Coordinated control method of excitation system compatible with double working conditions of power generation and phase modifier

Abstract

This application discloses a collaborative control method for an excitation system compatible with both generator and synchronous condenser operating conditions. By decoupling and reconstructing electromechanical sensor data and performing pattern recognition, it captures the time-varying inertial parameters of the unit during operating condition transitions in real time. Based on this, a state-space observation mechanism based on flux linkage continuity prediction is constructed, directly transforming the evolution of the system's inertial boundary into feedforward compensation voltage. This collaborative control strategy, by introducing dynamic time-varying inertia as a compensation benchmark, can overcome the adjustment dead zone of traditional proportional-integral converters, directionally eliminating flux linkage deviations within milliseconds and effectively suppressing transient reactive power drops. The resulting smooth switching not only significantly improves the control robustness of the excitation system under extreme transient conditions but also fundamentally avoids actuator saturation risks and shaft derailment shutdown hazards, greatly enhancing the voltage disturbance rejection capability of synchronous units in new power systems.

Description

Technical Field

[0001] This application relates to the field of intelligent control, and more specifically, to a method for coordinated control of an excitation system that is compatible with both power generation and synchronous condenser operating conditions. Background Technology

[0002] With the rapid development of new power systems, the high proportion of renewable energy connected to the grid places higher demands on the voltage support capacity of the power grid. This necessitates that synchronous generator units frequently switch dynamically between generation and condenser modes to balance the dual functions of active power transmission and reactive power compensation. Therefore, constructing a coordinated control scheme for the excitation system that is deeply compatible with both operating modes and has smooth switching capabilities has become a core requirement for ensuring the transient stability and operational flexibility of the large power grid.

[0003] However, existing dual-condition excitation control schemes mostly rely on preset logic switching commands or fixed excitation control algorithms optimized for specific conditions. In actual operation, this traditional adaptive adjustment strategy often adopts step-like parameter changes, making it difficult to accurately characterize the highly nonlinear physical characteristics of the unit during the transition between operating modes. Especially at the moment of severe inertia collapse caused by the tripping of the unit's prime mover, existing strategies cannot capture the dynamic evolution trajectory of the system's inertial boundary in real time, resulting in a lack of deep integration of physical laws into the switching process. As a result, during the smooth transition from generator to synchronous condenser dual-condition, the sudden changes in the stator and rotor electromagnetic parameters and time-varying inertial boundaries make the excitation control system highly susceptible to technical problems such as sudden changes in air gap flux linkage and transient reactive power surge drop. This electromagnetic field imbalance caused by command adjustment lag and parameter mismatch can not only induce violent oscillations in control commands and actuator protective limiting, but also, in severe cases, lead to uncontrolled reactive power output and even trigger catastrophic consequences such as unit shaft disconnection and shutdown, making it difficult to meet the high robustness and high precision control requirements of modern power systems for unit operating mode transitions.

[0004] Therefore, an optimized collaborative control method for the excitation system that is compatible with both power generation and synchronous condenser operating conditions is desired. Summary of the Invention

[0005] To address the aforementioned technical problems, this application provides a coordinated control method for an excitation system that is compatible with both power generation and synchronous condenser operating conditions.

[0006] According to one aspect of this application, a coordinated control method for an excitation system compatible with both power generation and synchronous condenser operating conditions is provided, comprising: S1: Perform high-frequency stripping and coordinate system decoupling reconstruction on the acquired raw sensor data stream to obtain the synchronous fundamental frequency feature dataset; S2: Perform dual-condition state feature extraction and pattern recognition on the synchronous base frequency feature dataset to obtain the condition switching flag and the system time-varying inertial parameters; S3: Based on the system's time-varying inertial parameters and operating condition switching indicators, the flux continuity prediction and physical difference comparison are performed on the synchronous base frequency characteristic dataset to obtain the stator and rotor flux deviation matrix. S4: The time-varying inertial parameters of the system are used as integral time constants to process the stator and rotor flux linkage deviation matrix to obtain the feedforward compensation voltage vector; S5: Perform dual coordinated control command modulation on the synchronous base frequency characteristic dataset and the feedforward compensation voltage vector to obtain the coordinated excitation trigger pulse signal.

[0007] Compared with existing technologies, this application provides a cooperative control method for an excitation system compatible with both generator and synchronous condenser operating conditions. By decoupling and reconstructing electromechanical sensor data and performing pattern recognition, it captures the time-varying inertial parameters of the unit during operating condition transitions in real time. Based on this, a state-space observation mechanism based on flux linkage continuity prediction is constructed, directly transforming the evolution of the system's inertial boundary into feedforward compensation voltage. This cooperative control strategy, by introducing dynamic time-varying inertia as a compensation benchmark, can overcome the adjustment dead zone of traditional proportional-integral converters, directionally eliminating flux linkage deviations within milliseconds and effectively suppressing transient reactive power drops. The resulting smooth switching not only significantly improves the control robustness of the excitation system under extreme transient conditions but also fundamentally avoids actuator saturation risks and shaft derailment shutdown hazards, greatly enhancing the voltage disturbance rejection capability of synchronous units in new power systems. Attached Figure Description

[0008] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0009] Figure 1 This is a flowchart of a coordinated control method for an excitation system compatible with both generator and synchronous condenser operating conditions, according to an embodiment of this application. Figure 2 This is a schematic diagram of the data flow of the coordinated control method for an excitation system compatible with both power generation and synchronous condenser operation according to an embodiment of this application. Figure 3 This is a flowchart of step S2 in the coordinated control method for an excitation system compatible with both power generation and synchronous condenser operation according to an embodiment of this application; Figure 4 This is a flowchart of step S3 in the excitation system coordinated control method compatible with both power generation and synchronous condenser operating conditions according to an embodiment of this application. Detailed Implementation

[0010] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0011] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0012] While this application makes various references to certain modules of the systems according to embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The modules described are merely illustrative, and different aspects of the systems and methods may use different modules.

[0013] Flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0014] The technical solution of this application proposes a coordinated control method for an excitation system that is compatible with both power generation and synchronous condenser operating conditions. Figure 1 This is a flowchart of a coordinated control method for an excitation system compatible with both generator and synchronous condenser operating conditions, according to an embodiment of this application. Figure 2 This is a system architecture diagram of a coordinated control method for an excitation system compatible with both generator and synchronous condenser operating conditions, according to an embodiment of this application. Figure 1 and Figure 2As shown, the method for coordinated control of an excitation system compatible with both generator and synchronous condenser operating conditions according to an embodiment of this application includes the following steps: S1, performing high-frequency stripping and coordinate system decoupling reconstruction on the acquired raw sensor data stream to obtain a synchronous base frequency feature dataset; S2, performing dual-operating-condition state feature extraction and pattern recognition on the synchronous base frequency feature dataset to obtain operating condition switching flags and system time-varying inertial parameters; S3, based on the system time-varying inertial parameters and operating condition switching flags, performing flux continuity prediction and physical difference comparison on the synchronous base frequency feature dataset to obtain a stator and rotor flux deviation matrix; S4, processing the stator and rotor flux deviation matrix using the system time-varying inertial parameters as an integral time constant to obtain a feedforward compensation voltage vector; S5, performing dual coordinated control command modulation on the synchronous base frequency feature dataset and the feedforward compensation voltage vector to obtain a coordinated excitation trigger pulse signal.

[0015] Specifically, in step S1, the acquired raw sensor data stream undergoes high-frequency stripping and coordinate system decoupling reconstruction to obtain a synchronous fundamental frequency characteristic dataset. The raw sensor data stream includes stator instantaneous voltage, stator instantaneous current, rotor excitation current, rotor physical speed, active power, reactive power, and prime mover tripping status. It should be understood that during the complex transient phase of smooth switching between generator and synchronous condenser operating conditions, the unit operates in a highly nonlinear environment. The data stream directly acquired by the raw sensors not only contains the fundamental frequency signal characterizing the system state but also couples with a large amount of sensor clutter, inverter high-frequency switching harmonics, and random noise generated by drastic changes in the electromagnetic environment. Directly inputting this raw data, containing high-frequency noise and time-domain discontinuity conflicts, into subsequent flux prediction or command modulation stages will cause severe output jitter in the proportional-integral loop of the control system, and may even lead to actuator malfunctions and operational instability. Therefore, in the technical solution of this application, the acquired raw sensor data stream is subjected to high-frequency stripping and coordinate system decoupling reconstruction in order to eliminate high-frequency interference and complete complex coordinate reconstruction, thereby providing a pure physical reference with time-domain synchronization characteristics for subsequent working condition identification and feedforward compensation.

