A Fast Correction Method for Transient Safety Domain Based on Critical Injection Power

Abstract

This invention relates to a method for rapidly correcting the transient safety domain based on critical injected power, comprising the following steps: Step 1, predicting the critical injected power of the dominant unstable unit based on real-time measurement data acquired after a fault; Step 2, using the proportional relationship of the hyperplane coefficients of each generator in the offline simulation, and combining it with the predicted value of the critical injected power of the dominant unstable unit obtained in Step 1, rapidly correcting the transient safety domain boundary. This invention can determine the critical injected power of the dominant unstable unit based on real-time measurement data and rapidly correct and update the transient safety domain boundary.

Description

Technical Field

[0001] This invention belongs to the field of power grid security and stability analysis technology, and relates to a fast correction method for transient security domain, especially a fast correction method for transient security domain based on critical injected power. Background Technology

[0002] Traditional transient stability assessment methods are mainly based on time-domain simulation and energy functions. Due to their respective shortcomings, they are difficult to apply directly to online transient stability assessment. Synchronous phasor technology, with its unique ability to synchronously sample voltage and current waveform data with the Global Positioning System (GPS), is considered one of the most important measurement technologies in power systems today. With the gradual deployment of phasor measurement units (PMUs), data-driven and machine learning-based transient stability assessment methods have rapidly developed and become a hot topic of research. PMUs, by synchronizing sampling with the microprocessor system, can place widely distributed phasor calculations on a common time scale. The emergence of Wide Area Measurement Systems (WAMS) based on this technology enables tasks such as real-time dynamic tracking of the system, which were previously impossible. This helps operators accurately understand the grid state, quickly predict whether the power system is trending towards instability, and further delve into the evolution of transient processes.

[0003] Machine learning methods can be used to effectively perform online transient stability analysis. However, the analysis results of existing machine learning methods are difficult to apply to new operating states. When the generator output changes, it is necessary to reassess the transient stability.

[0004] Transient Security Regions (TSRs) are typically constructed based on steady-state system data for specific faults, and do not require changes in transient stability assessments as generator output fluctuates. However, this method lacks verification and correction based on actual fault information, resulting in a lack of reliable guarantees for online applications. While the TSR can be modified by searching for power critical points, real-world power grids have numerous nodes, leading to a very high computational burden. Searching for power critical points across the entire injection space is time-consuming and unsuitable for online applications.

[0005] A search revealed no published patent documents that are identical or similar to this invention. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and propose a rapid correction method for the transient safety domain based on the critical injection power. This method can determine the critical injection power of the dominant unstable unit based on real-time collected measurement data and quickly correct and update the transient safety domain boundary.

[0007] The present invention solves its practical problem by adopting the following technical solution:

[0008] A fast correction method for the transient safety domain based on critical injection power includes the following steps:

[0009] Step 1: Based on the real-time measurement data collected after the fault, predict the critical injection power of the main unstable unit;

[0010] Step 2: Using the proportional relationship of the hyperplane coefficients of each generator in the offline simulation, and combined with the predicted value of the critical injected power of the dominant unstable unit obtained in Step 1, the transient safety domain is quickly corrected.

[0011] Furthermore, the specific steps of step 1 include:

[0012] (1) Based on the real-time measurement data collected after the fault, predict the critical resection time;

[0013] (2) Under the same terminal voltage level, the critical cut-off time and the power output of the dominant unstable generator vary according to a sensitivity coefficient, which is defined as K, representing the change in critical cut-off time under unit power change. This sensitivity coefficient can be obtained through offline simulation, as shown in equation (2):

[0014]

[0015] In the formula, Δf represents the change in critical resection time, and ΔP represents the change in power.

[0016] (3) The critical injection power can be calculated using equation (3):

[0017] P c =P0+K(T) cr -y CCT (3)

[0018] In equation (3), P c It is the critical injection power, P0 is the power at the current operating point, and T cr This is the fault clearing time.

[0019] Moreover, the specific method of step (1) of step 1 is as follows:

[0020] The learning process for the critical resection time can be viewed as a mapping function. Assuming this mapping function is f, the predicted value of the critical resection time can be obtained through equation (1):

[0021] y CCT =f(x) (1)

[0022] In the formula, x is the quantity measured by the input of the critical resection time prediction network, and y CCT This is a predicted value for the critical resection time;

[0023] Furthermore, the specific method for step 2 is as follows:

[0024] After obtaining the critical injection power of the dominant unstable generator in step 1, the new TSR hyperplane coefficient is obtained according to equation (4), and then the TSR is corrected:

[0025]

[0026] In the above formula, n-1 is the dimension of the injection space excluding the balancing machine, and P i Let α be the injected power at node i. i Let be the hyperplane coefficient of node i; from the above equation, we can see that after obtaining the new power critical point P, α can be obtained. i ', and then use this formula to correct the TSR boundary.

