Offshore wind power fault analysis and positioning method based on residual error correction

Abstract

The invention relates to an offshore wind power fault analysis and positioning method based on residual correction, and the method comprises the following steps: building a digital twinborn model of an offshore wind power grid based on the offshore wind power grid; fusing the measurement value of the measurement system of the offshore wind power grid and the digital twinborn model, and establishing a state estimation model; predicted power of each node of an offshore wind power grid is predicted through a state estimation model, a fault section is determined by collecting actual power of each node of a measurement system, calculating weighted residual errors of the predicted power and the actual power, virtual nodes are arranged in the fault section, and weighted residual errors of adjacent virtual node sections are compared. And an adjacent virtual node section with the maximum weighted residual error is obtained, so that the accurate position of the fault is determined. Compared with the prior art, the method has the advantages of high positioning precision, wide applicability and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid fault location, and in particular to an offshore wind power fault analysis and location method based on residual correction. Background Art

[0002] The offshore wind power grid fault location method effectively improves the monitoring performance of offshore wind power by using high-precision measurement equipment, such as PMU, SCADA, AMI measurement, etc. Fault location based on measurement equipment or systems can provide reliable protection for the safe operation of the power grid by updating data in real time, which has also become a hot topic of current research. The current power grid fault location method installs a large number of sensors and intelligent devices in the distribution network, uses the measured voltage and current values, and calculates the fault distance through Ohm's law according to the electrical characteristics of the line. This method is simple to operate, but the accuracy is easily affected by factors such as current transformers and fault types.

[0003] With the continuous development of offshore wind power automation technology, the topology of the power grid has become increasingly complex, bringing more severe challenges to offshore wind power fault location. Existing fault location methods may have insufficient sensitivity when dealing with complex fault types (such as high-impedance grounding faults and asymmetric faults), making it difficult to quickly and accurately locate the fault point.

[0004] Therefore, it is necessary to design a set of intelligent fault location technologies for offshore wind power so that the system can accurately and quickly detect the fault area, which has important practical significance. Summary of the invention

[0005] The purpose of the present invention is to provide an offshore wind power fault analysis and positioning method based on residual correction in order to overcome the defect of insufficient positioning accuracy in the above-mentioned prior art.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] A method for analyzing and locating offshore wind power faults based on residual correction includes the following steps:

[0008] S1: Based on the offshore wind power grid, establish a digital twin model of the offshore wind power grid;

[0009] S2: Integrate the measurement values ​​of the measurement system of the offshore wind power grid and the digital twin model to establish a state estimation model;

[0010] S3: The predicted power of each node of the offshore wind power grid is predicted through the state estimation model. The weighted residual of the predicted power and the actual power is calculated by collecting the actual power of each node of the measurement system, and the fault section is determined. A virtual node is set in the fault section, and the weighted residuals of adjacent virtual node sections are compared to obtain the adjacent virtual node section with the largest weighted residual, thereby determining the precise location of the fault.

[0011] Furthermore, the measurement system includes a smart meter and a μPMU, the measurement values ​​of the μPMU include the amplitude and phase angle of the three-phase node voltage and the branch current, and the smart meter converts the measurement value into the measurement value of the μPMU through the node information.

[0012] Furthermore, the measurement value of the measurement system and the power indirectly measured are used to supplement the measurement result of the energy meter. The calculation expression for converting the power into the current measurement value of the smart meter is:

[0013]

[0014] Where P i.pse With Q i.pse is the measured value of injected power at node i, is the real part of the current measurement value of the smart meter, is the imaginary part of the current measured by the smart meter, U i,re is the real part of the voltage vector of the smart meter, U i.im is the imaginary part of the voltage vector of the smart meter.

[0015] Furthermore, the expression of the state estimation model is:

[0016] x k =f(x k-1 ,u k-1 )+w k-1

[0017] z k =h(x k ,u k )+v k-1

[0018] In the formula, x is the state quantity; u is the input quantity; z is the measurement value; k is the step size, w and v are process and observation noise with zero mean; the system is represented as an n×N matrix in the form of X=[x1,...,xN], where n is the state dimension and N is the number of samples.

