• 2017 Jimenez, Victor A. and Lizondo, Diego and Will, Adrian and Rodriguez, Sebastian, Short-Term Load Forecasting for Low Voltage Distribution Lines in Tucumán, Argentina, In 5to Congreso Nacional de Ingeniería Informática / Sistemas de Información (CoNaIISI 2017), 2017.
    • Tipo: En congreso


      Resumen: Load forecasting is a critical technique in the decision-making process for the proper management of different aspects of the electricity distribution network. Load Forecasting is a very complex problem, and there is a wide variety of techniques to address it. However, it is the data and local conditions that determine what technique is the best. In this work, we compare different One-Day-Ahead Load Forecasting methods, applied to data from low voltage distribution lines in Tucumán, Argentina, to determine which one is the most suitable for this region. We describe in detail each step of the methodology, including the application of a variable selection method based on Genetic Algorithms. In the forecast model building stage, we used and compared three algorithms: Multi-Linear Regression, Radial Basis Function Neural Network, and Feed-forward Neural Network. Results show that Radial Basis Neural Networks is the best in our case, which allows predictions with a Mean Absolute Percentage Error (MAPE) between 6.9% and 10.1% in the selected cases. This accuracy is sufficient for the management improvement purposes of the local electricity company.