A dynamic short-term price forecasting framework for oilfield materials

This paper proposes an effective forecasting framework for the Oilfield Class-A materials, which take a large scale of proportion in the total purchasing cost. Based on the ARIMA model in time series analysis method, the dynamic forecasting framework is constructed to make short-time price prediction for Class-A materials in Oilfield, which only needs a small sample set to obtain high prediction accuracy. Considering the same kind of time series data usually having some fixed and similar features, the construction of the dynamic ARIMA model database in the framework can reduce the complexity of computing, by providing the next prediction step with reference historical models. According to the framework, in China Daqing Oilfield, the four places market-price of the Small Deformed Steel Bar (20-HRB335) in the Class-A materials from January to December in 2011 are dynamically predicted, including Tianjin, Shijiazhuang, Harbin and Chengdu. The accuracies of the prediction are no more than 0.96%, 0.69%, 0.55% and 0.34%, respectively. The result is given a high evaluation by Daqing Oilfiled Materials Corporation. The conclusion section points out that the forecasting framework has to call the ARIMA module repeatedly in a complete prediction process, which causes a large calculation, and the feedback mechanism of the ARIMA model database also needs a further improvement.