Predicting Unpredictable, Controlling Uncontrollable

Selection of activated sludge process parameters that provide required sludge settling and effluent quality at the lowest operating cost is a challenging task. There are no mechanistic models that can be used for predicting sludge settling, and until now there were no reliable methods that can help to quantify process parameters under which the desired sludge settling is achieved. Since sludge settling is one of the key indicators of activated sludge performance it was impossible to automate selection of optimum process parameters.The paper describes a new method that allows selection of process parameters under which the desirable settling can be achieved. The method is based on data mining technique and utilizes neither mechanistic nor stochastic models. The method was tested using databases from two activated sludge systems. Results showed that optimized process parameters allowed achieving desirable sludge volume index (SVI) values 72% of the time at one system and 100% of the time at another. Prior to optimization these SVI values were achieved only 32% and 39% of the time, correspondingly. The developed method in a combination with mechanistic models of the activated sludge process provides a good theoretical basis for automating process parameters selection.