If it is unsupervised clustering then you may try a few modifications of the FCM algorithm, and select the one with best performance. You could also try the evolving clustering method.

But, if your task is to carry out classification, then you may try neural network or support vector machine.

With best wishes,

For clustering, you may see the paper 'Clustering by fast search and find of density peaks' published in Science two year ago. I think this method has a good performance to address the clustering problem, but I am sorry that I do not understand electric customer load profiles.

I'm also on the search! K-Means is definitely very prevalent, maybe because of the relative good performance vs. complexity. I haven't myself found the ONE clustering algorithm for load profiles, and don't expect to. But I will be testing k-means and other unsupervised methods in my work, the SOM sounds interesting. I think it boils down to how you aggregate and feature extract from your load profiles.

Just my 2 cent.

Dear Sir

Following methods of Artificial Neural Networks are best suited for clustering/Classification

1- Adaptive resonance theory neural network, an unsupervised learning technique can be used for classification. It is categorized in two categories: 1-ART1 (for binary Input vectors) 2- ART2 (for continuous values). But ART1 is very popular.

2-Self Organizing Map(SOM) or Konhenen's SOM- Another unsupervised learning technique which widely used for clustering. Correlation and regression can also be measured which are helpful in analysis.

Regards

Dr K Kumar

India

----------------