Coherency identification using growing self organizing feature maps [power system stability]

Coherency Identification in Interconnected Power System—An Independent Component Analysis Approach

“…The methodologies underlying these approaches are diverse, e.g. artificial neural network (ANN) [11], particle swarm optimization and k-means (PSO-KM) algorithm [12], graph theory [13], partitioning around medoids (PAM) [14], hierarchical clustering [15], self-organizing feature maps [16], Fast-Fourier transform (FFT) [17], [18], Hilbert-Huang transform (HHT) [19] and Principal Component Analysis (PCA) [20], [21]. Despite having the advantages over model-based approaches, these techniques have some limitations, for instance, the ANN algorithm requires excessive off-line training data to train the neural network in order to identify the coherent generators.…”

Taxonomy of coherency detection and coherency‐based methods for generators grouping and power system partitioning

“…In the first step, the degree of coherency between generators is calculated and, then, in the second step, algorithms such as clustering methods are used to group the generators. Examples of such clustering methods are fuzzy c‐means (FCM) [27, 75, 79, 83, 84], KM [44, 46, 111, 116], k‐harmonic means [117], subtractive clustering [47], partitioning around medoids [78], support vector clustering [73, 74], HC [80, 103], artificial neural network (ANN) [72], growing self‐organising feature maps [104, 115] and multi‐flock‐based approach [76, 77]. As an example, the KM algorithm uses an iterative procedure to place the generators in a predefined number of groups as follows: 1: Initialise c i cluster centres randomly. 2: Determine the membership of each datapoint (bus or generator) with respect to each cluster centre using the equation below: 3: Calculate the cost function defined in (21).…”