ABSTRACT
 
Wastewater treatment process is a complicated industry process which is difficult
to control，because it has the characters of multi-variables，non-linear，time-varying，randomicity  and  complexity.  Therefore,  it  is  difficult  to  implement  steady  running  of the system and the effluent according with the standard by the traditional control theories and  technologies.  Currently,  the  application  research  of  the  artificial  intelligence technique combined with other application fields in engineering systems has became the hotspot, and obtained greater achievement. However, it is still on the stage of groping research  at  home  in  the  AI  applications  of  wastewater  treatment.  The  urgent  problem are  not  only  how  to  apply  artificial  intelligence  technique  realizing  steady  and dependable  running  of  the  process  of  wastewater  treatment,  also  how  to  ensure  the effluent  according  with  the  standard  and  energy-saving.  This  dissertation,  analyzed  the status quo of the AI application in wastewater treatment fields aboard and home, studied the  multi-agent  theory  of  distributed  artificial  intelligence,  fuzzy  set  theory,  expert system  theory,  and  artificial  neural  network  theory,  put  forwards  a  new  idea  of  a distributed  intelligent  system  of  wastewater  treatment  based  on  multi-agent  theory  in order to realize the intelligent control of wastewater treatment process, on-line measure of  the  technologic  parameters  which  is  difficult  to  measure  and  real-time  diagnosis  of the  exceptional  phenomena,  developed  the  prototype  system  based  on  the  opening JADE developing platform according to the FIPA standard by Fuzzy JESS, Hibernate and Spring framework   techniques etc, and offered feasible route for the application of
the Multi-Agent theory.  Main results of the research as following:
   ①  Through  the  analysis  of  agent,  MAS  theory  and  the  characters  of  wastewater treatment process，brought forward the hierarchical structure and the system architecture of the intelligent distributed wastewater treatment system based on multi-agent theory, defined the inner logic frame and function of each agent of the system model. The  research  results  indicate  that  the  hierarchy  of  the  agents  of  the  intelligent system is clear, modularized, the agents is independent as well as contacted with each other organically, and the robustness and the expansibility of the system are excellent.
②  In order to actualize the multi-agent system, guarantee the opening trait of the
system,  interoperate  among  the  agents  each  other,  through  the  research  on  the  FIPA  standard, JADE Agent developing platform, Fuzzy JESS expert system developing tool and Hibernate, Spring Framework database technology, utilized the automations system platform of the wastewater treatment, the intelligent system has been implemented.
  ③  Through the analysis of the exceptional phenomena of the wastewater treatment
for the activated sludge process, established the fault tree of the aeration tank and the fault tree of the secondary clarifier for the wastewater treatment system of the activated sludge  process,  adopted  the  forward  inference  mechanism,  used  Fuzzy  JESS  expert system  developing  tool,  realized  the  agent  of  the  fault  diagnosis  of  the  exceptional phenomena of the wastewater treatment for the activated sludge process. The agent of the fault diagnosis has the characteristics of knowledge base, inference machine, and the main  program  separated  mutually,  therefore,  it  is  good  to  update  knowledge  fountain,
and  guarantee  the  trait  of  clarity,  flexibility,  migration  of  the  system.  Platform- independent  and  multithreading  mechanism  allows  the  Fuzzy  JESS  to  execute concurrently applications.
     ④  Some key variants（such as SVI）can’t be obtained directly by online measure
or  as  the  longer  delay  times  of  getting  measure  results,  the  control  of  the  variants  is difficult.  After  studying  the  principle  of  soft-measure  techniques, gives  a  new  model method  of  the  parameters  soft-measure  of  wastewater  treatment  based  on  artificial neural networks. To enhance the real-time and validity of the artificial neural network, using  rough  set  theory  to  preprocess  the  model,  after  reducing  the  attributes  and
redundant data, set up the soft-measure model of the parameters difficultly to measure
of wastewater treatment based on rough set theory and artificial neural network. Then
using  the  real  parameters  from  a  wastewater  treatment  plant,  made  the  simulation experiment. The result shows, as the model without preprocess by rough set, the error of the measure values is not only less, but the data dimensions, and the training time and the training steps of the artificial neural network are reduced. So, this is favorable to the practicability the soft-measure model. 
     ⑤  In order to realize the process control of the wastewater treatment system, the intelligent system has set three executable agents include the agent of DO control, the agent  of  sludge  circumfluence,  and  the  agent  of  excess  sludge  discharge.  DO  control agent used Fuzzy control method which was fit to the characters of the big inertia, big lag  and  indetermination  of  DO  control.  The  agent  of  sludge  circumfluence  measures real-timely the MLSS of the aeration tank，according to the difference between the value of MLSS by real-time measure and the given value, adopted PID arithmetic to control  the running speed of the pump for the sludge circumfluence. By calculating the value of the  daily  excess  sludge  discharge,  the  agent of  excess  sludge  discharge  control  makes use of the liquidometer to control running of the pump.
     ⑥  The  interaction  among  the  agents  is  an  important  character  of  MAS,  so  the conflict  of  agents  is  unavoidable.  By  the  process  of  coordination  and  troubleshooting, agents fulfilled a problem solution. Through the analysis of the possible confliction and the  distinguish  methods  among  the  agents  in  the  wastewater  treatment  of  activated sludge process, and according to the theory of conflict negotiation and task cooperation, brought forward the troubleshooting strategy for the intelligent system, and realized the troubleshooting agent.
     ⑦  The functions of the agent of data service are to provide uniform and flexible interfaces  for  other  Agents  to  access  real-time  database.  The  reading  and  writing operation of the agents in the MAS to the real-time database are realized by using the correlative  service  provided  by  the  agent  of  data  service.  The  dissertation  adopted Hibernate  and  Spring  Framework  database  access  technology,  made  the  real-time database in the system is transparent for the agents, at the same time, the system is of outstanding migration, and long lifecycle.
     ⑧  Owing to MAS is a lax coupling system, which is composed of many agents,
communication is the foundation of realizing interaction and social organization, so the communications  among  agents  is  one  of  the  important  problems  of  MAS  research. Through  the  analysis  of  the  agent  communications  pattern,  adopted  FIPA-ACL  as  the communication language in this intelligent system, also ascertained the mode, strategy, protocol, and content of the communication, implemented the communications among the agents of this intelligent system.
The  distributed  intelligent  system  of  wastewater  treatment  based  on  multi-agent theory  could  improve  the  decision-making  level  of  the  whole  control  system,  thereby realizes  intelligent  running,  control  and  monitor  in  the  wastewater  treatment  process, enhances the water quality and implements energy saving. The software and hardware platform  of  Modern  wastewater  water  control  system,  efficient  java language environment,  and  JADE,  FuzzyJESS,  Hibernate,  Spring  Framework  etc  technology make the realization of the intelligent system is feasible.
 
Keywords: Agent,   MAS,   Intelligent System,   Activated Sludge Process, 
Wastewater Treatment