论文部分内容阅读
电厂给水水质的提高使得低磷酸盐处理成为炉水调节的常用手段。为便于管理,炉水水样一般通过取样管传送至化学车间进行统一检测和分析。由于水样的传送和分析需要耗时,炉水磷酸根控制成为典型的时滞系统控制。现有的自动加药系统多采用PI控制,该方案能基本保证加药控制的稳定性,但在取样管较长,时滞明显的场合,不能对磷酸根浓度进行精确调节,这一缺陷在系统调峰和变负荷运行时更加明显。为改善磷酸根浓度的控制效果,本文提出了炉水低磷酸盐处理的LM-Smith控制方案。该方案根据经典的Smith控制,在取样点实施管道改造,增设现场磷酸根表,通过电信号反馈补偿管道传送的时延。为克服传统Smith控制抗干扰性差的弱点,引入Levnberg-Marquardt神经网络算法,在线辨识Smith预估模型,实现对被控对象的准确跟踪,仿真结果表明该方法响应迅速,辨识精度高,输出信号准确,鲁棒性较强,改善了Smith控制的效果,提高了磷酸盐自动加药系统的控制性能。
The improvement of water quality in power plants makes low-phosphate treatment a common method for boiler water regulation. For ease of management, boiler water samples are generally sent through the sampling tube to the chemical plant for uniform detection and analysis. Due to the time-consuming transport and analysis of water samples, the control of furnace phosphate has become a typical time-lag system control. The existing automatic dosing system to use more PI control, the program can basically guarantee the stability of the dosing control, but the sampling tube is longer, the time delay is obvious, can not be accurately adjusted for the concentration of phosphate, a disadvantage of Peak shaving and variable load operation is more obvious. In order to improve the control effect of phosphate concentration, a LM-Smith control scheme of low phosphoric acid salt treatment was proposed in this paper. The program according to the classic Smith control, the implementation of pipelines in the sampling point transformation, additional field phosphate table, through the electrical signal feedback compensation pipeline delay. To overcome the weakness of traditional Smith control, the Levnberg-Marquardt neural network algorithm was introduced to identify the Smith predictor model on-line and realize the accurate tracking of the controlled object. The simulation results show that this method has the advantages of fast response, high recognition accuracy and accurate output signal , Strong robustness, improve the Smith control effect, improve the phosphate dosing system control performance.