Distributed stochastic economic dispatch for smart grids: a model predictive control approach

• Autores: Miguel Andrés Velásquez Motta
• Directores de la Tesis: Mohammad Shahidehpour (dir. tes.), Angela I. Cadena (dir. tes.), Nicanor Quijano (dir. tes.)
• Lectura: En la Universidad de los Andes (Colombia) ( Colombia ) en 2018
• Idioma: español
• Enlaces
  • Tesis en acceso abierto en: hdl.handle.net
• Resumen

  • Power systems have experienced several changes since smart grids and renewable resources increased their penetration. Traditionally, power systems operation has been addressed with unit commitment and economic dispatch problems that rely on a centralized operator. These operation methods are usually performed on a day-ahead basis, i.e., every 24 hours. As a result of volatility in renewable resources and demand, it is better to shorten the operation period, e.g., every hour. Centralized methods might not be feasible for solving short-term economic dispatch, especially in systems with several agents. Thereby, the research questions this thesis are what method can be used for solving short-term economic dispatch in the presence of smart grid elements? Second, what models can be designed in order to optimally dispatch power plants and operate different agents in a smart grid environment? Third, how uncertainty can be considered in such models without increasing dimensionality and keeping tractability? Fourth, what is the best way to operate power systems with smart grid elements? In order to solve all these questions, we deeply analyzed economic dispatch methods and smart grid elements. Next, we proposed two distributed economic dispatch methods that are feasible for hourly and ultra-short term periods. In addition, we integrated stochastic programming through a data-driven scenario generation in order to include randomness of power system variables. Finally, a hierarchical operation of hourly and ultra-short term was proposed to enhance operation performance. The results obtained in this thesis show that proposed methods answer our research questions and serve as a basis for operating power systems more efficiently. Under uncertainty framework, it is better to use stochastic approaches rather than deterministic ones. For using stochastic approaches, it is necessary to pass from centralized controllers to distributed architectures as it has been proposed in this work