[0027] Advantages and beneficial effects of the present invention:

[0028] 1. This invention provides a method for rapid correction of transient safety domain based on critical injected power. It utilizes machine learning to explore the mapping relationship between the critical fault clearing time and the critical injected power of the dominant unstable generator group, thereby enabling rapid calculation of the hyperplane coefficients of the transient safety domain of the dominant unstable generator group and all generators. This allows for rapid correction of the transient safety domain, achieving transient stability assessment and providing support for overall system stability control.

[0029] 2. This invention only searches for the critical injection power of the dominant unstable generator group corresponding to the system instability mode, which greatly reduces the dimensionality of the search space and improves computational efficiency.

[0030] 3. The fault clearing time used in this invention is an important reference indicator in transient stability assessment and is often used to evaluate the transient stability margin of a system. Therefore, the results obtained have universal applicability. In addition, the fault clearing time can not only be used to determine whether the system is stable, but also clearly show the distance of each generator output to the stability boundary, providing overall measurement and margin information of transient stability.

[0031] 4. The method proposed in this invention considers post-fault measurement information, thus reducing reliance on generator models and parameter information to a certain extent. It not only effectively corrects erroneous predictions of the current fault, but also allows the corrected TSR to be applied to similar fault scenarios in the future. Furthermore, this invention provides a new approach to TSR application models. The learning process of this invention can be continuously conducted online, and it is foreseeable that with the increase of time, the corrected TSR will gradually approach the true TSR, which is consistent with the characteristic of continuously improving accuracy in online learning.

[0032] 5. This invention combines offline simulation with real-time measurement data, resulting in highly accurate results and providing a new approach for the rapid correction of transient safety domain boundaries. Attached Figure Description

[0033] Figure 1 This is a graph showing the relationship between the critical cut-off time and the power output of the dominant unstable generator, provided by the present invention.

[0034] Figure 2 This is the wiring diagram of the New England 10-machine 39-node system provided by the present invention;

[0035] Figure 3 This is a diagram of the power angle curves of each generator after a fault, provided by the present invention.

[0036] Figure 4 This invention provides a transient security domain diagram dominated by G33 and G34.

[0037] Figure 5 This is a comparison chart of the errors of different methods provided by the present invention. Detailed Implementation

[0038] The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings:

[0039] A fast correction method for the transient safety domain based on critical injection power includes the following steps:

[0040] Step 1: Based on the real-time measurement data collected after the fault, predict the critical injection power of the main unstable unit;

[0041] The specific steps of step 1 include:

[0042] (1) Based on the real-time measurement data collected after the fault, predict the critical resection time:

[0043] The learning process for the critical resection time can be viewed as a mapping function. Assuming this mapping function is f, the predicted value of the critical resection time can be obtained through equation (1):

[0044] y CCT =f(x) (1)

[0045] In the formula, x is the quantity measured by the input of the critical resection time prediction network, and y CCT This is a predicted value for the critical resection time;

[0046] (2) Under the same terminal voltage level, the critical cut-off time and the power output of the dominant unstable generator vary according to a sensitivity coefficient, which is defined as K, representing the change in critical cut-off time under unit power change. This sensitivity coefficient can be obtained through offline simulation, as shown in equation (2):

[0047]

[0048] In the formula, Δf represents the change in critical resection time, and ΔP represents the change in power.

[0049] (3) The critical injection power can be calculated using equation (3).

[0050] P c =P0+K(T) cr -y CCT (3)

[0051] In equation (3), P c It is the critical injection power, P0 is the power at the current operating point, and T cr This is the fault clearing time.

[0052] In this embodiment, the working principle of step 1 is as follows:

[0053] TSRs are sets built on the active power injection space, whose boundaries consist of a series of power critical points. Accurately calculating the values ​​of these power critical points is crucial for understanding the system's stability level and ensuring transient stability. Determining the boundary injection power points within the safe domain has long been a challenging problem for researchers. Recent advancements in wide-area measurement systems and the emergence of artificial intelligence algorithms have provided new approaches to solving this problem.

