Load forecasting based on short-term correlation clustering

• Published in: 2017 IEEE Innovative Smart Grid Technologies - Asia (ISGT-Asia) pp. 1 - 7
• Main Authors: Tao, Shun, Li, Yongtong, Xiao, Xiangning, Yao, Liting
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-12-2017
• Subjects: Clustering methods; Correlation; correlation analysis; Forecasting; Itemsets; load clustering; Load forecasting; Load modeling; short-term
• ISSN: 2378-8542
• DOI: 10.1109/ISGT-Asia.2017.8378416

Load forecasting is the basis not only of power system stable and safe operation, but also of power demand side intelligent electricity management. Short-term correlation analysis can be used to mine the electricity consumption of a period of time. The analysis of similar electricity consumption can improve the effect of load forecasting. Therefore, this paper proposes a load forecasting method based on short-time correlation clustering. First, a method is proposed to analyze the correlation matrix of power sequences, and then the effects of noise information of the correlation matrix is eliminated. Then, based on the correlation matrix, we group feeders with different electricity consumption by using the fuzzy C-means clustering algorithm. After each feeder is assigned to a special group, we forecast each group load based on artificial neural network. The load forecasting results from each group are summed to obtain total load forecast. Finally, the case studies with instance data verify the load clustering based on short- term correlation analysis for power sequences can improve load forecasting.