With the advent of smart grids, a growing need to maintain real-time balance between electric energy production and consumption emerges due to high cost of electric energy storage. Successful trading on the wholesale market is an important prerequisite for achieving that balance. In this paper we describe the wholesale market bidding strategies implemented in the intelligent software agent CrocodileAgent 2012 that acts as a broker in the smart grid power market simulator Power Trading Agent Competition (Power TAC) [1].

The Power TAC Wholesale Market is modelled as a periodic double auction which clears every simulated timeslot. In the Power TAC Wholesale Market brokers buy and sell electric energy at the market clearing price (i.e., equilibrium price). Brokers can simultaneously send orders (i.e., bids or asks) to 24 different markets, each representing one of the 24 upcoming simulated timeslots (i.e., day-ahead market). Broker's main goal is to ensure the sufficient amount of energy for customers in its portfolio by buying energy on the Power TAC Wholesale Market from producers and other brokers. In case of energy imbalance the broker must pay a balancing fee.

Modules of the CrocodileAgent 2012 are shown in Figure 1. The Forecasting Module uses historic data (e.g., past energy production/consumption, previous clearing prices) to predict the amount of energy customers will need in the next 24 timeslots and to estimate clearing prices on the Power TAC Wholesale Market. For calculating predictions we use the triple exponential smoothing technique better known as the Holt-Winters Method [2]. The predicted values are defined as:

(1) where [0.0, 1.0] are parameters used to calculate the triple exponential smoothing of past values consisted in the parameter (i.e., past clearing prices), the parameter is the periodicity of those values (e.g., daily periodicityperiodicity of 24 timeslots) while is the number of values to be predicted. The CrocodileAgent 2012 retrieves the information about the amount of energy it needs to buy or sell for each of the 24 upcoming timeslots and computes the price offered on the Power TAC Wholesale Market depending on the number of tries left (e.g., if the current timeslot is and the broker wants to buy the energy needed for the timeslot , it can do so by biding in the current and 9 upcoming timeslots). As the respective timeslot approaches, the CrocodileAgent 2012 increases the price it is willing to pay in order to buy the needed amount of energy. If there is only one try left (i.e., the current timeslot is ), the broker offers a predefined maximum price. More detailed description of the CrocodileAgent 2012 can be found in [3].