A primary current boost protection system for large-capacity transformers

Abstract

This invention relates to the field of transformer technology, specifically to a primary current boost protection system for a large-capacity transformer. The system includes: a support vector model (SVM) section for collecting parameters of the large-capacity transformer at multiple time points during operation; a protection section that, in response to received signals, activates a protection mechanism for the large-capacity transformer, including: an overcurrent rapid trip protection section, a negative sequence protection section, and a frequency response protection section; and a performance evaluation section for calculating the performance parameters of the large-capacity transformer based on the adjusted current, negative sequence current, and frequency response. If the performance parameters do not exceed a set threshold, an adjustment command is sent to the SVM section; upon receiving the adjustment command, the SVM section adjusts the intrinsic parameters of the SVM. This invention improves the flexibility and adaptability of primary current boost protection for large-capacity transformers, enhances accuracy and reliability, and achieves automation and intelligence.

Description

Technical Field

[0001] The invention relates to the technical field of transformers, and in particular to a primary current boost protection system for a large-capacity transformer. Background Art

[0002] With the continuous development and expansion of power systems, large-capacity transformers, as one of the core components of power systems, undertake the key tasks of power transmission and distribution. However, in power systems, large-capacity transformers face various potential operating problems and risks, such as overcurrent, negative sequence current, frequency disturbance, etc. In order to ensure the safe and stable operation of the power system, current boost protection is a vital task. Current boost protection aims to monitor current abnormalities and take appropriate measures to prevent equipment from overheating, damage, and even power system collapse. Traditional large-capacity transformer current boost protection usually adopts fixed set value protection methods, which are based on experience and the rated parameters of the equipment to determine when to trigger the protection mechanism. For example, overcurrent protection usually sets a fixed current protection value based on the rated current of the transformer. Once the current exceeds this set value, the protection system will be triggered, cutting off the power supply or taking other protective measures.

[0003] Although these traditional methods can work well in some cases, they have the following problems: Inadaptability of fixed thresholds: Traditional fixed-value protection methods cannot adapt to changes in the operating status of the power system. The load and operating conditions of the power system may change due to seasons, weather or other factors, so the fixed threshold may not accurately reflect the current operating conditions. Because traditional fixed-value protection methods cannot flexibly adjust the threshold, they are prone to false alarms or missed alarms. False alarms can cause unnecessary protection actions, while missed alarms may cause equipment damage or power system instability. Traditional fixed-value protection methods can usually only handle simple overcurrent problems, but cannot effectively identify and respond to complex fault conditions such as negative sequence current, frequency disturbances, etc. In order to use fixed-value protection methods, protection parameters need to be set based on experience. This requires operation and maintenance personnel to have rich experience and expertise, otherwise they may be improperly set. Traditional fixed-value protection methods cannot adjust protection parameters in real time, so they cannot adapt to instantaneous changes in the power system. Traditional current-upgrading protection methods usually lack a comprehensive evaluation mechanism for the performance of large-capacity transformers, and cannot detect problems in time and take measures to repair them. Summary of the invention

[0004] The main purpose of the present invention is to provide a large-capacity transformer primary current boost protection system, which improves the flexibility and adaptability of the large-capacity transformer primary current boost protection, improves the accuracy and reliability, and realizes automation and intelligence.

[0005] In order to solve the above technical problems, the present invention provides

[0006] A primary current boost protection system for a large-capacity transformer, the system comprising: a support vector model part, used to collect parameters of a large-capacity transformer during operation at multiple moments, the parameters at each moment being used as sample points for constructing a support vector model; a label is marked for each sample point, the label value being -1 or 1; when the label is 1, it represents that a protection mechanism needs to be triggered; when the label is -1, it represents that the protection mechanism does not need to be triggered; a support vector machine is trained using these sample points and corresponding labels to obtain a support vector machine model, the support vector machine model determines whether a protection mechanism needs to be triggered based on real-time parameters, and when it is determined that the protection mechanism needs to be triggered, a signal is sent to a protection part; the protection part responds to the received signal to start a protection mechanism for the large-capacity transformer, the protection part comprising: an overcurrent rapid removal protection part, a negative sequence protection part and a frequency response protection part; the overcurrent rapid removal protection part The protection part is used to set a reference current protection value according to the rated current and preset coefficient of the large-capacity transformer, the resistance of the transformer, the ambient temperature, and the influence of external interference, and use the support vector machine model to adjust the reference current protection value to obtain the current setting value. After dynamically adjusting the current setting value, the adjusted current is obtained for overcurrent protection; the negative sequence protection part is used to calculate the negative sequence current of the transformer based on the three-phase current of the large-capacity transformer for negative sequence protection; the frequency response protection part is used to calculate the frequency response according to the reference frequency and the frequency for frequency response protection; the performance evaluation part is used to calculate the performance parameters of the large-capacity transformer according to the adjusted current, negative sequence current and frequency response, and if the performance parameters do not exceed the set threshold, an adjustment instruction is sent to the support vector model part; after receiving the adjustment instruction, the support vector model part adjusts the internal parameters of the support vector model.

