The Use of Singular Spectrum Analysis and K-Means Clustering-Based Bootstrap to Improve Multistep Ahead Load Forecasting

In general, studies on short-term hourly electricity load modeling and forecasting do not investigate in detail the sources of uncertainty forecasting. This study aims to evaluate impact benefits applying bootstrap aggregation overcoming time series forecasting, thereby increasing accuracy multistep ahead point forecasts. We implemented existing proposed clustering-based bootstrapping methods generate new series. method, we use singular spectrum analysis decompose between signal noise reduce variance bootstrapped The is then by K-means generation Gaussian normal distribution (KM.N) before adding it back signal, resulting apply benchmark models for SARIMA, NNAR, TBATS, DSHW, model all determine forecast values obtained from original are compared with mean median across forecasts calculated using Malaysian, Polish, Indonesian 12, 24, 36 steps ahead. conclude that, this case, method improves multistep-ahead values, especially when considering SARIMA NNAR models.