基于人工神经网络的输变电工程造价预测研究

发布时间：2018-09-04 19:43

【摘要】：电网工程的造价是一个多变量、高度非线性的问题。过去对于输变电工程造价的预测主要靠在该领域用有多年实践经验的技术人员的实际分析和操作。但是当工程情况复杂多变时,很难通过经验估计得到单项工程可靠的结果用以指导工程造价控制。在变电工程中的单位容量造价和输电工程中的单位长度造价是两个投资方和施工单位最为关切的主要技术经济指标,也是输变电工程造价管理与控制的核心指标。因此,投资方和施工单位迫切需要一种理想的预测方法能够利用已建工程的历史造价资料,快速预测出新建电力工程的主要技术经济指标,以便合理制定建造方案,为电力工程建设争取主动时间,提高项目资金投入的审查效率和项目的质量,指导新电力建工程的造价。论文对输变电工程造价数据预处理技术进行研究,结合工程造价历史数据的具体特点,提出包括数据清洗、数据转换和数据约简等内容的输变电工程造价数据预处理方法,并且以电力输电工程为案例进行仿真,验证数据预处理方法的有效性。然后,针对输变电工程造价数据数据,提出一种易于操作、快速有效的输变电工程造价预测模型,即MEA-BP造价预测模型。其中,人工神经网络算法在小样本学习领域表现十分优越,也适合于对工程数目有限、影响因素颇多的输变电工程造价数数据进行学习。思维进化算法作为强大的参数优化算法,在模型中对BP人工神经网络参数进行优化。最后利用实例对其进行了仿真模拟。证明其预测模型在精度方面相比于传统的造价计量方法有了较大的提升,以电力输电和变电工程为案例的仿真表明,该系统可以稳定有效地实现工程造价管理,为工程建设的顺利实施提供技术支持。
[Abstract]:The cost of power grid engineering is a multivariable and highly nonlinear problem. In the past, the cost prediction of transmission and transformation projects mainly depended on the practical analysis and operation of technicians with many years of practical experience in this field. However, when the engineering situation is complex and changeable, it is difficult to obtain reliable results of single project by empirical estimation to guide the project cost control. The unit capacity cost and the unit length cost in the power transmission project are the main technical and economic indexes concerned by the two investors and the construction units, as well as the core indexes of the cost management and control of the transmission and transformation projects. Therefore, the investors and construction units urgently need an ideal forecasting method which can quickly predict the main technical and economic indexes of the newly built electric power projects by using the historical cost data of the existing construction projects, so as to reasonably formulate the construction plans. It can gain active time for electric power engineering construction, improve the efficiency of project capital investment and project quality, and guide the cost of new electric power construction project. In this paper, the preprocessing technology of transmission and transformation project cost data is studied, and the pretreatment method of transmission and transformation engineering cost data including data cleaning, data conversion and data reduction is put forward according to the specific characteristics of historical data of project cost. The simulation of power transmission project is carried out to verify the validity of the data preprocessing method. Then, based on the cost data of transmission and transformation project, a fast and effective cost prediction model of transmission and transformation project, namely MEA-BP cost prediction model, is proposed. Among them, the artificial neural network algorithm is very superior in the field of small sample learning, and it is also suitable for learning the cost data of transmission and transformation projects with limited number of projects and many influential factors. As a powerful parameter optimization algorithm, the mental evolution algorithm optimizes the parameters of BP artificial neural network in the model. Finally, an example is used to simulate it. Compared with the traditional cost measurement method, the prediction model is proved to be more accurate. The simulation results of electric power transmission and substation projects show that the system can realize project cost management stably and effectively. To provide technical support for the smooth implementation of engineering construction.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F426.61;TP183;TU723.3

【参考文献】