Time series analysis of JEPX spot price with the box-Jenkins method

Abstract

Following the examples of other countries, in April 2005 Japan launched wholesale electric power exchange operations as a primary item of system reform in line with electric liberalization. Only two years have passed since the initiation of these operations. However, in the summer of 2005, the surge in market prices was evident, which suggested that certain measures should be taken to confront potential market risks. Establishing a useful system for forecasting market prices through the modeling of price fluctuations in the wholesale electric market became essential. Currently, various price models are being proposed. Taking both the limited amount of data and the model's purpose into consideration, this study adopted the univariate time series model. We conducted a time series analysis on the open price indexes in the JEPX spot market with the Box-Jenkins method. Since a seven-day cycle can be observed in the data, we adopted the seasonal ARIMA model. In accordance with the procedures of the Box-Jenkins method, we determined the degree of the model's polynomial using the autocorrelation and partial autocorrelation of the data and estimated the parameters of the model with the maximum likelihood method. We conducted a forecast on next day JEPX spot market prices with this time series model and examined its validity and utility as a forecasting tool. Price forecasts made with this model require only a small amount of data and will save substantial analysis work. Consequently, this method is expected to be widely used by market participants as the reference data for their bid pricing.