A wind farm sending line time domain distance protection method fusing composite weight matrix

Abstract

This invention relates to a time-domain distance protection method for wind farm transmission lines based on a composite weight matrix. The method includes: after a fault occurs, extracting the instantaneous values ​​of line voltage and current, performing low-pass filtering, constructing a discrete form of the differential equation, constructing a composite weight matrix integrating two types of weights, embedding the composite weight matrix into a recursive least squares framework to obtain recursive least squares parameters after incorporating the composite weight matrix, identifying the parameters, calculating the measured impedance, using the directional circle characteristic as the distance protection criterion, comparing the measured impedance with the set impedance, and determining whether the protection should operate. This method constructs a time-domain fault difference equation based on the electrical quantities at both ends of the line, avoiding the explicit influence of transition resistance; designing a composite weight matrix integrating fitting error and transient change points, and embedding it into a recursive least squares framework, improves the convergence speed, steady-state accuracy, and transition resistance tolerance of the protection during transient processes.

Description

Technical Field

[0001] This invention relates to the field of power system relay protection technology, specifically to a time-domain distance protection method for wind farm transmission lines that integrates a composite weight matrix. Background Technology

[0002] New energy power generation technologies, represented by photovoltaics and wind power, have been developed and widely applied on a large scale in my country. Among them, doubly-fed induction generators (DFIGs) dominate large-scale wind farms due to their advantages such as good power decoupling control capabilities, a wide operating speed range, and high energy conversion efficiency. However, the grid-connected operation of DFIGs via power electronic converters fundamentally differs from traditional synchronous generators in their electromechanical transient characteristics, resulting in typical features such as weak inertia, lack of support, and weak feeder characteristics. During faults in the transmission lines of centralized DFIG wind farms, the fault current output by the wind farm exhibits complex characteristics such as limited amplitude, controlled phase, rich harmonic content, and slow decay of aperiodic components due to the activation of the rotor-side converter control and crowbar protection circuits of the DFIGs. This poses a severe challenge to fault analysis methods based on the linear superposition principle and synchronous power supply, as well as the operational performance of traditional relay protection principles. For example, due to the conservation of rotor flux, doubly-fed wind turbines feed transient inrush currents with rich frequency bands into the power grid. Subsequently, the switching of the crowbar circuit limits the amplitude of the fault current, which may lead to a decrease in the sensitivity of overcurrent protection or even failure to operate. The high harmonic content and non-power frequency phase characteristics of short-circuit current can cause serious errors in the measurement impedance of distance protection, resulting in false tripping or failure to operate. When a high transition resistance fault occurs, its measurement impedance is severely distorted, the effective protection range is shortened, and the risk of false tripping or failure to operate increases significantly, seriously threatening the safety of the power grid.

[0003] To address these challenges, scholars both domestically and internationally have proposed various novel protection methods, such as frequency domain and model matching methods, intelligent algorithm-driven methods, and control-protection synergy and active injection methods. Time-domain distance protection methods, theoretically unaffected by system frequency shifts and attenuating aperiodic components, have become a hot topic in current novel protection research. However, traditional time-domain distance protection uses a lumped-parameter RL model that ignores the distributed capacitance effect of the line; simultaneously, its parameter identification algorithm processes all data points with equal weight, making it extremely sensitive to current and voltage signals containing abundant high-frequency transient components and noise in the early stages of a fault. Therefore, when this method is applied to wind power transmission lines, parameter estimation exhibits significant fluctuations and errors. Summary of the Invention

[0004] This invention aims to address the reliability issues of traditional distance protection for wind farm transmission lines under conditions of high transition resistance, strong transient interference, and distributed capacitance. A time-domain distance protection method incorporating a composite weight matrix is ​​designed. This method constructs a time-domain fault difference equation based on electrical quantities at both ends of the line, avoiding the explicit influence of transition resistance. A composite weight matrix integrating fitting error and transient abrupt change points is designed and embedded within a recursive least squares framework, improving the convergence speed, steady-state accuracy, and transition resistance tolerance of the protection during transient processes.

