An expert system for specifying the major substation components: machine learning approach

Abstract

A learnable expert system for substations (LESS) has been developed that will learn design rules of the substations from examples. The resulting rules and concepts reflect the philosophy of design embedded in the provided examples. Thus, the substation design will be consistent with the design of the other substations owned and operated by the company if the examples consist of descriptions of their existing substations. LESS is combined with two learning modules to acquire the design knowledge required for specifying the major components and the layout of an electric substation. LESS has the capability of accepting rules dictated by laws, codes or standards, as well as formulating rules based on design practice which are learned from positive examples. It can also avoid certain problematic situations in past designs by observing negative examples. LESS has graphical capability and will produce a plan view of the substation.<>