A method for energy storage grid-connected protection based on time-frequency domain waveform analysis and dynamic impedance domain

Abstract

The present application relates to a kind of energy storage grid-connected protection methods based on time-frequency domain waveform analysis and dynamic impedance domain, the present application solves the severe challenge that existing protection technology faces in new energy storage access line, especially in short-circuit current limited, high resistance ground fault, operating mode switching and transient process complex, single protection principle is difficult to take into account sensitivity, selectivity and reliability problem, main protection fusion time domain waveform and the multi-source information of frequency domain energy, significantly improve the detection ability and reliability of in-zone fault under weak feed power condition, based on time-frequency domain correlation criterion design, realize the millisecond level fast removal of in-zone fault, meet the requirement of system stability to protection speed, through main-backup collaborative mechanism and adaptive dynamic impedance modeling, effectively solve the low detection rate of high resistance ground fault and the problem of misoperation of out-zone fault.

Description

Technical Field

[0001] This invention relates to the field of power system relay protection, and in particular to a method for grid-connected energy storage protection based on time-frequency domain waveform analysis and dynamic impedance domain. Background Technology

[0002] With the significant increase in the penetration rate of new energy power generation such as wind power and photovoltaics, as well as large-scale energy storage systems in the power grid, the proportion of power sources based on power electronic inverters has increased significantly. As a core support for building a new power system, the grid-connected scale of energy storage systems continues to expand. However, their unique operating characteristics, distinct from traditional synchronous generators and pure new energy power sources, have led to a fundamental change in the dynamic fault response mechanism. Specifically, energy storage inverters exhibit strictly controlled current source characteristics during faults—the output current amplitude is controlled by active control strategies (such as millisecond-level current limiting mechanisms). Their short-circuit current level is far lower than that of traditional synchronous power sources, and they lack the strong transient characteristics (such as DC offset components and slow decay characteristics) generated by the large rotor inductance of synchronous generators.

[0003] The aforementioned characteristics, combined with the unique operating scenarios of energy storage systems (frequent switching between charging and discharging modes, islanded operation in weak power grids, and parallel connection of multiple energy storage units), lead to a double failure risk for traditional relay protection systems that rely on high-amplitude transient currents:

[0004] 1. Potential failure to operate: When a high-impedance grounding fault occurs in the energy storage transmission line or a remote fault occurs in the weak feeder system, the limited output characteristics of the controlled current source result in insufficient fault current amplitude, and the sensitivity of traditional overcurrent protection is significantly reduced.

[0005] 2. Risk of false tripping: When an external fault occurs at the access point of the energy storage station, the low amplitude and low harmonic content of its output current may not meet the blocking conditions of traditional transient quantity protection criteria (such as harmonic braking and DC component blocking), leading to false tripping of the protection.

[0006] In recent years, while some progress has been made in improving protection schemes for new energy scenarios (such as waveform correlation analysis, dynamic impedance calculation, or improved DTW traveling wave technology), four major bottlenecks still exist in grid-connected systems dominated by energy storage:

[0007] (1) Lack of adaptability to sudden changes in fault characteristics caused by active control strategies of energy storage systems (such as PQ / Vf mode switching);

[0008] (2) It is difficult to effectively cover the protection requirements of the multi-terminal topology of the energy storage cluster;

[0009] (3) The adaptive adjustment mechanism of the protection criterion under a wide range of charging and discharging conditions is not considered;

[0010] (4) Existing algorithms cannot balance the contradiction between the millisecond-level response requirement and the computation time of multiple features (waveform difference, impedance trajectory, power gradient, etc.).

[0011] The above problems indicate that existing technologies are insufficient to meet the stringent requirements of energy storage grid-connected systems for protection speed, selectivity and reliability. There is an urgent need to build a new protection system that integrates energy storage-specific fault characteristic quantities (time-domain waveform difference, impedance dynamic trajectory and power direction gradient) to break through the protection performance bottleneck under complex operating scenarios. Summary of the Invention

[0012] In view of the above situation and to overcome the shortcomings of the existing technology, the purpose of this invention is to provide a method for grid-connected energy storage based on time-frequency domain waveform correlation analysis and dynamic impedance domain, which can effectively solve the problem of rapid identification and reliable isolation of internal faults in grid-connected energy storage lines.