[0005] The technical solution adopted by this invention to solve its technical problem is: a time-domain distance protection method for wind farm transmission lines that integrates a composite weight matrix, comprising: S1. After a fault occurs, extract the instantaneous values ​​of voltage and current on the M and N sides of the bus. S2. Perform low-pass filtering on the collected voltage and current. S3. Establish an improved time-domain differential equation that takes into account the voltage and current at both ends of the line, and discretize the differential equation. S4. Construct a composite weight matrix that integrates the two types of weights, namely the transient change point detection weight matrix and the fitting error weight matrix; S5. Composite weight calculation: Combine the two types of weights in step S4 to form a composite weight matrix. S6. Embed the composite weight matrix into the recursive least squares framework for parameter identification. Specifically, the recursive least squares method after introducing the composite weight matrix is ​​used to identify the resistance and inductance values ​​from the fault point to the bus M protection installation location in real time, and to calculate the measured impedance. S7. Using the directional circular characteristic as the distance protection criterion, the measured impedance is compared with the set impedance to determine whether the protection should operate.

[0006] The beneficial effects of this invention are as follows: Compared with existing technologies, the wind farm transmission line time-domain distance protection method integrating a composite weight matrix of the present invention has the following advantages: (1) The improved time-domain fault difference equation avoids the explicit influence of transition resistance, transforms the high transition resistance problem into a model robustness problem, and has strong resistance to transition resistance. (2) A composite weight matrix construction method that integrates fitting error weight and transient mutation point weight, wherein the fitting error weight is dynamically allocated based on historical residual statistics, and the transient mutation point weight is identified and penalized by three complementary detection methods: sliding window change rate, difference sign change, and adjacent point difference threshold. The composite weight matrix dynamically allocates weights from two dimensions: fitting confidence and transient features, effectively suppressing transient interference in the early stage of the fault, improving the convergence speed and steady-state accuracy of parameter identification in the transient process, and making the impedance trajectory smoother. (3) The algorithm implementation of embedding the composite weight matrix into the recursive least squares framework for parameter identification, and the overall process of impedance calculation and direction circle characteristic protection judgment based on the convergence criterion of parameter change rate. (4) It has low dependence on the accuracy of line distributed parameters and system impedance, can tolerate certain parameter deviations, reduces excessive requirements on on-site measurement configuration, and has strong engineering adaptability. (5) Modular design, clear algorithm structure, easy to implement digitally, applicable to short transmission line scenarios of wind farms at 220kV and below. Attached Figure Description

[0007] Figure 1 Equivalent circuit diagram for faults in the transmission lines of a doubly fed wind farm.

[0008] Figure 2 This is the simulation model for the present invention.

[0009] Figure 3 This is a flowchart of the present invention.

[0010] Figure 4 The impedance measurement results for different phase-to-phase faults in this invention are shown.

[0011] Figure 5 The impedance measurement results for grounding faults at different distances are presented in this invention.

[0012] Figure 6 The results of phase-to-phase fault measurement impedance under different methods of the present invention are shown.

[0013] Figure 7 The results of ground fault measurement impedance under different methods of the present invention are shown.

[0014] Figure 8 The measurement error of phase-to-phase faults at different distances in this invention is (Rg=3Ω).

[0015] Figure 9 The measurement error of ground faults at different distances in this invention is (Rg=3Ω).

[0016] Figure 10 The measurement error of phase-to-phase faults with different transition resistances in this invention is (Z1=3.67Ω).

[0017] Figure 11 The measurement error of grounding faults with different transition resistances in this invention is (Z1=3.67Ω). Detailed Implementation

[0018] The invention will now be described in further detail with reference to the accompanying drawings.

