Double-fed variable speed pumped storage unit rotor current fast estimation method based on double subspace

Abstract

The application relates to a double-fed variable-speed pumped storage unit rotor current rapid estimation method based on a double subspace. The method comprises the following steps: obtaining three-phase current signals of a rotor side in a unit; performing a Clarke coordinate transformation on the three-phase current signals, and constructing a Hankel matrix based on a coordinate transformation result, as a time sequence observation data matrix; performing spatial decomposition based on the time sequence observation data matrix to obtain a main frequency subspace and a harmonic subspace; determining a main frequency direct estimation value from the main frequency subspace based on a frequency continuity constraint and a pole modulus constraint; performing adaptive harmonic identification based on the main frequency direct estimation value and the harmonic subspace to obtain a target harmonic frequency; and performing joint estimation based on the main frequency direct estimation value, the target harmonic frequency and the time sequence observation data matrix to obtain a transient signal estimation result of the rotor current. The method can improve the estimation accuracy of the pumped storage unit rotor current, and realizes rapid overcurrent protection of the pumped storage unit rotor.

Description

Technical Field

[0001] This application relates to the field of fault protection and power signal processing technology for pumped storage units, and in particular to a method for rapid estimation of rotor current of a doubly fed variable speed pumped storage unit based on dual subspace. Background Technology

[0002] With the large-scale construction of new power systems, doubly-fed variable-speed pumped-storage units undertake key control tasks such as frequency regulation, voltage regulation, and fault voltage support for the power grid, serving as the core unit for system stability control. After a power system fault occurs, the rotor-side current of the unit exhibits significant non-stationary characteristics. The signal frequency, amplitude, and phase fluctuate rapidly with the fault progression, while being superimposed with multiple interferences such as sudden increases in harmonic energy, frequency drift, three-phase imbalance, and measurement noise. Traditional steady-state signal detection methods struggle to output continuous and stable electrical quantity feedback, necessitating transient estimation to provide data support for the unit's rotor overcurrent protection.

[0003] In related technologies, commonly used transient estimation methods mainly include Fast Fourier Transform (FFT), Discrete Fourier Transform (DFT), Phase-Locked Loop (PLL), and parameterization methods such as Estimation of Signal Parameters via Rotational Invariance Techniques (ESPRIT). Among these, FFT and DFT rely on fixed-frequency window functions, and their frequency resolution is limited by the window length, making them prone to spectral leakage and estimation lag during rapid transient changes. While PLLs offer good real-time performance, they are susceptible to oscillations and phase-locked loop shifts under fault impacts and strong harmonic conditions. Although the traditional ESPRIT method has high frequency resolution, its subspace estimation is typically based on fixed-window singular value decomposition, making it sensitive to noise, harmonics, and mode crossovers, and prone to pole misselection and frequency jumps during fault transients.

[0004] Therefore, the accuracy of the frequency and amplitude estimation results obtained by the above method under complex transient conditions is low, which affects the reliability of the rotor overcurrent protection of pumped storage units. Summary of the Invention

[0005] Therefore, it is necessary to provide a method, apparatus, and equipment for rapid estimation of rotor current of doubly fed variable speed pumped storage units based on twin space, which can improve the estimation accuracy of rotor current of pumped storage units and address the above-mentioned technical problems.

[0006] In a first aspect, this application provides a method for rapid estimation of rotor current in a doubly-fed variable-speed pumped-storage unit based on dual-subspace, including:

[0007] Acquire the three-phase current signal on the rotor side of the doubly-fed variable-speed pumped-storage unit;

[0008] The three-phase current signal is subjected to Clark coordinate transformation, and a Hankel matrix of complex space vector is constructed based on the coordinate transformation result as a time-series observation data matrix; the time-series observation data matrix is ​​used to describe the time shift information and frequency mode information of the three-phase current signal.

[0009] Spatial decomposition is performed based on the time-series observation data matrix to obtain the main frequency subspace and harmonic subspace;

[0010] Based on frequency continuity constraints and pole magnitude constraints, a direct estimate of the dominant frequency is determined from the dominant frequency subspace; the direct estimate of the dominant frequency is used to describe the fundamental frequency variation in the three-phase current signal.

[0011] Adaptive harmonic identification is performed based on the direct estimate of the main frequency and the harmonic subspace to obtain the target harmonic frequency; the target harmonic frequency is used to describe the transient broadband interference in the three-phase current signal;

[0012] Based on the direct estimate of the main frequency, the target harmonic frequency, and the time-series observation data matrix, a joint estimation is performed to obtain the transient signal estimation result of the rotor current; the transient signal estimation result includes at least the signal frequency and the signal amplitude.

[0013] Secondly, this application also provides a rapid estimation device for rotor current of a doubly-fed variable-speed pumped-storage unit based on twin-space, comprising:

[0014] The acquisition module is used to acquire the three-phase current signal on the rotor side of the doubly-fed variable-speed pumped-storage unit;

[0015] The matrix construction module is used to perform Clark coordinate transformation on the three-phase current signal and construct the Hankel matrix of the complex space vector based on the coordinate transformation result, which serves as the time-series observation data matrix; the time-series observation data matrix is ​​used to describe the time shift information and frequency mode information of the three-phase current signal.

[0016] The spatial decomposition module is used to perform spatial decomposition based on the time-series observation data matrix to obtain the main frequency subspace and harmonic subspace.

[0017] The main frequency estimation module is used to determine the direct estimate of the main frequency from the main frequency subspace based on frequency continuity constraints and pole magnitude constraints; the direct estimate of the main frequency is used to describe the fundamental frequency variation in the three-phase current signal;

[0018] The harmonic identification module is used to perform adaptive harmonic identification based on the direct estimate of the main frequency and the harmonic subspace to obtain the target harmonic frequency; the target harmonic frequency is used to describe the transient broadband interference in the three-phase current signal;

[0019] The output module is used to perform joint estimation based on the direct estimate of the main frequency, the target harmonic frequency, and the time-series observation data matrix to obtain the transient signal estimation result of the rotor current; the transient signal estimation result includes at least the signal frequency and the signal amplitude.

[0021] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method.

[0022] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0023] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method.