[0016] In practice, the raw sensor data stream is first subjected to Kalman denoising and fundamental frequency reconstruction to obtain a clean electromechanical dataset. Since the raw sensor data stream is contaminated by measurement noise, conversion noise, and high-order harmonics introduced by power electronic switches during acquisition, a Kalman filter is used to preprocess the raw data to separate the core fundamental components characterizing the unit's electromechanical dynamics. As an optimal recursive state estimator, the Kalman filter can linearly estimate the internal state of the system from noisy measurements with minimum mean square error. In this process, the fundamental positive-sequence components of the terminal voltage / current are used as the state variables to be estimated. By establishing a state-space model that includes fundamental dynamics and noise statistics, the Kalman filter can recursively derive clean fundamental voltage and current estimates in real time. This process effectively removes high-frequency noise and harmonics, resulting in a clean electromechanical dataset containing a sequence of estimated fundamental voltage, current, and speed signals after preliminary purification.

[0017] Next, zero compensation for orthogonal axis virtual electrical voltage and time delay tracking of the fundamental frequency reference electrical angle are performed on the clean electromechanical dataset to obtain the time-domain synchronization correlation vector. Here, it should be understood that the positive-sequence component of the stator voltage can be extracted from the clean electromechanical dataset. However, in order to transform the signal in the stationary coordinate system to a coordinate system that rotates synchronously with the rotor magnetic field, the synchronous rotation electrical angle must be accurately obtained. Therefore, in the embodiments of this application, the fundamental reference electrical angle is first tracked in real time from the clean stator fundamental voltage using a phase-locked loop or similar algorithm. However, considering the sampling and calculation delays of the digital control system and the phase lag introduced by the filter, directly using this angle for coordinate transformation will lead to inaccurate decoupling of the coordinate axis components and steady-state errors. Therefore, an orthogonal axis virtual electrical voltage zero compensation strategy is introduced. That is, in an ideal synchronous rotating coordinate system, the quadrature axis component of the stator voltage should be controlled to zero. Therefore, a virtual quadrature axis voltage error signal can be constructed, which reflects the coordinate transformation deviation caused by the angle tracking delay. Then, a compensator processes the quadrature-axis voltage error signal, and its output fine-tunes the reference electrical angle to correct the angle deviation caused by system delay. The compensated synchronous electrical angle is more accurate, ensuring the correctness of subsequent coordinate transformations. Using this accurate synchronous electrical angle, the voltage and current vectors in the pure electromechanical dataset can be transformed into a synchronously rotating direct-axis quadrature-axis coordinate system, obtaining the corresponding direct-axis and quadrature-axis voltage and current vectors. These vectors are strictly synchronized with the rotor magnetic field in the time domain, ultimately forming a time-domain synchronous correlation vector.

[0018] Furthermore, the time-domain synchronization correlation vector is decoupled and mapped by reference electrical angle projection, and then assembled with normalized parameters to obtain the synchronous baseband feature dataset. The reference electrical angle refers to the angular position variable value, locked in real-time by the front-end time delay tracking algorithm, representing the instantaneous phase state of the grid baseband voltage; it serves as the time axis reference for spatial coordinate transformation. The normalized parameter is a dimensionless value reflecting the relative proportion of the equipment state, obtained by dividing the measured values ​​of stator voltage, current, and rotor excitation current by the system's preset rated or reference values. The final synchronous baseband feature dataset is a structured data dictionary containing the voltage and current vectors projected in the direct and quadrature coordinate systems, as well as normalized active and reactive power characteristics; it is the sole logical input for all subsequent adaptive control algorithms. In this process, firstly, the time-domain synchronization correlation vector generated in the previous steps is retrieved, and the electromagnetic parameters in the three-phase stationary coordinate system are projected onto the synchronous rotating coordinate system using the real-time tracked baseband reference electrical angle. Specifically, the instantaneous three-phase sampled values ​​are weighted and summed with the sine and cosine functions of the reference electrical angle, and a specific scaling factor with constant amplitude or power is introduced. Through this transformation logic, the three-phase AC magnetic field, which originally rotated drastically over time, is decomposed into mutually orthogonal direct-axis excitation components and quadrature-axis torque components, thus achieving decoupling of electromagnetic parameters at the physical level. Next, the decoupled components are compared point-to-point with their corresponding reference parameters. Specifically, the measured amplitude obtained from the projection is used as the divisor, and the rated voltage or rated current reference constant pre-configured in the controller memory is used as the divisor. The percentage or multiple of the physical quantity relative to the rated value is obtained through division. This processing method can eliminate the differences in absolute values ​​between generator sets of different capacities, ensuring that all input features are mapped within a specific engineering computing power safety domain, enhancing the algorithm's versatility. Finally, the data streams of each channel, after decoupling and normalization processing, are synchronously encapsulated according to a preset priority time scale. Specifically, the direct-axis components, quadrature-axis components, and power feature bits are serialized and arranged according to fixed memory addresses, and logical criterion flag bits required for condition switching are added to form a unified synchronous baseband feature dataset.

[0019] Specifically, in step S2, dual-condition state feature extraction and pattern recognition are performed on the synchronous baseband feature dataset to obtain the condition switching flag and the system's time-varying inertial parameters. It should be understood that existing dual-condition excitation control schemes largely rely on preset logic switching commands or fixed excitation control algorithms optimized for specific conditions. This traditional adaptive adjustment strategy often employs step-like parameter changes, making it difficult to accurately characterize the highly nonlinear physical characteristics of the unit during the operating mode transition. Especially during the moment of severe inertia collapse caused by the unit's prime mover tripping, existing strategies cannot capture the dynamic evolution trajectory of the system's inertial boundary in real time, resulting in a lack of deep integration of physical laws during the switching process. Therefore, in the technical solution of this application, dual-condition state feature extraction and pattern recognition are performed on the synchronous baseband feature dataset to capture in real time the physical characteristic distortions induced by unit tripping, load changes, etc., thereby providing parameter support with a real-time physical background for subsequent flux deviation calculation and feedforward compensation voltage generation output.

[0020] Figure 3 This is a flowchart of step S2 in the coordinated control method for an excitation system compatible with both generator and synchronous condenser operating conditions according to an embodiment of this application. Figure 3 As shown, S2 includes: S21, performing active power drop comparison, phase detection, and tripping logic judgment on the synchronous base frequency feature dataset to obtain the operating condition switching flag and the transition state feature subset for longitudinal transition; S22, using the digital polarity transition command of the operating condition switching flag as the activation wake-up condition, performing virtual inertia mapping and time-varying parameter deduction on the transition state feature subset to obtain the system time-varying inertial parameters.

[0021] Specifically, in step S21, active power sag comparison, phase detection, and tripping logic are performed on the synchronous base frequency feature dataset to obtain the operating condition switching flag and a transitional feature subset for longitudinal transition. Those skilled in the art should know that under power generation conditions, the unit outputs positive active power to the grid. When switching to synchronous condenser operating conditions, the unit coupling needs to be released, i.e., the turbine trips. Once the prime mover trips, the unit can no longer obtain mechanical power input from the turbine, and its ability to transmit active power to the grid as a generator will be rapidly lost. Therefore, active power sag is the most direct and significant electrical indicator of operating condition switching.

