Consensus Clustering for the Identification of Coherent Regions with Varied Generation Mix
Kiran Kumar Challa, Alok Kumar Bharati, Venkataramana Ajjarapu · 2026
Read Consensus Clustering for the Identification of Coherent Regions with Varied Generation Mix by Kiran Kumar Challa, Alok Kumar Bharati, Venkataramana Ajjarapu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-16222.txt.
Introduction
Challa, Bharati, and Ajjarapu present a multi-view consensus method for finding coherent regions in power grids with varied generation mixes. As inverter-based sources grow, grid frequency response gets less predictable, and coherent region borders shift with disturbances and operating conditions. Most older methods assume one large generator outage. Their method handles a wide range of disturbances, shown on the miniWECC 240-bus test system.
Catalog record
| CREATOR | Kiran Kumar Challa, Alok Kumar Bharati, Venkataramana Ajjarapu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 23109 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.16222
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Consensus Clustering for the Identification of Coherent Regions with Varied Generation Mix Kiran Kumar Challa, Alok Kumar Bharati, Venkataramana Ajjarapu Abstract With a steady increase in the inverter technology integration to the grid, frequency response of the large interconnection system becomes more unpredictable. This leads to a significant change in the boundaries of the coherent region, which highly depends on the changing disturbance locations and operating conditions. While most of the existing coherency identification is based on a single large generator outage, it is important to identify these boundaries in view of wide range of disturbances. With large amount of inverters in the system, there is increase in the dynamic interactions of the various grid components leading to a need for such boundary identification. This paper presents the multi-view consensus algorithm to identify coherency in the case of variable grid operating conditions and wide range of disturbances. The proposed approach is demonstrated by identifying the coherent regions in the miniWECC 240 bus test system. I Introduction Coherency identification is essential in modern grid operations to ensure that generation meets the demand, frequency remains stable, and tie-line flows are controlled, preventing major disturbances and enabling seamless grid operation [ 9 ] . With the increasing inverter based resource (IBR) integration, decreasing system inertia makes it critical to maintain inertial support in every coherent region. Coherency identification also has several applications such as dynamic model reduction [ 12 , 11 ] , control islanding [ 2 , 5 ] , and wide area control [ 3 , 4 ] . Several model-based and measurement-based approaches are presented in literature to identify coherent generators in a large interconnected system. Model-based approaches rely on the linear/nonlinear simulation of the power system and uses the time domain responses for the CI. The concepts of slow coherency [ 15 ] , electromechanical distance in the transient period [ 10 ] and relative coupling between the generators [ 6 ] are presented using the linearized power system model. These methods suffer accuracy due to the model approximations. These methods are associated with the inter-area oscillations which suffer inefficient when there are significant inter-area oscillations [ 14 ] . Measurement-based approaches employ the PMUs to identify coherency through Fourier transform [ 13 ] and dynamic model decomposition [ 1 ] of the frequency responses. These approaches are unreliable in the changing grid conditions due to their sensitivity to noises, spurious signal components and depends on the type of disturbance. The online measurement based methods do have limitations on the data-collection infrastructure, excessive computational burden, and bandwidth requirements that affect coherency grouping’s success. While all the existing coherency identification approaches consider the occurrence of a single large contingency during the process, the clustering results under different contingencies such as disturbance location and type of disturbance have significant impact on the coherency boundaries. With increasing penetration of IBRs, frequency deviation in large-interconnected power systems is unpredictable. Coherency between the bus frequencies also changes with generation mix in the system at the time of disturbance and network which makes the identification of coherent regions more difficult. We propose to identify the coherent regions for a set of contingencies covering the location of generators, size of generators, type of source for a given operating condition. We use multi-view consensus clustering that aims to integrate information from different frequency-related contingencies to arrive at a single, unified clustering solution. We propose to utilize the advantage of effective ensemble learning presented in [ 7 ] to achieve the unified clustering. Ensemble learning-based clustering is a technique where multiple clustering results are combined into a single, more reliable, and stable clustering solution. The goal is to find a consensus among different views that enhances the quality of the clustering results. This approach can be particularly effective when each contingency provides complementary information about the underlying coherency between the bus frequencies. II Variability in Coherency The penetration of varied generation resulted in the multiple small generations being interconnected to the grid instead of larger generators in the case of synchronous generator-based resources. These generators do spread across the large interconnected system and impact the coherency of the grid. Identification of coherency using the system dynamics during the largest generator outage becomes infeasible as the new interconnected generators are not larger in size and largest synchronous generators are retiring. Importance of the online coherency identification is emphasized in [ 8 ] . It is also identified in [ 8 ] that the coherency changes with each disturbance in the system. Though the online coherency identification helps in the control, off-line identification in view of multiple disturbances is critical in planning of the real power reserves. Frequency responses in the case of different generator outages is studied using miniWECC 240 system to demonstrate the critical aspects of coherency variation and results are presented in Fig. 1 and Fig. 2 . The miniWECC 240-bus system is a reduced-order power system model representing a simplified version of the Western Electricity Coordinating Council (WECC) grid. In this model, the total generation capacity is approximately 291 gigawatts (GW), with about 59 GW (roughly 20%) attributed to grid-following inverter technology, that includes variable output generation. Frequency response of this system is observed for two different generator outages to understand the variation in the coherency with the change in the disturbance. Generator outage disturbances are created with a single large generator in the system at bus 4131 of 7418 MW capacity and another large generator at bus 5032 of 3646 MW capacity. Fig. 1 and Fig. 2 present the frequency with a 7418 MW (at bus 4131) hydro generator and 3646MW (at bus 5032) coal generator unit. Fig. 1 : Frequency responses for a trip of generator at bus-4131 Fig. 2 : Frequency responses for a trip of generator at bus-5032 It is evident from these responses that the frequency nadir, ROCOF and time of frequency nadir are completely different. The trip of generator at 5032 causes a similar frequency nadir in spite of the amount of generator outage is almost half. Similarity index matrix for the first disturbance is compared with that of the second disturbance, and it is observed that the coherency between the bus frequencies is different in the case of the two disturbances. This also indicates that the frequency is nonlinear and unpredictable in a system with IBR penetrations which leads to the variations in the coherency. This similarity matrix for the wide range of generator outages will be utilized in the clustering algorithm to identify the coherent regions. We propose to use multi-view consensus clustering algorithm to identify the coherent regions in view of varying coherency for different generator outages. III Consensus Clustering for Coherent Region Identification Identifying frequency coherent regions that correspond to multiple generator outages at different locations of a large interconnected network required the clustering of the nodes in view of multiple contingencies. Spectral clustering is a powerful graph-based clustering technique that is widely used to extract the eigenvalues of a node similarity matrix to get the low dimensional information before applying a standard clustering method. We propose to use consensus learning method to the spectral clustering approach to identify the unified clusters in view of multiple