Demand Response and Energy Portfolio Optimization for Smart Grid using Machine Learning and Cooperative Game Theory

Adriana Chis

Research output: ThesisDoctoral ThesisCollection of Articles


Widespread availability of electricity is a hallmark of civilization. A reliable electricity supply is fundamental for the social and technological development of the world. To cope with the growing electricity demand and other challenges associated with energy delivery today, technological advancements towards a modern updated power grid are needed. The development of a smart grid is a solution to enable a more stable, reliable, efficient, economical and sustainable energy generation, transmission, distribution and usage. One drawback of the traditional power grid is the mismatch between energy supply and demand. The solution to this problem is the deployment of a more flexible energy generation system, together with a balanced electricity consumption. This could be achieved by means of demand side management (DSM). The focus of this thesis is to model efficient DSM methods for optimizing electricity consumption. In particular, price-based demand response (DR) methods that require the active participation of electricity users are developed. Price-based DR methods allow for energy users to optimize their energy consumption and reduce their costs. This occurs if they adjust and change their electricity consumption patterns in response to dynamic prices applied by utility companies. One problem tackled in this thesis is that of optimizing the charging of electric vehicles (EVs). More and more people are interested in purchasing EVs. The EVs however, will significantly increase their electricity consumption and cost. Using machine learning techniques, efficient methods that optimize the home charging of an EV and reduce the long term cost of charging for the owner are developed. The EV charging is scheduled by taking advantage of the time-varying electricity prices within a day, but also of the dynamic nature of prices on different days. In the traditional power grid, the role of the energy consumers was that of price takers with no other involvement in the energy sector. The smart grid however, will support consumers also in owning renewable energy sources (RESs) and energy storing systems (ESSs). Local energy generation and ownership of ESSs opens opportunities for new energy strategies and markets. By enabling cooperation among energy producers and consumers,  they would be able to manage and use their renewable energy resources and storage spaces more efficiently and reduce their electricity consumption costs even more. In this thesis, collaborative models for exchange and trade of energy within communities of households owning RESs and ESSs are developed. Using a mathematical model from cooperative game theory, the community energy portfolio optimization problem is formulated as a coalitional game for the households to minimize their costs, individually and collectively. Moreover, using a concept from microeconomics, a DSM method is also developed from the perspective of the utility company to balance the community's grid energy consumption.
Translated title of the contributionDemand Response and Energy Portfolio Optimization for Smart Grid using Machine Learning and Cooperative Game Theory
Original languageEnglish
QualificationDoctor's degree
Awarding Institution
  • Aalto University
  • Koivunen, Visa, Supervising Professor
Print ISBNs978-952-60-8121-2
Electronic ISBNs978-952-60-8122-9
Publication statusPublished - 2018
MoE publication typeG5 Doctoral dissertation (article)


  • smart grids
  • demand side management
  • demand response
  • machine learning
  • game theory
  • electric vehicles
  • smart community


Dive into the research topics of 'Demand Response and Energy Portfolio Optimization for Smart Grid using Machine Learning and Cooperative Game Theory'. Together they form a unique fingerprint.

Cite this