← Back to Search
Estimation Of Hydrolysis Parameters In Full-scale Anerobic Digesters.
Published 2009 · Mathematics, Medicine
Reduce the time it takes to create your bibliography by a factor of 10 by using the world’s favourite reference manager
Time to take this seriously.
In hydrolysis-limited anerobic systems, the key parameters describing degradation are degradability extent (f(d)), and the lumped apparent first order coefficient (k(hyd)). These are often measured in biological methane potential (BMP) tests. Using modern techniques, it should also be possible to estimate these parameters in full-scale systems, especially where inputs are dynamic. In this study, we evaluated f(d) and k(hyd) values and uncertainty based on nonlinear parameter estimation from (i) BMP tests and (ii) effluent gas and solids from two full-scale digesters fed with highly variable feed flows and concentrations (up to 6 kg COD m(-3) day(-1)). The substrate was thermally hydrolyzed activated sludge, and the inoculum for BMP tests was from the full-scale digesters. While identifiability of both parameters in the BMP tests was generally good, only f(d) could be well identified using continuous data. For k(hyd) using continuous data, normally only a lower limit could be found (upper was unbounded). In addition, parameters as estimated on different outputs (VS and gasflow) and two different digesters were consistent, with an f(d) value of 0.45-0.55, and a k(hyd) value of >5 day(-1). Gradual changes in f(d) over the 450 days could be related to upstream changes. f(d) values as estimated in BMP tests were consistent (if conservative) with continuous estimates, with a f(d) in BMP of 0.4-0.5. k(hyd) values were an order of magnitude lower (0.15-0.25 day(-1) vs. >5 day(-1)), and this translated to very poor model performance when BMP-estimated values were used in the continuous model. This means that while BMP testing may be used for project feasibility analysis, values obtained should not be used for dynamic modeling. The parameter confidence regions found were highly nonlinear, especially for continuous systems, indicating that iterative or sampling techniques are required for an estimate of real parameter uncertainty.