, and dataset division.

[0113] Adopt a deep recurrent neural network (DRNN) and introduce external variables, fuse the real - time monitored signal and historical operation data obtained by decomposition, and jointly use the historical operation bias current data, current real - time monitored data and external variables of the power transformer as the input of the DRNN model, capture the long - term dependencies provided by the historical data and the short - term dynamic changes provided by the real - time data, and accurately predict the DC bias current data at the next moment.

[0114] Specifically, the DRNN model enhances the characteristics of the bias magnetizing current by introducing external variables to achieve accurate prediction of the bias magnetizing current. The external variables include:

[0115] (a) Global magnetic field index

[0116] Geomagnetic storms generated by the interaction between the geomagnetic field and the solar plasma wind cause changes in the geomagnetic field, generating a potential gradient on the Earth's surface, and low-frequency induced current flows into the transformer winding. The global magnetic field index reflects the DC bias magnetic source caused by geomagnetic storms, and the global magnetic field index is measured through geomagnetic stations.

[0117] (b) Grounding electrode current

[0118] In a high-voltage DC transmission system, when operating in a monopolar ground return mode, the DC current flowing back in the ground flows into the transformer with a neutral point grounded. The grounding electrode current reflects the bias magnetization of the transformer core exacerbated by the monopolar ground return of the high-voltage DC transmission system, and the grounding electrode current is measured through a DC shunt.

[0119] (c) Core temperature

[0120] DC bias causes severe saturation of the transformer core, increases the loss of metal structural components, and causes local overheating. The core temperature reflects the abnormal temperature rise caused by core saturation due to bias magnetization, and the core temperature is measured through a temperature sensor.

[0121] The present invention adopts a deep recurrent neural network (DRNN) and introduces external variables, fuses the real-time monitoring signals and historical operation data obtained by decomposition, jointly uses the historical operation bias magnetizing current data, current real-time monitoring data and external variables of the power transformer as the input of the DRNN model, captures the long-term dependencies provided by historical data and the short-term dynamic changes provided by real-time data, and accurately predicts the DC bias magnetizing current data at the next moment.

[0122] Embodiment 2

[0123] On the basis of Embodiment 1, the present invention also considers measuring from the perspective of multiple power transformers, constrains the output data of the DRNN according to the upper and lower limit conditions of the DC bias magnetizing current, that is, processes the data predicted by the model. Specifically, the objective function considered is the sum of the DC bias magnetizing currents of the monitoring nodes of all power transformers and the total series resistance, and the improved exterior penalty function method is adopted to solve its nonlinear programming problem to find the global optimal solution, that is, the neutral point series resistance, avoiding global conflicts caused by local optimization. Specifically, it includes:

[0124] S5 adopts the improved exterior penalty function method to solve the nonlinear programming problem of the bias magnetizing current suppression parameter;

[0125] Specifically, in this embodiment, the improved exterior penalty function method is used to solve the nonlinear programming problem of the DC bias current suppression parameter. That is, the penalty function is used to handle the constraint conditions, and an adaptive penalty parameter is adopted. According to the constraint violation degree, that is, the penalty function value, the penalty parameter is adjusted to improve the solution efficiency of the algorithm.

[0126] The objective function of the original constrained programming is the combination of the total DC bias current and the total series resistance; the improved exterior penalty function method converts the constrained programming into an unconstrained programming, converts the original constraints into penalty terms and adds them to the original objective function as the augmented objective function. Specifically:

[0127] The minimization objective function of the nonlinear programming of the DC bias current suppression parameter is: In the formula, is the decision variable; is the series resistance of the th monitoring node of the power transformer; is the DC bias current of the th monitoring node of the power transformer, that is, the current data predicted by the above-mentioned deep neural network model; is the number of monitoring nodes of the power transformer; , are weight coefficients, which are adjusted according to needs;

[0128] The constraint conditions of the nonlinear programming of the DC bias current suppression parameter are: In the formula, is the upper and lower limit constraint of the DC bias current of each power transformer node.