[0007] Furthermore, the parameters include: preset coefficient, rated current, resistance, ambient temperature, frequency, reference frequency, angular frequency and three-phase current.

[0008] Furthermore, suppose that the N sample points are (x1, y1), (x2, y2), …, (x N ,y N ), where x i is the parameter at each moment, which is an M-dimensional feature vector, y i The value is 1 or -1, indicating two categories; the goal is to find the parameters w and intercept b of a hyperplane to minimize the following optimization problem:

[0009] minimize

[0010] Constraint w T Φ(x i )+b≥1-ξ i ,i=1,2,…,N;

[0011] ξ i ≥0,i=1,2,…,N

[0012] Specifically, w is the weight vector; b is the intercept; ξ i is a slack variable; C is a regularization parameter used to balance the penalty of maximizing the margin and classification errors; y i is the label of the sample point, which takes the value of 1 or -1, indicating two categories; Φ(x i ) represents the feature vector x i The feature vector is mapped to a high-dimensional space through a kernel function.

[0013] Furthermore, the support vector machine model is expressed using the following formula:

[0014]

[0015] Among them, α i is the Lagrange multiplier;

[0016] K(x i ,x t )=Φ(x t ) T Φ(x i );

[0017] Among them, x t is a real-time parameter; according to the calculated f(x t ) value to determine whether the protection mechanism needs to be triggered.

[0018] Furthermore, the reference current protection value is calculated using the following formula:

[0019]

[0020] Where ω is the angular frequency; |x i | is the feature vector x i The modulus value; R 変 is the resistance; I 额定 is the rated current; K is the set coefficient; p i is the power of the i-th sample point.

[0021] Furthermore, the following formula is used to adjust the reference current protection value using the support vector machine model to obtain the current setting value:

[0022]

[0023] Among them, α is the Gaussian distribution coefficient; x is the Gaussian distribution variable; C is the regularization parameter.

[0024] Furthermore, the current setting value is dynamically adjusted using the following formula to obtain the adjusted current:

[0025]

[0026] Among them, f 基准 is the reference frequency; γ is the first adjustment coefficient, ranging from 0.3 to 0.6; δ is the second adjustment coefficient, ranging from 1.2 to 1.4; f i is the frequency of the i-th sample point; I t The real-time current.

[0027] Furthermore, the negative sequence protection part uses the following formula to calculate the negative sequence current of the transformer:

[0028]

[0029] Among them, n is the serial index of the three-phase current; j is the imaginary number symbol.

[0030] Furthermore, the frequency response protection part calculates the frequency response according to the reference frequency and the frequency using the following formula:

[0031]

[0032] Where Δf is the frequency response; f t The real-time frequency.

[0033] Furthermore, the performance evaluation part calculates the weighted average sum of the difference between the adjusted current and the mean current of the sample point, the difference between the variance of the negative sequence current and the current of the sample point, and the difference between the frequency response and the mean frequency of the sample point according to the set three weighting coefficients, and obtains the performance parameters of the large-capacity transformer; if the performance parameter does not exceed the set threshold, an adjustment instruction is sent to the support vector model part; after receiving the adjustment instruction, the support vector model part multiplies the weight vector of the support vector model by the performance parameter for adjustment.

[0034] A large-capacity transformer primary current boost protection system of the present invention has the following beneficial effects: the support vector machine model combines the features and label information of multiple sample points, and obtains an accurate judgment of the current power system state by calculating complex mathematical relationships. This data-driven method is based on actual observation data and has higher accuracy. Unlike the traditional fixed value protection method, the support vector machine model can dynamically determine whether the protection mechanism needs to be triggered according to the distribution of actual data. This reduces the probability of false alarms and missed alarms, and ensures timely and accurate protection. The support vector machine model comprehensively considers multiple parameters and features when judging whether the protection mechanism needs to be triggered. This includes information on multiple aspects such as current, frequency, and negative sequence current. By integrating this information together, the support vector machine model can more comprehensively evaluate the state of the power system, not only being able to handle overcurrent problems, but also dealing with more complex fault conditions such as negative sequence current and frequency disturbances. This improves the reliability and comprehensiveness of the system. The performance evaluation part of the present invention evaluates the performance state of the large-capacity transformer by calculating the difference and weighted average of multiple performance parameters. If the performance does not reach the set threshold, the system will send an adjustment instruction to the support vector model part to optimize the parameters of the model. This feedback loop mechanism enables the system to continuously optimize itself and adapt to different operating conditions, improving the protection and performance of large-capacity transformers. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A schematic diagram of the system structure of a large-capacity transformer primary current boost protection system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0036] The method of the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments of the present invention.