[0024] The aforementioned method, apparatus, and equipment for rapid estimation of rotor current in a doubly-fed variable-speed pumped-storage unit based on dual-subspace first performs Clark coordinate transformation on the three-phase current on the rotor side and constructs a Hankel matrix as the time-series observation data matrix. Then, through spatial decomposition, the signal is split into independent primary frequency subspaces and harmonic subspaces, achieving decoupling between the fundamental frequency and broadband harmonic interference, avoiding coupling pollution of fundamental frequency parameters by harmonic components, and improving the overall robustness of signal estimation. Subsequently, the direct estimate of the primary frequency is determined using pole magnitude constraints and frequency continuity constraints, effectively filtering out noise. The acoustic signature introduces false poles and suppresses frequency jumps during transient processes, ensuring a continuous and stable output frequency sequence and providing a stable and reliable frequency reference for subsequent identification. Then, adaptive harmonic identification is performed based on the direct estimate of the dominant frequency, achieving the separation of various transient broadband interference components. Finally, joint estimation is constructed using the direct estimate of the dominant frequency, harmonic frequencies, and time-series observation matrices to avoid harmonic energy absorption problems associated with single-fundamental fitting, accurately extracting the fundamental signal frequency and amplitude. Based on the determined signal frequency and amplitude, rapid overcurrent protection of the unit rotor is implemented. In this process, the complete workflow of hierarchical subspace decoupling, dual-constraint frequency screening, and multi-component joint estimation effectively improves the accuracy of frequency and amplitude estimation results under complex transient conditions, thereby achieving rapid overcurrent protection for the pumped storage unit rotor. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is an application environment diagram of a fast rotor current estimation method for a doubly-fed variable-speed pumped-storage unit based on dual subspace in one embodiment.

[0027] Figure 2 This is a flowchart illustrating a method for rapid estimation of rotor current in a doubly-fed variable-speed pumped-storage unit based on dual subspace in one embodiment.

[0028] Figure 3 This is a schematic diagram illustrating the construction process of a time-series observation data matrix in one embodiment;

[0029] Figure 4 This is a schematic diagram of the three-phase rotor current flow during a short-circuit fault in one embodiment.

[0030] Figure 5 This refers to the main frequency and three-phase amplitude of the three-phase rotor current in one embodiment;

[0031] Figure 6 This is a schematic diagram of instantaneous current fitting in one embodiment;

[0032] Figure 7 This is a structural block diagram of a rapid rotor current estimation device for a doubly fed variable speed pumped-storage unit based on bi-subspace in one embodiment.

[0033] Figure 8 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0035] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0036] The fast rotor current estimation method for doubly-fed variable-speed pumped-storage units based on twin-subspace provided in this application can be applied to, for example... Figure 1 The application environment shown is illustrated. Terminal 101 communicates with server 102 via a network. A data storage system can store the data that server 102 needs to process. The data storage system can be integrated onto server 102, or it can be located in the cloud or on another network server.

[0037] The unit management personnel initiate a fault protection task for the doubly-fed variable-speed pumped-storage unit through terminal 101. After receiving the fault protection task, server 102 acquires the three-phase current signals on the rotor side of the doubly-fed variable-speed pumped-storage unit in the event of an operational fault. It performs a Clarke coordinate transformation on the three-phase current signals and constructs a Hankel matrix of complex space vectors based on the transformation results, which serves as the time-series observation data matrix. This time-series observation data matrix describes the time shift information and frequency mode information of the three-phase current signals. Based on the time-series observation data matrix, spatial decomposition is performed to obtain the dominant frequency subspace. Harmonic subspace; based on frequency continuity constraints and pole magnitude constraints, the direct estimate of the dominant frequency is determined from the dominant frequency subspace; the direct estimate of the dominant frequency is used to describe the fundamental frequency variation in the three-phase current signal; adaptive harmonic identification is performed based on the direct estimate of the dominant frequency and the harmonic subspace to obtain the target harmonic frequency; the target harmonic frequency is used to describe the transient broadband interference in the three-phase current signal; based on the direct estimate of the dominant frequency, the target harmonic frequency, and the time-series observation data matrix, a joint estimation is performed to obtain the transient signal estimation result of the rotor current; the transient signal estimation result includes at least the signal frequency and the signal amplitude.

[0038] The terminal 101 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, etc. The server 102 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services.

[0039] In one exemplary embodiment, such as Figure 2 As shown, a fast estimation method for rotor current of a doubly-fed variable-speed pumped-storage unit based on dual subspace is provided, and this method is applied to... Figure 1Taking the server in the example, the explanation includes the following steps 201 to 207. Wherein:

[0040] Step 201: Obtain the three-phase current signal on the rotor side of the doubly-fed variable-speed pumped-storage unit.

[0041] Among them, the rotor side of the doubly-fed variable speed pumped storage unit refers to the electrical circuit where the rotor winding of the doubly-fed variable speed pumped storage unit and the rotor-side converter connected thereto are located; furthermore, the three-phase current on the rotor side is the three-phase AC current connected to the rotor-side converter of the rotor winding, which is the core electrical observation reflecting the transient operating status of the unit and the impact of grid fault disturbances.

[0042] The three-phase current signal includes the three discrete time-domain currents of phase a, phase b, and phase c corresponding to the three-phase windings on the rotor side. Under normal operating conditions, the three-phase current signal exhibits a three-phase symmetrical sinusoidal characteristic.

[0043] It should be noted that, in the embodiments of this application, the three-phase current signal on the rotor side of the doubly-fed variable-speed pumped-storage unit refers to the fault transient current collected when a fault occurs in the unit. During a fault transient, non-stationary changes such as amplitude distortion, phase shift, harmonic superposition, and three-phase imbalance occur, causing the fault transient current to typically contain fundamental, harmonic, attenuated DC components, and noise. The discrete-time model of the three-phase current signal can be expressed as:

[0044] ;

[0045] ;

[0046] .

[0047] in, The amplitude of the k-th frequency component; For the corresponding frequency; This is the initial phase; To attenuate the DC component; Noise term; The sampling period is n; n is the index of the discrete sampling point. , and These represent the three discrete time-domain currents of phase a, phase b, and phase c, respectively, and K represents the total number of independent sinusoidal modes within the current signal.

[0048] In some embodiments, the server can use the fault signal of the doubly fed variable speed pumped storage unit as the starting time for data acquisition and collect the three-phase current signal on the rotor side through a sensor acquisition device; wherein, the fault signal can be the tripping signal of the unit's protection switch or a signal triggered when the real-time acquisition value of a preset electrical indicator is greater than a preset threshold.

[0049] Step 202: Perform Clark coordinate transformation on the three-phase current signal, and construct the Hankel matrix of the complex space vector based on the coordinate transformation result as the time-series observation data matrix; the time-series observation data matrix is ​​used to describe the time shift information and frequency mode information of the three-phase current signal.

[0050] Clarke coordinate transformation refers to the conversion of the coupled rotor-side three-phase current in a three-phase coordinate system into two independent real current components in a two-phase stationary orthogonal coordinate system by relying on a fixed orthogonal transformation matrix, thereby eliminating the natural phase coupling constraint of the three phases and separating the positive-sequence and negative-sequence disturbance components.

