摘要
对历史负荷数据进行处理是提高电力系统负荷预测精度首先要解决的问题。脏数据处理的过程就是对于含有脏数据的负荷曲线模式的辨识过程。首先利用自适应共振网络(ART网)对日负荷曲线进行分类,确定出每一类负荷曲线的特征曲线;然后用超圆神经网络(CC网)对特征曲线进行脏数据辨识;最后利用扩展短期负荷预测方法对脏数据进行修正。对某市2002年8月份的数据进行脏数据辨识,结果证明所提出的模型对脏数据的平均检测率为92.11%,效果令人满意。采用该处理过的历史数据对某市2002年8月14日的负荷进行预测,结果表明,利用该方法处理后的数据进行负荷预测提高了负荷预测的精度。
Processing historical load data is first to solve the problem in improving the load forecasting accuracy in power system. The process of dirty data processing is the recognition process of load curve pattern containing dirty data. To begin with, self-adaptive resonance networks can be used to classify the daily load curves so as to determine the feature curves of each kind of load curve patterns; and then, the ultra-circle neural networks is used to recognize dirty data in the feature curves; and finally, the expanding short-term load forecasting approach is adopted to modify the dirty data. The dirty data recognition is made in the data of August in the year of 2002 of some city. The results prove that the average detection rate of dirty data by the suggested model is 92.11%0 so that the effect is satisfactory. Also, the processed historical data are used to forecast the load of August 14, 2002 in some city; and the results indicate that the data processed by this approach in carrying out load forecast can improve the accuracy of load forecast.
出处
《西安理工大学学报》
CAS
2007年第3期277-281,共5页
Journal of Xi'an University of Technology
关键词
数据预处理
负荷预测
模式识别
data pre-processing
load forecast
pattern-recognition