[0129] Combining the minimization objective function of the nonlinear programming of the DC bias current suppression parameter with the penalty function corresponding to the constraint conditions of the nonlinear programming of the DC bias current suppression parameter as the objective function, which is expressed as: In the formula, is the penalty parameter; .

[0130] The steps of the improved exterior penalty function method include:

[0131] Step 1: Set the initial penalty parameter , the allowable error , any initial point , ;

[0132] Step 2: Taking as the initial point, using the gradient descent method, calculate the optimal solution of the unconstrained problem ;

[0133] Step 3: If , then end and output the optimal solution of the original problem , that is, a neutral point series resistor; otherwise, execute step 4;

[0134] Step 4: Update the penalty parameter, , and execute step 2.

[0135] The penalty parameter update formula in step 4 is: In the formula, is the gain coefficient, which is adjusted as needed; is the smoothing term to avoid a zero denominator.

[0136] When the constraint violation degree is large, the penalty parameter increases faster, accelerating the influence of the increased penalty term and quickly iterating the infeasible solution into a feasible solution; when the constraint violation degree is small, the penalty parameter increases slower, avoiding the difficulty in solving caused by the too fast increase of the penalty parameter and gradually transitioning the feasible solution to the optimal solution.

[0137] S6 The neutral point series resistor predicted according to the DC bias current constraint programming problem, so as to adjust the DC bias current suppression strategy in real time and reduce the DC bias component.

[0138] According to the neutral point series resistor solved by the above analysis method, the DC bias current suppression strategy is adjusted in real time to reduce the DC bias component. The dynamic governance mainly adopts the method of connecting a small resistor in series at the neutral point of the power transformer. The selected device mainly consists of an adjustable resistor and a protective gap. When the bias current flows into the transformer, the series resistor limits the magnitude of the DC bias current, and the protective gap serves as a protection device for the small resistor. The expression of the series small resistor is: In the formula, is the set value of the adjustable resistor at the th power transformer node; is the series resistor at the th power transformer node solved by nonlinear programming.

[0139] Embodiment 3

[0140] The present invention also provides a power transformer DC bias current monitoring and suppression system based on a magnetoelectric coupling sensor. As Figure 3 shown, the system includes:

[0141] Measurement module: Install the magnetoelectric coupling sensor at the key monitoring nodes of the power transformer to capture the magnetic field signal during the operation of the power transformer in real time;

[0142] The magnetoelectric coupling sensor of the measurement module adopts a multi-channel array layout, which is distributed on the surface of the iron core of the power transformer to realize the conversion from the magnetic field signal to the mechanical signal and then to the electrical signal.

[0143] Optionally, the magnetoelectric coupling sensor is composed of a magnetostrictive material and a piezoelectric material. Under the magnetization of an external magnetic field, the size of the magnetostrictive material changes in one or some directions to generate force, specifically one of Fe-Ga, Terfenol-D, Metglas, and NiFe2O4.

[0144] Optionally, an external force is applied to the piezoelectric material in a specific direction, causing the material to deform and generate charges, specifically one of quartz, LiNbO3, BaTiO3, PZT, PVDF, and PMN-PT.

[0145] Optionally, the structure of the magnetoelectric coupling sensor is determined by the composite form of the two materials, specifically one of 0-3 type granular composite materials, 2-2 type laminated composite materials, and 1-3 type columnar composite materials.

[0146] Signal processing module: Connected to the magnetoelectric coupling sensor, it is used to filter, gain amplify, synchronize the phase, and digitize the original current signal to eliminate noise interference and improve the signal-to-noise ratio of the signal;

[0147] The signal processing module includes:

[0148] By injecting a known current into the transformer winding using a standard current source and recording the output current of the magnetoelectric coupling sensor, a relationship curve between the two is established. And based on the relationship curve between the actual current of the transformer winding and the output current of the magnetoelectric coupling sensor, the current signal output by the magnetoelectric coupling sensor is converted into the actual current signal of the transformer. On this basis, the signal is filtered, gain amplified, synchronized in phase, and digitized.