[0013] The technical solution solved by this invention is:

[0014] A grid-connected protection method for energy storage based on time-frequency domain waveform correlation analysis and dynamic impedance domain includes a power line and three-phase AC power sources and energy storage stations at both ends of the line. On the line near the three-phase AC power source, from near to far, are sequentially arranged a step-down transformer, bus A, a first circuit breaker, and a first current transformer for collecting the voltage values ​​at both ends of bus A and the current values ​​at the first circuit breaker. On the line near the energy storage station, from near to far, are sequentially arranged bus B, a second circuit breaker, and a second current transformer for collecting the voltage values ​​at both ends of bus B and the current values ​​at the second circuit breaker. The method also includes a first processor, a first controller, a second processor, and a second controller. The signal output terminal of the first current transformer is connected to the signal input terminal of the first processor, the signal output terminal of the first processor is connected to the signal input terminal of the first controller, the signal output terminal of the first controller is connected to the signal input terminal of the first circuit breaker, the signal output terminal of the second current transformer is connected to the signal input terminal of the second processor, the signal output terminal of the second processor is connected to the signal input terminal of the second controller, and the signal output terminal of the second controller is connected to the signal input terminal of the second circuit breaker. The specific method includes the following steps:

[0015] Step 1: Collect the voltage value at both ends of bus A and the current value at the first circuit breaker through the first current transformer and input the voltage and current signals to the first processor; collect the voltage value at both ends of bus B and the current value at the second circuit breaker through the second current transformer and input the voltage and current signals to the second processor.

[0016] Step 2: The first processor and the second processor calculate the time-domain waveform similarity and wavelet packet entropy of the currents at both ends of the line using the currents of the two circuit breakers at both ends of the line obtained in Step 1, and obtain the comprehensive criterion K. If the value of K is greater than the threshold, the protection will be activated and the main protection will be completed; if the value of K is less than the threshold, the main protection will not be activated.

[0017] Step 3: The first processor and the second processor use the voltage and current at both ends of the line obtained in Step 1 to construct a dynamic impedance. When the measured impedance track M deviates from the elliptical impedance domain corresponding to the current operating mode for more than 80ms, the protection action trips and completes the backup protection.

[0018] This invention provides a transmission line protection method suitable for large-scale grid integration of new energy storage. Its core lies in constructing a primary-backup collaborative protection architecture:

[0019] in:

[0020] (1) Main protection (step 2): Longitudinal protection constructed based on time-frequency domain signal correlation analysis. This method comprehensively utilizes time-domain waveforms and frequency-domain energy characteristics, and achieves rapid and sensitive fault identification by analyzing the correlation of electrical quantities at both ends of the line.

[0021] (2) Backup protection (step 3): Distance protection based on adaptive dynamic impedance modeling. This method constructs a dynamic impedance model that reflects the line's operating status, serving as an effective supplement to the main protection.

[0022] Main protection and backup protection work together:

[0023] The primary protection aims for extremely rapid action (typical action time ≤ 40ms) to clear faults within the protection zone as quickly as possible. The backup protection serves as redundancy (typical action time ≤ 80ms), providing reliable protection in complex conditions such as primary protection failure (e.g., high-impedance fault scenarios) or communication anomalies, covering any potential blind spots in the primary protection's operation.

[0024] Compared with the prior art, the present invention has the following beneficial technical effects:

[0025] This invention addresses the severe challenges faced by existing protection technologies in new energy storage access lines, particularly in conditions such as limited short-circuit current (controlled current source characteristics), high-resistance grounding faults, complex operating mode switching, and transient processes, where a single protection principle struggles to simultaneously address sensitivity, selectivity, and reliability. Its significant advantages include:

[0026] 1. High sensitivity and speed:

[0027] The main protection integrates multi-source information from time-domain waveforms and frequency-domain energy, significantly improving the detection capability and reliability of faults (especially minor faults) within the zone under weak power supply conditions.

[0028] Based on the time-frequency domain correlation criterion design, millisecond-level (≤40ms) rapid clearing of faults within the zone is achieved, meeting the system stability requirements for protection speed.

[0029] 2. High reliability and strong adaptability:

[0030] By employing a primary-backup collaborative mechanism and adaptive dynamic impedance modeling (combined with adaptive adjustment of SOC and other states), the problem of low detection rate of high-resistance grounding faults and false tripping due to faults outside the zone is effectively solved.

