Short-term load forecasting is the basis for the safe operation of power systems. The accuracy of forecasting will have a direct impact on the load distribution of the entire power grid. There are many factors affecting the load, while the method based on similar historical days' data can fully consider these factors. It forecasts load by selecting similar historical days' data and then obtaining a weighted average from them. However, in previous studies, the weights of similar days selected are not obvious, which cannot reflect the importance of the most similar days, and results in a big forecasting error. In this paper, the weight of the most similar days is increased so as to embody the influence of the most similar days on the forecasting load,and then weighted average of the selected similar days is used to predict the load of 96 periods. At the same time, it makes an analysis on how to select similar days and situations without similar days. Moreover, it forecasts the load of a certain week of June in Hainan, and the forecasting results are more desirable than previous methods.
[1]
T. Funabashi,et al.
One-Hour-Ahead Load Forecasting Using Neural Networks
,
2002
.
[2]
Liu Yi.
Impact Analysis of Hourly Weather Factors in Short-Term Load Forecasting and Its Processing Strategy
,
2006
.
[3]
Moon-Hee Park,et al.
Short-term Load Forecasting Using Artificial Neural Network
,
1992
.
[4]
Y. Shimakura,et al.
Short-term load forecasting using an artificial neural network
,
1993,
[1993] Proceedings of the Second International Forum on Applications of Neural Networks to Power Systems.
[5]
Hu Ziheng.
Method to select similar days for short-term load forecasting
,
2004
.
[6]
Chen Zhi.
PRINCIPLE COMPONENT ANALYSIS OF DATA PROCESSING FOR SHORT\|TERM LOAD FORECASTING
,
2000
.