Modeling competition of virtual power plants via deep learning

Traditionally, models pooling flexible demand and generation units into Virtual Power Plants have been solved via separated approaches, decomposing the problem parts dedicated to market clearing separate managing state-constraints. The reason for this is high computational complexity of solving dynamic, i.e. multi-stage, problems under competition. Such approaches downside not adequately modeling direct competition between these agents over entire considered time period. This paper approximates decisions players ‘actor networks’ assumptions on future realizations uncertainties as ‘critic networks’, approaching tractability issues multi-period optimization at same time. Mathematical proof solution converging a Nash equilibrium provided supported by case studies IEEE 30 118 bus systems. Utilizing approach, framework able cope with uncertainty spaces extending beyond traditional approximations such scenario trees. In addition, suggests various possibilities parallelization in order increase efficiency. Applying process allows parallel all periods training parallel, previously only succession.