Our paper Benchmarking Foundation Models for Zero-Shot Water Demand Forecasting has been accepted at the Encontro Nacional de Inteligência Artificial e Computacional (ENIAC), which will be held alongside BRACIS in Cuiabá, Mato Grosso, Brazil, from October 19 to 22.

This publication is the result of the undergraduate thesis by Luana Rathunde Sandler, lead author of the work.

The study evaluates the zero-shot application of time series foundation models, specifically Chronos and Moirai, for daily water demand forecasting. Using data from a coastal city in Southern Brazil characterized by high seasonal volatility due to tourism, we compared these approaches against established statistical and machine learning baselines. Our findings demonstrate that Transformer-based foundation models achieve competitive performance, highlighting their viability for rapid deployment or data-scarce scenarios.

As detailed in the conference website: “ENIAC 2026 is the 23rd edition of a series of successful meetings bringing together Artificial Intelligence and Computational Intelligence, supported by Brazilian Special Interest Groups on Artificial Intelligence and Computational Intelligence from the Brazilian Computer Society. ENIAC is part of the 36th Brazilian Conference on Intelligent Systems (BRACIS). This Brazilian meeting provides a forum for researchers, practitioners, educators, and students to present and discuss innovations, trends, experiences, and developments in the Artificial Intelligence and Computational Intelligence fields. “In particular, it is the ideal event for postgraduate and undergraduate students to submit and present their ongoing results, even their first papers.”

As soon as published, the paper will be linked at the Publications page.