[0019] A time-domain distance protection method for wind farm transmission lines based on a composite weight matrix includes: S1. After a fault occurs, extract the instantaneous values ​​of voltage and current on the M and N sides of the line. The specific method is as follows: Taking a 220kV wind farm transmission line as an example, its simplified model is as follows: Figure 1 As shown, the line protection installation points on the wind farm side and the system side are located at bus M and bus N, respectively. The instantaneous voltage and current values ​​measured at bus M are... , The instantaneous values ​​of voltage and current measured at bus N are , The outgoing line is characterized by a hybrid π-type model consisting of a concentrated resistor, an inductor, and distributed capacitances to ground at both ends, in order to approximately reflect the characteristics of the distributed parameters. The protection installation point here refers to the line protection system, where relays are installed to collect the instantaneous values ​​of voltage and current. S2. Perform low-pass filtering on the collected voltage and current. S3. Construct the discrete form of the differential equation. The specific method is as follows: S31. Establish an improved time-domain differential equation considering the voltage and current at both ends of the line. Assuming a fault occurs at point F on the line, based on Kirchhoff's voltage law, improved time-domain differential equations considering the voltage and current at both ends of the line are established between bus M and the fault point, as shown in Equations 1 and 2 respectively: (1) (2) In the formula: R 1 、L 1 R and L represent the resistance and inductance values ​​from the fault point to the installation location of the busbar M protection, respectively; R and L represent the resistance and inductance values ​​of the entire transmission line; R g The value of the transition resistor. This refers to the short-circuit current flowing into the faulty branch.

[0020] Substituting Formula 1 into Formula 2 and rearranging, we get Formula 3: (3) Formula 3 transforms the direct influence of transition resistance on parameter identification into an indirect constraint on the measurements at both ends of the line. S32. Discretize the improved time-domain differential equation. At the protection installation location, the relay extracts discrete sampled data. After discretizing the differential equation, Equation 3 can be expressed as Equation 4: (4) In the formula: h To protect the sampling period, kUsing the sampling point number as an example, Equation 4 is rearranged into a linear regression form, resulting in the improved time-domain fault difference equation, which is Equation 5: (5) The parameters in Formula 5 are: y(k) These are observed values. h(k) It is a regression vector. θ It is a parameter vector, and the formulas for each parameter are: (6) (7) (8) For relays with zero-sequence current compensation and 0° wiring configuration, formulas 6 and 7 are replaced by formulas 9 and 10: (9) (10) In the formula: ; K r , K l These are the compensation coefficients for zero-sequence resistance and inductance, respectively. i m0 ( k ), i n0 ( k These are the zero-sequence currents measured at bus M and bus N, respectively. For relays with 0° wiring, formulas 6 and 7 are modified to formulas 11 and 12: (11) (12) In the formula: .

[0021] S4. To improve the robustness of the parameter identification algorithm to transient processes and noise in wind power faults, a composite weight matrix integrating two types of weights is constructed, namely the transient change point detection weight matrix and the fitting error weight matrix. The specific method is as follows: S41. By statistically analyzing the fitting error and calculating the weights, a fitting error weight matrix is ​​constructed. Based on the statistical analysis of historical fitting residuals, the weights of data points are dynamically adjusted, prioritizing points with smaller fitting errors. The method is as follows: S411. First, calculate the fitting residuals: (13) S412. Then update the error statistics: (14) (15) In the formula: This is an estimate of the mean error; For error variance estimation; α The forgetting factor is used to control the memory length of the statistical statistic.

[0022] S413, then calculate the standardized error score. z ( k ): (16) In the formula: To prevent division by zero of constants.

