摘要
针对水厂净水过程中混凝投药控制过程具有影响因素多、滞后性大和非线性的特点,依托实际项目,结合BP神经网络建立了短程反馈混凝投药自控模型。模拟水厂混凝沉淀工艺开发了中试装置,并对模型进行离线检测和连续运行,结果表明:BP神经网络模型的混凝剂投加量预测值与实际值的相对误差不超过6%;在模拟斜管中取水检测沉后水浊度,可缩短停留时间约20min;短程反馈BP神经网络混凝投药自控模型对不同季节的长江水均具有良好的适应性和较高的灵敏度,能控制沉后水浊度稳定在目标范围内。
Considering multiple influence factors, large delay and nonlinearity of coagulant dosage control at water treatment plants, an automatic control model of coagulant dosage with short-range feedback was established using BP neural network. The pilot plant was developed by simulating the coagula- tion and sedimentation process of a water treatment plant, while offline detection and continuous operation of the model were carried on. The results showed that the relative error was less than 6% between the predicted and the actual coagulant dosages. Water samples were taken from the simulated inclined pipe and analyzed for effluent turbidity, which could shorten the residence time by approximately 20 rain. The automatic control model of coagulant dosage with short-range feedback BP neural network showed a high sensitivity and could be applied to treatment of to the Yangtze River source water in different seasons. The effluent turbidity could be stably controlled within the target range.
出处
《中国给水排水》
CAS
CSCD
北大核心
2013年第11期26-29,共4页
China Water & Wastewater
基金
国家水体污染控制与治理科技重大专项(2009ZX07424-004)
国家科技支撑计划项目(2012BAJ25B06-001)
关键词
混凝投药控制
BP神经网络
模拟斜管
反馈控制
中试
coagulant dosage control
BP neural network
simulated inclined pipe
feed- back control
pilot-scale test