[0019] Furthermore, the power prediction is performed by integrating the Kalman filter method, iterating the prediction process, and obtaining the system transfer equation as follows:

[0020]

[0021] In the formula, and are the system state vector and observation vector at time k respectively; and Represent the noise and error introduced by the system during state transfer and observation respectively. The expression of the prediction result is:

[0022]

[0023] In the formula, and They represent the predicted and observed states of the j-th grid node at time k, respectively.

[0024] Furthermore, the modified expression of the integrated Kalman filter is:

[0025]

[0026] Where P k is the error covariance matrix, H k is the Jacobian matrix, K k is the Kalman gain, N is the number of samples, the upper horizontal line represents the mean, R d is the line resistance.

[0027] Furthermore, the fault section is determined by the residual normalized sum of squares, and the calculation expression of the residual normalized sum of squares is:

[0028]

[0029] In the formula, σ i is the standard deviation of the ith observation, z k is the measured value, ζ is the normalized sum of squares of the residuals, when the normalized sum of squares of the residuals conforms to the chi-square distribution, it indicates that the power grid is normal; otherwise, it indicates a fault section.

[0030] Furthermore, the virtual nodes are set in such a way that a certain number of virtual nodes are set on the road between two nodes with the largest weighted residuals.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] 1) The present invention establishes a digital twin model, fuses the measurement system with the digital twin model, predicts the power grid nodes through the state estimation model and Kalman filter correction, amplifies the abnormal data, calculates the residual between the predicted value and the actual value to identify the fault section, and accurately locates the fault by setting virtual nodes, with high accuracy.

[0033] 2) The present invention can analyze different fault types, different transition resistances and high-resistance faults, has high positioning accuracy for common faults, amplifies the display of fault characteristics, and is widely applicable to a variety of power grid environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a schematic diagram of the process of the present invention.

[0035] Figure 2 A schematic diagram of a virtual node.

[0036] Figure 3 This is a schematic diagram of IEEE33 node simulation.

[0037] Figure 4 This is the residual normalized sum of squares result graph. DETAILED DESCRIPTION

[0038] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0039] Example 1

[0040] The present invention is a method for analyzing and locating offshore wind power faults based on residual correction, comprising the following steps:

[0041] S1: Based on the offshore wind power grid, establish a digital twin model of the offshore wind power grid;

[0042] S2: Integrate the measurement values ​​of the measurement system of the offshore wind power grid and the digital twin model to establish a state estimation model;

[0043] S3: The predicted power of each node of the offshore wind power grid is predicted through the state estimation model. The weighted residual of the predicted power and the actual power is calculated by collecting the actual power of each node of the measurement system, and the fault section is determined. A virtual node is set in the fault section, and the weighted residuals of adjacent virtual node sections are compared to obtain the adjacent virtual node section with the largest weighted residual, thereby determining the precise location of the fault.

[0044] In order to clearly explain the specific process of fault location, Figure 1 The various steps of fault location are presented in detail, including the integration, conversion, time synchronization of measurement data, and the final fault state estimation.

[0045] Different measurement systems have certain differences in access system location, sampling frequency, measurement state, etc. In order to make full use of the data from different measurement systems, this paper will pre-process the obtained measurements from two aspects: measurement data conversion and time normalization, laying the foundation for subsequent high-precision and reliable fault location.

[0046] Since the measurements of μPMU and smart meters are nonlinear, in order to facilitate calculation, this paper converts their initial measurements into corresponding linear measurements. μPMU can directly measure the amplitude and phase angle of three-phase node voltage and branch current. The real and imaginary parts of voltage and current are obtained through linear transformation as follows:

[0047]

[0048] Where: U i.P ,θ i.P are the amplitude and phase angle of the node voltage at the measurement point respectively; I ij.P ,θ ij.P are the amplitude and phase angle of the branch current measured at the measuring point respectively, and j is the imaginary unit in the complex number.