[0051] A complex space vector is a vector obtained by uniformly encapsulating two sets of independent real current components obtained by the Clark coordinate transformation in complex form.

[0052] The Hankel matrix is ​​a special structured matrix in which the sampling time sequence of each row is sequentially extended and the sampling values ​​on each anti-diagonal line are kept consistent. It is used to batch fill time-by-time complex vector sampling data in a delayed arrangement.

[0053] It is understandable that, due to the coupling relationship between the three-phase current signals, direct frequency estimation of a single phase is easily affected by negative sequence and zero sequence components. Therefore, the Clark coordinate transformation is used to convert it into two-phase stationary coordinates, and then a complex space vector is constructed to decouple the three-phase signals, while reflecting the degree of system imbalance during fault transients.

[0054] Meanwhile, the Hankel matrix allows for the orderly arrangement of discrete sampling data at different delay positions, preserving the dynamic evolution of the signal throughout the entire fault process. This enables the obtained time-series observation data matrix to present the time-shift characteristics of the signal and the inherent distribution characteristics of various frequency modes in a standardized manner, providing a standardized matrix input for subsequent processing and reducing interference from three-phase imbalance and measurement noise.

[0055] Step 203: Perform spatial decomposition based on the time-series observation data matrix to obtain the main frequency subspace and harmonic subspace.

[0056] Spatial decomposition refers to splitting the eigenvectors from the time-series observation data matrix that carries all signal components according to the strength of the signal energy, thereby obtaining multiple independent subspaces.

[0057] The main frequency subspace is a matrix formed by splicing together a set of eigenvectors with the highest energy within the overall signal subspace, representing the fundamental main signal component with the highest energy proportion under fault conditions.

[0058] The harmonic subspace is a matrix formed by splicing together subsequent eigenvectors of intermediate energy levels, representing broadband disturbance components such as harmonics and interharmonics induced by faults.

[0059] In some embodiments, the time series observation data matrix can be spatially decomposed by global singular value decomposition (SVD) to obtain the dominant frequency subspace and harmonic subspace. For example, singular value decomposition is performed on the complete time series observation matrix, the eigenvectors are sorted according to the singular values ​​from largest to smallest, the eigenvectors are segmented according to the modal energy range, and the dominant frequency subspace and harmonic subspace are assembled respectively, while the noise vectors with corresponding weak singular values ​​are discarded.

[0060] In other embodiments, the time-series observation data matrix can be spatially decomposed using the Projection Approximation Subspace Tracking (PAST) algorithm to obtain the dominant frequency subspace and harmonic subspace. For example, a preset forgetting factor is configured to iteratively update the signal subspace window by window, and the overall signal subspace matrix is ​​output in real time in a recursive manner. Then, feature vectors are extracted in layers according to the modal energy level, and two types of dedicated subspaces, namely dominant frequency and harmonic, are constructed in blocks.

[0061] Understandably, in related technologies, a unified subspace is typically used to simultaneously estimate the fundamental and harmonic modes. When harmonic energy increases during transient processes, the dominant frequency mode is easily contaminated, leading to frequency misjudgment. However, by spatially decomposing the time-series observation data matrix, a dominant frequency subspace and a harmonic subspace are obtained. The dominant frequency subspace is used only to track the fundamental dominant mode, while the harmonic subspace is used to identify transient interference harmonics. By modeling the dominant frequency and harmonics separately in different subspaces, the coupling effect of harmonics on the dominant frequency estimation can be effectively reduced, and the frequency stability during fault transients can be improved.

[0062] Step 204: Based on frequency continuity constraints and pole magnitude constraints, determine the direct estimate of the main frequency from the main frequency subspace; the direct estimate of the main frequency is used to describe the fundamental frequency variation in the three-phase current signal.

[0063] Among them, the frequency continuity constraint refers to the constraint on the difference between adjacent candidate mode frequencies among multiple candidate mode frequencies contained in the main frequency subspace. It can be understood that for adjacent candidate mode frequencies, the fundamental frequency under real operating conditions will not undergo drastic changes. Therefore, the smaller the difference between adjacent candidate mode frequencies, the lower the possibility that the candidate mode frequency belongs to the abnormal frequency hopping result caused by noise or pseudo poles.

[0064] The pole magnitude constraint means that the poles corresponding to the effective sinusoidal signals will be close to the unit circle in the complex plane, while spurious poles will deviate significantly from the unit circle due to noise disturbances. Therefore, the greater the deviation of the candidate mode frequency from the unit circle, the more likely the candidate mode frequency is a noise-derived spurious mode pole.

[0065] The direct estimate of the main frequency refers to the optimal window fundamental frequency result determined after screening by frequency continuity constraints and pole magnitude constraints. Furthermore, the direct estimate of the main frequency represents the instantaneous frequency of the fundamental component of the rotor current under fault transient conditions. By determining the direct estimate of the main frequency, a benchmark reference can be provided for subsequent harmonic identification, and data support can also be provided for subsequent overcurrent protection of the unit rotor.

[0066] In some embodiments, a scoring model can be constructed based on frequency continuity constraints and pole magnitude constraints. The scoring model is used to determine the scores of multiple candidate main frequencies in the main frequency subspace, and then the candidate main frequency with the highest score is determined as the direct estimate of the main frequency.

[0067] Step 205: Adaptive harmonic identification is performed based on the direct estimate of the main frequency and the harmonic subspace to obtain the target harmonic frequency; the target harmonic frequency is used to describe the transient broadband interference in the three-phase current signal.

[0068] Among them, harmonics are various broadband disturbance components in three-phase current that are distinct from the fundamental frequency and are generated by transient impacts of power grid faults and nonlinear loads of lines.

[0069] In this embodiment of the application, adaptive harmonic identification refers to automatically filtering out harmonic components that meet the deviation conditions based on the direct estimate of the dominant frequency. For example, the process of adaptive harmonic identification may be to first read all candidate harmonic frequencies obtained by harmonic subspace analysis, calculate the frequency difference between each candidate harmonic frequency and the direct estimate of the dominant frequency, retain the candidate results whose absolute difference with the direct estimate of the dominant frequency is greater than a preset discrimination threshold, and obtain the target harmonic frequency.

[0070] Understandably, adaptive harmonic identification can overcome the limitations of traditional fixed-parameter notch filters, which can only adapt to preset integer harmonics. It can adapt to the dynamic drift of harmonic frequencies and mode crossover during the transient stage of a fault, automatically complete the dynamic detection of broadband disturbances, and provide a more accurate reference for subsequent harmonic component suppression.

