Data-driven stochastic modelling of manufacturing systems for sustainable energy management

Relevance: the critical issue of reducing the environmental impact of manufacturing factories through sustainable energy efficiency. It emphasizes the importance of energy conservation technologies in achieving energy efficiency and introduces a novel, data-driven control setup for optimizing energy consumption in complex systems. The study is particularly relevant to the system engineering community as it explores the use of machine learning, specifically Gaussian Processes, to model and control energy consumption in manufacturing systems. Aim: to develop and demonstrate a new approach for enhancing energy efficiency in manufacturing systems by using machine learning, specifically Gaussian Processes Regression (GPR). The study seeks to correlate the dynamics, complexity, and interrelated energy consumption recordings of production machines and apply this model within a Model Predictive Control (MPC) framework to achieve optimized energy-saving solutions. Methods: building a model using Gaussian Processes Regression based on historical sensor data collected from production machines. This model captures the dynamics and energy consumption patterns of the machines. The GPR model is then integrated into a Model Predictive Control loop to generate optimal control actions that minimize energy consumption at each control time step. A numerical example is provided to demonstrate the effectiveness of the proposed method. Results: the proposed GPR-based modelling approach is effective in identifying energy-saving opportunities and quantifying their potential. The model provides a trust region with 95% confidence, allowing for the identification of previously unseen energy-saving challenges. The study suggests that this approach can serve as a valuable tool for companies with established energy improvement programs, offering a new perspective for further enhancement of energy efficiency. Keywords: Gaussian processes, machine learning, model predictive control, stochastic model, sustainable manufacturing.