[0049] The measured value of the smart meter is converted and derived to the node where the smart meter is located using the known node information, and the form of the smart meter measured value is consistent with that of the μPMU for data fusion. The voltage vector reference value is shown in the following formula:

[0050]

[0051] Where U i.r and U i.i is the state estimation value of the voltage phasor; U i.S It is the voltage amplitude measurement value of the smart meter.

[0052] The power measurement is converted to the equivalent branch current:

[0053]

[0054] Where P ij.S With Q ij.S is the measured power value at the beginning and end of branch ij.

[0055] In the distribution system, the node injected power also includes pseudo-measurements of indirect measurements to supplement the measurements of smart meters. The supplementary measured power is converted into current measurements at the smart meter installation:

[0056]

[0057] Where P i.pse With Q i.pseis the measured value of the injected power of node i; the power injected by the missing smart meter is supplemented by the injection of this pseudo measurement.

[0058] The fused measurement is used as the measurement of the state estimation model, and the measurement vector includes measurement information such as branch power, node injection power, branch current and node voltage.

[0059] The weighted residual between the measured value and the actual measured value is estimated by integrated Kalman filtering. When the system topology and measurement conditions are known, the hybrid measurement equation is constructed using integrated Kalman filtering.

[0060] The integrated Kalman filter represents the distribution of propagation states through a set. Unlike the least squares method, it avoids deriving the Jacobian matrix of the nonlinear model, thereby reducing the complexity of the algorithm implementation. In addition, this method uses the nonlinear characteristics and error statistics of the model to make the state estimation more accurate. Based on the digital twin model for fault location constructed in this paper, the hybrid measurement value is combined to establish the equation:

[0061] x k =f(x k-1 ,u k-1 )+w k-1 (7)

[0062] z k =h(x k ,u k )+v k-1 (8)

[0063] In the formula, x is the state quantity; u is the input quantity; z is the measurement value; k is the step size, w and v are the process and observation noise with zero mean; the system is represented as an n×N matrix, in the form of X=[x1, ..., xN], where n is the state dimension, N is the number of samples, and where there is no measurement point, the current and voltage are calculated using the distribution network state at the previous moment and the current and voltage of the adjacent nodes at that moment.

[0064]

[0065] When α is introduced as a state, the nonlinearity of the model increases, and the extended vector of the state is defined as y = (x, α); Formulas (7)-(8) are the discretization of the linear model differential algebraic equations given in Formula (9). Formulas (7) and (8) can be rewritten as Formulas (10) and (11):

[0066] y k =f(y k-1 ,u k-1 )+w k-1 (10)

[0067] zk =h(y k ,u k )+v k-1 (11)

[0068] The method based on integrated Kalman filtering can use an iterative prediction process, and the system transfer equations are shown in equations (12) and (13):

[0069]

[0070] In the formula, and are the system state vector and observation vector at time k respectively; and They represent the noise or error introduced by the system during state transfer and observation respectively.

[0071] Predict the status of each grid node:

[0072]

[0073] In the formula, and They represent the predicted and observed states of the j-th grid node at time k, respectively.

[0074] Compute the mean of the predicted states:

[0075]

[0076] In the formula, and Represent the means of the predicted and observed states of all ensemble members at time k, respectively.

[0077] Compute the covariance matrix of the predicted states:

[0078]

[0079] Where P k|k-1 is the predicted state covariance matrix at time k;

[0080] The integrated Kalman filter correction expression is:

[0081]

[0082] Where P k is the error covariance matrix, H k is the Jacobian matrix, K k is the Kalman gain, the Jacobian matrix H k No explicit derivation from nonlinear equations is required.

[0083] The residual between the updated observed and predicted values ​​is expressed as:

[0084]

[0085] After correction by integrated Kalman filtering, the data will be smoother and random fluctuations will be removed, making the corrected r k More stable and accurate response to fault characteristics.

