Artificial Neural Networks for Water Inflow Forecasting - The power of Machine Learning to Real Life Problems
Published:
Approximately 4 years ago, we embarked on a journey around the world of machine learning and artificial neural networks to seek ways where we could achieve improvements associated with the process of forecasting natural water inflows at hydroelectric power plants. Victor A. D. Faria, at that time still a graduate student in the electrical engineering course at the Federal University of Itajubá, and today a Ph.D. at North Carolina State University, accepted the research challenge with me, and professors José Wanderley Marangon Lima and Luana Medeiros Marangon Lima. Since then, a lot has happened and many fruits are being harvested with an unprecedented and important publication in the International Journal of Environmental Science and Technology that deals with water inflow forecasting using multi-layer Perceptrons neural networks for all hydroelectric plants that participate in centralized dispatch in the Brazilian power system. Results obtained in this study point to greater accuracy and precision of the neural network models developed in relation to those obtained by the models in use in the Brazilian electricity sector. This work comes at a very important time for the sector, where concerns associated with droughts, hydrothermal dispatch and energy prices are in the spotlight. In order to plan and make more robust decisions, it is necessary to improve the representation of future uncertainties (water inflows, electricity demand, wind and solar generation, climate/weather, etc). Congratulations Victor for your dedication and hard work and also to everyone else involved in the work for their important contributions. Article link: An assessment of multi-layer perceptron networks for streamflow forecasting in large-scale interconnected hydrosystems