Transient Stability-Based Security State Classification of Power System Networks Using Kohonen’s Neural Network

Mukherjee, Rituparna; De, Abhinandan

doi:10.1007/978-981-15-4692-1_5

Cited by 1 publication

(1 citation statement)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To enhance the prediction accuracy of machine learning in dealing with complex and large feature space, the same authors (Mukherjee and De, 2021) proposed a methodology integrating a self-organizing map (SOM), which aims at both reducing the data cardinality and enhancing the classification accuracy. However, SOMs are unsupervised algorithms, which require the deployment of heuristic techniques for transforming the data clusters into classes.…”

Section: Introductionmentioning

confidence: 99%

The role of data-driven methods in power system security assessment from aggregated grid data

De Caro,

Giannuzzi,

Pisani

et al. 2024

Front. Smart Grids

View full text Add to dashboard Cite

Data-driven techniques have been considered as an enabling technology for reducing the computational burden of both static and dynamic power system security analysis. Anyway, the studies reported in the literature mainly focused on inferring from historical data the mapping between the bus variables before and after a certain contingencies set, while, to the best of the Author's knowledge, limited contributions have been devoted to try and classify the power system security state by processing aggregated grid data. This is a relevant issue to address for a Transmission System Operator since it could allow a sensible decrease in the computational burden and, considering that aggregated grid data can be reliably predicted from several hours to one day ahead, it may enable the evolution of security assessment to security forecasting. In trying and filling this research gap, this paper explores the role of machine learning and feature selection algorithms. A realistic case study involving 2 years of synthetic grid data simulated on the Italian power system model against future potential operational scenarios characterized by a high share of renewables is presented and discussed to identify the most promising computing paradigms, analyzing the criticality of tuning the feature selection and classifier algorithms.

show abstract

Section: Introductionmentioning

confidence: 99%