粒子群算法及电厂若干问题的研究

发布时间：2019-01-03 13:09

【摘要】：高能耗与重污染是制约火力发电发展的重要因素,在保证安全生产和满足负荷需求的前提下,应该尽量降低电厂的煤耗和污染物排放。基于现代智能算法的优化方法是一种提高电厂运行经济性、降低污染物排放的有效手段。因此,需要广大的电力科技工作者对智能算法以及电厂问题本身展开研究。本文在全面总结粒子群算法(PSO)、负荷分配优化和NOx减排优化的发展历程及其研究现状的基础上,开展了相应的研究并取得了一些有益的成果。本文主要内容如下：(1)针对粒子群算法早熟问题和收敛问题,引入并改进了正交实验设计学习策略,从群最优粒子和次优粒子中提取有价值的信息,组成新的指引粒子；为了加强群最优粒子跳出局部最优点的能力,引入并改进了模拟退火搜索策略,以一定的概率对群最优粒子进行模拟退火搜索。在标准测试函数上测试了改进的算法,结果表明改进的算法在平均值、标准差、评价次数、成功率和成功表现等方面的性能得到了提升。(2)为克服搜索的早熟问题,根据粒子群的多样性指导惯性权重的调整,提出了一种新的多样性与惯性权重映射方法；将本文的几种改进进行综合,利用线性系统稳定性判据以及概率论从理论上论证了改进算法的收敛性。T检验和Wilcoxon检验表明改进后的算法效果优于改进以前的算法,改进的算法在平均值、标准差、评价次数、成功率和成功表现等方面的性能优于典型粒子群算法和其它若干典型算法。(3)为提高建模预测的准确性,利用改进的粒子群算法优化了支持向量机的核参数和惩罚参数；引入了处理大样本的集成支持向量机模型,并针对子支持向量机的参数选取问题进行了改进；根据不同负荷的摸底试验、燃烧器设计数据及锅炉以往运行特性等,在高中低3种负荷阶段下对某600MW机组进行了相关的调整试验,获得了与NOx排放特性和锅炉热效率相关的特性数据,基于改进的支持向量机建立了以负荷、煤质、总风量、总燃料量、二次风风门开度、燃尽风风门开度、风箱和炉膛差压以及燃烧器摆角等变量为输入,NOx排放量为输出的排放特性模型。预测结果表明该模型预测准确性较高,泛化能力较强。(4)为处理负荷分配优化问题中的多机组和操作禁区问题,引进了改进的粒子群算法,算例表明改进算法可以给出更优秀的分配方案；提出了飞灰含碳量循环确定法,从而改进了基于实时数据的煤质计算方法,在此基础上利用机组的运行数据计算相应的煤质；在考虑煤质变化、环境变化、设备老化以及运行水平等不受控制的因素的基础上,提出了基于改进支持向量机的机组煤耗特性预测模型,并通过算例验证了该模型良好的预测性能；基于所建立的煤耗特性模型,进行了某电厂的负荷分配优化,验证了煤耗特性模型的可行性和有效性。(5)对NOx减排优化模型进行了深入的研究,在考虑煤质变化以及机组负荷输出的基础上,改进了NOx减排优化模型,实例分析验证了该模型的有效性和可靠性；基于改进的NOx减排优化模型,利用改进的粒子群算法,研究了NOx减排优化问题,并和相关的典型算法进行比较。利用算例验证了改进算法的良好性能,优化结果与抑制NOx生成的原则具有一致性。
[Abstract]:High energy consumption and heavy pollution are important factors that restrict the development of thermal power generation. Under the premise of ensuring safe production and meeting the demand of load, the coal consumption and pollutant discharge of the power plant should be reduced as much as possible. The optimization method based on the modern intelligent algorithm is an effective way to improve the operation economy of the power plant and reduce the pollutant discharge. Therefore, a wide range of power and technology workers are required to study the intelligent algorithms and the power plant problems themselves. Based on the comprehensive summary of the development of particle swarm optimization (PSO), load distribution optimization and the optimization of NOx emission reduction, the corresponding research has been carried out and some useful results have been obtained. The main contents of this paper are as follows: (1) In view of the precocious problem and the convergence problem of the particle swarm optimization, the orthogonal experiment design learning strategy is introduced and improved, the valuable information is extracted from the group optimal particle and the sub-optimal particle, and the new guiding particles are formed; In order to enhance the ability of the best particle to jump out of the local best, the simulated annealing search strategy is introduced and improved, and the simulated annealing search is carried out on the group's optimal particles with a certain probability. The improved algorithm is tested on the standard test function. The results show that the improved algorithm is improved in the aspects of average value, standard deviation, number of times of evaluation, success rate and successful performance. (2) in order to overcome the precocious problem of the search, according to the diversity of the particle swarm, the adjustment of the inertia weight is guided, a new method for mapping the diversity and the inertia weight is proposed, The convergence of the improved algorithm is theoretically demonstrated by the stability criterion of the linear system and the probability theory. T-test and Wilcoxon test show that the improved algorithm is superior to the previous algorithm, and the improved algorithm is superior to the typical particle swarm algorithm and some other typical algorithms in the aspects of average value, standard deviation, number of times of evaluation, success rate and success performance. (3) In order to improve the accuracy of the modeling prediction, the kernel parameters and the penalty parameters of the support vector machine are optimized by the improved particle swarm optimization algorithm; the integrated support vector machine model for processing the large samples is introduced, and the problem of the parameter selection of the sub-support vector machine is improved; According to the simulation test of different loads, the design data of the combustor and the running characteristics of the boiler, a 600MW unit has been adjusted and tested at the low 3 load stage of the high school, and the characteristic data related to the NOx emission characteristics and the boiler thermal efficiency are obtained. Based on the improved support vector machine, the emission characteristic model is established based on the variables of load, coal quality, total air volume, total fuel quantity, secondary air-wind door opening degree, burn-out wind-air door opening degree, air box and furnace difference pressure and burner swing angle as input, and the NOx emission amount is the output. The prediction results show that the prediction accuracy of the model is high and the generalization ability is strong. (4) In order to solve the problem of multi-unit and operation restricted area in the optimization problem of load distribution, an improved particle swarm algorithm is introduced, which shows that the improved algorithm can give a better allocation scheme, and a method for determining the carbon content of the fly ash is put forward. so that the coal quality calculation method based on the real-time data is improved, the corresponding coal quality is calculated on the basis of the operation data of the unit, The prediction model of the coal consumption characteristic of the unit based on the improved support vector machine is put forward, and the good prediction performance of the model is verified by the calculation example; based on the established model of the coal consumption characteristic, the load distribution optimization of a certain power plant is carried out, and the feasibility and the effectiveness of the coal consumption characteristic model are verified. (5) An in-depth study of the NOx emission reduction optimization model is carried out, on the basis of considering the change of coal quality and the output of the unit load, the optimization model of NOx emission reduction is improved, and the effectiveness and the reliability of the model are verified by the example analysis; and on the basis of the improved NOx emission reduction optimization model, An improved particle swarm algorithm is used to study the optimization of NOx emission reduction and to compare with the relevant typical algorithms. The good performance of the improved algorithm is verified by the calculation example, and the result is consistent with the principle of inhibiting the generation of NOx.
【学位授予单位】：东南大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TM621;TP18

【相似文献】