[0037] Example 1: Reference Figure 1A primary current boost protection system for a large-capacity transformer, the system comprising: a support vector model part, used to collect parameters of a large-capacity transformer during operation at multiple moments, the parameters at each moment being used as sample points for constructing a support vector model; for each sample point, a label is marked, the label value is -1 or 1; when the label is 1, it means that a protection mechanism needs to be triggered; when the label is -1, it means that the protection mechanism does not need to be triggered; a support vector machine is trained using these sample points and corresponding labels to obtain a support vector machine model, the support vector machine model determines whether a protection mechanism needs to be triggered based on real-time parameters, and when it is determined that the protection mechanism needs to be triggered, a signal is sent to the protection part; the protection part responds to the received signal to start the protection mechanism for the large-capacity transformer, the protection part comprises: an overcurrent rapid removal protection part, a negative sequence protection part and a frequency response protection part; the overcurrent rapid removal protection part The protection part is used to set a reference current protection value according to the rated current and preset coefficient of the large-capacity transformer, the resistance of the transformer, the ambient temperature, and the influence of external interference, and use the support vector machine model to adjust the reference current protection value to obtain the current setting value. After dynamically adjusting the current setting value, the adjusted current is obtained for overcurrent protection; the negative sequence protection part is used to calculate the negative sequence current of the transformer based on the three-phase current of the large-capacity transformer for negative sequence protection; the frequency response protection part is used to calculate the frequency response according to the reference frequency and the frequency for frequency response protection; the performance evaluation part is used to calculate the performance parameters of the large-capacity transformer according to the adjusted current, negative sequence current and frequency response, and if the performance parameters do not exceed the set threshold, an adjustment instruction is sent to the support vector model part; after receiving the adjustment instruction, the support vector model part adjusts the internal parameters of the support vector model.

[0038] Specifically, SVM is a supervised learning algorithm used for binary classification problems (in this case, to determine whether the protection mechanism needs to be triggered). The basic idea is to map the sample data into a high-dimensional space and find a hyperplane so that the sample points of different categories have the largest interval in the high-dimensional space, so that the classification is more accurate. The goal of SVM is to find a hyperplane with the largest interval, which refers to the sum of the distances of the sample points of two different categories from the hyperplane. Such a hyperplane can better separate the two categories and has better generalization ability. Support vectors are those sample points closest to the hyperplane, which play a key role in defining the hyperplane. These sample points determine the position and direction of the hyperplane, so they are called support vectors.

[0039] The first task of the support vector model part is to train the SVM model using the collected bulk transformer operating parameters. Each sample point includes a set of parameters and a label indicating whether the protection mechanism needs to be triggered. During the training process, the SVM will find a hyperplane to maximize the interval between sample points and ensure that the sample points are correctly classified. Once the SVM model is trained, it can determine whether the protection mechanism needs to be triggered based on the transformer parameters collected in real time. Specifically, it maps the real-time parameters to the same high-dimensional space as the training data, and then determines whether they are on one side of the hyperplane based on the position of these parameters in the high-dimensional space. The output of the SVM model is a decision function that returns a value based on the input parameters. If the value is greater than a certain threshold, the system will determine that the protection mechanism needs to be triggered, otherwise it does not. This decision is made based on the distribution and interval of the sample points learned by the model during training.

[0040] In the primary current-raising protection system for large-capacity transformers, the primary role of SVM is to perform parameter training. This training process is based on the collected operating parameters of large-capacity transformers at multiple moments, and a label is marked for the parameters at each moment, indicating whether the protection mechanism needs to be triggered. Through training, SVM learns how to make decisions based on these parameters to determine whether the protection needs to be triggered. Once the SVM model is trained, it can make judgments based on the transformer parameters collected in real time. Specifically, SVM maps the real-time parameters into a high-dimensional space, and then determines whether these parameters are on one side of the hyperplane based on their position in the high-dimensional space. If they are on one side of the hyperplane, SVM determines that the protection mechanism needs to be triggered; otherwise, it does not need to be triggered. Unlike traditional fixed-value protection methods, SVM has greater flexibility. It can adaptively adjust protection decisions based on actual conditions instead of relying on static thresholds. This enables the system to protect large-capacity transformers more accurately based on actual changes, improving the reliability and safety of the power system.

[0041] Overcurrent rapid removal protection is based on current measurement. It first considers the rated current and preset coefficient of the large-capacity transformer, as well as factors such as the transformer resistance and ambient temperature. Then, a reference current protection value is set based on these parameters. This reference current protection value is a reference value for detecting overcurrent conditions. Negative sequence current usually refers to the current component under asymmetrical working conditions, which can generate negative sequence voltage and may damage the transformer. The principle of negative sequence protection is to calculate the negative sequence current of the transformer by obtaining the three-phase current of the large-capacity transformer. If the amplitude of the negative sequence current exceeds a certain threshold, the protection action will be triggered. Frequency response protection involves frequency calculation. It uses the reference frequency and the actual measured frequency data to calculate the frequency response. If the frequency deviates from the normal range, the protection action will be triggered. The main function of overcurrent rapid removal protection is to prevent large-capacity transformers from being damaged by overcurrent. It dynamically sets the reference current protection value based on the actual operating parameters and environmental conditions. If the current exceeds this protection value, the system will quickly remove the power supply to protect the transformer from damage. The function of negative sequence protection is to detect negative sequence current under asymmetrical working conditions, which may be caused by faults. If the negative sequence current is abnormally high, it may mean that there is a problem with the transformer, and the triggering protection mechanism can prevent further damage. Frequency response protection is used to detect frequency anomalies in the power system. The deviation of frequency may be caused by power system faults or other problems. The triggering protection mechanism can limit the impact of abnormal frequency on the transformer and ensure the stability of the power system. Their main function is to monitor the operating status of large-capacity transformers and take protective measures in time to ensure the safe and stable operation of the power system. This dynamic adjustment and multiple protection level design can effectively respond to different types of faults and abnormal situations, and improve the reliability of the protection system.

