Comparison of CART- and MLP-based power system transient stability preventive control

This paper compares the power system transient stability preventive controls based on classification and regression tree (CART) and multilayer perceptron (MLP). CART is a branch of the decision trees (DTs) algorithms, which have been applied extensively in the literature for preventive control design due to their transparent character of interpreting inference procedures. Recent research has shown that multilayer perceptron (MLP), a member of artificial neural networks (ANNs) family, can also be employed to design preventive control since explaining internal mapping relationship becomes possible. This paper first briefs the essentials of CART and MLP, and then follows the numerical study on a three-generator six-bus power system. Transient stability preventive controls are designed by CART and MLP, and are compared afterwards. The finding shows that preventive controls developed by both approaches think alike. They are even complementary.

► Decision trees algorithms have been commonly mentioned for designing power systems preventive controls.
► Recent research shows that multilayer perceptron can also be used to design preventive control.
► This paper compares the power systems transient stability preventive controls based on these two approaches.
► The finding shows that preventive controls based on these two approaches see eye to eye to some extent.
► They can even help each other and compensate their weaknesses.