[0086] The residual data processed by the above steps are used to significantly amplify the fault characteristics through the residual normalized sum of squares method, so as to infer the fault section:

[0087]

[0088] In the formula, σ i is the standard deviation of the ith observation or the standard deviation of the error.

[0089] When there is no fault in the offshore wind power, ζ is close to 0, which means that the fitting confidence of the digital twin model and the RTDS model is very high, which means that ζ conforms to the chi-square distribution. On the contrary, when a fault occurs in the distribution network, ζ will change significantly, and the fitting confidence is low at this time, which means that the probability of conforming to the chi-square distribution is extremely small. The probability density of the chi-square distribution is shown as follows:

[0090]

[0091] After determining the fault segment, in order to accurately locate the fault, n virtual nodes are added to the fault segment, such as Figure 2 shown.

[0092] Figure 2 The number of the additional virtual node is b (l,r) , where l represents the fault branch and r represents the additional virtual fault point. Then the weighted residuals of the virtual nodes are compared to finally get the precise location of the fault. The number of virtual nodes is determined according to the specific accuracy required.

[0093] The more virtual nodes there are, the higher the positioning accuracy. In fact, the distribution network lines are short, and the setting of a small number of virtual nodes can also meet the distribution network positioning needs.

[0094] Simulation test: In order to verify the effectiveness of the proposed method in fault location, a real-time simulation system is built as follows: Figure 3 The IEEE33-node standard system shown in the figure performs distribution network fault location test and runs the simulation model and digital twin model at the same time. The real-time synchronous operation of the RTDS simulation model and the digital twin model makes the fault location effect more reliable and accurate.

[0095] The system is connected to the large power grid at the left bus and connected to the distributed power source. The specific configuration is shown in Table 1.

[0096] Table 1 Distributed power capacity and access location

[0097]

[0098] Scenario 1: Location analysis of different fault types

[0099] In order to test the influence of different types of faults on fault location, this paper selects multiple sections to test the fault location method. The transition resistance is set to 50Ω. The location results are shown in Table 2.

[0100] Table 2 Fault location error under different fault types

[0101]

[0102] Experimental results show that the method proposed in the present invention can perform high-precision fault location at different fault locations under single-phase grounding, two-phase grounding, two-phase short circuit and three-phase short circuit faults.

[0103] Scenario 2: Positioning analysis under different transition resistances

[0104] In offshore wind power, changes in transition resistance will significantly affect the accuracy and effectiveness of fault location. By testing different transition resistances, the performance of the fault location algorithm under different conditions can be evaluated to ensure its stability in practical applications. Among the faults that often occur in distribution networks, transition resistance has the greatest impact on ground faults. Therefore, this paper selected three transition resistances of 10Ω, 50Ω, and 100Ω to test the accuracy of the fault location method proposed in this paper in the scenario of LLG faults. The positioning results are shown in Table 3.

[0105] The results show that the method proposed in the present invention can achieve a positioning accuracy of more than 97% when the transition resistance is 100Ω or less, which can well meet the fault location requirements of offshore wind power.

[0106] Table 3 Positioning results under different transition resistances

[0107]

[0108] Scenario 3: Positioning analysis under high-resistance fault

[0109] High-resistance faults are common and their detection and location are relatively complex, so special attention needs to be paid to ensure the safety and reliability of the distribution network.

[0110] High-resistance faults will result in relatively low fault currents. Traditional methods are difficult to identify fault characteristics, which may lead to inaccurate positioning or delays. However, the corrected residual normalized sum of squares in this paper can well amplify the high-resistance fault characteristics. An LG fault with a transition resistance of 1000Ω is set at 25% of lines 9 and 10. The residual normalized sum of squares is as follows: Figure 4 shown.

[0111] The experimental results show that the residual corrected by the method in this paper can significantly amplify the fault characteristics in the case of high-resistance faults, laying a good foundation for subsequent accurate fault location.