[0023] S414, Error Fitting Weights Using piecewise functions for calculation and applying boundary constraints to the final weights, we obtain formulas 17 and 18: (17) (18) In the formula: The maximum fit weights; The minimum fitting weights are used to prevent numerical instability. S42. Construct a transient mutation point detection weight matrix. The transient mutation point weight matrix identifies transient mutation points through three complementary methods: sliding window rate of change detection, difference sign change detection, and adjacent point difference threshold detection, and assigns penalized weights accordingly. Specifically: The sliding window rate of change detection uses the following formula 19. (19) In the formula: It is a sliding window of length m; γ is the rate of change threshold; The difference sign change detection uses the following formula 20. (20) In the formula: ; The adjacent point difference threshold detection uses the following formula 21. (twenty one) In the formula: The difference threshold is typically set to... ; By combining the results of three detection methods—window change rate detection, difference sign change detection, and adjacent point difference threshold detection—a voting mechanism is used to determine the transient abrupt change point, as shown in Formula 22. (twenty two) For data identified as transient mutation points, a penalty weight is assigned, as shown in Formula 23. (twenty three) In the formula: This is the weighting penalty factor for transient mutation points; S5. Composite weight calculation: Combine the two types of weights from step S4 to form a composite weight matrix, as shown in Formula 24. (twenty four) S6. During the parameter identification process, the composite weight matrix is ​​embedded into the recursive least squares framework to achieve adaptive suppression of bad data. The parameters of the recursive least squares method after introducing the composite weight matrix are updated as follows: (25) (26) (27) In the formula: Forgetting factor; Real-time identification using recursive least squares method with composite weight matrix R 1. L 1. Calculate the measured impedance.

[0024] S7. Using the directional circular characteristic as the distance protection criterion, the measured impedance is compared with the set impedance to determine whether the protection should operate.

[0025] The comparison formula is:

[0026] Among them, the measured impedance is Z m The set impedance refers to the impedance value calculated for the relay. Z set yes The preset action threshold of the relay protection device is used to determine whether a fault has occurred in the transmission line.

[0027] Its flowchart is as follows Figure 3 As shown, when a fault occurs in the wind farm's transmission line, the relay collects time-domain sampling data of the voltage and current at both ends of the line at a sampling rate of 10kHz and a window length of 5ms. After low-pass filtering, the data is substituted into the improved time-domain fault difference equation. The differential equation is then discretized to construct a composite weight matrix that integrates two types of weights. The composite weight matrix is ​​then fused using a recursive least squares algorithm for parameter identification. The forgetting factor λ is set to 0.95, and the transient change point penalty factor is... Taking 0.1, after calculating the measured impedance, the directional circular characteristic is used as the distance protection criterion. The protection is determined to operate by judging whether the measured impedance obtained by fitting is within the operating range of the characteristic.

[0028] The following verification is based on actual situations: Based on the actual network structure and parameters of the transmission lines of a doubly-fed wind power grid-connected system in a certain region, a simulation model was built using the Matlab / Simulink simulation platform, as follows: Figure 2 As shown, the wind farm has 238 2MW doubly-fed wind turbines installed. In the event of a serious fault, the internal crowbar protection of the turbines will activate, locking the converters. The 220kV transmission line is 23.51km long, with the following parameters per unit length: r 1 = 0.05Ω / km l 1 = 0.981 mH / km r 0 = 0.15Ω / km l 0 = 2.943 mH / km c 1 = 1.16 μF c 0 = 0.65 μF. The data sampling rate is 10 kHz, and the data sampling time window length is 5 ms.

[0029] In the simulation, metallic two-phase short-circuit faults and single-phase ground faults were set at 20%, 50%, and 80% of the total length of the transmission line near bus M, respectively. The fault impedance was extracted using a time-domain distance protection method with a fused composite weight matrix. The results are as follows: Figure 4 , 5 As shown, in the case of a metallic fault in the transmitting line, the fault impedance obtained by this method has a small error compared with the theoretical value.

[0030] Further, the accuracy of the proposed method is compared with traditional power frequency protection methods and traditional time-domain protection methods. Fault tests are set at specific key points of the line. Taking protection stage I as an example, the phase-to-phase element is set to reliably avoid faults at 75% of the total length of the transmission line, with an impedance value of 5.5Ω; the grounding element is set to reliably avoid faults at 60% of the total length of the line, with an impedance value of 4.4Ω. Accordingly, a metallic two-phase short circuit is set at 75% of the line length in the simulation, and a metallic single-phase ground fault is set at 60% of the line length. The impedance measurement results of the three methods are as follows: Figure 6 , 7 As shown, traditional power frequency protection methods exhibit significant impedance fluctuations, easily leading to protection failure under two-phase short-circuit fault scenarios and large measurement errors under single-phase ground fault scenarios, failing to meet protection reliability requirements. In contrast, both traditional time-domain protection methods and the method presented in this paper show improvements in measurement accuracy. Moreover, the calculation results of the method presented in this paper have smaller errors than those of traditional time-domain protection methods, and the measured impedance curve is smoother, exhibiting higher measurement stability and robustness.