[0031] The backup protection design ensures that the protection system can still operate reliably and protect the system safety when the main protection fails to operate, communication errors occur, or the operating conditions of the energy storage power station (such as switching between charging and discharging modes or power output fluctuations) change drastically.

[0032] Dynamic impedance characteristic modeling enhances the adaptability to complex transient processes.

[0033] 3. Comprehensive solutions to problems in complex working conditions:

[0034] The method of this invention can effectively distinguish between internal and external faults in energy storage access lines, and exhibits excellent performance under various complex operating conditions such as short-circuit current limitation, high-resistance faults, complex transients, and operation mode switching. It overcomes the technical bottlenecks of existing technologies, such as poor adaptability to the above-mentioned complex operating conditions, decreased protection performance, and even failure.

[0035] 4. Enhance system security margins:

[0036] The dual-criteria collaborative and primary-backup coordinated architecture provides a higher level of line protection solution for power grids with large-scale new energy storage, significantly improving the overall reliability and adaptability of protection and meeting the development needs of new power systems. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the energy storage grid-connected longitudinal protection system of the present invention.

[0038] Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0039] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0040] like Figures 1-2As shown, this invention discloses an energy storage grid-connected protection method based on time-frequency domain waveform analysis and dynamic impedance domain. The method includes an energy storage grid-connected longitudinal protection system. This system comprises a line 5 and three-phase AC power supplies 1 and energy storage power stations 14 at both ends of the line. On the line near the three-phase AC power supply 1, from near to far, are sequentially arranged a step-down transformer 2, bus A3, a first circuit breaker 4, and a first current transformer 6 for collecting the voltage values ​​at both ends of bus A3 and the current values ​​at the first circuit breaker. On the line near the energy storage power station 14, from near to far, are sequentially arranged a bus B13, a second circuit breaker 9, and a current transformer 6 for collecting the voltage values ​​at both ends of bus B3 and the current values ​​at the second circuit breaker 14. The second current transformer 10, which measures the current at the circuit breaker, also includes a first processor 7, a first controller 8, a second processor 11, and a second controller 12. The signal output terminal of the first current transformer 6 is connected to the signal input terminal of the first processor 7, the signal output terminal of the first processor 7 is connected to the signal input terminal of the first controller 8, the signal output terminal of the first controller 8 is connected to the signal input terminal of the first circuit breaker 4, the signal output terminal of the second current transformer 10 is connected to the signal input terminal of the second processor 11, the signal output terminal of the second processor 11 is connected to the signal input terminal of the second controller 12, and the signal output terminal of the second controller 12 is connected to the signal input terminal of the second circuit breaker 9.

[0041] The specific method includes the following steps:

[0042] Step 1: Collect the voltage value at both ends of bus A and the current value at the first circuit breaker through the first current transformer and input the voltage and current signals to the first processor 7; collect the voltage value at both ends of bus B and the current value at the second circuit breaker through the second current transformer 10 and input the voltage and current signals to the second processor 11.

[0043] Step 2: The first processor and the second processor calculate the time-domain waveform similarity and wavelet packet entropy of the currents at both ends of the line using the currents of the two circuit breakers at both ends of the line obtained in Step 1, and obtain the comprehensive criterion K. If the value of K is greater than the threshold, the protection will be activated and the main protection will be completed; if the value of K is less than the threshold, the main protection will not be activated.

[0044] Step 3: The first processor and the second processor use the voltage and current at both ends of the line obtained in Step 1 to construct a dynamic impedance. When the measured impedance track M deviates from the elliptical impedance domain corresponding to the current operating mode for more than 80ms, the protection action trips and completes the backup protection.

[0045] Step 1 uses a frequency of 5kHz. The two current transformers respectively collect the three-phase voltage and current values ​​at both ends of the line, and input the collected voltage and current signals to the corresponding processors. For each phase current, the following is executed:

[0046]

[0047] in:

[0048] Δi(t): The sudden change in single-phase current at time t;

[0049] i(t): The current value of a certain phase of the three-phase current collected at the current time (t) (obtained by real-time acquisition of three-phase voltage and current signals through sampling devices on both sides of the line);

[0050] i(tk×1ms): The current value at the kth 1ms before time t (k=1,2,3,4,5, i.e., take the historical current data of the 5 1ms before the current time);

[0051] When any phase Δi continuously exceeds 0.2 times the rated current (0.2I) N If the time reaches 2ms, it is determined that a fault has occurred, and steps 2 and 3 are performed.