[0112] The following is a multiple verification test of the positioning effect under high-resistance faults. Using different fault types, 726 fault locations were set for simulation tests for 4 high-resistance faults of 400Ω, 600Ω, 800Ω, and 1000Ω. The experimental results are shown in the table. The positioning accuracy is expressed by the following formula:

[0113]

[0114] Where, L A is the positioning accuracy, L f is the distance between the actual fault location and the measured location result; L t is the total length of the faulty line.

[0115] Table 4 Fault location accuracy under high transition resistance

[0116]

[0117] Experimental results show that when the fault characteristics are weak, the fault location accuracy of the method proposed in the present invention can still reach more than 94%.

[0118] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A method for analyzing and locating offshore wind power faults based on residual correction, characterized in that: The following steps are involved: S1: Based on the offshore wind power grid, establish a digital twin model of the offshore wind power grid; S2: Integrate the measurement values ​​of the measurement system of the offshore wind power grid and the digital twin model to establish a state estimation model; S3: The predicted power of each node of the offshore wind power grid is predicted through the state estimation model. The weighted residual of the predicted power and the actual power is calculated by collecting the actual power of each node of the measurement system, and the fault section is determined. A virtual node is set in the fault section, and the weighted residuals of adjacent virtual node sections are compared to obtain the adjacent virtual node section with the largest weighted residual, thereby determining the precise location of the fault.

2. The method for analyzing and locating offshore wind power faults based on residual correction according to claim 1 is characterized in that: The measurement system includes a smart meter and a μPMU. The measurement values ​​of the μPMU include the three-phase node voltage and the amplitude and phase angle of the branch current. The smart meter converts the measurement values ​​into the measurement values ​​of the μPMU through the node information.

3. The method for analyzing and locating offshore wind power faults based on residual correction according to claim 1 is characterized in that: The measurement value of the measurement system also includes the indirectly measured power, which is used to supplement the measurement result of the energy meter. The calculation expression for converting the power into the current measurement value of the smart meter is: Where P i.pse With Q i.pse is the measured value of injected power at node i, is the real part of the current measurement value of the smart meter, is the imaginary part of the current measured by the smart meter, U i,re is the real part of the voltage vector of the smart meter, U i.im is the imaginary part of the voltage vector of the smart meter.

4. The method for analyzing and locating offshore wind power faults based on residual correction according to claim 1 is characterized in that: The expression of the state estimation model is: x k =f(x k-1 ,u k-1 )+w k-1 z k =h(x k ,u k )+v k-1 In the formula, x is the state quantity; u is the input quantity; z is the measurement value; k is the step size, w and v are process and observation noise with zero mean; the system is represented as an n×N matrix in the form of X=[x1,...,xN], where n is the state dimension and N is the number of samples.

5. The method for analyzing and locating offshore wind power faults based on residual correction according to claim 1, characterized in that: The predicted power is obtained by iterating the prediction process through the integrated Kalman filter method, and the system transfer equation is: In the formula, and are the system state vector and observation vector at time k respectively; and Represent the noise and error introduced by the system during state transfer and observation respectively. The expression of the prediction result is: In the formula, and They represent the predicted and observed states of the j-th grid node at time k, respectively.

6. The method for analyzing and locating offshore wind power faults based on residual correction according to claim 5 is characterized in that: The modified expression of the integrated Kalman filter is: Where P k is the error covariance matrix, H k is the Jacobian matrix, K k is the Kalman gain, N is the number of samples, the upper horizontal line represents the mean, R d is the line resistance.

7. The method for analyzing and locating offshore wind power faults based on residual correction according to claim 6 is characterized in that: The fault section is determined by the residual normalized sum of squares, and the calculation expression of the residual normalized sum of squares is: In the formula, σ i is the standard deviation of the ith observation, z k is the measured value, ζ is the normalized sum of squares of the residuals, when the normalized sum of squares of the residuals conforms to the chi-square distribution, it indicates that the power grid is normal; otherwise, it indicates a fault section.

8. The method for analyzing and locating offshore wind power faults based on residual correction according to claim 1, characterized in that: The virtual nodes are set in such a way that a certain number of virtual nodes are set on the road between two nodes with the largest weighted residuals.

Images