Cost forecast model for power engineering based on based on back propagation neural networks

A novel power engineering cost forecast model was proposed by combining feature extraction and small sample learning. The initial data was preprocessed with principal component analysis to remove the correlation among the original indexes and get the potential independent indexes. The new indexes acted as the input set to build a new forecast model based on least squares support vector machines. The results of this model were compared with the forecast results getting from artificial neural network. By comparing the forecast results with different principal components number, the optimal number was determined to achieve the desired forecast effect. The prediction results indicate that the method can extract the feature of initial data effectively and is good at small sample learning . The expected forecasting results can be reached.