【作者】 张楠；

【导师】 钱江波；

【作者基本信息】 华北电力大学 ， 动力机械及工程， 2018， 硕士

【摘要】 当前,随着我国经济的突飞猛进以及人民生活水平快速的提高,对能源的需求日益增加,随之而来的能源资源供需矛盾与生态环境污染问题都已经成为近年来我国面临的重大挑战。针对这些问题,我国开始实施全民节能行动,将节能减排作为重点工作。而身为能源消耗大户的火力发电厂,是开展节能降耗工作的重要关注对象。在众多指标中,煤耗是最能体现火电厂节能效果的指标。现今电厂煤耗主要依靠能量平衡的正平衡法与反平衡法进行计算,不仅计算过程繁琐复杂,而且还需要对计算结果进行大量的后期修正,无法真实的反映电厂煤耗。本文通过将支持向量机算法创新性地应用于煤耗预测中,不仅可以为火电机组经济性诊断提供新的思路和方法,也对解决电力系统中面临的其它具有更复杂条件的诊断问题有非常重大的启发意义。首先由机器学习与统计学习理论入手,从线性回归和非线性回归两个方面介绍支持向量机回归问题,并进一步分析最小二乘支持向量机算法的原理。其次,选取部分对发电煤耗率影响较大的运行参数进行了具体研究,并应用最小二乘支持向量机方法进行预测模型建立;该模型可以依托部分训练样本建立判别函数和相应的非线性变换,把机组煤耗与其影响因素之间复杂的非线性关系较好地表现出来。在此基础上,选取某电厂600MW机组为实例,结果表明该模型训练时间短、收敛性好,对于小样本有较强的处理能力,在样本数较少的情形中也可以很好地进行统计学习。最后,基于该LS-SVM模型开发了“基于精细化运行的火电机组运行指导系统”,系统界面简单、使用方便,能够提前为火力发电厂经营管理者提供详细的关于机组煤耗的信息库,同时基于耗差分析理论对机组参数进行耗差计算并提供运行指导意见,对提高电厂管理的智能化水平具有重大的现实意义。更多还原

【Abstract】 With the rapid development of the economics and people’s living standards in China,the demands for energy are expanding increase,therefore,the contradiction between supply and demand of energy resources and the pollution of ecological environment has become a major challenge in recent years.To solve the problems above,China began to implement the national energy saving action,and put energy saving and emission reduction as the key work.The thermal power plant,which is a large energy consumer,is an important concern for energy saving and consumption reduction.Among the many indexes,coal consumption is the best indicator of energy saving effect in thermal power plant.The coal consumption of the power plant is mainly calculated by the positive balance method and the counter balance method,which involves a complex calculation procedure and requires a large number of later corrections for the results,cannot truly reflect the coal consumption of the power plant.In this paper,the support vector machine(SVM)algorithm is innovatively applied to the coal consumption prediction,which can not only provide new ideas and methods for the economic diagnosis of thermal power units,but also has great enlightening significance to solve the other complicated diagnosis problems.First,starting from machine learning and statistical learning theory,we introduce the support vector machine regression problem from two aspects of linear regression and nonlinear regression,and further analyze the principle of least squares support vector machine(LS-SVM).Secondly,we select part of operating parameters which have great influence on the coal consumption rate and carry out a specific study,then establish the prediction model using the LS-SVM method.The model can build discrimination function and corresponding nonlinear transformation based on some training samples,showing the complex nonlinear relationship between coal consumption and its influencing factors.On this basis,taking a 600 MW unit of a power plant as an example,the results show that the model not only has short training time and good convergence,but also has strong handling capacity for small samples,thus statistical learning can be done well in the case of a small number of samples.Finally,we develop ‘the operation guidance system of thermal power unit based on fine operation’ based on the LS-SVM model,which has new and friendly interface,operating briefly and conveniently.The system can provide detailed information database for coal consumption of thermal power plants in advance,calculate the loss of the unit paiameters based on the energy loss analysis theory and provide guidance for operation,and have great practical significance to improve the level of intelligent management of power plants.更多还原