The pattern of "technological change - energy efficiency" is essential to identify the possible future evolution and conceive sound policy measures to reorient the energy - economic system. Being technology a fundamental driving factor of the evolution of energy economic system, it is essential to study the basic mechanisms of technological change what has an important role in de-coupling energy demand from economic growth. Technological learning is the main way to induce technological change. Recent literature has emphasized that two factors - learning by R&D (research and development) and learning by doing should play a major role when modeling technical change in energy - economy models. And learning by R&D and learning by doing are the main drivers of energy saving technical change that eventually affect energy intensity. In this paper, the model that we present is a productive function based on Cobb-Douglas production function in which both learning by R&D and learning by doing are explicitly accounted for through an index of endogenous technological change and technological cost reduction. Moreover, exogenous technological improvement induced by autonomous technological change which involves an autonomous exponential improvement. In our model, the different components of technical change have a differentiated impact on energy efficiency effects. Firstly, for learning by R&D, R&D expenditures induce the developments of energy-saving technologies, so that energy elasticity will be reduced, which decouples energy demand from economic growth. Secondly, for learning by doing, energy-saving technological cost is reduced with the experience growing (we use the cumulative capital as the driver). Thirdly, autonomous technological change promotes exogenous technological improvement which induces improvements in labor productivity growth, capital productivity, and energy use productivity. Thus, our new formulation hinges on the relationship between technology change and all of the three factors as learning by R&D, learning by doing and autonomous change at the same time. In addition, we investigate the effects of technological learning on energy efficiency and highlight the dynamic interrelationships between the different variables and their role in the model.