[0149] (a) Filter circuit, whose input is the acquisition signal of the magnetoelectric coupling sensor. A low-pass filter is used, and its cut-off frequency is set to 2 kHz to eliminate high-frequency noise in the acquisition signal, such as switching noise, etc.;

[0150] (b) Gain amplifier, whose input is the noise-reduced signal passing through the filter circuit. It is used to amplify the weak current signal in multiple stages, dynamically adjust the gain according to the signal amplitude to ensure that the input signal of the analog-to-digital converter (ADC) is within the optimal range, and its gain is set to 10 dB - 40 dB;

[0151] (c) Signal synchronization unit, whose input is the amplified signal passing through the gain amplifier. A phase-locked detection circuit is used. By detecting the phase information with the signal source and comparing and analyzing the phase difference between the two signals, the signal phase is adjusted to achieve signal synchronization;

[0152] (d) A modulus conversion unit with the input being the synchronized signal that has passed through the signal synchronization unit, used for converting analog signals into digital signals. The AD7606 chip is adopted. This chip is of the SAR type, with 16-bit synchronous sampling, powered by a single 5V power supply, and the throughput rate of all channels being 200 kSPS.

[0153] Data processing module: Decompose the output of the signal processing module, extract the intrinsic mode functions of the signal, and seek the characteristics of the bias current of the power transformer;

[0154] The data processing module adopts the improved successive variational mode decomposition based on improved information acquisition optimization (improved IAO-SVMD). By using the improved IAO algorithm to seek the optimal penalty factor to improve the decomposition effect of SVMD, it adaptively decomposes the complex signal output by the signal processing module into several intrinsic mode functions, and extracts the characteristics of the DC bias current of the power transformer caused by stray current from them, that is, the direction and amplitude of the DC bias current.

[0155] The data processing module adopts the improved IAO algorithm based on Circle chaotic mapping to obtain the best penalty factor of SVMD. The steps include:

[0156] Step 1: Set the objective function of the IAO algorithm as the fitness function and initialize relevant parameters, including population size, number of iterations, etc.;

[0157] Step 2: Initialize the population position by using Circle chaotic mapping, replacing the random initialization of the population position in the original IAO algorithm. It effectively covers each area and will not cause the initial state to repeat. And when the algorithm is run multiple times, it avoids falling into the same local optimum and improves the optimization efficiency of the algorithm, expressed as: Among them, represents the chaotic value corresponding to the i -th population individual; , represent constant factors, taking , ; represents the modulo operator, used to limit the value range of variables; represents the initial position of the -th population individual; , represent the upper and lower limits of the search space; Step 3: Calculate the fitness value according to the objective function, and seek the current optimal solution according to the calculated fitness value;

[0158] Step 4: Optimize the IAO algorithm in two stages, including the exploration stage and the exploitation stage;

[0159] (a)The filtering and evaluation of information are regarded as the exploration stage of the IAO algorithm, expressed as: ; In the formula, represents the current iteration number; represents the th iteration number, and is the position of the th individual in the population; represents the updated position of the th individual in the population; is a random number between [0, 1]; represents the random position of an individual in the population; In the formula, represents the subjective influence factor, which affects the final acquisition and application of information, and is defined by the following formula: In the formula, , , and are random numbers between [0, 1]; represents the modulo operator;

[0160] represents the reliability factor, which describes the ability of the algorithm to self-adjust according to information quality at different iteration stages to optimize its behavior, and is defined by the following formula: In the formula, represents the maximum iteration number; represents the information quality factor, which is defined by the following formula: In the formula, is a random number between [0, 1]; (b) The process of information analysis and organization is regarded as the development stage of the IAO algorithm, expressed as: In the formula, represents the position of the best individual in the population generated in the previous iteration; represents the average value of the positions of the best individuals in the population generated in the previous iteration; , , and are random numbers between [0, 1]; represents the control factor in information analysis and organization, and is defined by the following formula: Step 5: Update the fitness value through the IAO optimization process and seek the current optimal solution according to the calculated fitness value;

[0161] Step 6: Determine whether the given maximum number of iterations is reached. If not, go to Step 3; otherwise, output the global optimal solution , that is, the optimal penalty factor.