[0054] As defined by the safety domain definition, fault clearing time is a crucial factor influencing its construction. The fault clearing time in a system depends on the protection and control time and is generally a constant. If, under a certain set of injected power, the critical clearing time is exactly equal to the fault clearing time, then this set of injected power represents a power critical point in a high-dimensional space. Among these, the dominant unstable generator, due to severe disturbances, is more likely to have its injected power exceed its critical injected power, requiring special attention. If the critical clearing time can be quickly determined from post-fault measurements, and a mapping relationship between the critical clearing time and the critical injected power can be established, the critical injected power of the dominant unstable generator can be quickly calculated.

[0055] Many scholars' research has laid the foundation for solving this problem. Existing literature has proven that artificial neural networks (ANNs) can be used to estimate the critical clearing time of power system faults. Other scholars have further improved upon this by combining probabilistic neural networks and radial basis function networks, fully utilizing the advantages of both to enhance the predictive ability of critical clearing time and effectively improve prediction accuracy. These research results have made it possible to quickly calculate the critical injection power online. This invention predicts the critical clearing time based on previous research. The learning process of the critical clearing time can be viewed as a mapping function. Assuming this mapping function is f, the predicted value of the critical clearing time can be obtained through equation (1).

[0056] y CCT =f(x) (1)

[0057] In the formula, x is the quantity measured by the input of the critical resection time prediction network, and y CCT This is a predicted value for the critical cutoff time. This invention studies the mapping relationship between the critical cutoff time and the injected power of the dominant unstable generator. When other control variables are kept constant and only the injected power of the dominant unstable generator is changed, the change in critical cutoff time is as follows: Figure 1 As shown. Figure 1 The study demonstrates how the critical cut-off time varies with the power output of the dominant unstable generator under different terminal voltage levels.

[0058] As can be seen from the figure above, under the same terminal voltage level, the critical cut-off time varies with the power output of the dominant unstable generator according to a sensitivity coefficient, which has a good linear effect. This invention defines this sensitivity coefficient as K, representing the change in critical cut-off time per unit power change. This sensitivity coefficient can be obtained through offline simulation, as shown in equation (2):

[0059]

[0060] In the formula, Δf represents the change in critical resection time, and ΔP represents the change in power. The critical injection power can then be calculated using formula (3).

[0061] P c =P0+K(T) cr -y CCT (3)

[0062] In equation (3), P c It is the critical injection power, P0 is the power at the current operating point, and T cr This is the fault clearing time.

[0063] Step 2: Using the proportional relationship of the hyperplane coefficients of each generator in the offline simulation, and combined with the predicted value of the critical injected power of the dominant unstable unit obtained in Step 1, the transient safety domain is quickly corrected.

[0064] The specific method for step 2 is as follows:

[0065] After obtaining the critical injection power of the dominant unstable generator in step 1, the new TSR hyperplane coefficient is obtained according to equation (4), and then the TSR is corrected:

[0066]

[0067] In the above formula, n-1 is the dimension of the injection space excluding the balancing machine, and P i Let α be the injected power at node i. i Let be the hyperplane coefficient of node i (' represents the online corrected TSR). From the above equation, it can be seen that after obtaining the new power critical point P, α can be calculated. i ', and then use this formula to correct the TSR boundary.

[0068] In this embodiment, the working principle of step 2 is as follows:

[0069] Currently, the main application mode of transient security domains (TSRs) is "offline calculation, online application," meaning that after constructing the TSR offline, it is directly applied to online stability assessment. This method relies excessively on the accuracy of the model. During normal operation of a power system, it is subjected to various small disturbances almost constantly, such as transformer tap adjustment and overhead line swaying due to wind. Simulation calculations can generate batches of fault data, but it is difficult to guarantee the consistency between simulation and actual data. For accident analysis of actual power grids, inconsistencies between simulation results and actual fault results frequently occur.

[0070] Therefore, this invention proposes a novel approach for rapidly correcting TSR. In practical applications, a mapping relationship is established between the critical cut-off time and the critical injected power of the dominant unstable generator. The proportional relationship of the hyperplane coefficients of each generator in the offline simulation is maintained, meaning the critical injected power of each generator changes at the same proportional rate. Thus, after obtaining the critical injected power of the dominant unstable generator, the new TSR hyperplane coefficients can be calculated according to equation (4), thereby correcting the TSR.