[0022] In this process, the algorithm first presets a power drop threshold and a power change rate threshold. When it detects that the per-unit active power drops sharply from a positive value to near zero or even negative (motor state, absorbing a small amount of active power to compensate for losses) within a very short period of time, and the rate of power drop exceeds the preset change rate threshold, the algorithm determines that an active power drop event has occurred. Simultaneously, to increase the reliability of the criterion, the algorithm combines the hard-wired signal of the prime mover tripping status from the original sensor data stream for tripping logic determination. Specifically, when the active power drop event logically matches the prime mover tripping signal, the algorithm determines that a switch from generation to phase modulation has occurred. At this time, the algorithm immediately generates a switching flag, for example, flipping it from logic "0" (generation state) to logic "1" (switching active state or target phase modulation state). This flag is a Boolean logic or digital instruction representing the current operating mode of the generator, indicating whether the system is currently in generator mode, synchronous condenser mode, or a transitional state between the two, used to activate all subsequent adaptive compensation algorithms. Simultaneously, starting from the current moment, the algorithm caches a short time window of the synchronous base frequency feature dataset, especially the dynamic change sequence of key quantities such as speed, power, voltage, and current. This data subset is the transition state feature subset used for longitudinal transition. This subset refers to a set of high-dimensional matrix data intelligently extracted from the synchronous base frequency feature dataset during a period before and after the polarity reversal moment. It reflects the details of the evolution of the electromagnetic parameters of the stator and rotor of the unit. It covers key features such as power drop slope and speed deviation gradient. It completely records the process of switching transient and will be passed to the next sub-step for inertial parameter deduction.

[0023] Specifically, in step S22, the digital polarity transition command of the operating condition switching flag is used as the activation and wake-up condition to perform virtual inertia mapping and time-varying parameter deduction on the transient state feature subset to obtain the system's time-varying inertia parameters. The digital polarity transition command refers to the step-like signal flip of the operating condition switching flag from a low level representing the steady state to a high level representing the startup transient when the system determines a qualitative change in the operating mode through active power drop phase detection and tripping logic. This signal serves as the trigger for the entire compensation logic. This step aims to quantify the changes in the system's dynamic characteristics caused by the operating condition switching. Its core is to estimate the system's time-varying inertia parameters, which are typically related to the unit's rotational inertia and damping coefficient, determining the system's speed change characteristics against power disturbances. Specifically, under generator operating conditions, this parameter value is relatively large (including turbine inertia); under synchronous condenser operating conditions, this parameter value is significantly reduced (containing only generator rotor inertia). The process is an event-triggered estimation algorithm that performs analysis on a subset of the transitional features of the cache in response to a digital polarity transition from 0 to 1 in response to a condition switching flag.

[0024] In this process, firstly, the algorithm digitally scans the logic state of the operating condition switching flag within each control pulse cycle. Specifically, it subtracts the flag's decision bit from the previous decision bit. When the difference exceeds a preset logic step threshold, the system determines a valid transition and broadcasts an activation signal to the memory bus. Secondly, upon activation, the algorithm immediately locates the pointer in the high-speed buffer and extracts a change sequence containing dynamic power drop characteristics. Specifically, using a preset time window, it extracts specific-dimensional parameter features from the cyclic data pool and performs a projection mapping based on acceleration energy loss on this data segment using a virtual inertia weight factor, transforming it into a virtual deviation vector characterizing the system's energy decay rate. Subsequently, based on the extracted feature subset, the controller uses the Euler nonlinear exponential function to perform a closed-loop solution for the system's disturbance tolerance margin at that moment, deriving the system's time-varying inertial parameters with time-domain decay properties. Specifically, firstly, the first derivative (i.e., rate of change) of active power over time is extracted from the transient state feature subset, and its absolute value is taken. Then, this absolute value is multiplied by a preset penalty factor scalar, and a time-cumulative integration is performed within the transient time window of the transition. Finally, the negative value of the integral result is used as the power of an exponential function with a natural constant (such as the Euler number) as its base, and multiplied by the initial nominal inertia value of the generator set. Through this nonlinear mapping logic, the system achieves instantaneous digital quantification of sudden changes in physical state, transforming the abstract power drop into a time-varying inertial parameter of the system with decay properties. This process can be expressed by the following formula: in, This represents the time-varying inertial parameters of the final generated system. This is the nominal initial inertia constant value of the unit when it leaves the factory. This represents a scalar quantity used to characterize the degree to which abrupt power changes violate the system's inertial boundaries. The first derivative of the active power in the characteristic subset of the transition state as a function of time. This represents exponential operations. Through this nonlinear mapping logic, the system achieves instantaneous digital quantification of sudden changes in physical state, transforming the abstract power drop into time-varying inertial parameters of the system with decay properties.

[0025] Specifically, in step S3, based on the system's time-varying inertial parameters and operating condition switching indicators, the flux linkage continuity prediction and physical difference comparison are performed on the synchronous fundamental frequency characteristic dataset to obtain the stator-rotor flux linkage deviation matrix. It should be understood that when a generator set smoothly switches from generator operation to synchronous condenser operation due to prime mover tripping, the stator-rotor electromagnetic parameters and time-varying inertial axis boundaries will fluctuate drastically. Traditional proportional-integral controllers, due to their inherent logic lag, cannot eliminate the stator-rotor electromagnetic field imbalance induced by flux linkage mutations in real time, while model predictive control algorithms are difficult to converge in real time within millisecond-level microcomputer devices due to excessive computational requirements. Therefore, in the technical solution of this application, by introducing a flux linkage observation mechanism with physical continuity constraints, the transfer array of the observer is dynamically corrected using real-time captured inertial parameters, thereby directly capturing transient flux linkages and generating anti-disturbance feedforward signals, fundamentally suppressing the transient reactive power impact caused by air gap flux linkage mutations.

[0026] Figure 4 This is a flowchart of step S3 in the coordinated control method for an excitation system compatible with both generator and synchronous condenser operating conditions according to an embodiment of this application. Figure 4 As shown, S3 includes: S31, in response to the operating condition switching flag being in an active state, the system's time-varying inertial parameters are imported as the nonlinear attenuation base into the factory-specified constant transfer coefficients of the full-dimensional state-space observer for adaptive topology scaling allocation to obtain the reshaped state-space matrix; S32, using the reshaped state-space matrix as the core prediction dominant term, the synchronous fundamental frequency characteristic dataset is subjected to state accumulation error drift suppression and multi-step continuous definite integral solution to obtain the theoretical coherent flux reference value; S33, the synchronous fundamental frequency characteristic dataset and the theoretical coherent flux reference value are subjected to full-dimensional reduced-order measured mapping and differential subtraction abrupt span calculation to obtain the stator and rotor flux deviation matrix.

[0027] Specifically, in step S31, in response to the operating condition switching flag being in an active state, the system's time-varying inertial parameters are imported as the nonlinear decay base into the factory-specified constant transfer coefficients of the full-dimensional state-space observer for adaptive topology scaling allocation to obtain a reshaped state-space matrix. Here, the full-dimensional state-space observer refers to a digital model constructed based on the generator's fifth-order nonlinear equations, used for online derivation of electromagnetic state vectors (such as air gap flux components) that cannot be directly observed within the stator and rotor; the factory-specified constant transfer coefficients refer to the linear state transition matrix pre-coded into the controller program at the time of equipment shipment, determined based on the unit's rated physical inertia and impedance characteristics.

[0028] In this process, firstly, the algorithm intercepts the condition switching flag bit within each sampling cycle. The trigger immediately opens the transmission bus channel for injecting dynamic parameters into the observer kernel if and only if the flag bit presents an active logic value representing startup compensation in the memory address. Secondly, the system time-varying inertial parameters calculated in the previous steps are retrieved, and the nonlinear characteristics of the exponential decay function are used to process them into scaling relaxation factors with physical penalty significance. Specifically, the system time-varying inertial parameters are used as the nonlinear decay base, combined with a preset topology precision constant, to deduce the scaling weights for the matrix elements. This process can be expressed by the following formula: in, The real-time topology scaling factor derived from the deduction. These are the topology accuracy constant coefficients, which are pre-stored in the configuration table and are used to characterize the sensitivity of a particular unit to physical dimensions. This refers to the real-time input of the system's time-varying inertial parameters. Furthermore, the scaling factor calculated above is applied to each non-zero element of the factory-specified constant transition coefficient matrix of the full-dimensional state-space observer. Through scalar-matrix multiplication, adaptive allocation of the state transition intensity of the observation model is achieved. The resulting reconstructed state-space matrix will serve as the core dominant term for subsequent multi-step continuous definite integral solutions. Its synthesis process can be expressed by the following formula: in, The final output is the reshaped state space matrix. This is the original factory-specified constant transfer coefficient matrix. Through this mapping process, the state evolution rate inside the observer can be flexibly self-corrected based on the actual collapse level of the unit's inertia.