[0052] The main protection process in step 2 is as follows:

[0053] After determining the fault in step 1, current sampling data from both sides of the line within a 40ms time window following the fault triggering are obtained. The Dynamic Time Warping (DTW) algorithm is used to calculate the similarity of the current waveforms at both ends of the line and the wavelet packet energy entropy, constructing a comprehensive criterion, specifically:

[0054] Dynamic time warping (RTW) algorithms can find the shortest cumulative distance path between two sequences while allowing for flexible scaling of the time axis. This path is used to measure the difference in waveform shape. Its recursive calculation is defined as follows: Let the discrete sampling sequences of the corresponding phases at both ends be:

[0055] X = {x1, ..., x} N}、Y={y1,...,y N}

[0056] Define local distance:

[0057] d(x_i, y_j) = |x i -y j |,

[0058] Then the cumulative distance D DTW for:

[0059] D(i,j)=d(x) i y j )+min{D(i-1,j-1)},D(i-1,j),D(i,j-1)}(2)

[0060] in:

[0061] xi y j : These are the current values ​​at the i-th and j-th sampling points in the sequences at both ends, respectively;

[0062] d(x i y j ): The absolute value of the difference between the current values ​​at the i-th and j-th sampling points in the two-end sequences;

[0063] D(i, j): The cumulative distance from the starting points of the two ends of the sequence to the i-th and j-th points;

[0064] D DTW : DTW distance between the two complete sequences (N and M are the lengths of the two sequences);

[0065] Ultimately, DTW is far from D. DTW The smaller the value, the higher the similarity of the waveforms at both ends, and vice versa. Compared with simple linear measures such as the Pearson correlation coefficient, DTW can handle the details of fault waveforms on both sides and nonlinear mismatches in time scales, thus improving the robustness of correlation discrimination.

[0066] The discrete current sampling sequence of the corresponding phase at one end of the line, X = {x1, ..., x2} N}: The discrete current sampling sequence corresponding to one end of the line (N is the number of sampling points); Y = {y1, ..., y2} N}: The discrete current sampling sequence corresponding to the other end of the line; Substitute the discrete current sampling sequences on both sides of the line into formula (2) to calculate the DTW distance D of the current on both sides of the line. DTW ;

[0067] Wavelet packet entropy feature extraction:

[0068] To enhance the detection capability for transient faults and improve sensitivity to high-resistance and nonlinear faults, wavelet packet decomposition is performed on the acquired current signal to obtain multi-band energy distribution. The energy distribution characteristics of each band are extracted, and the entropy value of the signal is calculated. Specifically, J-level wavelet packet transform is performed on each phase current to obtain 2... J Individual frequency band coefficients:

[0069] Calculate the energy E of each subband i And normalized to a probability distribution

[0070]

[0071] in:

[0072] E i : Energy of the i-th sub-band.

[0073] Then, the wavelet packet entropy is calculated using the Shannon entropy formula, and the energy of each subband is normalized to p. i , Shannon entropy formula:

[0074] H=-∑p i ×ln(p i (4)

[0075] in:

[0076] p i : The time-frequency distribution of energy at a point in the i-th sub-band;

[0077] Comprehensive criteria:

[0078] Different faults inside and outside the fault zone result in different time-frequency structures of the current signals at both ends of the line. When a fault occurs outside the fault zone, the transient current waveforms at both ends are highly similar. When a fault occurs inside the fault zone, the current waveforms at both ends of the line differ due to the fault. Dynamic Time Warping (DTW) uses flexible path matching to eliminate sampling asynchrony errors and accurately quantifies waveform similarity, which can effectively identify most faults within the fault zone. However, when a high-resistance fault (such as arc resistance) or a nonlinear component (such as an energy storage inverter) occurs within the fault zone, time-domain analysis alone is insufficient to capture the fault characteristics. However, at this time, the frequency domain energy distribution is distorted. Wavelet packet energy entropy (WPE) can be used to calculate the energy entropy difference between each frequency band to capture the frequency domain characteristic differences, thereby enhancing the detection sensitivity for high-resistance and nonlinear faults and improving the fault detection capability under complex operating conditions. Therefore, time-domain waveform similarity is an intuitive and universal characteristic of faults within the zone (covering most fault scenarios) and can be given high weight; frequency-domain energy difference, for special operating conditions such as high impedance and inverter limiting, serves as a supplement to improve protection robustness, while reflecting both waveform consistency (judging as a fault within the zone when it approaches 1) and frequency-domain difference (strengthening high impedance fault identification when it increases). Together, they form a collaborative strategy of "time domain as the main focus and frequency domain as the auxiliary focus," resulting in the following comprehensive judgment criteria. First, the comprehensive judgment index is calculated:

[0079]

[0080] in:

[0081] K: Comprehensive judgment indicator;

[0082] D DTW DTW distance between the two current waveforms;

[0083] D MAX : The maximum DTW distance calculated when there is a fault outside the zone;

[0084] H j1 H j2 : Wavelet packet entropy of the corresponding phase current signals at both ends of the line;

[0085] When a fault occurs outside the line zone, the voltage and current waveforms at both ends show a high degree of consistency. At this time, the waveform difference value obtained by dynamic time warping (DTW) analysis is small, so the characteristic index value approaches 0. At the same time, the sum of the spectrum differences obtained by frequency domain analysis at both ends is significantly smaller, resulting in a smaller comprehensive index K value. Conversely, when a fault occurs inside the line zone, the similarity of the waveforms at both ends decreases, the difference value of DTW analysis results increases, the waveform matching degree decreases significantly, and the frequency domain characteristic differences are also more obvious, resulting in a significant increase in the K value.

[0086] Through extensive simulation calculations, this application sets the action threshold to 0.75, that is:

[0087] When the real-time calculated K value exceeds 0.75, the main protection is activated and trips; when K is less than 0.75, it does not operate.

[0088] The backup protection process in step 3 is as follows:

[0089] Based on the voltage and current obtained in step 1, a dynamic impedance model is constructed. When the measured impedance trajectory deviates from the elliptical impedance domain corresponding to the current operating mode for more than 80ms, the protection system will activate and trip. Specifically:

[0090] Based on the voltage and current signals sampled from both sides in step 1, the interference of the zero-sequence component is eliminated through αβ coordinate transformation to obtain impedance data in the two-phase stationary coordinate system. Signal preprocessing and feature extraction are then performed. The acquired three-phase signals are subsequently subjected to αβ coordinate transformation (Clarke transform), and the specific calculation formula is as follows:

[0091]

[0092] in:

[0093] u a u b u c Instantaneous values ​​of three-phase voltage (obtained from current transformers);

[0094] u α u β Voltage components in the transformed αβ coordinate system;

[0095] In addition, the SOC (State of Charge) of the energy storage system is estimated in real time using a combination of the ampere-hour integral method and the open-circuit voltage method. The calculation formula is as follows:

[0096]

[0097] in:

[0098] Q n For rated capacity, ΔSOC OCVThis is an open-circuit voltage correction term; the SOC is updated in real time every second, providing a basis for status identification and subsequent protection criteria.

[0099] SOC(t): State of charge at time t;

[0100] i(t): Charging or discharging current;

[0101] Q n Rated capacity of the energy storage system;

[0102] ΔSOC OCV Open-circuit voltage correction term;

[0103] Given that the operating status of an energy storage system, such as its State of Charge (SOC) and power flow direction, has a significant impact on its impedance characteristics, this application uses the SOC range as the main framework and, combined with the charging and discharging power flow direction, employs a k-means clustering algorithm to automatically divide complex operating conditions into three typical operating modes: low SOC charging, floating, and high SOC discharging. Under different operating modes, the line impedance exhibits different mean values ​​and fluctuation ranges. This method allows for real-time and accurate matching of the line's operating status, laying the foundation for subsequent dynamic impedance domain modeling and adaptive protection.

[0104] Dynamic impedance calculation and elliptic domain update:

[0105] In each identified operating mode, the voltage and current signals in the αβ coordinate system are subjected to Fourier transform in real time to extract the fundamental phasor, and the line impedance is calculated in complex plane form.

[0106]

[0107] in:

[0108] Z: The magnitude of the line impedance;

[0109] U α U β Voltage components in the αβ coordinate system; (obtained through αβ transformation)

[0110] I α I β : Current components in the αβ coordinate system; (obtained through αβ transformation, the current transformation formula is the same as that for voltage, only the voltage is converted into current);

[0111] Calculate the ratio of the combined magnitudes of voltage and current in the αβ coordinate system to obtain the magnitude Z of the impedance;

[0112] To address the statistical distribution of impedance under different modes, an elliptic modeling method is used to describe the impedance active boundary, with the mathematical expression being:

[0113]

[0114] in:

[0115] θ: Phase angle of the impedance;

[0116] The discriminant for the elliptic impedance domain is:

[0117]

[0118] in:

[0119] Z0 and θ0 are the centers of the ellipse, and a and b are the major and minor semi-axes, respectively.

