System and method for controlling effluents in treatment systems

A system and method is provided that predicts operational parameters for all unit operations in water treatment plants or the like. Initial training with historical operations data, for example, allows the system and method to develop equations that can in turn predict the present and future performance of the plant in real time. In addition, the system and method can control operations of the plant in real time. The system improves the performance of the plant to meet predetermined subpoints of various parameters. For example, the predetermined subpoints can be used to enable the plant to meet regulatory needs while controlling for other parameters such as cost, chemical fees, flow rates and power consumption. The system and method include a non-linear predictive model for turbidity. The system considers the influent water quality and analyzes treatment options available to predict the dose of various chemicals required to get desired treatment. It will then predict plant performance resulting from intended operator changes in real time. The system preferably includes general regression neural networks with modeling modifications to learn if the works including learning patterns to make predictions and cost for operations control of unit operations and/or the system. The system includes virtual sensors for parameters that cannot be detected on-line. The system and method determine sufficient data to monitor and control all water quality parameters in the water treatment plant. The water treatment plant operations can be predicted and controlled as a plurality of coupled unit operations. In one embodiment, a unit operation block consist of a power mixer, a rapid mix basin, flocculation basin, and settling tank controlled as a coagulation control loop.