Application of artificial neural networks to stochastic electric power production production costing and operations planning

摘要

The paper presents a new method for stochastic electric power systme prodcution costing and operations planning. In the method we formulate a stochastic Hofield/Chua-Kennedy neural network in which unit availability and systme load demand are random parameters with known statistics. Unit outages are modelled as Markov processes. The unit commitment-status variables are (0, 1) integers while the unit dispatch/loading levels take on decimal values. The unit commitments-tatus variables together with the unit dispatch/loading levels are random processes satisfying appropriatel derived deterministic equivalent differential equations. A review of the techniques that are currently employed by utilities for production costing and operations planning with particular emphasis on those for the stochastic problem is also presented. Among the shortcomings of these techniques are their inability to commit and dispatch units simultaneously, and to account for the effect of forced outages on unit availability in a chronological manner. Our method addresses these short-comings.