Island new energy power grid control and protection collaborative system and method based on hierarchical feature space

Abstract

The application provides an island new energy power grid control and protection collaborative system and method based on a hierarchical feature space. The method uses an operation core in an asymmetric multiprocessor architecture to collect power grid transient waveforms at a high frequency, constructs a hierarchical feature space containing time and frequency dimensions, accurately determines system faults through Mahalanobis distance, wakes up a management core through a high-speed inter-core bus within 1.5 ms at the moment of fault removal by protection, directly calls a preloaded source storage collaborative control strategy by skipping bottom layer compilation, and issues a power jump instruction. The application belongs to the technical field of power system automation and relay protection, breaks the traditional state of mutual isolation of protection and control, realizes that the response delay of full-link control and protection linkage is less than 10 ms, breaks the response bottleneck of more than 50 ms of traditional devices, and provides faster steady-state support capability for an island power grid.

Description

Technical Field

[0001] This invention belongs to the field of power system automation and relay protection technology. Specifically, it relates to a control and protection collaborative system and method for isolated new energy power grids based on hierarchical feature space. Background Technology

[0002] Driven by the global energy transition and the "dual-carbon" goal, the penetration rate of new energy sources such as wind power and photovoltaic power in modern power systems continues to rise. However, some remote or specific regional new energy power plants, limited by their construction location and the grid structure of the local main power grid, are prone to disconnection from the main grid when the grid experiences operational anomalies or external faults, thus forming regional "island-operating" microgrid systems. In island mode, the energy supply of the microgrid is highly dependent on local photovoltaic arrays, wind turbines, and energy storage systems. Without the support of the main grid, the system as a whole exhibits typical characteristics of "low inertia and weak damping." This change in physical characteristics makes islanded power grids face severe challenges, including extremely complex energy dispatching and an exponentially increasing difficulty in stability control.

[0003] Currently, traditional secondary systems and monitoring and protection devices used in renewable energy power plants generally exhibit the following serious shortcomings when dealing with the complex operating conditions of isolated microgrids: First, the traditional programming model has a lengthy cycle. Existing secondary device strategy configurations heavily rely on professional technicians to develop tedious low-level code, often taking months from logic design and code writing to final on-site debugging. This long-cycle, inefficient model is simply unsuitable for the flexible operation requirements of renewable energy power plants, which require "multi-scenario, rapid iteration." When faced with scenarios such as instantaneous peak photovoltaic output or drastic weather changes, it is highly susceptible to severe wind and solar curtailment due to untimely strategy updates, resulting in huge economic losses. Second, fixed logic devices have extremely poor adaptability. Existing monitoring and protection devices mostly adopt factory-fixed logic designs, while the technical routes of renewable energy inverters and energy storage converters from different manufacturers vary significantly. Whenever a power plant is expanded or a new equipment model is replaced, repeated customized development and protocol integration of the secondary system are required, which not only infinitely extends the project construction cycle but also keeps equipment modification costs high. Third, the response delay is too large and the control and protection systems are isolated from each other. Currently, the total response delay of most embedded secondary devices is generally above 50ms. In islanded mode, when faced with sudden voltage drops or rapid fluctuations in renewable energy output, the traditional relay protection mode, relying on a single component, suffers from a complete disconnect between protection actions (fault clearing) and control actions (active / reactive power regulation) due to the lack of global spatiotemporal characteristic coordination. This isolation prevents the device from issuing timely control commands to compensate for power deficits upon fault clearing, easily triggering microgrid frequency collapse and cascading grid disconnection accidents at power plants. Therefore, designing an optimized control system that can accurately identify faults based on transient characteristics, overcome the traditional 50ms delay bottleneck, and achieve multi-device source-storage coordinated linkage, has become a pressing technical challenge in the field of smart grids, addressing the "low inertia" characteristics of islanded grids. Summary of the Invention

[0004] To address the technical problems of large response delays, isolated control and protection systems, and susceptibility to interference in fault identification in existing technologies, this invention provides a control and protection coordination system and method for isolated new energy power grids based on hierarchical feature space. This method enables isolated microgrids to accurately identify internal faults and external interferences under different degrees of disturbance, relying on a multi-core AMP hardware architecture, and to complete the rapid linkage of protection actions and source-storage power compensation control within milliseconds, thereby ensuring the stability of the microgrid system's frequency and voltage and improving the overall power supply reliability.