[0071]

[0072] In the above formula, n-1 is the dimension of the injection space excluding the balancing machine, and P i Let α be the injected power at node i. i Let be the hyperplane coefficient of node i (' represents the online corrected TSR). From the above equation, it can be seen that after obtaining the new power critical point P, α can be calculated. i', and then use this formula to correct the TSR boundary.

[0073] Traditional safety domain (TSR) construction relies solely on offline computation, while the method proposed in this invention considers post-fault measurement information, thus reducing reliance on generator models and parameter information to some extent. This not only effectively corrects erroneous predictions of current faults but also allows the corrected TSR to be applied to similar fault scenarios in the future. Furthermore, this invention provides a new approach to TSR application models. The learning process of this invention can be continuously conducted online, and it is foreseeable that with increasing time, the corrected TSR will gradually approach the true TSR, which aligns with the characteristic of continuously improving accuracy in online learning.

[0074] The working principle of this invention is:

[0075] Different generators inject power corresponding to different hyperplane coefficients mean that different generators have varying impacts on the system's transient stability. During TSR (Transient Stability Retention) construction, the current operating point is at different distances from each hyperplane. In unstable samples, the current operating point exceeds the TSR boundaries dominated by the dominant unstable generators; in stable samples, the current operating point is very close to the TSR boundaries dominated by the dominant unstable generators. Therefore, operators often focus more on the safety domain boundaries of the dominant unstable generators, making the rapid determination and correction of these boundaries crucial. Furthermore, determining the critical injected power of the generators is also a challenge in TSR construction. Previous methods relied on time-domain simulations, consuming significant computation time and unsuitable for online applications. Therefore, researching how to quickly determine the critical injected power of the dominant unstable generators in the system is of significant guiding importance for rapidly correcting TSR.

[0076] This invention proposes a novel method for rapidly correcting the transient safety domain based on the critical injection power. This method can predict the critical injection power of the dominant unstable unit based on real-time measurement information after a fault, and further realize the rapid updating of the transient safety domain boundary.

[0077] (1) First, it was theoretically analyzed that different generators have different effects on the transient stability of the system. The boundary of the transient safety domain mainly depends on the dominant unstable unit. Finding the critical injection power of the dominant unstable unit in the system is of great guiding significance for quickly correcting the transient safety domain.

[0078] (2) Then, based on wide-area measurement data and artificial intelligence algorithms, a mapping relationship between critical cut-off time and critical injection power of dominant unstable unit was established, and a fast solution method for critical injection power of dominant unstable unit was proposed.

[0079] (3) Finally, using the proportional relationship of the hyperplane coefficients of each generator in the offline simulation and combined with the predicted value of the critical injected power of the dominant unstable unit, a fast correction method for the transient safety domain is proposed.

[0080] The invention will be further illustrated below with specific examples:

[0081] This invention uses a New England 10-machine 39-node system (such as...) Figure 2 Taking the example shown, the specific embodiments of the present invention will be described in detail below. A method for fast correction of transient safety domain based on critical injection power includes the following steps:

[0082] (1) Fault settings

[0083] The fault is set as a three-phase ground fault at the beginning of line 21-22, with a fault clearing time of 0.14 seconds. After a system fault occurs, it is difficult to directly determine the system stability and instability mode in the early stages of the fault through measurement. Taking the power angle curve as an example... Figure 3 The diagram shows the changes in the power angle of each generator before and after the fault. Figure 3 As can be seen, in the early stage of the fault, the power angles of each generator were not set and the differences between them were not large, making it difficult to directly judge the transient stability of the system and the dominant unit of instability.

[0084] (2) Transient security domain boundary correction

[0085] The method proposed in this invention further mines transient information from the measurement data. Through instability mode identification, the instability mode is classified as type 3, namely G32, G33, G34, G35, G36, G37, and G38, indicating unstable generator units. Two-dimensional TSRs are constructed using generators G33 and G34, and the critical injection power for G33 and G34 is searched. The K values ​​are 1.44 × 10⁻⁴ s / MW and 9.84 × 10⁻⁵ s / MW, respectively. The final calculated critical injection power results are 620MW and 552MW, respectively. The boundary coefficient ratios of the TSRs of each generator obtained from offline simulation are applied online using equation (2-29). The hyperplane coefficients of the transient safety domain dominated by G33 and G34 after offline construction and online correction are shown in Tables 1 and 2, respectively.

