【作者】 廖峰；

【导师】 姚建刚；

【作者基本信息】 湖南大学 ， 电力系统及其自动化， 2012， 硕士

【摘要】 母线负荷预测的结果是实现调度精益化管理的基础，是开展安全校核和阻塞管理、电力电量平衡、检修计划编制等工作的前提。开展母线负荷预测能够实现电网安全校核方式的转变，提高日前安全校核和阻塞管理的精确性，提升电网在线安全稳定分析水平。本文通过分析湖南电网220kV母线负荷特性，按照其日负荷曲线的不同，采用模糊C均值聚类的方法对母线进行了分类。之后，以长沙地区母线为进一步的分析对象，按聚类类型选取各类的典型母线进行详细、深入的分析。根据各类型母线的影响因素不同，分别从母线与气象因素的相关性、日类型对母线负荷的影响等方面有针对性地展开分析，总结出各类母线的具体影响因素与负荷特点，为下一步的预测提供参考。根据母线负荷与系统负荷的差异——母线负荷基数小、易受气象要素变化影响等，本文在充分考虑气象要素、日类型、小电源等因素对母线负荷预测的影响下，提出了基于改进灰色模型的母线负荷预测方法。该方法采用指数加权法对原始母线负荷序列进行改造，减弱异常值的影响，既可以充分利用样本数据中的有用信息，又大大减少了样本数据的随机性，强化原始数列的趋势。对于受气象因素和日类型影响较大的母线，建立日特征向量后利用灰色关联度理论寻找若干最优相似日，以选取日的母线负荷作为历史负荷样本。而对影响较小的母线则根据近大远小的规律就近选取负荷样本，采用加权移动平均的方法，对较近的负荷样板附以较大的权重进行预测。通过算例表明了分类处理的合理性和必要性，证明了作者所提出的方法具有较高的预测精度。然而，各级变电站的母线数量繁多，且负荷特性差异较大，预测人员的工作量大，难以保证预测精度。本文在前文研究的基础上，结合开展母线负荷预测工作的目的、母线的实际情况，开发了支持安全校核的集中式一体化母线负荷预测系统。系统基于.NET平台和Microsoft SQL Server数据库，采用多层体系Browser/Server（B/S）结构，以原有的负荷预测系统为依托，系统采用先进的预测策略，在提高预测准确率的同时，为发电计划安全校核与调度的精细化管理提供数据支持和技术保障。既调动各级负荷预测人员的积极性，保证了负荷人员的工作效率，又提高了母线负荷预测的准确率。更多还原

【Abstract】 Bus load forecasting results are the realization of the foundation of leanoperation management, and are the premise of carrying out security check andcongestion management, electric power and energy balance and maintenanceplanning. Carrying out bus load forecasting can change the way of power gridsecurity check, and improve the accuracy of the security check and the congestionmanagement, together with the grid online security and stability analysis level.This paper analyzed the220kV bus load characteristics in Hunan provincePower Grid, and put them into different categories by using fuzzy C-clusteringmethod. Then the paper selected classical buses from each category within Changshaarea and studied them in detail. The correlations between bus load and such factorsas weather, the types of days were considered, and corresponding loadcharacteristics are noted for further reference.With the meteorological and day-type factors thoroughly considered, this paperproposed a new bus load forecasting method based on improved gray model. Thisnew method used weighted exponential method to transform the original bus loadsequence so that we could weaken the impact of outliers, fully utilize inherent usefulinformation, reduce the randomness of the sample data and strengthen the originalsequence trend. Because the buses severely subjected to the meteorological andday-type factors, this paper selected several most similar day loads as historicalsamples by building classic day feature vector. While there was a weak affection ofthese factors, this paper, by applying the weighted moving average method, usedsuch a way so that recent day loads samples were given a bigger weights and oldersamples smaller weights. Example had shown that the classification was reasonableand necessary, and of high prediction accuracy.However, the buses of each substation are numerous and the load characteristicsare different from each other. Therefore, the workload of prediction is heavy and theaccuracy of the prediction is very low. In this paper, a centralized bus loadforecasting system was developed which supports the security checking of powergeneration program management according to the practical situation of bus load andthe expectations of bus load prediction. The software was developed withMicrosoft.NET platform and Microsoft SQL Server database, using the most advanced and sophisticated multi-layer system of Browser/Server (B/S) structurewith the basis of the existing load forecasting system. The centralized integratedsystem consists of bus load forecasting unit, load statistics and analysis unit,prediction and assessment-based assessment unit, data collection unit, forecastingdata reporting unit, basic information management and systems managementprocesses unit which provides security checking for generating schemes andscheduling fine management to provide support of data and technical protection.Advanced prediction strategy was proposed which can improve the predictionaccuracy.更多还原