[0071] Step 206: Based on the direct estimate of the main frequency, the target harmonic frequency, and the time-series observation data matrix, a joint estimation is performed to obtain the transient signal estimation result of the rotor current; the transient signal estimation result includes at least the signal frequency and the signal amplitude.

[0072] The signal frequency includes the fundamental frequency and harmonic frequency, and the signal amplitude includes the fundamental amplitude and harmonic amplitude.

[0073] In some embodiments, a joint orthogonal basis matrix can be constructed first based on the direct estimate of the dominant frequency and the target harmonic frequency to resolve the signal coupling problem between the fundamental frequency and broadband harmonic disturbances, thereby avoiding distortion in the fundamental frequency amplitude measurement caused by harmonic disturbances. Then, least squares projection is performed based on the joint orthogonal basis matrix and the time-series observation data matrix. Matrix fitting is performed on the three-phase current samples sampled in windows within the time-series observation data matrix to solve for the amplitude parameters corresponding to the dominant frequency component and each harmonic component. Finally, the complete transient signal estimation result of the rotor current is obtained by combining the direct estimate of the dominant frequency and the target harmonic frequency.

[0074] In some embodiments, rapid overcurrent protection for the rotor of a pumped storage unit can be achieved based on transient signal estimation results.

[0075] In the aforementioned rapid estimation method for doubly-fed variable-speed pumped-storage units, the three-phase current on the rotor side is first subjected to Clark coordinate transformation and a Hankel matrix is ​​constructed as the time-series observation data matrix. Then, through spatial decomposition, the signal is split into independent primary frequency subspaces and harmonic subspaces, achieving decoupling of the fundamental frequency and broadband harmonic interference, avoiding coupling pollution of fundamental parameter estimation by harmonic components, and improving the overall robustness of signal estimation. Next, the direct estimate of the primary frequency is determined using pole magnitude constraints and frequency continuity constraints, effectively filtering out spurious poles caused by noise, suppressing frequency jumps during transient processes, ensuring a continuous and stable output frequency sequence, and providing a stable and reliable frequency reference for subsequent identification. Then, adaptive harmonic identification is completed based on the direct estimate of the primary frequency, achieving the separation of various transient broadband interference components. Finally, a joint estimation is constructed by combining the direct estimate of the primary frequency, harmonic frequencies, and the time-series observation matrix, avoiding the harmonic energy absorption problem existing in single-fundamental fitting, accurately extracting the fundamental signal frequency and signal amplitude, and realizing rapid overcurrent protection of the unit rotor based on the determined signal frequency and signal amplitude. In the above process, the complete process of hierarchical subspace decoupling, dual-constraint frequency screening, and multi-component joint estimation effectively improves the accuracy of frequency and amplitude estimation results under complex transient conditions, thereby achieving rapid overcurrent protection for the rotor of pumped storage units.

[0076] In one exemplary embodiment, such as Figure 3 As shown, the construction process of the time-series observation data matrix includes steps 301 to 303. Wherein:

[0077] Step 301: The three-phase current signal is transformed from three-phase stationary coordinates to two-phase stationary coordinates through Clark coordinate transformation to obtain two-phase orthogonal current components.

[0078] For example, the server transforms the three-phase current signal into a two-phase orthogonal coordinate system by performing a Clarke coordinate transformation, as shown in the following expression:

[0079] .

[0080] in, , and These represent the three discrete-time currents of phase a, phase b, and phase c, respectively. and These are the two-phase orthogonal current components in an orthogonal coordinate system.

[0081] Step 302: Perform a composite operation of the real and imaginary parts based on the two-phase orthogonal current components to obtain a complex space vector signal.

[0082] For example, it can be based on and Constructing complex space vectors ,Right now ;in, It is the imaginary unit.

[0083] Understandably, under ideal equilibrium conditions, the trajectory of the complex space vector is a standard circle; however, when an imbalance fault occurs in the system, its trajectory gradually deviates from the circle and exhibits elliptical characteristics. Therefore, the complex space vector can not only decouple the three-phase signals, but also reflect the degree of system imbalance during the fault transient period.

[0084] Step 303: Construct the Hankel matrix based on the complex space vector signal to obtain the time-series observation data matrix.

[0085] For example, the Hankel matrix, which is also the time-series observation data matrix, is constructed based on complex space vectors, and its expression is as follows:

[0086] .

[0087] in, For data length, The number of rows in the matrix. is the number of columns in the matrix.

[0088] In some embodiments, considering that the choice of matrix dimension affects the algorithm's adaptability to noise and short-window data: too few rows will weaken the subspace resolution, while too many rows may increase the computational load and ill-conditioned risk, L=N-floor(N / 3) and K=floor(N / 3)+1 can be set. This preserves enough rows to form a shifted subspace and ensures that the number of columns can cover multiple delay segments within the window. Here, floor is the floor function.

[0089] In this embodiment, the dimensionality of the three-phase current signal is reduced by Clarke transformation, and the sampling data at different delay times are encapsulated in an orderly manner by Hankel matrix, which fully preserves the hidden frequency modes and time-series shift characteristics of the signal. This provides a time-aligned and reliable matrix input for subsequent subspace recursive decomposition and main frequency harmonic decoupling identification, reducing the identification error caused by three-phase imbalance and phase shift in subsequent algorithms.

[0090] In some embodiments, spatial decomposition is performed based on the time-series observation data matrix to obtain the dominant frequency subspace and harmonic subspace, including:

[0091] Based on a preset forgetting factor and a projection approximation subspace tracking algorithm, the time-series observation data matrix is ​​spatially recursively tracked to obtain the overall signal subspace matrix; the overall signal subspace matrix is ​​then decomposed based on modal energy to obtain the main frequency subspace and harmonic subspace.

[0092] For example, the projection approximation subspace tracking algorithm obtains the overall signal subspace matrix by minimizing the cost function. , where the cost function It can be represented as: .

[0093] in, Represents the preset forgetting factor. This represents the complex space vector data at the nth sampling time in the time-series observation data matrix. This represents the global signal subspace matrix to be solved. This represents the conjugate transpose of the overall signal subspace matrix. The projected residual vector represents the noise and residual perturbation components that cannot be represented by the current estimation subspace. This represents the 2-norm operation of vectors.

[0094] The process of spatially recursively calculating the time-series observation data matrix based on a preset forgetting factor and a projection approximation subspace tracking algorithm may include:

[0095] For the data in the k-th column x of the time-series observation data matrix (i.e., the Hankel matrix) k Its projection onto the signal subspace is represented as This allows us to determine the reconstruction error. Through error Update the overall signal subspace at step k+1, i.e. ,in, Let be the gain vector at step k. ,in, Represents the preset forgetting factor. Let be the covariance inverse matrix at step k, which follows the update formula. .

