DAI-DEPUR: An Environmental Decision Support System for the control and supervison of Municipal WasteWater Treatment Plants

This paper introduces DAI-Depur, an Environmental Decision Support System (EDSS). This systems supports the supervision of Municipal Waste Water Treatment Plants (WWTP). The paper highlights the succesful application of AI techniques to a complex environmental problem as the wastewater treatment that has both social and economic impact as well as in the environment. DAIDepur is successfully performing real-time support to the operation of the Granollers facility, in Catalonia, since September 1999. 1 GENERAL SPECIFICATIONS The increasing degradation of the environment has forced the society to consider changes in human behaviour for ensuring the essential conditions for the life in the Earth. This consideration has encouraged research and a great effort has been placed on understanding, preventing and correcting environmental degradation. In this sense, the treatment of water and wastewater has become one of the most important environmental issues. Wastewater treatment is fundamental to keep the water natural resources (rivers, lakes and seas) in as high quality as possible. Not only for this environmental reason, but also due to the more and more restrictive social regulations, the correct management of wastewater treatment facilities has become very important during the last 20 years. Environmental systems possess several inherent characteristics which make their understanding and control difficult: they evolve over time, involve processes which take place in a 3-dimensional space, are complex, involve interactions between physical-chemical and biological processes, are stochastic, and, very often, are periodic in time [6]. The complexity and the multi-faceted nature of many environmental problems, suggests that their suitable management cannot be based on a single technique. Municipal Waste Water Treatment Plants (WWTP) are clear examples of complex processes which meet all these distinctive features of environmental systems. Though an enormous amount of research effort has gone trying to improve control and supervision of WWTP, its correct management is still far from beings successfully solved by the research community. Classical control methods based on mathematical models, have been successfully used to improve and optimize WWTP operation. However, these classical control methods show some limitations when trying to control the activated sludge process of WWTP, mainly 1 Technical University of Catalonia. c/Jordi Girona 1 & 3. 08034 Barcelona, CAT. email: [email protected] 2 University of Girona. LEQUIA. Campus Montilivi s/n. 17071 Girona. CAT. email: [email protected] when the plant is not working under the ideal (normal) state. There are many characteristics of the process that difficult the success of classical methods, e.g. the inflow is variable (both in quantity and in quality); not only there is a living catalyst (the microorganisms) but also a population that varies over time both in quantity and in the relative number of species; the knowledge of the process is scarce; there are few and unreliable on-line analyzers; and most of the data related to the process is subjective and can not be numerically quantified. Although progress in control engineering, computer technology, and process sensors has enabled automatic control improvement, integrated operation of WWTP is still far from being solved. The number of measured variables in a WWTP is increasing and the need and possibility to control the process is becoming greater. With this increasing instrumentation there is certainly more information available, but it must be reminded that data rich is not the same as information rich [13]. It is not an easy task for operators and process engineers to acquire, to integrate and to understand all this day-today increasing amount of information. The solution could arrive with the development of knowledge-based decision support systems that handle the particular characteristics of the process, using this available but incomplete information to guarantee the quality of the discharged water. Although knowledge-based systems came into picture in the 1980s, some authors suggest that they never succeeded for two reasons: they were too complex, and the available knowledge could not be captured in reliable models and advisory systems [9]. A support system cannot only be based on mathematical modelling, but must also take advantage of heuristic knowledge from literature and experts, while including specific experiences accumulated through years of experience in the facility itself. There have been some approaches to improve WWTP operation using single knowledge-based techniques, such as [1, 11]. However, an effective decision support system to supervise the operation of the process should be described as a hierarchical multi-level structure that integrates different concurrent modules, overcoming the limitations in the use of each single technique, and providing higher accuracy, reliability and usefulness [4]. A reasonable proposal should link advanced and robust control algorithms to some knowledge-based techniques, allocating the detailed engineering to numerical computations, while delegating the logical analysis and reasoning to supervisory intelligent systems [15]. This paper describes the implementation of an Environmental Decision Support System (EDSS) to supervise and control the operation of a real WWTP. This EDSS integrates advanced control algorithms with knowledge based and cased based systems in an hybrid architecture. The EDSS is performing real-time support to process operation since September 1999. The development of this architecure has been previously described in [10]. 1.1 Plan of the paper Section §2 describes the implementation of DAI-Depur the proposed EDSS into a real WWTP. Results of the first four-month validation period are also shown and discussed in §3. These results are compared with those of year 2001. The facility selected to implement the EDSS is located in Granollers, in the Besòs river basin (Spain). The water treatment line encompasses preliminary, primary and secondary treatment to remove the organic matter, the suspended solids and, under some conditions, the nitrogen contained in the raw water of about 130,000 equivalent-inhabitants. The sludge treatment line encompasses thickening, anaerobic digestion and dewatering. The raw influent comes from a combined sewer. In §4 we give the final conclusions and talk about the key steps which are necessary to transfer the system to another facility and the economic and social impact of this process.