[0042] Embodiment 2: The parameters include: preset coefficient, rated current, resistance, ambient temperature, frequency, reference frequency, angular frequency and three-phase current.

[0043] Specifically, the preset coefficients are used to calculate the reference current protection value. By considering factors such as rated current, resistance, ambient temperature, etc., the preset coefficients can be determined according to the specific transformer design and operation conditions to ensure appropriate overcurrent protection. The rated current is the maximum current value that the transformer can operate safely. It is used to calculate the reference current protection value to ensure that the overcurrent protection is not falsely triggered under normal operating conditions. Resistance is one of the electrical characteristics of the transformer, which is used to calculate the reference current protection value. The larger the resistance, the lower the voltage at the same current, which may cause false overcurrent alarms. The ambient temperature can affect the electrical characteristics of the transformer, especially the value of the resistance. A high temperature environment may cause an increase in resistance, so the temperature factor needs to be considered to calculate the reference current protection value. Frequency is an important parameter of the power system and is crucial to the frequency response protection mechanism. Abnormal frequency deviation may trigger the frequency response protection. The reference frequency is the normal operating frequency of the power system, usually 50Hz or 60Hz. It is used to calculate the reference value of the frequency response protection. The angular frequency is the angular part of the frequency, usually expressed in radians per second (rad / s). It is used to calculate the angular frequency deviation in the frequency response protection. The three-phase current is a key parameter for the operation status of large-capacity transformers and is used to monitor negative-sequence current and real-time current. Negative-sequence current protection calculates the negative-sequence current of the transformer based on the three-phase current.

[0044] Example 3: Assume that the N sample points are (x1, y1), (x2, y2), ..., (x N ,y N ), where x i is the parameter at each moment, which is an M-dimensional feature vector, y i The value is 1 or -1, indicating two categories; the goal is to find the parameters w and intercept b of a hyperplane to minimize the following optimization problem:

[0045] minimize

[0046] Constraint w T Φ(x i )+b≥1-ξ i ,i=1,2,…,N;

[0047] ξ i ≥0,i=1,2,…,N

[0048] Specifically, w is the weight vector; b is the intercept; ξ i is a slack variable; C is a regularization parameter used to balance the penalty of maximizing the margin and classification errors; y i is the label of the sample point, which takes the value of 1 or -1, indicating two categories; Φ(x i ) represents the feature vector x i The feature vector is mapped to a high-dimensional space through a kernel function.

[0049] Specifically, the main goal of SVM is to maximize the interval of the hyperplane. The hyperplane represents the boundary of the current state, and maximizing the interval means ensuring that under normal operating conditions, the current waveform does not touch the vicinity of the hyperplane. This provides additional security for the system to prevent false alarms or misjudgments. The loss function L(y,f(x)) in the formula is responsible for penalizing classification errors. In the current boost protection system, classification errors may lead to unnecessary protection actions on large-capacity transformers, thereby affecting the normal operation of the power system. Through the loss function, SVM forces the model to be more cautious in classification to reduce the possibility of false alarms. Relaxation variable ξ i represents the distance between the current waveform and the hyperplane. These slack variables allow for some fluctuations or noise, but are also controlled by the regularization parameter C. The choice of regularization parameters allows the system to strike a balance between flexibility and precision, ensuring that the protection system can adapt to different current conditions and environments. The main role of SVM is to improve the safety of the protection system. It ensures that protection actions are only triggered when the current is abnormal by maximizing the interval of the hyperplane and the penalty for classification errors. This reduces the possibility of false alarms and reduces unnecessary intervention. SVM is adaptive and can adjust protection decisions based on different current waveforms and parameters. This means that the system can adapt to different working conditions and operating conditions, improving the robustness of the large-capacity transformer protection system. SVM uses the kernel function Φ(x i ) can handle linearly inseparable current waveforms. This is very important for detecting complex fault conditions and abnormal conditions, as the current waveform may exhibit nonlinear characteristics under different operating conditions. The regularization parameters and loss function adjustment of SVM can be optimized according to the specific application scenario to ensure that the performance of the protection system can reach the best state under different conditions. The principle and function of the SVM formula in the primary current boost protection system of large-capacity transformers is to achieve accurate classification and reliable protection of current states by maximizing the classification interval, reducing false alarms, improving safety and adaptability, and handling linear inseparable problems, thereby improving the performance and reliability of the protection system.