[0029] Specifically, in step S32, the reshaped state-space matrix is ​​used as the core predictive term. State accumulation error drift suppression and multi-step continuous definite integral solving are performed on the synchronous fundamental frequency characteristic dataset to obtain a theoretically coherent flux linkage reference value. Here, it should be understood that at the moment of severe inertia collapse caused by prime mover tripping, the delayed feedback regulation cannot immediately respond to the sudden change in stator and rotor flux linkage. Simultaneously, although model predictive control has predictive capabilities, the model predictive control algorithm faces significant obstacles to industrial implementation due to the computational demands of the generator's fifth-order nonlinear equations, which are difficult to converge in real-time within millisecond-level microcomputer devices. Therefore, this application constructs a theoretical flux linkage reference value with physical continuity characteristics, which can directly compensate for the missing reactive power equation balancing boundary caused by energy drop in physics, fundamentally intercepting the sudden change in air gap flux linkage caused by abrupt changes in electromagnetic parameters and time-varying inertial boundaries. In this step, state accumulation error drift suppression refers to a closed-loop correction mechanism introduced during the definite integral solution process to prevent the flux linkage state from deviating over time due to sensor noise or calculation rounding errors. Multi-step continuous definite integral solving refers to the infinitesimal integral operation of state variables using the reshaped state transition logic within a single simulation prediction cycle. The final theoretical coherent flux linkage reference value refers to the target electromagnetic field component with time-domain coherence, deduced in mathematical space based on the current physical energy state of the system after eliminating the influence of transient nonlinear disturbances.

[0030] In this process, firstly, the algorithm retrieves real-time physical measurements such as terminal voltage and current from the synchronous fundamental frequency feature dataset, compares them with the predicted output of the previous cycle inside the observer, and introduces feedback correction gain. This process ensures that the observer's evolution is always locked within the true physical observation boundary, and its drift suppression correction logic can be expressed by the formula: in, This represents the state derivative vector after drift suppression correction. This is the reshaped state space matrix generated in the previous step, incorporating the system's time-varying inertia. For the system in The state vector at any given time, in the excitation system control, usually contains internal electromagnetic parameters that cannot be directly observed, such as the flux linkage components of the stator and rotor. The input influence matrix characterizes the weights of the influence of external excitation inputs on changes in the internal state of the system. This is the real-time excitation input vector of the system. The preset error convergence gain matrix, The output is a physical measurement, that is, a real-time electromechanical characteristic signal such as voltage and current directly captured from the unit's sensor network after decoupling; The output mapping matrix is ​​used to map the internally invisible state vectors. The projection transformation yields a theoretically observable output value. Next, using the corrected derivative vector as the integrand, a definite integral based on high-order numerical analysis is performed within the prediction step window. This process, by reshaping the dominance of the state-space matrix, accurately inverts the dynamic evolution trajectory of the stator and rotor flux linkages during the inertia collapse period. The solution logic for its multi-step definite integral can be expressed by the following formula: in, Predict the target value of the magnetic flux after the step window ends; This represents the known theoretical flux linkage reference starting value at the current moment, which is the initial state of this integral calculation step sequence; These are the integration weighting coefficients, used to assign the contribution weight of different sub-sampling points to the final integration result; Indicates the time of subdivision sampling Below, the state derivative vector after drift suppression correction is the core integrand of the entire integral, representing the instantaneous rate of change of the magnetic flux in each infinitesimal time interval; This represents the timestamp of each discrete integral sampling point within the prediction step size span; This represents the prediction step size (or integration time window) of the flux linkage observer. Since the reshaping matrix reflects the collapse of the system's physical inertia in real time, this integration process achieves seamless connection to the physical boundary during the energy drop period. Furthermore, the high-dimensional state variables obtained from the integration are decoupled spatially to extract the magnitude and phase information of the stator and rotor flux linkages, ultimately encapsulating them into a theoretically coherent flux linkage reference value. This value directly characterizes the theoretical lifeline that the excitation system must maintain to ensure smooth magnetic field switching under current physical constraints.

[0031] Specifically, in step S33, the stator and rotor flux linkage deviation matrix is ​​obtained by performing a full-dimensional reduced-order measured mapping and differential subtraction abrupt change span calculation on the synchronous fundamental frequency characteristic dataset and the theoretical coherent flux linkage reference value. Here, it should be understood that when the unit experiences severe inertia collapse due to prime mover tripping and switches to synchronous condenser operation, the magnetic field balance of the stator and rotor will momentarily become unstable due to the sudden change in physical boundaries. Specifically, the sudden changes in stator and rotor electromagnetic parameters and time-varying inertial boundaries will cause abrupt changes in air gap flux linkage and transient reactive power impacts in the excitation control system. Since traditional proportional-integral control schemes have inherent logical lag, and the computational power requirements of the generator's fifth-order nonlinear equations are difficult to converge in real time within a millisecond-level microcomputer device, the technical solution of this application, by differentially comparing the measured state mapped from the characteristic dataset with the theoretical coherent trajectory derived from inertia, can accurately quantify the span of the air gap flux linkage abrupt change. The final stator-rotor flux linkage deviation matrix can directly compensate for the missing reactive power equation boundary balancing caused by energy drop from a physical perspective, providing an absolute difference reference for the subsequent generation of high-performance feedforward disturbance rejection commands, thereby fundamentally intercepting transient reactive power surge drop. Here, full-dimensional reduced-order measured mapping refers to mapping the high-dimensional synchronous fundamental frequency characteristic dataset (including stator voltage, current, and rotor current, etc.) to a low-order reduced-order coordinate space that can directly express the physical quantity of flux linkage through specific electromechanical parameter association logic; differential subtraction abrupt change span calculation refers to obtaining the instantaneous jump amplitude value (i.e., abrupt change span) of flux linkage within a micro-element time period through dynamic weighted subtraction operation of two data sources; the final stator-rotor flux linkage deviation matrix refers to the final encapsulated multi-dimensional data set containing the residuals of magnetic circuit evolution in each dimension of the direct axis (d-axis) and quadrature axis (q-axis).

[0032] In this process, the kernel first retrieves the synchronous fundamental frequency characteristic dataset. Utilizing the physical constraints of the generator's electromagnetic circuit, the current and voltage vectors are mapped to the measured transient flux linkage phase vectors. The calculation logic for this reduced-order mapping can be expressed by the following formula: in, This represents the measured transient flux linkage phase vector obtained after mapping. This represents the full-dimensional reduced-order mapping operator matrix constructed based on the mutual inductance and leakage inductance parameters of the generator stator and rotor. This is a real-time input synchronous fundamental frequency characteristic dataset. Through this mapping, complex electrical parameters are transformed into comparison components with magnetic field physical properties. Next, the measured transient flux linkage phase vector and the theoretical coherent flux linkage reference value are synchronously sampled, and differential subtraction logic is performed to capture abrupt deviations in the flux linkage during its evolution. The calculation process can be expressed by the following formula: in, This is the flux linkage abrupt change difference vector, which accurately reflects the energy gap in the magnetic circuit caused by the prime mover tripping. Furthermore, the calculated flux linkage abrupt change difference vector is orthogonally decoupled and encapsulated according to its direct axis (d-axis) and quadrature axis (q-axis) components, and a normalization coefficient is introduced to finally generate the stator and rotor flux linkage deviation matrix. Its generation logic can be expressed by the formula: in, The normalized parameter matrix representing the diagonal distribution. The stator and rotor flux linkage deviation matrix is ​​used as the sole dominant quantity in the subsequent S4 step for calculating the feedforward compensation voltage vector.

