Climate change has stimulated the energy sector, which is the largest emitter of global greenhouse gases (~40% in 2019), to transition to low-carbon energies. Europe, being one of the highest historical emitters of greenhouse gases, sits at the forefront of the energy transition. With a growing share of variable renewable power systems in the electricity mix on the one hand, and changing frequency and intensity of extreme events on the other, the weather-sensitive European energy sector is continuously on the lookout for accurate forecasts of essential climate variables on a continuum of timescales. The weather forecasts on short- to medium-range (i.e., from a few minutes ahead to at most two weeks) are reliable and essentially deterministic, and hence their operational use within the energy sector is well established. However, on timescales beyond two weeks and up to two months, i.e. in the sub-seasonal range, the predictions are necessarily probabilistic, and their reliability is far from that offered by short- and medium-range forecasts. Consequently, the operational use of sub-seasonal predictions within the energy sector is still in its infancy.Having accurate information about the expected renewable energy production and electricity consumption on sub-seasonal timescales can help the energy sector in determining required reserve levels, scheduling maintenance, assessing and allocating risks attributed to extreme events, and estimating grid transmission capacity. In this regard, the main objective of this thesis is to provide more reliable information on sub-seasonal timescales, relative to climatology, to aid the energy sector in operational decision-making. We focus this research on 100-m wind speed and 2-m temperature over Europe.As an essential first step, we rigorously assess the skill of sub-seasonal dynamical predictions of these two variables to quantify their predictability limits as they are delivered in a given forecasting system (the extended-range predictions of the European Centre for Medium-Range Weather Forecasts). We show that the weekly mean predictions of gridded temperature are more reliable than climatology for up to six weeks, and those of wind speed for up to three weeks. As a second step, we develop a statistical downscaling technique to reconstruct sub-seasonal predictions of wind speed and temperature using predictions of large-scale atmospheric circulation. We summarize the large-scale atmospheric state in a few indices by employing a dimension reduction methodology conditioned on wind speed and temperature over Europe. In other words, we use historical, observationally derived data to capture the relationship between the large-scale atmospheric circulation and our variables of interest (100-m wind speed and 2-m temperature). We then employ this relationship on sub-seasonal predictions of large-scale circulation, which are more reliable than surface variables, to deduce information about our variables of interest. This method allows us to produce, from a given ensemble of sub-seasonal predictions of large-scale circulation, a new ensemble of sub-seasonal predictions of our variables of interest. We demonstrate that the information thus extracted has value, as the hybrid ensemble combining both the dynamical and the statistical predictions of our variables of interest are more reliable than the dynamical predictions alone. As a final study, we investigate episodes of wind drought over Europe, because of their importance to the energy sector. A case study of the July 2018 episode of weak winds and the associated predictions, with and without our statistical downscaling methodology, illustrates the persistent difficulties of sub-seasonal predictions in predicting extreme events, in this case, due to the long-lasting challenge of forecasting blocking events.