【作者】 聂秋平；

【导师】 吴敏；

【作者基本信息】 中南大学 ， 计算机应用技术， 2011， 博士

【摘要】 加强能源管理是实现循环经济的重要手段。钢铁企业的能耗费用约占工业部门能耗总量的15%-20%,高能耗造成对环境的污染和经济效益的负面效应长期以来一直是我国钢铁企业所面临的重大难题之一。能源管理模式分散粗放、能源计划与平衡安排不当、调度系统简单低效等问题,已成为我国钢铁企业低能源利用率、高能耗的重要原因。针对上述问题,本文结合某大型钢铁联合企业能源中心建设项目,从钢铁企业能源系统分析、能源产消预测、能源实绩平衡和能源优化调度模型及其应用等方面展开研究,主要研究内容和创新性研究成果如下。(1)指出了钢铁企业能源管理所普遍存在的主要问题分析了典型钢铁联合企业的烧结、焦炉炼焦、高炉炼铁、转炉炼钢、连续铸造和轧钢等主要生产工序所使用的煤气、电力、水、水蒸气、氧、氩和氮气等能源的分布状况以及主要能源的管理流程,指出能源产消量的预测、能源的实绩平衡与数据校正、能源的计划管理与优化调度等问题是当前钢铁企业能源管理工作中所普遍存在和亟待解决的研究问题。(2)能源产消特性分析与不同特性下的产消预测模型基于煤气和电力两种主要能源,结合生产工艺分析和灰色关联度分析方法,分析了钢铁企业主要生产工序及各工序主要能源用户的能源产消特性,并根据不同的产消特性,建立不同的能源产消预测模型对能源的产生或消耗量进行预测。针对产消量基本处于一定范围内的用户,建立基于生产计划的预测模型,以非能源的计划产量作为依据进行预测；针对产消量没有明显的规律,存在随机性、灰色特性,且与其影响因素之间呈非线性关系的用户,建立基于灰色RBF神经网络模型的预测模型,将与产消量相关联的数据进行灰色累加处理后,作为模型的训练数据,以时间序列作为模型输入进行预测,并以预测误差作为反馈对神经网络结构进行修正；针对与其影响因素之间存在明显规律或线性关系的用户,建立多层递阶回归分析预测模型,将线性回归方法与多层递阶方法相结合,将预报对象看成是随机动态的时变系统进行预测；针对短时间内呈线性变化,长时间内呈现连续性和周期性变化的用户,建立基于时间序列的自回归滑动平均预测模型,利用前段时间若干能源产消量的线性组合对后段时间进行预测。工业现场数据的仿真实验表明,所建立的四类预测模型对于具有不同产消特性的用户具有较好的针对性和适应性,预测精度较高。(3)能源数据校正与煤气实绩平衡策略基于所提出的产消预测模型,提出了一种能源数据校正方法,以产消预测模型得出的预测值为参考值,以校正值与该预测值之间差值的平方和最小为目标函数,以能源平衡方程作为约束条件,建立优化模型,通过求解模型最优解,得到能源数据的校正值,并采用基于污染正态分布的过失误差诊断方法对能源计量数据中的过失误差进行诊断,对含有过失误差数据再重新求解校正值。过失误差的检测,能为检测设备的诊断提供指导。利用该数据校正方法,针对煤气这一钢铁企业最重要的能源,设计了一种煤气自动实绩平衡策略,由职能部门设置补偿参数,煤气用户设置仪表停计时间,平衡管理员设置平衡的运行参数,以测量数据的校正值代替原始测量值,实现了煤气系统的自动实绩平衡。工业生产过程数据的仿真实验表明,能源数据校正方法合理可行,煤气实绩平衡策略可提高煤气平衡的科学性和自动化程度。(4)两种煤气优化调度模型分析了企业内部能源系统的网络结构,推导了基于数学规划的能源优化调度总体模型的数学表达式。针对煤气系统,提出基于单元分类的和基于产消预测的煤气优化调度模型。1)基于单元分类的煤气调度模型根据各煤气用户在生产过程中的作用与煤气产消方式的不同对用户进行分类,以统一的调度目标和不同类别的约束方程对各用户煤气进行优化调度。模型主要从宏观角度着眼于钢铁企业全流程,可以适用于所有的用户；2)基于产消预测的调度模型利用煤气产消预测模型的预测值,然后使用能源总体优化调度总体模型对预测结果进行相应修正,并作为优化调度值。模型增加了对关键影响因素的考虑,调度结果具有相对较高的精确度,但仅适用于具有预测模型的用户,且受预测精度影响较大。工业过程数据仿真结果表明,调度模型可显著减少煤气的放散量。(5)能源产消预测与优化调度模型在能源管理系统中的应用与实现使用Windows平台,Visual studio开发环境,Oracle数据库,基于所提出的能源产消预测模型和优化调度模型,设计实现了煤气自动实绩平衡系统和煤气优化调度系统,并应用于论文背景钢铁企业,作为其能源管理系统的一部分。两系统投入运行后,可以在5分钟内实现按班为最小周期的全厂煤气的自动实绩平衡,可以在2分钟内自动生成煤气的班调度计划；有效提高了实际平衡的效率和煤气利用率,经济效益显著。更多还原

