基于多变量的火力发电厂烟气脱硫PH值智能检测方法研究

发布时间：2018-06-07 21:14

本文选题：湿法烟气脱硫 + PH值　；参考：《西南石油大学》2015年硕士论文

【摘要】：火力发电厂烟气脱硫系统中,吸收塔浆液PH值的大小直接关系到脱硫效率的高低。脱硫效率是评价火电厂S02排放是否达标的唯一指标,因此,对吸收塔浆液PH值进行实时、精确的检测显得尤为重要。在工程实践中,吸收塔浆液PH值代表吸收塔溶液的酸碱度,通常是采用PH值测试仪进行测量,但PH测试仪失效的情况时有发生。当PH值测试仪失效时,易造成吸收塔浆液酸碱度的失衡。酸碱度失衡的吸收塔浆液会加速脱硫系统设备的损坏,严重时会造成整个脱硫系统瘫痪。为了将PH测试仪失效对整个脱硫系统的损害降到最小,脱硫系统在吸收塔浆液PH测试仪失效时会采用临时方法对吸收塔浆液PH值进行测量。一般情况下,常用的临时方法是采用冗余设备或现场取样对吸收塔浆液PH值进行测量。鉴于传统的临时方法存在着经济成本高、时滞性大的缺点,研发一种新型的PH值检测方法就变得尤为重要。本文针对石灰石-石膏湿法烟气脱硫工艺进行研究,提出一种将人工智能模型应用于吸收塔浆液PH值检测的方法。首先,将湿法脱硫系统中吸收塔浆液PH值的影响因素烟气流量、S02浓度、02含量、粉尘含量、烟气温度、吸收塔浆液密度、石灰石浆液密度等多个主要运行指标作为输入变量,吸收塔浆液的PH值作为输出变量,分别建立偏最小二乘回归(PLS)、粒子群优化的BP神经网络(PSO-BP)、模拟退火优化的支持向量机(SA-SVM)以及遗传优化的最小二乘支持向量机(GALS-SVM)等检测模型。其次,取西南地区某装机容量为600MW的火力发电厂的实时检测数据,用已建立四种模型对脱硫系统中的吸收塔浆液PH值进行检测,确定人工智能模型的可行性。第三,选取同一组样本数据对上述四种人工智能模型进行检验,比较几个模型的精确度。经研究分析发现：相对于其他三种人工智能检测模型,PSO-BP神经网络检测模型的相对误差最小,具有更好的检测性能。第四,为了得到更加精确的结果,引入裁剪平均法对四种人工智能模型检测结果进行处理。最后,将优化后的人工智能模型应用于该电厂,证实了使用该人工智能模型对脱硫系统运行成本的控制有显著的效果。研究结果表明,人工智能检测模型能够用于石灰石-石膏湿法烟气脱硫系统吸收塔浆液PH值检测的研究,并为脱硫系统的安全生产、节能减排与成本控制提供更加可靠的保障。
[Abstract]:In the flue gas desulfurization system of thermal power plant, the PH value of absorber slurry is directly related to the desulfurization efficiency. Desulfurization efficiency is the only index to evaluate whether S02 emission is up to standard, so it is very important to measure the PH value of absorber slurry in real time and accurately. In engineering practice, the pH value of the slurry of the absorber represents the pH of the solution of the absorber, which is usually measured by using the PH tester, but the failure of the PH tester occurs from time to time. When the PH tester fails, it is easy to cause the imbalance of pH and alkalinity of the absorber slurry. The slurry of absorber with unbalanced acidity and alkalinity will accelerate the damage of desulfurization system equipment and cause the whole desulfurization system to be paralyzed. In order to minimize the damage to the whole desulfurization system caused by the failure of the PH tester, a temporary method will be used to measure the PH value of the slurry in the absorber during the failure of the PH tester of the absorber slurry. In general, the commonly used temporary method is to measure the PH value of the absorber slurry by using redundant equipment or field sampling. In view of the disadvantages of the traditional temporary method, such as high economic cost and large delay, it is very important to develop a new PH detection method. Based on the study of limestone gypsum wet flue gas desulphurization process, an artificial intelligence model is proposed to detect PH value of slurry in absorber. Firstly, the main operating parameters, such as S02 concentration, dust content, flue gas temperature, absorption tower slurry density, limestone slurry density and so on, are taken as input variables, which are the influencing factors of PH value of absorber slurry in wet desulphurization system, such as the content of S02, dust content, flue gas temperature, limestone slurry density and so on. The PH value of the slurry of the absorber is taken as the output variable, and the detection models such as partial least square regression, PSO BP neural network, simulated annealing optimized support vector machine (SA-SVM) and genetic optimization least squares support vector machine (GALS-SVM) are established, respectively. Secondly, taking the real-time detection data of a 600MW thermal power plant in southwest China, the PH value of absorber slurry in desulfurization system is detected by using four established models, and the feasibility of artificial intelligence model is determined. Thirdly, four artificial intelligence models mentioned above are tested with the same set of sample data, and the accuracy of several models is compared. It is found that compared with the other three artificial intelligence detection models, the PSO-BP neural network has the least relative error and has better detection performance. Fourthly, in order to obtain more accurate results, clipping average method is introduced to deal with the results of four artificial intelligence models. Finally, the optimized artificial intelligence model is applied to the power plant, and it is proved that the artificial intelligence model has a remarkable effect on the operation cost control of desulfurization system. The results show that the artificial intelligence detection model can be used to detect the PH value of slurry in the absorption tower of limestone gypsum wet flue gas desulfurization system, and provide more reliable guarantee for the safe production, energy saving and emission reduction and cost control of the desulfurization system.
【学位授予单位】：西南石油大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：X773;X831

【相似文献】