[0050] The first part of the optimization problem in formula Indicates minimizing the norm of the weight vector, that is, maximizing the interval of the hyperplane. This interval represents the distance from the sample point to the hyperplane. Maximizing the interval means that the hyperplane can better separate samples of two different categories. The loss function L(y,f(x)) is used to penalize classification errors, that is, when y i and w T Φ(x i)+b have different signs, there will be a loss. This encourages the SVM to find a hyperplane so that the sample points are correctly classified. The relaxed variable ξ in the constraint condition i Some sample points are allowed to be in the misclassified area of ​​the hyperplane, but they will be penalized in the loss function. The regularization parameter C controls the degree of penalty for classification errors. When C is larger, SVM will emphasize classification accuracy more, and when C is smaller, SVM will focus more on maximizing the margin. The feature mapping function Φ(x i ) maps the sample points from the original feature space to a high-dimensional feature space, which can handle linearly inseparable situations. The choice of kernel function can be adjusted according to the characteristics of the problem. Common kernel functions include linear kernels, polynomial kernels, and Gaussian kernels. The main function of SVM is to perform binary classification, that is, to divide the sample points into two different categories. It finds a hyperplane so that the sample points are correctly classified and the classification interval is maximized. The role of the regularization parameter C is to balance the maximization of the interval and the penalty for classification errors. By adjusting the value of C, the classification accuracy and tolerance of SVM can be controlled to adapt it to different application scenarios. The use of kernel functions allows SVM to handle nonlinear problems and map sample points into high-dimensional space, thereby finding a linearly separable hyperplane in the high-dimensional space. The feature mapping function Φ(x i ) can highlight important features in the feature space and help improve classification performance. SVM pursues maximizing the classification interval, which helps to improve the generalization ability of the model and better classify new data. The principle and function of the SVM model represented by the formula in the primary current boost protection system of large-capacity transformers is to find a hyperplane to best separate samples of different categories by maximizing the classification interval, penalizing classification errors, adjusting regularization parameters, and using kernel functions, thereby achieving accurate classification and protection of the transformer status.

[0051] Embodiment 4: The support vector machine model is represented by the following formula:

[0052]

[0053] Among them, α i is the Lagrange multiplier;

[0054] K(x i ,x t )=Φ(x t ) T Φ(x i );

[0055] Among them, x t is a real-time parameter; according to the calculated f(x t ) value to determine whether the protection mechanism needs to be triggered.

[0056] Specifically, the output f(x t ) is a function used to determine whether the protection mechanism needs to be triggered. It is calculated through a combination of a series of parameters and weights. i are the Lagrange multipliers obtained during the SVM training process, which are used to assign weights to sample points. Each training sample has a corresponding Lagrange multiplier, which plays an important role in the SVM model. i ,x t ) represents the sample x i and real-time parameter x t It does this by mapping the two parameters into a high-dimensional feature space and then calculating their inner product in that feature space. This allows the SVM to perform linear classification in a high-dimensional space, even when it is not linearly separable in the original feature space. The intercept term b is a constant in the SVM model that is used to translate and adjust the position of the hyperplane when determining whether to trigger the protection mechanism. The main function of the SVM model is to calculate the value of the hyperplane according to the real-time parameter x. t The value of f(x t ), and determine whether to trigger the protection mechanism based on the calculated value. t ) is greater than a certain threshold, indicating that the current state is abnormal and the protection mechanism needs to be triggered, otherwise it maintains normal operation. i are Lagrange multipliers, which are critical for the calculation and classification decision of SVM. Their values ​​depend on the training data and the choice of kernel function, which is obtained through optimization algorithm. These multipliers determine the weights of support vectors, which play a key role in the SVM model and help determine the location of the hyperplane. The kernel function K(x i ,x t ) allows SVM to classify in high-dimensional feature space, which is the key to SVM in dealing with nonlinear problems. The kernel function can be selected according to the characteristics of the problem to ensure that the SVM can adapt when dealing with different types of current waveforms. The intercept term b can be used to fine-tune the position of the hyperplane to ensure the adaptability of the SVM model to actual conditions. It allows the SVM to find a suitable boundary in high-dimensional space to distinguish between normal and abnormal current states. The formula in the SVM model is used to determine in real time whether it is necessary to trigger the primary current boost protection mechanism of a large-capacity transformer. By considering factors such as Lagrange multipliers, kernel functions, and intercept terms, the SVM model can flexibly classify and protect current states, thereby improving the accuracy and reliability of the large-capacity transformer protection system.

[0057] Embodiment 5: The reference current protection value is calculated using the following formula:

[0058]

[0059] Where ω is the angular frequency; |x i | is the feature vector x i The modulus value; R 変 is the resistance; I 额定 is the rated current; K is the set coefficient; p i is the power of the i-th sample point.

[0060] Embodiment 6: Use the following formula and support vector machine model to adjust the reference current protection value to obtain the current setting value:

[0061]

[0062] Among them, α is the Gaussian distribution coefficient; x is the Gaussian distribution variable; C is the regularization parameter.