[0162] The data processing module uses the improved IAO-SVMD to perform adaptive mode decomposition on the complex signal output by the signal processing module. The steps include:

[0163] Step 1: Input the preprocessed current signal and set relevant parameters, including the penalty factor , discrimination accuracy , discrimination accuracy , noise variance , etc.;

[0164] Step 2: Initialize the mode components , Lagrange multipliers , center frequencies ;

[0165] Step 3: Update the parameters respectively through the following formulas , , ; In the formula, represents the noise tolerance; is defined by the following formula: Step 4: Repeat Step 3 until the convergence condition is met, and output the th mode component and its center frequency. The convergence condition is defined by the following formula: ; Step 5: Determine whether all the obtained modes meet the constraint conditions. If not, return to Step 2; otherwise, the iteration ends and all the mode components and their center frequencies are output. The constraint condition is defined by the following formula: In the formula, represents the sampling length of; Step 6: When the power transformer has DC bias, there is a DC component in the exciting current, which also includes fundamental, second, and higher harmonic components. Among them, the characteristics of the DC component can effectively characterize whether the transformer has DC bias and its severity. Analyze all the mode components obtained by decomposition and extract the direction and amplitude characteristics of the DC bias current.

[0166] The data processing module uses a multi-objective fitness function as the objective function, which is expressed as follows:

[0167] Envelope entropy represents the sparsity of the original signal. The smaller the value, the stronger the sparsity of the signal and the less noise. The original signal obtains the envelope signal after Hilbert demodulation , is defined by the following formula: ;

[0168] In the formula, represents the normalized form of the envelope signal;

[0169] Energy difference represents the accuracy of the main resonance component extraction. The closer the energy difference is to zero, the better the decomposition effect. Assume is the original signal, is the modal components after decomposition, is defined by the following formula: In the formula, is the energy of the original signal; is the sum of the energies of the modal components after decomposition;

[0170] Combining the envelope entropy and the energy difference into a multi-objective fitness function , which can not only measure the complexity of each modal component after signal decomposition, but also ensure that the total energy of each modal component after decomposition is consistent with the original signal, avoiding energy loss, is defined by the following formula: In the formula, , are weight coefficients, which are adjusted according to needs.

[0171] The data processing module uses SVMD to decompose complex signals. Regarding the parameter selection of SVMD, the discrimination accuracy , the discrimination accuracy , and the noise tolerance have relatively limited influence on the decomposition result. Select the commonly used discrimination accuracy , values, that is , and set the noise tolerance to 0. The result of SVMD decomposition is mainly affected by the penalty factor . By introducing a constraint criterion, the intrinsic mode function (IMF) decomposition is adaptively realized without the need to set the decomposition layer , and the penalty factor is optimized and selected by the improved IAO algorithm.

[0172] Data prediction module: The deep recurrent neural network (DRNN) is used to realize the fusion prediction of the historical operation data and real-time monitoring data of the bias magnetizing current. Based on the system perspective, the optimal bias magnetizing current suppression parameter is solved by nonlinear programming to achieve global co-optimization, thus realizing the transformation from passive response to active suppression.

[0173] The data prediction module adopts the deep recurrent neural network (DRNN) and introduces external variables to fuse the real-time monitoring signals and historical operation data obtained by decomposition. The historical operation bias magnetizing current data, current real-time monitoring data and external variables of the power transformer are jointly used as the input of the DRNN model to capture the long-term dependencies provided by the historical data and the short-term dynamic changes provided by the real-time data, and accurately predict the DC bias magnetizing current data at the next moment.

