Quarter-hour-ahead load forecasting for microgrid energy management system

Abstract

This paper presents a quarter-hourly ahead load forecasting method using the Artificial Neural Network (ANN). The proposed method was designed and programmed using the National Instrument (NI) LabVIEW software tool. The architecture of the ANN is a three-layer feed forward neural network and its estimation technique is based on backpropagation (BP). The performance of the network is enhanced by implementing an Early Stopping (ES) algorithm to avoid the overfitting of the training data. In order to test the performance of the algorithm, historical load data obtained from Energy Market Company (EMC) in Singapore were used in training the ANN and satisfactory results were obtained. The proposed ANN-based load forecast technique was tested with two different types of training data namely the actual load and the relative incremental of the actual load. The results are discussed in the paper. The algorithm is integrated into the Microgrid Energy Management System (MG-EMS) at Laboratory for Clean Energy Research (LaCER), School of Electrical & Electronic Engineering in Nanyang Technological University (NTU). The daily forecast data is useful to Unit Commitment (UC) for minimizing schedule cost or maximizing revenue of MG-EMS.