Abstract
Knowledge acquisition and extension of expert systems for electric power systems are becoming difficult, and this paper proposes a solution by symbolic machine learning. The authors have developed a learning method HCL (Hierarchical Concept based Learning), which enables us to acquire knowledge to make switching sequences of power system operation.
HCL builds up knowledge base of an expert system from operational experiences in the past. The learning process of HCL is divided into organization phase and generalization phase. In the organization phase, HCL analyzes each experience and structures it by recognizing several aims of switching operations, and individual knowledge is produced. In the generalization phase, HCL integrates similar experiences into common operational rules by translating their individual names to variables.
HCL features the use of some conceptual hierarchies for organization, for instance the hierarchy of domain model which expresses the structural relationships between devices or equipment, such as ‘a_part_of’ or ‘connected_to’ relations. Based on the conceptual hierarchy, HCL enrolls all the operations in a tree structure, of which the nodes express their several aims.
As application examples of HCL, knowledge acquisition of operational sequences for bus-bar switching is shown.
