Sand production in oil and gas wells can occur if fluid flow exceeds a certain threshold governed by factors such as consistency of the reservoir rock, stress state and the type of completion used around the well. The amount of solids can be less than a few grams per cubic meter of reservoir fluid, posing only minor problems, or a substantial amount over a short period of time, resulting in erosion and in some cases filling and blocking of the wellbore. This paper provides a review of selected approaches and models that have been developed for sanding prediction. Most of these models are based on the continuum assumption, while a few have recently been developed based on discrete element model. Some models are only capable of assessing the conditions that lead to the onset of sanding, while others are capable of making volumetric predictions. Some models use analytical formulae, particularly those for estimating the onset of sanding while others use numerical models, particularly in calculating sanding rate. Although major improvements have been achieved in the past decade, sanding tools are still unable to predict the sand mass and the rate of sanding for all field problems in a reliable form.

No abstract

We performed a comprehensive sand prediction study of several deep, HPHT wells within a large reservoir and the findings revealed that, for these wells, common criteria based on critical drawdown, minimum bottomhole pressure, depletion or fluid velocity failed to predict the outcome by a relatively large margin. All these wells were subjected to relatively high levels of drawdown and also very high fluid velocities and, with the exception of one well, none showed any sanding until water production was encountered. In this paper, we provide a rationale for why water can be highly effective in inducing sand production and we support our argument using advanced numerical modeling. This exercise also ranks the performance of some of the common tools and theories that are conventionally used for sand prediction. We also provide reasons why some of these models do not perform satisfactorily for the cases studied. The originality of the work is in demonstrating that prior to sand production, the dis-aggregated rock (i.e., individual sand particles) around the wellbore is basically held together by the capillary tension which is destroyed by water flow. While the capillary tension appears to be insignificant (as it is in the order of 1 psi or so), it provides a significant resistance against sand mobilization. The importance of this issue has been quantified using advanced numerical modeling. This concept is vastly different from the previous theories that propose water weakens the rock through chemical interaction or changing the relative permeability. Introduction While a great deal of work has been done in the general area of sand production1–15, approaches used to quantify the volumes of the produced sand have faced challenges in the validation process (Class A prediction). In fairness, it is difficult to firmly single out the deficiencies when predictions do not materialize in a consistent fashion as the quality of the input data, monitoring of sanding events as well as the assumptions and physics used for modeling sanding can all be potential culprits. Following up on this line of reasoning, it is not the intention of this paper to prove or refute any previous work done in sand production studies nor to show that our method is universally superior. The primary intention is to provide a deeper insight into the mechanisms of sanding, in general, and water-production induced sanding, in particular. We try to support the views presented using basic fundamentals along with field observations. Significance and Potential Applications Conventional sand production models2,3,15–17 predict the onset of sanding which in practice is presumed to signify large-scale sanding. This single case solution scenario does not give operators options to assess risks and benefits which is becoming increasingly more relevant under the currently optimized completion and production practices. In essence, operators would like to know, at any stage in a well's life, how much sand will be produced (rate and duration) for a given production strategy (e.g., maximum drawdown, effects of bean-up and shut-in cycles, impact of water). By better understanding the role of various variables one is enabled to choose the optimal completion method for the life of the well (which may exclude installation or deferring sand control measures) and quantify the impact of aggressive fluid production strategies in terms of volume and rate of sanding.

