Forecasting Short-Term Electricity Load Using Validated Ensemble Learning

As short-term load forecasting is essential for the day-to-day operation planning of power systems, we built an ensemble learning model to perform such Thai data. The proposed uses voting regression (VR), producing forecasts with weighted averages from five individual models: three parametric multiple linear regressors and two non-parametric machine-learning models. are models gradient-descent (LR), ordinary least-squares (OLS) estimators, generalized auto-regression (GLSAR) In contrast, decision trees (DT) random forests (RF). To select best variables hyper-parameters, used cross-validation (CV) performance instead test data performance, which yielded overly good performance. We compared various validation schemes found that Blocked-CV scheme gives error closest error. Using Blocked-CV, results show VR outperforms all its predictors.