[0096] In some embodiments, the overall signal subspace matrix can be arranged according to modal energy, and the main frequency subspace and harmonic subspace can be obtained according to a preset partitioning rule.

[0097] For example, determine the dimension r of the dominant frequency subspace. main =2, the dimension of the harmonic subspace is r. h =8, meaning that after arranging the overall signal subspace matrix according to modal energy from high to low, the first two columns are taken to form the main frequency subspace, i.e., the main frequency subspace. It can be represented as L*r main The matrix; the eight consecutive columns form the harmonic subspace, i.e., the harmonic subspace. It can be represented as L*r h , where L is the number of rows in the overall signal subspace matrix, which is also the number of rows in the time series observation data matrix.

[0098] In the above embodiments, the overall signal subspace matrix is ​​updated iteratively by setting a projection approximation subspace tracking algorithm with a preset forgetting factor. This eliminates the need to perform computationally intensive global singular value decomposition on the entire time-series observation data. At the same time, the forgetting factor can weaken the weight of old sampled data, adapting to the characteristics of fault condition signals that evolve dynamically over time.

[0099] In some embodiments, based on frequency continuity constraints and pole magnitude constraints, the direct estimate of the dominant frequency is determined from the dominant frequency subspace, including:

[0100] The dominant frequency subspace is split by shifting to obtain multiple dominant frequency shift subspace matrices. The multiple dominant frequency shift subspace matrices are weighted based on the first preset weight to obtain the dominant frequency state matrix corresponding to the dominant frequency subspace. Based on the frequency continuity constraint and the pole magnitude constraint, the direct estimate of the dominant frequency is determined from the dominant frequency state matrix.

[0101] For example, the dominant frequency subspace can be shifted and split to obtain two dominant frequency shifted subspace matrices corresponding to the dominant frequency subspace. and ,in, This represents extracting the first line to the second-to-last line. This means extracting the last line of the first row. This is the total number of rows in the main frequency subspace, which is also the total number of rows in the time-series observation data matrix. Represents the main frequency subspace.

[0102] In some embodiments, the main frequency state matrix It can be obtained through the following calculation formula: .

[0103] in, and These are the two main frequency shift subspace matrices corresponding to the main frequency subspace. This is the first preset weight.

[0104] Furthermore, in some embodiments, the first preset weight can be a fourth-power weight, i.e., the first preset weight. ,in, Represents the diagonal matrix operator. This is the total number of rows in the main frequency subspace.

[0105] In some embodiments, multiple signal pole estimates and candidate frequency values ​​corresponding to each signal pole estimate can be determined based on the frequency state matrix. A scoring function is constructed through frequency continuity constraints and pole magnitude constraints. Each candidate frequency value is scored using the scoring function. Finally, the direct estimate of the frequency is determined from multiple candidate frequencies based on the score of each candidate frequency value.

[0106] In the above embodiments, the main frequency state matrix is ​​constructed by performing weighted operations on each group of shift subspace matrices according to the first preset weight, which can weaken the mode shift problem caused by single-point sampling noise and local three-phase imbalance disturbance. The main frequency direct estimate is determined by pole magnitude constraint and frequency continuity constraint. On the one hand, the inherent signal characteristics of poles close to the unit circle are used to eliminate noise-induced pseudo-pole modes. On the other hand, the time prior that the frequency of adjacent windows will not change drastically suppresses the unreasonable main frequency jump phenomenon in the fault transient process, thereby improving the recognition accuracy of the main frequency direct estimate.

[0107] In some embodiments, based on frequency continuity constraints and pole magnitude constraints, the direct estimate of the dominant frequency is determined from the dominant frequency state matrix, including:

[0108] Based on the eigenvalues ​​of the dominant frequency state matrix, multiple candidate modal frequencies corresponding to the dominant frequency subspace are determined; based on frequency continuity constraints and pole magnitude constraints, the target modal frequency is determined from the multiple candidate modal frequencies; based on the direct estimate of the historical dominant frequency, the target modal frequency is subjected to continuous smoothing and jump suppression to obtain the direct estimate of the dominant frequency.

[0109] In some embodiments, the eigenvalues ​​of the dominant frequency state matrix can be used as signal pole estimates, and the corresponding candidate mode frequencies can be obtained further based on the pole phase angles.

[0110] For example, determining the main frequency state matrix eigenvalues , , Representing the poles, then determining the eigenvalues. Corresponding candidate mode frequencies , ;in, Represents the signal sampling period. Representative eigenvalues Argument angle.

[0111] In some embodiments, a comprehensive scoring function can be constructed based on frequency continuity constraints and pole magnitude constraints. Each candidate modal frequency is scored using the comprehensive scoring function, and the target modal frequency is determined based on the scores of each candidate modal frequency.

[0112] For example, a comprehensive scoring function can be constructed. , ,in, and To preset weights, This indicates a pole magnitude constraint. , This indicates a frequency continuity constraint. , Represents the frequency of the k-th candidate mode. This represents the frequency of the (k-1)th candidate mode.

[0113] Then, the candidate modal frequency with the highest score is selected as the target modal frequency, i.e., the target modal frequency. .

[0114] Clearly, the comprehensive scoring function reflects, on the one hand, that the poles of the ideal steady-state sinusoidal mode should be close to the unit circle; on the other hand, it reflects that the modal frequencies of adjacent windows should not undergo unreasonable abrupt changes. Finally, the candidate frequency with the largest score is selected as the direct estimate of the dominant frequency, thereby prioritizing the selection of the dominant mode that is close to the historical frequency, thus effectively suppressing frequency jumps.

[0115] In some embodiments, an adaptive smoothing coefficient can be calculated based on the difference between the target modal frequency and the direct estimate of the historical dominant frequency. This allows the output to inherit more historical values, i.e., be closer to the direct estimate of the historical dominant frequency, when the difference is small. When the difference is large but still within a reasonable range, the output is allowed to move closer to the target modal frequency.

[0116] For example, the direct estimate of the clock frequency can be expressed as:

[0117] .

[0118] in, This represents the direct estimate of the main frequency corresponding to the current time window k. This represents the direct estimate of the historical dominant frequency corresponding to the previous time window k-1. Represents the smoothing coefficient. This represents a preset difference threshold. This represents the target modal frequency corresponding to the current time window k.

[0119] in, , ,in, This represents the historical target mode frequency corresponding to the previous time window k-1.

