Intelligent system for power load forecasting in off-grid platform

Accurate and reliable load forecasting is a very important and required task conditioning the operation and management of electrical power generation systems. It is a key issue especially in planning and controlling the power grid system. The load forecasting process makes part of a smart control system. In off-grid platforms, smart control systems are needed to keep the consumed power equal to generated power as well as to maintain the power quality at standard levels of power quality parameters. Many mathematical models have been designed for load forecasting, including artificial neural network (ANN), decision tree (DT), support vector machine (SVM), fuzzy sets, etc. Still, the power load forecasting remains an open issue. In this article, we introduce an intelligent approach that predicts electrical load using data taken from an off-grid platform. The proposed approach builds on four models, namely K-means with ANN, K-means with DT, K-medoids with ANN, and K- medoids with DT. The article describes the design of these four forecasting models and compares them. The simulation results of the four models were evaluated and compared using mean absolute percentage error (MAPE) criteria. The best forecasting results were obtained using K-medoids clustering combined with ANN, where the MAPE was about 8%.