Optimization Model for Biomass Plant Control

Development and validation of a computational operation model for an industrial biomass plant, capable of processing data from multiple sensors in real time and recommending operating configurations that maximize combustion efficiency. Given the intrinsic variability of organic residues and fluctuations in their moisture content, the ideal combustion model is highly dynamic in nature. Solving this problem requires the execution of the following fundamental steps:

1. Multimodal data ingestion and processing: designing an architecture capable of fusing data from heterogeneous sensors (numerical signals for temperature and pressure, and unstructured data from video) into a unified data stream.

2. Prediction/simulation model development: creating a model (based on Artificial Intelligence, Machine Learning, or physical-mathematical models) capable of predicting the thermal efficiency of the boiler based on current state variables.

3. Operating parameter optimization: implementing an optimization algorithm that identifies, in a multidimensional space, the ideal combination of inputs for the current scenario.

4. Image analysis for combustion control: developing computer vision algorithms to extract flame characteristics (color, position, height, etc.) from the video feed, using them as early indicators of combustion quality.

5. Decision support system prototyping: developing an interface that provides operators with actionable real-time recommendations, enabling dynamic adjustments to boiler operation.

6. Validation and testing: evaluating model performance against historical data and validating the accuracy of suggestions against pre-established efficiency metrics.

Cooperation with company or external entity: The work will be developed with Renova's teams.