Design Of Remote Monitoring System, Research And Application Of Support Vector Machine In Sewage Treatment Process

At present, some large and medium scale water enterprises are responsible for operating and maintaining a number of sewage treatment plants and each of them is operating independently. The operation data and craft parameters could not be collected to corporate center immediately. The situation of information lag is caused for enterprises. It failed to real-time centralized management and it is also inconvenient to control. With the increasing developing of communication and network technology, the traditional management model has been changed greatly and technology of remote monitoring is feasible. The process of sewage treatment has various characteristics, such as nonlinearity, time-variety and complexity, so it is difficult to detect several critical variables in real time. Although many on-line detectors are developed successfully at present stage, but many problems still exist, such as expensive price, high maintenance costs and long detection period. The presence of soft sensors technique provides the possibility for solving these problems.My work starts from actual project, based on the research of a sewage treatment plant, the remote monitoring system has been designed and developed, soft sensor technology has been researched and applied in the process of sewage treatment. The main tasks and innovations are as following:(1) The remote monitoring system has been designed and developed, which contains several remote monitoring stations and a remote monitoring center. A remote monitoring station is developed based on OPC and ZMQ interface, several functions are fulfilled such as field data collecting, data retaining, data encoding, data compressing, data encrypting, remote commands sending and receiving. In order to improve system reliability, two sets of identical and independent configuration are designed and the technology of Master-Slave backup is used.(2) The remote monitoring center distinguishes various sewage treatment plants from different factory_id. Various fields data is received and plotted in real-time trend graph. Meanwhile, commands to each plant are sent, remote control is achieved.(3) Libsvm is reexploited, which is a kind of open source soft sensor toolkit, data preprocessing function and grid search optimization function are rewritten, soft sensor module is integrated into system. Modeling and prediction for COD and TN are researched based on the real-time and history data of remote monitoring system. The conditions and procedures of model updating are raised. Thereby, soft sensor could be adapted to different operating conditions.