[0063] Specifically, setting the coefficient K allows the base current to be adjusted according to system requirements. It acts like a current multiplier. According to the principle of current protection, the rated current is usually the normal operating current of the transformer. By multiplying by the coefficient K, the base current can be adjusted to suit different operating conditions. 変 The exponential term with time T represents the decay of the current in the resistor. This follows from Ohm's law, which relates current to voltage and resistance: The exponential term takes into account the effect of resistance on the current and changes over time. This reflects the transient characteristics of the current. The summation of power and sine function adds the power and sine function values ​​at multiple sample points. Power p i It can represent the energy distribution in the current waveform. i |) may reflect the periodic changes in the current waveform. These terms take into account various characteristics of the current waveform, including oscillations, peaks and valleys. Ohm's law describes the relationship between current, voltage and resistance. According to Ohm's law, the current I is equal to the voltage U divided by the resistance R, that is, Therefore, the decay of current in a resistor can be expressed by an exponential term, namely, e (R×T) . The sine function sin(ωt) describes the changes in periodic signals. In power systems, AC currents usually have periodic characteristics, so the sine function is often used to describe the oscillations of current waveforms. ω represents the angular frequency, which is usually related to the frequency of the power system. The main principle of current protection is to monitor abnormal current conditions to prevent damage to equipment or instability in the power system. The reference current protection value is an important reference value used to determine whether to trigger the protection mechanism. The various terms in the formula take into account different characteristics of the current, including the current magnitude, transient characteristics, and periodic changes, to more comprehensively evaluate the current state.

[0064] Embodiment 7: After dynamically adjusting the current setting value using the following formula, the adjusted current is obtained:

[0065]

[0066] Among them, f 基准 is the reference frequency; γ is the first adjustment coefficient, ranging from 0.3 to 0.6; δ is the second adjustment coefficient, ranging from 1.2 to 1.4; f i is the frequency of the i-th sample point; I t The real-time current.

[0067] Specifically, the current setting value I 设定 Used to set the target value of current protection. Based on this value, the system will dynamically adjust to adapt to the actual current situation. Second adjustment coefficient γ: This is an adjustment coefficient used to adjust the current setting value. Its value range is 1.2 to 1.4 and can be adjusted according to system requirements. A larger γ value will increase the current setting value, while a smaller γ value will reduce the current setting value. Each sample point has an associated frequency value f i Frequency is usually used to indicate the periodicity of the current waveform. These frequency values ​​are used to adjust the current setting value to adapt to the frequency changes of the current waveform. t Represents the real-time current value. It is used to compare with the current setting value to determine whether the current setting value needs to be adjusted. Lagrange multiplier α i :Each sample point has an associated Lagrange multiplier α i These multipliers are used to penalize errors in the current setpoint. Larger α i A value of 0 will increase the penalty for error, requiring the error to be closer to the set value. This is a measure of the error between the actual current and the set value. The error calculation is subject to The effect of the parameters. A smaller error will result in a current setting value that is closer to the set value.

[0068] The first term in the formula The frequency variation of the current waveform is taken into account. This is based on the relationship between frequency and current waveform in the power system. In the power system, the frequency variation may reflect the characteristics of the current waveform under different working conditions. The Lagrange multiplier α is introduced i , which is used to penalize the error between the current setting value and the actual current. This reflects the importance of the error in the control system. By increasing α i The value of can increase the sensitivity to errors. $ measures the error between the real-time current and the set value. The parameter controls the weight of the error. Smaller errors will result in current setpoints that are closer to the setpoint to reduce the error. This dynamic current setpoint adjustment formula takes into account multiple factors, including frequency adjustment, error penalty, and real-time error, to dynamically adjust the current setpoint to adapt to the current waveform under different operating conditions.

[0069] Embodiment 8: The negative sequence protection part uses the following formula to calculate the negative sequence current of the transformer:

[0070]

[0071] Among them, n is the serial index of the three-phase current; j is the imaginary number symbol.

[0072] Specifically, the formula The part represents the superposition of currents in complex form. This is based on Euler's formula for complex numbers, where e jθ represents a complex number, and θ is the phase angle. In the power system, there are usually three phases of current, namely phase A, phase B and phase C. The currents of these three phases can be expressed in complex form and the negative sequence current can be calculated by superposition. The complex form representation makes the calculation more convenient because it can combine the phase information and amplitude information in one mathematical expression. In the power system, the three-phase current is provided by three-phase power supplies or power lines. There is a phase difference of 120 degrees between the three phase currents. The formula The Lagrange multiplier α is the sum of the three phase currents and takes into account the phase difference between them. This is to calculate the negative sequence current, which is usually caused by the imbalance of the three phase currents. i and sample point x i Consider additional factors in the negative sequence current calculation. Each sample point x i There is an associated Lagrange multiplier α i , which are coefficients used to penalize errors in the current calculation. This allows the system to take into account the contribution of different sample points to the negative sequence current calculation and to weight different sample points as needed. Error term It is used to measure the error between the actual three-phase current and the set value. The calculation of the error is affected by the Euclidean distance, which is a method of measuring the distance between two vectors. Here, it is used to measure the difference between the three-phase current and the set value. A smaller error value means that the actual current is closer to the set value.