[0086] Table 1. Coefficients of the G33-dominated hyperplane equation

[0087]

[0088] Table 2. Coefficients of the G34-dominated hyperplane equations

[0089]

[0090] Choosing G33 and G34 as coordinate axes, the corrected dimensionality-reduced TSR is as follows: Figure 4As shown.

[0091] As can be seen, the current injected power point is outside the TSR, so the system will experience transient instability. From Figure 4 It can be seen that the injected power of different generators collectively affects the transient stability of the system, and inaccurate regulation can occur when the stability control of each generator is performed individually. The method proposed in this invention allows operators to understand the overall impact of the injected power of each generator on the transient stability of the system, thereby providing support for more accurate and rapid stability control.

[0092] (3) Method Validation

[0093] To verify the correctness of the method of this invention, the results obtained by the method of this invention are compared and analyzed with those of the traditional fitting method. This invention uses the error parameter err. j This parameter represents the ratio of the distance from the critical injection point to the boundary hyperplane in the sample space to the magnitude of the injection vector, and is calculated using the following formula:

[0094]

[0095] In the formula, d j Indicates the critical injection power point x j Distance to the hyperplane, ||x j || represents the magnitude of the critical injection power modulus, α i For hyperplane coefficients.

[0096] Comparison of errors from different methods Figure 5 As shown:

[0097] Figure 5 The horizontal axis represents the 20 selected power critical points, and the vertical axis represents the percentage error calculated according to equation (5). It can be seen that at all power critical points, both the method of this invention and the fitting method are within the 5% error range allowed by engineering calculations. However, the method of this invention can use measurement data to correct TSR in real time, which significantly improves the calculation efficiency and can be applied online.

[0098] In this invention, the PC configuration is: Intel Core i7-4710 CPU / 8.00GB RAM. This invention utilizes machine learning methods to identify instability modes. After model training, only the characteristics of the generator after the fault need to be used as input data. The identification time is independent of system size and consumes almost no time. During the safety domain boundary correction process, the method of this invention does not require analytical calculation of the critical injected power, thus significantly shortening the computation time to only 3.2ms, fully meeting the requirements for online real-time operation. Compared to traditional methods, which require approximately 12 seconds to rebuild the TSR, the method proposed in this invention effectively promotes the online application of TSR.

[0099] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0100] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0101] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0102] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

Claims

1. A fast correction method for the transient safety domain based on critical injection power, characterized in that: Includes the following steps: Step 1: Based on the real-time measurement data collected after the fault, predict the critical injection power of the main unstable unit; Step 2: Using the proportional relationship of the hyperplane coefficients of each generator in the offline simulation, and combined with the predicted value of the critical injected power of the dominant unstable unit obtained in Step 1, the transient safety domain is quickly corrected. The specific steps of step 1 include: (1) Based on the real-time measurement data collected after the fault, predict the critical resection time; (2) Under the same terminal voltage level, the critical cut-off time and the power output of the dominant unstable generator vary according to a sensitivity coefficient, which is defined as K, representing the change in critical cut-off time under unit power change. This sensitivity coefficient can be obtained through offline simulation, as shown in equation (2): In the formula, Δf represents the change in critical resection time, and ΔP represents the change in power. (3) The critical injection power can be calculated using equation (3): P c =P0+K(T cr -y CCT ) (3) In equation (3), P c It is the critical injection power, P0 is the power at the current operating point, and T cr This is the fault clearing time; The specific method for step 2 is as follows: After obtaining the critical injection power of the dominant unstable generator in step 1, the new TSR hyperplane coefficient is obtained according to equation (4), and then the TSR is corrected: In the above formula, n-1 is the dimension of the injection space excluding the balancing machine, and P i Let α be the injected power at node i. i Let be the hyperplane coefficient of node i; from the above equation, we can see that after obtaining the new power critical point P, α can be obtained. i ', and then use this formula to correct the TSR boundary.

2. The method for fast correction of transient safety domain based on critical injection power according to claim 1, characterized in that: The specific method for step (1) of step 1 is as follows: The learning process for the critical resection time can be viewed as a mapping function. Assuming this mapping function is f, the predicted value of the critical resection time can be obtained through equation (1): y CCT =f(x) (1) In the formula, x is the quantity measured by the input of the critical resection time prediction network, and y CCT This is a predicted value for the critical resection time.

Images