Summary This paper introduces a predictive tool that forecasts the drawdown associated with the onset of sanding as well as it predicts the sanding rate in real time. Experimental data on hollow cylinder samples (HCS) are used to support the validity of the numerical model. Experiments on hollow-cylinder synthetic-sandstone specimens were conducted, involving real-time sand-production measurement under various conditions. A numerical approach was used for simulating the experimental results. The material behavior was simulated using an elastoplastic stress-strain relationship. The model simulated the interaction between fluid flow and mechanical deformation of the medium in predicting sand production. The model simulated strain softening of the material accompanied with shear-bands formation as well as tensile failure. In the post-disaggregation phase, additional features were considered, including allowing for the removal of the disaggregated elements that have satisfied the sanding criteria and, consequently, making the necessary adjustments to the size and properties of the domain under consideration. The model can be used for time-dependent analysis of wellbore stability as it undergoes disaggregation and sand production induced by depletion, drawdown, and water cut. Such numerical tools can be used in designing the completion by identifying the critical operational conditions associated with severe sanding over the lifetime of the wellbore. The model showed a reasonable agreement with experimental results in terms of rock deformation and sanding rate. Further validation of the model against experimental and field data is necessary for its potential field applications. Introduction It is estimated that 70% of the total world's oil and gas reserves are found in poorly consolidated reservoirs (Bianco and Halleck 2001). Poorly consolidated formations are the most common solid producers. Several sand-production prediction methods have been proposed using geotechnical models. Existing models can effectively predict the onset of sand production and analyze cavity stability and rock failure; however, there still is room for improvement in predicting the volumetric sand production over the lifetime of the wellbore as a function of the completion strategy, drawdown, depletion, and water-cut. In the following, a brief description of the existing models is introduced. Modeling Strategies. Several analytical and numerical models have been proposed for the prediction of sanding (Risnes et al. 1982; Perkins and Weingarten 1988; Sanfilippo et al. 1995; Vaziri et al. 1997; Vaziri and Palmer 1998; Morita and Fuh 1998). Most predict only the onset of sanding (Sanfilippo et al. 1995; Morita et al. 1989a; Morita et al. 1989b; Veeken et al. 1991; Weingarten and Perkins 1995; Kessler et al. 1993; Tronvoll and Halleck 1994; Wang and Dusseault 1996). There are only a few that give an indication of the severity of sanding (Papamichos and Malmanger 1999; Nouri et al. 2003; van den Hoek and Geilikman 2003). Some models view sand production as a mixed hydromechanical process (Papamichos and Malmanger 1999; Tronvoll et al. 1992; Tronvoll et al. 1997a; Tronvoll et al. 1997b; Charlez 1997). Some others base their sanding model solely on a cavity's mechanical stability (Antheunis et al. 1976a).

HPHT wells represent one class of problems where the performance of current sand prediction models has not been properly evaluated. We applied three different sanding models to a cluster of six HPHT wells. In all these wells, no sanding was observed under considerable levels of drawdown, certainly far surpassing conditions required for sand failure. Moreover, under failed conditions, the wells were producing high rates of gas for very long periods. The approaches used included popular analytical shear failure and tensile failure based models which showed an unusually high level of conservatism in their prediction of sanding in high pressure wells. We provide a plausible explanation for this behavior which is attributed to the underlying proposition that sand production occurs when sand fails. While this supposition is used in all applications of such models, in deep, high stress and pressure systems, the problem is magnified due to the fact that failure occurs relatively early in the operating life of the reservoir. We propose an alternate approach where sand production criterion is extended to include not only sand failure but also adequate seepage forces to liquefy the sand and hence mobilize it. This approach is shown to better capture the observed response in the field. As an added bonus, the proposed approach quantifies the volume of sand rather than simply give indications of the onset of sand production. This information is helpful in developing the optimal production strategy throughout the life of field. In this paper, we discuss the pros and cons of the commonly used models for sand prediction and provide examples to validate the newly proposed concepts for quantifying sand production. Introduction This paper supplements a recently published paper1 which was focused on the issue of water impact on sand production. While the main emphasis of the current paper is centered around the validity of common analytical sand prediction models in application to high pressure, high temperature (HPHT) wells, some duplication of the material presented earlier is necessary otherwise the main framework for discussion of the findings will be lost and insufficient background and data would be directly available for a meaningful presentation of the findings. The objective of this paper is to show the performance of several common tools and techniques for sand prediction analysis in application to HPHT wells. There are two basic reasons for focusing the main efforts on HPHT wells:previous reported studies have not covered this class of problems, anddeep, over-pressured gas reservoirs more clearly and convincingly bring out the limitations of some of the assumptions embodied in the current approaches used for sand prediction. The basic discussions with respect to pros and cons and validations of different techniques for sand prediction ranging from purely empirical to mid-range analytical and rigorous numerical methods have been presented in a number of publications(e.g., 2–6) and we will do our utmost not to repeat the same messages or findings. Critical Review of Common Methods of Analysis For the sake of objectivity, purely empirical methods (those based on fitting future predictions based on existing trends observed in the field) are not covered here. We will discuss numerical and analytical methods with emphasis on the latter as these constitute the most commonly used ones.