[0174] In the data prediction module, the DRNN model enhances the characteristics of the bias magnetizing current by introducing external variables to achieve accurate prediction of the bias magnetizing current. The external variables include:

[0175] (a) Global magnetic field index

[0176] Geomagnetic storms generated by the interaction between the geomagnetic field and the solar plasma wind change the geomagnetic field, generating a potential gradient on the earth's surface, and low-frequency induced current flows into the transformer winding. The global magnetic field index reflects the DC bias source caused by geomagnetic storms, and the global magnetic field index is measured by geomagnetic stations.

[0177] (b) Grounding electrode current

[0178] In a high-voltage DC transmission system, when the monopole ground return operation mode is adopted, the DC current flowing back in the ground flows into the transformer with a neutral point grounded. The grounding electrode current reflects the bias magnetization of the transformer core aggravated by the monopole ground return of the high-voltage DC transmission system, and the grounding electrode current is measured by a DC shunt.

[0179] (c) Core temperature

[0180] DC bias causes serious saturation of the transformer core, increases the loss of metal structural parts, and causes local overheating. The core temperature reflects the abnormal temperature rise caused by core saturation due to bias magnetization, and the core temperature is measured by a temperature sensor.

[0181] Data analysis module: Measured from the perspective of multiple power transformers in the system, the output data of the DRNN is constrained according to the upper and lower limits of the DC bias magnetizing current. Considering that the objective function is the sum of the DC bias magnetizing currents of the monitoring nodes of all power transformers and the total series resistance, the improved external penalty function method is used to solve its nonlinear programming problem to find the global optimal solution, that is, the neutral point series resistance, avoiding global conflicts caused by local optimization.

[0182] The data analysis module uses an improved external penalty function method to solve the nonlinear programming problem of the bias current suppression parameter, that is, uses a penalty function to handle the constraint conditions, and adopts an adaptive penalty parameter. According to the constraint violation degree, that is, the penalty function value, the penalty parameter is adjusted to improve the solution efficiency of the algorithm. Specifically,

[0183] The nonlinear programming minimization objective function of the bias current suppression parameter is: ; where is the decision variable; is the series resistance of the monitoring node of the th power transformer; DC bias current of the monitoring node of the th power transformer; , are the weight coefficients, which are adjusted according to needs.

[0184] The nonlinear programming constraint conditions of the bias current suppression parameter are: where is the upper and lower limit constraint of the DC bias current of each power transformer node. Combining the nonlinear programming minimization objective function of the bias current suppression parameter with the penalty function corresponding to the nonlinear programming constraint conditions of the bias current suppression parameter as the objective function, it is expressed as: where is the penalty parameter; .

[0185] The steps of the improved external penalty function method include:

[0186] Step 1: Set the initial penalty parameter , the allowable error , any initial point , ;

[0187] Step 2: Using as the initial point, adopt the gradient descent method to calculate the optimal solution of the unconstrained problem ;

[0188] Step 3: If , then end and output the optimal solution of the original problem , that is, the neutral point series resistance; otherwise, execute Step 4;

[0189] Step 4: Update the penalty parameter, , and execute Step 2.

[0190] The penalty parameter update formula in Step 4 is: where is the gain factor, adjust as needed; is a smoothing term to avoid the denominator being zero.

[0191] When the constraint violation is large, the penalty parameter increases faster, accelerating the impact of the penalty term and quickly iterating the infeasible solution to a feasible solution; when the constraint violation is small, the penalty parameter increases slower, avoiding the difficulty of solving the problem caused by the penalty parameter increasing too quickly, and gradually transitioning the feasible solution to the optimal solution.

[0192] Dynamic management module: According to the optimization parameters output by the data analysis module, the DC bias current suppression strategy is adjusted in real time to reduce the DC bias component in real time.

[0193] The dynamic management module adjusts the DC bias current suppression strategy in real time according to the neutral point series resistance solved by the data analysis module to reduce the DC bias component. The dynamic management mainly adopts the method of connecting a small resistor in series with the neutral point of the power transformer. The selected device is mainly composed of an adjustable resistor and a protection gap. When the bias current flows into the transformer, the series resistance limits the size of the DC bias current. The protection gap serves as a protection device for the small resistor. The expression of the series small resistor is: In the formula, For the The adjustable resistor setting value of each power transformer node; The first The series resistance of each power transformer node.