[0033] Specifically, in S4, the system's time-varying inertial parameters are used as integral time constants to process the stator and rotor flux linkage deviation matrix to obtain the feedforward compensation voltage vector. It should be understood that at the moment of severe inertia collapse switching caused by the generator set tripping, the feedforward compensation voltage vector exhibits extremely high-frequency, severe transient impact execution characteristics. Traditional fixed-gain compensation schemes, by neglecting the core time-varying boundary of "inertia" in physical properties, cannot physically scale the intensity of the control pulse during the rapid energy dissipation period. By introducing real-time calculated system time-varying inertial parameters, the abstract physical inertia collapse can be transformed into an integral time reference in control theory. This allows for dynamic anchoring of the compensation voltage's dynamic margin based on the generator set's real-time resistance to electromagnetic fluctuations, eliminating transient conflicts and saturation cutoff risks at the substantive physical execution layer, and completely bypassing the execution divergence trap caused by the strong nonlinear cross-coupling of numerous physical variables. The stator and rotor flux linkage deviation matrix refers to the dynamic residual matrix between the theoretical flux linkage predicted by the preceding full-dimensional state-space observer and the measured mapped flux linkage; it directly reflects the magnetic circuit gap caused by energy drops in the system. The resulting feedforward compensation voltage vector is the final generated digital instruction credential used to capture and execute flux collapse extremum erasure operations in real time.

[0034] In specific implementation, firstly, the stator and rotor flux linkage deviation matrix is ​​subjected to multidimensional orthogonal decoupling extraction to obtain the biaxial disturbance component sequence. Since the stator and rotor flux linkage deviation matrix contains multidimensional, highly coupled raw residual data, without decoupling extraction, the control system cannot identify whether the disturbance occurs in the excitation magnetic field component or the electromagnetic damping component. This physical coupling leads to severe cross-orthogonal or unidirectional conflicts in the control commands in the spatial direction, resulting in severe integral wind saturation and making it difficult for the system to converge after the physical transient impact ends. Therefore, in the technical solution of this application, a spatial orthogonal decoupling mechanism is established to decompose the complex flux linkage deviation span into mutually independent control axial components with clear physical meaning, thereby providing a clean and orthogonal input signal for subsequent feedforward compensation based on time-varying inertia, eliminating transient conflicts and saturation truncation risks from the physical execution layer.

[0035] In this process, firstly, reference electrical angles, time-locked, are retrieved from the preceding synchronous fundamental frequency feature dataset to construct an orthogonal projection reference that rotates synchronously with the rotor's physical position. Secondly, the input stator and rotor flux linkage deviation matrix is ​​used as the source vector, and a spatial rotation and eigenvalue extraction are performed on it using an orthogonal decoupling mapping operator. This reshapes the deviation from a static, mutually coupled winding dimension into direct-axis and quadrature-axis components that can directly characterize the intensity and direction of the magnetic circuit deviation. The core orthogonal decoupling extraction logic can be expressed by the following formula: in, This represents the generated sequence of biaxial perturbation components; Indicates the real-time reference electrical angle; This represents an orthogonal decoupling operator matrix constructed based on the generator inductance parameter matrix and electrical angles. It ensures that the mapped components have equivalent consistency in the energy dimension through conformal transformation. This refers to the input stator and rotor flux linkage deviation matrix. Furthermore, due to sampling noise, the decoupled components may contain high-frequency glitches. The algorithm utilizes windowed moving average logic to perform feature denoising on the decoupled direct-axis and quadrature-axis components. The calculation logic for this smoothing process can be expressed by the following formula: in, The sampling depth of the sliding window; This represents the smoothed dual-axis perturbation component sequence of the final output, which is a clean feature signal after noise stripping and is used for subsequent anti-perturbation propagation calculations. Indicates the index of the current discrete clock step; This represents the historical duration offset index. Through this calculation process, the system can extract the true biaxial disturbance component sequence from the transient deviation matrix without deviation, thereby ensuring that an absolutely accurate physical excitation source can be obtained when using the system's time-varying inertial parameters for disturbance rejection.

[0036] Next, the system's time-varying inertial parameters are used as the denominator for inverse proportional time correction to perform time-varying inertia-based disturbance rejection propagation on the biaxial disturbance component sequence to obtain the transient compensation sequence. It should be understood that during the complex transient phase of smooth switching from generator to synchronous condenser operation, the physical inertial boundary of the unit is not constant but exhibits drastic time-varying characteristics with the prime mover tripping and energy release process. If only conventional fixed proportional gain is used to compensate for flux disturbances, it will be insufficient to meet the unit's disturbance rejection requirements during severe inertia collapse, easily leading to excessive compensation strength or lag. By introducing the system's time-varying inertial parameters into the computational structure, a disturbance rejection propagation mechanism based on physical properties can be established, allowing the generation rate and amplitude of the compensation voltage to be adjusted in real time according to the unit's inertial potential to resist electromagnetic fluctuations. When inertia decreases, the system automatically amplifies the compensation weight, thereby selectively eliminating flux fluctuations within milliseconds, ensuring the unit has extremely high survivability robustness during peak transition periods. The inverse proportional time correction denominator refers to placing the system's time-varying inertial parameters in the denominator of the division operation, so that their values ​​form an inverse proportional mapping relationship with the output compensation intensity, thereby adjusting the integral time scale.

[0037] In this process, firstly, the real-time calculated dual-axis disturbance component sequence and the system time-varying inertial parameters derived from previous steps are retrieved. Secondly, an inverse proportional transfer function with numerical protection is constructed, using the system time-varying inertial parameters as the denominator of the inverse proportional time correction, to perform point-to-point disturbance rejection transfer on the aforementioned dual-axis disturbance component sequence. This process can be expressed by the following formula: in, and These represent the transient compensation amounts generated in the direct and quadrature directions, respectively; and This represents the real-time input value in the dual-axis perturbation component sequence; The system's time-varying inertial parameters are used to correct the denominator of the inverse proportional time correction. A tiny numerical compensation bias constant is pre-configured in the algorithm's memory to prevent the denominator from encountering absolute zero and causing a division-by-zero crash in the underlying system. Through this inverse proportional operation, an adaptive gain mapping of the flux linkage deviation to the compensation level is achieved. When the inertia parameter decreases with changing operating conditions, the decrease in the denominator automatically increases the intensity of the compensation voltage. Furthermore, the calculated direct-axis and quadrature-axis transient compensation quantities are vector-merged to form a time-domain continuous transient compensation sequence, which serves as the raw input for subsequent scaling and normalization processing. The final generated transient compensation sequence is represented as follows: in, This is a transient compensation sequence. Through this calculation logic, the excitation system establishes a real-time dynamic correlation between deviation and compensation at the physical execution level.

[0038] Furthermore, the transient compensation sequence is calibrated, normalized, and repackaged with feedforward vectors to obtain the feedforward compensation voltage vector. It should be understood that the transient compensation sequence obtained from the front-end decoupling derivation is essentially a raw numerical stream calculated based on flux linkage deviation and time-varying inertial parameters. Its amplitude range and data format have not yet been mapped and aligned with the physical characteristics and mathematical model of the underlying excitation converter. To achieve full compatibility of the excitation system with both generation and phase-modulation operating conditions, the excitation parameter configuration is adjusted and the control strategy is optimized. Since the generator is converted to synchronous condenser operation, the phase-modulation mode requires high-performance excitation indicators such as 3.5 times the rated excitation voltage. If the raw compensation sequence is directly input into the modulation stage, the inconsistent numerical dimensions will lead to inaccurate control gain, or even digital overflow or excessive output limits of the physical actuator. Therefore, a calibration reference system is established to map the transient calculation results to a unified per-unit value domain and encapsulate them into vectors according to the interface requirements of dual-coordinated control command scheduling, thereby ensuring that the compensation command can accurately and safely drive the excitation power unit. The calibration normalization process involves using a pre-set rated excitation voltage reference to perform a linear mapping on each component of the transient compensation sequence, transforming it into a proportionality coefficient relative to the reference value. This aims to eliminate the interference of physical dimensions on the robustness of the algorithm. The feedforward vector reassembly and encapsulation process involves aligning the calculated and calibrated direct-axis and quadrature-axis voltage increments with clock signals and attribute annotations according to the data format specified by the control system bus protocol, constructing a standard high-dimensional state vector matrix. The final feedforward compensation voltage vector is the final digital instruction output, containing transient terminal voltage correction information required to directly compensate for flux collapse, serving as the core input for coordinated instruction modulation in the subsequent S5 step.