[0005] The technical problem to be solved by the present invention is achieved through the following technical solution:

[0006] A hierarchical feature space-based control and protection coordination method for isolated renewable energy grids is applied to isolated renewable energy grids that include photovoltaic, wind power, energy storage devices, and devices employing an asymmetric multiprocessing architecture. The asymmetric multiprocessing architecture is explicitly divided into a computational core subsystem and a management core subsystem. The method includes the following steps:

[0007] S1: Based on the multi-core asymmetric multiprocessing architecture, the computing core subsystem performs full-domain data acquisition to obtain the steady-state data and transient waveforms of the power grid, and constructs an equivalent model processed by the management core subsystem;

[0008] S2: Based on the steady-state data and transient waveforms collected in step S1, transient feature parameters are extracted from the computing core subsystem to construct a hierarchical feature space. The transient fault identification and relay protection action determination are performed by comparing the calculated Mahalanobis distance with the dynamic adaptive threshold.

[0009] S3: When step S2 determines that there is an internal system fault and triggers the cut-off action, the management core subsystem is woken up through the high-speed inter-core bus, and the pre-loaded source-storage collaborative control strategy is called to generate energy storage dynamic compensation instructions to perform high-speed linkage of protection and control, so that the microgrid system can be restored to stable operation.

[0010] Furthermore, step S1 specifically includes: constructing an asymmetric multiprocessing hardware architecture consisting of a management core and a computing core. The computing core directly connects to the underlying power grid electrical measurement interface and is responsible for driving 100kHz high-frequency sampling, simultaneously acquiring steady-state data and transient waveforms of photovoltaic, wind turbine, and energy storage devices. The management core is responsible for the analysis of visualization graphics components, system communication, and background scheduling, mapping the complex multi-device system to the equivalent model.

[0011] Furthermore, step S2 specifically includes: extracting transient characteristic parameters of the device from the computing core, the transient characteristic parameters including transient energy and mutation rate, thereby constructing a hierarchical feature space matrix containing time and frequency dimensions; using a transient differential protection algorithm to calculate the Mahalanobis distance from the feature vector at the time of fault occurrence to the feature plane of normal operation; when the distance exceeds the dynamic adaptive threshold controlled by the real-time operation mode of the station, distinguishing between internal and external faults, and executing relay protection action judgment and underlying FPGA hardware-level trip output.

[0012] Furthermore, the expressions for calculating the transient characteristic parameters and the Mahalanobis distance are as follows:

[0013] ;

[0014] ;

[0015] ;

[0016] ;

[0017] Among them, E trans The transient characteristic energy within the time window Δt, trans (t) represents the transient high-frequency components extracted by the sliding window Fourier transform, ΔI is the current mutation rate, THD2 is the second harmonic distortion rate, I1 is the RMS value of the fundamental current, I2 is the RMS value of the second harmonic current, F is the multidimensional hierarchical eigenvector, and D... M The Mahalanobis distance, Let ∑ be the expected mean of the historical eigenvectors under normal steady-state operating conditions, and let ∑ be the covariance matrix.

[0018] Furthermore, step S3 specifically includes: when the computing core subsystem determines that there is an internal system fault and triggers a disconnection action, the computing core subsystem directly wakes up the linkage control strategy in the management core through a high-speed inter-core shared memory mechanism with a delay of less than 1.5ms; the management core subsystem performs steady-state active and reactive power balance calculations based on the fault disconnection result of step S2, directly calls the pre-loaded source-storage collaborative control strategy, generates dynamic energy storage compensation instructions, dynamically adjusts the output of healthy energy storage devices, and rapidly issues power jump instructions through the underlying hardware to smooth grid frequency fluctuations.

[0019] Furthermore, the calculation formula for the energy storage dynamic compensation command is as follows:

[0020] ;

[0021] in, The corrected instantaneous compensation target value, This represents the initial energy storage power before the fault is cleared. To cut off the lost active power of photovoltaic and other new energy sources, K p This is the frequency droop adjustment coefficient. This represents the real-time frequency of an isolated power grid after being disturbed. It serves as the reference frequency for the internal rated frequency of the microgrid.

[0022] A hierarchical feature space-based control and protection coordination system for isolated renewable energy power grids, applicable to the method described above, includes:

[0023] The computing core system has an input terminal for direct connection to the underlying power grid electrical measurement interface. The computing core system includes a digital signal processor for performing 100kHz high-frequency sampling, hierarchical feature space mapping, Mahalanobis distance fault identification model calculation, and relay protection action determination.