[0194] Finally, the present invention also provides a computer-readable storage medium storing a computer program, characterized in that when the computer program is executed by a processor, the above-mentioned method for monitoring and suppressing the DC bias current of a power transformer based on a magnetoelectric coupling sensor is implemented.

[0195] In the present invention, unless otherwise clearly specified and limited, a first feature being "above" or "below" a second feature may mean that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. Moreover, a first feature being "above", "above" or "above" a second feature may mean that the first feature is directly above or obliquely above the second feature, or simply means that the first feature is higher in level than the second feature. A first feature being "below", "below" or "below" a second feature may mean that the first feature is directly below or obliquely below the second feature, or simply means that the first feature is lower in level than the second feature.

[0196] In the description of this specification, the description referring to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0197] Any process or method description represented in a flowchart or described in other ways herein can be understood to represent a module, segment, or portion of code including one or more executable instructions for implementing a specific logical function or process, and the scope of the preferred embodiments of the present invention includes additional implementations, where the functions can be executed in a manner not shown or discussed, including in a substantially simultaneous manner according to the functions involved or in a reverse order, which should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0198] The logic and / or steps represented in a flowchart or described in other ways herein, for example, can be considered as an ordered list of executable instructions for implementing a logical function, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or in connection with these instruction execution systems, apparatus, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection portion having one or more wirings (electronic device), a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or other suitable processing as necessary, and then stored in a computer memory.

[0199] It should be understood that each part of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logic functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0200] Those of ordinary skill in the art can understand that all or part of the steps carried by the method of the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0201] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications to these embodiments once they know the basic creative concept. Therefore, the appended claims are intended to be construed as including the preferred embodiments and all changes and modifications falling within the scope of the present invention.

[0202] Obviously, those skilled in the art can make various changes and modifications to the embodiments of the present invention without departing from the spirit and scope of the embodiments of the present invention. Thus, if these modifications and variations of the embodiments of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these changes and modifications.

Claims

1. A method for monitoring and suppressing DC bias current of a power transformer based on a magnetoelectric coupling sensor, characterized in that, The method includes: Install magnetoelectric coupling sensors at the monitoring nodes of each power transformer to capture the magnetic field signals during the operation of the power transformer in real time, thereby obtaining the original current signals, and preprocessing the collected original current signals; Adopt the successive variational mode decomposition method based on improved information acquisition optimization to adaptively decompose the preprocessed current signal into several intrinsic mode functions, and extract the DC bias current data of the power transformer caused by stray current from them. The successive variational mode decomposition method based on improved information acquisition optimization first seeks the optimal penalty factor through the improved information acquisition optimization algorithm IAO, and applies the optimal penalty factor to the successive variational mode decomposition method. Finally, analyze and extract the direction and amplitude characteristics of the DC bias current according to all the decomposed mode components; Jointly use the historical DC bias current data of the power transformer, the real-time monitored DC bias current data and external variables as the input of the deep recurrent neural network model to capture the long-term dependencies provided by historical data and the short-term dynamic changes provided by real-time data, so as to complete the prediction of the DC bias current data at the next moment. The external variables include: global magnetic field index, grounding electrode current and core temperature; In the process of seeking the optimal penalty factor through the improved information acquisition optimization algorithm IAO in the successive variational mode decomposition method based on improved information acquisition optimization, it specifically includes: Use the Circle chaotic map to initialize the population position, calculate the fitness value according to the objective function, and seek the current optimal solution according to the current fitness value; Use the exploration stage and exploitation stage of the IAO algorithm to optimize the position of the population individuals, update the fitness value during the optimization process, and seek the current optimal solution according to the calculated fitness value; Iteratively loop according to the given maximum number of iterations, and output the global optimal solution, that is, the optimal penalty factor; The initialization of the population position using the Circle chaotic map is expressed as: ； Among them, represents the chaotic value corresponding to the i th individual in the population; , represent constant factors; represents the modulo operator, which is used to limit the value range of variables; represents the th initial position of the individual in the population; , represent the upper and lower limits of the search space.

