火电厂污染物排放量的预测与控制

发布时间：2018-07-30 08:35

【摘要】：目前环境污染问题已经成为全世界的焦点。火力发电厂在保障电力充足供应的同时,也成为了重要的污染源。如何采取有效的控制措施来限制污染物的排放已经成为了当前亟待解决的问题,特别是"超低排放"的概念提出之后,针对火电厂污染物排放量的先进控制策略也相继出现。但是对于污染物排放量的预测研究还为数不多,一些先进的智能预测方法还未能应用到该领域。比如说在其他领域都有着很好的应用效果的神经网络预测方法,如果能将神经网络的预测能力应用于实际生产过程当中,对污染物排放量的控制将大有裨益。Elman神经网络本身具有动态建模的功能,非常适合对未知模型的预测估计。本文应用遗传算法优化Elman神经网络权重和阈值,对污染物排放量进行预测。证明了改进的Elman神经网络相比于传统的Elman神经网络可以更好的预测污染物排放量,预测误差更小,预测模型更加精确。小波神经网络是神经网络学习的一种,网络结构与典型的BP神经网络类似,隐含层所用激活函数为小波基函数。在既要加快学习速度又要保持系统稳定的前提下,可以在权重和参数修的学习规则中添加一项动量项,起到保证系统稳定性和避免局部最小化的作用。学习规则修改后相比于之前在数据预测的精度方面有了明显的提高。如何在兼顾企业经济效益的同时,又不对环境造成污染,这是在完成污染物排放量预测之后要解决的问题。本文先将非线性规划法与遗传算法相结合,改善了遗传算法局部寻优效果差的问题。再将遗传非线性规划法与理想点法搭配使用,用于优化多目标函数。并针对目前火电行业要减少污染与降低发电成本之间的矛盾,将改进的理想点法应用于对火电厂污染物排放的优化上。尽量做到在不影响发电负荷,并减少一次性投资的前提下,同时减少污染物的排放。并且通过现场采集数据,在Matlab软件上进行仿真,针对传统方法与改进方法的效果进行了对比,结果表明改进后的办法能够明显的降低污染物排放量。能应用到实际当中的话,每年可减少的污染物排放量相当可观。
[Abstract]:At present, the problem of environmental pollution has become the focus of the world. Thermal power plant has become an important source of pollution while ensuring sufficient power supply. How to take effective control measures to limit the emission of pollutants has become an urgent problem, especially after the concept of "ultra-low emission" was put forward, advanced control strategies for pollutant emissions from thermal power plants have emerged one after another. However, there are few researches on pollutant emission prediction, and some advanced intelligent forecasting methods have not been applied to this field. For example, the neural network prediction method, which has good application effect in other fields, if the prediction ability of the neural network can be applied to the actual production process, The control of pollutant emissions will be of great benefit. Elman neural network itself has the function of dynamic modeling, which is very suitable for the prediction and estimation of unknown models. In this paper, genetic algorithm is used to optimize the weight and threshold of Elman neural network to predict pollutant emissions. It is proved that the improved Elman neural network can predict pollutant emission better than the traditional Elman neural network, the prediction error is smaller, and the prediction model is more accurate. Wavelet neural network is a kind of neural network learning. The network structure is similar to the typical BP neural network. The activation function of the hidden layer is wavelet basis function. On the premise of speeding up the learning speed and maintaining the stability of the system, a momentum term can be added to the learning rules of weight and parameter revision, which can guarantee the stability of the system and avoid the local minimization. The accuracy of data prediction is improved obviously after the modification of the learning rules. How to take into account the economic benefits of enterprises without causing pollution to the environment is a problem to be solved after the discharge forecast of pollutants is completed. In this paper, the nonlinear programming method and genetic algorithm are combined to improve the problem of poor local optimization of genetic algorithm. Then the genetic nonlinear programming method and the ideal point method are used to optimize the multiobjective function. In view of the contradiction between reducing pollution and reducing the cost of power generation in thermal power industry, the improved ideal point method is applied to the optimization of pollutant emission from thermal power plants. At the same time, reduce the emission of pollutants without affecting the generation load and reducing the one-time investment. The results show that the improved method can obviously reduce the emission of pollutants by comparing the effect of the traditional method with the improved method through the field data collection and simulation on the Matlab software. The amount of pollutant emissions that can be reduced each year is considerable if it can be applied in practice.
【学位授予单位】：山西大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP183;X773

【相似文献】