Valuing the electricity produced locally in renewable energy communities through noncooperative resources scheduling games

This paper investigates local energy communities composed of end-users connected to the public electricity distribution network and sharing common resources such as the grid and their local generation. Two market designs are proposed for the optimal day-ahead scheduling of energy exchanges within these communities. The first implements a collaborative demand-side management scheme where members’ objectives are coupled through grid tariffs; the second allows for the valuation of excess generation in the community and on the retail market. Two grid tariff structures are tested, one academic and one reflecting current Belgian regulations. Individuals’ bills are obtained through four cost allocation methods. Both designs are formulated as optimization problems and noncooperative games. In the latter case, the existence and efficiency of the corresponding (generalized) Nash equilibria are studied and solution algorithms are used. The models are tested on a use-case with 55 members and compared to a benchmark where members act individually. Global Renewable Energy Community and individual costs are computed, inefficiencies of the decentralized models are evaluated against social optima, and technical indices are assessed. Our main contributions are threefold. First, the existence of an equilibrium that is a social optimum is established. Secondly, it is shown, analytically when possible, and empirically otherwise, that community bills obtained with the centralized and decentralized approaches are equivalent (or exhibit negligible differences) for the studied cost allocation methods. Similarly, individuals’ bills obtained ex-post from the faster centralized formulation closely match those (maximum negligible deviation of 1.8 %) from decentralized models in most studied configurations, except for the continuous billing with the academic grid tariff. In the latter case, 65 % of members have deviations of less than 5.6 %, with maximum deviations of 31 %.