Forecasting multiple-well flow rates using a novel space-time modeling approach

A different approach based on existing spatiotemporal models is developed in this research for forecasting the production flow rates of a group of adjacent wells in a reservoir. Unlike the previous single-well analysis techniques, this new methodology tries to forecast the flow rates of a number of producing wells located in a certain neighborhood. It also takes into account the variety of flow rate patterns that may take place during the production period of the wells. Regarding the time-series nature of flow rate data and the spatial relationships between multiple wells in a proximity, the Space-Time Auto-Regressive Moving-Average technique, abbreviated as STARMA, is borrowed to employ in this research to capture the spatial inter-relationships of the wells and model the flow rate data of multiple wells simultaneously. Partitioning the field into different groups of adjacent wells based on the similarity of flow rate patterns is also recommended using the spatial clustering algorithms as an enhancing part of the proposed methodology. The capability of the proposed approach in forecasting the flow rates of multiple wells is verified using the real data of a case study including 38 active oil producing wells located in a reservoir in South West Iran. The forecasts generated by the proposed method are finally compared with the predictions obtained by the current ARIMA techniques. Based on findings of this research, the proposed space time modeling of multiple well flow rates helps in estimating the more accurate forecasts, when the single well analysis techniques fail in predicting the real trend of future flow rates.