[0120] In the above embodiments, continuous smoothing and jump suppression are carried out by combining the direct estimation value of historical main frequency, which effectively weakens the instantaneous main frequency spike jump of the fault, improves the stability and measurement accuracy of the fundamental frequency timing output, and provides a stable main frequency reference for subsequent processing.

[0121] In some embodiments, adaptive harmonic identification is performed based on the direct estimate of the dominant frequency and the harmonic subspace to obtain the target harmonic frequency, including:

[0122] The harmonic subspace is split by shifting to obtain multiple harmonic shift subspace matrices. The multiple harmonic shift subspace matrices are weighted based on a second preset weight to obtain the harmonic state matrix corresponding to the harmonic subspace. Based on the eigenvalues ​​of the harmonic state matrix, multiple candidate harmonic frequencies corresponding to the harmonic subspace are determined. Based on a preset discrimination threshold and the frequency difference between the direct estimate of the dominant frequency and the multiple candidate harmonic frequencies, the target harmonic frequency is determined from the multiple candidate harmonic frequencies.

[0123] For example, the harmonic subspace can be shifted and decomposed to obtain two dominant frequency shifted subspace matrices corresponding to the harmonic subspace. and ,in, This represents extracting the first line to the second-to-last line. This means extracting the last line of the first row. This represents the total number of rows in the harmonic subspace, which is also the total number of rows in the time-series observation data matrix. Represents the harmonic subspace.

[0124] In some embodiments, the harmonic state matrix It can be obtained through the following calculation formula: .

[0125] in, and These are the two dominant frequency shift subspace matrices corresponding to the harmonic subspace. This is the second preset weight.

[0126] Furthermore, in some embodiments, the second preset weight may be a quadratic weight, i.e., the second preset weight. ,in, Represents the diagonal matrix operator. This is the total number of rows in the main frequency subspace.

[0127] In some embodiments, the eigenvalues ​​of the harmonic state matrix can be used as signal pole estimates, and the corresponding candidate mode frequencies can be obtained based on the pole phase angles.

[0128] For example, determining the harmonic state matrix eigenvalues , , Representing the poles, then determining the eigenvalues. Corresponding candidate harmonic frequencies , ;in, Represents the signal sampling period. Representative eigenvalues Argument angle.

[0129] For example, the target harmonic frequency can be determined by the following expression:

[0130] .

[0131] in, This represents the target harmonic frequency obtained through screening. Represents candidate harmonic frequencies. This represents the direct estimate of the main frequency, and 0.5 is the preset discrimination threshold.

[0132] In the above embodiments, the timing noise interference is suppressed by shifting and splitting the harmonic subspace and constructing a harmonic state matrix by weighting. Then, the candidate harmonic frequencies are screened by using the direct estimation value of the main frequency as a benchmark and a threshold. This can eliminate false modes generated by pseudo poles, decouple and complete harmonic identification, reduce the probability of harmonic missed detection and misjudgment, and improve the identification accuracy of harmonic frequencies under fault transient conditions.

[0133] In some embodiments, a joint estimation is performed based on the direct estimate of the main frequency, the target harmonic frequency, and the time-series observation data matrix to obtain the transient signal estimation result of the rotor current, including:

[0134] A joint orthogonal basis matrix is ​​constructed based on the direct estimate of the main frequency and the target harmonic frequency; regularized least squares estimation is performed based on the time-series observation data matrix and the joint orthogonal basis matrix to obtain the initial estimation result; the initial estimation result is smoothed based on adaptive Kalman filtering to obtain the transient signal estimation result of the rotor current.

[0135] For example, let's take phase A of a three-phase current as an example for illustration, and its signal of one frequency. It can be represented as ,in, , This represents a direct estimate of the clock speed. Representing the harmonic frequencies of each target, and Representing the cosine and sine coefficients respectively, a joint orthogonal basis matrix is ​​constructed based on this. ,in:

[0136] ;

[0137] .

[0138] Represents the fundamental submatrix, which is arranged according to N sampling times. arrive Substitute the direct estimated value of the main frequency into each column and fill in the values ​​of cosine and sine. Each row corresponds to a single sampling time, and the two columns store the calculation results of cosine and sine respectively. This represents the i-th harmonic sub-matrix, which is similar to the fundamental sub-matrix, except that the corresponding target harmonic frequency needs to be substituted in.

[0139] For example, regularized least squares estimation can be expressed as ,in, The regularization coefficient is... It is the identity matrix. This is the observation vector corresponding to the phase A current in the time-series observation data matrix.

[0140] The initial estimation results may include the fundamental amplitude of the signal. , and They are respectively The included fundamental cosine component coefficients and fundamental sine component coefficients, and harmonic amplitude estimation results. , and They are respectively The cosine component coefficients and sine component coefficients corresponding to the nth harmonic are included.

[0141] For example, Kalman filtering can be used to smooth the initial estimation results to smooth the amplitude output and obtain the transient signal estimation results.

[0142] The Kalman state equation is as follows:

[0143] ;

[0144] ;

[0145] ;

[0146] .

[0147] in, The prior covariance is the predicted amplitude error value obtained from the previous calculation. This is the posterior covariance after the previous update. This is the process noise variance, used to characterize the uncertainty of the amplitude itself as it naturally fluctuates with operating conditions; For Kalman gain, To measure the noise variance and characterize the observation errors introduced during the sampling and identification processes, This is the output after k-filtering correction for the current time window. This is the output after filtering by k-1 in the previous time window. This is the original amplitude calculated by least squares, i.e., the initial estimate.

[0148] The above filtering can effectively suppress drastic fluctuations in frequency and amplitude caused by fault impacts, noise, and model mismatch.

[0149] Furthermore, the filtering process can be further optimized using adaptive Kalman filtering, taking into account both "tracking capability" and "noise resistance," for example: ; ; .

[0150] in, Represents the adaptive weighting coefficient. This represents the difference between the direct estimate of the main frequency of the current window and the previous window. A preset baseline value representing the variance of process noise. This represents the process noise variance after dynamic adaptive updating. This represents a preset reference value representing the variance of the measured noise. This represents the measurement noise variance after dynamic adaptive updating.

[0151] When the frequency change is small When the value is close to 1, Q and R are close to their base values, resulting in relatively stable filtering. However, when the frequency changes significantly, As the frequency decreases, both process noise and measurement noise increase. The physical meaning is that when the frequency changes significantly, the system may enter a transient state, and the amplitude may also change, but the measurement uncertainty also increases.

[0152] In the above embodiments, by building an orthogonal basis based on the identified direct estimate of the main frequency and the target harmonic frequency, and by performing regularized least squares solution in combination with the time series observation data matrix, the amplitude calculation deviation caused by observation noise can be suppressed. Then, by using adaptive Kalman filtering for time series smoothing correction, the interference of random disturbances under fault transients can be further weakened, and the estimation accuracy and time series stability of the transient signal amplitude and phase parameters of rotor current can be improved.

