Wastewater treatment plants are systems, in which physical, chemical and biological processes run in combination. Due to increasing demands in terms of efficiency with, at the same time, increasing pressure on costs process optimisations are constantly required.

The problems with such optimisations lie, above all, in the complexity and the high degree of non-linearity of the biochemical processes. These follow exponential growth laws. Activated sludge, which is subject to fluctuations on a seasonal basis, and even depending on the time of day, can serve as an example. For these reasons, a system can react very differently to the same control command at different times. If one takes into account as well that several parameters have to be controlled at all times, then methods of linear control technology (e.g. PI controllers) – which are widespread in traditional automation technology – cannot ensure optimum operation.

>

The fundamental principle of an intelligent, optimum control system is based on the following three steps. Various computational intelligence (CI) processes or statistical processes are thereby combined in order to jointly exploit their advantages.

I. Classification of the Operational Situation:

The measured values for the system are analysed and summarised in classes. This can be prescribed by a control system or determined automatically by cluster methods such as k-mean clustering. Thus the class, in each case, describes a specific operational situation in relation to the plant.

II. State Transfer:

Following classification of the operational situation the latter must be translated into a state for the plant. In addition to the current operational situation, the state for the plant also takes into account all previous states. As a result, operational situations over a fairly lengthy period of time can be evaluated. In addition, the state transfer allows synchronisation with other parts of the system. The state transfer is translated with the aid of state machines.

III. Selection of the Control Strategy:

After the state of the plant has been determined, the selection of suitable control strategies can be made in relation to the given state. This can, for example, be made through the selection of set values for the subdivided control system.