[0073] Specifically, the main function of the formula is to calculate the negative sequence current in the power system. Negative sequence current is a current component that is usually caused by unbalanced problems in the power system, such as phase difference, asymmetric load or fault. The presence of negative sequence current may cause equipment overheating, damage or system instability, so it needs to be monitored and evaluated. The complex form superposition part in the formula takes into account the phase difference between the three phase currents and combines them into a comprehensive negative sequence current value. This helps to comprehensively consider the influence of the three phase currents to accurately calculate the magnitude and direction of the negative sequence current. The Lagrange multiplier α in the formula i and sample point x i Allows the system to adjust the weight of the negative sequence current calculation based on sample data. This is very useful for considering the importance of different sample points and handling additional constraints. Through Lagrange multipliers, the system can adapt to different power system conditions more flexibly. The error term in the formula Used to measure the error between the actual current and the set value. A smaller error value means that the actual current is closer to the set value. This error measurement is very important for monitoring the stability and accuracy of the current. If the error exceeds a certain threshold, it may mean that there is a problem with the power system and protective measures need to be taken. When the negative sequence current exceeds the set threshold or the error reaches a certain level, the system can trigger corresponding protection measures, such as disconnecting the circuit or cutting off the power supply to avoid equipment overload or power system failure. This helps ensure the stability and safety of the power system. The purpose of this negative sequence current calculation formula is to help the power system protection system monitor the negative sequence current, evaluate the stability of the system, perform negative sequence current calculations based on actual conditions, and trigger protective measures when necessary to ensure the normal operation and safety of the power system.

[0074] Embodiment 9: The frequency response protection part calculates the frequency response according to the reference frequency and the frequency using the following formula:

[0075]

[0076] Where Δf is the frequency response; f t The real-time frequency.

[0077] Specifically, the frequency response Δf is used to represent the variable of the frequency response. The frequency response is usually used to describe the frequency change in the power system. It reflects the change of the system frequency relative to the reference frequency. The real-time frequency f t Indicates the real-time frequency value. It is used to compare with the reference frequency to determine the frequency response. The real-time frequency usually changes due to changes in the operating conditions of the power system. Frequency change rate This term represents the rate of change of frequency, i.e. the derivative of frequency with time. It takes into account the sum of the frequencies of multiple sample points and calculates its change with time. This is to capture the instantaneous change of frequency. Lagrange multiplier α i and sample point f i These terms are used to account for additional factors in the frequency response calculation. i There is an associated Lagrange multiplier α i , which are coefficients used to penalize errors in the frequency calculation. This allows the system to take into account the weights of different sample points when calculating the frequency response. It is used to measure the error between the actual frequency and the reference frequency. The calculation of the error is affected by the Euclidean distance, which is used to measure the difference between two frequencies. A smaller error value means that the actual frequency is closer to the reference frequency.

[0078] In the formula The part represents the instantaneous rate of change of frequency. Frequency changes are usually affected by load changes, faults or other factors in the power system. By calculating the rate of change of frequency, the instantaneous changes of frequency and possible frequency disturbances can be monitored. Lagrange multipliers are a method for dealing with constraints, which are used here to penalize the error in frequency calculation. It allows the system to make flexible trade-offs when considering different sample points to adapt to changes in the power system. Error term Used to measure the error between the actual frequency and the reference frequency. Euclidean distance is a commonly used method to measure the distance between two frequencies. Here, it is used to measure the difference between the actual frequency and the reference frequency. The purpose of this frequency response calculation formula is to help the power system monitor and evaluate the frequency response so as to promptly identify frequency changes and disturbances in the power system. It takes into account the frequency change rate, Lagrange multiplier and error term in order to more accurately evaluate the frequency response and take appropriate measures to ensure the stability and performance of the power system.

[0079] Embodiment 10: The performance evaluation part calculates the weighted average sum of the difference between the adjusted current and the mean current of the sample point, the difference between the variance of the negative sequence current and the current of the sample point, and the difference between the frequency response and the mean frequency of the sample point according to the set three weighting coefficients, and obtains the performance parameters of the large-capacity transformer; if the performance parameters do not exceed the set threshold, an adjustment instruction is sent to the support vector model part; after receiving the adjustment instruction, the support vector model part multiplies the weight vector of the support vector model by the performance parameter for adjustment.

[0080] Although the specific embodiments of the present invention are described above, it should be understood by those skilled in the art that these specific embodiments are only illustrative, and those skilled in the art may omit, replace, and change the details of the above methods and systems in various ways without departing from the principles and essence of the present invention. For example, merging the above method steps so as to perform substantially the same functions in substantially the same manner to achieve substantially the same results is within the scope of the present invention. Therefore, the scope of the present invention is limited only by the appended claims.

