Short-Term Load Forecasting for Campus Building with Small-Scale Loads by Types Using Artificial Neural Network

摘要

Nowadays, the large portion of total energy is consumed by buildings, so it is essential to use the energy efficiency in buildings. With the deployment of a smart meter in the building, a small-scale load data are now available such as lighting, general and the other loads. We first analyze the small-scale loads through the fixed k-means clustering algorithm. Based on the analysis of load characteristics, we propose five artificial neural network-based forecasting models for campus buildings. The five models use total load, temperature and small-scale loads as input data with different input combinations. The case studies show that using small-scale loads improves the forecasting performance.