TEMPUS: a machine learning tool

摘要

The use Knowledge-Based Systems (KBS) in Power System Control Centers has been introduced to assist operators in network operation [CIGRE-93]. The success of KBS requires them to be kept efficient over time, even when modifications in the electrical networks and in its operating methods are undertaken. Real situations handled by the KBS can be used to enhance its knowledge making them more useful and efficient. This can be achieved trough the use of adequate machine learning techniques. This paper deals with knowledge maintenance and machine learning issues presenting the achievements of SPARSE [Vale-93,Vale-97b, Vale-99a], a KBS used in Portuguese Transmission Control Centers for intelligent alarm processing. A Knowledge Automatic Extraction tool was developed to confirm or suggest modifications in the Knowledge Base of SPARSE. A Rule Editor module was also developed which allows to change the Rule Base of SPARSE. The Machine Learning tool is named TEMPUS, while the Rule Editor is named EDIREGRA. TEMPUS has been applied, with success, to the expert system SPARSE. The results obtained so far show the need and utility of this tool. On the other hand, EDIREGRA makes the process of changing time parameters associated to some premises of the SPARSE rules as well as the creation of new rules much easier.