Interval Prediction Method Based on Neural Networks for Short-Term Load Forecasting

摘要

Short-term load forecasting (STLF) is one of the most important issues in power system operation. With the penetration of renewable energies, the uncertainty levels of power systems have increased. Uncertainty often affects the accuracy of load forecasting models. Besides, there is no index available indicating reliability of predicted values in point forecasting. Prediction intervals (PIs) can provide more information and quantify the level of uncertainty. By combining neural network (NN) models with scalar method, a new method called PSO-based scalar method is proposed to construct PIs for target values in this paper. A new evaluation index is adopted to translate the primary multi-objective problem into a constrained single-objective problem. Particle swarm optimization (PSO) is used to solve the problem.