[0039] In this process, firstly, the excitation conversion coefficient adapted to the current operating condition is retrieved from the controller's non-volatile memory to calibrate the amplitude of the discrete transient compensation. This process can be expressed by the formula: in, This represents the calibrated transient compensation vector; The input is the transient compensation sequence; This represents the conversion gain matrix based on the preset transformation ratio of the physical actuator. Secondly, the rated nominal voltage measured on the secondary side of the excitation transformer is used as the calibration denominator, and the calibrated vector is normalized to ensure its amplitude meets the per-unit range requirements. The normalization calculation process can be expressed by the formula: in, This is the generated standardized sequence vector; This is the reference constant for the system's rated excitation voltage; and Represents respectively in The system generates a per-unit direct-axis (d-axis) voltage compensation component and a per-unit quadrature-axis (q-axis) voltage compensation component, obtained after scaling. Through this logic, the system achieves absolute dimensional consistency between the compensation signal and the base voltage command. Finally, the per-unit components are spatially aligned according to the timestamp of the control cycle and encapsulated into a feedforward feature data packet to form the final feedforward compensation voltage vector. The encapsulation and reassembly process can be expressed by the following formula: in, This is the final generated feedforward compensation voltage vector; This indicates a concatenated logical operator, representing the serialization and concatenation of different data bits, feature components, and verification fields according to a preset physical memory address or communication frame format during the digital encapsulation process. The header information indicates that in software engineering data protocols, the header typically includes data identifiers, characteristic attribute labels, and synchronization clock tags to ensure that the control network can correctly identify and process the feedforward command signal in real time. This field represents the check bit. In real-time industrial control systems, it is used to detect whether bit flips or data loss occur during the transmission of feedforward branch signals on the high-speed data bus, ensuring the integrity and security of control commands. Through this encapsulation logic, complex physical compensation characteristics are transformed into a standardized feedforward compensation voltage vector, providing a reliable data carrier for achieving smooth coordinated control between generator and synchronous condenser operating conditions.

[0040] Specifically, in step S5, dual coordinated control command modulation is applied to the synchronous base frequency characteristic dataset and the feedforward compensation voltage vector to obtain a coordinated excitation trigger pulse signal. It should be understood that the excitation system, through parameter configuration and optimization strategies, must meet the independent operation capabilities of both generation and phase regulation modes, as well as the strong excitation requirements. Since the excitation regulation system includes a low-frequency basic steady-state regulation term and a high-frequency feedforward transient compensation term, failure to perform nonlinear coordinated modulation on these two commands will lead to command truncation, spatial phase distortion, and severe integral wind saturation at the voltage ceiling boundary of the physical actuator. To effectively improve the robustness of the unit under extreme and inertia-prone conditions, dual coordinated control command modulation is applied to the synchronous base frequency characteristic dataset and the feedforward compensation voltage vector to achieve seamless fusion of metastable state commands. This generates a smooth excitation trigger pulse group through spatial vector allocation, eliminating the physical traps that induce large generator desynchronization from within the final channel.

[0041] In the first embodiment of this application, step S5 includes: S51, performing terminal voltage tracking error comparison and integral accumulation correction to eliminate steady-state error on the synchronous base frequency feature dataset to obtain the base voltage duty cycle command; S52, performing metastable state command feedforward decoupling and fusion on the base voltage duty cycle command and the feedforward compensation voltage vector to obtain the target cooperative voltage vector; S53, performing space vector pulse width digital modulation on the target cooperative voltage vector to obtain the cooperative excitation trigger pulse signal.

[0042] Specifically, in step S51, the synchronous baseband feature dataset is compared with the terminal voltage tracking error and integrated cumulative correction is performed to eliminate steady-state error, thereby obtaining the base voltage duty cycle command. Since the power grid has strict limits on the fluctuation of the generator terminal voltage, if the synchronous baseband sampling data is not deeply tracked and its steady-state error eliminated, voltage drift will occur during steady-state operation, thus affecting the reliability of grid-connected operation. Therefore, by establishing a long-period steady-state control benchmark based on a proportional-integral (PI) regulation mechanism, a convergent base duty cycle command is provided for subsequent multi-dimensional coordinated modulation.

[0043] In this process, firstly, the per-unit direct axis (d-axis) and quadrature axis (q-axis) voltage components obtained through coordinate transformation are retrieved from the input synchronous fundamental frequency feature dataset, and vector magnitude reconstruction is performed to obtain a composite parameter representing the effective value of the actual terminal voltage. The calculation logic is as follows: in, This represents the per-unit component of the reconstructed measured terminal voltage. Next, the voltage reference value stored in non-volatile memory is retrieved and subtracted from the measured value. The calculation logic is as follows: in, To generate the terminal voltage tracking error, The preset synchronous reference voltage per unit is used as the standard. Furthermore, the aforementioned tracking error is used as the input to the proportional and integral terms of the excitation regulator. By integrating the error over time, a base voltage duty cycle command capable of offsetting system disturbances is established. The calculation principle is as follows: in, The duty cycle command is the base voltage for the final output. This is the proportional adjustment coefficient. This is the integral adjustment coefficient. Through this closed-loop calculation logic, the system establishes a steady-state excitation energy level that matches the current operating conditions.

[0044] Specifically, in step S52, the base voltage duty cycle command and the feedforward compensation voltage vector are decoupled and fused using a transient steady-state command feedforward to obtain the target cooperative voltage vector. Since the excitation system involves both steady-state terminal voltage regulation requirements and transient flux collapse compensation requirements, a single control quantity cannot simultaneously address the unit's operational stability during the switching between power generation and phase regulation. Therefore, in the technical solution of this application, the base voltage duty cycle command, representing steady-state control, and the feedforward compensation voltage vector, representing transient disturbance rejection, are fused to meet the operational characteristic requirements under different operating conditions, ensuring smooth unit switching and operational stability. The base voltage duty cycle command refers to the base pulse width control quantity output by the proportional-integral stage of the excitation regulator, used to maintain the steady-state operation of the generator terminal voltage; it reflects the low-frequency steady-state regulation law of the control system. The feedforward compensation voltage vector is a high-frequency disturbance rejection command derived based on the system's time-varying inertial parameters and flux deviation matrix, designed to instantly eliminate sudden electromagnetic field deviations. The final target coordinated voltage vector, as the output of this step, is the final digital control certificate after the composite superposition of the two command information in the physical voltage vector domain, which directly guides the subsequent space vector modulation.

[0045] In this process, firstly, the algorithm synchronously retrieves the base voltage duty cycle instruction for the current control cycle and the feedforward compensation voltage vector calculated in the previous steps from the high-speed memory bus. Secondly, it performs arithmetic accumulation operations on the direct axis (d-axis) and quadrature axis (q-axis) components of the two instructions in the synchronously rotating coordinate system, and combines the steady-state reference and transient correction values. The specific calculation principle of the combination is as follows: in, and These represent the components of the generated target cooperative voltage vector on the d-axis and q-axis, respectively. and The duty cycle command component is the input base voltage; and This is the feedforward compensation voltage vector component. Furthermore, to prevent the total vector magnitude from exceeding the rigid hardware defenses of the physical converter due to blind hard addition, the system executes saturation interception logic based on the accumulation result to ensure that the synthesized cooperative instruction does not exceed the physical limit bit of the actuator. The specific limiting logic is expressed as follows: in, This is the preset per-unit threshold constant for the maximum output voltage of the excitation converter. This is the feedforward compensation voltage vector. Through this linear superposition path, the system completes the initial convergence of the metastable state command.

[0046] Specifically, in step S53, the target coordinated voltage vector is subjected to space vector pulse width digital modulation (SVPWM) to obtain a coordinated excitation trigger pulse signal. To ensure stable excitation support for the unit under both generation and synchronous condenser operating conditions, and to guarantee smooth switching and operational stability, the technical solution of this application utilizes space vector pulse width modulation (SVPWM) technology. This enables the voltage vector output by the excitation converter to closely approximate the target coordinated voltage vector in both amplitude and phase, thereby accurately achieving the excitation control target compatible with both operating conditions and supporting high-performance excitation indicators such as 3.5 times the rated excitation voltage within the high-frequency control cycle. Space vector pulse width digital modulation refers to a digital control technology that generates a rotating magnetic field vector in the stator-rotor gap of the motor by controlling the duration of different switching states of the power converter. The final coordinated excitation trigger pulse signal refers to the pulse-form digital level signal generated to directly control the switching logic of the excitation converter (such as a thyristor or inverter bridge).