[0031] Introducing fault impedance calculation error δ: (28) In the formula: Z1 is the fault impedance; Z is the total impedance of the transmission line.

[0032] Phase-to-phase faults and ground faults were set at different locations on the transmission line. The results are summarized in Table 1, and the error curve is shown in Figure 1. Figure 8 , 9 As shown.

[0033] Table 1. Comparison of fault impedance measurement errors between traditional time-domain protection methods and the method presented in this paper at different fault locations.

[0034] According to Table 1 and Figure 8 , 9 It can be seen that the errors of the proposed method are all smaller than those of traditional time-domain protection methods, and the error distribution of both methods exhibits a characteristic of being high at both ends and low in the middle. During near-end faults, although the line inductive reactance is small, the distributed capacitance shunting is significant; during mid-section faults, the capacitance effect weakens, signal quality is good, model matching is highest, and error is smallest; during far-end faults, the line capacitance effect is complex, voltage drops lead to a decrease in signal-to-noise ratio, model error and signal attenuation are amplified, and impedance calculation accuracy decreases. Traditional time-domain protection methods, due to the mismatch between their lumped parameter models and the actual distributed parameter characteristics of the line, and the lack of adaptive ability to data quality changes, struggle to cope with the aforementioned nonlinear changes. The proposed method, however, can effectively suppress abnormal data point interference caused by capacitance effects and signal attenuation by dynamically evaluating data reliability, thus exhibiting more stable and accurate impedance calculation capabilities at each fault location.

[0035] In the simulation, phase-to-phase and ground faults were simulated at 50% of the transmission line under different transition resistances. The fault impedance measurement results are summarized in Table 2, and the error curve is shown in the figure. Figure 10 , 11 As shown.

[0036] Table 2 Comparison of fault impedance measurement errors between traditional time-domain protection methods and the method presented in this paper under different transition resistances.

[0037] According to Table 2 and Figure 10 , 11It is known that the error of traditional time-domain protection methods increases significantly with increasing transition resistance, posing a risk of failure to operate in high-resistance fault scenarios. This is because the method is based on the assumption that the impedance angle of the fault loop is constant. When the transition resistance is high, the resistive component of the loop increases significantly, causing the total impedance angle to deviate severely from the line inductive reactance angle, thus leading to model mismatch and calculation errors. In ground faults, the introduction of zero-sequence parameters further exacerbates this problem. In contrast, the method presented in this paper has an error within 3%, ensures reliable protection operation, effectively overcomes the model mismatch problem under high transition resistance, and demonstrates stronger resistance tolerance.

[0038] This invention is not limited to the described embodiments. Anyone should know that any structural changes made under the guidance of this invention, and any technical solutions that are the same as or similar to this invention, fall within the protection scope of this invention.

[0039] The technologies, shapes, and structures not described in detail in this invention are all known technologies.

Claims

1. A time-domain distance protection method for wind farm transmission lines integrating a composite weight matrix, characterized in that: include: S1. After a fault occurs, extract the instantaneous values ​​of voltage and current on the M and N sides of the bus. S2. Perform low-pass filtering on the collected voltage and current. S3. Establish an improved time-domain differential equation that takes into account the voltage and current at both ends of the line, and discretize the differential equation. S4. Construct a composite weight matrix that integrates the two types of weights, namely the transient change point detection weight matrix and the fitting error weight matrix; S5. Composite weight calculation: Combine the two types of weights in step S4 to form a composite weight matrix. S6. Embed the composite weight matrix into the recursive least squares framework, and use the recursive least squares method after introducing the composite weight matrix to identify the resistance and inductance values ​​from the fault point to the bus M protection installation point in real time, and calculate the measured impedance. S7. Using the directional circular characteristic as the distance protection criterion, the measured impedance is compared with the set impedance to determine whether the protection should operate.