[0024] The management core system has an output terminal for connecting to the control port of energy storage and new energy equipment. The management core system includes a non-real-time processor for performing visualization graphics component analysis, steady-state active / reactive power balance calculation, and pre-loading source-storage collaborative control strategy.

[0025] A high-speed inter-core bus uses a shared memory mechanism to connect the computing core subsystem and the management core system bidirectionally, enabling ultra-fast physical-level communication.

[0026] The underlying field-programmable gate array module has its input terminal connected to the output terminal of the computing core subsystem and the management core subsystem, and is used to receive instructions and execute hardware-level trip output and rapidly issue energy storage power jump instructions.

[0027] Furthermore, the computing core subsystem includes:

[0028] The high-frequency sampling module is used to execute 100kHz sampling drive to synchronously acquire steady-state data and transient waveforms of photovoltaic, wind turbine, and energy storage devices.

[0029] The feature space construction module is used to extract transient energy and mutation rate to construct a hierarchical feature space matrix;

[0030] The Mahalanobis distance calculation module is used to calculate the Mahalanobis distance from the fault feature vector to the normal feature plane.

[0031] The fault determination and tripping module is used to compare the Mahalanobis distance with the dynamic adaptive threshold, and to execute hardware-level tripping output through the underlying field-programmable gate array module when an internal fault is determined.

[0032] Furthermore, the management core subsystem includes:

[0033] The steady-state calculation module is used to perform steady-state active and reactive power balance calculations based on the fault clearing results determined by the computing core subsystem, so as to determine the power deficit of the system.

[0034] The preloading strategy module is used to skip the underlying compilation and directly call the preloaded source storage coordination control strategy;

[0035] The compensation instruction generation module is used to generate a dynamic energy storage compensation instruction that includes a frequency droop correction term based on the power deficit.

[0036] Furthermore, the underlying field-programmable gate array module is specifically used to execute hardware-level trip output when receiving the trip instruction from the computing core subsystem, and to rapidly issue a power jump instruction to the healthy energy storage converter when receiving the compensation instruction from the management core subsystem; the high-speed inter-core bus is a physical-level shared memory bus with a latency strictly less than 1.5ms.

[0037] The present invention has the following beneficial effects:

[0038] Based on existing microgrid protection and control theories, this invention innovatively analyzes the decisive influence of multi-dimensional hierarchical transient characteristics on fault identification accuracy, and the significant advantage of multi-core AMP architecture physical-level communication in shortening protection-control linkage delay time. Compared to traditional measurement and control protection devices based on fixed logic or single-core architecture, this invention fully leverages the collaborative discrimination advantages of feature space Mahalanobis distance and adaptive action threshold, providing more accurate fault / disturbance boundary delineation (resistance to high transition resistance interference) for isolated power grids facing complex operating conditions such as high load switching and high-resistance grounding. Furthermore, the ultra-fast hardware and software linkage mechanism based on the AMP architecture ensures that the total response time from fault identification and clearing to energy storage power compensation is strictly controlled within 10ms, breaking the bottleneck of response times greater than 50ms caused by the heavy reliance on network protocol stacks in traditional devices, and greatly improving the power supply reliability and disturbance resistance of isolated renewable energy power plants. Attached Figure Description

[0039] Figure 1 This is a hardware platform architecture diagram of a control and protection collaborative system based on a multi-core AMP architecture in an embodiment of the present invention.

[0040] Figure 2 This is a 3D scatter plot of Mahalanobis distance for fault identification in the hierarchical feature space of this invention.

[0041] Figure 3 This is a comparison diagram of the step response latency between the traditional architecture and the multi-core AMP architecture in this embodiment of the invention.

[0042] Figure 4 This is a timing logic diagram of the linkage between islanded power grid protection actions and energy storage control in an embodiment of the present invention.

