Methods for anaerobic digestion model fitting—comparison between heuristic and automatic approach

Abstract

The article demonstrates if automatic optimization can be better than manual adjustment. The subject of optimization was the temperature-phased anaerobic digestion (TPAD) model. A selection of 3 parameters per each reactor in the process chain was appointed—reaction rate for propionate conversion, acetate conversion, and hydrolysis. Overall, both methods provided very convergent results. However, the total summary error (TSE) for the automatic algorithm was always moderately lower than for manual—the difference varied between 16.16 and 57.05 percentage points. Although the manual method has significant advantages—adjustment was more homogenous and gave more uniform fitting. Finally, cross-validation was performed to unify the values between the experimental series. The result was a total number of 4 values for each optimized constant—for two temperature points in each of two methods. Due to inconclusive information about the accuracy, averaged values were calculated to use in further researches. The recommendation from this article is to connect the best aspect of both methods to achieve the most accurate results.