Short Term Price Forecasting Using Adaptive Generalized Neuron Model

Deregulation in the electricity industry has made price forecasting the basis for maximizing profit of the different market players in the competitive market. The profit of market player depends on the bidding strategy and the successful bidding strategy requires accurate price forecasting of electricity price. The existing methods of price forecasting can be broadly classified into (i) statistical methods (ii) simulation-based methods and (iii) soft computing methods. The conventional neural networks were used for price forecasting due to their ability to find an accurate relation between the historical data and the forecasted price without any system knowledge. They suffer from major drawbacks like training time dependency on complexity of the system, huge data requirement, ANN structure is not fixed, hidden neurons requirement is large relatively, local minima. In the proposed work, the problems associated with conventional ANN trained using back-propagation are solved using improved generalized neuron model. The genetic algorithm along with fuzzy tuning is used for training the free parameters of the proposed forecasting model.