電気学会論文誌B(電力・エネルギー部門誌)
Online ISSN : 1348-8147
Print ISSN : 0385-4213
ISSN-L : 0385-4213
特集論文
状態空間モデルを用いたJEPX価格モデリングの基礎検討
大藤 建太兼本 茂
著者情報
ジャーナル フリー

2008 年 128 巻 1 号 p. 57-65

詳細
抄録
Time series models are popular in attempts to model and forecast price dynamics in various markets. In this paper, we have formulated two state space models and tested them for its applicability to power price modeling and forecasting using JEPX (Japan Electric Power eXchange) data. The state space models generally have a high degree of flexibility with its time-dependent state transition matrix and system equation configurations. Based on empirical data analysis and past literatures, we used calculation assumptions to a) extract stochastic trend component to capture non-stationarity, and b) detect structural changes underlying in the market. The stepwise calculation algorithm followed that of Kalman Filter. We then evaluated the two models' forecasting capabilities, in comparison with ordinary AR (autoregressive) and ARCH (autoregressive conditional heteroskedasticity) models. By choosing proper explanatory variables, the latter state space model yielded as good a forecasting capability as that of the AR and the ARCH models for a short forecasting horizon.
著者関連情報
© 電気学会 2008
前の記事 次の記事
feedback
Top