Optimization of Domestic and Industrial Biodigestors Based on Machine Learning Techniques

Abstract

Purpose: Development of an application for determining the technical and economic feasibility of implementing and operating domestic biodigesters using rigorous mathematical modeling of the anaerobic digestion process in conjunction with Machine Learning techniques to obtain reduced metamodels. Theoretical Framework: The generation of biodegradable waste results from human activities and has detrimental environmental impacts. To mitigate this problem, anaerobic digestion in biodigesters emerges as a viable solution, promoting the production of biogas and biofertilizers, generating economic and environmental benefits. However, implementing and operating this system requires significant investments. Method/Design/Approach: The combination of the ADM1 model with Machine Learning techniques is used to create simplified metamodels, allowing for more feasible simulations and optimizations, thereby developing an application to assess the technical and economic feasibility of biodigesters. This application is obtained by packaging the reduced metamodel using the MATLAB Compiler, which will be made available as an Excel add-in. Results and Conclusion: The reduced metamodel effectively represented the rigorous Simulink model, and the optimization of the process proved satisfactory. Furthermore, the add-in generated through the MATLAB Compiler met expectations. Research Implications: Enhanced understanding of the waste biodigestion process, demonstrating the economic and environmental returns achieved when focusing more on this area. Originality/Value: Development of a tool that enables the simulation and evaluation of a biodigestion process without the need to purchase expensive software.