基于相似时间序列检索的超短期负荷预测

针对目前超短期负荷预测算法存在的预测精度不稳定、实时性能不强等问题，从时序数据挖掘的重要方法——相似时间序列的检索出发，结合负荷自身的周期性变化规律，提出了一种新的超短期负荷预测方法。该方法具有简单实用的坏数据处理机制；通过扩展负荷序列相似的概念有效地增加了预测样本的数量，提高了预测样本的质量；对预测值的加权处理抵御了单样本预测带来的风险，使预测的精度稳定在一个较高水平。实际应用结果表明，该方法的预测精度高、稳定性强，能较好地满足电力系统各方面的需求。

Ultra-Short Term Load Forecasting Based on Similarity Search in Time-Series

In current ultra-short term load forecasting algorithms,such disadvantages as unstable forecasting accuracy, weak real-time performance and so on,are to be avoided. Based on the important method of time-series data mining,i.e., the similarity search in time-series and combining with periodical variation law of power load itself,a new ultra-short term load forecasting method is proposed.The proposed method possesses simple and practical bad data processing mechanism;the number of forecast samples is effectively increased by use of the concept of the similarity of load series expanded;the risk brought about by single sample forecasting is reduced by weighted processing of forecasted results and make the forecasting accuracy stable in a higher level. Applying the proposed method to an actual power system,it is shown that the proposed method possesses the features of high forecasting accuracy and strong stability,so it can satisfy the requirements of power system in various aspects well.