Claims

1. A large-capacity transformer primary current boost protection system, characterized in that: The system includes: a support vector model part, which is used to collect the parameters of the large-capacity transformer during operation at multiple moments, and the parameters at each moment are used as sample points for constructing the support vector model; a label is marked for each sample point, and the label value is -1 or 1; when the label is 1, it means that the protection mechanism needs to be triggered; when the label is -1, it means that the protection mechanism does not need to be triggered; these sample points and corresponding labels are used to train a support vector machine to obtain a support vector machine model, and the support vector machine model determines whether the protection mechanism needs to be triggered according to the real-time parameters, and when it is determined that the protection mechanism needs to be triggered, a signal is sent to the protection part; the protection part responds to the received signal and starts the protection mechanism for the large-capacity transformer, and the protection part includes: an overcurrent rapid removal protection part, a negative sequence protection part and a frequency response protection part; the overcurrent rapid removal protection part is used to The rated current and preset coefficient of the large-capacity transformer, the resistance of the transformer, the ambient temperature, and the influence of external interference are taken into consideration to set a reference current protection value, and a support vector machine model is used to adjust the reference current protection value to obtain a current setting value. After dynamically adjusting the current setting value, an adjusted current is obtained for overcurrent protection; the negative sequence protection part is used to calculate the negative sequence current of the transformer based on the three-phase current of the large-capacity transformer, which is used for negative sequence protection; the frequency response protection part is used to calculate the frequency response based on the reference frequency and the frequency, which is used for frequency response protection; the performance evaluation part is used to calculate the performance parameters of the large-capacity transformer based on the adjusted current, negative sequence current and frequency response, and if the performance parameters do not exceed the set threshold, an adjustment instruction is sent to the support vector model part; after receiving the adjustment instruction, the support vector model part adjusts the internal parameters of the support vector model.

2. The large-capacity transformer primary current rising protection system according to claim 1, characterized in that: The parameters include: preset coefficient, rated current, resistance, ambient temperature, frequency, reference frequency, angular frequency and three-phase current.

3. The large-capacity transformer primary current rise protection system according to claim 2 is characterized in that: The N sample points are (x1,y1),(x2,y2),…,(x N ,y N ), where x i is the parameter at each moment, which is an M-dimensional feature vector, y i The value is 1 or -1, indicating two categories; the goal is to find the parameters w and intercept b of a hyperplane to minimize the following optimization problem: Minimize Constraint w T Φ(x i )+b≥1-ξ i ,i=1,2,…,N; x i ≥0,i=1,2,…,N Specifically, w is the weight vector; b is the intercept; ξ i is a slack variable; C is a regularization parameter used to balance the penalty of maximizing the margin and classification errors; y i is the label of the sample point, which takes the value of 1 or -1, indicating two categories; Φ(x i ) means that the feature vector x i The feature vector is mapped to a high-dimensional space through a kernel function.

4. The large-capacity transformer primary current boost protection system according to claim 3, characterized in that: The support vector machine model is represented by the following formula: Among them, α i is the Lagrange multiplier; K(x i ,x t )=Φ(x t ) T Φ(x i ); Among them, x t is a real-time parameter; according to the calculated f(x t ) value to determine whether the protection mechanism needs to be triggered.

5. The large-capacity transformer primary current rising protection system according to claim 4, characterized in that: The reference current protection value is calculated using the following formula: Where ω is the angular frequency; |x i | is the feature vector x i The modulus value; R 変 is the resistance; I 额定 is the rated current; K is the set coefficient; p i is the power of the i-th sample point.

6. The large-capacity transformer primary current rising protection system according to claim 5, characterized in that: Use the following formula and support vector machine model to adjust the reference current protection value and obtain the current setting value: Among them, α is the Gaussian distribution coefficient; x is the Gaussian distribution variable; C is the regularization parameter.

7. The large-capacity transformer primary current boost protection system according to claim 6, characterized in that: Use the following formula to dynamically adjust the current setting value to obtain the adjusted current: Among them, f 基准 is the reference frequency; γ is the first adjustment coefficient, ranging from 0.3 to 0.6; δ is the second adjustment coefficient, ranging from 1.2 to 1.4; f i is the frequency of the i-th sample point; I t The real-time current.

8. The large-capacity transformer primary current boost protection system according to claim 7, characterized in that: The negative sequence protection part uses the following formula to calculate the negative sequence current of the transformer: Among them, n is the serial index of the three-phase current; j is the imaginary number symbol.

9. The large-capacity transformer primary current boost protection system according to claim 8, characterized in that: The frequency response protection part calculates the frequency response according to the reference frequency and the frequency using the following formula: Where Δf is the frequency response; f t The real-time frequency.

10. The large-capacity transformer primary current boost protection system according to claim 9, characterized in that: The performance evaluation part calculates the weighted average sum of the difference between the adjusted current and the mean current of the sample point, the difference between the variance of the negative sequence current and the current of the sample point, and the difference between the frequency response and the mean frequency of the sample point according to the set three weighting coefficients, and obtains the performance parameters of the large-capacity transformer; if the performance parameter does not exceed the set threshold, an adjustment instruction is sent to the support vector model part; after receiving the adjustment instruction, the support vector model part multiplies the weight vector of the support vector model by the performance parameter for adjustment.

Images