2. The wind farm transmission line time-domain distance protection method based on a fused composite weight matrix according to claim 1, characterized in that: The improved time-domain differential equations in step 3 are Equations 1 and 2, respectively: (1) (2) In the formula: R 1 、L 1 These are the resistance and inductance values ​​from the fault point to the installation location of the busbar M protection; R, L The resistance and inductance values ​​for the entire length of the output line; R g This is the resistance value of the transition resistor. This refers to the short-circuit current flowing into the faulty branch; Substituting Formula 1 into Formula 2 and rearranging, we get Formula 3: (3) Formula 3 transforms the direct influence of the transition resistance on parameter identification into an indirect constraint on the measurements at both ends of the line.

3. The wind farm transmission line time-domain distance protection method based on a fused composite weight matrix according to claim 1, characterized in that: Discretizing the improved time-domain differential equation yields Formula 4: (4) In the formula: h To protect the sampling period, k Using the sampling point number as an example, Equation 4 is rearranged into a linear regression form, resulting in the improved time-domain fault difference equation, which is Equation 5: (5) The parameters in Formula 5 are: y(k) These are observed values. h(k) Here, θ is the regression vector, and θ is the parameter vector. The formulas for each parameter are as follows: (6) (7) (8) For relays with zero-sequence current compensation and 0° wiring configuration, formulas 6 and 7 are replaced by formulas 9 and 10: (9) (10) In the formula: ; K r 、K 1 These are the compensation coefficients for zero-sequence resistance and inductance, respectively. i m0 ( k ), i n0 ( k These are the zero-sequence currents measured at bus M and bus N, respectively. For relays with 0° wiring, formulas 6 and 7 are modified to formulas 11 and 12: (11) (12) In the formula: .

4. The wind farm transmission line time-domain distance protection method based on a fused composite weight matrix according to claim 1, characterized in that: The method for constructing the fitting error weight matrix in step 4 is as follows: S411. First, calculate the fitting residuals: (13) S412. Then update the error statistics: (14) (15) In the formula: This is an estimate of the mean error; For error variance estimation; α The forgetting factor is used to control the memory length of the statistical statistic. S413, then calculate the standardized error score. z(k) : (16) In the formula: To prevent division by zero of constants; S414, Error Fitting Weights Using piecewise functions for calculation and applying boundary constraints to the final weights, we obtain formulas 17 and 18: (17) (18) In the formula: The maximum fit weights; The minimum fit weights.

5. The wind farm transmission line time-domain distance protection method based on a fused composite weight matrix according to claim 1, characterized in that: In step 4, the transient mutation point weight matrix identifies transient mutation points and assigns penalty weights through three complementary methods: sliding window rate of change detection, difference sign change detection, and adjacent point difference threshold detection.

6. The wind farm transmission line time-domain distance protection method based on a fused composite weight matrix according to claim 5, characterized in that: The sliding window rate of change detection uses the following formula 19. (19) In the formula: It is a sliding window of length m; γ is the rate of change threshold; The difference sign change detection uses the following formula 20. (20) In the formula: ; The adjacent point difference threshold detection adopts the following formula 21. (21) In the formula: The difference threshold is typically set to... ; By combining the results of three detection methods—window change rate detection, difference sign change detection, and adjacent point difference threshold detection—a voting mechanism is used to determine the transient abrupt change point, as shown in Formula 22. (22) For data identified as transient mutation points, a penalty weight is assigned, as shown in Formula 23. (23) In the formula: This is the weighting penalty factor for transient mutation points.

7. The wind farm transmission line time-domain distance protection method based on a fused composite weight matrix according to claim 1, characterized in that: In step 6, the recursive least squares parameters after introducing the composite weight matrix are updated as follows: (25) (26) (27) In the formula: It is a forgetting factor.

Images