AbstractThe desirability of effective control of anaerobic digesters as a means of avoiding imbalance in the microbial population has become clearer and this can be seen from the literature in recent years. A number of published control strategies have been encouragingly successful, however the non-linear and time varying nature of the process generally requires a bespoke, engineered system dependant on the characteristics of the system.
The 'cost of knowing' in employing control systems, is generally high. The ideal scenario for operators would be the availability of a generic control system at reasonable cost, which would be applicable to a large group of high rate reactor designs, operating on a limited (but broad) variety of waste streams.
The system would be able to control from commissioning through to steady state and should be able to cope with reasonable expected shock loading conditions, albeit perhaps at some degree of sub-optimality. The aim of this work is to develop a control strategy, which will lead in future to this end.
Bicarbonate alkalinity (BA) is a key parameter which indicates the buffering capacity of the anaerobic digestion system and which has the potential for helping to maintain a stable system in the face of changing organic and toxic load. This is particularly the case when used in association with other informative on-line parameters such as gas production rate, %CO2 concentration in the
gas, TOC, pH and volatile fatty acids.
All but the last of these have been investigated using a fluidised bed reactor and the degree to which the anaerobic process is non-linear and time varying has been assessed, as the level of complexity required to represent anaerobic digestion 'well enough' was not clear. Simple linear black box models of low order were investigated, predicting over a limited horizon and relying on current and recent data values to refine the prediction.
Independent black box ARX models were identified for gas production rate, % CO 2 , bicarbonate alkalinity and Total Organic Carbon using on-line data from a fluidised bed reactor at varying organic load. Model predictions looked ahead one sample step (30 minutes) and when validated using data obtained in a different time period (separated by 4-8 weeks) gave significant predictions in each case.
The non-linear nature of the process was found to have little effect over the operating conditions investigated. Also the variation of the process within 4-8 weeks period was not sufficient to cause the models to predict badly.
The performance of three black box models which were parameterised and validated using data collected from the same laboratory scale fluidised bed anaerobic digester, were compared. The models investigated were all ARX (auto regressive with exogenous input) models, the first being a linear single input single output (SISO) model, the second a linear multi-input multi-output (MIMO) model and the third a non-linear neural network based model. The performances of the models were compared and it was found that the SISO model was the least able to predict the changes in the reactor parameters (bicarbonate alkalinity, gas production rate and % CO2 ).
The MIMO and neural models both performed reasonably well. Though the neural model was shown to be superior overall to the MIMO model, the simplicity of the latter should be a consideration in choosing between them. A simulation with a horizon approaching 48 hours was performed using this model and showed that the method was not sufficiently accurate for use in situations where pure simulation was required.
This thesis includes the use of a two population deterministic model calibrated using data from a fluidised bed reactor operating on a simulated yeast waste, in the development of a Model Reference Adaptive Control (MRAC) strategy. The strategy uses a three term adaption mechanism, which is described in the thesis as a Fast Adaption Trajectory (FAT) strategy, as it was found to be necessary to respond to catastrophic events over short time scales, in order to maintain the viability of the bacterial population.
Numerical optimisation in a simulation environment was used to parameterise the controller, and this was done on the basis of only basic design information being available for the reactor which was to be controlled.
The controller was tested on a significantly different Expanded Granular Sludge Blanket (EGSB) reactor operated on a sucrose based feed and which did not inform the controller design process beyond basic physical information. Two actuation strategies were explored over several months of operation, using a single on-line bicarbonate alkalinity monitor, which in the event proved to have significant reliability problems.
Not withstanding the problems with the alkalinity monitor, which was dominant in determining the success or failure of the control strategy, it was found that the control strategy was able to maintain control during start-up, which was the ambition of this part of the experimentation. Both actuation methodologies showed promise although the variation of loading rate was not adequately tested by the experimentation, which was conducted.
The actuation by dosing with bicarbonate proved to be better at maintaining control in the face of repeated and severe perturbations caused by failure in the bicarbonate monitor system. It is believed that the FAT controller is likely to be a transferable technique provided that unmodelled dynamics are not excessively dominant and that the reactor system is comparable to a CSTR design with predominantly soluble waste in the feed.
|Date of Award||Feb 2003|
|Supervisor||Dennis Hawkes (Supervisor) & Freda Hawkes (Supervisor)|
- activated sludge process