TABLE OF CONTENTS Artificial intelligence and machine learning techniques provide a qualitative as well as a quantitative assessment of the power system security state. As discussed above, the quantitative results for single power system operating points have to be used to understand the global behaviour of the system state qualitatively. Traditionally this is the task of the human operator or planning engineer.
Inspired by the successful application of artificial intelligence techniques in other technical domains like component diagnosis and VLSI synthesis [4], several research groups attempted to implement the reasoning of the power system operator with the help of expert systems, heuristic search techniques or fuzzylogic: see the special issue of the IEEE Proceedings edited by Liu [5] for a review.
A second class of tools are machine learning techniques which develop a global description of the power system state and operating space using statistical data analysis. Pattern recognition and artificial neural networks (ANNs) belong to this category.
Although initially very different in means, both techniques are capable of dealing to some extent with unknown, that is not-analysed, system states. Using symbolic reasoning techniques, specific engineering 'rules' have to be implemented in a sufficiently abstract manner to generalise acquired knowledge while the machine learning techniques achieve generalisation by interpolating analysed data. Note that some successful implementations of an expert system use statistical case analysis as an essential tool in order to collect case-independent quantitative knowledge of the power system.
Both groups of techniques...
