|Prediction of Representative Waveform of Surplus PV Power Based on Cluster Analysis of Power Demand and PV Output|
|Akito Kawanobe1, Shinji Wakao1, Tomoya Taima2, Norihiro Kanno2, Hiroyuki Mabuchi2
1Waseda University, Shinjyuku-ku, Japan
/2Tokyo Electric Power Company Holdings, Inc., Yokohama-shi, Japan
In some Japanese consumers with photovoltaic (PV) power generation, the valid 10-year period of the Feed-in tariff program started in 2009 will be over in 2019. As a result, they will expect the merit of self-consumption of PV power from the economic point of view instead of selling it and will gradually introduce storage batteries for surplus PV power. At present, the demand curve of the consumer has already changed apparently due to the reversal power flow from PVs massively introduced. In the future, additional uncertainties such as storage battery operation in consumers will result in the more difficulty in predicting the apparent demand of consumer and operating the distribution system. Therefore, the precise prediction method of the apparent demand curve will become more important to ensure stable supply of electric power.
With the above background, in this paper, we perform the cluster analysis of measured data obtained by smart meter, i.e., actual demand, PV output, and surplus PV power, to reveal their representative waveforms. We carry out the clustering of the above data normalized so that their average equals to 0 and the variance to 1 to focus on how they fluctuate with time. The fluctuation of PV output depends on the daily weather condition. On the other hand, the variation of actual demand is subject to the type of power contract. Finally, we propose a convenience reproduction method of the surplus PV power waveform, i.e., apparent demand seen from distribution system, by linearly combining the actual demand and PV output representative waveforms by each consumer and report high accuracy in reproduction.
It is considered that the proposed method can be useful for estimating the apparent demand curve of the entire distribution system, which will contribute greatly to the stable operation of the system under the condition of mass introduction of PVs with batteries in the future.
Area: Sub-Area 11.1: Solar Resource Assessment and Variability Modeling