[0153] In some embodiments, in order to verify the effectiveness of the method provided in this application, a model of an actual 336MW doubly-fed variable-speed pumped-storage unit was modeled in the internal fault calculation program of MATLAB.

[0154] The parameters of the modeled doubly-fed variable-speed pumped-storage unit are shown in the table below.

[0155] parameter value capacity 336MW Stator rated voltage 15.75kV Stator / rotor turns ratio 0.414 Stator leakage reactance 0.139pu Stator resistance 0.0018pu Rotor leakage reactance 0.189pu Rotor resistance 0.0018pu Magnetizing Reactor 3.49pu time constant 9.46s

[0156] When a short-circuit fault occurs between the first turn of branch 1 of phase A and the fifth turn of branch 2 of phase A on the rotor side at a rotor frequency of 1Hz, the three-phase rotor current is as follows: Figure 4 As shown, the three-phase rotor current main frequency and three-phase amplitude are then calculated using the method provided in this application with a data window of 0.1s (i.e., 0.1 cycles). Figure 5 As shown, the instantaneous current fitting situation is as follows: Figure 6 As shown, the method provided in this application can accurately calculate the current frequency and amplitude even within a data window of 1 / 10 of a cycle. The current value fitted by the calculation results is close to the actual value, thereby enabling rapid measurement of low-frequency signals on the rotor side of the doubly-fed variable-speed pumped-storage unit. This achieves maximum grid voltage support and ensures hydraulic stability, while strictly suppressing rotor overvoltage and overcurrent within the hardware safety rating range.

[0157] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0158] Based on the same inventive concept, this application also provides a rapid estimation device for a doubly-fed variable-speed pumped-storage unit, used to implement the rapid estimation method for the doubly-fed variable-speed pumped-storage unit described above. The solution provided by this device is similar to the implementation described in the above method. Therefore, the specific limitations in one or more embodiments of the rapid estimation device for doubly-fed variable-speed pumped-storage units provided below can be found in the limitations of the rapid estimation method for doubly-fed variable-speed pumped-storage units described above, and will not be repeated here.

[0159] In one exemplary embodiment, such as Figure 7 As shown, a fast estimation device for rotor current of a doubly-fed variable-speed pumped-storage unit based on dual-subspace is provided. The fast estimation device 700 for rotor current of a doubly-fed variable-speed pumped-storage unit based on dual-subspace includes:

[0160] The acquisition module 701 is used to acquire the three-phase current signal on the rotor side of the doubly-fed variable-speed pumped-storage unit.

[0161] The matrix construction module 702 is used to perform Clark coordinate transformation on the three-phase current signal and construct the Hankel matrix of the complex space vector based on the coordinate transformation result, which serves as the time-series observation data matrix. The time-series observation data matrix is ​​used to describe the time shift information and frequency mode information of the three-phase current signal.

[0162] The spatial decomposition module 703 is used to perform spatial decomposition based on the time-series observation data matrix to obtain the main frequency subspace and harmonic subspace.

[0163] The main frequency estimation module 704 is used to determine the direct estimate of the main frequency from the main frequency subspace based on frequency continuity constraints and pole magnitude constraints; the direct estimate of the main frequency is used to describe the fundamental frequency variation in the three-phase current signal.

[0164] The harmonic identification module 705 is used to perform adaptive harmonic identification based on the direct estimate of the main frequency and the harmonic subspace to obtain the target harmonic frequency; the target harmonic frequency is used to describe the transient broadband interference in the three-phase current signal.

[0165] The output module 706 is used to perform joint estimation based on the direct estimate of the main frequency, the target harmonic frequency, and the time-series observation data matrix to obtain the transient signal estimation result of the rotor current; the transient signal estimation result includes at least the signal frequency and the signal amplitude.

[0166] In some embodiments, the matrix construction module 702 is used to transform the three-phase current signal from three-phase stationary coordinates to two-phase stationary coordinates through Clark coordinate transformation to obtain two-phase orthogonal current components; perform real and imaginary part composite operation based on the two-phase orthogonal current components to obtain complex space vector signal; and construct Hankel matrix based on complex space vector signal to obtain time series observation data matrix.

[0167] In some embodiments, the spatial decomposition module 703 is used to perform spatial recursive tracking of the time-series observation data matrix based on a preset forgetting factor and a projection approximation subspace tracking algorithm to obtain the overall signal subspace matrix; and to decompose the overall signal subspace matrix based on modal energy to obtain the main frequency subspace and harmonic subspace.

[0168] In some embodiments, the main frequency estimation module 704 is used to perform shifting and splitting processing on the main frequency subspace to obtain multiple main frequency shifted subspace matrices corresponding to the main frequency subspace; to weight the multiple main frequency shifted subspace matrices based on a first preset weight to obtain the main frequency state matrix corresponding to the main frequency subspace; and to determine the direct estimate of the main frequency from the main frequency state matrix based on frequency continuity constraints and pole magnitude constraints.

[0169] In some embodiments, the dominant frequency estimation module 704 is used to determine multiple candidate modal frequencies corresponding to the dominant frequency subspace based on the eigenvalues ​​of the dominant frequency state matrix; determine the target modal frequency from the multiple candidate modal frequencies based on frequency continuity constraints and pole magnitude constraints; and perform continuous smoothing and jump suppression on the target modal frequency based on the historical dominant frequency direct estimate to obtain the dominant frequency direct estimate.

[0170] In some embodiments, the harmonic identification module 705 is used to perform shifting and splitting processing on the harmonic subspace to obtain multiple harmonic shifted subspace matrices corresponding to the harmonic subspace; to weight the multiple harmonic shifted subspace matrices based on a second preset weight to obtain the harmonic state matrix corresponding to the harmonic subspace; to determine multiple candidate harmonic frequencies corresponding to the harmonic subspace based on the eigenvalues ​​of the harmonic state matrix; and to determine the target harmonic frequency from the multiple candidate harmonic frequencies based on a preset discrimination threshold and the frequency difference between the direct estimate of the dominant frequency and the multiple candidate harmonic frequencies.

[0171] In some embodiments, the output module 706 is used to construct a joint orthogonal basis matrix based on the direct estimate of the main frequency and the target harmonic frequency; perform regularized least squares estimation based on the time-series observation data matrix and the joint orthogonal basis matrix to obtain an initial estimation result; and perform smoothing processing on the initial estimation result based on adaptive Kalman filtering to obtain the transient signal estimation result of the rotor current.

