Short-Term Load Forecasting Enhanced With Statistical Data-Filtering Method

Abstract

Short-term load forecasting (STLF) plays an increasingly important role in power-flow dispatch, design and planning of power systems. To enhance the accuracy of STLF, in recent years, time-series approaches and other intelligent methods are strongly exploited in load forecasting models of distribution networks. In developing countries, the load demand can be suddenly changed by power consumption from different customers, e.g. industrial customers, residential customers, so the load dataset is often unstable. As a result, reliability assessment of the load dataset is very necessary for the pre-processing of data-filtering methods. This paper introduces a novel data-filtering method that considers the reliability of dataset by analyzing a wide range of pre-defined confidence levels. This proposed data-filtering method is also compared to other existing data-filtering methods (e.g. Kalman filter, Density-Based Spatial Clustering of Applications with Noise (DBSCAN), Wavelet transform and Singular Spectrum Analysis (SSA)). Case studies of short-term load forecasting using the real data from the Supervisory Control and Data Acquisition (SCADA) system of a typical distribution network in Vietnam are conducted with an Artificial Neural Network (ANN) model and a conventional (ARIMA) statistical model in order to validate the novel statistical data-filtering method proposed. The achieved results demonstrate which according to the most confidence level of the load dataset, the STLF using ANN and ARIMA models with the proposed statistical data-filtering method can outperform that with the existing data-filtering methods. For example, regarding the typical distribution network in Vietnam, the highest confidence level of the load dataset is about 95% when doing the STLF.