History Matching of the Valhall Field Using a Global Optimization Method and Uncertainty Assessment

Abstract

This paper provides a study of a history match on a complex reservoir model using a global optimization method. This is done by applying Evolutionary Algorithms to the problem of history matching. The results of the history match are then used to carry out an uncertainty assessment on variables of interest. The main parameters used in the history match included: horizontal permeabilities, porosities and vertical transmissibilities. This study also made use of methods for improving the convergence of the optimization cycle, which included using correlations, adopting a Bayesian approach and exploring the search space. The results obtained over the optimization cycle, are used to identify sensitivity parameters, correlations and parameter trends in a global search space. In addition the original manual history match was further improved by adopting a pressure match using an Evolutionary Strategy. Best matched cases were selected based on the global and partial objective values of each match. Predictions runs were performed in order to investigate the effect on the cumulative oil produced and the STOIIP. Finally an uncertainty assessment of the most recent history match was carried out using an experimental design matrix. The results of the experimental design were used to generate a proxy, which is used in a Monte Carlo simulation to develop P10/P50/P90 oil forecasts. Introduction The focus of this study is to assess the Valhall field reservoir simulation model using Evolutionary Algorithms and uncertainty assessments, in order to improve the manual history match of this model; this is done in turn to increase the reliability and confidence of the simulation model utilized by Total E&P Norge AS and its partners. The Valhall field is a complex field with rock compaction as its main energy drive. This contributes to many challenges which include the reduction of porosity and permeability with time and the continued change of the reservoir thickness due to compaction. The compaction can also affect different parts of the reservoir differently, for example the reduction in the pore volume of one region due to compaction can vary from other regions. This can cause many problems to the history matching process, in which changing a particular parameter can have opposite effects on a well or region, therefore, it is extremely difficult to history match this type of field. Other challenges include a variable production profile due to well instabilities, chalk production and downtime, which contribute to the history match and prediction complications. An Evolutionary Algorithm will be utilized in order to enhance the history match of this field. This algorithm will assist in exploring different methods and strategies that have never been tested on this field before. The algorithm will be used as part of the relatively new software, Multipurpose Environment for Parallel Optimization (Mepo®). Quantification of uncertainty is common practice and the most important parameters should be assessed over a range of uncertainty. Uncertainty assessment in the past has been slow and inefficient; it was normally done by varying one parameter at a time and running simulations, in order to assess the results[15]. One of the most important uncertainties is the oil production forecasts which are evaluated as the oil recovery factor or the cumulative oil production[14]-[21], another important parameter is the Stock Tank Oil Initial in Place. In order to increase the acceptance of the results from the history match, the uncertainty of the history match has to be quantified. Oil, water and gas production forecasts for economical decisions are normally carried out. Important decisions, such as well placement and locations are made with respect to the predictions conducted using the history matched reservoir model, and therefore, it is important to implement an uncertainty assessment to such models. The results from the optimized history match model are used to create the Experimental Design Matrix (EDM). The uncertainty assessment was carried out using MEPOrisk® software, still under development by Scandpower Petroleum Technology. Figure 1, shows the uncertainty assessment workflow carried out using MEPOrisk®.

A Systematic Approach to Uncertainty Reduction with a Probabilistic and Multi-Objective History Matching

“…Different approaches can be mentioned: the Stochastic Gaussian Search Direction (SGSD) algorithm (Li and Reynolds 2011), the evolutionary algorithm (Al-Shamma and Teigland 2006), the Bayesian optimization algorithm (Abdollahzadeh et al 2011), the ant colony optimization (Hajizadeh et al 2011), and the particle swarm optimization method (Martínez et al 2012). Most of these methods require a high number of numerical simulations, so, in order to make their application computationally feasible, they commonly simplify the physics of the reservoir model by representing them with proxy models.…”

Retaining geological realism in dynamic modelling: a channelized turbidite reservoir example from West Africa

“…Christie et al (2006) used neural networks in combination with the neighbourhood algorithm for history matching. Recently, a variety of commercial algorithms have been introduced in the literature to aid engineers and geoscientists in the history-matching process (Schulze-Riegert et al 2002;Bustamante et al 2005;Al-Shamma & Teigland 2006;Cullick et al 2006). The main theme of such algorithms is time savings through automation of tedious and repetitive tasks, e.g.…”

Desenvolvimento de indicadores para sistematização do ajuste de histórico de produção probabilístico

“…Exemplos de aplicação deste procedimento são encontrados nos trabalhos de Bissel (1997), Bennett & Graf (2000) e Jenni et al (2004). Uma aplicação de algoritmo evolucionário para obter vários modelos ajustados e avaliar incertezas na previsão de produção foi publicada por Al-Shamma & Teigland (2006).…”