【Abstract】 Strengthening energy management is an important means to achieve the circular economy. The energy consumption cost of iron and steel enterprise takes up approximately 15%-20%of the total energy consumption cost of industrial sectors. The high energy consumption results in environmental pollution and negative effects on economic benefit, which have been a difficult problem for the iron and steel enterprises. The mains factors that result in the high energy consumption include the confused energy management, improper energy arrangement, inefficient energy scheduling and so forth. This paper aims to solve these problems based on the energy center construction project of a large-scale steel joint enterprise. The energy system, energy product-consume prediction, energy performance balancing, energy scheduling models and its applications are analyzed and studied. The main original research achievements are as follows:(1) Indicate the main problems existing in energy management systemThe distribution and management process of all kinds of energy such as gas, power, water, vapor, oxygen, argon and nitrogen consumed in the main production processes of sintering, coke oven coking, blast furnace ironmaking, converter steelmaking, continuous casting and rolling in a typical steel joint enterprise have been analyzed. It points out that energy product-consume prediction, energy performance balancing and energy scheduling remain to be improved in the current energy management.(2) Energy product-consume characteristic analysis and prediction models based on different characteristicConcerning the two main energy resources, gas and power, and combining the process characteristics analysis with grey relational analysis, the energy consumption characteristics of the users in the main production processes are analyzed. Then, the energy product-consume prediction models are developed according to the product-consume characteristics. For the users whose production and consumption are within a certain scope, the prediction model based on the production plan is built. It forecasts based on the planned production of non-energy. For the users whose production and consumption is of randomness, grey characteristics and has nonlinear relation with its influencing factors, the prediction model based on the RBF neural network is constructed. It adopts the grey accumulated data as the trained data, the time series as the input and the prediction error as the feedback to modify the neural network structure. For the users whose production and consumption are of obvious or linear relation with its influencing factors, the prediction model based on the multi-layer hierarchical regression is built. It combines the linear regression with multi-layer hierarchical, and takes the predicted object as the random dynamic time-varying system. For the users whose production and consumption are of linear change in a short time, but of continuity and cyclical change in a long time, the forecasting model based on the auto-regressive and moving average are developed. It predicts the changes in a latter period with the linear combination of energy product-consume in a former period.The results of simulation based on the practical industrial data show that the four models are of good pertinence, flexibility and accuracy.(3) Energy data reconciliation and performance balancing strategyBased on the prediction models, an energy data reconciliation method is proposed. It adopts the predicted value as the reference value, the balance equation and reconciliation range as the constraint condition, and sets the objective function as minimizing the sum-of-square of differential between the rectified and predicted value to construct the optimization model. The rectified value of energy data is obtained according to the optimal solution of the optimization model. Meanwhile, the method based on the polluted normal distribution gross error diagnosis is adopted to diagnose the error in the energy metrological data. The rectified value of the erotic data is re-solved. A gas automatic performance balancing strategy has been designed based on the data reconciliation method. The functional departments set the compensation parameters, gas users set the instrument stop time, and administrators set the balance operation parameters. Rectified value of metrological data replaces the original measured value to achieve automatic performance balancing of gas system. The results of simulation based on the practical industrial data show that the data reconciliation model and balance strategy are applicable and could improve the scientific and automation level of gas balance.(4) Two energy optimal scheduling modelsBy analyzing the energy network structure, the expression of energy optimal scheduling based on the mathematical programming is deduced. Then, the gas scheduling models based on unit classification and product-consume forecasting are proposed.1) The gas scheduling model based on unit classification classifies the gas users into different units according to their roles and functions in the production process. The optimal scheduling is achieved with unified scheduling goal and different constraint conditions. The model focuses on the entire process from the macroscopic angle and could be used by all gas users.2) The scheduling model based on product-consume prediction adopts the predicted value of the gas product-consume prediction model, fixes the value with the overall energy optimal scheduling model, and sets the fixed value as optimal scheduling value. This model is of higher accuracy by taking the key factors into consideration. However, it can only be used by the users with forecasting model and is influenced by the predicting accuracy. The results of simulation based on the practical industrial data show that the two models could significantly reduce the amount of radiation gas.(5) Implementation and application of energy product-consume predicting and optimal scheduling model in energy management systemBased on the energy product-consume predicting model and optimal scheduling model, the gas auto performance balancing and optimal scheduling system using Windows XP platform, Visual studio integrated development environment and Oracle database is built. It is applied in a iron and steel enterprise as a part of the energy management system, and could achieve auto performance balance of the gas of the whole factory by taking a class as the minimal cycle within five minutes and could generate class scheduling plan automatically within two minutes. It effectively improves the efficiency of the performance balance and the gas utilization, and remarkably increases the economic benefit.更多还原