In this paper, we use neural network and Markov theory
to set up a model for power load forecasting, and the
application of the forecasting model can take full advantage of
the time series of load information, while avoiding
unnecessary interference factors with a strong scientific. This
method is not only high accuracy but also simple, practical
and workable. In addition, the paper first attempt to use
Golden Section method to divide the state interval, and it is
proved the method is feasible.