Capacity price prediction in the German Reserve Control market with AI

The German electricity Transmission System Operators are sourcing their demand of Secondary Control Reserve (SCR) in a single-round pay-as-bid auction. Rational bidders are regarding both their costs for the SCR provision and an estimate of the maximum achievable auction price in order to maximise their revenues. Such price estimation is a complex game theory problem and cannot be done analytically. In this work, a Multilayer Feedforward Network is used for capacity price estimation in the German SCR market. A high forecast quality is achieved by using a Genetic Algorithm for topology optimization of these networks. The forecast quality of the developed approach is compared with an Ordinary Least Squares linear regression model.