Automatic Learning Approaches for Electric Power Systems

This chapter describes a methodology based on the combination of probabilistic reasoning, automatic learning, and Monte Carlo simulations, which has been used extensively for the study of electric power systems. Electric power systems are essentially large, complex nonlinear systems that have grown significantly in size and importance during the last 50 years. The very large size of electric power systems makes understanding them and the mastering of their reliability a quite complex problem. An interconnected system is operated by a number of independent companies that have to make decisions without knowing precisely the strategy of their neighbors. Thus, electric power systems are also a good example of distributed decision making under uncertainties. The methodology described in this chapter is a “computer experiment” type of method. This chapter describes the generic approach together with the principles of the main classes of automatic learning methods while also discussing a few real-life applications and some new research directions. This chapter concludes with a discussion of the usefulness of the proposed approach and its applicability to the study of complex systems in general.