[0043] Figure 5 This is a flowchart of a control and protection coordination method for isolated new energy power grids based on hierarchical feature space, according to an embodiment of the present invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] like Figure 1The diagram shows the hardware platform architecture of the control and protection coordination system based on a multi-core AMP architecture in this embodiment of the invention. At the physical level, the control and protection coordination embedded device abandons the traditional single-core or homogeneous multi-core architecture and adopts an asymmetric multiprocessing (AMP) architecture, clearly dividing it into two independent and collaborative parts: a "computation core subsystem (DSP / real-time core)" and a "management core subsystem (ARM / non-real-time core)". The computation core subsystem directly connects to the underlying power grid electrical measurement interface (such as analog quantities from CT / PT transformers), responsible for 100kHz high-frequency sampling drive and the extremely time-consuming extraction of transient energy and mutation rates, calculation of the Mahalanobis distance fault identification model, and ultimately executing relay protection action determination and FPGA hardware-level trip output. The management core subsystem connects to isolated energy storage / new energy equipment (such as the PCS / photovoltaic inverter control port), focusing on high-level visualization of graphical components, steady-state active / reactive power balance calculation, and pre-loading of source-storage coordinated control strategies. The two subsystems abandon the traditional time-consuming network protocol and adopt a "high-speed inter-core bus (shared memory mechanism)" for ultra-fast physical-level communication. The inherent latency of this mechanism is strictly less than 1.5ms, which constitutes the physical basis for the millisecond-level control and protection linkage of this invention.

[0046] like Figure 5 As shown, this embodiment of the invention provides a control and protection coordination method for isolated renewable energy grids based on hierarchical feature space. This method is applied to isolated renewable energy grids that include photovoltaic, wind power, and energy storage devices, and employ an asymmetric multiprocessing architecture. The asymmetric multiprocessing architecture is clearly divided into a computational core subsystem and a management core subsystem. The method includes the following steps:

[0047] S1: Based on the multi-core asymmetric multiprocessing architecture, the computing core subsystem performs full-domain data acquisition to obtain the steady-state data and transient waveforms of the power grid, and constructs an equivalent model processed by the management core subsystem;

[0048] S2: Based on the steady-state data and transient waveforms collected in step S1, transient feature parameters are extracted from the computing core subsystem to construct a hierarchical feature space. The transient fault identification and relay protection action determination are performed by comparing the calculated Mahalanobis distance with the dynamic adaptive threshold.

[0049] S3: When step S2 determines that there is an internal system fault and triggers the cut-off action, the management core subsystem is woken up through the high-speed inter-core bus, and the pre-loaded source-storage collaborative control strategy is called to generate energy storage dynamic compensation instructions to perform high-speed linkage of protection and control, so that the microgrid system can be restored to stable operation.

[0050] Furthermore, step S1 specifically includes: constructing an asymmetric multiprocessing hardware architecture consisting of a "management core (non-real-time core) + computing core (real-time core)" on the hardware platform. The computing core directly interfaces with the underlying instrument transformer analog signals, responsible for driving 100kHz high-frequency sampling and simultaneously acquiring steady-state data and transient waveforms from photovoltaic, wind turbine, and energy storage devices. The management core, on the other hand, interfaces with the device's control port, responsible for visual graphics component analysis, system communication, and background scheduling, mapping the complex multi-device physical system into an equivalent model that is easy to compute.

[0051] Furthermore, the transient fault identification in step S2 is specifically obtained and determined through the following steps: When a disturbance occurs in the isolated power grid, the computing core collects the three-phase current at high frequency, extracts the transient high-frequency components using sliding window Fourier transform, and then extracts the transient characteristic parameters of the equipment, including transient energy, current mutation rate, and second harmonic distortion rate within the time window. A hierarchical multidimensional feature space matrix containing time and frequency dimensions is constructed from the above parameters. A transient differential protection algorithm is adopted to calculate the Mahalanobis distance from the current sampled feature vector to the normal operating state feature plane in real time when the fault occurs; the Mahalanobis distance is compared with the dynamic adaptive action threshold controlled by the real-time operation mode of the substation; when the Mahalanobis distance exceeds the set critical threshold, it is accurately identified as an internal fault of the microgrid, and the underlying FPGA hardware executes the trip output; if the threshold is not exceeded, it is regarded as a normal external disturbance (such as large load switching), and no action is taken.

[0052] Furthermore, the mathematical expressions required for extracting transient feature parameters and calculating Mahalanobis distance are as follows:

[0053]

[0054]

[0055]

[0056]

[0057] Among them, E trans The transient characteristic energy within the time window Δt, trans (t) represents the transient high-frequency components extracted by the sliding window Fourier transform, ΔI is the current mutation rate, THD2 is the second harmonic distortion rate, I1 is the RMS value of the fundamental current, I2 is the RMS value of the second harmonic current, F is the multidimensional hierarchical eigenvector, and D... M The Mahalanobis distance, Let ∑ be the expected mean of the historical eigenvectors under normal steady-state operating conditions, and let ∑ be the covariance matrix.

