The recent changes to power systems have opened up a frontier of challenges and potential benefits for all the participants. The smart grid paradigm helps to support better decisions, to improve efficiency, and to provide a more reliable, economic, and sustainable energy supply. This is particularly important given the growing interest in user participation via demand-response programs. The research presented in this thesis contributes to the development and penetration of demand-response programs. These approaches have been developed from the user perspective, taking into account elements of the power system such as peak reduction and cost. This research combines elements from the real world and novel ideas to balance conflicting goals and move forward in the evolution of the grid. One of the main contributions is the use and estimation of capacity profiles. The idea behind setting a capacity profile is to establish a compromise between the expected demand and the level of service perceived by the user. The capacity profiles are determined in advance and account for the users' future energy requirements in a demand-response context. In the first contribution (Chapter 4) the methodology focuses on heating and cooling devices. The capacity profile works in combination with an existing admission controller to guide the control of the building temperature. In this case the capacity profile is estimated by a data-fitting approach and a multiclass classifier. In the second contribution (Chapter 5) the framework is based on specific features of a stochastic demand. This demand is generated by the aggregation of activity-based loads that are triggered by the user behavior. Here, the capacity profiles are determined as a function of the available information. First, we use a heuristic approach when we have limited information about the consumption patterns. Second, we use two-stage optimization when we have the demand scenarios. One highlight of this contribution is the inclusion of a flexible time-and-level-of-use pricing. This pricing policy allows the user to determine the capacity profile and its corresponding tariffs from a set of options provided by the grid. These customized tariffs are aligned with the normal user behavior and eventually guide a learning process to optimize the consumption. The third contribution (Chapter 6) aggregates several users, different demand response programs, and various shared resources to plan the consumption, load shifting, and peak reduction. This approach takes information such as the demand profiles and user preferences directly from the housing units. It solves a biobjective optimization problem to find a trade-off between the total user satisfaction and the total cost of the energy consumption. This contribution includes a fixed cost structure similar to that of the second contribution to encourage load shifting. Finally, experiments are reported in each contribution to validate the performance of the proposed approaches. We provide sensitivity analysis, benchmark comparisons, and a discussion of real-world features to help clarify the strengths, limitations, and possibilities for future research.