基于云计算的组合短期负荷预测方法研究

发布时间：2018-04-16 05:13

本文选题：负荷预测 + 云计算　；参考：《兰州理工大学》2017年硕士论文

【摘要】：近年来,随着计算机技术普遍应用,智能电网迅速发展。电力部门工作人员为保证系统安全、经济的运行,对短期负荷预测结果的稳定性、准确性、高效性提出了更高的要求。目前短期负荷预测的研究方向主要集中在对预测模型的整体优化上,这在一定程度上提高了负荷预测的工作速度和计算精度。但是这些预测方法大多建立在对影响因素的整体分析之上,对各个因素的自身特性考虑的不够全面,导致预测模型的准确性难以进一步提高,通用性较差。本文根据各个影响因素相关性的不同,对具有更高预测效率的组合预测模型进行了相关研究。首先,本文分析了短期负荷预测的实际应用背景,对当前该领域内的国内外研究现状进行了归纳总结,在充分对比多种传统智能预测算法的优缺点之后,对未来短期负荷预测的研究重点进行分析。本次选用浙江省某地区的历史数据作为训练样本和预测样本,对预测地区的负荷特性、经济特性、气象因素等进行了深入分析,针对原始数据自身存在的不足,对其进行数据预处理,采用双向比较法筛选修复问题数据,增强了预测结果的可靠性和准确性。其次,为了对预测过程进行精细化研究,本文对影响负荷大小的各个因素进行了确定性相关的分类,利用细菌觅食算法优化极限学习机预测模型对确定性相关影响因素负荷进行预测,利用云模型优化核极限学习机预测模型对非确定性相关影响因素负荷进行预测,通过对两种预测模型的预测结果加权求和,得到最终的负荷大小。最后,由于该组合预测模型运算复杂,大大增加了运算的难度,为了解决单机计算资源不足的问题,本文引入云计算对组合预测模型进行并行化改造,提高了预测模型的大数据处理能力,增强了这一新模型的实际应用效果。结果发现,相比于传统预测方法,本文通过引入云模型优化核极限学习机预测模型,增加了对非确定性相关影响因素的考虑范围,提高了预测结果的准确性,使预测精度提高了0.23%。通过引入云计算,提高了预测模型的并行计算性能,使单次预测时间减少了大约900s,加快了计算的速度,提高了工作人员的工作效率。
[Abstract]:In recent years, with the widespread application of computer technology, smart grid has developed rapidly.In order to ensure the safe and economical operation of the system, the power department staff put forward higher requirements for the stability, accuracy and efficiency of the short-term load forecasting results.At present, the research direction of short-term load forecasting is mainly focused on the overall optimization of forecasting model, which improves the working speed and calculation accuracy of load forecasting to a certain extent.However, most of these prediction methods are based on the overall analysis of the influencing factors, and the characteristics of each factor are not fully considered, resulting in the accuracy of the prediction model is difficult to further improve, and the generality is poor.In this paper, a combination forecasting model with higher prediction efficiency is studied according to the different correlation between different factors.First of all, this paper analyzes the practical application background of short-term load forecasting, summarizes the current domestic and foreign research status in this field, after fully comparing the advantages and disadvantages of many traditional intelligent forecasting algorithms.The emphasis of future short-term load forecasting is analyzed.In this paper, the historical data of a certain area of Zhejiang Province are selected as training samples and forecasting samples, and the load characteristics, economic characteristics and meteorological factors of the predicted area are analyzed in depth, aiming at the shortcomings of the original data itself.The data preprocessing and bidirectional comparison method are used to screen the restoration data, which enhances the reliability and accuracy of the prediction results.Secondly, in order to study the forecasting process in detail, this paper classifies the factors that affect the load size by deterministic correlation.The bacterial foraging algorithm was used to optimize the prediction model of deterministic factors, and the cloud model was used to predict the load of non-deterministic factors.The final load size is obtained by weighted summation of the forecasting results of the two forecasting models.Finally, due to the complex operation of the combined prediction model, it greatly increases the difficulty of calculation. In order to solve the problem of insufficient computing resources, this paper introduces cloud computing to transform the composite prediction model into parallel.The ability of big data to deal with the prediction model is improved, and the practical application effect of the new model is enhanced.The results show that compared with the traditional prediction methods, the cloud model is introduced to optimize the prediction model of the kernel limit learning machine, which increases the scope of consideration of the non-deterministic related factors and improves the accuracy of the prediction results.The prediction accuracy is improved by 0.23.By introducing cloud computing, the parallel computing performance of the prediction model is improved, the time of single prediction is reduced about 900s, the speed of calculation is accelerated, and the work efficiency of staff is improved.
【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM715

【参考文献】