Intelligent Assistant for Smart Factory Power Management

Abstract

This article presents the development and implementation of an intelligent assistant for smart energy management in an industrial setting. The project, developed in collaboration with Bosch Termotecnologia Aveiro, aims to leverage the advancements of Industry 4.0 and address the increasing demands for efficiency, quality and sustainability in industrial processes. The intelligent assistant implements machine learning algorithms, specifically autoencoders, to predict energy consumption related variables and detect anomalies in data collected from the factory floor. In the models’ building process, an AutoML hyperparameter optimization tool is utilized to further enhance their predictive abilities. The implementation of the assistant showcases its capabilities not only regarding the predictions, but also in terms of data visualization and alarm generation. The results obtained through the use of Auto ML showcase remarkable improvements compared to traditional regression models and basic autoencoders, proving its effectiveness in terms of making accurate predictions, with improvements up to 90% in some indicators when compared to those basic models. Alsointerms of the anomaly detection task, the chosen methodology proves to be effective in detecting different kinds of anomalies. The integration of the assistant with a Grafana dashboard and the exploration of additional communication tools, such as a Telegram bot, further improve the user experience and expand the potential applications of the assistant.