0      0

2022 ISPE Annual Meeting & Expo

SES202c - Machine Learning to Sustainable Production: A Case Study

Nov 1, 2022 10:45am ‐ Nov 1, 2022 12:15pm


The presentation describes a case-study showing machine-learning capabilities in reducing environmental impacts and costs, with a very limited investment in instrumentation and software. VTU Engineering has designed the expansion of a cooling towers system for an API plant in Northern Italy (~1200 MWh/year, OPEX>100 000 /year): targets are to maintain the T-range of the cooling water, reduce pressure fluctuations in the distribution network. Six 6 induced draft cooling towers, two pumps (without VSD), and little instrumentation resulted in a plant with little controllability, available 24/7, not energy efficient.

The proposed (and simulated) solution installs a weather station, few temperature and pressure sensors, and will equip the motors with VSD to manage the power absorption. The whole system will be connected to an AI unit which follows the L.I.C.O. approach: Learn: the AI unit monitors the system. Integrate: the algorithm is tested against different scenarios (off-line), identifies variables correlation, defines control strategies (data-analysis, machine training). Control: AI unit is put on-line as a surrogate model for the plant and is used in an optimal control scheme with a tailored objective function (energy saving). New set-points are proposed to the DCSOptimize: the control strategy is continuously improved.


You must be logged in and own this session in order to post comments.