[0058] Furthermore, the rapid linkage of protection and control in step S3 specifically includes: establishing a "protection-control linkage" mechanism the instant the computing core determines a fault and triggers a disconnection action. The computing core directly issues a hardware interrupt through a high-speed inter-core physical bus (shared memory mechanism) with a delay strictly less than 1.5ms, waking up the linkage control strategy in the management core, eliminating the cumbersome transmission of traditional network protocol stacks. The management core calculates the system's power deficit based on the steady-state active / reactive power balance before and after fault disconnection, directly skips the compilation of the underlying code, calls the pre-loaded source-storage collaborative control strategy, and generates a dynamic energy storage compensation instruction containing a frequency droop correction term. This instruction is issued through the underlying FPGA hardware-level acceleration channel, dynamically adjusting the output of the healthy energy storage converter (PCS) to perform instantaneous discharge, thus mitigating the frequency and voltage drops caused by the loss of source-side support in the isolated power grid.

[0059] Furthermore, the calculation formula for the energy storage dynamic compensation command is based on the steady-state active power balance equation of the islanded microgrid, and its final compensation target value expression is:

[0060]

[0061] in, The corrected instantaneous compensation target value, This represents the initial energy storage power before the fault is cleared. To eliminate the lost active power from renewable energy sources such as photovoltaics, Kp is the frequency droop adjustment coefficient. This represents the real-time frequency of an isolated power grid after being disturbed. It serves as the reference frequency for the internal rated frequency of the microgrid.

[0062] The following example illustrates this:

[0063] First, during normal operation or disturbance of the isolated microgrid, the computational core subsystem obtains the three-phase current i(t) through 100kHz high-frequency sampling, and extracts the transient high-frequency components hidden in the fundamental frequency using sliding window Fourier transform. trans (t). Since a single parameter cannot fully reflect complex faults, the actual power calculation module will combine the following formula to calculate the transient characteristic energy E within the time window Δt. trans The three parameters—current mutation rate ΔI, which characterizes the instantaneous severity of a fault, and second harmonic distortion rate THD2, which reflects the degree of power quality degradation—together construct a multidimensional hierarchical feature vector F.

[0064]

[0065]

[0066] Among them, Etrans The transient characteristic energy within the time window Δt, trans (t) represents the transient high-frequency component extracted by sliding window Fourier transform, ΔI is the current mutation rate, THD2 is the second harmonic distortion rate, I1 is the effective value of the fundamental current, I2 is the effective value of the second harmonic current, and F is the multidimensional hierarchical feature vector.

[0067] Subsequently, based on historical data accumulated under normal steady-state operating conditions, the system obtains the expected mean μ and covariance matrix ∑ of the eigenvectors. According to the Mahalanobis distance criterion, the Mahalanobis distance D of the currently sampled eigenvector F in the hierarchical feature space is calculated in real time. M .

[0068]

[0069] Among them, D M The Mahalanobis distance, Let ∑ be the expected mean of the historical eigenvectors under normal steady-state operating conditions, and let ∑ be the covariance matrix.

[0070] The system will introduce an adaptive action threshold D based on the real-time proportion of renewable energy output within the microgrid. set .like Figure 2 The image shown is a 3D scatter plot of Mahalanobis distance for fault identification in a hierarchical feature space according to an embodiment of the present invention. From this three-dimensional mapping space, the significant differences in spatial distribution between steady-state operation, external disturbances (large load switching), and internal actual faults can be clearly seen. When the system experiences load switching such as the starting of a large motor, although transient fluctuations will occur, the calculated Mahalanobis distances tend to cluster near the steady-state cluster and will not exceed the adaptive action threshold D. set The system determines that it is in an anti-interference state and does not operate. However, once a highly concealed fault occurs, such as a high-resistance grounding fault, the position of its comprehensive characteristic vector in space will deviate sharply, causing the calculated D to... M Rapidly greater than D set At this point, the voltage sag and fault detection model is triggered, the system clearly determines it to be an internal fault, and executes the protection trip command. Once the computational core triggers the protection disconnection action, the system immediately enters the second stage of control and protection coordination—rapid linkage. In islanded operation mode, the microgrid must maintain strict active power balance, that is:

[0071]

[0072] In the above formula, the first term on the left is the total active power output of all grid-connected photovoltaic branches in the isolated microgrid, the second term is the total active power output of all wind turbine units, the third term is the current active power output of the energy storage system, and the right side is the active power demand of the microgrid load. This formula is used to determine the steady-state active power balance benchmark of the source side, energy storage side and load side during normal operation.

