In the paper we analyze the evolution of methods and models for long-term forecasting of energy development. A conflict between two trends is shown: the complication of the applied methodological tools and objective acceleration of an increase in the uncertainty. The second trend diminishes the practical value of «accurate» forecasts but raises the importance of identifying the field of forecasting (the cone of uncertainty) and comprehensively assessing the options this region consists of, including risk analysis. The paper demonstrates the advisability of using such an analysis of stochastic models.

Key words: forecasting, energy, models, uncertainty, optimization, stochastic, investment risks.