[0047] In this process, firstly, the algorithm receives the target coordinated voltage vector and, in conjunction with the real-time measured reference electrical angle, performs a conformal mapping from the rotating coordinate system to the two-phase stationary coordinate system. The calculation logic is as follows: in, and Represents the components in the stationary coordinate system. For real-time electrical angle, and The target voltage vector is represented by its two-axis component. Next, the physical hexagonal sector where the output voltage vector resides is determined based on its angle. Based on this, the action time of two adjacent fundamental spatial voltage vectors and the zero vector is calculated using the linear decomposition principle, thus equivalently synthesizing the target vector in the time domain. The decomposition logic of its action time is as follows: in, To control the cycle, and The duration of action of two adjacent fundamental vectors. The zero vector action time, , and These represent the corresponding space voltage vectors. Furthermore, based on the calculated time allocation results, and following the rules of symmetrical PWM sequences, a set of binary logic signals is generated to control the switching of the converter bridge arm switches, namely, the coordinated excitation trigger pulse signals. This process ensures that the average voltage output across the converter terminals remains physically consistent with the target coordinated voltage vector.

[0048] However, analysis revealed that in the complex transient scenario of smooth switching between generator and synchronous condenser operating conditions, the first embodiment used a direct and crude pure arithmetic addition to merge the base voltage duty cycle command and the feedforward compensation voltage vector. This blindly superimposed control logic failed to fully consider the special relationship between frequency standard time-domain fault conflicts and the coupling of physical actuator saturation boundaries, resulting in a significant risk of operational instability in the actual physical engineering implementation of the control system.

[0049] Specifically, at the moment of severe inertia collapse switching caused by the tripping of the prime mover, the feedforward compensation voltage vector exhibits extremely high-frequency, violent transient impact execution characteristics, while the base voltage duty cycle command output by the proportional-integral loop shows a low-frequency steady-state regulation pattern. These two commands exhibit severe cross-orthogonal and in-direction control conflicts within the control cycle. When the high- and low-frequency commands are opposite in spatial direction, the low-frequency base response is forcibly canceled by the transient command, resulting in regulation delay. When they are in the same direction, blindly adding them together causes the integral term of the low-frequency proportional-integral controller to grow rapidly in a very short time, leading to severe integral wind saturation. This causes the system to fall into a difficult-to-converge reverse oscillation after the physical transient impact ends. Furthermore, this superposition logic ignores the risk of nonlinear actuator saturation boundary coupling. Because the physical DC bus voltage of the underlying excitation converter has an absolutely rigid ceiling physical limit, unconditional linear superposition can easily cause the magnitude of the synthesized target voltage vector to break through the physical hexagonal defense of the inverter topology. Once the command encounters a rigid amplitude truncation, not only will the output amplitude be limited, but the absolute spatial phase of the command vector will also undergo severe uncontrollable distortion and shift. This will directly lead to the spatial angle mapping disorder of the stator and rotor compensation magnetic fields, ultimately triggering a catastrophic reactive power drop crisis or even causing actual grid connection disconnection and shutdown.

[0050] In order to eliminate the aforementioned transient conflicts and saturation truncation risks from the actual physical execution layer, this application proposes an optimal mechanism that replaces the original absolute alignment accumulation module with an adaptive instruction fusion mechanism based on dynamic boundary and time-varying conflict decoupling.

[0051] Specifically, firstly, transient conflict coupling degree identification is performed on the base voltage duty cycle command and feedforward compensation voltage vector to obtain the control conflict coupling degree. This step aims to accurately quantify the collinearity interference level of low-frequency steady-state regulation command and high-frequency transient feedforward impulse command in the multidimensional topological space and avoid error accumulation. Specifically, by performing spatial dimension reduction reconstruction on the intercepted base voltage duty cycle command and feedforward compensation voltage vector, and utilizing the vector inner product feature structure in mathematical spatial projection, forward and reverse dot product operations based on high-order modulus normalization are performed, thereby deriving time-varying scalar data that can characterize the degree of spatial angle geometric interference to establish the control conflict coupling degree. This step breaks through the limitations of direct algebraic comparison, historically upgrading it to continuous dynamic analysis of spatial geometric angle features, which can be expressed by the formula: in, The control conflict coupling degree parameter is dynamically calculated as the real-time time variable evolves, and its final eigenvalue range is safely normalized and limited to the range of negative one to positive one. This represents the base voltage duty cycle command vector used to dominate the typical time-varying steady-state mild disturbance response characteristics. This represents the feedforward compensation voltage vector used for instantaneous capture and execution of flux collapse extremum erasure operations; This represents performing a modulo operation on the inner enclosed vector matrix to obtain the absolute mathematical scalar magnitude; This represents the inner product dot operation between two matrix-vector components that are in the same time domain; the formula is appended with... This indicates a tiny numerical compensation bias constant pre-configured in the algorithm's memory, specifically designed to prevent the denominator from encountering an absolute zero value, which could lead to a division-by-zero crash in the underlying system.

[0052] The closer the coupling conflict value extracted through this calculation process is to the positive limit boundary, the more destructive the contention and interference between the two commands within the same physical execution domain in the current control cycle becomes. This mapping deduction perfectly matches the scenario requirements of deeply capturing the high-frequency internal friction defects of internal commands during the nonlinear transient order reduction period, and provides an extremely sensitive digital adjudication benchmark for subsequent intelligent command yielding decisions.

[0053] Next, the base voltage duty cycle command, feedforward compensation voltage vector, and control conflict coupling degree are cross-decoupled and adjusted with integral wind saturation suppression to obtain the corrected base voltage command. This step aims to establish a physical firewall to strongly intercept the cumulative backflow of errors caused by high-dimensional parallel control conflicts that penetrates into the underlying physical closed-loop converter, thereby preserving the global convergence stability of the proportional-integral control system under extreme cutoff conditions. Specifically, the controller simultaneously retrieves the base voltage duty cycle command, feedforward compensation voltage vector, and the newly calculated control conflict coupling degree from the previous step via the high-speed memory bus. Then, it extracts the positive interference component that deeply characterizes the destructive force of system coupling conflicts, and combines this with the actual occupancy ratio of the feedforward high-frequency voltage intensity to the underlying device's physical limit voltage pool to precisely construct a nonlinear exponential decay function containing a dynamic reverse penalty gene. Furthermore, this dynamic penalty function is used to apply a soft decoupling multiplicative decay correction operation with forced suppression to the base voltage duty cycle command carrying low-frequency characteristics, thereby deriving a safe and usable corrected base voltage command through order reduction. This process can be expressed by the formula: in, This represents the corrected base voltage command safety vector output after the anti-integral wind saturation suppression logic has been cleared. This represents an exponentially decaying action operator with nonlinear scaling capability, constructed using Euler's natural constant as the integration base; This represents the penalty exponential constant factor; This represents a positive-restrained, negative-restrained logical operator that forces the selection of the largest value from two parallel values ​​enclosed in parentheses. It profoundly describes the maximum physical highest transient voltage per unit threshold constant that the AC side of the converter hardware topology associated with the generator body can allow for drive projection.

[0054] It is understandable that when the prime mover disengages, causing a sharp collapse in the transient flux linkage of the stator and rotor, and the feedforward compensation disturbance rejection demand explodes, even nearly eroding the entire hardware voltage control margin of the frequency converter, the exponential decay trigger term inside the formula will irreversibly cause the closed-loop command required for normal operation to implement smooth and flexible damping decay, approaching a minimum zero value. This flexible dynamic yield calculation strategy based on the depth index of the system's disturbance rejection state effectively avoids the execution divergence trap caused by the strong nonlinear cross-coupling of numerous physical variables, enabling the control system survivability robustness and convergence disturbance rejection of the generator excitation system to achieve a generational improvement in extreme conditions and oscillations during periods of inertia loss.