[0120] Z0, θ0: Impedance magnitude and phase angle at the center of the ellipse; (obtained through real-time statistical analysis of running data)

[0121] a, b: the major and minor axes of the ellipse;

[0122] Mapping relationship between SOC and elliptic parameters

[0123]

[0124] The criteria for obtaining backup protection are as follows:

[0125]

[0126] The rules for dynamically determining parameters include:

[0127] The baseline value a0 = 0.2 × Z line (Z line (Positive sequence impedance of the line)

[0128] Reference value b0 = 15°

[0129] a0 and b0 are the reference values ​​of the major and minor semi-axis of the elliptical impedance domain of the energy storage system under standard operating mode.

[0130] Scaling factor is adaptively adjusted according to SOC:

[0131]

[0132] Among them, Z0 and θ0 are updated online using a recursive least squares algorithm, with the initial value being the current;

[0133] The core criterion for backup protection is the relationship between the impedance trajectory and the elliptical domain of the real-time operation mode. The current impedance is checked every 20ms to see if it falls within the elliptical domain. If the impedance trajectory deviates from the elliptical domain for four consecutive cycles, it is considered to be abnormal or a potential fault. That is, when the impedance trajectory Z is greater than 1 within an 80ms time window, the backup protection is activated and trips. When the impedance trajectory Z is less than or equal to 1 within an 80ms time window, it does not operate.

[0134] Figure 1 In this embodiment, a 10km three-phase transmission line was built using the Simscape Electrical module library on the Simulink platform. A three-phase AC power supply and an energy storage station were connected to opposite ends of the line. The three-phase AC power supply was configured with a rated voltage of 10kV and a frequency of 50Hz using the ACVoltage Source module. The energy storage station was modeled using a combination of the Controlled CurrentSource and Universal Bridge modules, with a rated capacity of 50MW / 100MWh, capable of switching between PQ and Vf modes. The transmission line was constructed using the Transmission Line module, with parameters set as follows: resistance 0.15Ω / km, inductance 1.2mH / km, and capacitance 0.01μF / km. Circuit breakers were installed at both ends using the Circuit Breaker module. Current Measurement modules were installed on both sides of the line to collect three-phase current signals, and the Voltage Measurement module was used to synchronously collect the bus voltages at both ends. The sampling frequency was set to 5kHz (sampling period 1ms). All signals were exported to the workspace via the To Workspace module and named IA, IB, IC, VA, VB, and VC.

[0135] Based on the above model, different fault types and fault locations are set through the Three-Phase Fault module to verify the operation of the main protection and backup protection.

[0136] First, the main protection is verified. The specific process is as follows:

[0137] When energy storage is in charging mode, different types of faults are set at different fault locations on the line, and multiple simulations are performed. The simulation results are shown in the table below. The table shows that, regardless of where any fault occurs along the entire line, the protection action value is greater than 0.75 after the fault occurs, indicating that the proposed protection scheme can operate reliably and quickly.

[0138]

[0139] When the energy storage is in the discharge state, different types of faults are set at different fault locations on the line, and multiple simulations are performed. The simulation results are shown in the table below. The table shows that, at any point along the entire line, regardless of the fault, the protection action value is greater than 0.75 after the fault occurs, indicating that the proposed protection scheme can operate reliably and quickly.

[0140]

[0141] The backup protection was verified. Specifically, assuming the main protection did not operate, different types of faults were set at different fault locations on the line, and multiple simulations were performed under different energy storage SOC states. The simulation results are shown in the table below.

