Generation scheduling with demand bids

The structure of a market for energy is a fundamental design problem in the move towards competition in the electricity supply industry. Many systems have moved away from the traditional methods of central optimisation towards auction mechanisms. However, there are problems associated with the incorporation of system and unit constraints, e.g., minimum up- and down-times and ramping limits, aspects that are dealt with more easily using central optimisation. This paper describes a centralised pool electricity market that optimises the system based on bids from the supply and the demand side. A simultaneous market for reserve that considers spinning reserve from generators and nonspinning reserve in the form of interruptible customer load is included. The problem is formulated as an augmented Lagrangian function and a recurrent neural network is used to solve it. The method is also applicable to a more general auction mechanism.