[0055] Furthermore, the modified base voltage command and feedforward compensation voltage vector are subjected to conformal orthogonal synthesis based on dynamic margin to obtain the target cooperative voltage vector. This step aims to converge and accommodate all transient execution interferences within the safe boundary of the physically feasible operating domain of the hardware equipment. Given that the final output maximum withstand drive voltage of the inverter full-bridge converter belongs to a rigid boundary that is absolutely inviolable from a physical perspective, in order to ensure the core premise of ensuring the strength of feedforward compensation that has lifeline significance, and to absolutely guarantee that the three-phase spatial drive phase angle used to support the spatial magnetic field is not distorted by violent limiting, the multi-dimensional fusion synthesizer solves online in a dynamic mathematical algebraic form for the actual level remaining available dynamic margin of the underlying rigid voltage ceiling after forcibly deducting the magnitude of the feedforward compensation command. After the system firmly establishes the sole fusion criterion of absolutely prioritizing the urgent needs of the upper-side feedforward, it transforms the acquired available residual space margin into a smooth scalar scaling relaxation factor specifically used for constraining and correcting the base voltage command. Following strict geometric conformal properties, it uses pure radial amplitude constraint adjustment as the entry point to advance the command confluence mechanism, thereby generating a unique digital credential target cooperative voltage vector that possesses both composite control effects and is free from any hardware saturation distortion or breakdown risk. The synthesizer engine uses a nonlinear dynamic fusion conformal process for the final control command determination, which can be expressed by the formula: in, The target collaborative final voltage vector; This represents a protective blocking and limiting logic operator that rigorously intercepts overflow errors and extracts the smaller parameter of the two parallel parameters within parentheses without bias.

[0056] In particular, the routine execution of this fusion process closely aligns with the industrial mandatory constraint execution requirements for high-voltage converter saturation prevention. It directly replaces the violent forced phase angle splitting and limiting mechanism that may cause abrupt changes in phasor phase angle by adopting a multi-dimensional radial geometry scaling calculation based on pure scalars. While perfectly ensuring the strength of the pre-dominant compensation support magnetic field necessary for high-voltage transient crossing, it achieves absolute angle preservation and fidelity preservation of the core collaborative excitation phase angle from the most stringent stator and rotor magnetic interaction level. It also eliminates the root cause of desynchronization that may induce large generators to detach from the grid-connected rotating shaft system during the transformation of the synchronous condenser into a non-functional source provider from within the final control digital wave generation source channel.

[0057] In summary, the collaborative control method for excitation systems compatible with both generator and synchronous condenser operating conditions according to the embodiments of this application is explained. It captures the time-varying inertial parameters of the unit during operating condition transitions in real time through decoupling, reconstruction, and pattern recognition of electromechanical sensor data. Based on this, a state-space observation mechanism based on flux linkage continuity prediction is constructed, thereby directly transforming the evolution of the system's inertial boundary into feedforward compensation voltage. This collaborative control strategy, by introducing dynamic time-varying inertia as a compensation benchmark, can overcome the adjustment dead zone of traditional proportional-integral converters, directionally eliminating flux linkage deviations within milliseconds, effectively suppressing transient reactive power drops. The resulting smooth switching not only significantly improves the control robustness of the excitation system under extreme transient conditions but also fundamentally avoids actuator saturation risks and shaft disconnection shutdown hazards, greatly enhancing the voltage disturbance rejection capability of synchronous units in new power systems.

[0058] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for coordinated control of an excitation system compatible with both power generation and synchronous condenser operating conditions, characterized in that, include: S1: Perform high-frequency stripping and coordinate system decoupling reconstruction on the acquired raw sensor data stream to obtain the synchronous fundamental frequency feature dataset; S2: Perform dual-condition state feature extraction and pattern recognition on the synchronous base frequency feature dataset to obtain the condition switching flag and the system time-varying inertial parameters; S3: Based on the system's time-varying inertial parameters and operating condition switching indicators, the flux continuity prediction and physical difference comparison are performed on the synchronous base frequency characteristic dataset to obtain the stator and rotor flux deviation matrix. S4: The time-varying inertial parameters of the system are used as integral time constants to process the stator and rotor flux linkage deviation matrix to obtain the feedforward compensation voltage vector; S5: Perform dual coordinated control command modulation on the synchronous base frequency characteristic dataset and the feedforward compensation voltage vector to obtain the coordinated excitation trigger pulse signal.

2. The method for coordinated control of an excitation system compatible with both generator and synchronous condenser operating conditions according to claim 1, characterized in that, The raw sensor data stream includes stator instantaneous voltage, stator instantaneous current, rotor excitation current, rotor physical speed, active power, reactive power, and prime mover tripping status.

3. The method for coordinated control of an excitation system compatible with both generator and synchronous condenser operating conditions according to claim 2, characterized in that, Step S1 includes: Kalman denoising and fundamental wave reconstruction are performed on the raw sensor data stream to obtain a clean electromechanical dataset; Orthogonal axis virtual electric compression zero compensation and fundamental frequency reference electric angle time delay tracking are performed on the clean electromechanical dataset to obtain the time-domain synchronization correlation vector; The time-domain synchronization correlation vector is decoupled by reference electrical angle projection and assembled with normalized parameters to obtain the synchronization fundamental frequency feature dataset.

4. The method for coordinated control of an excitation system compatible with both generator and synchronous condenser operating conditions according to claim 1, characterized in that, Step S2 includes: Active power drop comparison phase detection and tripping logic determination are performed on the synchronous base frequency feature dataset to obtain the operating condition switching flag and the transition state feature subset for longitudinal transmission; Using the digital polarity transition command of the operating condition switching flag as the activation and wake-up condition, virtual inertia mapping and time-varying parameter deduction are performed on the transition state feature subset to obtain the system's time-varying inertial parameters.

5. The method for coordinated control of an excitation system compatible with both generator and synchronous condenser operating conditions according to claim 1, characterized in that, Step S3 includes: In response to the operating condition switching flag being in the active state, the system's time-varying inertial parameters are imported as the nonlinear decay base into the factory-specified constant transfer coefficients of the full-dimensional state-space observer for adaptive topology scaling allocation to obtain the reshaped state-space matrix; Using the reshaped state space matrix as the core prediction term, state accumulation error drift suppression and multi-step continuous definite integral solution are performed on the synchronous fundamental frequency feature dataset to obtain the theoretical coherent flux reference value. The stator and rotor flux deviation matrix is ​​obtained by performing full-dimensional reduced-order measured mapping and differential subtraction abrupt span calculation on the synchronous fundamental frequency characteristic dataset and theoretical coherent flux reference value.

6. The method for coordinated control of an excitation system compatible with both generator and synchronous condenser operating conditions according to claim 1, characterized in that, Step S4 includes: Multidimensional orthogonal decoupling extraction of flux linkage deviation is performed on the stator and rotor flux linkage deviation matrix to obtain the biaxial disturbance component sequence; Using the system's time-varying inertial parameters as the denominator of the inverse proportional time correction, the biaxial disturbance component sequence is subjected to disturbance rejection propagation based on time-varying inertia to obtain the transient compensation sequence. The transient compensation sequence is calibrated and normalized and repackaged with the feedforward vector to obtain the feedforward compensation voltage vector.

7. The method for coordinated control of an excitation system compatible with both generator and synchronous condenser operating conditions according to claim 1, characterized in that, Step S5 includes: The synchronous base frequency characteristic dataset is compared with the terminal voltage tracking error and the integral accumulation correction to eliminate the steady error is performed to obtain the base voltage duty cycle command. The base voltage duty cycle command and the feedforward compensation voltage vector are decoupled and fused by metastable state command feedforward to obtain the target cooperative voltage vector. Space vector pulse width digital modulation is performed on the target coordinated voltage vector to obtain the coordinated excitation trigger pulse signal.

8. The method for coordinated control of an excitation system compatible with both generator and synchronous condenser operating conditions according to claim 7, characterized in that, The base voltage duty cycle command and the feedforward compensation voltage vector are decoupled and fused using a metastable state command feedforward method to obtain the target cooperative voltage vector, including: Transient conflict coupling degree identification is performed on the base voltage duty cycle command and the feedforward compensation voltage vector to obtain the control conflict coupling degree; Cross-decoupling and integral wind saturation suppression adjustment are performed on the base voltage duty cycle command, feedforward compensation voltage vector and control conflict coupling degree to obtain the corrected base voltage command; The target cooperative voltage vector is obtained by conformal orthogonal synthesis based on dynamic margin of the modified base voltage command and the feedforward compensation voltage vector.

Images