Avrupa gün öncesi piyasalarında bulunan piyasa takas problemleri

Electricity is a commodity that must be constantly kept in balance. Electricity markets have therefore evolved accordingly. The day-ahead electricity market (DAM) is a spot market where the electricity is traded one day before the actual delivery date. Each day, there is a two sided closed auction in a DAM and the corresponding market clearing problem needs to be solved in a limited amount of time due to the strict regulations. In this thesis, we study the European DAM clearing problem. We consider hourly, block, linked block, and flexible orders in our model. We first start with a stylized market clearing problem, which is a total surplus maximization problem with hourly orders that can be represented by piecewise linear functions. The total surplus is the sum of the producer and consumer surpluses. In the stylized problem, consumer and producer surpluses are determined by using the demand and the supply curves of the hourly orders. We propose an optimization model that instead employs total curves represented by the difference between the demand and supply curves. We show that the set of optimal solutions of the proposed model is precisely the same as that of the stylized problem. We next incorporate block, linked block and flexible orders and present an optimization model, denoted by (MC), based on the total curves. The model (MC) is a mixed integer quadratic programming problem. By fixing the binary variables of the model (MC) to their optimal values, we investigate the optimality conditions of the resulting problem. We identify two sets of optimality conditions. We show that these two sets of conditions may not simultaneously be satisfied. By imposing each of these two sets of conditions in the model (MC), we obtain two different variants, one of which is already used in the European power exchanges and the other one is a new contribution to the literature. For the two problem variants, we propose several solution methodologies. Our computational experiments illustrate that the proposed solution methods are capable of computing optimal solutions on real and synthetic data sets in a reasonably small amount of time in most of the cases.