Complex Power System Status Monitoring and Evaluation Using Big Data Platform and Machine Learning Algorithms: A Review and a Case Study

Table 5

Comparisons of the non-Gaussian data methods.

Method	Data assumption	Parameters	Disadvantages
ICA	Can be described as a linear combination of non-Gaussian variables	Number of ICs	(1) High computational cost; (2) Hard to determine the control limit
GMM	Can be described by local linear models	Multiple parameters in the model	(1) Complicated to train the models; (2) Hard to determine the number of local models
SVDD	No strict assumption of data distribution	Kernel parameters in the model	(1) Hard to tune the kernel parameters; (2) Trade-off between accurate boundary and low false alarm control limit