2. The method for monitoring and suppressing the DC bias current of a power transformer based on a magnetoelectric coupling sensor according to claim 1, wherein The method further includes: According to the augmented objective function without constrained programming set, adopt the improved exterior penalty function method to solve the corresponding nonlinear programming problem to find the global optimal solution and obtain the neutral point series resistance of each power transformer; the augmented objective function without constrained programming is obtained based on the nonlinear programming minimization objective function and penalty term of the bias current suppression parameter; Adjust the DC bias current suppression strategy in real time according to the obtained neutral point series resistance to reduce the DC bias component.

3. The method for monitoring and suppressing the DC bias current of a power transformer based on a magnetoelectric coupling sensor according to claim 1, wherein Calculating the fitness value according to the objective function and seeking the current optimal solution according to the current fitness value, the objective function is the envelope entropy and the energy difference to form a multi-objective fitness function , expressed as: ; In the formula, , are weight coefficients.

4. The method for monitoring and suppressing the DC bias current of a power transformer based on a magnetoelectric coupling sensor according to claim 3, wherein Applying the optimal penalty factor to the successive variational mode decomposition method and analyzing and extracting the direction and amplitude characteristics of the DC bias current according to all the decomposed mode components includes: Input the preprocessed current signal and set relevant parameters. The relevant parameters include the optimal penalty factor, discrimination accuracy, discrimination accuracy, noise variance, and initialize the mode components, Lagrange multipliers, and central frequencies; Iteratively update the parameter mode components, Lagrange multipliers, and central frequencies until the convergence condition is reached, and output the mode components and their central frequencies corresponding to the current iteration; Determine whether all the acquired modes meet the constraints. If not, reinitialize the modal components, Lagrange multipliers, and center frequencies. Otherwise, the iteration ends and all the modal components and their center frequencies are output. When DC bias occurs in the power transformer, there is a DC component in the excitation current, including the fundamental wave, second harmonic components and above. The characteristics of the DC component can effectively characterize whether DC bias occurs in the transformer and its severity. All the modal components obtained by decomposition are analyzed and the direction and amplitude characteristics of the DC bias current are extracted.

5. The method for monitoring and suppressing the DC bias current of a power transformer based on a magnetoelectric coupling sensor according to claim 2, wherein The augmented objective function without constraint programming according to the setting specifically includes: adding the nonlinear programming minimization objective function of the bias current suppression parameter and the penalty function corresponding to the nonlinear programming constraint condition of the bias current suppression parameter as the augmented objective function; The non-linear programming minimization objective function of the bias current suppression parameter is expressed as: ; Among them, is a decision variable; is the series resistance of the th monitoring node of the power transformer; is the DC bias current of the th monitoring node of the power transformer; is the number of monitoring nodes of the power transformer; , are weight coefficients; The non - linear programming constraint conditions of the bias current suppression parameters are expressed as: ; where are the upper and lower limit constraints of the DC bias current of each power transformer node; The augmented objective function is expressed as: ; Among them, is the penalty parameter; .

6. The method for monitoring and suppressing the DC bias current of a power transformer based on a magnetoelectric coupling sensor according to claim 5, characterized in that, The improved external penalty function method is adopted to solve the corresponding nonlinear programming problem to find the global optimal solution and obtain the neutral point series resistance of each power transformer, including: Step 1: Set the initial penalty parameter , allowable error , any initial point , ; Step 2: Starting from as the initial point, use the gradient descent method to calculate the optimal solution of the unconstrained problem that minimizes the augmented objective function ; ; Step 3: If , then end and output the optimal solution to the original problem , that is, the neutral point series resistance; otherwise, execute Step 4; Step 4: Update the penalty parameter using the penalty parameter update formula, , and execute Step 2; The penalty parameter update formula is as follows: ; Among them, is the gain coefficient; is the smoothing term; When the constraint violation is large, the penalty parameter increases faster. At this time, the impact of the penalty term is increased, and the infeasible solution is quickly iterated to a feasible solution. When the constraint violation is small, the penalty parameter increases slower. At this time, the difficulty in solving the problem caused by the rapid increase of the penalty parameter is avoided, and the feasible solution is gradually transitioned to the optimal solution.