[0073] If a fault causes a photovoltaic branch to be disconnected, the system will instantly experience an active power deficit. Upon receiving an interrupt signal from the processing core via shared memory, the management core immediately activates the preloading strategy, based on the calculation formula:

[0074]

[0075] In the above formula, the left side is the corrected dynamic compensation power command for energy storage; the first term on the right side is the energy storage output before fault clearing, the second term is the active power deficit of new energy caused by the cleared photovoltaic branch, and the third term is the frequency droop correction term, K. p f is the frequency droop adjustment coefficient. ref f is the reference frequency for the microgrid's rated frequency. sys This refers to the real-time frequency of the isolated power grid after a disturbance. The aforementioned active power balance formula is used to provide the steady-state power reference for the source, storage, and load before the fault. When the eigenvector formed by the transient characteristic parameters exceeds the adaptive action threshold and triggers protection disconnection according to the Mahalanobis distance formula, the management core calculates the active power deficit after disconnection based on this steady-state reference, and then superimposes the frequency deviation to form the corrected dynamic compensation power command for energy storage. Therefore, the transient characteristic parameters and the Mahalanobis distance formula are used to determine whether protection is triggered and the triggering time, the active power balance formula is used to determine the power deficit after disconnection, and the energy storage...

[0076] The dynamic compensation formula is used to convert the deficit into a compensation instruction issued to the healthy energy storage converter.

[0077] To more clearly demonstrate the breakthrough in response speed of this invention, a 100% renewable energy islanded microgrid model incorporating photovoltaic, wind power, and energy storage was constructed using MATLAB, and Dset was defined. Specific data for different scenarios were obtained through experiments (see Table 1 below):

[0078] Table 1. Hierarchical characteristic parameters and fault identification results of isolated microgrids under different disturbance scenarios.

[0079]

[0080] Combining the data in Table 1 and Figure 3 (Comparison of response latency step between traditional and multi-core AMP architectures) Figure 4 (Timing Logic Diagram) Taking the most severe single-phase metallic short circuit in the photovoltaic area as an example, the linkage timing is broken down in detail:

[0081] T = 0 ms: A fault occurs in the photovoltaic area, and the computing core collects transient voltage and current at high frequency.

[0082] T = 3.5 ms: The computational kernel calculates the hierarchical characteristics and Mahalanobis distance, determines it to be an internal fault, and then triggers the protection trip to disconnect the faulty photovoltaic system. This process takes T seconds. detect = 3.5 ms.

[0083] T = 5.0 ms: The processing core triggers a hardware interrupt, pushing the power deficit to the management core via the high-speed inter-core bus. This process takes T seconds. ipc = 1.5 ms.

[0084] T = 6.5 ms: Management core wake-up preloading strategy, calculate the energy storage dynamic compensation amount according to formula (6).

[0085] T < 9.0 ms: FPGA hardware accelerates the issuance of power boost commands, causing the energy storage converter (PCS) to discharge instantaneously, smoothing grid frequency fluctuations, and the control execution time T exec ≈ 2.5 ms.

[0086] Therefore, the total latency of the entire control and protection coordination process in this embodiment is as follows: .like Figure 3 As shown in the comparison curves, when faced with the same fault, the traditional single-core architecture often experiences energy storage hysteresis compensation around T=75ms due to the layered interaction of the network protocol stack, resulting in a gap of Δt > 50ms, which can easily lead to grid collapse. In contrast, this invention employs a multi-core AMP architecture with physical-level shared memory linkage, achieving ultra-fast compensation of Δt < 10ms, breaking through traditional technical barriers.

[0087] Therefore, the AMP hardware architecture system based on hierarchical feature space proposed in this invention enables isolated microgrids to resist interference and accurately identify internal faults in complex electrical environments by relying solely on the built-in transient measurement algorithm. At the same time, the rapid power support for source-storage collaboration is achieved by relying on the ultra-fast inter-core communication mechanism, which greatly reduces the impact of transient faults on important and sensitive loads in the isolated area and significantly improves the power quality and power supply resilience of the microgrid.