[0142] 1. Charging status (SOC < 20%)

[0143] Elliptic domain parameters:

[0144] a = 0.8a0 = 0.65Ω

[0145] b = 0.9b0 = 13.5°

[0146]

[0147] 2. Float charge state (20% ≤ SOC ≤ 80%)

[0148] Elliptic domain parameters:

[0149] a = 0.8a0 = 0.812Ω

[0150] b = 0.9b0 = 15°

[0151]

[0152] 3. State of discharge (SOC > 80%)

[0153] Elliptic domain parameters

[0154] a = 1.2a0 = 0.974Ω

[0155] b = 1.1b0 = 16.59°

[0156]

[0157]

[0158] In summary, simulations on the MATLAB / Simulink platform fully validate the energy storage grid-connected protection method based on time-frequency domain waveform analysis and dynamic impedance domain. This method enables high-speed, sensitive, and reliable identification of internal faults under energy storage access and complex operating conditions, effectively preventing maloperation or failure to operate. This invention not only possesses good engineering feasibility and repeatability but also has strong practical application value, providing a safe and reliable relay protection solution for new power systems and large-scale energy storage grid-connected scenarios.

Claims

1. A grid-connected protection method for energy storage based on time-frequency domain waveform analysis and dynamic impedance domain, comprising a grid-connected longitudinal protection system for energy storage, characterized in that, The grid-connected protection system for energy storage includes a line (5) and three-phase AC power supplies (1) and energy storage power stations (14) at both ends of the line. On the line near the three-phase AC power supply (1), a step-down transformer (2), bus A (3), a first circuit breaker (4), and a first current transformer (6) for collecting the voltage value at both ends of bus A and the current value at the first circuit breaker are installed sequentially from near to far on the line near the energy storage power station (14). On the line near the energy storage power station (14), a bus B (13), a second circuit breaker (9), and a second current transformer (10) for collecting the voltage value at both ends of bus B and the current value at the second circuit breaker are installed sequentially from near to far on the line. It also includes a first processor (7) and a first controller. (8) The second processor (11) and the second controller (12), the signal output terminal of the first current transformer (6) is connected to the signal input terminal of the first processor (7), the signal output terminal of the first processor (7) is connected to the signal input terminal of the first controller (8), the signal output terminal of the first controller (8) is connected to the signal input terminal of the first circuit breaker (4), the signal output terminal of the second current transformer (10) is connected to the signal input terminal of the second processor (11), the signal output terminal of the second processor (11) is connected to the signal input terminal of the second controller (12), and the signal output terminal of the second controller (12) is connected to the signal input terminal of the second circuit breaker (9). The specific method includes the following steps: Step 1: Collect the voltage value at both ends of bus A and the current value at the first circuit breaker through the first current transformer (6) and input the voltage and current signals to the first processor (7); collect the voltage value at both ends of bus B and the current value at the second circuit breaker through the second current transformer (10) and input the voltage and current signals to the second processor (11). Step 2: The first processor and the second processor calculate the time-domain waveform similarity and wavelet packet entropy of the currents at both ends of the line using the currents of the two circuit breakers at both ends of the line obtained in Step 1, and obtain the comprehensive criterion K. If the value of K is greater than the threshold, the protection will be activated and the main protection will be completed; if the value of K is less than the threshold, the main protection will not be activated. Step 3: The first processor and the second processor use the voltage and current at both ends of the line obtained in Step 1 to construct a dynamic impedance. When the measured impedance trajectory deviates from the elliptical impedance domain corresponding to the current operating mode for more than 80ms, the protection action trips and completes the backup protection. Step 1 uses a frequency of 5kHz. The two current transformers respectively collect the three-phase voltage and current values ​​at both ends of the line, and input the collected voltage and current signals to the corresponding processors. For each phase current, the following is executed: （1） in: The sudden change in single-phase current at time t; The value of a specific phase of the three-phase current collected at the current time (t); The current value at the kth 1ms interval before time t; When any phase If the current exceeds 0.2 times the rated current for 2ms, it is determined that a fault has occurred, and steps 2 and 3 are performed. The main protection process in step 2 is as follows: Based on the fault determination in step 1, current sampling data from both sides of the line within a 40ms time window after the fault is triggered are obtained. The dynamic time warping algorithm is used to calculate the similarity of the current waveforms at both ends of the line and the wavelet packet energy entropy, constructing a comprehensive criterion, specifically: Dynamic time warping (RTW) algorithms can find the shortest cumulative distance path between two sequences while allowing for flexible scaling of the time axis. This path is used to measure the difference in waveform shape. Its recursive calculation is defined as follows: Let the discrete sampling sequences of the corresponding phases at both ends be: 、 Define local distance: d(x_i,y_j)=|x i -y j |, Then the cumulative distance for: （2） in: : These are the current values ​​at the i-th and j-th sampling points in the sequences at both ends, respectively; The absolute value of the difference between the current values ​​at the i-th and j-th sampling points in the two-end sequences; : The cumulative distance from the starting points of both ends of the sequence to the i-th and j-th points; DTW distance between the two complete sequences; Discrete current sampling sequence of the corresponding phase at one end of the line. : The discrete current sampling sequence corresponding to one end of the line; The discrete current sampling sequence at the other end of the line; substituting the discrete current sampling sequences on both sides of the line into formula (2), the DTW distance between the currents on both sides of the line is calculated. ; Wavelet packet entropy feature extraction: To enhance the detection capability for transient faults and improve sensitivity to high-resistance and nonlinear faults, wavelet packet decomposition is performed on the acquired current signal to obtain multi-band energy distribution. The energy distribution characteristics of each band are extracted, and the entropy value of the signal is calculated. Specifically, J-level wavelet packet transform is performed on each phase current to obtain 2... J Individual frequency band coefficients: Calculate the energy of each subband And normalized to a probability distribution （3） in: Energy of the i-th sub-band; Then, the wavelet packet entropy is calculated using the Shannon entropy formula, and the energy of each subband is normalized to... ,Shannon entropy formula: （4） in: : The time-frequency distribution of energy at a point in the i-th sub-band; The comprehensive judgment criteria are as follows. First, the comprehensive judgment index is calculated: （5） in: Comprehensive judgment indicators; DTW distance between the two ends of the current waveform; : The maximum DTW distance calculated when there is a fault outside the zone; : Wavelet packet entropy of the corresponding phase current signals at both ends of the line; When the real-time calculated K value exceeds 0.75, the main protection is activated and trips; when K is less than 0.75, it does not operate.