[0172] The modules in the rapid estimation device of the aforementioned doubly-fed variable-speed pumped-storage unit can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0173] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 8 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs in the non-volatile storage media to run. The database stores data related to the doubly-fed variable-speed pumped-storage unit. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a rapid estimation method for the rotor current of a doubly-fed variable-speed pumped-storage unit based on dual-subspace.

[0174] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0175] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0176] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0177] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the method described above.

[0178] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0179] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0180] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0181] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for rapid estimation of rotor current in a doubly-fed variable-speed pumped-storage unit based on dual-subspace, characterized in that, The method includes: Acquire the three-phase current signal on the rotor side of the doubly-fed variable-speed pumped-storage unit; The three-phase current signal is subjected to Clark coordinate transformation, and a Hankel matrix of complex space vector is constructed based on the coordinate transformation result as a time-series observation data matrix; the time-series observation data matrix is ​​used to describe the time shift information and frequency mode information of the three-phase current signal. Spatial decomposition is performed based on the time-series observation data matrix to obtain the main frequency subspace and harmonic subspace; Based on frequency continuity constraints and pole magnitude constraints, a direct estimate of the dominant frequency is determined from the dominant frequency subspace; the direct estimate of the dominant frequency is used to describe the fundamental frequency variation in the three-phase current signal. Adaptive harmonic identification is performed based on the direct estimate of the main frequency and the harmonic subspace to obtain the target harmonic frequency; the target harmonic frequency is used to describe the transient broadband interference in the three-phase current signal; Based on the direct estimate of the main frequency, the target harmonic frequency, and the time-series observation data matrix, a joint estimation is performed to obtain the transient signal estimation result of the rotor current; the transient signal estimation result includes at least the signal frequency and the signal amplitude.

2. The method according to claim 1, characterized in that, The process of performing Clarke coordinate transformation on the three-phase current signals and constructing a Hankel matrix of complex space vectors based on the transformation results, as a time-series observation data matrix, includes: The three-phase current signal is transformed from three-phase stationary coordinates to two-phase stationary coordinates by Clark coordinate transformation to obtain two-phase quadrature current components. A complex space vector signal is obtained by performing a composite operation of the real and imaginary parts based on the two-phase orthogonal current components. Based on the complex space vector signal, a Hankel matrix is ​​constructed to obtain the time-series observation data matrix.

3. The method according to claim 1, characterized in that, The spatial decomposition based on the time-series observation data matrix to obtain the dominant frequency subspace and harmonic subspace includes: Based on a preset forgetting factor and a projection approximation subspace tracking algorithm, the time-series observation data matrix is ​​spatially recursively tracked to obtain the overall signal subspace matrix. The overall signal subspace matrix is ​​decomposed based on modal energy to obtain the main frequency subspace and harmonic subspace.

4. The method according to claim 1, characterized in that, The process of determining the direct estimate of the dominant frequency from the dominant frequency subspace based on frequency continuity constraints and pole magnitude constraints includes: The main frequency subspace is shifted and split to obtain multiple main frequency shift subspace matrices corresponding to the main frequency subspace; Based on a first preset weight, multiple main frequency shift subspace matrices are weighted to obtain the main frequency state matrix corresponding to the main frequency subspace; Based on the frequency continuity constraint and the pole magnitude constraint, the direct estimate of the main frequency is determined from the main frequency state matrix.

5. The method according to claim 4, characterized in that, The process of determining the direct estimate of the dominant frequency from the dominant frequency state matrix based on the frequency continuity constraint and the pole magnitude constraint includes: Based on the eigenvalues ​​of the dominant frequency state matrix, multiple candidate mode frequencies corresponding to the dominant frequency subspace are determined; Based on the frequency continuity constraint and the pole magnitude constraint, the target modal frequency is determined from the multiple candidate modal frequencies; Based on the direct estimate of the historical dominant frequency, the target modal frequency is continuously smoothed and jump-suppressed to obtain the direct estimate of the dominant frequency.

6. The method according to claim 1, characterized in that, The adaptive harmonic identification based on the direct estimate of the dominant frequency and the harmonic subspace to obtain the target harmonic frequency includes: The harmonic subspace is shifted and decomposed to obtain multiple harmonic shift subspace matrices corresponding to the harmonic subspace; Based on the second preset weight, the multiple harmonic shift subspace matrices are weighted to obtain the harmonic state matrix corresponding to the harmonic subspace; Based on the eigenvalues ​​of the harmonic state matrix, multiple candidate harmonic frequencies corresponding to the harmonic subspace are determined. Based on a preset discrimination threshold and the frequency difference between the direct estimate of the main frequency and multiple candidate harmonic frequencies, the target harmonic frequency is determined from the multiple candidate harmonic frequencies.

7. The method according to any one of claims 1 to 6, characterized in that, The method of jointly estimating the transient signal of the rotor current based on the direct estimate of the main frequency, the target harmonic frequency, and the time-series observation data matrix includes: Construct a joint orthogonal basis matrix based on the direct estimate of the dominant frequency and the target harmonic frequency; Based on the time-series observation data matrix and the joint orthogonal basis matrix, regularized least squares estimation is performed to obtain the initial estimation result; Based on adaptive Kalman filtering, the initial estimation results are smoothed to obtain the transient signal estimation results of the rotor current.

8. A rapid estimation device for rotor current of a doubly-fed variable-speed pumped-storage unit based on twin-subspace, characterized in that, The device includes: The acquisition module is used to acquire the three-phase current signal on the rotor side of the doubly-fed variable-speed pumped-storage unit; The matrix construction module is used to perform Clark coordinate transformation on the three-phase current signal and construct the Hankel matrix of the complex space vector based on the coordinate transformation result, which serves as the time-series observation data matrix; the time-series observation data matrix is ​​used to describe the time shift information and frequency mode information of the three-phase current signal. The spatial decomposition module is used to perform spatial decomposition based on the time-series observation data matrix to obtain the main frequency subspace and harmonic subspace. The main frequency estimation module is used to determine the direct estimate of the main frequency from the main frequency subspace based on frequency continuity constraints and pole magnitude constraints; the direct estimate of the main frequency is used to describe the fundamental frequency variation in the three-phase current signal; The harmonic identification module is used to perform adaptive harmonic identification based on the direct estimate of the main frequency and the harmonic subspace to obtain the target harmonic frequency; the target harmonic frequency is used to describe the transient broadband interference in the three-phase current signal; The output module is used to perform joint estimation based on the direct estimate of the main frequency, the target harmonic frequency, and the time-series observation data matrix to obtain the transient signal estimation result of the rotor current; the transient signal estimation result includes at least the signal frequency and the signal amplitude.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images