[0088] This invention also provides a hierarchical feature space-based islanded renewable energy grid control and protection coordination system applicable to the method described above, comprising:

[0089] The computing core system has an input terminal for direct connection to the underlying power grid electrical measurement interface. The computing core system includes a digital signal processor for performing 100kHz high-frequency sampling, hierarchical feature space mapping, Mahalanobis distance fault identification model calculation, and relay protection action determination.

[0090] The management core system has an output terminal for connecting to the control port of energy storage and new energy equipment. The management core system includes a non-real-time processor for performing visualization graphics component analysis, steady-state active / reactive power balance calculation, and pre-loading source-storage collaborative control strategy.

[0091] A high-speed inter-core bus uses a shared memory mechanism to connect the computing core subsystem and the management core system bidirectionally, enabling ultra-fast physical-level communication.

[0092] The underlying field-programmable gate array module has its input terminal connected to the output terminal of the computing core subsystem and the management core subsystem, and is used to receive instructions and execute hardware-level trip output and rapidly issue energy storage power jump instructions.

[0093] Compared with the prior art, the present invention has the following features and effects:

[0094] This invention, based on a fault identification method using hierarchical feature space and Mahalanobis distance, can accurately distinguish between internal faults and external interference (such as high-load switching and high-resistance grounding), effectively solving the technical problem of traditional protection devices being susceptible to misjudgment or failure to operate due to transition resistance. Furthermore, this invention fully utilizes the physical-level division of labor between the computational core and management core in the asymmetric multiprocessing (AMP) architecture, establishing a "protection-control rapid linkage" mechanism through a high-speed inter-core bus with a latency strictly less than 1.5ms. This ensures that the total response time of the entire link, from fault identification and protection disconnection to the issuance of energy storage power compensation, is strictly controlled within 10ms, completely breaking the performance bottleneck of traditional devices relying on network protocol stacks, which results in response latency greater than 50ms. This invention overcomes the shortcomings of traditional isolated protection and control systems, achieving millisecond-level source-storage collaborative support, significantly improving the power supply reliability and disturbance resistance resilience of isolated renewable energy grids under transient disturbances.

[0095] The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for coordinated control and protection of isolated renewable energy power grids based on hierarchical feature space, characterized in that, The method is applied to an islanded renewable energy grid that includes photovoltaic, wind power, energy storage equipment, and devices employing an asymmetric multiprocessing architecture, wherein the asymmetric multiprocessing architecture is clearly divided into a computing core subsystem and a management core subsystem; the method includes the following steps: S1: Based on the multi-core asymmetric multiprocessing architecture, the computing core subsystem performs full-domain data acquisition to obtain the steady-state data and transient waveforms of the power grid, and constructs an equivalent model processed by the management core subsystem; S2: Based on the steady-state data and transient waveforms collected in step S1, transient feature parameters are extracted from the computing core subsystem to construct a hierarchical feature space. The transient fault identification and relay protection action determination are performed by comparing the calculated Mahalanobis distance with the dynamic adaptive threshold. S3: When step S2 determines that there is an internal system fault and triggers the cut-off action, the management core subsystem is woken up through the high-speed inter-core bus, and the pre-loaded source-storage collaborative control strategy is called to generate energy storage dynamic compensation instructions to perform high-speed linkage of protection and control, so that the microgrid system can be restored to stable operation.

2. The method according to claim 1, characterized in that, Step S1 specifically includes: constructing an asymmetric multiprocessing hardware architecture consisting of a management core and a computing core. The computing core directly connects to the underlying power grid electrical measurement interface and is responsible for driving 100kHz high-frequency sampling, simultaneously acquiring steady-state data and transient waveforms of photovoltaic, wind turbine, and energy storage devices. The management core is responsible for the parsing of visualization graphics components, system communication, and background scheduling, mapping the complex multi-device system to the equivalent model.

3. The method according to claim 1, characterized in that, Step S2 specifically includes: extracting transient characteristic parameters of the equipment from the computing core, the transient characteristic parameters including transient energy and mutation rate, thereby constructing a hierarchical feature space matrix containing time and frequency dimensions; using a transient differential protection algorithm to calculate the Mahalanobis distance from the feature vector at the time of the fault to the feature plane of the normal operating state; when the distance exceeds the dynamic adaptive threshold controlled by the real-time operation mode of the station, distinguishing between internal and external faults, and executing relay protection action judgment and underlying FPGA hardware-level trip output.