2. The energy storage grid-connected protection method based on time-frequency domain waveform analysis and dynamic impedance domain as described in claim 1, characterized in that, The backup protection process in step 3 is as follows: Based on the voltage and current obtained in step 1, a dynamic impedance model is constructed. When the measured impedance trajectory deviates from the elliptical impedance domain corresponding to the current operating mode for more than 80ms, the protection system will activate and trip. Specifically: Based on the voltage and current signals sampled from both sides in step 1, the interference of the zero-sequence component is eliminated through αβ coordinate transformation to obtain impedance data in the two-phase stationary coordinate system. Signal preprocessing and feature extraction are then performed. The acquired three-phase signals are subsequently subjected to αβ coordinate transformation, and the specific calculation formula is as follows: （6） （7） in: Instantaneous values ​​of three-phase voltage; Voltage components in the transformed αβ coordinate system; Furthermore, the State of Charge (SOC) of the energy storage system is estimated in real time using a combination of the ampere-hour integral method and the open-circuit voltage method. The calculation formula is as follows: （8） in: For rated capacity, This is an open-circuit voltage correction term; State of charge at time t; Charging or discharging current; Rated capacity of the energy storage system; Dynamic impedance calculation and elliptic domain update: Real-time Fourier transform of voltage and current signals in the αβ coordinate system is performed to extract the fundamental phasor, and the line impedance is calculated in complex plane form. （9） in: The magnitude of the line impedance; , Voltage components in the αβ coordinate system; , Current components in the αβ coordinate system; Calculate the ratio of the combined magnitudes of voltage and current in the αβ coordinate system to obtain the magnitude Z of the impedance; To address the statistical distribution of impedance under different modes, an elliptic modeling method is used to describe the impedance active boundary, with the mathematical expression being: (10) in: Phase angle of impedance; The discriminant for the elliptic impedance domain is: （11） in: , Center of the ellipse , These are the major and minor semi-axles, respectively; , The impedance magnitude and phase angle at the center of the ellipse; , The major and minor axes of an ellipse; The criteria for obtaining backup protection are as follows: （12） The rules for dynamically determining parameters include: benchmark value ; benchmark value ; Scaling factor is adaptively adjusted according to SOC: （13） （14） in, The current is updated online using a recursive least squares algorithm. The core criterion for backup protection is the relationship between the impedance trajectory and the elliptical domain of the real-time operation mode. The current impedance is checked every 20ms to see if it falls within the elliptical domain. If the impedance trajectory deviates from the elliptical domain for four consecutive cycles, it is considered to be abnormal or a potential fault. That is, when the impedance trajectory Z is greater than 1 within an 80ms time window, the backup protection is activated and trips. When the impedance trajectory Z is less than or equal to 1 within an 80ms time window, it does not operate.

Images