A simulation approach of a smart grid by cooperative bargaining is presented in this paper. Each participant of the smart grid determines its optimal schedule to meet its power and heating demand at minimal costs employing solar panels, fuel cells and batteries. This is done by solving a quadratic optimisation problem which takes the energy prices and the available devices into account. The energy prices are related to the demand and supply in the smart grid, so that a lower demand yields lower prices. The cooperative bargaining game is used to tune the participants' optimal solution to obtain a Nash equilibrium. The computed solutions of the participants are validated against the capacities and structure of the smart grid by solving a multi-commodity flow problem. The presented model features multiple types of energy, so that they may be substituted to meet the participants' demand. Furthermore, the participants may also act as supplier and not only as consumer, which allows decentralised generation of energy. The approach is validated in several experiments where effects like negative energy prices if generated energy exceeds the smart grid's total demand and peak-shaving with even small-capacity batteries are exhibited.

No abstract

The approach in this paper renders it possible to simulate large-scale smart grids by efficient parallel computations. This permits a detailed analysis of the consumption behaviours, efficiency and impact of green energies, and self-sustainability of a smart grid. The smart grid is modelled as a multi-agent system. Each agent represents a building which is optimally controlled. That is, an agents meets its prescribed energy demand by trading energy or applying devices, e.g. solar panels and fuel cells, minimising its costs. A cooperative bargaining game is devised in which the agents participate to obtain a global optimal solution. In this paper, this inherently serial bargaining game is parallelised. The parallelisation is necessary to be able to deal with the large amount of data and computations which need to be performed. In the experiments the validity of the presented approach is shown and as a proof of concept a large smart grid of over 40 million agents is simulated.