7. The method for monitoring and suppressing the DC bias current of a power transformer based on a magnetoelectric coupling sensor according to claim 6, wherein The method of adjusting the DC bias current suppression strategy in real time according to the obtained neutral point series resistance to reduce the DC bias component includes: Connect a small resistor in series with the neutral point of the power transformer. When the bias magnetic current flows into the transformer, the series resistor limits the magnitude of the DC bias magnetic current, and a protection gap is used as the protection device for the small resistor. The expression for the series small resistor is as follows: ; In the formula, is the set value of the adjustable resistor of the th power transformer node; is the series resistance of the th power transformer node for solving the non-linear programming problem.

8. A DC bias current monitoring and suppression system for a power transformer based on a magnetoelectric coupling sensor, characterized in that The system includes: The measurement module is used to install the magnetoelectric coupling sensor in the monitoring node of each power transformer to capture the magnetic field signal of the power transformer in real time during operation, and then obtain the original current signal; A signal processing module, used for preprocessing the collected original current signal; A data processing module is used to use a successive variational modal decomposition method based on improved information acquisition optimization to adaptively decompose the preprocessed current signal into a plurality of intrinsic mode functions, and extract the DC bias current data of the power transformer caused by the stray current, wherein the successive variational modal decomposition method based on improved information acquisition optimization seeks an optimal penalty factor through an improved information acquisition optimization algorithm IAO, and applies the optimal penalty factor to the successive variational modal decomposition method, and extracts the direction and amplitude characteristics of the DC bias current according to all modal components obtained by the decomposition; The data prediction module is used to combine the historical DC bias current data of the power transformer, the real-time monitored DC bias current data and external variables as the input of the deep recurrent neural network model, capture the long-term dependency provided by the historical data and the short-term dynamic changes provided by the real-time data, and complete the prediction of the DC bias current data at the next moment. The external variables include: global magnetic field index, grounding pole current and core temperature; The improved information acquisition optimization successive variational mode decomposition method seeks the optimal penalty factor through an improved information acquisition optimization algorithm IAO, specifically including: The Circle chaotic map is used to initialize the population position, the fitness value is calculated according to the objective function, and the current optimal solution is sought according to the current fitness value; Optimize the positions of population individuals in the exploration and exploitation phases using the IAO algorithm, update the fitness value during the optimization process, and seek the current optimal solution according to the calculated fitness value; Iterate cyclically according to the given maximum number of iterations, and output the global optimal solution, that is, the optimal penalty factor; The initialization of the population position using the Circle chaotic map is expressed as: ； Among them, represents the chaotic value corresponding to the i th individual in the population; , represent constant factors; represents the modulo operator, which is used to limit the value range of variables; represents the initial position of the th individual in the population; , represent the upper and lower limits of the search space.

9. The direct current bias current monitoring and suppression system for a power transformer based on a magnetoelectric coupling sensor according to claim 8, wherein The system also includes: A data analysis module, which is used to solve the corresponding non-linear programming problem by adopting an improved external penalty function method according to the augmented objective function of the unconstrained programming set, so as to find the global optimal solution and obtain the neutral point series resistance of each power transformer; the augmented objective function of the unconstrained programming is obtained based on the non-linear programming minimization objective function and penalty term of the DC bias current suppression parameter; A dynamic governance module, which is used to adjust the DC bias current suppression strategy in real time according to the obtained neutral point series resistance and reduce the DC bias component.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method for monitoring and suppressing the DC bias current of a power transformer based on a magnetoelectric coupling sensor as described in any one of claims 1 to 7.

Images