Table 7 Key SML techniques

Large-scale renewable energy grid-connection

(a) High dimensional correlation modeling

(1) A large amount of renewable energy bring dimension disaster to uncertain modeling.

Singular value decomposition && principal component analysis [157]

Uncertainty modeling of decision variables, which includes capacity or location.

(2) The high dimension reduces the accuracy of probability modeling.

Convolutional neural network[158]

(3) The correlation coefficient can only grasp the overall characteristics, and the correlation simulation is not accurate.

Extreme operation scenarios

(b) Small probability estimation

Small probability event is difficult to estimate accurately, but it affects the electric network reliability.

Response surface methodology [153]

Uncertainty constraint modeling, which includes voltage amplitude and static voltage index.

First-order reliability method [159]

Second-order reliability method [160]

Analytical method [161]

Central moment method [162]

Typical operation scenarios

(c) Classification

Typical scenario extraction is usually based on clustering algorithm, but the correctness of unsupervised learning is difficult to verify.

The nearest neighbor approach && nonnegative matrix factorization [96]

Uncertainty objective modeling, which includes network loss and return on investment.

Wasserstein distance metric [155]

k-means clustering algorithm [163]