Next Day Load Forecasting Using Artificial Neural Network Models With Autoregression and Weighted Frequency Bin Blocks

Abstract

In this study, two different hybrid approaches based on Artificial Neural Network (ANN) models with Autoregressive (AR) method and Weighted Frequency Bin Blocks (WFBB), are used for next day load forecasting. To compare with the hybrid approaches and conventional models, the next day load forecasting is also performed by using AR and ANN models, separately. In the first hybrid approach, ANN model with AR method, the results of the AR method applied to all data taken from Turkish Electric Power Company and Electricity Generation Company, is used as an only additional input for ANN model. In this approach, the ANN structure has two layers composed of 49 and 24 neurons for input and output layers, respectively. In the second hybrid approach, ANN model with WFBB, the results obtained from WFBB are used for all inputs in the ANN. model. In this approach, input and output layers in the ANN structure are composed of 48 and 24 neurons, respectively. Feed Forward Back Propagation (FFBP) is chosen for all neural network models in this study. The forecasting results obtained from AR, ANN and the two hybrid models are compared to each other in the sense of root mean square error (RMSE). It is observed that the RMSE values for the hybrid approaches are smaller titan the conventional models. Then, the hybrid models forecast better than the conventional models.