Hybrid methodology for short-term load forecasting

The main objective of this paper is to accurately forecast the short-term loads using Discrete Wavelet Transform (DWT) in combination with Artificial Neural Network/ Support Vector Machine. The complete analysis has been carried out using Temperature, Humidity, Dew Point and Actual loads as features. Here, 8-level DWT decomposition has been done to extract the 8 detailed and approximation coefficients, which are also used as features. Thereafter to enhance the accuracy, four optimal features are selected from the total feature set using Forward Feature Selection Algorithm during the training process during ANN/ SVM. The test data with the optimal features were then fed to the ANN or SVM for load forecasting. Here MAPE has been considered as the performance index. The test results demonstrate that the proposed technique is quite accurate to forecast the loads.