Modelling and control of the activated sludge process : towards a systematic framework

In spite of over two decades of development, dynamic models of the activated sludge process have enjoyed only limited success in their application for control purposes. Two deficiencies are especially notable. First, the models themselves are neither comprehensive nor fully reliable; second, there is a lack of suitable model-based algorithms by which to derive control strategies. Consequently, the activated sludge process is still operated in a largely conventional manner and prone to abrupt failure. The objectives of this thesis are accordingly to develop (i) a systematic procedure by which models of the activated sludge process can be identified and (ii) a more complete framework (than hitherto) within which suitable control strategies for the activated sludge process can be derived. The thesis begins with an extensive review of the three related components of activated sludge process control: (i) understanding of the normal and abnormal behaviour of the process; (ii) the control schemes specific to these various types of behaviour; and (iii) dynamic modelling of the activated sludge process. From this integrated survey, a systematic procedure is established for developing a mathematical model, its identification and the estimation of its parameters. A framework for control is then proposed, with special reference to the incorporation of dynamic models in regulating both the long-term and short-term behaviour of the activated sludge process. Based on the proposed identification procedure, a dynamic multiple-species model for simulating both normal and abnormal behaviour of the activated sludge system is developed, its structure then identified, and its parameters estimated with reference to two sets of field data (from the Norwich and Davyhulme Wastewater Treatment Works). Despite the inevitable limitations of the model as found in this identification process (in particular, its over-parameterisation), the model is considered to be broadly acceptable in its performance and thus adequate for its subsequent application in simulating both longterm and (average) short-term behaviour of the activated sludge process. Using the identified model, a comprehensive assessment of control strategies for the activated sludge process is conducted. This is focused on storm-event and sludge bulking control and, significantly, assessment of interactions among the control strategies for these two types of problem. The tested strategies include conventional procedures, various feedback controllers, and some optimal feedforward control algorithms. The manipulated control variables are confined to those routinely used, i.e. recycle rate, wastage rate, stepfeed and step-sludge. Even so, the operational flexibility of the activated sludge process can thus be fully exploited. It is found that the control of abnormal behaviour requires as early an implementation of control actions as possible. A simplified version of the model is thus subsequently developed and incorporated within an on-line estimation algorithm in order to provide a deeper insight into the behaviour of the biomass in real-time, especially in respect of its growth rate. Using this less conventional operating information, the prediction of sludge bulking is further explored.