4. The method according to claim 3, characterized in that, The expressions for calculating the transient characteristic parameters and the Mahalanobis distance are as follows: ； ； ； ； Among them, E trans The transient characteristic energy within the time window Δt, trans (t) represents the transient high-frequency components extracted by the sliding window Fourier transform, ΔI is the current mutation rate, THD2 is the second harmonic distortion rate, I1 is the RMS value of the fundamental current, I2 is the RMS value of the second harmonic current, F is the multidimensional hierarchical eigenvector, and D... M The Mahalanobis distance, Let ∑ be the expected mean of the historical eigenvectors under normal steady-state operating conditions, and let ∑ be the covariance matrix.

5. The method according to claim 1, characterized in that, Step S3 specifically includes: when the computing core subsystem determines that there is an internal system fault and triggers a disconnection action, the computing core subsystem directly wakes up the linkage control strategy in the management core through a high-speed inter-core shared memory mechanism with a delay of less than 1.5ms; the management core subsystem performs steady-state active and reactive power balance calculations based on the fault disconnection result of step S2, directly calls the pre-loaded source-storage collaborative control strategy, generates dynamic energy storage compensation instructions, dynamically adjusts the output of healthy energy storage devices, and rapidly issues power jump instructions through the underlying hardware to smooth grid frequency fluctuations.

6. The method according to claim 5, characterized in that, The calculation formula for the energy storage dynamic compensation command is as follows: ； in, The corrected instantaneous compensation target value, This represents the initial energy storage power before the fault is cleared. To cut off the lost active power of photovoltaic and other new energy sources, K p This is the frequency droop adjustment coefficient. This represents the real-time frequency of an isolated power grid after being disturbed. It serves as the reference frequency for the internal rated frequency of the microgrid.

7. A hierarchical feature space-based islanded renewable energy grid control and protection coordination system applicable to the method described in any one of claims 1 to 6, characterized in that, include: The computing core system has an input terminal that is used to directly connect to the underlying power grid electrical measurement interface. The computing core system includes a digital signal processor for performing 100kHz high-frequency sampling, hierarchical feature space mapping, Mahalanobis distance fault identification model calculation, and relay protection action determination. The management core system has an output terminal for connecting to the control port of energy storage and new energy equipment. The management core system includes a non-real-time processor for performing visualization graphics component analysis, steady-state active / reactive power balance calculation, and pre-loading source-storage collaborative control strategy. A high-speed inter-core bus uses a shared memory mechanism to connect the computing core subsystem and the management core system bidirectionally, enabling ultra-fast physical-level communication. The underlying field-programmable gate array module has its input terminal connected to the output terminal of the computing core subsystem and the management core subsystem, and is used to receive instructions and execute hardware-level trip output and rapidly issue energy storage power jump instructions.

8. The system according to claim 7, characterized in that, The computational core subsystem includes: The high-frequency sampling module is used to execute 100kHz sampling drive to synchronously acquire steady-state data and transient waveforms of photovoltaic, wind turbine, and energy storage devices. The feature space construction module is used to extract transient energy and mutation rate to construct a hierarchical feature space matrix. The Mahalanobis distance calculation module is used to calculate the Mahalanobis distance from the fault feature vector to the normal feature plane. The fault determination and tripping module is used to compare the Mahalanobis distance with the dynamic adaptive threshold, and to execute hardware-level tripping output through the underlying field-programmable gate array module when an internal fault is determined.

9. The system according to claim 7, characterized in that, The management subsystem includes: The steady-state calculation module is used to perform steady-state active and reactive power balance calculations based on the fault clearing results determined by the computing core subsystem, so as to determine the power deficit of the system. The preloading strategy module is used to skip the underlying compilation and directly call the preloaded source storage coordination control strategy; The compensation instruction generation module is used to generate a dynamic energy storage compensation instruction that includes a frequency droop correction term based on the power deficit.

10. The system according to claim 7, characterized in that, The underlying field-programmable gate array module is specifically used to execute hardware-level trip output when it receives the trip instruction from the computing core subsystem, and to rapidly issue a power jump instruction to the healthy energy storage converter when it receives the compensation instruction from the management core subsystem; the high-speed inter-core bus is a physical-level shared memory bus with a latency strictly less than 1.5ms.

Images