| With the deregulation of power industry allover the world, establishing electricity markets to optimize distribution of resources is the trend of power industry. Electricity price issues are the key problems in the markets and how to price the special commodity-electricity is essential for the smooth market operation. So using the relative historic data to predict the future electricity price is a very meaningful work for every participator in the power market.After introducing the constituents, the characteristics and the influence factors of the electricity price in power system, this paper discusses the merits and shortcomings of some current forecasting methods. Based on some basic theoretical aspects of Neural Networks, its advantages in the electricity price forecasting are analyzed, the methods of neural network-based short-term electricity price forecasting are described. To improve the accuracy of price forecasting, three aspects of ameliorations are proposed: 1. selecting the factors which can show the change of the price sequence accurately; 2. the pretreatment of the input variables; 3. ameliorations to the configuration of neural network.A price forecasting model using RBF neural network is proposed. Moreover, similarity searching technique is applied to build the training squadron and input variables. The electricity price of next day is forecasted successfully by this model. The result of forecasting shows that RBF neural network is more steady and accurate than BP neural network. The precision of forecasting by similarity searching technique is higher than traditional methods.Finally, this paper points out that in order to improve the accuracy, based on the given power market the constituents and characteristics of the electricity price must be analyzed first. To deal the choice of the input variables well and to design the model matching the characteristics of the electricity price are also important. The data used in this paper come from Australia Power Market. |
PDF Full Text Request